sys/netpfil/pf/pf_ioctl.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-2-Clause
   3  *
   4  * Copyright (c) 2001 Daniel Hartmeier
   5  * Copyright (c) 2002,2003 Henning Brauer
   6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  *
  13  *    - Redistributions of source code must retain the above copyright
  14  *      notice, this list of conditions and the following disclaimer.
  15  *    - Redistributions in binary form must reproduce the above
  16  *      copyright notice, this list of conditions and the following
  17  *      disclaimer in the documentation and/or other materials provided
  18  *      with the distribution.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  31  * POSSIBILITY OF SUCH DAMAGE.
  32  *
  33  * Effort sponsored in part by the Defense Advanced Research Projects
  34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  36  *
  37  *      $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
  38  */
  39
  40 #include <sys/cdefs.h>
  41 __FBSDID("$FreeBSD$");
  42
  43 #include "opt_inet.h"
  44 #include "opt_inet6.h"
  45 #include "opt_bpf.h"
  46 #include "opt_pf.h"
  47
  48 #include <sys/param.h>
  49 #include <sys/_bitset.h>
  50 #include <sys/bitset.h>
  51 #include <sys/bus.h>
  52 #include <sys/conf.h>
  53 #include <sys/endian.h>
  54 #include <sys/fcntl.h>
  55 #include <sys/filio.h>
  56 #include <sys/hash.h>
  57 #include <sys/interrupt.h>
  58 #include <sys/jail.h>
  59 #include <sys/kernel.h>
  60 #include <sys/kthread.h>
  61 #include <sys/lock.h>
  62 #include <sys/mbuf.h>
  63 #include <sys/module.h>
  64 #include <sys/proc.h>
  65 #include <sys/smp.h>
  66 #include <sys/socket.h>
  67 #include <sys/sysctl.h>
  68 #include <sys/md5.h>
  69 #include <sys/ucred.h>
  70
  71 #include <net/if.h>
  72 #include <net/if_var.h>
  73 #include <net/vnet.h>
  74 #include <net/route.h>
  75 #include <net/pfil.h>
  76 #include <net/pfvar.h>
  77 #include <net/if_pfsync.h>
  78 #include <net/if_pflog.h>
  79
  80 #include <netinet/in.h>
  81 #include <netinet/ip.h>
  82 #include <netinet/ip_var.h>
  83 #include <netinet6/ip6_var.h>
  84 #include <netinet/ip_icmp.h>
  85
  86 #ifdef INET6
  87 #include <netinet/ip6.h>
  88 #endif /* INET6 */
  89
  90 #ifdef ALTQ
  91 #include <net/altq/altq.h>
  92 #endif
  93
  94 static struct pf_kpool  *pf_get_kpool(char *, u_int32_t, u_int8_t, u_int32_t,
  95                             u_int8_t, u_int8_t, u_int8_t);
  96
  97 static void              pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
  98 static void              pf_empty_kpool(struct pf_kpalist *);
  99 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
 100                             struct thread *);
 101 #ifdef ALTQ
 102 static int               pf_begin_altq(u_int32_t *);
 103 static int               pf_rollback_altq(u_int32_t);
 104 static int               pf_commit_altq(u_int32_t);
 105 static int               pf_enable_altq(struct pf_altq *);
 106 static int               pf_disable_altq(struct pf_altq *);
 107 static u_int32_t         pf_qname2qid(char *);
 108 static void              pf_qid_unref(u_int32_t);
 109 #endif /* ALTQ */
 110 static int               pf_begin_rules(u_int32_t *, int, const char *);
 111 static int               pf_rollback_rules(u_int32_t, int, char *);
 112 static int               pf_setup_pfsync_matching(struct pf_kruleset *);
 113 static void              pf_hash_rule(MD5_CTX *, struct pf_krule *);
 114 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 115 static int               pf_commit_rules(u_int32_t, int, char *);
 116 static int               pf_addr_setup(struct pf_kruleset *,
 117                             struct pf_addr_wrap *, sa_family_t);
 118 static void              pf_addr_copyout(struct pf_addr_wrap *);
 119 static void              pf_src_node_copy(const struct pf_ksrc_node *,
 120                             struct pf_src_node *);
 121 #ifdef ALTQ
 122 static int               pf_export_kaltq(struct pf_altq *,
 123                             struct pfioc_altq_v1 *, size_t);
 124 static int               pf_import_kaltq(struct pfioc_altq_v1 *,
 125                             struct pf_altq *, size_t);
 126 #endif /* ALTQ */
 127
 128 VNET_DEFINE(struct pf_krule,    pf_default_rule);
 129
 130 #ifdef ALTQ
 131 VNET_DEFINE_STATIC(int,         pf_altq_running);
 132 #define V_pf_altq_running       VNET(pf_altq_running)
 133 #endif
 134
 135 #define TAGID_MAX        50000
 136 struct pf_tagname {
 137         TAILQ_ENTRY(pf_tagname) namehash_entries;
 138         TAILQ_ENTRY(pf_tagname) taghash_entries;
 139         char                    name[PF_TAG_NAME_SIZE];
 140         uint16_t                tag;
 141         int                     ref;
 142 };
 143
 144 struct pf_tagset {
 145         TAILQ_HEAD(, pf_tagname)        *namehash;
 146         TAILQ_HEAD(, pf_tagname)        *taghash;
 147         unsigned int                     mask;
 148         uint32_t                         seed;
 149         BITSET_DEFINE(, TAGID_MAX)       avail;
 150 };
 151
 152 VNET_DEFINE(struct pf_tagset, pf_tags);
 153 #define V_pf_tags       VNET(pf_tags)
 154 static unsigned int     pf_rule_tag_hashsize;
 155 #define PF_RULE_TAG_HASH_SIZE_DEFAULT   128
 156 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
 157     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
 158     "Size of pf(4) rule tag hashtable");
 159
 160 #ifdef ALTQ
 161 VNET_DEFINE(struct pf_tagset, pf_qids);
 162 #define V_pf_qids       VNET(pf_qids)
 163 static unsigned int     pf_queue_tag_hashsize;
 164 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT  128
 165 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
 166     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
 167     "Size of pf(4) queue tag hashtable");
 168 #endif
 169 VNET_DEFINE(uma_zone_t,  pf_tag_z);
 170 #define V_pf_tag_z               VNET(pf_tag_z)
 171 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 172 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 173
 174 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 175 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 176 #endif
 177
 178 static void              pf_init_tagset(struct pf_tagset *, unsigned int *,
 179                             unsigned int);
 180 static void              pf_cleanup_tagset(struct pf_tagset *);
 181 static uint16_t          tagname2hashindex(const struct pf_tagset *, const char *);
 182 static uint16_t          tag2hashindex(const struct pf_tagset *, uint16_t);
 183 static u_int16_t         tagname2tag(struct pf_tagset *, char *);
 184 static u_int16_t         pf_tagname2tag(char *);
 185 static void              tag_unref(struct pf_tagset *, u_int16_t);
 186
 187 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 188
 189 struct cdev *pf_dev;
 190
 191 /*
 192  * XXX - These are new and need to be checked when moveing to a new version
 193  */
 194 static void              pf_clear_states(void);
 195 static int               pf_clear_tables(void);
 196 static void              pf_clear_srcnodes(struct pf_ksrc_node *);
 197 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 198 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 199
 200 /*
 201  * Wrapper functions for pfil(9) hooks
 202  */
 203 #ifdef INET
 204 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 205     int dir, int flags, struct inpcb *inp);
 206 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 207     int dir, int flags, struct inpcb *inp);
 208 #endif
 209 #ifdef INET6
 210 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 211     int dir, int flags, struct inpcb *inp);
 212 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 213     int dir, int flags, struct inpcb *inp);
 214 #endif
 215
 216 static int              hook_pf(void);
 217 static int              dehook_pf(void);
 218 static int              shutdown_pf(void);
 219 static int              pf_load(void);
 220 static void             pf_unload(void);
 221
 222 static struct cdevsw pf_cdevsw = {
 223         .d_ioctl =      pfioctl,
 224         .d_name =       PF_NAME,
 225         .d_version =    D_VERSION,
 226 };
 227
 228 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
 229 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 230
 231 /*
 232  * We need a flag that is neither hooked nor running to know when
 233  * the VNET is "valid".  We primarily need this to control (global)
 234  * external event, e.g., eventhandlers.
 235  */
 236 VNET_DEFINE(int, pf_vnet_active);
 237 #define V_pf_vnet_active        VNET(pf_vnet_active)
 238
 239 int pf_end_threads;
 240 struct proc *pf_purge_proc;
 241
 242 struct rmlock                   pf_rules_lock;
 243 struct sx                       pf_ioctl_lock;
 244 struct sx                       pf_end_lock;
 245
 246 /* pfsync */
 247 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
 248 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
 249 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
 250 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
 251 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
 252 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
 253 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
 254
 255 /* pflog */
 256 pflog_packet_t                  *pflog_packet_ptr = NULL;
 257
 258 extern u_long   pf_ioctl_maxcount;
 259
 260 static void
 261 pfattach_vnet(void)
 262 {
 263         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 264
 265         pf_initialize();
 266         pfr_initialize();
 267         pfi_initialize_vnet();
 268         pf_normalize_init();
 269
 270         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 271         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 272
 273         RB_INIT(&V_pf_anchors);
 274         pf_init_kruleset(&pf_main_ruleset);
 275
 276         /* default rule should never be garbage collected */
 277         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 278 #ifdef PF_DEFAULT_TO_DROP
 279         V_pf_default_rule.action = PF_DROP;
 280 #else
 281         V_pf_default_rule.action = PF_PASS;
 282 #endif
 283         V_pf_default_rule.nr = -1;
 284         V_pf_default_rule.rtableid = -1;
 285
 286         V_pf_default_rule.evaluations = counter_u64_alloc(M_WAITOK);
 287         for (int i = 0; i < 2; i++) {
 288                 V_pf_default_rule.packets[i] = counter_u64_alloc(M_WAITOK);
 289                 V_pf_default_rule.bytes[i] = counter_u64_alloc(M_WAITOK);
 290         }
 291         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 292         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 293         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 294
 295         /* initialize default timeouts */
 296         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 297         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 298         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 299         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 300         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 301         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 302         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 303         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 304         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 305         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 306         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 307         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 308         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 309         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 310         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 311         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 312         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 313         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 314         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 315         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 316
 317         bzero(&V_pf_status, sizeof(V_pf_status));
 318         V_pf_status.debug = PF_DEBUG_URGENT;
 319
 320         V_pf_pfil_hooked = 0;
 321
 322         /* XXX do our best to avoid a conflict */
 323         V_pf_status.hostid = arc4random();
 324
 325         for (int i = 0; i < PFRES_MAX; i++)
 326                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 327         for (int i = 0; i < LCNT_MAX; i++)
 328                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 329         for (int i = 0; i < FCNT_MAX; i++)
 330                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 331         for (int i = 0; i < SCNT_MAX; i++)
 332                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 333
 334         if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
 335             INTR_MPSAFE, &V_pf_swi_cookie) != 0)
 336                 /* XXXGL: leaked all above. */
 337                 return;
 338 }
 339
 340 static struct pf_kpool *
 341 pf_get_kpool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 342     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 343     u_int8_t check_ticket)
 344 {
 345         struct pf_kruleset      *ruleset;
 346         struct pf_krule         *rule;
 347         int                      rs_num;
 348
 349         ruleset = pf_find_kruleset(anchor);
 350         if (ruleset == NULL)
 351                 return (NULL);
 352         rs_num = pf_get_ruleset_number(rule_action);
 353         if (rs_num >= PF_RULESET_MAX)
 354                 return (NULL);
 355         if (active) {
 356                 if (check_ticket && ticket !=
 357                     ruleset->rules[rs_num].active.ticket)
 358                         return (NULL);
 359                 if (r_last)
 360                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 361                             pf_krulequeue);
 362                 else
 363                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 364         } else {
 365                 if (check_ticket && ticket !=
 366                     ruleset->rules[rs_num].inactive.ticket)
 367                         return (NULL);
 368                 if (r_last)
 369                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 370                             pf_krulequeue);
 371                 else
 372                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 373         }
 374         if (!r_last) {
 375                 while ((rule != NULL) && (rule->nr != rule_number))
 376                         rule = TAILQ_NEXT(rule, entries);
 377         }
 378         if (rule == NULL)
 379                 return (NULL);
 380
 381         return (&rule->rpool);
 382 }
 383
 384 static void
 385 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
 386 {
 387         struct pf_kpooladdr     *mv_pool_pa;
 388
 389         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 390                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 391                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 392         }
 393 }
 394
 395 static void
 396 pf_empty_kpool(struct pf_kpalist *poola)
 397 {
 398         struct pf_kpooladdr *pa;
 399
 400         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 401                 switch (pa->addr.type) {
 402                 case PF_ADDR_DYNIFTL:
 403                         pfi_dynaddr_remove(pa->addr.p.dyn);
 404                         break;
 405                 case PF_ADDR_TABLE:
 406                         /* XXX: this could be unfinished pooladdr on pabuf */
 407                         if (pa->addr.p.tbl != NULL)
 408                                 pfr_detach_table(pa->addr.p.tbl);
 409                         break;
 410                 }
 411                 if (pa->kif)
 412                         pfi_kkif_unref(pa->kif);
 413                 TAILQ_REMOVE(poola, pa, entries);
 414                 free(pa, M_PFRULE);
 415         }
 416 }
 417
 418 static void
 419 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
 420 {
 421
 422         PF_RULES_WASSERT();
 423
 424         TAILQ_REMOVE(rulequeue, rule, entries);
 425
 426         PF_UNLNKDRULES_LOCK();
 427         rule->rule_flag |= PFRULE_REFS;
 428         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 429         PF_UNLNKDRULES_UNLOCK();
 430 }
 431
 432 void
 433 pf_free_rule(struct pf_krule *rule)
 434 {
 435
 436         PF_RULES_WASSERT();
 437
 438         if (rule->tag)
 439                 tag_unref(&V_pf_tags, rule->tag);
 440         if (rule->match_tag)
 441                 tag_unref(&V_pf_tags, rule->match_tag);
 442 #ifdef ALTQ
 443         if (rule->pqid != rule->qid)
 444                 pf_qid_unref(rule->pqid);
 445         pf_qid_unref(rule->qid);
 446 #endif
 447         switch (rule->src.addr.type) {
 448         case PF_ADDR_DYNIFTL:
 449                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 450                 break;
 451         case PF_ADDR_TABLE:
 452                 pfr_detach_table(rule->src.addr.p.tbl);
 453                 break;
 454         }
 455         switch (rule->dst.addr.type) {
 456         case PF_ADDR_DYNIFTL:
 457                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 458                 break;
 459         case PF_ADDR_TABLE:
 460                 pfr_detach_table(rule->dst.addr.p.tbl);
 461                 break;
 462         }
 463         if (rule->overload_tbl)
 464                 pfr_detach_table(rule->overload_tbl);
 465         if (rule->kif)
 466                 pfi_kkif_unref(rule->kif);
 467         pf_kanchor_remove(rule);
 468         pf_empty_kpool(&rule->rpool.list);
 469         counter_u64_free(rule->evaluations);
 470         for (int i = 0; i < 2; i++) {
 471                 counter_u64_free(rule->packets[i]);
 472                 counter_u64_free(rule->bytes[i]);
 473         }
 474         counter_u64_free(rule->states_cur);
 475         counter_u64_free(rule->states_tot);
 476         counter_u64_free(rule->src_nodes);
 477         free(rule, M_PFRULE);
 478 }
 479
 480 static void
 481 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
 482     unsigned int default_size)
 483 {
 484         unsigned int i;
 485         unsigned int hashsize;
 486
 487         if (*tunable_size == 0 || !powerof2(*tunable_size))
 488                 *tunable_size = default_size;
 489
 490         hashsize = *tunable_size;
 491         ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
 492             M_WAITOK);
 493         ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
 494             M_WAITOK);
 495         ts->mask = hashsize - 1;
 496         ts->seed = arc4random();
 497         for (i = 0; i < hashsize; i++) {
 498                 TAILQ_INIT(&ts->namehash[i]);
 499                 TAILQ_INIT(&ts->taghash[i]);
 500         }
 501         BIT_FILL(TAGID_MAX, &ts->avail);
 502 }
 503
 504 static void
 505 pf_cleanup_tagset(struct pf_tagset *ts)
 506 {
 507         unsigned int i;
 508         unsigned int hashsize;
 509         struct pf_tagname *t, *tmp;
 510
 511         /*
 512          * Only need to clean up one of the hashes as each tag is hashed
 513          * into each table.
 514          */
 515         hashsize = ts->mask + 1;
 516         for (i = 0; i < hashsize; i++)
 517                 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
 518                         uma_zfree(V_pf_tag_z, t);
 519
 520         free(ts->namehash, M_PFHASH);
 521         free(ts->taghash, M_PFHASH);
 522 }
 523
 524 static uint16_t
 525 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
 526 {
 527         size_t len;
 528
 529         len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
 530         return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
 531 }
 532
 533 static uint16_t
 534 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
 535 {
 536
 537         return (tag & ts->mask);
 538 }
 539
 540 static u_int16_t
 541 tagname2tag(struct pf_tagset *ts, char *tagname)
 542 {
 543         struct pf_tagname       *tag;
 544         u_int32_t                index;
 545         u_int16_t                new_tagid;
 546
 547         PF_RULES_WASSERT();
 548
 549         index = tagname2hashindex(ts, tagname);
 550         TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
 551                 if (strcmp(tagname, tag->name) == 0) {
 552                         tag->ref++;
 553                         return (tag->tag);
 554                 }
 555
 556         /*
 557          * new entry
 558          *
 559          * to avoid fragmentation, we do a linear search from the beginning
 560          * and take the first free slot we find.
 561          */
 562         new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
 563         /*
 564          * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
 565          * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
 566          * set.  It may also return a bit number greater than TAGID_MAX due
 567          * to rounding of the number of bits in the vector up to a multiple
 568          * of the vector word size at declaration/allocation time.
 569          */
 570         if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
 571                 return (0);
 572
 573         /* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
 574         BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
 575
 576         /* allocate and fill new struct pf_tagname */
 577         tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
 578         if (tag == NULL)
 579                 return (0);
 580         strlcpy(tag->name, tagname, sizeof(tag->name));
 581         tag->tag = new_tagid;
 582         tag->ref = 1;
 583
 584         /* Insert into namehash */
 585         TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
 586
 587         /* Insert into taghash */
 588         index = tag2hashindex(ts, new_tagid);
 589         TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
 590
 591         return (tag->tag);
 592 }
 593
 594 static void
 595 tag_unref(struct pf_tagset *ts, u_int16_t tag)
 596 {
 597         struct pf_tagname       *t;
 598         uint16_t                 index;
 599
 600         PF_RULES_WASSERT();
 601
 602         index = tag2hashindex(ts, tag);
 603         TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
 604                 if (tag == t->tag) {
 605                         if (--t->ref == 0) {
 606                                 TAILQ_REMOVE(&ts->taghash[index], t,
 607                                     taghash_entries);
 608                                 index = tagname2hashindex(ts, t->name);
 609                                 TAILQ_REMOVE(&ts->namehash[index], t,
 610                                     namehash_entries);
 611                                 /* Bits are 0-based for BIT_SET() */
 612                                 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
 613                                 uma_zfree(V_pf_tag_z, t);
 614                         }
 615                         break;
 616                 }
 617 }
 618
 619 static u_int16_t
 620 pf_tagname2tag(char *tagname)
 621 {
 622         return (tagname2tag(&V_pf_tags, tagname));
 623 }
 624
 625 #ifdef ALTQ
 626 static u_int32_t
 627 pf_qname2qid(char *qname)
 628 {
 629         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 630 }
 631
 632 static void
 633 pf_qid_unref(u_int32_t qid)
 634 {
 635         tag_unref(&V_pf_qids, (u_int16_t)qid);
 636 }
 637
 638 static int
 639 pf_begin_altq(u_int32_t *ticket)
 640 {
 641         struct pf_altq  *altq, *tmp;
 642         int              error = 0;
 643
 644         PF_RULES_WASSERT();
 645
 646         /* Purge the old altq lists */
 647         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 648                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 649                         /* detach and destroy the discipline */
 650                         error = altq_remove(altq);
 651                 }
 652                 free(altq, M_PFALTQ);
 653         }
 654         TAILQ_INIT(V_pf_altq_ifs_inactive);
 655         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 656                 pf_qid_unref(altq->qid);
 657                 free(altq, M_PFALTQ);
 658         }
 659         TAILQ_INIT(V_pf_altqs_inactive);
 660         if (error)
 661                 return (error);
 662         *ticket = ++V_ticket_altqs_inactive;
 663         V_altqs_inactive_open = 1;
 664         return (0);
 665 }
 666
 667 static int
 668 pf_rollback_altq(u_int32_t ticket)
 669 {
 670         struct pf_altq  *altq, *tmp;
 671         int              error = 0;
 672
 673         PF_RULES_WASSERT();
 674
 675         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 676                 return (0);
 677         /* Purge the old altq lists */
 678         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 679                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 680                         /* detach and destroy the discipline */
 681                         error = altq_remove(altq);
 682                 }
 683                 free(altq, M_PFALTQ);
 684         }
 685         TAILQ_INIT(V_pf_altq_ifs_inactive);
 686         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 687                 pf_qid_unref(altq->qid);
 688                 free(altq, M_PFALTQ);
 689         }
 690         TAILQ_INIT(V_pf_altqs_inactive);
 691         V_altqs_inactive_open = 0;
 692         return (error);
 693 }
 694
 695 static int
 696 pf_commit_altq(u_int32_t ticket)
 697 {
 698         struct pf_altqqueue     *old_altqs, *old_altq_ifs;
 699         struct pf_altq          *altq, *tmp;
 700         int                      err, error = 0;
 701
 702         PF_RULES_WASSERT();
 703
 704         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 705                 return (EBUSY);
 706
 707         /* swap altqs, keep the old. */
 708         old_altqs = V_pf_altqs_active;
 709         old_altq_ifs = V_pf_altq_ifs_active;
 710         V_pf_altqs_active = V_pf_altqs_inactive;
 711         V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
 712         V_pf_altqs_inactive = old_altqs;
 713         V_pf_altq_ifs_inactive = old_altq_ifs;
 714         V_ticket_altqs_active = V_ticket_altqs_inactive;
 715
 716         /* Attach new disciplines */
 717         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
 718                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 719                         /* attach the discipline */
 720                         error = altq_pfattach(altq);
 721                         if (error == 0 && V_pf_altq_running)
 722                                 error = pf_enable_altq(altq);
 723                         if (error != 0)
 724                                 return (error);
 725                 }
 726         }
 727
 728         /* Purge the old altq lists */
 729         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 730                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 731                         /* detach and destroy the discipline */
 732                         if (V_pf_altq_running)
 733                                 error = pf_disable_altq(altq);
 734                         err = altq_pfdetach(altq);
 735                         if (err != 0 && error == 0)
 736                                 error = err;
 737                         err = altq_remove(altq);
 738                         if (err != 0 && error == 0)
 739                                 error = err;
 740                 }
 741                 free(altq, M_PFALTQ);
 742         }
 743         TAILQ_INIT(V_pf_altq_ifs_inactive);
 744         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 745                 pf_qid_unref(altq->qid);
 746                 free(altq, M_PFALTQ);
 747         }
 748         TAILQ_INIT(V_pf_altqs_inactive);
 749
 750         V_altqs_inactive_open = 0;
 751         return (error);
 752 }
 753
 754 static int
 755 pf_enable_altq(struct pf_altq *altq)
 756 {
 757         struct ifnet            *ifp;
 758         struct tb_profile        tb;
 759         int                      error = 0;
 760
 761         if ((ifp = ifunit(altq->ifname)) == NULL)
 762                 return (EINVAL);
 763
 764         if (ifp->if_snd.altq_type != ALTQT_NONE)
 765                 error = altq_enable(&ifp->if_snd);
 766
 767         /* set tokenbucket regulator */
 768         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 769                 tb.rate = altq->ifbandwidth;
 770                 tb.depth = altq->tbrsize;
 771                 error = tbr_set(&ifp->if_snd, &tb);
 772         }
 773
 774         return (error);
 775 }
 776
 777 static int
 778 pf_disable_altq(struct pf_altq *altq)
 779 {
 780         struct ifnet            *ifp;
 781         struct tb_profile        tb;
 782         int                      error;
 783
 784         if ((ifp = ifunit(altq->ifname)) == NULL)
 785                 return (EINVAL);
 786
 787         /*
 788          * when the discipline is no longer referenced, it was overridden
 789          * by a new one.  if so, just return.
 790          */
 791         if (altq->altq_disc != ifp->if_snd.altq_disc)
 792                 return (0);
 793
 794         error = altq_disable(&ifp->if_snd);
 795
 796         if (error == 0) {
 797                 /* clear tokenbucket regulator */
 798                 tb.rate = 0;
 799                 error = tbr_set(&ifp->if_snd, &tb);
 800         }
 801
 802         return (error);
 803 }
 804
 805 static int
 806 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
 807     struct pf_altq *altq)
 808 {
 809         struct ifnet    *ifp1;
 810         int              error = 0;
 811
 812         /* Deactivate the interface in question */
 813         altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 814         if ((ifp1 = ifunit(altq->ifname)) == NULL ||
 815             (remove && ifp1 == ifp)) {
 816                 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 817         } else {
 818                 error = altq_add(ifp1, altq);
 819
 820                 if (ticket != V_ticket_altqs_inactive)
 821                         error = EBUSY;
 822
 823                 if (error)
 824                         free(altq, M_PFALTQ);
 825         }
 826
 827         return (error);
 828 }
 829
 830 void
 831 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 832 {
 833         struct pf_altq  *a1, *a2, *a3;
 834         u_int32_t        ticket;
 835         int              error = 0;
 836
 837         /*
 838          * No need to re-evaluate the configuration for events on interfaces
 839          * that do not support ALTQ, as it's not possible for such
 840          * interfaces to be part of the configuration.
 841          */
 842         if (!ALTQ_IS_READY(&ifp->if_snd))
 843                 return;
 844
 845         /* Interrupt userland queue modifications */
 846         if (V_altqs_inactive_open)
 847                 pf_rollback_altq(V_ticket_altqs_inactive);
 848
 849         /* Start new altq ruleset */
 850         if (pf_begin_altq(&ticket))
 851                 return;
 852
 853         /* Copy the current active set */
 854         TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
 855                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 856                 if (a2 == NULL) {
 857                         error = ENOMEM;
 858                         break;
 859                 }
 860                 bcopy(a1, a2, sizeof(struct pf_altq));
 861
 862                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 863                 if (error)
 864                         break;
 865
 866                 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
 867         }
 868         if (error)
 869                 goto out;
 870         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 871                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 872                 if (a2 == NULL) {
 873                         error = ENOMEM;
 874                         break;
 875                 }
 876                 bcopy(a1, a2, sizeof(struct pf_altq));
 877
 878                 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 879                         error = EBUSY;
 880                         free(a2, M_PFALTQ);
 881                         break;
 882                 }
 883                 a2->altq_disc = NULL;
 884                 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
 885                         if (strncmp(a3->ifname, a2->ifname,
 886                                 IFNAMSIZ) == 0) {
 887                                 a2->altq_disc = a3->altq_disc;
 888                                 break;
 889                         }
 890                 }
 891                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 892                 if (error)
 893                         break;
 894
 895                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 896         }
 897
 898 out:
 899         if (error != 0)
 900                 pf_rollback_altq(ticket);
 901         else
 902                 pf_commit_altq(ticket);
 903 }
 904 #endif /* ALTQ */
 905
 906 static int
 907 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 908 {
 909         struct pf_kruleset      *rs;
 910         struct pf_krule         *rule;
 911
 912         PF_RULES_WASSERT();
 913
 914         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 915                 return (EINVAL);
 916         rs = pf_find_or_create_kruleset(anchor);
 917         if (rs == NULL)
 918                 return (EINVAL);
 919         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 920                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 921                 rs->rules[rs_num].inactive.rcount--;
 922         }
 923         *ticket = ++rs->rules[rs_num].inactive.ticket;
 924         rs->rules[rs_num].inactive.open = 1;
 925         return (0);
 926 }
 927
 928 static int
 929 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 930 {
 931         struct pf_kruleset      *rs;
 932         struct pf_krule         *rule;
 933
 934         PF_RULES_WASSERT();
 935
 936         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 937                 return (EINVAL);
 938         rs = pf_find_kruleset(anchor);
 939         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 940             rs->rules[rs_num].inactive.ticket != ticket)
 941                 return (0);
 942         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 943                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 944                 rs->rules[rs_num].inactive.rcount--;
 945         }
 946         rs->rules[rs_num].inactive.open = 0;
 947         return (0);
 948 }
 949
 950 #define PF_MD5_UPD(st, elm)                                             \
 951                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 952
 953 #define PF_MD5_UPD_STR(st, elm)                                         \
 954                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 955
 956 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 957                 (stor) = htonl((st)->elm);                              \
 958                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 959 } while (0)
 960
 961 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 962                 (stor) = htons((st)->elm);                              \
 963                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 964 } while (0)
 965
 966 static void
 967 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 968 {
 969         PF_MD5_UPD(pfr, addr.type);
 970         switch (pfr->addr.type) {
 971                 case PF_ADDR_DYNIFTL:
 972                         PF_MD5_UPD(pfr, addr.v.ifname);
 973                         PF_MD5_UPD(pfr, addr.iflags);
 974                         break;
 975                 case PF_ADDR_TABLE:
 976                         PF_MD5_UPD(pfr, addr.v.tblname);
 977                         break;
 978                 case PF_ADDR_ADDRMASK:
 979                         /* XXX ignore af? */
 980                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 981                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 982                         break;
 983         }
 984
 985         PF_MD5_UPD(pfr, port[0]);
 986         PF_MD5_UPD(pfr, port[1]);
 987         PF_MD5_UPD(pfr, neg);
 988         PF_MD5_UPD(pfr, port_op);
 989 }
 990
 991 static void
 992 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
 993 {
 994         u_int16_t x;
 995         u_int32_t y;
 996
 997         pf_hash_rule_addr(ctx, &rule->src);
 998         pf_hash_rule_addr(ctx, &rule->dst);
 999         PF_MD5_UPD_STR(rule, label);
1000         PF_MD5_UPD_STR(rule, ifname);
1001         PF_MD5_UPD_STR(rule, match_tagname);
1002         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1003         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1004         PF_MD5_UPD_HTONL(rule, prob, y);
1005         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1006         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1007         PF_MD5_UPD(rule, uid.op);
1008         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1009         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1010         PF_MD5_UPD(rule, gid.op);
1011         PF_MD5_UPD_HTONL(rule, rule_flag, y);
1012         PF_MD5_UPD(rule, action);
1013         PF_MD5_UPD(rule, direction);
1014         PF_MD5_UPD(rule, af);
1015         PF_MD5_UPD(rule, quick);
1016         PF_MD5_UPD(rule, ifnot);
1017         PF_MD5_UPD(rule, match_tag_not);
1018         PF_MD5_UPD(rule, natpass);
1019         PF_MD5_UPD(rule, keep_state);
1020         PF_MD5_UPD(rule, proto);
1021         PF_MD5_UPD(rule, type);
1022         PF_MD5_UPD(rule, code);
1023         PF_MD5_UPD(rule, flags);
1024         PF_MD5_UPD(rule, flagset);
1025         PF_MD5_UPD(rule, allow_opts);
1026         PF_MD5_UPD(rule, rt);
1027         PF_MD5_UPD(rule, tos);
1028 }
1029
1030 static int
1031 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1032 {
1033         struct pf_kruleset      *rs;
1034         struct pf_krule         *rule, **old_array;
1035         struct pf_krulequeue    *old_rules;
1036         int                      error;
1037         u_int32_t                old_rcount;
1038
1039         PF_RULES_WASSERT();
1040
1041         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1042                 return (EINVAL);
1043         rs = pf_find_kruleset(anchor);
1044         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1045             ticket != rs->rules[rs_num].inactive.ticket)
1046                 return (EBUSY);
1047
1048         /* Calculate checksum for the main ruleset */
1049         if (rs == &pf_main_ruleset) {
1050                 error = pf_setup_pfsync_matching(rs);
1051                 if (error != 0)
1052                         return (error);
1053         }
1054
1055         /* Swap rules, keep the old. */
1056         old_rules = rs->rules[rs_num].active.ptr;
1057         old_rcount = rs->rules[rs_num].active.rcount;
1058         old_array = rs->rules[rs_num].active.ptr_array;
1059
1060         rs->rules[rs_num].active.ptr =
1061             rs->rules[rs_num].inactive.ptr;
1062         rs->rules[rs_num].active.ptr_array =
1063             rs->rules[rs_num].inactive.ptr_array;
1064         rs->rules[rs_num].active.rcount =
1065             rs->rules[rs_num].inactive.rcount;
1066         rs->rules[rs_num].inactive.ptr = old_rules;
1067         rs->rules[rs_num].inactive.ptr_array = old_array;
1068         rs->rules[rs_num].inactive.rcount = old_rcount;
1069
1070         rs->rules[rs_num].active.ticket =
1071             rs->rules[rs_num].inactive.ticket;
1072         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1073
1074
1075         /* Purge the old rule list. */
1076         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1077                 pf_unlink_rule(old_rules, rule);
1078         if (rs->rules[rs_num].inactive.ptr_array)
1079                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1080         rs->rules[rs_num].inactive.ptr_array = NULL;
1081         rs->rules[rs_num].inactive.rcount = 0;
1082         rs->rules[rs_num].inactive.open = 0;
1083         pf_remove_if_empty_kruleset(rs);
1084
1085         return (0);
1086 }
1087
1088 static int
1089 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1090 {
1091         MD5_CTX                  ctx;
1092         struct pf_krule         *rule;
1093         int                      rs_cnt;
1094         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
1095
1096         MD5Init(&ctx);
1097         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1098                 /* XXX PF_RULESET_SCRUB as well? */
1099                 if (rs_cnt == PF_RULESET_SCRUB)
1100                         continue;
1101
1102                 if (rs->rules[rs_cnt].inactive.ptr_array)
1103                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1104                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1105
1106                 if (rs->rules[rs_cnt].inactive.rcount) {
1107                         rs->rules[rs_cnt].inactive.ptr_array =
1108                             malloc(sizeof(caddr_t) *
1109                             rs->rules[rs_cnt].inactive.rcount,
1110                             M_TEMP, M_NOWAIT);
1111
1112                         if (!rs->rules[rs_cnt].inactive.ptr_array)
1113                                 return (ENOMEM);
1114                 }
1115
1116                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1117                     entries) {
1118                         pf_hash_rule(&ctx, rule);
1119                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1120                 }
1121         }
1122
1123         MD5Final(digest, &ctx);
1124         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1125         return (0);
1126 }
1127
1128 static int
1129 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1130     sa_family_t af)
1131 {
1132         int error = 0;
1133
1134         switch (addr->type) {
1135         case PF_ADDR_TABLE:
1136                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1137                 if (addr->p.tbl == NULL)
1138                         error = ENOMEM;
1139                 break;
1140         case PF_ADDR_DYNIFTL:
1141                 error = pfi_dynaddr_setup(addr, af);
1142                 break;
1143         }
1144
1145         return (error);
1146 }
1147
1148 static void
1149 pf_addr_copyout(struct pf_addr_wrap *addr)
1150 {
1151
1152         switch (addr->type) {
1153         case PF_ADDR_DYNIFTL:
1154                 pfi_dynaddr_copyout(addr);
1155                 break;
1156         case PF_ADDR_TABLE:
1157                 pf_tbladdr_copyout(addr);
1158                 break;
1159         }
1160 }
1161
1162 static void
1163 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1164 {
1165         int     secs = time_uptime, diff;
1166
1167         bzero(out, sizeof(struct pf_src_node));
1168
1169         bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1170         bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1171
1172         if (in->rule.ptr != NULL)
1173                 out->rule.nr = in->rule.ptr->nr;
1174
1175         for (int i = 0; i < 2; i++) {
1176                 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1177                 out->packets[i] = counter_u64_fetch(in->packets[i]);
1178         }
1179
1180         out->states = in->states;
1181         out->conn = in->conn;
1182         out->af = in->af;
1183         out->ruletype = in->ruletype;
1184
1185         out->creation = secs - in->creation;
1186         if (out->expire > secs)
1187                 out->expire -= secs;
1188         else
1189                 out->expire = 0;
1190
1191         /* Adjust the connection rate estimate. */
1192         diff = secs - in->conn_rate.last;
1193         if (diff >= in->conn_rate.seconds)
1194                 out->conn_rate.count = 0;
1195         else
1196                 out->conn_rate.count -=
1197                     in->conn_rate.count * diff /
1198                     in->conn_rate.seconds;
1199 }
1200
1201 #ifdef ALTQ
1202 /*
1203  * Handle export of struct pf_kaltq to user binaries that may be using any
1204  * version of struct pf_altq.
1205  */
1206 static int
1207 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1208 {
1209         u_int32_t version;
1210
1211         if (ioc_size == sizeof(struct pfioc_altq_v0))
1212                 version = 0;
1213         else
1214                 version = pa->version;
1215
1216         if (version > PFIOC_ALTQ_VERSION)
1217                 return (EINVAL);
1218
1219 #define ASSIGN(x) exported_q->x = q->x
1220 #define COPY(x) \
1221         bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1222 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1223 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1224
1225         switch (version) {
1226         case 0: {
1227                 struct pf_altq_v0 *exported_q =
1228                     &((struct pfioc_altq_v0 *)pa)->altq;
1229
1230                 COPY(ifname);
1231
1232                 ASSIGN(scheduler);
1233                 ASSIGN(tbrsize);
1234                 exported_q->tbrsize = SATU16(q->tbrsize);
1235                 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1236
1237                 COPY(qname);
1238                 COPY(parent);
1239                 ASSIGN(parent_qid);
1240                 exported_q->bandwidth = SATU32(q->bandwidth);
1241                 ASSIGN(priority);
1242                 ASSIGN(local_flags);
1243
1244                 ASSIGN(qlimit);
1245                 ASSIGN(flags);
1246
1247                 if (q->scheduler == ALTQT_HFSC) {
1248 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1249 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1250                             SATU32(q->pq_u.hfsc_opts.x)
1251
1252                         ASSIGN_OPT_SATU32(rtsc_m1);
1253                         ASSIGN_OPT(rtsc_d);
1254                         ASSIGN_OPT_SATU32(rtsc_m2);
1255
1256                         ASSIGN_OPT_SATU32(lssc_m1);
1257                         ASSIGN_OPT(lssc_d);
1258                         ASSIGN_OPT_SATU32(lssc_m2);
1259
1260                         ASSIGN_OPT_SATU32(ulsc_m1);
1261                         ASSIGN_OPT(ulsc_d);
1262                         ASSIGN_OPT_SATU32(ulsc_m2);
1263
1264                         ASSIGN_OPT(flags);
1265
1266 #undef ASSIGN_OPT
1267 #undef ASSIGN_OPT_SATU32
1268                 } else
1269                         COPY(pq_u);
1270
1271                 ASSIGN(qid);
1272                 break;
1273         }
1274         case 1: {
1275                 struct pf_altq_v1 *exported_q =
1276                     &((struct pfioc_altq_v1 *)pa)->altq;
1277
1278                 COPY(ifname);
1279
1280                 ASSIGN(scheduler);
1281                 ASSIGN(tbrsize);
1282                 ASSIGN(ifbandwidth);
1283
1284                 COPY(qname);
1285                 COPY(parent);
1286                 ASSIGN(parent_qid);
1287                 ASSIGN(bandwidth);
1288                 ASSIGN(priority);
1289                 ASSIGN(local_flags);
1290
1291                 ASSIGN(qlimit);
1292                 ASSIGN(flags);
1293                 COPY(pq_u);
1294
1295                 ASSIGN(qid);
1296                 break;
1297         }
1298         default:
1299                 panic("%s: unhandled struct pfioc_altq version", __func__);
1300                 break;
1301         }
1302
1303 #undef ASSIGN
1304 #undef COPY
1305 #undef SATU16
1306 #undef SATU32
1307
1308         return (0);
1309 }
1310
1311 /*
1312  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1313  * that may be using any version of it.
1314  */
1315 static int
1316 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1317 {
1318         u_int32_t version;
1319
1320         if (ioc_size == sizeof(struct pfioc_altq_v0))
1321                 version = 0;
1322         else
1323                 version = pa->version;
1324
1325         if (version > PFIOC_ALTQ_VERSION)
1326                 return (EINVAL);
1327
1328 #define ASSIGN(x) q->x = imported_q->x
1329 #define COPY(x) \
1330         bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1331
1332         switch (version) {
1333         case 0: {
1334                 struct pf_altq_v0 *imported_q =
1335                     &((struct pfioc_altq_v0 *)pa)->altq;
1336
1337                 COPY(ifname);
1338
1339                 ASSIGN(scheduler);
1340                 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1341                 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1342
1343                 COPY(qname);
1344                 COPY(parent);
1345                 ASSIGN(parent_qid);
1346                 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1347                 ASSIGN(priority);
1348                 ASSIGN(local_flags);
1349
1350                 ASSIGN(qlimit);
1351                 ASSIGN(flags);
1352
1353                 if (imported_q->scheduler == ALTQT_HFSC) {
1354 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1355
1356                         /*
1357                          * The m1 and m2 parameters are being copied from
1358                          * 32-bit to 64-bit.
1359                          */
1360                         ASSIGN_OPT(rtsc_m1);
1361                         ASSIGN_OPT(rtsc_d);
1362                         ASSIGN_OPT(rtsc_m2);
1363
1364                         ASSIGN_OPT(lssc_m1);
1365                         ASSIGN_OPT(lssc_d);
1366                         ASSIGN_OPT(lssc_m2);
1367
1368                         ASSIGN_OPT(ulsc_m1);
1369                         ASSIGN_OPT(ulsc_d);
1370                         ASSIGN_OPT(ulsc_m2);
1371
1372                         ASSIGN_OPT(flags);
1373
1374 #undef ASSIGN_OPT
1375                 } else
1376                         COPY(pq_u);
1377
1378                 ASSIGN(qid);
1379                 break;
1380         }
1381         case 1: {
1382                 struct pf_altq_v1 *imported_q =
1383                     &((struct pfioc_altq_v1 *)pa)->altq;
1384
1385                 COPY(ifname);
1386
1387                 ASSIGN(scheduler);
1388                 ASSIGN(tbrsize);
1389                 ASSIGN(ifbandwidth);
1390
1391                 COPY(qname);
1392                 COPY(parent);
1393                 ASSIGN(parent_qid);
1394                 ASSIGN(bandwidth);
1395                 ASSIGN(priority);
1396                 ASSIGN(local_flags);
1397
1398                 ASSIGN(qlimit);
1399                 ASSIGN(flags);
1400                 COPY(pq_u);
1401
1402                 ASSIGN(qid);
1403                 break;
1404         }
1405         default:
1406                 panic("%s: unhandled struct pfioc_altq version", __func__);
1407                 break;
1408         }
1409
1410 #undef ASSIGN
1411 #undef COPY
1412
1413         return (0);
1414 }
1415
1416 static struct pf_altq *
1417 pf_altq_get_nth_active(u_int32_t n)
1418 {
1419         struct pf_altq          *altq;
1420         u_int32_t                nr;
1421
1422         nr = 0;
1423         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1424                 if (nr == n)
1425                         return (altq);
1426                 nr++;
1427         }
1428
1429         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1430                 if (nr == n)
1431                         return (altq);
1432                 nr++;
1433         }
1434
1435         return (NULL);
1436 }
1437 #endif /* ALTQ */
1438
1439 static void
1440 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1441     struct pf_pooladdr *pool)
1442 {
1443
1444         bzero(pool, sizeof(*pool));
1445         bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1446         strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1447 }
1448
1449 static void
1450 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1451     struct pf_kpooladdr *kpool)
1452 {
1453
1454         bzero(kpool, sizeof(*kpool));
1455         bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1456         strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1457 }
1458
1459 static void
1460 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1461 {
1462
1463         bzero(rule, sizeof(*rule));
1464
1465         bcopy(&krule->src, &rule->src, sizeof(rule->src));
1466         bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1467
1468         for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1469                 if (rule->skip[i].ptr == NULL)
1470                         rule->skip[i].nr = -1;
1471                 else
1472                         rule->skip[i].nr = krule->skip[i].ptr->nr;
1473         }
1474
1475         strlcpy(rule->label, krule->label, sizeof(rule->label));
1476         strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1477         strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1478         strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1479         strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1480         strlcpy(rule->match_tagname, krule->match_tagname,
1481             sizeof(rule->match_tagname));
1482         strlcpy(rule->overload_tblname, krule->overload_tblname,
1483             sizeof(rule->overload_tblname));
1484
1485         bcopy(&krule->rpool, &rule->rpool, sizeof(krule->rpool));
1486
1487         rule->evaluations = counter_u64_fetch(krule->evaluations);
1488         for (int i = 0; i < 2; i++) {
1489                 rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1490                 rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1491         }
1492
1493         /* kif, anchor, overload_tbl are not copied over. */
1494
1495         rule->os_fingerprint = krule->os_fingerprint;
1496
1497         rule->rtableid = krule->rtableid;
1498         bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1499         rule->max_states = krule->max_states;
1500         rule->max_src_nodes = krule->max_src_nodes;
1501         rule->max_src_states = krule->max_src_states;
1502         rule->max_src_conn = krule->max_src_conn;
1503         rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1504         rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1505         rule->qid = krule->qid;
1506         rule->pqid = krule->pqid;
1507         rule->rt_listid = krule->rt_listid;
1508         rule->nr = krule->nr;
1509         rule->prob = krule->prob;
1510         rule->cuid = krule->cuid;
1511         rule->cpid = krule->cpid;
1512
1513         rule->return_icmp = krule->return_icmp;
1514         rule->return_icmp6 = krule->return_icmp6;
1515         rule->max_mss = krule->max_mss;
1516         rule->tag = krule->tag;
1517         rule->match_tag = krule->match_tag;
1518         rule->scrub_flags = krule->scrub_flags;
1519
1520         bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1521         bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1522
1523         rule->rule_flag = krule->rule_flag;
1524         rule->action = krule->action;
1525         rule->direction = krule->direction;
1526         rule->log = krule->log;
1527         rule->logif = krule->logif;
1528         rule->quick = krule->quick;
1529         rule->ifnot = krule->ifnot;
1530         rule->match_tag_not = krule->match_tag_not;
1531         rule->natpass = krule->natpass;
1532
1533         rule->keep_state = krule->keep_state;
1534         rule->af = krule->af;
1535         rule->proto = krule->proto;
1536         rule->type = krule->type;
1537         rule->code = krule->code;
1538         rule->flags = krule->flags;
1539         rule->flagset = krule->flagset;
1540         rule->min_ttl = krule->min_ttl;
1541         rule->allow_opts = krule->allow_opts;
1542         rule->rt = krule->rt;
1543         rule->return_ttl = krule->return_ttl;
1544         rule->tos = krule->tos;
1545         rule->set_tos = krule->set_tos;
1546         rule->anchor_relative = krule->anchor_relative;
1547         rule->anchor_wildcard = krule->anchor_wildcard;
1548
1549         rule->flush = krule->flush;
1550         rule->prio = krule->prio;
1551         rule->set_prio[0] = krule->set_prio[0];
1552         rule->set_prio[1] = krule->set_prio[1];
1553
1554         bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1555
1556         rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1557         rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1558         rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1559 }
1560
1561 static void
1562 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1563 {
1564
1565         bzero(krule, sizeof(*krule));
1566
1567         bcopy(&rule->src, &krule->src, sizeof(rule->src));
1568         bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1569
1570         strlcpy(krule->label, rule->label, sizeof(rule->label));
1571         strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1572         strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1573         strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1574         strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1575         strlcpy(krule->match_tagname, rule->match_tagname,
1576             sizeof(rule->match_tagname));
1577         strlcpy(krule->overload_tblname, rule->overload_tblname,
1578             sizeof(rule->overload_tblname));
1579
1580         bcopy(&rule->rpool, &krule->rpool, sizeof(krule->rpool));
1581
1582         /* Don't allow userspace to set evaulations, packets or bytes. */
1583         /* kif, anchor, overload_tbl are not copied over. */
1584
1585         krule->os_fingerprint = krule->os_fingerprint;
1586
1587         krule->rtableid = rule->rtableid;
1588         bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1589         krule->max_states = rule->max_states;
1590         krule->max_src_nodes = rule->max_src_nodes;
1591         krule->max_src_states = rule->max_src_states;
1592         krule->max_src_conn = rule->max_src_conn;
1593         krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1594         krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1595         krule->qid = rule->qid;
1596         krule->pqid = rule->pqid;
1597         krule->rt_listid = rule->rt_listid;
1598         krule->nr = rule->nr;
1599         krule->prob = rule->prob;
1600         krule->cuid = rule->cuid;
1601         krule->cpid = rule->cpid;
1602
1603         krule->return_icmp = rule->return_icmp;
1604         krule->return_icmp6 = rule->return_icmp6;
1605         krule->max_mss = rule->max_mss;
1606         krule->tag = rule->tag;
1607         krule->match_tag = rule->match_tag;
1608         krule->scrub_flags = rule->scrub_flags;
1609
1610         bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1611         bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1612
1613         krule->rule_flag = rule->rule_flag;
1614         krule->action = rule->action;
1615         krule->direction = rule->direction;
1616         krule->log = rule->log;
1617         krule->logif = rule->logif;
1618         krule->quick = rule->quick;
1619         krule->ifnot = rule->ifnot;
1620         krule->match_tag_not = rule->match_tag_not;
1621         krule->natpass = rule->natpass;
1622
1623         krule->keep_state = rule->keep_state;
1624         krule->af = rule->af;
1625         krule->proto = rule->proto;
1626         krule->type = rule->type;
1627         krule->code = rule->code;
1628         krule->flags = rule->flags;
1629         krule->flagset = rule->flagset;
1630         krule->min_ttl = rule->min_ttl;
1631         krule->allow_opts = rule->allow_opts;
1632         krule->rt = rule->rt;
1633         krule->return_ttl = rule->return_ttl;
1634         krule->tos = rule->tos;
1635         krule->set_tos = rule->set_tos;
1636         krule->anchor_relative = rule->anchor_relative;
1637         krule->anchor_wildcard = rule->anchor_wildcard;
1638
1639         krule->flush = rule->flush;
1640         krule->prio = rule->prio;
1641         krule->set_prio[0] = rule->set_prio[0];
1642         krule->set_prio[1] = rule->set_prio[1];
1643
1644         bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1645 }
1646
1647 static int
1648 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1649 {
1650         int                      error = 0;
1651         PF_RULES_RLOCK_TRACKER;
1652
1653         /* XXX keep in sync with switch() below */
1654         if (securelevel_gt(td->td_ucred, 2))
1655                 switch (cmd) {
1656                 case DIOCGETRULES:
1657                 case DIOCGETRULE:
1658                 case DIOCGETADDRS:
1659                 case DIOCGETADDR:
1660                 case DIOCGETSTATE:
1661                 case DIOCSETSTATUSIF:
1662                 case DIOCGETSTATUS:
1663                 case DIOCCLRSTATUS:
1664                 case DIOCNATLOOK:
1665                 case DIOCSETDEBUG:
1666                 case DIOCGETSTATES:
1667                 case DIOCGETTIMEOUT:
1668                 case DIOCCLRRULECTRS:
1669                 case DIOCGETLIMIT:
1670                 case DIOCGETALTQSV0:
1671                 case DIOCGETALTQSV1:
1672                 case DIOCGETALTQV0:
1673                 case DIOCGETALTQV1:
1674                 case DIOCGETQSTATSV0:
1675                 case DIOCGETQSTATSV1:
1676                 case DIOCGETRULESETS:
1677                 case DIOCGETRULESET:
1678                 case DIOCRGETTABLES:
1679                 case DIOCRGETTSTATS:
1680                 case DIOCRCLRTSTATS:
1681                 case DIOCRCLRADDRS:
1682                 case DIOCRADDADDRS:
1683                 case DIOCRDELADDRS:
1684                 case DIOCRSETADDRS:
1685                 case DIOCRGETADDRS:
1686                 case DIOCRGETASTATS:
1687                 case DIOCRCLRASTATS:
1688                 case DIOCRTSTADDRS:
1689                 case DIOCOSFPGET:
1690                 case DIOCGETSRCNODES:
1691                 case DIOCCLRSRCNODES:
1692                 case DIOCIGETIFACES:
1693                 case DIOCGIFSPEEDV0:
1694                 case DIOCGIFSPEEDV1:
1695                 case DIOCSETIFFLAG:
1696                 case DIOCCLRIFFLAG:
1697                         break;
1698                 case DIOCRCLRTABLES:
1699                 case DIOCRADDTABLES:
1700                 case DIOCRDELTABLES:
1701                 case DIOCRSETTFLAGS:
1702                         if (((struct pfioc_table *)addr)->pfrio_flags &
1703                             PFR_FLAG_DUMMY)
1704                                 break; /* dummy operation ok */
1705                         return (EPERM);
1706                 default:
1707                         return (EPERM);
1708                 }
1709
1710         if (!(flags & FWRITE))
1711                 switch (cmd) {
1712                 case DIOCGETRULES:
1713                 case DIOCGETADDRS:
1714                 case DIOCGETADDR:
1715                 case DIOCGETSTATE:
1716                 case DIOCGETSTATUS:
1717                 case DIOCGETSTATES:
1718                 case DIOCGETTIMEOUT:
1719                 case DIOCGETLIMIT:
1720                 case DIOCGETALTQSV0:
1721                 case DIOCGETALTQSV1:
1722                 case DIOCGETALTQV0:
1723                 case DIOCGETALTQV1:
1724                 case DIOCGETQSTATSV0:
1725                 case DIOCGETQSTATSV1:
1726                 case DIOCGETRULESETS:
1727                 case DIOCGETRULESET:
1728                 case DIOCNATLOOK:
1729                 case DIOCRGETTABLES:
1730                 case DIOCRGETTSTATS:
1731                 case DIOCRGETADDRS:
1732                 case DIOCRGETASTATS:
1733                 case DIOCRTSTADDRS:
1734                 case DIOCOSFPGET:
1735                 case DIOCGETSRCNODES:
1736                 case DIOCIGETIFACES:
1737                 case DIOCGIFSPEEDV1:
1738                 case DIOCGIFSPEEDV0:
1739                         break;
1740                 case DIOCRCLRTABLES:
1741                 case DIOCRADDTABLES:
1742                 case DIOCRDELTABLES:
1743                 case DIOCRCLRTSTATS:
1744                 case DIOCRCLRADDRS:
1745                 case DIOCRADDADDRS:
1746                 case DIOCRDELADDRS:
1747                 case DIOCRSETADDRS:
1748                 case DIOCRSETTFLAGS:
1749                         if (((struct pfioc_table *)addr)->pfrio_flags &
1750                             PFR_FLAG_DUMMY) {
1751                                 flags |= FWRITE; /* need write lock for dummy */
1752                                 break; /* dummy operation ok */
1753                         }
1754                         return (EACCES);
1755                 case DIOCGETRULE:
1756                         if (((struct pfioc_rule *)addr)->action ==
1757                             PF_GET_CLR_CNTR)
1758                                 return (EACCES);
1759                         break;
1760                 default:
1761                         return (EACCES);
1762                 }
1763
1764         CURVNET_SET(TD_TO_VNET(td));
1765
1766         switch (cmd) {
1767         case DIOCSTART:
1768                 sx_xlock(&pf_ioctl_lock);
1769                 if (V_pf_status.running)
1770                         error = EEXIST;
1771                 else {
1772                         int cpu;
1773
1774                         error = hook_pf();
1775                         if (error) {
1776                                 DPFPRINTF(PF_DEBUG_MISC,
1777                                     ("pf: pfil registration failed\n"));
1778                                 break;
1779                         }
1780                         V_pf_status.running = 1;
1781                         V_pf_status.since = time_second;
1782
1783                         CPU_FOREACH(cpu)
1784                                 V_pf_stateid[cpu] = time_second;
1785
1786                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1787                 }
1788                 break;
1789
1790         case DIOCSTOP:
1791                 sx_xlock(&pf_ioctl_lock);
1792                 if (!V_pf_status.running)
1793                         error = ENOENT;
1794                 else {
1795                         V_pf_status.running = 0;
1796                         error = dehook_pf();
1797                         if (error) {
1798                                 V_pf_status.running = 1;
1799                                 DPFPRINTF(PF_DEBUG_MISC,
1800                                     ("pf: pfil unregistration failed\n"));
1801                         }
1802                         V_pf_status.since = time_second;
1803                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1804                 }
1805                 break;
1806
1807         case DIOCADDRULE: {
1808                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1809                 struct pf_kruleset      *ruleset;
1810                 struct pf_krule         *rule, *tail;
1811                 struct pf_kpooladdr     *pa;
1812                 struct pfi_kkif         *kif = NULL;
1813                 int                      rs_num;
1814
1815                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1816                         error = EINVAL;
1817                         break;
1818                 }
1819                 if (pr->rule.src.addr.p.dyn != NULL ||
1820                     pr->rule.dst.addr.p.dyn != NULL) {
1821                         error = EINVAL;
1822                         break;
1823                 }
1824 #ifndef INET
1825                 if (pr->rule.af == AF_INET) {
1826                         error = EAFNOSUPPORT;
1827                         break;
1828                 }
1829 #endif /* INET */
1830 #ifndef INET6
1831                 if (pr->rule.af == AF_INET6) {
1832                         error = EAFNOSUPPORT;
1833                         break;
1834                 }
1835 #endif /* INET6 */
1836
1837                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1838                 pf_rule_to_krule(&pr->rule, rule);
1839
1840                 if (rule->ifname[0])
1841                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1842                 rule->evaluations = counter_u64_alloc(M_WAITOK);
1843                 for (int i = 0; i < 2; i++) {
1844                         rule->packets[i] = counter_u64_alloc(M_WAITOK);
1845                         rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1846                 }
1847                 rule->states_cur = counter_u64_alloc(M_WAITOK);
1848                 rule->states_tot = counter_u64_alloc(M_WAITOK);
1849                 rule->src_nodes = counter_u64_alloc(M_WAITOK);
1850                 rule->cuid = td->td_ucred->cr_ruid;
1851                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1852                 TAILQ_INIT(&rule->rpool.list);
1853
1854 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1855
1856                 PF_RULES_WLOCK();
1857                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1858                 ruleset = pf_find_kruleset(pr->anchor);
1859                 if (ruleset == NULL)
1860                         ERROUT(EINVAL);
1861                 rs_num = pf_get_ruleset_number(pr->rule.action);
1862                 if (rs_num >= PF_RULESET_MAX)
1863                         ERROUT(EINVAL);
1864                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1865                         DPFPRINTF(PF_DEBUG_MISC,
1866                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1867                             ruleset->rules[rs_num].inactive.ticket));
1868                         ERROUT(EBUSY);
1869                 }
1870                 if (pr->pool_ticket != V_ticket_pabuf) {
1871                         DPFPRINTF(PF_DEBUG_MISC,
1872                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
1873                             V_ticket_pabuf));
1874                         ERROUT(EBUSY);
1875                 }
1876
1877                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1878                     pf_krulequeue);
1879                 if (tail)
1880                         rule->nr = tail->nr + 1;
1881                 else
1882                         rule->nr = 0;
1883                 if (rule->ifname[0]) {
1884                         rule->kif = pfi_kkif_attach(kif, rule->ifname);
1885                         pfi_kkif_ref(rule->kif);
1886                 } else
1887                         rule->kif = NULL;
1888
1889                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1890                         error = EBUSY;
1891
1892 #ifdef ALTQ
1893                 /* set queue IDs */
1894                 if (rule->qname[0] != 0) {
1895                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1896                                 error = EBUSY;
1897                         else if (rule->pqname[0] != 0) {
1898                                 if ((rule->pqid =
1899                                     pf_qname2qid(rule->pqname)) == 0)
1900                                         error = EBUSY;
1901                         } else
1902                                 rule->pqid = rule->qid;
1903                 }
1904 #endif
1905                 if (rule->tagname[0])
1906                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1907                                 error = EBUSY;
1908                 if (rule->match_tagname[0])
1909                         if ((rule->match_tag =
1910                             pf_tagname2tag(rule->match_tagname)) == 0)
1911                                 error = EBUSY;
1912                 if (rule->rt && !rule->direction)
1913                         error = EINVAL;
1914                 if (!rule->log)
1915                         rule->logif = 0;
1916                 if (rule->logif >= PFLOGIFS_MAX)
1917                         error = EINVAL;
1918                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1919                         error = ENOMEM;
1920                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1921                         error = ENOMEM;
1922                 if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
1923                         error = EINVAL;
1924                 if (rule->scrub_flags & PFSTATE_SETPRIO &&
1925                     (rule->set_prio[0] > PF_PRIO_MAX ||
1926                     rule->set_prio[1] > PF_PRIO_MAX))
1927                         error = EINVAL;
1928                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1929                         if (pa->addr.type == PF_ADDR_TABLE) {
1930                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
1931                                     pa->addr.v.tblname);
1932                                 if (pa->addr.p.tbl == NULL)
1933                                         error = ENOMEM;
1934                         }
1935
1936                 rule->overload_tbl = NULL;
1937                 if (rule->overload_tblname[0]) {
1938                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1939                             rule->overload_tblname)) == NULL)
1940                                 error = EINVAL;
1941                         else
1942                                 rule->overload_tbl->pfrkt_flags |=
1943                                     PFR_TFLAG_ACTIVE;
1944                 }
1945
1946                 pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
1947                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1948                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1949                     (rule->rt > PF_NOPFROUTE)) &&
1950                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1951                         error = EINVAL;
1952
1953                 if (error) {
1954                         pf_free_rule(rule);
1955                         PF_RULES_WUNLOCK();
1956                         break;
1957                 }
1958
1959                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1960                 counter_u64_zero(rule->evaluations);
1961                 for (int i = 0; i < 2; i++) {
1962                         counter_u64_zero(rule->packets[i]);
1963                         counter_u64_zero(rule->bytes[i]);
1964                 }
1965                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1966                     rule, entries);
1967                 ruleset->rules[rs_num].inactive.rcount++;
1968                 PF_RULES_WUNLOCK();
1969                 break;
1970
1971 #undef ERROUT
1972 DIOCADDRULE_error:
1973                 PF_RULES_WUNLOCK();
1974                 counter_u64_free(rule->evaluations);
1975                 for (int i = 0; i < 2; i++) {
1976                         counter_u64_free(rule->packets[i]);
1977                         counter_u64_free(rule->bytes[i]);
1978                 }
1979                 counter_u64_free(rule->states_cur);
1980                 counter_u64_free(rule->states_tot);
1981                 counter_u64_free(rule->src_nodes);
1982                 free(rule, M_PFRULE);
1983                 if (kif)
1984                         free(kif, PFI_MTYPE);
1985                 break;
1986         }
1987
1988         case DIOCGETRULES: {
1989                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1990                 struct pf_kruleset      *ruleset;
1991                 struct pf_krule         *tail;
1992                 int                      rs_num;
1993
1994                 PF_RULES_WLOCK();
1995                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1996                 ruleset = pf_find_kruleset(pr->anchor);
1997                 if (ruleset == NULL) {
1998                         PF_RULES_WUNLOCK();
1999                         error = EINVAL;
2000                         break;
2001                 }
2002                 rs_num = pf_get_ruleset_number(pr->rule.action);
2003                 if (rs_num >= PF_RULESET_MAX) {
2004                         PF_RULES_WUNLOCK();
2005                         error = EINVAL;
2006                         break;
2007                 }
2008                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2009                     pf_krulequeue);
2010                 if (tail)
2011                         pr->nr = tail->nr + 1;
2012                 else
2013                         pr->nr = 0;
2014                 pr->ticket = ruleset->rules[rs_num].active.ticket;
2015                 PF_RULES_WUNLOCK();
2016                 break;
2017         }
2018
2019         case DIOCGETRULE: {
2020                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
2021                 struct pf_kruleset      *ruleset;
2022                 struct pf_krule         *rule;
2023                 int                      rs_num;
2024
2025                 PF_RULES_WLOCK();
2026                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
2027                 ruleset = pf_find_kruleset(pr->anchor);
2028                 if (ruleset == NULL) {
2029                         PF_RULES_WUNLOCK();
2030                         error = EINVAL;
2031                         break;
2032                 }
2033                 rs_num = pf_get_ruleset_number(pr->rule.action);
2034                 if (rs_num >= PF_RULESET_MAX) {
2035                         PF_RULES_WUNLOCK();
2036                         error = EINVAL;
2037                         break;
2038                 }
2039                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2040                         PF_RULES_WUNLOCK();
2041                         error = EBUSY;
2042                         break;
2043                 }
2044                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2045                 while ((rule != NULL) && (rule->nr != pr->nr))
2046                         rule = TAILQ_NEXT(rule, entries);
2047                 if (rule == NULL) {
2048                         PF_RULES_WUNLOCK();
2049                         error = EBUSY;
2050                         break;
2051                 }
2052
2053                 pf_krule_to_rule(rule, &pr->rule);
2054
2055                 if (pf_kanchor_copyout(ruleset, rule, pr)) {
2056                         PF_RULES_WUNLOCK();
2057                         error = EBUSY;
2058                         break;
2059                 }
2060                 pf_addr_copyout(&pr->rule.src.addr);
2061                 pf_addr_copyout(&pr->rule.dst.addr);
2062
2063                 if (pr->action == PF_GET_CLR_CNTR) {
2064                         counter_u64_zero(rule->evaluations);
2065                         for (int i = 0; i < 2; i++) {
2066                                 counter_u64_zero(rule->packets[i]);
2067                                 counter_u64_zero(rule->bytes[i]);
2068                         }
2069                         counter_u64_zero(rule->states_tot);
2070                 }
2071                 PF_RULES_WUNLOCK();
2072                 break;
2073         }
2074
2075         case DIOCCHANGERULE: {
2076                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
2077                 struct pf_kruleset      *ruleset;
2078                 struct pf_krule         *oldrule = NULL, *newrule = NULL;
2079                 struct pfi_kkif         *kif = NULL;
2080                 struct pf_kpooladdr     *pa;
2081                 u_int32_t                nr = 0;
2082                 int                      rs_num;
2083
2084                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
2085                     pcr->action > PF_CHANGE_GET_TICKET) {
2086                         error = EINVAL;
2087                         break;
2088                 }
2089                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2090                         error = EINVAL;
2091                         break;
2092                 }
2093
2094                 if (pcr->action != PF_CHANGE_REMOVE) {
2095 #ifndef INET
2096                         if (pcr->rule.af == AF_INET) {
2097                                 error = EAFNOSUPPORT;
2098                                 break;
2099                         }
2100 #endif /* INET */
2101 #ifndef INET6
2102                         if (pcr->rule.af == AF_INET6) {
2103                                 error = EAFNOSUPPORT;
2104                                 break;
2105                         }
2106 #endif /* INET6 */
2107                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2108                         pf_rule_to_krule(&pcr->rule, newrule);
2109
2110                         if (newrule->ifname[0])
2111                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2112                         newrule->evaluations = counter_u64_alloc(M_WAITOK);
2113                         for (int i = 0; i < 2; i++) {
2114                                 newrule->packets[i] =
2115                                     counter_u64_alloc(M_WAITOK);
2116                                 newrule->bytes[i] =
2117                                     counter_u64_alloc(M_WAITOK);
2118                         }
2119                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
2120                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
2121                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2122                         newrule->cuid = td->td_ucred->cr_ruid;
2123                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2124                         TAILQ_INIT(&newrule->rpool.list);
2125                 }
2126
2127 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
2128
2129                 PF_RULES_WLOCK();
2130                 if (!(pcr->action == PF_CHANGE_REMOVE ||
2131                     pcr->action == PF_CHANGE_GET_TICKET) &&
2132                     pcr->pool_ticket != V_ticket_pabuf)
2133                         ERROUT(EBUSY);
2134
2135                 ruleset = pf_find_kruleset(pcr->anchor);
2136                 if (ruleset == NULL)
2137                         ERROUT(EINVAL);
2138
2139                 rs_num = pf_get_ruleset_number(pcr->rule.action);
2140                 if (rs_num >= PF_RULESET_MAX)
2141                         ERROUT(EINVAL);
2142
2143                 if (pcr->action == PF_CHANGE_GET_TICKET) {
2144                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2145                         ERROUT(0);
2146                 } else if (pcr->ticket !=
2147                             ruleset->rules[rs_num].active.ticket)
2148                                 ERROUT(EINVAL);
2149
2150                 if (pcr->action != PF_CHANGE_REMOVE) {
2151                         if (newrule->ifname[0]) {
2152                                 newrule->kif = pfi_kkif_attach(kif,
2153                                     newrule->ifname);
2154                                 pfi_kkif_ref(newrule->kif);
2155                         } else
2156                                 newrule->kif = NULL;
2157
2158                         if (newrule->rtableid > 0 &&
2159                             newrule->rtableid >= rt_numfibs)
2160                                 error = EBUSY;
2161
2162 #ifdef ALTQ
2163                         /* set queue IDs */
2164                         if (newrule->qname[0] != 0) {
2165                                 if ((newrule->qid =
2166                                     pf_qname2qid(newrule->qname)) == 0)
2167                                         error = EBUSY;
2168                                 else if (newrule->pqname[0] != 0) {
2169                                         if ((newrule->pqid =
2170                                             pf_qname2qid(newrule->pqname)) == 0)
2171                                                 error = EBUSY;
2172                                 } else
2173                                         newrule->pqid = newrule->qid;
2174                         }
2175 #endif /* ALTQ */
2176                         if (newrule->tagname[0])
2177                                 if ((newrule->tag =
2178                                     pf_tagname2tag(newrule->tagname)) == 0)
2179                                         error = EBUSY;
2180                         if (newrule->match_tagname[0])
2181                                 if ((newrule->match_tag = pf_tagname2tag(
2182                                     newrule->match_tagname)) == 0)
2183                                         error = EBUSY;
2184                         if (newrule->rt && !newrule->direction)
2185                                 error = EINVAL;
2186                         if (!newrule->log)
2187                                 newrule->logif = 0;
2188                         if (newrule->logif >= PFLOGIFS_MAX)
2189                                 error = EINVAL;
2190                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2191                                 error = ENOMEM;
2192                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2193                                 error = ENOMEM;
2194                         if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2195                                 error = EINVAL;
2196                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2197                                 if (pa->addr.type == PF_ADDR_TABLE) {
2198                                         pa->addr.p.tbl =
2199                                             pfr_attach_table(ruleset,
2200                                             pa->addr.v.tblname);
2201                                         if (pa->addr.p.tbl == NULL)
2202                                                 error = ENOMEM;
2203                                 }
2204
2205                         newrule->overload_tbl = NULL;
2206                         if (newrule->overload_tblname[0]) {
2207                                 if ((newrule->overload_tbl = pfr_attach_table(
2208                                     ruleset, newrule->overload_tblname)) ==
2209                                     NULL)
2210                                         error = EINVAL;
2211                                 else
2212                                         newrule->overload_tbl->pfrkt_flags |=
2213                                             PFR_TFLAG_ACTIVE;
2214                         }
2215
2216                         pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2217                         if (((((newrule->action == PF_NAT) ||
2218                             (newrule->action == PF_RDR) ||
2219                             (newrule->action == PF_BINAT) ||
2220                             (newrule->rt > PF_NOPFROUTE)) &&
2221                             !newrule->anchor)) &&
2222                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2223                                 error = EINVAL;
2224
2225                         if (error) {
2226                                 pf_free_rule(newrule);
2227                                 PF_RULES_WUNLOCK();
2228                                 break;
2229                         }
2230
2231                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2232                 }
2233                 pf_empty_kpool(&V_pf_pabuf);
2234
2235                 if (pcr->action == PF_CHANGE_ADD_HEAD)
2236                         oldrule = TAILQ_FIRST(
2237                             ruleset->rules[rs_num].active.ptr);
2238                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
2239                         oldrule = TAILQ_LAST(
2240                             ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2241                 else {
2242                         oldrule = TAILQ_FIRST(
2243                             ruleset->rules[rs_num].active.ptr);
2244                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2245                                 oldrule = TAILQ_NEXT(oldrule, entries);
2246                         if (oldrule == NULL) {
2247                                 if (newrule != NULL)
2248                                         pf_free_rule(newrule);
2249                                 PF_RULES_WUNLOCK();
2250                                 error = EINVAL;
2251                                 break;
2252                         }
2253                 }
2254
2255                 if (pcr->action == PF_CHANGE_REMOVE) {
2256                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2257                             oldrule);
2258                         ruleset->rules[rs_num].active.rcount--;
2259                 } else {
2260                         if (oldrule == NULL)
2261                                 TAILQ_INSERT_TAIL(
2262                                     ruleset->rules[rs_num].active.ptr,
2263                                     newrule, entries);
2264                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2265                             pcr->action == PF_CHANGE_ADD_BEFORE)
2266                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2267                         else
2268                                 TAILQ_INSERT_AFTER(
2269                                     ruleset->rules[rs_num].active.ptr,
2270                                     oldrule, newrule, entries);
2271                         ruleset->rules[rs_num].active.rcount++;
2272                 }
2273
2274                 nr = 0;
2275                 TAILQ_FOREACH(oldrule,
2276                     ruleset->rules[rs_num].active.ptr, entries)
2277                         oldrule->nr = nr++;
2278
2279                 ruleset->rules[rs_num].active.ticket++;
2280
2281                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2282                 pf_remove_if_empty_kruleset(ruleset);
2283
2284                 PF_RULES_WUNLOCK();
2285                 break;
2286
2287 #undef ERROUT
2288 DIOCCHANGERULE_error:
2289                 PF_RULES_WUNLOCK();
2290                 if (newrule != NULL) {
2291                         counter_u64_free(newrule->evaluations);
2292                         for (int i = 0; i < 2; i++) {
2293                                 counter_u64_free(newrule->packets[i]);
2294                                 counter_u64_free(newrule->bytes[i]);
2295                         }
2296                         counter_u64_free(newrule->states_cur);
2297                         counter_u64_free(newrule->states_tot);
2298                         counter_u64_free(newrule->src_nodes);
2299                         free(newrule, M_PFRULE);
2300                 }
2301                 if (kif != NULL)
2302                         free(kif, PFI_MTYPE);
2303                 break;
2304         }
2305
2306         case DIOCCLRSTATES: {
2307                 struct pf_state         *s;
2308                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2309                 u_int                    i, killed = 0;
2310
2311                 for (i = 0; i <= pf_hashmask; i++) {
2312                         struct pf_idhash *ih = &V_pf_idhash[i];
2313
2314 relock_DIOCCLRSTATES:
2315                         PF_HASHROW_LOCK(ih);
2316                         LIST_FOREACH(s, &ih->states, entry)
2317                                 if (!psk->psk_ifname[0] ||
2318                                     !strcmp(psk->psk_ifname,
2319                                     s->kif->pfik_name)) {
2320                                         /*
2321                                          * Don't send out individual
2322                                          * delete messages.
2323                                          */
2324                                         s->state_flags |= PFSTATE_NOSYNC;
2325                                         pf_unlink_state(s, PF_ENTER_LOCKED);
2326                                         killed++;
2327                                         goto relock_DIOCCLRSTATES;
2328                                 }
2329                         PF_HASHROW_UNLOCK(ih);
2330                 }
2331                 psk->psk_killed = killed;
2332                 if (V_pfsync_clear_states_ptr != NULL)
2333                         V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2334                 break;
2335         }
2336
2337         case DIOCKILLSTATES: {
2338                 struct pf_state         *s;
2339                 struct pf_state_key     *sk;
2340                 struct pf_addr          *srcaddr, *dstaddr;
2341                 u_int16_t                srcport, dstport;
2342                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2343                 u_int                    i, killed = 0;
2344
2345                 if (psk->psk_pfcmp.id) {
2346                         if (psk->psk_pfcmp.creatorid == 0)
2347                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2348                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2349                             psk->psk_pfcmp.creatorid))) {
2350                                 pf_unlink_state(s, PF_ENTER_LOCKED);
2351                                 psk->psk_killed = 1;
2352                         }
2353                         break;
2354                 }
2355
2356                 for (i = 0; i <= pf_hashmask; i++) {
2357                         struct pf_idhash *ih = &V_pf_idhash[i];
2358
2359 relock_DIOCKILLSTATES:
2360                         PF_HASHROW_LOCK(ih);
2361                         LIST_FOREACH(s, &ih->states, entry) {
2362                                 sk = s->key[PF_SK_WIRE];
2363                                 if (s->direction == PF_OUT) {
2364                                         srcaddr = &sk->addr[1];
2365                                         dstaddr = &sk->addr[0];
2366                                         srcport = sk->port[1];
2367                                         dstport = sk->port[0];
2368                                 } else {
2369                                         srcaddr = &sk->addr[0];
2370                                         dstaddr = &sk->addr[1];
2371                                         srcport = sk->port[0];
2372                                         dstport = sk->port[1];
2373                                 }
2374
2375                                 if ((!psk->psk_af || sk->af == psk->psk_af)
2376                                     && (!psk->psk_proto || psk->psk_proto ==
2377                                     sk->proto) &&
2378                                     PF_MATCHA(psk->psk_src.neg,
2379                                     &psk->psk_src.addr.v.a.addr,
2380                                     &psk->psk_src.addr.v.a.mask,
2381                                     srcaddr, sk->af) &&
2382                                     PF_MATCHA(psk->psk_dst.neg,
2383                                     &psk->psk_dst.addr.v.a.addr,
2384                                     &psk->psk_dst.addr.v.a.mask,
2385                                     dstaddr, sk->af) &&
2386                                     (psk->psk_src.port_op == 0 ||
2387                                     pf_match_port(psk->psk_src.port_op,
2388                                     psk->psk_src.port[0], psk->psk_src.port[1],
2389                                     srcport)) &&
2390                                     (psk->psk_dst.port_op == 0 ||
2391                                     pf_match_port(psk->psk_dst.port_op,
2392                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
2393                                     dstport)) &&
2394                                     (!psk->psk_label[0] ||
2395                                     (s->rule.ptr->label[0] &&
2396                                     !strcmp(psk->psk_label,
2397                                     s->rule.ptr->label))) &&
2398                                     (!psk->psk_ifname[0] ||
2399                                     !strcmp(psk->psk_ifname,
2400                                     s->kif->pfik_name))) {
2401                                         pf_unlink_state(s, PF_ENTER_LOCKED);
2402                                         killed++;
2403                                         goto relock_DIOCKILLSTATES;
2404                                 }
2405                         }
2406                         PF_HASHROW_UNLOCK(ih);
2407                 }
2408                 psk->psk_killed = killed;
2409                 break;
2410         }
2411
2412         case DIOCADDSTATE: {
2413                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
2414                 struct pfsync_state     *sp = &ps->state;
2415
2416                 if (sp->timeout >= PFTM_MAX) {
2417                         error = EINVAL;
2418                         break;
2419                 }
2420                 if (V_pfsync_state_import_ptr != NULL) {
2421                         PF_RULES_RLOCK();
2422                         error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2423                         PF_RULES_RUNLOCK();
2424                 } else
2425                         error = EOPNOTSUPP;
2426                 break;
2427         }
2428
2429         case DIOCGETSTATE: {
2430                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
2431                 struct pf_state         *s;
2432
2433                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2434                 if (s == NULL) {
2435                         error = ENOENT;
2436                         break;
2437                 }
2438
2439                 pfsync_state_export(&ps->state, s);
2440                 PF_STATE_UNLOCK(s);
2441                 break;
2442         }
2443
2444         case DIOCGETSTATES: {
2445                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
2446                 struct pf_state         *s;
2447                 struct pfsync_state     *pstore, *p;
2448                 int i, nr;
2449
2450                 if (ps->ps_len <= 0) {
2451                         nr = uma_zone_get_cur(V_pf_state_z);
2452                         ps->ps_len = sizeof(struct pfsync_state) * nr;
2453                         break;
2454                 }
2455
2456                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2457                 nr = 0;
2458
2459                 for (i = 0; i <= pf_hashmask; i++) {
2460                         struct pf_idhash *ih = &V_pf_idhash[i];
2461
2462                         PF_HASHROW_LOCK(ih);
2463                         LIST_FOREACH(s, &ih->states, entry) {
2464
2465                                 if (s->timeout == PFTM_UNLINKED)
2466                                         continue;
2467
2468                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
2469                                         PF_HASHROW_UNLOCK(ih);
2470                                         goto DIOCGETSTATES_full;
2471                                 }
2472                                 pfsync_state_export(p, s);
2473                                 p++;
2474                                 nr++;
2475                         }
2476                         PF_HASHROW_UNLOCK(ih);
2477                 }
2478 DIOCGETSTATES_full:
2479                 error = copyout(pstore, ps->ps_states,
2480                     sizeof(struct pfsync_state) * nr);
2481                 if (error) {
2482                         free(pstore, M_TEMP);
2483                         break;
2484                 }
2485                 ps->ps_len = sizeof(struct pfsync_state) * nr;
2486                 free(pstore, M_TEMP);
2487
2488                 break;
2489         }
2490
2491         case DIOCGETSTATUS: {
2492                 struct pf_status *s = (struct pf_status *)addr;
2493
2494                 PF_RULES_RLOCK();
2495                 s->running = V_pf_status.running;
2496                 s->since   = V_pf_status.since;
2497                 s->debug   = V_pf_status.debug;
2498                 s->hostid  = V_pf_status.hostid;
2499                 s->states  = V_pf_status.states;
2500                 s->src_nodes = V_pf_status.src_nodes;
2501
2502                 for (int i = 0; i < PFRES_MAX; i++)
2503                         s->counters[i] =
2504                             counter_u64_fetch(V_pf_status.counters[i]);
2505                 for (int i = 0; i < LCNT_MAX; i++)
2506                         s->lcounters[i] =
2507                             counter_u64_fetch(V_pf_status.lcounters[i]);
2508                 for (int i = 0; i < FCNT_MAX; i++)
2509                         s->fcounters[i] =
2510                             counter_u64_fetch(V_pf_status.fcounters[i]);
2511                 for (int i = 0; i < SCNT_MAX; i++)
2512                         s->scounters[i] =
2513                             counter_u64_fetch(V_pf_status.scounters[i]);
2514
2515                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2516                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2517                     PF_MD5_DIGEST_LENGTH);
2518
2519                 pfi_update_status(s->ifname, s);
2520                 PF_RULES_RUNLOCK();
2521                 break;
2522         }
2523
2524         case DIOCSETSTATUSIF: {
2525                 struct pfioc_if *pi = (struct pfioc_if *)addr;
2526
2527                 if (pi->ifname[0] == 0) {
2528                         bzero(V_pf_status.ifname, IFNAMSIZ);
2529                         break;
2530                 }
2531                 PF_RULES_WLOCK();
2532                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2533                 PF_RULES_WUNLOCK();
2534                 break;
2535         }
2536
2537         case DIOCCLRSTATUS: {
2538                 PF_RULES_WLOCK();
2539                 for (int i = 0; i < PFRES_MAX; i++)
2540                         counter_u64_zero(V_pf_status.counters[i]);
2541                 for (int i = 0; i < FCNT_MAX; i++)
2542                         counter_u64_zero(V_pf_status.fcounters[i]);
2543                 for (int i = 0; i < SCNT_MAX; i++)
2544                         counter_u64_zero(V_pf_status.scounters[i]);
2545                 for (int i = 0; i < LCNT_MAX; i++)
2546                         counter_u64_zero(V_pf_status.lcounters[i]);
2547                 V_pf_status.since = time_second;
2548                 if (*V_pf_status.ifname)
2549                         pfi_update_status(V_pf_status.ifname, NULL);
2550                 PF_RULES_WUNLOCK();
2551                 break;
2552         }
2553
2554         case DIOCNATLOOK: {
2555                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
2556                 struct pf_state_key     *sk;
2557                 struct pf_state         *state;
2558                 struct pf_state_key_cmp  key;
2559                 int                      m = 0, direction = pnl->direction;
2560                 int                      sidx, didx;
2561
2562                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
2563                 sidx = (direction == PF_IN) ? 1 : 0;
2564                 didx = (direction == PF_IN) ? 0 : 1;
2565
2566                 if (!pnl->proto ||
2567                     PF_AZERO(&pnl->saddr, pnl->af) ||
2568                     PF_AZERO(&pnl->daddr, pnl->af) ||
2569                     ((pnl->proto == IPPROTO_TCP ||
2570                     pnl->proto == IPPROTO_UDP) &&
2571                     (!pnl->dport || !pnl->sport)))
2572                         error = EINVAL;
2573                 else {
2574                         bzero(&key, sizeof(key));
2575                         key.af = pnl->af;
2576                         key.proto = pnl->proto;
2577                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2578                         key.port[sidx] = pnl->sport;
2579                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2580                         key.port[didx] = pnl->dport;
2581
2582                         state = pf_find_state_all(&key, direction, &m);
2583
2584                         if (m > 1)
2585                                 error = E2BIG;  /* more than one state */
2586                         else if (state != NULL) {
2587                                 /* XXXGL: not locked read */
2588                                 sk = state->key[sidx];
2589                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2590                                 pnl->rsport = sk->port[sidx];
2591                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2592                                 pnl->rdport = sk->port[didx];
2593                         } else
2594                                 error = ENOENT;
2595                 }
2596                 break;
2597         }
2598
2599         case DIOCSETTIMEOUT: {
2600                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
2601                 int              old;
2602
2603                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2604                     pt->seconds < 0) {
2605                         error = EINVAL;
2606                         break;
2607                 }
2608                 PF_RULES_WLOCK();
2609                 old = V_pf_default_rule.timeout[pt->timeout];
2610                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2611                         pt->seconds = 1;
2612                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2613                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2614                         wakeup(pf_purge_thread);
2615                 pt->seconds = old;
2616                 PF_RULES_WUNLOCK();
2617                 break;
2618         }
2619
2620         case DIOCGETTIMEOUT: {
2621                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
2622
2623                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2624                         error = EINVAL;
2625                         break;
2626                 }
2627                 PF_RULES_RLOCK();
2628                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2629                 PF_RULES_RUNLOCK();
2630                 break;
2631         }
2632
2633         case DIOCGETLIMIT: {
2634                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
2635
2636                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2637                         error = EINVAL;
2638                         break;
2639                 }
2640                 PF_RULES_RLOCK();
2641                 pl->limit = V_pf_limits[pl->index].limit;
2642                 PF_RULES_RUNLOCK();
2643                 break;
2644         }
2645
2646         case DIOCSETLIMIT: {
2647                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
2648                 int                      old_limit;
2649
2650                 PF_RULES_WLOCK();
2651                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2652                     V_pf_limits[pl->index].zone == NULL) {
2653                         PF_RULES_WUNLOCK();
2654                         error = EINVAL;
2655                         break;
2656                 }
2657                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2658                 old_limit = V_pf_limits[pl->index].limit;
2659                 V_pf_limits[pl->index].limit = pl->limit;
2660                 pl->limit = old_limit;
2661                 PF_RULES_WUNLOCK();
2662                 break;
2663         }
2664
2665         case DIOCSETDEBUG: {
2666                 u_int32_t       *level = (u_int32_t *)addr;
2667
2668                 PF_RULES_WLOCK();
2669                 V_pf_status.debug = *level;
2670                 PF_RULES_WUNLOCK();
2671                 break;
2672         }
2673
2674         case DIOCCLRRULECTRS: {
2675                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2676                 struct pf_kruleset      *ruleset = &pf_main_ruleset;
2677                 struct pf_krule         *rule;
2678
2679                 PF_RULES_WLOCK();
2680                 TAILQ_FOREACH(rule,
2681                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2682                         counter_u64_zero(rule->evaluations);
2683                         for (int i = 0; i < 2; i++) {
2684                                 counter_u64_zero(rule->packets[i]);
2685                                 counter_u64_zero(rule->bytes[i]);
2686                         }
2687                 }
2688                 PF_RULES_WUNLOCK();
2689                 break;
2690         }
2691
2692         case DIOCGIFSPEEDV0:
2693         case DIOCGIFSPEEDV1: {
2694                 struct pf_ifspeed_v1    *psp = (struct pf_ifspeed_v1 *)addr;
2695                 struct pf_ifspeed_v1    ps;
2696                 struct ifnet            *ifp;
2697
2698                 if (psp->ifname[0] != 0) {
2699                         /* Can we completely trust user-land? */
2700                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2701                         ifp = ifunit(ps.ifname);
2702                         if (ifp != NULL) {
2703                                 psp->baudrate32 =
2704                                     (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2705                                 if (cmd == DIOCGIFSPEEDV1)
2706                                         psp->baudrate = ifp->if_baudrate;
2707                         } else
2708                                 error = EINVAL;
2709                 } else
2710                         error = EINVAL;
2711                 break;
2712         }
2713
2714 #ifdef ALTQ
2715         case DIOCSTARTALTQ: {
2716                 struct pf_altq          *altq;
2717
2718                 PF_RULES_WLOCK();
2719                 /* enable all altq interfaces on active list */
2720                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2721                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2722                                 error = pf_enable_altq(altq);
2723                                 if (error != 0)
2724                                         break;
2725                         }
2726                 }
2727                 if (error == 0)
2728                         V_pf_altq_running = 1;
2729                 PF_RULES_WUNLOCK();
2730                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2731                 break;
2732         }
2733
2734         case DIOCSTOPALTQ: {
2735                 struct pf_altq          *altq;
2736
2737                 PF_RULES_WLOCK();
2738                 /* disable all altq interfaces on active list */
2739                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2740                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2741                                 error = pf_disable_altq(altq);
2742                                 if (error != 0)
2743                                         break;
2744                         }
2745                 }
2746                 if (error == 0)
2747                         V_pf_altq_running = 0;
2748                 PF_RULES_WUNLOCK();
2749                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2750                 break;
2751         }
2752
2753         case DIOCADDALTQV0:
2754         case DIOCADDALTQV1: {
2755                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
2756                 struct pf_altq          *altq, *a;
2757                 struct ifnet            *ifp;
2758
2759                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2760                 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2761                 if (error)
2762                         break;
2763                 altq->local_flags = 0;
2764
2765                 PF_RULES_WLOCK();
2766                 if (pa->ticket != V_ticket_altqs_inactive) {
2767                         PF_RULES_WUNLOCK();
2768                         free(altq, M_PFALTQ);
2769                         error = EBUSY;
2770                         break;
2771                 }
2772
2773                 /*
2774                  * if this is for a queue, find the discipline and
2775                  * copy the necessary fields
2776                  */
2777                 if (altq->qname[0] != 0) {
2778                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2779                                 PF_RULES_WUNLOCK();
2780                                 error = EBUSY;
2781                                 free(altq, M_PFALTQ);
2782                                 break;
2783                         }
2784                         altq->altq_disc = NULL;
2785                         TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2786                                 if (strncmp(a->ifname, altq->ifname,
2787                                     IFNAMSIZ) == 0) {
2788                                         altq->altq_disc = a->altq_disc;
2789                                         break;
2790                                 }
2791                         }
2792                 }
2793
2794                 if ((ifp = ifunit(altq->ifname)) == NULL)
2795                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2796                 else
2797                         error = altq_add(ifp, altq);
2798
2799                 if (error) {
2800                         PF_RULES_WUNLOCK();
2801                         free(altq, M_PFALTQ);
2802                         break;
2803                 }
2804
2805                 if (altq->qname[0] != 0)
2806                         TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2807                 else
2808                         TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2809                 /* version error check done on import above */
2810                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2811                 PF_RULES_WUNLOCK();
2812                 break;
2813         }
2814
2815         case DIOCGETALTQSV0:
2816         case DIOCGETALTQSV1: {
2817                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
2818                 struct pf_altq          *altq;
2819
2820                 PF_RULES_RLOCK();
2821                 pa->nr = 0;
2822                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2823                         pa->nr++;
2824                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2825                         pa->nr++;
2826                 pa->ticket = V_ticket_altqs_active;
2827                 PF_RULES_RUNLOCK();
2828                 break;
2829         }
2830
2831         case DIOCGETALTQV0:
2832         case DIOCGETALTQV1: {
2833                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
2834                 struct pf_altq          *altq;
2835
2836                 PF_RULES_RLOCK();
2837                 if (pa->ticket != V_ticket_altqs_active) {
2838                         PF_RULES_RUNLOCK();
2839                         error = EBUSY;
2840                         break;
2841                 }
2842                 altq = pf_altq_get_nth_active(pa->nr);
2843                 if (altq == NULL) {
2844                         PF_RULES_RUNLOCK();
2845                         error = EBUSY;
2846                         break;
2847                 }
2848                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2849                 PF_RULES_RUNLOCK();
2850                 break;
2851         }
2852
2853         case DIOCCHANGEALTQV0:
2854         case DIOCCHANGEALTQV1:
2855                 /* CHANGEALTQ not supported yet! */
2856                 error = ENODEV;
2857                 break;
2858
2859         case DIOCGETQSTATSV0:
2860         case DIOCGETQSTATSV1: {
2861                 struct pfioc_qstats_v1  *pq = (struct pfioc_qstats_v1 *)addr;
2862                 struct pf_altq          *altq;
2863                 int                      nbytes;
2864                 u_int32_t                version;
2865
2866                 PF_RULES_RLOCK();
2867                 if (pq->ticket != V_ticket_altqs_active) {
2868                         PF_RULES_RUNLOCK();
2869                         error = EBUSY;
2870                         break;
2871                 }
2872                 nbytes = pq->nbytes;
2873                 altq = pf_altq_get_nth_active(pq->nr);
2874                 if (altq == NULL) {
2875                         PF_RULES_RUNLOCK();
2876                         error = EBUSY;
2877                         break;
2878                 }
2879
2880                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2881                         PF_RULES_RUNLOCK();
2882                         error = ENXIO;
2883                         break;
2884                 }
2885                 PF_RULES_RUNLOCK();
2886                 if (cmd == DIOCGETQSTATSV0)
2887                         version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2888                 else
2889                         version = pq->version;
2890                 error = altq_getqstats(altq, pq->buf, &nbytes, version);
2891                 if (error == 0) {
2892                         pq->scheduler = altq->scheduler;
2893                         pq->nbytes = nbytes;
2894                 }
2895                 break;
2896         }
2897 #endif /* ALTQ */
2898
2899         case DIOCBEGINADDRS: {
2900                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2901
2902                 PF_RULES_WLOCK();
2903                 pf_empty_kpool(&V_pf_pabuf);
2904                 pp->ticket = ++V_ticket_pabuf;
2905                 PF_RULES_WUNLOCK();
2906                 break;
2907         }
2908
2909         case DIOCADDADDR: {
2910                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2911                 struct pf_kpooladdr     *pa;
2912                 struct pfi_kkif         *kif = NULL;
2913
2914 #ifndef INET
2915                 if (pp->af == AF_INET) {
2916                         error = EAFNOSUPPORT;
2917                         break;
2918                 }
2919 #endif /* INET */
2920 #ifndef INET6
2921                 if (pp->af == AF_INET6) {
2922                         error = EAFNOSUPPORT;
2923                         break;
2924                 }
2925 #endif /* INET6 */
2926                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2927                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2928                     pp->addr.addr.type != PF_ADDR_TABLE) {
2929                         error = EINVAL;
2930                         break;
2931                 }
2932                 if (pp->addr.addr.p.dyn != NULL) {
2933                         error = EINVAL;
2934                         break;
2935                 }
2936                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2937                 pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2938                 if (pa->ifname[0])
2939                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2940                 PF_RULES_WLOCK();
2941                 if (pp->ticket != V_ticket_pabuf) {
2942                         PF_RULES_WUNLOCK();
2943                         if (pa->ifname[0])
2944                                 free(kif, PFI_MTYPE);
2945                         free(pa, M_PFRULE);
2946                         error = EBUSY;
2947                         break;
2948                 }
2949                 if (pa->ifname[0]) {
2950                         pa->kif = pfi_kkif_attach(kif, pa->ifname);
2951                         pfi_kkif_ref(pa->kif);
2952                 } else
2953                         pa->kif = NULL;
2954                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2955                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2956                         if (pa->ifname[0])
2957                                 pfi_kkif_unref(pa->kif);
2958                         PF_RULES_WUNLOCK();
2959                         free(pa, M_PFRULE);
2960                         break;
2961                 }
2962                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2963                 PF_RULES_WUNLOCK();
2964                 break;
2965         }
2966
2967         case DIOCGETADDRS: {
2968                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2969                 struct pf_kpool         *pool;
2970                 struct pf_kpooladdr     *pa;
2971
2972                 PF_RULES_RLOCK();
2973                 pp->nr = 0;
2974                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2975                     pp->r_num, 0, 1, 0);
2976                 if (pool == NULL) {
2977                         PF_RULES_RUNLOCK();
2978                         error = EBUSY;
2979                         break;
2980                 }
2981                 TAILQ_FOREACH(pa, &pool->list, entries)
2982                         pp->nr++;
2983                 PF_RULES_RUNLOCK();
2984                 break;
2985         }
2986
2987         case DIOCGETADDR: {
2988                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2989                 struct pf_kpool         *pool;
2990                 struct pf_kpooladdr     *pa;
2991                 u_int32_t                nr = 0;
2992
2993                 PF_RULES_RLOCK();
2994                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2995                     pp->r_num, 0, 1, 1);
2996                 if (pool == NULL) {
2997                         PF_RULES_RUNLOCK();
2998                         error = EBUSY;
2999                         break;
3000                 }
3001                 pa = TAILQ_FIRST(&pool->list);
3002                 while ((pa != NULL) && (nr < pp->nr)) {
3003                         pa = TAILQ_NEXT(pa, entries);
3004                         nr++;
3005                 }
3006                 if (pa == NULL) {
3007                         PF_RULES_RUNLOCK();
3008                         error = EBUSY;
3009                         break;
3010                 }
3011                 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3012                 pf_addr_copyout(&pp->addr.addr);
3013                 PF_RULES_RUNLOCK();
3014                 break;
3015         }
3016
3017         case DIOCCHANGEADDR: {
3018                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
3019                 struct pf_kpool         *pool;
3020                 struct pf_kpooladdr     *oldpa = NULL, *newpa = NULL;
3021                 struct pf_kruleset      *ruleset;
3022                 struct pfi_kkif         *kif = NULL;
3023
3024                 if (pca->action < PF_CHANGE_ADD_HEAD ||
3025                     pca->action > PF_CHANGE_REMOVE) {
3026                         error = EINVAL;
3027                         break;
3028                 }
3029                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3030                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3031                     pca->addr.addr.type != PF_ADDR_TABLE) {
3032                         error = EINVAL;
3033                         break;
3034                 }
3035                 if (pca->addr.addr.p.dyn != NULL) {
3036                         error = EINVAL;
3037                         break;
3038                 }
3039
3040                 if (pca->action != PF_CHANGE_REMOVE) {
3041 #ifndef INET
3042                         if (pca->af == AF_INET) {
3043                                 error = EAFNOSUPPORT;
3044                                 break;
3045                         }
3046 #endif /* INET */
3047 #ifndef INET6
3048                         if (pca->af == AF_INET6) {
3049                                 error = EAFNOSUPPORT;
3050                                 break;
3051                         }
3052 #endif /* INET6 */
3053                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3054                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3055                         if (newpa->ifname[0])
3056                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
3057                         newpa->kif = NULL;
3058                 }
3059
3060 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
3061                 PF_RULES_WLOCK();
3062                 ruleset = pf_find_kruleset(pca->anchor);
3063                 if (ruleset == NULL)
3064                         ERROUT(EBUSY);
3065
3066                 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3067                     pca->r_num, pca->r_last, 1, 1);
3068                 if (pool == NULL)
3069                         ERROUT(EBUSY);
3070
3071                 if (pca->action != PF_CHANGE_REMOVE) {
3072                         if (newpa->ifname[0]) {
3073                                 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3074                                 pfi_kkif_ref(newpa->kif);
3075                                 kif = NULL;
3076                         }
3077
3078                         switch (newpa->addr.type) {
3079                         case PF_ADDR_DYNIFTL:
3080                                 error = pfi_dynaddr_setup(&newpa->addr,
3081                                     pca->af);
3082                                 break;
3083                         case PF_ADDR_TABLE:
3084                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
3085                                     newpa->addr.v.tblname);
3086                                 if (newpa->addr.p.tbl == NULL)
3087                                         error = ENOMEM;
3088                                 break;
3089                         }
3090                         if (error)
3091                                 goto DIOCCHANGEADDR_error;
3092                 }
3093
3094                 switch (pca->action) {
3095                 case PF_CHANGE_ADD_HEAD:
3096                         oldpa = TAILQ_FIRST(&pool->list);
3097                         break;
3098                 case PF_CHANGE_ADD_TAIL:
3099                         oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3100                         break;
3101                 default:
3102                         oldpa = TAILQ_FIRST(&pool->list);
3103                         for (int i = 0; oldpa && i < pca->nr; i++)
3104                                 oldpa = TAILQ_NEXT(oldpa, entries);
3105
3106                         if (oldpa == NULL)
3107                                 ERROUT(EINVAL);
3108                 }
3109
3110                 if (pca->action == PF_CHANGE_REMOVE) {
3111                         TAILQ_REMOVE(&pool->list, oldpa, entries);
3112                         switch (oldpa->addr.type) {
3113                         case PF_ADDR_DYNIFTL:
3114                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
3115                                 break;
3116                         case PF_ADDR_TABLE:
3117                                 pfr_detach_table(oldpa->addr.p.tbl);
3118                                 break;
3119                         }
3120                         if (oldpa->kif)
3121                                 pfi_kkif_unref(oldpa->kif);
3122                         free(oldpa, M_PFRULE);
3123                 } else {
3124                         if (oldpa == NULL)
3125                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3126                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
3127                             pca->action == PF_CHANGE_ADD_BEFORE)
3128                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3129                         else
3130                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
3131                                     newpa, entries);
3132                 }
3133
3134                 pool->cur = TAILQ_FIRST(&pool->list);
3135                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3136                 PF_RULES_WUNLOCK();
3137                 break;
3138
3139 #undef ERROUT
3140 DIOCCHANGEADDR_error:
3141                 if (newpa != NULL) {
3142                         if (newpa->kif)
3143                                 pfi_kkif_unref(newpa->kif);
3144                         free(newpa, M_PFRULE);
3145                 }
3146                 PF_RULES_WUNLOCK();
3147                 if (kif != NULL)
3148                         free(kif, PFI_MTYPE);
3149                 break;
3150         }
3151
3152         case DIOCGETRULESETS: {
3153                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
3154                 struct pf_kruleset      *ruleset;
3155                 struct pf_kanchor       *anchor;
3156
3157                 PF_RULES_RLOCK();
3158                 pr->path[sizeof(pr->path) - 1] = 0;
3159                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3160                         PF_RULES_RUNLOCK();
3161                         error = ENOENT;
3162                         break;
3163                 }
3164                 pr->nr = 0;
3165                 if (ruleset->anchor == NULL) {
3166                         /* XXX kludge for pf_main_ruleset */
3167                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3168                                 if (anchor->parent == NULL)
3169                                         pr->nr++;
3170                 } else {
3171                         RB_FOREACH(anchor, pf_kanchor_node,
3172                             &ruleset->anchor->children)
3173                                 pr->nr++;
3174                 }
3175                 PF_RULES_RUNLOCK();
3176                 break;
3177         }
3178
3179         case DIOCGETRULESET: {
3180                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
3181                 struct pf_kruleset      *ruleset;
3182                 struct pf_kanchor       *anchor;
3183                 u_int32_t                nr = 0;
3184
3185                 PF_RULES_RLOCK();
3186                 pr->path[sizeof(pr->path) - 1] = 0;
3187                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3188                         PF_RULES_RUNLOCK();
3189                         error = ENOENT;
3190                         break;
3191                 }
3192                 pr->name[0] = 0;
3193                 if (ruleset->anchor == NULL) {
3194                         /* XXX kludge for pf_main_ruleset */
3195                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3196                                 if (anchor->parent == NULL && nr++ == pr->nr) {
3197                                         strlcpy(pr->name, anchor->name,
3198                                             sizeof(pr->name));
3199                                         break;
3200                                 }
3201                 } else {
3202                         RB_FOREACH(anchor, pf_kanchor_node,
3203                             &ruleset->anchor->children)
3204                                 if (nr++ == pr->nr) {
3205                                         strlcpy(pr->name, anchor->name,
3206                                             sizeof(pr->name));
3207                                         break;
3208                                 }
3209                 }
3210                 if (!pr->name[0])
3211                         error = EBUSY;
3212                 PF_RULES_RUNLOCK();
3213                 break;
3214         }
3215
3216         case DIOCRCLRTABLES: {
3217                 struct pfioc_table *io = (struct pfioc_table *)addr;
3218
3219                 if (io->pfrio_esize != 0) {
3220                         error = ENODEV;
3221                         break;
3222                 }
3223                 PF_RULES_WLOCK();
3224                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3225                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
3226                 PF_RULES_WUNLOCK();
3227                 break;
3228         }
3229
3230         case DIOCRADDTABLES: {
3231                 struct pfioc_table *io = (struct pfioc_table *)addr;
3232                 struct pfr_table *pfrts;
3233                 size_t totlen;
3234
3235                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3236                         error = ENODEV;
3237                         break;
3238                 }
3239
3240                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3241                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3242                         error = ENOMEM;
3243                         break;
3244                 }
3245
3246                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3247                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3248                     M_TEMP, M_WAITOK);
3249                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3250                 if (error) {
3251                         free(pfrts, M_TEMP);
3252                         break;
3253                 }
3254                 PF_RULES_WLOCK();
3255                 error = pfr_add_tables(pfrts, io->pfrio_size,
3256                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3257                 PF_RULES_WUNLOCK();
3258                 free(pfrts, M_TEMP);
3259                 break;
3260         }
3261
3262         case DIOCRDELTABLES: {
3263                 struct pfioc_table *io = (struct pfioc_table *)addr;
3264                 struct pfr_table *pfrts;
3265                 size_t totlen;
3266
3267                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3268                         error = ENODEV;
3269                         break;
3270                 }
3271
3272                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3273                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3274                         error = ENOMEM;
3275                         break;
3276                 }
3277
3278                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3279                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3280                     M_TEMP, M_WAITOK);
3281                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3282                 if (error) {
3283                         free(pfrts, M_TEMP);
3284                         break;
3285                 }
3286                 PF_RULES_WLOCK();
3287                 error = pfr_del_tables(pfrts, io->pfrio_size,
3288                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3289                 PF_RULES_WUNLOCK();
3290                 free(pfrts, M_TEMP);
3291                 break;
3292         }
3293
3294         case DIOCRGETTABLES: {
3295                 struct pfioc_table *io = (struct pfioc_table *)addr;
3296                 struct pfr_table *pfrts;
3297                 size_t totlen;
3298                 int n;
3299
3300                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3301                         error = ENODEV;
3302                         break;
3303                 }
3304                 PF_RULES_RLOCK();
3305                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3306                 if (n < 0) {
3307                         PF_RULES_RUNLOCK();
3308                         error = EINVAL;
3309                         break;
3310                 }
3311                 io->pfrio_size = min(io->pfrio_size, n);
3312
3313                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3314
3315                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3316                     M_TEMP, M_NOWAIT);
3317                 if (pfrts == NULL) {
3318                         error = ENOMEM;
3319                         PF_RULES_RUNLOCK();
3320                         break;
3321                 }
3322                 error = pfr_get_tables(&io->pfrio_table, pfrts,
3323                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3324                 PF_RULES_RUNLOCK();
3325                 if (error == 0)
3326                         error = copyout(pfrts, io->pfrio_buffer, totlen);
3327                 free(pfrts, M_TEMP);
3328                 break;
3329         }
3330
3331         case DIOCRGETTSTATS: {
3332                 struct pfioc_table *io = (struct pfioc_table *)addr;
3333                 struct pfr_tstats *pfrtstats;
3334                 size_t totlen;
3335                 int n;
3336
3337                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3338                         error = ENODEV;
3339                         break;
3340                 }
3341                 PF_RULES_WLOCK();
3342                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3343                 if (n < 0) {
3344                         PF_RULES_WUNLOCK();
3345                         error = EINVAL;
3346                         break;
3347                 }
3348                 io->pfrio_size = min(io->pfrio_size, n);
3349
3350                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3351                 pfrtstats = mallocarray(io->pfrio_size,
3352                     sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3353                 if (pfrtstats == NULL) {
3354                         error = ENOMEM;
3355                         PF_RULES_WUNLOCK();
3356                         break;
3357                 }
3358                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3359                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3360                 PF_RULES_WUNLOCK();
3361                 if (error == 0)
3362                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3363                 free(pfrtstats, M_TEMP);
3364                 break;
3365         }
3366
3367         case DIOCRCLRTSTATS: {
3368                 struct pfioc_table *io = (struct pfioc_table *)addr;
3369                 struct pfr_table *pfrts;
3370                 size_t totlen;
3371                 int n;
3372
3373                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3374                         error = ENODEV;
3375                         break;
3376                 }
3377
3378                 PF_RULES_WLOCK();
3379                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3380                 if (n < 0) {
3381                         PF_RULES_WUNLOCK();
3382                         error = EINVAL;
3383                         break;
3384                 }
3385                 io->pfrio_size = min(io->pfrio_size, n);
3386
3387                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3388                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3389                     M_TEMP, M_NOWAIT);
3390                 if (pfrts == NULL) {
3391                         error = ENOMEM;
3392                         PF_RULES_WUNLOCK();
3393                         break;
3394                 }
3395                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3396                 if (error) {
3397                         free(pfrts, M_TEMP);
3398                         PF_RULES_WUNLOCK();
3399                         break;
3400                 }
3401                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
3402                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3403                 PF_RULES_WUNLOCK();
3404                 free(pfrts, M_TEMP);
3405                 break;
3406         }
3407
3408         case DIOCRSETTFLAGS: {
3409                 struct pfioc_table *io = (struct pfioc_table *)addr;
3410                 struct pfr_table *pfrts;
3411                 size_t totlen;
3412                 int n;
3413
3414                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3415                         error = ENODEV;
3416                         break;
3417                 }
3418
3419                 PF_RULES_RLOCK();
3420                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3421                 if (n < 0) {
3422                         PF_RULES_RUNLOCK();
3423                         error = EINVAL;
3424                         break;
3425                 }
3426
3427                 io->pfrio_size = min(io->pfrio_size, n);
3428                 PF_RULES_RUNLOCK();
3429
3430                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3431                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3432                     M_TEMP, M_WAITOK);
3433                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3434                 if (error) {
3435                         free(pfrts, M_TEMP);
3436                         break;
3437                 }
3438                 PF_RULES_WLOCK();
3439                 error = pfr_set_tflags(pfrts, io->pfrio_size,
3440                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3441                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3442                 PF_RULES_WUNLOCK();
3443                 free(pfrts, M_TEMP);
3444                 break;
3445         }
3446
3447         case DIOCRCLRADDRS: {
3448                 struct pfioc_table *io = (struct pfioc_table *)addr;
3449
3450                 if (io->pfrio_esize != 0) {
3451                         error = ENODEV;
3452                         break;
3453                 }
3454                 PF_RULES_WLOCK();
3455                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3456                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
3457                 PF_RULES_WUNLOCK();
3458                 break;
3459         }
3460
3461         case DIOCRADDADDRS: {
3462                 struct pfioc_table *io = (struct pfioc_table *)addr;
3463                 struct pfr_addr *pfras;
3464                 size_t totlen;
3465
3466                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3467                         error = ENODEV;
3468                         break;
3469                 }
3470                 if (io->pfrio_size < 0 ||
3471                     io->pfrio_size > pf_ioctl_maxcount ||
3472                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3473                         error = EINVAL;
3474                         break;
3475                 }
3476                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3477                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3478                     M_TEMP, M_NOWAIT);
3479                 if (! pfras) {
3480                         error = ENOMEM;
3481                         break;
3482                 }
3483                 error = copyin(io->pfrio_buffer, pfras, totlen);
3484                 if (error) {
3485                         free(pfras, M_TEMP);
3486                         break;
3487                 }
3488                 PF_RULES_WLOCK();
3489                 error = pfr_add_addrs(&io->pfrio_table, pfras,
3490                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3491                     PFR_FLAG_USERIOCTL);
3492                 PF_RULES_WUNLOCK();
3493                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3494                         error = copyout(pfras, io->pfrio_buffer, totlen);
3495                 free(pfras, M_TEMP);
3496                 break;
3497         }
3498
3499         case DIOCRDELADDRS: {
3500                 struct pfioc_table *io = (struct pfioc_table *)addr;
3501                 struct pfr_addr *pfras;
3502                 size_t totlen;
3503
3504                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3505                         error = ENODEV;
3506                         break;
3507                 }
3508                 if (io->pfrio_size < 0 ||
3509                     io->pfrio_size > pf_ioctl_maxcount ||
3510                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3511                         error = EINVAL;
3512                         break;
3513                 }
3514                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3515                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3516                     M_TEMP, M_NOWAIT);
3517                 if (! pfras) {
3518                         error = ENOMEM;
3519                         break;
3520                 }
3521                 error = copyin(io->pfrio_buffer, pfras, totlen);
3522                 if (error) {
3523                         free(pfras, M_TEMP);
3524                         break;
3525                 }
3526                 PF_RULES_WLOCK();
3527                 error = pfr_del_addrs(&io->pfrio_table, pfras,
3528                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3529                     PFR_FLAG_USERIOCTL);
3530                 PF_RULES_WUNLOCK();
3531                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3532                         error = copyout(pfras, io->pfrio_buffer, totlen);
3533                 free(pfras, M_TEMP);
3534                 break;
3535         }
3536
3537         case DIOCRSETADDRS: {
3538                 struct pfioc_table *io = (struct pfioc_table *)addr;
3539                 struct pfr_addr *pfras;
3540                 size_t totlen, count;
3541
3542                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3543                         error = ENODEV;
3544                         break;
3545                 }
3546                 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3547                         error = EINVAL;
3548                         break;
3549                 }
3550                 count = max(io->pfrio_size, io->pfrio_size2);
3551                 if (count > pf_ioctl_maxcount ||
3552                     WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3553                         error = EINVAL;
3554                         break;
3555                 }
3556                 totlen = count * sizeof(struct pfr_addr);
3557                 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3558                     M_NOWAIT);
3559                 if (! pfras) {
3560                         error = ENOMEM;
3561                         break;
3562                 }
3563                 error = copyin(io->pfrio_buffer, pfras, totlen);
3564                 if (error) {
3565                         free(pfras, M_TEMP);
3566                         break;
3567                 }
3568                 PF_RULES_WLOCK();
3569                 error = pfr_set_addrs(&io->pfrio_table, pfras,
3570                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3571                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3572                     PFR_FLAG_USERIOCTL, 0);
3573                 PF_RULES_WUNLOCK();
3574                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3575                         error = copyout(pfras, io->pfrio_buffer, totlen);
3576                 free(pfras, M_TEMP);
3577                 break;
3578         }
3579
3580         case DIOCRGETADDRS: {
3581                 struct pfioc_table *io = (struct pfioc_table *)addr;
3582                 struct pfr_addr *pfras;
3583                 size_t totlen;
3584
3585                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3586                         error = ENODEV;
3587                         break;
3588                 }
3589                 if (io->pfrio_size < 0 ||
3590                     io->pfrio_size > pf_ioctl_maxcount ||
3591                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3592                         error = EINVAL;
3593                         break;
3594                 }
3595                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3596                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3597                     M_TEMP, M_NOWAIT);
3598                 if (! pfras) {
3599                         error = ENOMEM;
3600                         break;
3601                 }
3602                 PF_RULES_RLOCK();
3603                 error = pfr_get_addrs(&io->pfrio_table, pfras,
3604                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3605                 PF_RULES_RUNLOCK();
3606                 if (error == 0)
3607                         error = copyout(pfras, io->pfrio_buffer, totlen);
3608                 free(pfras, M_TEMP);
3609                 break;
3610         }
3611
3612         case DIOCRGETASTATS: {
3613                 struct pfioc_table *io = (struct pfioc_table *)addr;
3614                 struct pfr_astats *pfrastats;
3615                 size_t totlen;
3616
3617                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3618                         error = ENODEV;
3619                         break;
3620                 }
3621                 if (io->pfrio_size < 0 ||
3622                     io->pfrio_size > pf_ioctl_maxcount ||
3623                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3624                         error = EINVAL;
3625                         break;
3626                 }
3627                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
3628                 pfrastats = mallocarray(io->pfrio_size,
3629                     sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3630                 if (! pfrastats) {
3631                         error = ENOMEM;
3632                         break;
3633                 }
3634                 PF_RULES_RLOCK();
3635                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
3636                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3637                 PF_RULES_RUNLOCK();
3638                 if (error == 0)
3639                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
3640                 free(pfrastats, M_TEMP);
3641                 break;
3642         }
3643
3644         case DIOCRCLRASTATS: {
3645                 struct pfioc_table *io = (struct pfioc_table *)addr;
3646                 struct pfr_addr *pfras;
3647                 size_t totlen;
3648
3649                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3650                         error = ENODEV;
3651                         break;
3652                 }
3653                 if (io->pfrio_size < 0 ||
3654                     io->pfrio_size > pf_ioctl_maxcount ||
3655                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3656                         error = EINVAL;
3657                         break;
3658                 }
3659                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3660                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3661                     M_TEMP, M_NOWAIT);
3662                 if (! pfras) {
3663                         error = ENOMEM;
3664                         break;
3665                 }
3666                 error = copyin(io->pfrio_buffer, pfras, totlen);
3667                 if (error) {
3668                         free(pfras, M_TEMP);
3669                         break;
3670                 }
3671                 PF_RULES_WLOCK();
3672                 error = pfr_clr_astats(&io->pfrio_table, pfras,
3673                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3674                     PFR_FLAG_USERIOCTL);
3675                 PF_RULES_WUNLOCK();
3676                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3677                         error = copyout(pfras, io->pfrio_buffer, totlen);
3678                 free(pfras, M_TEMP);
3679                 break;
3680         }
3681
3682         case DIOCRTSTADDRS: {
3683                 struct pfioc_table *io = (struct pfioc_table *)addr;
3684                 struct pfr_addr *pfras;
3685                 size_t totlen;
3686
3687                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3688                         error = ENODEV;
3689                         break;
3690                 }
3691                 if (io->pfrio_size < 0 ||
3692                     io->pfrio_size > pf_ioctl_maxcount ||
3693                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3694                         error = EINVAL;
3695                         break;
3696                 }
3697                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3698                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3699                     M_TEMP, M_NOWAIT);
3700                 if (! pfras) {
3701                         error = ENOMEM;
3702                         break;
3703                 }
3704                 error = copyin(io->pfrio_buffer, pfras, totlen);
3705                 if (error) {
3706                         free(pfras, M_TEMP);
3707                         break;
3708                 }
3709                 PF_RULES_RLOCK();
3710                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
3711                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3712                     PFR_FLAG_USERIOCTL);
3713                 PF_RULES_RUNLOCK();
3714                 if (error == 0)
3715                         error = copyout(pfras, io->pfrio_buffer, totlen);
3716                 free(pfras, M_TEMP);
3717                 break;
3718         }
3719
3720         case DIOCRINADEFINE: {
3721                 struct pfioc_table *io = (struct pfioc_table *)addr;
3722                 struct pfr_addr *pfras;
3723                 size_t totlen;
3724
3725                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3726                         error = ENODEV;
3727                         break;
3728                 }
3729                 if (io->pfrio_size < 0 ||
3730                     io->pfrio_size > pf_ioctl_maxcount ||
3731                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3732                         error = EINVAL;
3733                         break;
3734                 }
3735                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3736                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3737                     M_TEMP, M_NOWAIT);
3738                 if (! pfras) {
3739                         error = ENOMEM;
3740                         break;
3741                 }
3742                 error = copyin(io->pfrio_buffer, pfras, totlen);
3743                 if (error) {
3744                         free(pfras, M_TEMP);
3745                         break;
3746                 }
3747                 PF_RULES_WLOCK();
3748                 error = pfr_ina_define(&io->pfrio_table, pfras,
3749                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3750                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3751                 PF_RULES_WUNLOCK();
3752                 free(pfras, M_TEMP);
3753                 break;
3754         }
3755
3756         case DIOCOSFPADD: {
3757                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3758                 PF_RULES_WLOCK();
3759                 error = pf_osfp_add(io);
3760                 PF_RULES_WUNLOCK();
3761                 break;
3762         }
3763
3764         case DIOCOSFPGET: {
3765                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3766                 PF_RULES_RLOCK();
3767                 error = pf_osfp_get(io);
3768                 PF_RULES_RUNLOCK();
3769                 break;
3770         }
3771
3772         case DIOCXBEGIN: {
3773                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3774                 struct pfioc_trans_e    *ioes, *ioe;
3775                 size_t                   totlen;
3776                 int                      i;
3777
3778                 if (io->esize != sizeof(*ioe)) {
3779                         error = ENODEV;
3780                         break;
3781                 }
3782                 if (io->size < 0 ||
3783                     io->size > pf_ioctl_maxcount ||
3784                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3785                         error = EINVAL;
3786                         break;
3787                 }
3788                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3789                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3790                     M_TEMP, M_NOWAIT);
3791                 if (! ioes) {
3792                         error = ENOMEM;
3793                         break;
3794                 }
3795                 error = copyin(io->array, ioes, totlen);
3796                 if (error) {
3797                         free(ioes, M_TEMP);
3798                         break;
3799                 }
3800                 PF_RULES_WLOCK();
3801                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3802                         switch (ioe->rs_num) {
3803 #ifdef ALTQ
3804                         case PF_RULESET_ALTQ:
3805                                 if (ioe->anchor[0]) {
3806                                         PF_RULES_WUNLOCK();
3807                                         free(ioes, M_TEMP);
3808                                         error = EINVAL;
3809                                         goto fail;
3810                                 }
3811                                 if ((error = pf_begin_altq(&ioe->ticket))) {
3812                                         PF_RULES_WUNLOCK();
3813                                         free(ioes, M_TEMP);
3814                                         goto fail;
3815                                 }
3816                                 break;
3817 #endif /* ALTQ */
3818                         case PF_RULESET_TABLE:
3819                             {
3820                                 struct pfr_table table;
3821
3822                                 bzero(&table, sizeof(table));
3823                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3824                                     sizeof(table.pfrt_anchor));
3825                                 if ((error = pfr_ina_begin(&table,
3826                                     &ioe->ticket, NULL, 0))) {
3827                                         PF_RULES_WUNLOCK();
3828                                         free(ioes, M_TEMP);
3829                                         goto fail;
3830                                 }
3831                                 break;
3832                             }
3833                         default:
3834                                 if ((error = pf_begin_rules(&ioe->ticket,
3835                                     ioe->rs_num, ioe->anchor))) {
3836                                         PF_RULES_WUNLOCK();
3837                                         free(ioes, M_TEMP);
3838                                         goto fail;
3839                                 }
3840                                 break;
3841                         }
3842                 }
3843                 PF_RULES_WUNLOCK();
3844                 error = copyout(ioes, io->array, totlen);
3845                 free(ioes, M_TEMP);
3846                 break;
3847         }
3848
3849         case DIOCXROLLBACK: {
3850                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3851                 struct pfioc_trans_e    *ioe, *ioes;
3852                 size_t                   totlen;
3853                 int                      i;
3854
3855                 if (io->esize != sizeof(*ioe)) {
3856                         error = ENODEV;
3857                         break;
3858                 }
3859                 if (io->size < 0 ||
3860                     io->size > pf_ioctl_maxcount ||
3861                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3862                         error = EINVAL;
3863                         break;
3864                 }
3865                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3866                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3867                     M_TEMP, M_NOWAIT);
3868                 if (! ioes) {
3869                         error = ENOMEM;
3870                         break;
3871                 }
3872                 error = copyin(io->array, ioes, totlen);
3873                 if (error) {
3874                         free(ioes, M_TEMP);
3875                         break;
3876                 }
3877                 PF_RULES_WLOCK();
3878                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3879                         switch (ioe->rs_num) {
3880 #ifdef ALTQ
3881                         case PF_RULESET_ALTQ:
3882                                 if (ioe->anchor[0]) {
3883                                         PF_RULES_WUNLOCK();
3884                                         free(ioes, M_TEMP);
3885                                         error = EINVAL;
3886                                         goto fail;
3887                                 }
3888                                 if ((error = pf_rollback_altq(ioe->ticket))) {
3889                                         PF_RULES_WUNLOCK();
3890                                         free(ioes, M_TEMP);
3891                                         goto fail; /* really bad */
3892                                 }
3893                                 break;
3894 #endif /* ALTQ */
3895                         case PF_RULESET_TABLE:
3896                             {
3897                                 struct pfr_table table;
3898
3899                                 bzero(&table, sizeof(table));
3900                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3901                                     sizeof(table.pfrt_anchor));
3902                                 if ((error = pfr_ina_rollback(&table,
3903                                     ioe->ticket, NULL, 0))) {
3904                                         PF_RULES_WUNLOCK();
3905                                         free(ioes, M_TEMP);
3906                                         goto fail; /* really bad */
3907                                 }
3908                                 break;
3909                             }
3910                         default:
3911                                 if ((error = pf_rollback_rules(ioe->ticket,
3912                                     ioe->rs_num, ioe->anchor))) {
3913                                         PF_RULES_WUNLOCK();
3914                                         free(ioes, M_TEMP);
3915                                         goto fail; /* really bad */
3916                                 }
3917                                 break;
3918                         }
3919                 }
3920                 PF_RULES_WUNLOCK();
3921                 free(ioes, M_TEMP);
3922                 break;
3923         }
3924
3925         case DIOCXCOMMIT: {
3926                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3927                 struct pfioc_trans_e    *ioe, *ioes;
3928                 struct pf_kruleset      *rs;
3929                 size_t                   totlen;
3930                 int                      i;
3931
3932                 if (io->esize != sizeof(*ioe)) {
3933                         error = ENODEV;
3934                         break;
3935                 }
3936
3937                 if (io->size < 0 ||
3938                     io->size > pf_ioctl_maxcount ||
3939                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3940                         error = EINVAL;
3941                         break;
3942                 }
3943
3944                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3945                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3946                     M_TEMP, M_NOWAIT);
3947                 if (ioes == NULL) {
3948                         error = ENOMEM;
3949                         break;
3950                 }
3951                 error = copyin(io->array, ioes, totlen);
3952                 if (error) {
3953                         free(ioes, M_TEMP);
3954                         break;
3955                 }
3956                 PF_RULES_WLOCK();
3957                 /* First makes sure everything will succeed. */
3958                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3959                         switch (ioe->rs_num) {
3960 #ifdef ALTQ
3961                         case PF_RULESET_ALTQ:
3962                                 if (ioe->anchor[0]) {
3963                                         PF_RULES_WUNLOCK();
3964                                         free(ioes, M_TEMP);
3965                                         error = EINVAL;
3966                                         goto fail;
3967                                 }
3968                                 if (!V_altqs_inactive_open || ioe->ticket !=
3969                                     V_ticket_altqs_inactive) {
3970                                         PF_RULES_WUNLOCK();
3971                                         free(ioes, M_TEMP);
3972                                         error = EBUSY;
3973                                         goto fail;
3974                                 }
3975                                 break;
3976 #endif /* ALTQ */
3977                         case PF_RULESET_TABLE:
3978                                 rs = pf_find_kruleset(ioe->anchor);
3979                                 if (rs == NULL || !rs->topen || ioe->ticket !=
3980                                     rs->tticket) {
3981                                         PF_RULES_WUNLOCK();
3982                                         free(ioes, M_TEMP);
3983                                         error = EBUSY;
3984                                         goto fail;
3985                                 }
3986                                 break;
3987                         default:
3988                                 if (ioe->rs_num < 0 || ioe->rs_num >=
3989                                     PF_RULESET_MAX) {
3990                                         PF_RULES_WUNLOCK();
3991                                         free(ioes, M_TEMP);
3992                                         error = EINVAL;
3993                                         goto fail;
3994                                 }
3995                                 rs = pf_find_kruleset(ioe->anchor);
3996                                 if (rs == NULL ||
3997                                     !rs->rules[ioe->rs_num].inactive.open ||
3998                                     rs->rules[ioe->rs_num].inactive.ticket !=
3999                                     ioe->ticket) {
4000                                         PF_RULES_WUNLOCK();
4001                                         free(ioes, M_TEMP);
4002                                         error = EBUSY;
4003                                         goto fail;
4004                                 }
4005                                 break;
4006                         }
4007                 }
4008                 /* Now do the commit - no errors should happen here. */
4009                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4010                         switch (ioe->rs_num) {
4011 #ifdef ALTQ
4012                         case PF_RULESET_ALTQ:
4013                                 if ((error = pf_commit_altq(ioe->ticket))) {
4014                                         PF_RULES_WUNLOCK();
4015                                         free(ioes, M_TEMP);
4016                                         goto fail; /* really bad */
4017                                 }
4018                                 break;
4019 #endif /* ALTQ */
4020                         case PF_RULESET_TABLE:
4021                             {
4022                                 struct pfr_table table;
4023
4024                                 bzero(&table, sizeof(table));
4025                                 strlcpy(table.pfrt_anchor, ioe->anchor,
4026                                     sizeof(table.pfrt_anchor));
4027                                 if ((error = pfr_ina_commit(&table,
4028                                     ioe->ticket, NULL, NULL, 0))) {
4029                                         PF_RULES_WUNLOCK();
4030                                         free(ioes, M_TEMP);
4031                                         goto fail; /* really bad */
4032                                 }
4033                                 break;
4034                             }
4035                         default:
4036                                 if ((error = pf_commit_rules(ioe->ticket,
4037                                     ioe->rs_num, ioe->anchor))) {
4038                                         PF_RULES_WUNLOCK();
4039                                         free(ioes, M_TEMP);
4040                                         goto fail; /* really bad */
4041                                 }
4042                                 break;
4043                         }
4044                 }
4045                 PF_RULES_WUNLOCK();
4046                 free(ioes, M_TEMP);
4047                 break;
4048         }
4049
4050         case DIOCGETSRCNODES: {
4051                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
4052                 struct pf_srchash       *sh;
4053                 struct pf_ksrc_node     *n;
4054                 struct pf_src_node      *p, *pstore;
4055                 uint32_t                 i, nr = 0;
4056
4057                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4058                                 i++, sh++) {
4059                         PF_HASHROW_LOCK(sh);
4060                         LIST_FOREACH(n, &sh->nodes, entry)
4061                                 nr++;
4062                         PF_HASHROW_UNLOCK(sh);
4063                 }
4064
4065                 psn->psn_len = min(psn->psn_len,
4066                     sizeof(struct pf_src_node) * nr);
4067
4068                 if (psn->psn_len == 0) {
4069                         psn->psn_len = sizeof(struct pf_src_node) * nr;
4070                         break;
4071                 }
4072
4073                 nr = 0;
4074
4075                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4076                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4077                     i++, sh++) {
4078                     PF_HASHROW_LOCK(sh);
4079                     LIST_FOREACH(n, &sh->nodes, entry) {
4080
4081                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4082                                 break;
4083
4084                         pf_src_node_copy(n, p);
4085
4086                         p++;
4087                         nr++;
4088                     }
4089                     PF_HASHROW_UNLOCK(sh);
4090                 }
4091                 error = copyout(pstore, psn->psn_src_nodes,
4092                     sizeof(struct pf_src_node) * nr);
4093                 if (error) {
4094                         free(pstore, M_TEMP);
4095                         break;
4096                 }
4097                 psn->psn_len = sizeof(struct pf_src_node) * nr;
4098                 free(pstore, M_TEMP);
4099                 break;
4100         }
4101
4102         case DIOCCLRSRCNODES: {
4103
4104                 pf_clear_srcnodes(NULL);
4105                 pf_purge_expired_src_nodes();
4106                 break;
4107         }
4108
4109         case DIOCKILLSRCNODES:
4110                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4111                 break;
4112
4113         case DIOCSETHOSTID: {
4114                 u_int32_t       *hostid = (u_int32_t *)addr;
4115
4116                 PF_RULES_WLOCK();
4117                 if (*hostid == 0)
4118                         V_pf_status.hostid = arc4random();
4119                 else
4120                         V_pf_status.hostid = *hostid;
4121                 PF_RULES_WUNLOCK();
4122                 break;
4123         }
4124
4125         case DIOCOSFPFLUSH:
4126                 PF_RULES_WLOCK();
4127                 pf_osfp_flush();
4128                 PF_RULES_WUNLOCK();
4129                 break;
4130
4131         case DIOCIGETIFACES: {
4132                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4133                 struct pfi_kif *ifstore;
4134                 size_t bufsiz;
4135
4136                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4137                         error = ENODEV;
4138                         break;
4139                 }
4140
4141                 if (io->pfiio_size < 0 ||
4142                     io->pfiio_size > pf_ioctl_maxcount ||
4143                     WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4144                         error = EINVAL;
4145                         break;
4146                 }
4147
4148                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4149                 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4150                     M_TEMP, M_NOWAIT);
4151                 if (ifstore == NULL) {
4152                         error = ENOMEM;
4153                         break;
4154                 }
4155
4156                 PF_RULES_RLOCK();
4157                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4158                 PF_RULES_RUNLOCK();
4159                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4160                 free(ifstore, M_TEMP);
4161                 break;
4162         }
4163
4164         case DIOCSETIFFLAG: {
4165                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4166
4167                 PF_RULES_WLOCK();
4168                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4169                 PF_RULES_WUNLOCK();
4170                 break;
4171         }
4172
4173         case DIOCCLRIFFLAG: {
4174                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4175
4176                 PF_RULES_WLOCK();
4177                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4178                 PF_RULES_WUNLOCK();
4179                 break;
4180         }
4181
4182         default:
4183                 error = ENODEV;
4184                 break;
4185         }
4186 fail:
4187         if (sx_xlocked(&pf_ioctl_lock))
4188                 sx_xunlock(&pf_ioctl_lock);
4189         CURVNET_RESTORE();
4190
4191         return (error);
4192 }
4193
4194 void
4195 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4196 {
4197         bzero(sp, sizeof(struct pfsync_state));
4198
4199         /* copy from state key */
4200         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4201         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4202         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4203         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4204         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4205         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4206         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4207         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4208         sp->proto = st->key[PF_SK_WIRE]->proto;
4209         sp->af = st->key[PF_SK_WIRE]->af;
4210
4211         /* copy from state */
4212         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4213         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4214         sp->creation = htonl(time_uptime - st->creation);
4215         sp->expire = pf_state_expires(st);
4216         if (sp->expire <= time_uptime)
4217                 sp->expire = htonl(0);
4218         else
4219                 sp->expire = htonl(sp->expire - time_uptime);
4220
4221         sp->direction = st->direction;
4222         sp->log = st->log;
4223         sp->timeout = st->timeout;
4224         sp->state_flags = st->state_flags;
4225         if (st->src_node)
4226                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4227         if (st->nat_src_node)
4228                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4229
4230         sp->id = st->id;
4231         sp->creatorid = st->creatorid;
4232         pf_state_peer_hton(&st->src, &sp->src);
4233         pf_state_peer_hton(&st->dst, &sp->dst);
4234
4235         if (st->rule.ptr == NULL)
4236                 sp->rule = htonl(-1);
4237         else
4238                 sp->rule = htonl(st->rule.ptr->nr);
4239         if (st->anchor.ptr == NULL)
4240                 sp->anchor = htonl(-1);
4241         else
4242                 sp->anchor = htonl(st->anchor.ptr->nr);
4243         if (st->nat_rule.ptr == NULL)
4244                 sp->nat_rule = htonl(-1);
4245         else
4246                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4247
4248         pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4249             sp->packets[0]);
4250         pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4251             sp->packets[1]);
4252         pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4253         pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4254
4255 }
4256
4257 static void
4258 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4259 {
4260         struct pfr_ktable *kt;
4261
4262         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4263
4264         kt = aw->p.tbl;
4265         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4266                 kt = kt->pfrkt_root;
4267         aw->p.tbl = NULL;
4268         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4269                 kt->pfrkt_cnt : -1;
4270 }
4271
4272 /*
4273  * XXX - Check for version missmatch!!!
4274  */
4275 static void
4276 pf_clear_states(void)
4277 {
4278         struct pf_state *s;
4279         u_int i;
4280
4281         for (i = 0; i <= pf_hashmask; i++) {
4282                 struct pf_idhash *ih = &V_pf_idhash[i];
4283 relock:
4284                 PF_HASHROW_LOCK(ih);
4285                 LIST_FOREACH(s, &ih->states, entry) {
4286                         s->timeout = PFTM_PURGE;
4287                         /* Don't send out individual delete messages. */
4288                         s->state_flags |= PFSTATE_NOSYNC;
4289                         pf_unlink_state(s, PF_ENTER_LOCKED);
4290                         goto relock;
4291                 }
4292                 PF_HASHROW_UNLOCK(ih);
4293         }
4294 }
4295
4296 static int
4297 pf_clear_tables(void)
4298 {
4299         struct pfioc_table io;
4300         int error;
4301
4302         bzero(&io, sizeof(io));
4303
4304         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4305             io.pfrio_flags);
4306
4307         return (error);
4308 }
4309
4310 static void
4311 pf_clear_srcnodes(struct pf_ksrc_node *n)
4312 {
4313         struct pf_state *s;
4314         int i;
4315
4316         for (i = 0; i <= pf_hashmask; i++) {
4317                 struct pf_idhash *ih = &V_pf_idhash[i];
4318
4319                 PF_HASHROW_LOCK(ih);
4320                 LIST_FOREACH(s, &ih->states, entry) {
4321                         if (n == NULL || n == s->src_node)
4322                                 s->src_node = NULL;
4323                         if (n == NULL || n == s->nat_src_node)
4324                                 s->nat_src_node = NULL;
4325                 }
4326                 PF_HASHROW_UNLOCK(ih);
4327         }
4328
4329         if (n == NULL) {
4330                 struct pf_srchash *sh;
4331
4332                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4333                     i++, sh++) {
4334                         PF_HASHROW_LOCK(sh);
4335                         LIST_FOREACH(n, &sh->nodes, entry) {
4336                                 n->expire = 1;
4337                                 n->states = 0;
4338                         }
4339                         PF_HASHROW_UNLOCK(sh);
4340                 }
4341         } else {
4342                 /* XXX: hash slot should already be locked here. */
4343                 n->expire = 1;
4344                 n->states = 0;
4345         }
4346 }
4347
4348 static void
4349 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4350 {
4351         struct pf_ksrc_node_list         kill;
4352
4353         LIST_INIT(&kill);
4354         for (int i = 0; i <= pf_srchashmask; i++) {
4355                 struct pf_srchash *sh = &V_pf_srchash[i];
4356                 struct pf_ksrc_node *sn, *tmp;
4357
4358                 PF_HASHROW_LOCK(sh);
4359                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4360                         if (PF_MATCHA(psnk->psnk_src.neg,
4361                               &psnk->psnk_src.addr.v.a.addr,
4362                               &psnk->psnk_src.addr.v.a.mask,
4363                               &sn->addr, sn->af) &&
4364                             PF_MATCHA(psnk->psnk_dst.neg,
4365                               &psnk->psnk_dst.addr.v.a.addr,
4366                               &psnk->psnk_dst.addr.v.a.mask,
4367                               &sn->raddr, sn->af)) {
4368                                 pf_unlink_src_node(sn);
4369                                 LIST_INSERT_HEAD(&kill, sn, entry);
4370                                 sn->expire = 1;
4371                         }
4372                 PF_HASHROW_UNLOCK(sh);
4373         }
4374
4375         for (int i = 0; i <= pf_hashmask; i++) {
4376                 struct pf_idhash *ih = &V_pf_idhash[i];
4377                 struct pf_state *s;
4378
4379                 PF_HASHROW_LOCK(ih);
4380                 LIST_FOREACH(s, &ih->states, entry) {
4381                         if (s->src_node && s->src_node->expire == 1)
4382                                 s->src_node = NULL;
4383                         if (s->nat_src_node && s->nat_src_node->expire == 1)
4384                                 s->nat_src_node = NULL;
4385                 }
4386                 PF_HASHROW_UNLOCK(ih);
4387         }
4388
4389         psnk->psnk_killed = pf_free_src_nodes(&kill);
4390 }
4391
4392 /*
4393  * XXX - Check for version missmatch!!!
4394  */
4395
4396 /*
4397  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4398  */
4399 static int
4400 shutdown_pf(void)
4401 {
4402         int error = 0;
4403         u_int32_t t[5];
4404         char nn = '\0';
4405
4406         do {
4407                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4408                     != 0) {
4409                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4410                         break;
4411                 }
4412                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4413                     != 0) {
4414                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4415                         break;          /* XXX: rollback? */
4416                 }
4417                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4418                     != 0) {
4419                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4420                         break;          /* XXX: rollback? */
4421                 }
4422                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4423                     != 0) {
4424                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4425                         break;          /* XXX: rollback? */
4426                 }
4427                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4428                     != 0) {
4429                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4430                         break;          /* XXX: rollback? */
4431                 }
4432
4433                 /* XXX: these should always succeed here */
4434                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4435                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4436                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4437                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4438                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4439
4440                 if ((error = pf_clear_tables()) != 0)
4441                         break;
4442
4443 #ifdef ALTQ
4444                 if ((error = pf_begin_altq(&t[0])) != 0) {
4445                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4446                         break;
4447                 }
4448                 pf_commit_altq(t[0]);
4449 #endif
4450
4451                 pf_clear_states();
4452
4453                 pf_clear_srcnodes(NULL);
4454
4455                 /* status does not use malloced mem so no need to cleanup */
4456                 /* fingerprints and interfaces have their own cleanup code */
4457         } while(0);
4458
4459         return (error);
4460 }
4461
4462 #ifdef INET
4463 static int
4464 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4465     struct inpcb *inp)
4466 {
4467         int chk;
4468
4469         chk = pf_test(PF_IN, flags, ifp, m, inp);
4470         if (chk && *m) {
4471                 m_freem(*m);
4472                 *m = NULL;
4473         }
4474
4475         if (chk != PF_PASS)
4476                 return (EACCES);
4477         return (0);
4478 }
4479
4480 static int
4481 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4482     struct inpcb *inp)
4483 {
4484         int chk;
4485
4486         chk = pf_test(PF_OUT, flags, ifp, m, inp);
4487         if (chk && *m) {
4488                 m_freem(*m);
4489                 *m = NULL;
4490         }
4491
4492         if (chk != PF_PASS)
4493                 return (EACCES);
4494         return (0);
4495 }
4496 #endif
4497
4498 #ifdef INET6
4499 static int
4500 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4501     struct inpcb *inp)
4502 {
4503         int chk;
4504
4505         /*
4506          * In case of loopback traffic IPv6 uses the real interface in
4507          * order to support scoped addresses. In order to support stateful
4508          * filtering we have change this to lo0 as it is the case in IPv4.
4509          */
4510         CURVNET_SET(ifp->if_vnet);
4511         chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4512         CURVNET_RESTORE();
4513         if (chk && *m) {
4514                 m_freem(*m);
4515                 *m = NULL;
4516         }
4517         if (chk != PF_PASS)
4518                 return (EACCES);
4519         return (0);
4520 }
4521
4522 static int
4523 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4524     struct inpcb *inp)
4525 {
4526         int chk;
4527
4528         CURVNET_SET(ifp->if_vnet);
4529         chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4530         CURVNET_RESTORE();
4531         if (chk && *m) {
4532                 m_freem(*m);
4533                 *m = NULL;
4534         }
4535         if (chk != PF_PASS)
4536                 return (EACCES);
4537         return (0);
4538 }
4539 #endif /* INET6 */
4540
4541 static int
4542 hook_pf(void)
4543 {
4544 #ifdef INET
4545         struct pfil_head *pfh_inet;
4546 #endif
4547 #ifdef INET6
4548         struct pfil_head *pfh_inet6;
4549 #endif
4550
4551         if (V_pf_pfil_hooked)
4552                 return (0);
4553
4554 #ifdef INET
4555         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
4556         if (pfh_inet == NULL)
4557                 return (ESRCH); /* XXX */
4558         pfil_add_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
4559         pfil_add_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
4560 #endif
4561 #ifdef INET6
4562         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
4563         if (pfh_inet6 == NULL) {
4564 #ifdef INET
4565                 pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
4566                     pfh_inet);
4567                 pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
4568                     pfh_inet);
4569 #endif
4570                 return (ESRCH); /* XXX */
4571         }
4572         pfil_add_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
4573         pfil_add_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
4574 #endif
4575
4576         V_pf_pfil_hooked = 1;
4577         return (0);
4578 }
4579
4580 static int
4581 dehook_pf(void)
4582 {
4583 #ifdef INET
4584         struct pfil_head *pfh_inet;
4585 #endif
4586 #ifdef INET6
4587         struct pfil_head *pfh_inet6;
4588 #endif
4589
4590         if (V_pf_pfil_hooked == 0)
4591                 return (0);
4592
4593 #ifdef INET
4594         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
4595         if (pfh_inet == NULL)
4596                 return (ESRCH); /* XXX */
4597         pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
4598             pfh_inet);
4599         pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
4600             pfh_inet);
4601 #endif
4602 #ifdef INET6
4603         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
4604         if (pfh_inet6 == NULL)
4605                 return (ESRCH); /* XXX */
4606         pfil_remove_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
4607             pfh_inet6);
4608         pfil_remove_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
4609             pfh_inet6);
4610 #endif
4611
4612         V_pf_pfil_hooked = 0;
4613         return (0);
4614 }
4615
4616 static void
4617 pf_load_vnet(void)
4618 {
4619         V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4620             NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4621
4622         pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4623             PF_RULE_TAG_HASH_SIZE_DEFAULT);
4624 #ifdef ALTQ
4625         pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4626             PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4627 #endif
4628
4629         pfattach_vnet();
4630         V_pf_vnet_active = 1;
4631 }
4632
4633 static int
4634 pf_load(void)
4635 {
4636         int error;
4637
4638         rm_init(&pf_rules_lock, "pf rulesets");
4639         sx_init(&pf_ioctl_lock, "pf ioctl");
4640         sx_init(&pf_end_lock, "pf end thread");
4641
4642         pf_mtag_initialize();
4643
4644         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
4645         if (pf_dev == NULL)
4646                 return (ENOMEM);
4647
4648         pf_end_threads = 0;
4649         error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4650         if (error != 0)
4651                 return (error);
4652
4653         pfi_initialize();
4654
4655         return (0);
4656 }
4657
4658 static void
4659 pf_unload_vnet(void)
4660 {
4661         int error;
4662
4663         V_pf_vnet_active = 0;
4664         V_pf_status.running = 0;
4665         error = dehook_pf();
4666         if (error) {
4667                 /*
4668                  * Should not happen!
4669                  * XXX Due to error code ESRCH, kldunload will show
4670                  * a message like 'No such process'.
4671                  */
4672                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4673                 return;
4674         }
4675
4676         PF_RULES_WLOCK();
4677         shutdown_pf();
4678         PF_RULES_WUNLOCK();
4679
4680         swi_remove(V_pf_swi_cookie);
4681
4682         pf_unload_vnet_purge();
4683
4684         pf_normalize_cleanup();
4685         PF_RULES_WLOCK();
4686         pfi_cleanup_vnet();
4687         PF_RULES_WUNLOCK();
4688         pfr_cleanup();
4689         pf_osfp_flush();
4690         pf_cleanup();
4691         if (IS_DEFAULT_VNET(curvnet))
4692                 pf_mtag_cleanup();
4693
4694         pf_cleanup_tagset(&V_pf_tags);
4695 #ifdef ALTQ
4696         pf_cleanup_tagset(&V_pf_qids);
4697 #endif
4698         uma_zdestroy(V_pf_tag_z);
4699
4700         /* Free counters last as we updated them during shutdown. */
4701         counter_u64_free(V_pf_default_rule.evaluations);
4702         for (int i = 0; i < 2; i++) {
4703                 counter_u64_free(V_pf_default_rule.packets[i]);
4704                 counter_u64_free(V_pf_default_rule.bytes[i]);
4705         }
4706         counter_u64_free(V_pf_default_rule.states_cur);
4707         counter_u64_free(V_pf_default_rule.states_tot);
4708         counter_u64_free(V_pf_default_rule.src_nodes);
4709
4710         for (int i = 0; i < PFRES_MAX; i++)
4711                 counter_u64_free(V_pf_status.counters[i]);
4712         for (int i = 0; i < LCNT_MAX; i++)
4713                 counter_u64_free(V_pf_status.lcounters[i]);
4714         for (int i = 0; i < FCNT_MAX; i++)
4715                 counter_u64_free(V_pf_status.fcounters[i]);
4716         for (int i = 0; i < SCNT_MAX; i++)
4717                 counter_u64_free(V_pf_status.scounters[i]);
4718 }
4719
4720 static void
4721 pf_unload(void)
4722 {
4723
4724         sx_xlock(&pf_end_lock);
4725         pf_end_threads = 1;
4726         while (pf_end_threads < 2) {
4727                 wakeup_one(pf_purge_thread);
4728                 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4729         }
4730         sx_xunlock(&pf_end_lock);
4731
4732         if (pf_dev != NULL)
4733                 destroy_dev(pf_dev);
4734
4735         pfi_cleanup();
4736
4737         rm_destroy(&pf_rules_lock);
4738         sx_destroy(&pf_ioctl_lock);
4739         sx_destroy(&pf_end_lock);
4740 }
4741
4742 static void
4743 vnet_pf_init(void *unused __unused)
4744 {
4745
4746         pf_load_vnet();
4747 }
4748 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4749     vnet_pf_init, NULL);
4750
4751 static void
4752 vnet_pf_uninit(const void *unused __unused)
4753 {
4754
4755         pf_unload_vnet();
4756 }
4757 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4758 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4759     vnet_pf_uninit, NULL);
4760
4761
4762 static int
4763 pf_modevent(module_t mod, int type, void *data)
4764 {
4765         int error = 0;
4766
4767         switch(type) {
4768         case MOD_LOAD:
4769                 error = pf_load();
4770                 break;
4771         case MOD_UNLOAD:
4772                 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4773                  * the vnet_pf_uninit()s */
4774                 break;
4775         default:
4776                 error = EINVAL;
4777                 break;
4778         }
4779
4780         return (error);
4781 }
4782
4783 static moduledata_t pf_mod = {
4784         "pf",
4785         pf_modevent,
4786         0
4787 };
4788
4789 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4790 MODULE_VERSION(pf, PF_MODVER);