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