sys/netpfil/pf/pf_ioctl.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-2-Clause
   3  *
   4  * Copyright (c) 2001 Daniel Hartmeier
   5  * Copyright (c) 2002,2003 Henning Brauer
   6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  *
  13  *    - Redistributions of source code must retain the above copyright
  14  *      notice, this list of conditions and the following disclaimer.
  15  *    - Redistributions in binary form must reproduce the above
  16  *      copyright notice, this list of conditions and the following
  17  *      disclaimer in the documentation and/or other materials provided
  18  *      with the distribution.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  31  * POSSIBILITY OF SUCH DAMAGE.
  32  *
  33  * Effort sponsored in part by the Defense Advanced Research Projects
  34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  36  *
  37  *      $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
  38  */
  39
  40 #include <sys/cdefs.h>
  41 __FBSDID("$FreeBSD$");
  42
  43 #include "opt_inet.h"
  44 #include "opt_inet6.h"
  45 #include "opt_bpf.h"
  46 #include "opt_pf.h"
  47
  48 #include <sys/param.h>
  49 #include <sys/_bitset.h>
  50 #include <sys/bitset.h>
  51 #include <sys/bus.h>
  52 #include <sys/conf.h>
  53 #include <sys/endian.h>
  54 #include <sys/fcntl.h>
  55 #include <sys/filio.h>
  56 #include <sys/hash.h>
  57 #include <sys/interrupt.h>
  58 #include <sys/jail.h>
  59 #include <sys/kernel.h>
  60 #include <sys/kthread.h>
  61 #include <sys/lock.h>
  62 #include <sys/mbuf.h>
  63 #include <sys/module.h>
  64 #include <sys/proc.h>
  65 #include <sys/smp.h>
  66 #include <sys/socket.h>
  67 #include <sys/sysctl.h>
  68 #include <sys/md5.h>
  69 #include <sys/ucred.h>
  70
  71 #include <net/if.h>
  72 #include <net/if_var.h>
  73 #include <net/vnet.h>
  74 #include <net/route.h>
  75 #include <net/pfil.h>
  76 #include <net/pfvar.h>
  77 #include <net/if_pfsync.h>
  78 #include <net/if_pflog.h>
  79
  80 #include <netinet/in.h>
  81 #include <netinet/ip.h>
  82 #include <netinet/ip_var.h>
  83 #include <netinet6/ip6_var.h>
  84 #include <netinet/ip_icmp.h>
  85
  86 #ifdef INET6
  87 #include <netinet/ip6.h>
  88 #endif /* INET6 */
  89
  90 #ifdef ALTQ
  91 #include <net/altq/altq.h>
  92 #endif
  93
  94 static struct pf_kpool  *pf_get_kpool(char *, u_int32_t, u_int8_t, u_int32_t,
  95                             u_int8_t, u_int8_t, u_int8_t);
  96
  97 static void              pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
  98 static void              pf_empty_kpool(struct pf_kpalist *);
  99 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
 100                             struct thread *);
 101 #ifdef ALTQ
 102 static int               pf_begin_altq(u_int32_t *);
 103 static int               pf_rollback_altq(u_int32_t);
 104 static int               pf_commit_altq(u_int32_t);
 105 static int               pf_enable_altq(struct pf_altq *);
 106 static int               pf_disable_altq(struct pf_altq *);
 107 static u_int32_t         pf_qname2qid(char *);
 108 static void              pf_qid_unref(u_int32_t);
 109 #endif /* ALTQ */
 110 static int               pf_begin_rules(u_int32_t *, int, const char *);
 111 static int               pf_rollback_rules(u_int32_t, int, char *);
 112 static int               pf_setup_pfsync_matching(struct pf_kruleset *);
 113 static void              pf_hash_rule(MD5_CTX *, struct pf_krule *);
 114 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 115 static int               pf_commit_rules(u_int32_t, int, char *);
 116 static int               pf_addr_setup(struct pf_kruleset *,
 117                             struct pf_addr_wrap *, sa_family_t);
 118 static void              pf_addr_copyout(struct pf_addr_wrap *);
 119 static void              pf_src_node_copy(const struct pf_ksrc_node *,
 120                             struct pf_src_node *);
 121 #ifdef ALTQ
 122 static int               pf_export_kaltq(struct pf_altq *,
 123                             struct pfioc_altq_v1 *, size_t);
 124 static int               pf_import_kaltq(struct pfioc_altq_v1 *,
 125                             struct pf_altq *, size_t);
 126 #endif /* ALTQ */
 127
 128 VNET_DEFINE(struct pf_krule,    pf_default_rule);
 129
 130 #ifdef ALTQ
 131 VNET_DEFINE_STATIC(int,         pf_altq_running);
 132 #define V_pf_altq_running       VNET(pf_altq_running)
 133 #endif
 134
 135 #define TAGID_MAX        50000
 136 struct pf_tagname {
 137         TAILQ_ENTRY(pf_tagname) namehash_entries;
 138         TAILQ_ENTRY(pf_tagname) taghash_entries;
 139         char                    name[PF_TAG_NAME_SIZE];
 140         uint16_t                tag;
 141         int                     ref;
 142 };
 143
 144 struct pf_tagset {
 145         TAILQ_HEAD(, pf_tagname)        *namehash;
 146         TAILQ_HEAD(, pf_tagname)        *taghash;
 147         unsigned int                     mask;
 148         uint32_t                         seed;
 149         BITSET_DEFINE(, TAGID_MAX)       avail;
 150 };
 151
 152 VNET_DEFINE(struct pf_tagset, pf_tags);
 153 #define V_pf_tags       VNET(pf_tags)
 154 static unsigned int     pf_rule_tag_hashsize;
 155 #define PF_RULE_TAG_HASH_SIZE_DEFAULT   128
 156 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
 157     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
 158     "Size of pf(4) rule tag hashtable");
 159
 160 #ifdef ALTQ
 161 VNET_DEFINE(struct pf_tagset, pf_qids);
 162 #define V_pf_qids       VNET(pf_qids)
 163 static unsigned int     pf_queue_tag_hashsize;
 164 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT  128
 165 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
 166     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
 167     "Size of pf(4) queue tag hashtable");
 168 #endif
 169 VNET_DEFINE(uma_zone_t,  pf_tag_z);
 170 #define V_pf_tag_z               VNET(pf_tag_z)
 171 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 172 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 173
 174 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 175 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 176 #endif
 177
 178 static void              pf_init_tagset(struct pf_tagset *, unsigned int *,
 179                             unsigned int);
 180 static void              pf_cleanup_tagset(struct pf_tagset *);
 181 static uint16_t          tagname2hashindex(const struct pf_tagset *, const char *);
 182 static uint16_t          tag2hashindex(const struct pf_tagset *, uint16_t);
 183 static u_int16_t         tagname2tag(struct pf_tagset *, char *);
 184 static u_int16_t         pf_tagname2tag(char *);
 185 static void              tag_unref(struct pf_tagset *, u_int16_t);
 186
 187 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 188
 189 struct cdev *pf_dev;
 190
 191 /*
 192  * XXX - These are new and need to be checked when moveing to a new version
 193  */
 194 static void              pf_clear_states(void);
 195 static int               pf_clear_tables(void);
 196 static void              pf_clear_srcnodes(struct pf_ksrc_node *);
 197 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 198 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 199
 200 /*
 201  * Wrapper functions for pfil(9) hooks
 202  */
 203 #ifdef INET
 204 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 205     int dir, int flags, struct inpcb *inp);
 206 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 207     int dir, int flags, struct inpcb *inp);
 208 #endif
 209 #ifdef INET6
 210 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 211     int dir, int flags, struct inpcb *inp);
 212 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 213     int dir, int flags, struct inpcb *inp);
 214 #endif
 215
 216 static int              hook_pf(void);
 217 static int              dehook_pf(void);
 218 static int              shutdown_pf(void);
 219 static int              pf_load(void);
 220 static void             pf_unload(void);
 221
 222 static struct cdevsw pf_cdevsw = {
 223         .d_ioctl =      pfioctl,
 224         .d_name =       PF_NAME,
 225         .d_version =    D_VERSION,
 226 };
 227
 228 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
 229 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 230
 231 /*
 232  * We need a flag that is neither hooked nor running to know when
 233  * the VNET is "valid".  We primarily need this to control (global)
 234  * external event, e.g., eventhandlers.
 235  */
 236 VNET_DEFINE(int, pf_vnet_active);
 237 #define V_pf_vnet_active        VNET(pf_vnet_active)
 238
 239 int pf_end_threads;
 240 struct proc *pf_purge_proc;
 241
 242 struct rmlock                   pf_rules_lock;
 243 struct sx                       pf_ioctl_lock;
 244 struct sx                       pf_end_lock;
 245
 246 /* pfsync */
 247 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
 248 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
 249 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
 250 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
 251 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
 252 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
 253 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
 254
 255 /* pflog */
 256 pflog_packet_t                  *pflog_packet_ptr = NULL;
 257
 258 extern u_long   pf_ioctl_maxcount;
 259
 260 static void
 261 pfattach_vnet(void)
 262 {
 263         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 264
 265         pf_initialize();
 266         pfr_initialize();
 267         pfi_initialize_vnet();
 268         pf_normalize_init();
 269
 270         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 271         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 272
 273         RB_INIT(&V_pf_anchors);
 274         pf_init_kruleset(&pf_main_ruleset);
 275
 276         /* default rule should never be garbage collected */
 277         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 278 #ifdef PF_DEFAULT_TO_DROP
 279         V_pf_default_rule.action = PF_DROP;
 280 #else
 281         V_pf_default_rule.action = PF_PASS;
 282 #endif
 283         V_pf_default_rule.nr = -1;
 284         V_pf_default_rule.rtableid = -1;
 285
 286         V_pf_default_rule.evaluations = counter_u64_alloc(M_WAITOK);
 287         for (int i = 0; i < 2; i++) {
 288                 V_pf_default_rule.packets[i] = counter_u64_alloc(M_WAITOK);
 289                 V_pf_default_rule.bytes[i] = counter_u64_alloc(M_WAITOK);
 290         }
 291         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 292         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 293         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 294
 295         /* initialize default timeouts */
 296         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 297         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 298         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 299         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 300         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 301         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 302         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 303         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 304         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 305         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 306         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 307         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 308         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 309         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 310         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 311         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 312         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 313         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 314         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 315         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 316
 317         bzero(&V_pf_status, sizeof(V_pf_status));
 318         V_pf_status.debug = PF_DEBUG_URGENT;
 319
 320         V_pf_pfil_hooked = 0;
 321
 322         /* XXX do our best to avoid a conflict */
 323         V_pf_status.hostid = arc4random();
 324
 325         for (int i = 0; i < PFRES_MAX; i++)
 326                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 327         for (int i = 0; i < LCNT_MAX; i++)
 328                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 329         for (int i = 0; i < FCNT_MAX; i++)
 330                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 331         for (int i = 0; i < SCNT_MAX; i++)
 332                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 333
 334         if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
 335             INTR_MPSAFE, &V_pf_swi_cookie) != 0)
 336                 /* XXXGL: leaked all above. */
 337                 return;
 338 }
 339
 340 static struct pf_kpool *
 341 pf_get_kpool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 342     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 343     u_int8_t check_ticket)
 344 {
 345         struct pf_kruleset      *ruleset;
 346         struct pf_krule         *rule;
 347         int                      rs_num;
 348
 349         ruleset = pf_find_kruleset(anchor);
 350         if (ruleset == NULL)
 351                 return (NULL);
 352         rs_num = pf_get_ruleset_number(rule_action);
 353         if (rs_num >= PF_RULESET_MAX)
 354                 return (NULL);
 355         if (active) {
 356                 if (check_ticket && ticket !=
 357                     ruleset->rules[rs_num].active.ticket)
 358                         return (NULL);
 359                 if (r_last)
 360                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 361                             pf_krulequeue);
 362                 else
 363                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 364         } else {
 365                 if (check_ticket && ticket !=
 366                     ruleset->rules[rs_num].inactive.ticket)
 367                         return (NULL);
 368                 if (r_last)
 369                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 370                             pf_krulequeue);
 371                 else
 372                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 373         }
 374         if (!r_last) {
 375                 while ((rule != NULL) && (rule->nr != rule_number))
 376                         rule = TAILQ_NEXT(rule, entries);
 377         }
 378         if (rule == NULL)
 379                 return (NULL);
 380
 381         return (&rule->rpool);
 382 }
 383
 384 static void
 385 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
 386 {
 387         struct pf_kpooladdr     *mv_pool_pa;
 388
 389         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 390                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 391                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 392         }
 393 }
 394
 395 static void
 396 pf_empty_kpool(struct pf_kpalist *poola)
 397 {
 398         struct pf_kpooladdr *pa;
 399
 400         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 401                 switch (pa->addr.type) {
 402                 case PF_ADDR_DYNIFTL:
 403                         pfi_dynaddr_remove(pa->addr.p.dyn);
 404                         break;
 405                 case PF_ADDR_TABLE:
 406                         /* XXX: this could be unfinished pooladdr on pabuf */
 407                         if (pa->addr.p.tbl != NULL)
 408                                 pfr_detach_table(pa->addr.p.tbl);
 409                         break;
 410                 }
 411                 if (pa->kif)
 412                         pfi_kkif_unref(pa->kif);
 413                 TAILQ_REMOVE(poola, pa, entries);
 414                 free(pa, M_PFRULE);
 415         }
 416 }
 417
 418 static void
 419 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
 420 {
 421
 422         PF_RULES_WASSERT();
 423
 424         TAILQ_REMOVE(rulequeue, rule, entries);
 425
 426         PF_UNLNKDRULES_LOCK();
 427         rule->rule_flag |= PFRULE_REFS;
 428         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 429         PF_UNLNKDRULES_UNLOCK();
 430 }
 431
 432 void
 433 pf_free_rule(struct pf_krule *rule)
 434 {
 435
 436         PF_RULES_WASSERT();
 437
 438         if (rule->tag)
 439                 tag_unref(&V_pf_tags, rule->tag);
 440         if (rule->match_tag)
 441                 tag_unref(&V_pf_tags, rule->match_tag);
 442 #ifdef ALTQ
 443         if (rule->pqid != rule->qid)
 444                 pf_qid_unref(rule->pqid);
 445         pf_qid_unref(rule->qid);
 446 #endif
 447         switch (rule->src.addr.type) {
 448         case PF_ADDR_DYNIFTL:
 449                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 450                 break;
 451         case PF_ADDR_TABLE:
 452                 pfr_detach_table(rule->src.addr.p.tbl);
 453                 break;
 454         }
 455         switch (rule->dst.addr.type) {
 456         case PF_ADDR_DYNIFTL:
 457                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 458                 break;
 459         case PF_ADDR_TABLE:
 460                 pfr_detach_table(rule->dst.addr.p.tbl);
 461                 break;
 462         }
 463         if (rule->overload_tbl)
 464                 pfr_detach_table(rule->overload_tbl);
 465         if (rule->kif)
 466                 pfi_kkif_unref(rule->kif);
 467         pf_kanchor_remove(rule);
 468         pf_empty_kpool(&rule->rpool.list);
 469
 470         pf_krule_free(rule);
 471 }
 472
 473 static void
 474 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
 475     unsigned int default_size)
 476 {
 477         unsigned int i;
 478         unsigned int hashsize;
 479
 480         if (*tunable_size == 0 || !powerof2(*tunable_size))
 481                 *tunable_size = default_size;
 482
 483         hashsize = *tunable_size;
 484         ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
 485             M_WAITOK);
 486         ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
 487             M_WAITOK);
 488         ts->mask = hashsize - 1;
 489         ts->seed = arc4random();
 490         for (i = 0; i < hashsize; i++) {
 491                 TAILQ_INIT(&ts->namehash[i]);
 492                 TAILQ_INIT(&ts->taghash[i]);
 493         }
 494         BIT_FILL(TAGID_MAX, &ts->avail);
 495 }
 496
 497 static void
 498 pf_cleanup_tagset(struct pf_tagset *ts)
 499 {
 500         unsigned int i;
 501         unsigned int hashsize;
 502         struct pf_tagname *t, *tmp;
 503
 504         /*
 505          * Only need to clean up one of the hashes as each tag is hashed
 506          * into each table.
 507          */
 508         hashsize = ts->mask + 1;
 509         for (i = 0; i < hashsize; i++)
 510                 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
 511                         uma_zfree(V_pf_tag_z, t);
 512
 513         free(ts->namehash, M_PFHASH);
 514         free(ts->taghash, M_PFHASH);
 515 }
 516
 517 static uint16_t
 518 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
 519 {
 520         size_t len;
 521
 522         len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
 523         return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
 524 }
 525
 526 static uint16_t
 527 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
 528 {
 529
 530         return (tag & ts->mask);
 531 }
 532
 533 static u_int16_t
 534 tagname2tag(struct pf_tagset *ts, char *tagname)
 535 {
 536         struct pf_tagname       *tag;
 537         u_int32_t                index;
 538         u_int16_t                new_tagid;
 539
 540         PF_RULES_WASSERT();
 541
 542         index = tagname2hashindex(ts, tagname);
 543         TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
 544                 if (strcmp(tagname, tag->name) == 0) {
 545                         tag->ref++;
 546                         return (tag->tag);
 547                 }
 548
 549         /*
 550          * new entry
 551          *
 552          * to avoid fragmentation, we do a linear search from the beginning
 553          * and take the first free slot we find.
 554          */
 555         new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
 556         /*
 557          * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
 558          * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
 559          * set.  It may also return a bit number greater than TAGID_MAX due
 560          * to rounding of the number of bits in the vector up to a multiple
 561          * of the vector word size at declaration/allocation time.
 562          */
 563         if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
 564                 return (0);
 565
 566         /* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
 567         BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
 568
 569         /* allocate and fill new struct pf_tagname */
 570         tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
 571         if (tag == NULL)
 572                 return (0);
 573         strlcpy(tag->name, tagname, sizeof(tag->name));
 574         tag->tag = new_tagid;
 575         tag->ref = 1;
 576
 577         /* Insert into namehash */
 578         TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
 579
 580         /* Insert into taghash */
 581         index = tag2hashindex(ts, new_tagid);
 582         TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
 583
 584         return (tag->tag);
 585 }
 586
 587 static void
 588 tag_unref(struct pf_tagset *ts, u_int16_t tag)
 589 {
 590         struct pf_tagname       *t;
 591         uint16_t                 index;
 592
 593         PF_RULES_WASSERT();
 594
 595         index = tag2hashindex(ts, tag);
 596         TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
 597                 if (tag == t->tag) {
 598                         if (--t->ref == 0) {
 599                                 TAILQ_REMOVE(&ts->taghash[index], t,
 600                                     taghash_entries);
 601                                 index = tagname2hashindex(ts, t->name);
 602                                 TAILQ_REMOVE(&ts->namehash[index], t,
 603                                     namehash_entries);
 604                                 /* Bits are 0-based for BIT_SET() */
 605                                 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
 606                                 uma_zfree(V_pf_tag_z, t);
 607                         }
 608                         break;
 609                 }
 610 }
 611
 612 static u_int16_t
 613 pf_tagname2tag(char *tagname)
 614 {
 615         return (tagname2tag(&V_pf_tags, tagname));
 616 }
 617
 618 #ifdef ALTQ
 619 static u_int32_t
 620 pf_qname2qid(char *qname)
 621 {
 622         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 623 }
 624
 625 static void
 626 pf_qid_unref(u_int32_t qid)
 627 {
 628         tag_unref(&V_pf_qids, (u_int16_t)qid);
 629 }
 630
 631 static int
 632 pf_begin_altq(u_int32_t *ticket)
 633 {
 634         struct pf_altq  *altq, *tmp;
 635         int              error = 0;
 636
 637         PF_RULES_WASSERT();
 638
 639         /* Purge the old altq lists */
 640         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 641                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 642                         /* detach and destroy the discipline */
 643                         error = altq_remove(altq);
 644                 }
 645                 free(altq, M_PFALTQ);
 646         }
 647         TAILQ_INIT(V_pf_altq_ifs_inactive);
 648         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 649                 pf_qid_unref(altq->qid);
 650                 free(altq, M_PFALTQ);
 651         }
 652         TAILQ_INIT(V_pf_altqs_inactive);
 653         if (error)
 654                 return (error);
 655         *ticket = ++V_ticket_altqs_inactive;
 656         V_altqs_inactive_open = 1;
 657         return (0);
 658 }
 659
 660 static int
 661 pf_rollback_altq(u_int32_t ticket)
 662 {
 663         struct pf_altq  *altq, *tmp;
 664         int              error = 0;
 665
 666         PF_RULES_WASSERT();
 667
 668         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 669                 return (0);
 670         /* Purge the old altq lists */
 671         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 672                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 673                         /* detach and destroy the discipline */
 674                         error = altq_remove(altq);
 675                 }
 676                 free(altq, M_PFALTQ);
 677         }
 678         TAILQ_INIT(V_pf_altq_ifs_inactive);
 679         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 680                 pf_qid_unref(altq->qid);
 681                 free(altq, M_PFALTQ);
 682         }
 683         TAILQ_INIT(V_pf_altqs_inactive);
 684         V_altqs_inactive_open = 0;
 685         return (error);
 686 }
 687
 688 static int
 689 pf_commit_altq(u_int32_t ticket)
 690 {
 691         struct pf_altqqueue     *old_altqs, *old_altq_ifs;
 692         struct pf_altq          *altq, *tmp;
 693         int                      err, error = 0;
 694
 695         PF_RULES_WASSERT();
 696
 697         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 698                 return (EBUSY);
 699
 700         /* swap altqs, keep the old. */
 701         old_altqs = V_pf_altqs_active;
 702         old_altq_ifs = V_pf_altq_ifs_active;
 703         V_pf_altqs_active = V_pf_altqs_inactive;
 704         V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
 705         V_pf_altqs_inactive = old_altqs;
 706         V_pf_altq_ifs_inactive = old_altq_ifs;
 707         V_ticket_altqs_active = V_ticket_altqs_inactive;
 708
 709         /* Attach new disciplines */
 710         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
 711                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 712                         /* attach the discipline */
 713                         error = altq_pfattach(altq);
 714                         if (error == 0 && V_pf_altq_running)
 715                                 error = pf_enable_altq(altq);
 716                         if (error != 0)
 717                                 return (error);
 718                 }
 719         }
 720
 721         /* Purge the old altq lists */
 722         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 723                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 724                         /* detach and destroy the discipline */
 725                         if (V_pf_altq_running)
 726                                 error = pf_disable_altq(altq);
 727                         err = altq_pfdetach(altq);
 728                         if (err != 0 && error == 0)
 729                                 error = err;
 730                         err = altq_remove(altq);
 731                         if (err != 0 && error == 0)
 732                                 error = err;
 733                 }
 734                 free(altq, M_PFALTQ);
 735         }
 736         TAILQ_INIT(V_pf_altq_ifs_inactive);
 737         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 738                 pf_qid_unref(altq->qid);
 739                 free(altq, M_PFALTQ);
 740         }
 741         TAILQ_INIT(V_pf_altqs_inactive);
 742
 743         V_altqs_inactive_open = 0;
 744         return (error);
 745 }
 746
 747 static int
 748 pf_enable_altq(struct pf_altq *altq)
 749 {
 750         struct ifnet            *ifp;
 751         struct tb_profile        tb;
 752         int                      error = 0;
 753
 754         if ((ifp = ifunit(altq->ifname)) == NULL)
 755                 return (EINVAL);
 756
 757         if (ifp->if_snd.altq_type != ALTQT_NONE)
 758                 error = altq_enable(&ifp->if_snd);
 759
 760         /* set tokenbucket regulator */
 761         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 762                 tb.rate = altq->ifbandwidth;
 763                 tb.depth = altq->tbrsize;
 764                 error = tbr_set(&ifp->if_snd, &tb);
 765         }
 766
 767         return (error);
 768 }
 769
 770 static int
 771 pf_disable_altq(struct pf_altq *altq)
 772 {
 773         struct ifnet            *ifp;
 774         struct tb_profile        tb;
 775         int                      error;
 776
 777         if ((ifp = ifunit(altq->ifname)) == NULL)
 778                 return (EINVAL);
 779
 780         /*
 781          * when the discipline is no longer referenced, it was overridden
 782          * by a new one.  if so, just return.
 783          */
 784         if (altq->altq_disc != ifp->if_snd.altq_disc)
 785                 return (0);
 786
 787         error = altq_disable(&ifp->if_snd);
 788
 789         if (error == 0) {
 790                 /* clear tokenbucket regulator */
 791                 tb.rate = 0;
 792                 error = tbr_set(&ifp->if_snd, &tb);
 793         }
 794
 795         return (error);
 796 }
 797
 798 static int
 799 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
 800     struct pf_altq *altq)
 801 {
 802         struct ifnet    *ifp1;
 803         int              error = 0;
 804
 805         /* Deactivate the interface in question */
 806         altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 807         if ((ifp1 = ifunit(altq->ifname)) == NULL ||
 808             (remove && ifp1 == ifp)) {
 809                 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 810         } else {
 811                 error = altq_add(ifp1, altq);
 812
 813                 if (ticket != V_ticket_altqs_inactive)
 814                         error = EBUSY;
 815
 816                 if (error)
 817                         free(altq, M_PFALTQ);
 818         }
 819
 820         return (error);
 821 }
 822
 823 void
 824 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 825 {
 826         struct pf_altq  *a1, *a2, *a3;
 827         u_int32_t        ticket;
 828         int              error = 0;
 829
 830         /*
 831          * No need to re-evaluate the configuration for events on interfaces
 832          * that do not support ALTQ, as it's not possible for such
 833          * interfaces to be part of the configuration.
 834          */
 835         if (!ALTQ_IS_READY(&ifp->if_snd))
 836                 return;
 837
 838         /* Interrupt userland queue modifications */
 839         if (V_altqs_inactive_open)
 840                 pf_rollback_altq(V_ticket_altqs_inactive);
 841
 842         /* Start new altq ruleset */
 843         if (pf_begin_altq(&ticket))
 844                 return;
 845
 846         /* Copy the current active set */
 847         TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
 848                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 849                 if (a2 == NULL) {
 850                         error = ENOMEM;
 851                         break;
 852                 }
 853                 bcopy(a1, a2, sizeof(struct pf_altq));
 854
 855                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 856                 if (error)
 857                         break;
 858
 859                 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
 860         }
 861         if (error)
 862                 goto out;
 863         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 864                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 865                 if (a2 == NULL) {
 866                         error = ENOMEM;
 867                         break;
 868                 }
 869                 bcopy(a1, a2, sizeof(struct pf_altq));
 870
 871                 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 872                         error = EBUSY;
 873                         free(a2, M_PFALTQ);
 874                         break;
 875                 }
 876                 a2->altq_disc = NULL;
 877                 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
 878                         if (strncmp(a3->ifname, a2->ifname,
 879                                 IFNAMSIZ) == 0) {
 880                                 a2->altq_disc = a3->altq_disc;
 881                                 break;
 882                         }
 883                 }
 884                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 885                 if (error)
 886                         break;
 887
 888                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 889         }
 890
 891 out:
 892         if (error != 0)
 893                 pf_rollback_altq(ticket);
 894         else
 895                 pf_commit_altq(ticket);
 896 }
 897 #endif /* ALTQ */
 898
 899 static int
 900 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 901 {
 902         struct pf_kruleset      *rs;
 903         struct pf_krule         *rule;
 904
 905         PF_RULES_WASSERT();
 906
 907         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 908                 return (EINVAL);
 909         rs = pf_find_or_create_kruleset(anchor);
 910         if (rs == NULL)
 911                 return (EINVAL);
 912         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 913                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 914                 rs->rules[rs_num].inactive.rcount--;
 915         }
 916         *ticket = ++rs->rules[rs_num].inactive.ticket;
 917         rs->rules[rs_num].inactive.open = 1;
 918         return (0);
 919 }
 920
 921 static int
 922 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 923 {
 924         struct pf_kruleset      *rs;
 925         struct pf_krule         *rule;
 926
 927         PF_RULES_WASSERT();
 928
 929         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 930                 return (EINVAL);
 931         rs = pf_find_kruleset(anchor);
 932         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 933             rs->rules[rs_num].inactive.ticket != ticket)
 934                 return (0);
 935         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 936                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 937                 rs->rules[rs_num].inactive.rcount--;
 938         }
 939         rs->rules[rs_num].inactive.open = 0;
 940         return (0);
 941 }
 942
 943 #define PF_MD5_UPD(st, elm)                                             \
 944                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 945
 946 #define PF_MD5_UPD_STR(st, elm)                                         \
 947                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 948
 949 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 950                 (stor) = htonl((st)->elm);                              \
 951                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 952 } while (0)
 953
 954 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 955                 (stor) = htons((st)->elm);                              \
 956                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 957 } while (0)
 958
 959 static void
 960 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 961 {
 962         PF_MD5_UPD(pfr, addr.type);
 963         switch (pfr->addr.type) {
 964                 case PF_ADDR_DYNIFTL:
 965                         PF_MD5_UPD(pfr, addr.v.ifname);
 966                         PF_MD5_UPD(pfr, addr.iflags);
 967                         break;
 968                 case PF_ADDR_TABLE:
 969                         PF_MD5_UPD(pfr, addr.v.tblname);
 970                         break;
 971                 case PF_ADDR_ADDRMASK:
 972                         /* XXX ignore af? */
 973                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 974                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 975                         break;
 976         }
 977
 978         PF_MD5_UPD(pfr, port[0]);
 979         PF_MD5_UPD(pfr, port[1]);
 980         PF_MD5_UPD(pfr, neg);
 981         PF_MD5_UPD(pfr, port_op);
 982 }
 983
 984 static void
 985 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
 986 {
 987         u_int16_t x;
 988         u_int32_t y;
 989
 990         pf_hash_rule_addr(ctx, &rule->src);
 991         pf_hash_rule_addr(ctx, &rule->dst);
 992         PF_MD5_UPD_STR(rule, label);
 993         PF_MD5_UPD_STR(rule, ifname);
 994         PF_MD5_UPD_STR(rule, match_tagname);
 995         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 996         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 997         PF_MD5_UPD_HTONL(rule, prob, y);
 998         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 999         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1000         PF_MD5_UPD(rule, uid.op);
1001         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1002         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1003         PF_MD5_UPD(rule, gid.op);
1004         PF_MD5_UPD_HTONL(rule, rule_flag, y);
1005         PF_MD5_UPD(rule, action);
1006         PF_MD5_UPD(rule, direction);
1007         PF_MD5_UPD(rule, af);
1008         PF_MD5_UPD(rule, quick);
1009         PF_MD5_UPD(rule, ifnot);
1010         PF_MD5_UPD(rule, match_tag_not);
1011         PF_MD5_UPD(rule, natpass);
1012         PF_MD5_UPD(rule, keep_state);
1013         PF_MD5_UPD(rule, proto);
1014         PF_MD5_UPD(rule, type);
1015         PF_MD5_UPD(rule, code);
1016         PF_MD5_UPD(rule, flags);
1017         PF_MD5_UPD(rule, flagset);
1018         PF_MD5_UPD(rule, allow_opts);
1019         PF_MD5_UPD(rule, rt);
1020         PF_MD5_UPD(rule, tos);
1021 }
1022
1023 static int
1024 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1025 {
1026         struct pf_kruleset      *rs;
1027         struct pf_krule         *rule, **old_array;
1028         struct pf_krulequeue    *old_rules;
1029         int                      error;
1030         u_int32_t                old_rcount;
1031
1032         PF_RULES_WASSERT();
1033
1034         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1035                 return (EINVAL);
1036         rs = pf_find_kruleset(anchor);
1037         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1038             ticket != rs->rules[rs_num].inactive.ticket)
1039                 return (EBUSY);
1040
1041         /* Calculate checksum for the main ruleset */
1042         if (rs == &pf_main_ruleset) {
1043                 error = pf_setup_pfsync_matching(rs);
1044                 if (error != 0)
1045                         return (error);
1046         }
1047
1048         /* Swap rules, keep the old. */
1049         old_rules = rs->rules[rs_num].active.ptr;
1050         old_rcount = rs->rules[rs_num].active.rcount;
1051         old_array = rs->rules[rs_num].active.ptr_array;
1052
1053         rs->rules[rs_num].active.ptr =
1054             rs->rules[rs_num].inactive.ptr;
1055         rs->rules[rs_num].active.ptr_array =
1056             rs->rules[rs_num].inactive.ptr_array;
1057         rs->rules[rs_num].active.rcount =
1058             rs->rules[rs_num].inactive.rcount;
1059         rs->rules[rs_num].inactive.ptr = old_rules;
1060         rs->rules[rs_num].inactive.ptr_array = old_array;
1061         rs->rules[rs_num].inactive.rcount = old_rcount;
1062
1063         rs->rules[rs_num].active.ticket =
1064             rs->rules[rs_num].inactive.ticket;
1065         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1066
1067
1068         /* Purge the old rule list. */
1069         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1070                 pf_unlink_rule(old_rules, rule);
1071         if (rs->rules[rs_num].inactive.ptr_array)
1072                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1073         rs->rules[rs_num].inactive.ptr_array = NULL;
1074         rs->rules[rs_num].inactive.rcount = 0;
1075         rs->rules[rs_num].inactive.open = 0;
1076         pf_remove_if_empty_kruleset(rs);
1077
1078         return (0);
1079 }
1080
1081 static int
1082 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1083 {
1084         MD5_CTX                  ctx;
1085         struct pf_krule         *rule;
1086         int                      rs_cnt;
1087         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
1088
1089         MD5Init(&ctx);
1090         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1091                 /* XXX PF_RULESET_SCRUB as well? */
1092                 if (rs_cnt == PF_RULESET_SCRUB)
1093                         continue;
1094
1095                 if (rs->rules[rs_cnt].inactive.ptr_array)
1096                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1097                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1098
1099                 if (rs->rules[rs_cnt].inactive.rcount) {
1100                         rs->rules[rs_cnt].inactive.ptr_array =
1101                             malloc(sizeof(caddr_t) *
1102                             rs->rules[rs_cnt].inactive.rcount,
1103                             M_TEMP, M_NOWAIT);
1104
1105                         if (!rs->rules[rs_cnt].inactive.ptr_array)
1106                                 return (ENOMEM);
1107                 }
1108
1109                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1110                     entries) {
1111                         pf_hash_rule(&ctx, rule);
1112                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1113                 }
1114         }
1115
1116         MD5Final(digest, &ctx);
1117         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1118         return (0);
1119 }
1120
1121 static int
1122 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1123     sa_family_t af)
1124 {
1125         int error = 0;
1126
1127         switch (addr->type) {
1128         case PF_ADDR_TABLE:
1129                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1130                 if (addr->p.tbl == NULL)
1131                         error = ENOMEM;
1132                 break;
1133         case PF_ADDR_DYNIFTL:
1134                 error = pfi_dynaddr_setup(addr, af);
1135                 break;
1136         }
1137
1138         return (error);
1139 }
1140
1141 static void
1142 pf_addr_copyout(struct pf_addr_wrap *addr)
1143 {
1144
1145         switch (addr->type) {
1146         case PF_ADDR_DYNIFTL:
1147                 pfi_dynaddr_copyout(addr);
1148                 break;
1149         case PF_ADDR_TABLE:
1150                 pf_tbladdr_copyout(addr);
1151                 break;
1152         }
1153 }
1154
1155 static void
1156 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1157 {
1158         int     secs = time_uptime, diff;
1159
1160         bzero(out, sizeof(struct pf_src_node));
1161
1162         bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1163         bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1164
1165         if (in->rule.ptr != NULL)
1166                 out->rule.nr = in->rule.ptr->nr;
1167
1168         for (int i = 0; i < 2; i++) {
1169                 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1170                 out->packets[i] = counter_u64_fetch(in->packets[i]);
1171         }
1172
1173         out->states = in->states;
1174         out->conn = in->conn;
1175         out->af = in->af;
1176         out->ruletype = in->ruletype;
1177
1178         out->creation = secs - in->creation;
1179         if (out->expire > secs)
1180                 out->expire -= secs;
1181         else
1182                 out->expire = 0;
1183
1184         /* Adjust the connection rate estimate. */
1185         diff = secs - in->conn_rate.last;
1186         if (diff >= in->conn_rate.seconds)
1187                 out->conn_rate.count = 0;
1188         else
1189                 out->conn_rate.count -=
1190                     in->conn_rate.count * diff /
1191                     in->conn_rate.seconds;
1192 }
1193
1194 #ifdef ALTQ
1195 /*
1196  * Handle export of struct pf_kaltq to user binaries that may be using any
1197  * version of struct pf_altq.
1198  */
1199 static int
1200 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1201 {
1202         u_int32_t version;
1203
1204         if (ioc_size == sizeof(struct pfioc_altq_v0))
1205                 version = 0;
1206         else
1207                 version = pa->version;
1208
1209         if (version > PFIOC_ALTQ_VERSION)
1210                 return (EINVAL);
1211
1212 #define ASSIGN(x) exported_q->x = q->x
1213 #define COPY(x) \
1214         bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1215 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1216 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1217
1218         switch (version) {
1219         case 0: {
1220                 struct pf_altq_v0 *exported_q =
1221                     &((struct pfioc_altq_v0 *)pa)->altq;
1222
1223                 COPY(ifname);
1224
1225                 ASSIGN(scheduler);
1226                 ASSIGN(tbrsize);
1227                 exported_q->tbrsize = SATU16(q->tbrsize);
1228                 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1229
1230                 COPY(qname);
1231                 COPY(parent);
1232                 ASSIGN(parent_qid);
1233                 exported_q->bandwidth = SATU32(q->bandwidth);
1234                 ASSIGN(priority);
1235                 ASSIGN(local_flags);
1236
1237                 ASSIGN(qlimit);
1238                 ASSIGN(flags);
1239
1240                 if (q->scheduler == ALTQT_HFSC) {
1241 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1242 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1243                             SATU32(q->pq_u.hfsc_opts.x)
1244
1245                         ASSIGN_OPT_SATU32(rtsc_m1);
1246                         ASSIGN_OPT(rtsc_d);
1247                         ASSIGN_OPT_SATU32(rtsc_m2);
1248
1249                         ASSIGN_OPT_SATU32(lssc_m1);
1250                         ASSIGN_OPT(lssc_d);
1251                         ASSIGN_OPT_SATU32(lssc_m2);
1252
1253                         ASSIGN_OPT_SATU32(ulsc_m1);
1254                         ASSIGN_OPT(ulsc_d);
1255                         ASSIGN_OPT_SATU32(ulsc_m2);
1256
1257                         ASSIGN_OPT(flags);
1258
1259 #undef ASSIGN_OPT
1260 #undef ASSIGN_OPT_SATU32
1261                 } else
1262                         COPY(pq_u);
1263
1264                 ASSIGN(qid);
1265                 break;
1266         }
1267         case 1: {
1268                 struct pf_altq_v1 *exported_q =
1269                     &((struct pfioc_altq_v1 *)pa)->altq;
1270
1271                 COPY(ifname);
1272
1273                 ASSIGN(scheduler);
1274                 ASSIGN(tbrsize);
1275                 ASSIGN(ifbandwidth);
1276
1277                 COPY(qname);
1278                 COPY(parent);
1279                 ASSIGN(parent_qid);
1280                 ASSIGN(bandwidth);
1281                 ASSIGN(priority);
1282                 ASSIGN(local_flags);
1283
1284                 ASSIGN(qlimit);
1285                 ASSIGN(flags);
1286                 COPY(pq_u);
1287
1288                 ASSIGN(qid);
1289                 break;
1290         }
1291         default:
1292                 panic("%s: unhandled struct pfioc_altq version", __func__);
1293                 break;
1294         }
1295
1296 #undef ASSIGN
1297 #undef COPY
1298 #undef SATU16
1299 #undef SATU32
1300
1301         return (0);
1302 }
1303
1304 /*
1305  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1306  * that may be using any version of it.
1307  */
1308 static int
1309 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1310 {
1311         u_int32_t version;
1312
1313         if (ioc_size == sizeof(struct pfioc_altq_v0))
1314                 version = 0;
1315         else
1316                 version = pa->version;
1317
1318         if (version > PFIOC_ALTQ_VERSION)
1319                 return (EINVAL);
1320
1321 #define ASSIGN(x) q->x = imported_q->x
1322 #define COPY(x) \
1323         bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1324
1325         switch (version) {
1326         case 0: {
1327                 struct pf_altq_v0 *imported_q =
1328                     &((struct pfioc_altq_v0 *)pa)->altq;
1329
1330                 COPY(ifname);
1331
1332                 ASSIGN(scheduler);
1333                 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1334                 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1335
1336                 COPY(qname);
1337                 COPY(parent);
1338                 ASSIGN(parent_qid);
1339                 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1340                 ASSIGN(priority);
1341                 ASSIGN(local_flags);
1342
1343                 ASSIGN(qlimit);
1344                 ASSIGN(flags);
1345
1346                 if (imported_q->scheduler == ALTQT_HFSC) {
1347 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1348
1349                         /*
1350                          * The m1 and m2 parameters are being copied from
1351                          * 32-bit to 64-bit.
1352                          */
1353                         ASSIGN_OPT(rtsc_m1);
1354                         ASSIGN_OPT(rtsc_d);
1355                         ASSIGN_OPT(rtsc_m2);
1356
1357                         ASSIGN_OPT(lssc_m1);
1358                         ASSIGN_OPT(lssc_d);
1359                         ASSIGN_OPT(lssc_m2);
1360
1361                         ASSIGN_OPT(ulsc_m1);
1362                         ASSIGN_OPT(ulsc_d);
1363                         ASSIGN_OPT(ulsc_m2);
1364
1365                         ASSIGN_OPT(flags);
1366
1367 #undef ASSIGN_OPT
1368                 } else
1369                         COPY(pq_u);
1370
1371                 ASSIGN(qid);
1372                 break;
1373         }
1374         case 1: {
1375                 struct pf_altq_v1 *imported_q =
1376                     &((struct pfioc_altq_v1 *)pa)->altq;
1377
1378                 COPY(ifname);
1379
1380                 ASSIGN(scheduler);
1381                 ASSIGN(tbrsize);
1382                 ASSIGN(ifbandwidth);
1383
1384                 COPY(qname);
1385                 COPY(parent);
1386                 ASSIGN(parent_qid);
1387                 ASSIGN(bandwidth);
1388                 ASSIGN(priority);
1389                 ASSIGN(local_flags);
1390
1391                 ASSIGN(qlimit);
1392                 ASSIGN(flags);
1393                 COPY(pq_u);
1394
1395                 ASSIGN(qid);
1396                 break;
1397         }
1398         default:
1399                 panic("%s: unhandled struct pfioc_altq version", __func__);
1400                 break;
1401         }
1402
1403 #undef ASSIGN
1404 #undef COPY
1405
1406         return (0);
1407 }
1408
1409 static struct pf_altq *
1410 pf_altq_get_nth_active(u_int32_t n)
1411 {
1412         struct pf_altq          *altq;
1413         u_int32_t                nr;
1414
1415         nr = 0;
1416         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1417                 if (nr == n)
1418                         return (altq);
1419                 nr++;
1420         }
1421
1422         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1423                 if (nr == n)
1424                         return (altq);
1425                 nr++;
1426         }
1427
1428         return (NULL);
1429 }
1430 #endif /* ALTQ */
1431
1432 void
1433 pf_krule_free(struct pf_krule *rule)
1434 {
1435         if (rule == NULL)
1436                 return;
1437
1438         counter_u64_free(rule->evaluations);
1439         for (int i = 0; i < 2; i++) {
1440                 counter_u64_free(rule->packets[i]);
1441                 counter_u64_free(rule->bytes[i]);
1442         }
1443         counter_u64_free(rule->states_cur);
1444         counter_u64_free(rule->states_tot);
1445         counter_u64_free(rule->src_nodes);
1446         free(rule, M_PFRULE);
1447 }
1448
1449 static void
1450 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1451     struct pf_pooladdr *pool)
1452 {
1453
1454         bzero(pool, sizeof(*pool));
1455         bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1456         strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1457 }
1458
1459 static void
1460 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1461     struct pf_kpooladdr *kpool)
1462 {
1463
1464         bzero(kpool, sizeof(*kpool));
1465         bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1466         strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1467 }
1468
1469 static void
1470 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1471 {
1472         bzero(pool, sizeof(*pool));
1473
1474         bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1475         bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1476
1477         pool->tblidx = kpool->tblidx;
1478         pool->proxy_port[0] = kpool->proxy_port[0];
1479         pool->proxy_port[1] = kpool->proxy_port[1];
1480         pool->opts = kpool->opts;
1481 }
1482
1483 static int
1484 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1485 {
1486         _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1487         _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1488
1489         bzero(kpool, sizeof(*kpool));
1490
1491         bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1492         bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1493
1494         kpool->tblidx = pool->tblidx;
1495         kpool->proxy_port[0] = pool->proxy_port[0];
1496         kpool->proxy_port[1] = pool->proxy_port[1];
1497         kpool->opts = pool->opts;
1498
1499         return (0);
1500 }
1501
1502 static void
1503 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1504 {
1505
1506         bzero(rule, sizeof(*rule));
1507
1508         bcopy(&krule->src, &rule->src, sizeof(rule->src));
1509         bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1510
1511         for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1512                 if (rule->skip[i].ptr == NULL)
1513                         rule->skip[i].nr = -1;
1514                 else
1515                         rule->skip[i].nr = krule->skip[i].ptr->nr;
1516         }
1517
1518         strlcpy(rule->label, krule->label, sizeof(rule->label));
1519         strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1520         strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1521         strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1522         strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1523         strlcpy(rule->match_tagname, krule->match_tagname,
1524             sizeof(rule->match_tagname));
1525         strlcpy(rule->overload_tblname, krule->overload_tblname,
1526             sizeof(rule->overload_tblname));
1527
1528         pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1529
1530         rule->evaluations = counter_u64_fetch(krule->evaluations);
1531         for (int i = 0; i < 2; i++) {
1532                 rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1533                 rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1534         }
1535
1536         /* kif, anchor, overload_tbl are not copied over. */
1537
1538         rule->os_fingerprint = krule->os_fingerprint;
1539
1540         rule->rtableid = krule->rtableid;
1541         bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1542         rule->max_states = krule->max_states;
1543         rule->max_src_nodes = krule->max_src_nodes;
1544         rule->max_src_states = krule->max_src_states;
1545         rule->max_src_conn = krule->max_src_conn;
1546         rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1547         rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1548         rule->qid = krule->qid;
1549         rule->pqid = krule->pqid;
1550         rule->rt_listid = krule->rt_listid;
1551         rule->nr = krule->nr;
1552         rule->prob = krule->prob;
1553         rule->cuid = krule->cuid;
1554         rule->cpid = krule->cpid;
1555
1556         rule->return_icmp = krule->return_icmp;
1557         rule->return_icmp6 = krule->return_icmp6;
1558         rule->max_mss = krule->max_mss;
1559         rule->tag = krule->tag;
1560         rule->match_tag = krule->match_tag;
1561         rule->scrub_flags = krule->scrub_flags;
1562
1563         bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1564         bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1565
1566         rule->rule_flag = krule->rule_flag;
1567         rule->action = krule->action;
1568         rule->direction = krule->direction;
1569         rule->log = krule->log;
1570         rule->logif = krule->logif;
1571         rule->quick = krule->quick;
1572         rule->ifnot = krule->ifnot;
1573         rule->match_tag_not = krule->match_tag_not;
1574         rule->natpass = krule->natpass;
1575
1576         rule->keep_state = krule->keep_state;
1577         rule->af = krule->af;
1578         rule->proto = krule->proto;
1579         rule->type = krule->type;
1580         rule->code = krule->code;
1581         rule->flags = krule->flags;
1582         rule->flagset = krule->flagset;
1583         rule->min_ttl = krule->min_ttl;
1584         rule->allow_opts = krule->allow_opts;
1585         rule->rt = krule->rt;
1586         rule->return_ttl = krule->return_ttl;
1587         rule->tos = krule->tos;
1588         rule->set_tos = krule->set_tos;
1589         rule->anchor_relative = krule->anchor_relative;
1590         rule->anchor_wildcard = krule->anchor_wildcard;
1591
1592         rule->flush = krule->flush;
1593         rule->prio = krule->prio;
1594         rule->set_prio[0] = krule->set_prio[0];
1595         rule->set_prio[1] = krule->set_prio[1];
1596
1597         bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1598
1599         rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1600         rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1601         rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1602 }
1603
1604 static int
1605 pf_check_rule_addr(const struct pf_rule_addr *addr)
1606 {
1607
1608         switch (addr->addr.type) {
1609         case PF_ADDR_ADDRMASK:
1610         case PF_ADDR_NOROUTE:
1611         case PF_ADDR_DYNIFTL:
1612         case PF_ADDR_TABLE:
1613         case PF_ADDR_URPFFAILED:
1614         case PF_ADDR_RANGE:
1615                 break;
1616         default:
1617                 return (EINVAL);
1618         }
1619
1620         if (addr->addr.p.dyn != NULL) {
1621                 return (EINVAL);
1622         }
1623
1624         return (0);
1625 }
1626
1627 static int
1628 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1629 {
1630         int ret;
1631
1632 #ifndef INET
1633         if (rule->af == AF_INET) {
1634                 return (EAFNOSUPPORT);
1635         }
1636 #endif /* INET */
1637 #ifndef INET6
1638         if (rule->af == AF_INET6) {
1639                 return (EAFNOSUPPORT);
1640         }
1641 #endif /* INET6 */
1642
1643         ret = pf_check_rule_addr(&rule->src);
1644         if (ret != 0)
1645                 return (ret);
1646         ret = pf_check_rule_addr(&rule->dst);
1647         if (ret != 0)
1648                 return (ret);
1649
1650         bzero(krule, sizeof(*krule));
1651
1652         bcopy(&rule->src, &krule->src, sizeof(rule->src));
1653         bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1654
1655         strlcpy(krule->label, rule->label, sizeof(rule->label));
1656         strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1657         strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1658         strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1659         strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1660         strlcpy(krule->match_tagname, rule->match_tagname,
1661             sizeof(rule->match_tagname));
1662         strlcpy(krule->overload_tblname, rule->overload_tblname,
1663             sizeof(rule->overload_tblname));
1664
1665         ret = pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1666         if (ret != 0)
1667                 return (ret);
1668
1669         /* Don't allow userspace to set evaulations, packets or bytes. */
1670         /* kif, anchor, overload_tbl are not copied over. */
1671
1672         krule->os_fingerprint = rule->os_fingerprint;
1673
1674         krule->rtableid = rule->rtableid;
1675         bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1676         krule->max_states = rule->max_states;
1677         krule->max_src_nodes = rule->max_src_nodes;
1678         krule->max_src_states = rule->max_src_states;
1679         krule->max_src_conn = rule->max_src_conn;
1680         krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1681         krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1682         krule->qid = rule->qid;
1683         krule->pqid = rule->pqid;
1684         krule->rt_listid = rule->rt_listid;
1685         krule->nr = rule->nr;
1686         krule->prob = rule->prob;
1687         krule->cuid = rule->cuid;
1688         krule->cpid = rule->cpid;
1689
1690         krule->return_icmp = rule->return_icmp;
1691         krule->return_icmp6 = rule->return_icmp6;
1692         krule->max_mss = rule->max_mss;
1693         krule->tag = rule->tag;
1694         krule->match_tag = rule->match_tag;
1695         krule->scrub_flags = rule->scrub_flags;
1696
1697         bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1698         bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1699
1700         krule->rule_flag = rule->rule_flag;
1701         krule->action = rule->action;
1702         krule->direction = rule->direction;
1703         krule->log = rule->log;
1704         krule->logif = rule->logif;
1705         krule->quick = rule->quick;
1706         krule->ifnot = rule->ifnot;
1707         krule->match_tag_not = rule->match_tag_not;
1708         krule->natpass = rule->natpass;
1709
1710         krule->keep_state = rule->keep_state;
1711         krule->af = rule->af;
1712         krule->proto = rule->proto;
1713         krule->type = rule->type;
1714         krule->code = rule->code;
1715         krule->flags = rule->flags;
1716         krule->flagset = rule->flagset;
1717         krule->min_ttl = rule->min_ttl;
1718         krule->allow_opts = rule->allow_opts;
1719         krule->rt = rule->rt;
1720         krule->return_ttl = rule->return_ttl;
1721         krule->tos = rule->tos;
1722         krule->set_tos = rule->set_tos;
1723         krule->anchor_relative = rule->anchor_relative;
1724         krule->anchor_wildcard = rule->anchor_wildcard;
1725
1726         krule->flush = rule->flush;
1727         krule->prio = rule->prio;
1728         krule->set_prio[0] = rule->set_prio[0];
1729         krule->set_prio[1] = rule->set_prio[1];
1730
1731         bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1732
1733         return (0);
1734 }
1735
1736 static int
1737 pf_killstates_row(struct pfioc_state_kill *psk, struct pf_idhash *ih)
1738 {
1739         struct pf_state         *s;
1740         struct pf_state_key     *sk;
1741         struct pf_addr          *srcaddr, *dstaddr;
1742         int                      killed = 0;
1743         u_int16_t                srcport, dstport;
1744
1745 relock_DIOCKILLSTATES:
1746         PF_HASHROW_LOCK(ih);
1747         LIST_FOREACH(s, &ih->states, entry) {
1748                 sk = s->key[PF_SK_WIRE];
1749                 if (s->direction == PF_OUT) {
1750                         srcaddr = &sk->addr[1];
1751                         dstaddr = &sk->addr[0];
1752                         srcport = sk->port[1];
1753                         dstport = sk->port[0];
1754                 } else {
1755                         srcaddr = &sk->addr[0];
1756                         dstaddr = &sk->addr[1];
1757                         srcport = sk->port[0];
1758                         dstport = sk->port[1];
1759                 }
1760
1761                 if (psk->psk_af && sk->af != psk->psk_af)
1762                         continue;
1763
1764                 if (psk->psk_proto && psk->psk_proto != sk->proto)
1765                         continue;
1766
1767                 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
1768                     &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
1769                         continue;
1770
1771                 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
1772                     &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
1773                         continue;
1774
1775                 if (psk->psk_src.port_op != 0 &&
1776                     ! pf_match_port(psk->psk_src.port_op,
1777                     psk->psk_src.port[0], psk->psk_src.port[1], srcport))
1778                         continue;
1779
1780                 if (psk->psk_dst.port_op != 0 &&
1781                     ! pf_match_port(psk->psk_dst.port_op,
1782                     psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
1783                         continue;
1784
1785                 if (psk->psk_label[0] && (! s->rule.ptr->label[0] ||
1786                     strcmp(psk->psk_label, s->rule.ptr->label)))
1787                         continue;
1788
1789                 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
1790                     s->kif->pfik_name))
1791                         continue;
1792
1793                 pf_unlink_state(s, PF_ENTER_LOCKED);
1794                 killed++;
1795                 goto relock_DIOCKILLSTATES;
1796         }
1797         PF_HASHROW_UNLOCK(ih);
1798
1799         return (killed);
1800 }
1801
1802 static int
1803 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1804 {
1805         int                      error = 0;
1806         PF_RULES_RLOCK_TRACKER;
1807
1808         /* XXX keep in sync with switch() below */
1809         if (securelevel_gt(td->td_ucred, 2))
1810                 switch (cmd) {
1811                 case DIOCGETRULES:
1812                 case DIOCGETRULE:
1813                 case DIOCGETADDRS:
1814                 case DIOCGETADDR:
1815                 case DIOCGETSTATE:
1816                 case DIOCSETSTATUSIF:
1817                 case DIOCGETSTATUS:
1818                 case DIOCCLRSTATUS:
1819                 case DIOCNATLOOK:
1820                 case DIOCSETDEBUG:
1821                 case DIOCGETSTATES:
1822                 case DIOCGETTIMEOUT:
1823                 case DIOCCLRRULECTRS:
1824                 case DIOCGETLIMIT:
1825                 case DIOCGETALTQSV0:
1826                 case DIOCGETALTQSV1:
1827                 case DIOCGETALTQV0:
1828                 case DIOCGETALTQV1:
1829                 case DIOCGETQSTATSV0:
1830                 case DIOCGETQSTATSV1:
1831                 case DIOCGETRULESETS:
1832                 case DIOCGETRULESET:
1833                 case DIOCRGETTABLES:
1834                 case DIOCRGETTSTATS:
1835                 case DIOCRCLRTSTATS:
1836                 case DIOCRCLRADDRS:
1837                 case DIOCRADDADDRS:
1838                 case DIOCRDELADDRS:
1839                 case DIOCRSETADDRS:
1840                 case DIOCRGETADDRS:
1841                 case DIOCRGETASTATS:
1842                 case DIOCRCLRASTATS:
1843                 case DIOCRTSTADDRS:
1844                 case DIOCOSFPGET:
1845                 case DIOCGETSRCNODES:
1846                 case DIOCCLRSRCNODES:
1847                 case DIOCIGETIFACES:
1848                 case DIOCGIFSPEEDV0:
1849                 case DIOCGIFSPEEDV1:
1850                 case DIOCSETIFFLAG:
1851                 case DIOCCLRIFFLAG:
1852                         break;
1853                 case DIOCRCLRTABLES:
1854                 case DIOCRADDTABLES:
1855                 case DIOCRDELTABLES:
1856                 case DIOCRSETTFLAGS:
1857                         if (((struct pfioc_table *)addr)->pfrio_flags &
1858                             PFR_FLAG_DUMMY)
1859                                 break; /* dummy operation ok */
1860                         return (EPERM);
1861                 default:
1862                         return (EPERM);
1863                 }
1864
1865         if (!(flags & FWRITE))
1866                 switch (cmd) {
1867                 case DIOCGETRULES:
1868                 case DIOCGETADDRS:
1869                 case DIOCGETADDR:
1870                 case DIOCGETSTATE:
1871                 case DIOCGETSTATUS:
1872                 case DIOCGETSTATES:
1873                 case DIOCGETTIMEOUT:
1874                 case DIOCGETLIMIT:
1875                 case DIOCGETALTQSV0:
1876                 case DIOCGETALTQSV1:
1877                 case DIOCGETALTQV0:
1878                 case DIOCGETALTQV1:
1879                 case DIOCGETQSTATSV0:
1880                 case DIOCGETQSTATSV1:
1881                 case DIOCGETRULESETS:
1882                 case DIOCGETRULESET:
1883                 case DIOCNATLOOK:
1884                 case DIOCRGETTABLES:
1885                 case DIOCRGETTSTATS:
1886                 case DIOCRGETADDRS:
1887                 case DIOCRGETASTATS:
1888                 case DIOCRTSTADDRS:
1889                 case DIOCOSFPGET:
1890                 case DIOCGETSRCNODES:
1891                 case DIOCIGETIFACES:
1892                 case DIOCGIFSPEEDV1:
1893                 case DIOCGIFSPEEDV0:
1894                         break;
1895                 case DIOCRCLRTABLES:
1896                 case DIOCRADDTABLES:
1897                 case DIOCRDELTABLES:
1898                 case DIOCRCLRTSTATS:
1899                 case DIOCRCLRADDRS:
1900                 case DIOCRADDADDRS:
1901                 case DIOCRDELADDRS:
1902                 case DIOCRSETADDRS:
1903                 case DIOCRSETTFLAGS:
1904                         if (((struct pfioc_table *)addr)->pfrio_flags &
1905                             PFR_FLAG_DUMMY) {
1906                                 flags |= FWRITE; /* need write lock for dummy */
1907                                 break; /* dummy operation ok */
1908                         }
1909                         return (EACCES);
1910                 case DIOCGETRULE:
1911                         if (((struct pfioc_rule *)addr)->action ==
1912                             PF_GET_CLR_CNTR)
1913                                 return (EACCES);
1914                         break;
1915                 default:
1916                         return (EACCES);
1917                 }
1918
1919         CURVNET_SET(TD_TO_VNET(td));
1920
1921         switch (cmd) {
1922         case DIOCSTART:
1923                 sx_xlock(&pf_ioctl_lock);
1924                 if (V_pf_status.running)
1925                         error = EEXIST;
1926                 else {
1927                         int cpu;
1928
1929                         error = hook_pf();
1930                         if (error) {
1931                                 DPFPRINTF(PF_DEBUG_MISC,
1932                                     ("pf: pfil registration failed\n"));
1933                                 break;
1934                         }
1935                         V_pf_status.running = 1;
1936                         V_pf_status.since = time_second;
1937
1938                         CPU_FOREACH(cpu)
1939                                 V_pf_stateid[cpu] = time_second;
1940
1941                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1942                 }
1943                 break;
1944
1945         case DIOCSTOP:
1946                 sx_xlock(&pf_ioctl_lock);
1947                 if (!V_pf_status.running)
1948                         error = ENOENT;
1949                 else {
1950                         V_pf_status.running = 0;
1951                         error = dehook_pf();
1952                         if (error) {
1953                                 V_pf_status.running = 1;
1954                                 DPFPRINTF(PF_DEBUG_MISC,
1955                                     ("pf: pfil unregistration failed\n"));
1956                         }
1957                         V_pf_status.since = time_second;
1958                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1959                 }
1960                 break;
1961
1962         case DIOCADDRULE: {
1963                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1964                 struct pf_kruleset      *ruleset;
1965                 struct pf_krule         *rule, *tail;
1966                 struct pf_kpooladdr     *pa;
1967                 struct pfi_kkif         *kif = NULL;
1968                 int                      rs_num;
1969
1970                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1971                         error = EINVAL;
1972                         break;
1973                 }
1974
1975                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1976                 error = pf_rule_to_krule(&pr->rule, rule);
1977                 if (error != 0) {
1978                         free(rule, M_PFRULE);
1979                         break;
1980                 }
1981
1982                 if (rule->ifname[0])
1983                         kif = pf_kkif_create(M_WAITOK);
1984                 rule->evaluations = counter_u64_alloc(M_WAITOK);
1985                 for (int i = 0; i < 2; i++) {
1986                         rule->packets[i] = counter_u64_alloc(M_WAITOK);
1987                         rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1988                 }
1989                 rule->states_cur = counter_u64_alloc(M_WAITOK);
1990                 rule->states_tot = counter_u64_alloc(M_WAITOK);
1991                 rule->src_nodes = counter_u64_alloc(M_WAITOK);
1992                 rule->cuid = td->td_ucred->cr_ruid;
1993                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1994                 TAILQ_INIT(&rule->rpool.list);
1995
1996 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1997
1998                 PF_RULES_WLOCK();
1999                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
2000                 ruleset = pf_find_kruleset(pr->anchor);
2001                 if (ruleset == NULL)
2002                         ERROUT(EINVAL);
2003                 rs_num = pf_get_ruleset_number(pr->rule.action);
2004                 if (rs_num >= PF_RULESET_MAX)
2005                         ERROUT(EINVAL);
2006                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
2007                         DPFPRINTF(PF_DEBUG_MISC,
2008                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
2009                             ruleset->rules[rs_num].inactive.ticket));
2010                         ERROUT(EBUSY);
2011                 }
2012                 if (pr->pool_ticket != V_ticket_pabuf) {
2013                         DPFPRINTF(PF_DEBUG_MISC,
2014                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
2015                             V_ticket_pabuf));
2016                         ERROUT(EBUSY);
2017                 }
2018
2019                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2020                     pf_krulequeue);
2021                 if (tail)
2022                         rule->nr = tail->nr + 1;
2023                 else
2024                         rule->nr = 0;
2025                 if (rule->ifname[0]) {
2026                         rule->kif = pfi_kkif_attach(kif, rule->ifname);
2027                         pfi_kkif_ref(rule->kif);
2028                 } else
2029                         rule->kif = NULL;
2030
2031                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2032                         error = EBUSY;
2033
2034 #ifdef ALTQ
2035                 /* set queue IDs */
2036                 if (rule->qname[0] != 0) {
2037                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2038                                 error = EBUSY;
2039                         else if (rule->pqname[0] != 0) {
2040                                 if ((rule->pqid =
2041                                     pf_qname2qid(rule->pqname)) == 0)
2042                                         error = EBUSY;
2043                         } else
2044                                 rule->pqid = rule->qid;
2045                 }
2046 #endif
2047                 if (rule->tagname[0])
2048                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2049                                 error = EBUSY;
2050                 if (rule->match_tagname[0])
2051                         if ((rule->match_tag =
2052                             pf_tagname2tag(rule->match_tagname)) == 0)
2053                                 error = EBUSY;
2054                 if (rule->rt && !rule->direction)
2055                         error = EINVAL;
2056                 if (!rule->log)
2057                         rule->logif = 0;
2058                 if (rule->logif >= PFLOGIFS_MAX)
2059                         error = EINVAL;
2060                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2061                         error = ENOMEM;
2062                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2063                         error = ENOMEM;
2064                 if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
2065                         error = EINVAL;
2066                 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2067                     (rule->set_prio[0] > PF_PRIO_MAX ||
2068                     rule->set_prio[1] > PF_PRIO_MAX))
2069                         error = EINVAL;
2070                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2071                         if (pa->addr.type == PF_ADDR_TABLE) {
2072                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
2073                                     pa->addr.v.tblname);
2074                                 if (pa->addr.p.tbl == NULL)
2075                                         error = ENOMEM;
2076                         }
2077
2078                 rule->overload_tbl = NULL;
2079                 if (rule->overload_tblname[0]) {
2080                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
2081                             rule->overload_tblname)) == NULL)
2082                                 error = EINVAL;
2083                         else
2084                                 rule->overload_tbl->pfrkt_flags |=
2085                                     PFR_TFLAG_ACTIVE;
2086                 }
2087
2088                 pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2089                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2090                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2091                     (rule->rt > PF_NOPFROUTE)) &&
2092                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
2093                         error = EINVAL;
2094
2095                 if (error) {
2096                         pf_free_rule(rule);
2097                         PF_RULES_WUNLOCK();
2098                         break;
2099                 }
2100
2101                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2102                 counter_u64_zero(rule->evaluations);
2103                 for (int i = 0; i < 2; i++) {
2104                         counter_u64_zero(rule->packets[i]);
2105                         counter_u64_zero(rule->bytes[i]);
2106                 }
2107                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2108                     rule, entries);
2109                 ruleset->rules[rs_num].inactive.rcount++;
2110                 PF_RULES_WUNLOCK();
2111                 break;
2112
2113 #undef ERROUT
2114 DIOCADDRULE_error:
2115                 PF_RULES_WUNLOCK();
2116                 pf_krule_free(rule);
2117                 pf_kkif_free(kif);
2118                 break;
2119         }
2120
2121         case DIOCGETRULES: {
2122                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
2123                 struct pf_kruleset      *ruleset;
2124                 struct pf_krule         *tail;
2125                 int                      rs_num;
2126
2127                 PF_RULES_WLOCK();
2128                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
2129                 ruleset = pf_find_kruleset(pr->anchor);
2130                 if (ruleset == NULL) {
2131                         PF_RULES_WUNLOCK();
2132                         error = EINVAL;
2133                         break;
2134                 }
2135                 rs_num = pf_get_ruleset_number(pr->rule.action);
2136                 if (rs_num >= PF_RULESET_MAX) {
2137                         PF_RULES_WUNLOCK();
2138                         error = EINVAL;
2139                         break;
2140                 }
2141                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2142                     pf_krulequeue);
2143                 if (tail)
2144                         pr->nr = tail->nr + 1;
2145                 else
2146                         pr->nr = 0;
2147                 pr->ticket = ruleset->rules[rs_num].active.ticket;
2148                 PF_RULES_WUNLOCK();
2149                 break;
2150         }
2151
2152         case DIOCGETRULE: {
2153                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
2154                 struct pf_kruleset      *ruleset;
2155                 struct pf_krule         *rule;
2156                 int                      rs_num;
2157
2158                 PF_RULES_WLOCK();
2159                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
2160                 ruleset = pf_find_kruleset(pr->anchor);
2161                 if (ruleset == NULL) {
2162                         PF_RULES_WUNLOCK();
2163                         error = EINVAL;
2164                         break;
2165                 }
2166                 rs_num = pf_get_ruleset_number(pr->rule.action);
2167                 if (rs_num >= PF_RULESET_MAX) {
2168                         PF_RULES_WUNLOCK();
2169                         error = EINVAL;
2170                         break;
2171                 }
2172                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2173                         PF_RULES_WUNLOCK();
2174                         error = EBUSY;
2175                         break;
2176                 }
2177                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2178                 while ((rule != NULL) && (rule->nr != pr->nr))
2179                         rule = TAILQ_NEXT(rule, entries);
2180                 if (rule == NULL) {
2181                         PF_RULES_WUNLOCK();
2182                         error = EBUSY;
2183                         break;
2184                 }
2185
2186                 pf_krule_to_rule(rule, &pr->rule);
2187
2188                 if (pf_kanchor_copyout(ruleset, rule, pr)) {
2189                         PF_RULES_WUNLOCK();
2190                         error = EBUSY;
2191                         break;
2192                 }
2193                 pf_addr_copyout(&pr->rule.src.addr);
2194                 pf_addr_copyout(&pr->rule.dst.addr);
2195
2196                 if (pr->action == PF_GET_CLR_CNTR) {
2197                         counter_u64_zero(rule->evaluations);
2198                         for (int i = 0; i < 2; i++) {
2199                                 counter_u64_zero(rule->packets[i]);
2200                                 counter_u64_zero(rule->bytes[i]);
2201                         }
2202                         counter_u64_zero(rule->states_tot);
2203                 }
2204                 PF_RULES_WUNLOCK();
2205                 break;
2206         }
2207
2208         case DIOCCHANGERULE: {
2209                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
2210                 struct pf_kruleset      *ruleset;
2211                 struct pf_krule         *oldrule = NULL, *newrule = NULL;
2212                 struct pfi_kkif         *kif = NULL;
2213                 struct pf_kpooladdr     *pa;
2214                 u_int32_t                nr = 0;
2215                 int                      rs_num;
2216
2217                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
2218                     pcr->action > PF_CHANGE_GET_TICKET) {
2219                         error = EINVAL;
2220                         break;
2221                 }
2222                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2223                         error = EINVAL;
2224                         break;
2225                 }
2226
2227                 if (pcr->action != PF_CHANGE_REMOVE) {
2228                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2229                         error = pf_rule_to_krule(&pcr->rule, newrule);
2230                         if (error != 0) {
2231                                 free(newrule, M_PFRULE);
2232                                 break;
2233                         }
2234
2235                         if (newrule->ifname[0])
2236                                 kif = pf_kkif_create(M_WAITOK);
2237                         newrule->evaluations = counter_u64_alloc(M_WAITOK);
2238                         for (int i = 0; i < 2; i++) {
2239                                 newrule->packets[i] =
2240                                     counter_u64_alloc(M_WAITOK);
2241                                 newrule->bytes[i] =
2242                                     counter_u64_alloc(M_WAITOK);
2243                         }
2244                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
2245                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
2246                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2247                         newrule->cuid = td->td_ucred->cr_ruid;
2248                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2249                         TAILQ_INIT(&newrule->rpool.list);
2250                 }
2251
2252 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
2253
2254                 PF_RULES_WLOCK();
2255                 if (!(pcr->action == PF_CHANGE_REMOVE ||
2256                     pcr->action == PF_CHANGE_GET_TICKET) &&
2257                     pcr->pool_ticket != V_ticket_pabuf)
2258                         ERROUT(EBUSY);
2259
2260                 ruleset = pf_find_kruleset(pcr->anchor);
2261                 if (ruleset == NULL)
2262                         ERROUT(EINVAL);
2263
2264                 rs_num = pf_get_ruleset_number(pcr->rule.action);
2265                 if (rs_num >= PF_RULESET_MAX)
2266                         ERROUT(EINVAL);
2267
2268                 if (pcr->action == PF_CHANGE_GET_TICKET) {
2269                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2270                         ERROUT(0);
2271                 } else if (pcr->ticket !=
2272                             ruleset->rules[rs_num].active.ticket)
2273                                 ERROUT(EINVAL);
2274
2275                 if (pcr->action != PF_CHANGE_REMOVE) {
2276                         if (newrule->ifname[0]) {
2277                                 newrule->kif = pfi_kkif_attach(kif,
2278                                     newrule->ifname);
2279                                 pfi_kkif_ref(newrule->kif);
2280                         } else
2281                                 newrule->kif = NULL;
2282
2283                         if (newrule->rtableid > 0 &&
2284                             newrule->rtableid >= rt_numfibs)
2285                                 error = EBUSY;
2286
2287 #ifdef ALTQ
2288                         /* set queue IDs */
2289                         if (newrule->qname[0] != 0) {
2290                                 if ((newrule->qid =
2291                                     pf_qname2qid(newrule->qname)) == 0)
2292                                         error = EBUSY;
2293                                 else if (newrule->pqname[0] != 0) {
2294                                         if ((newrule->pqid =
2295                                             pf_qname2qid(newrule->pqname)) == 0)
2296                                                 error = EBUSY;
2297                                 } else
2298                                         newrule->pqid = newrule->qid;
2299                         }
2300 #endif /* ALTQ */
2301                         if (newrule->tagname[0])
2302                                 if ((newrule->tag =
2303                                     pf_tagname2tag(newrule->tagname)) == 0)
2304                                         error = EBUSY;
2305                         if (newrule->match_tagname[0])
2306                                 if ((newrule->match_tag = pf_tagname2tag(
2307                                     newrule->match_tagname)) == 0)
2308                                         error = EBUSY;
2309                         if (newrule->rt && !newrule->direction)
2310                                 error = EINVAL;
2311                         if (!newrule->log)
2312                                 newrule->logif = 0;
2313                         if (newrule->logif >= PFLOGIFS_MAX)
2314                                 error = EINVAL;
2315                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2316                                 error = ENOMEM;
2317                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2318                                 error = ENOMEM;
2319                         if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2320                                 error = EINVAL;
2321                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2322                                 if (pa->addr.type == PF_ADDR_TABLE) {
2323                                         pa->addr.p.tbl =
2324                                             pfr_attach_table(ruleset,
2325                                             pa->addr.v.tblname);
2326                                         if (pa->addr.p.tbl == NULL)
2327                                                 error = ENOMEM;
2328                                 }
2329
2330                         newrule->overload_tbl = NULL;
2331                         if (newrule->overload_tblname[0]) {
2332                                 if ((newrule->overload_tbl = pfr_attach_table(
2333                                     ruleset, newrule->overload_tblname)) ==
2334                                     NULL)
2335                                         error = EINVAL;
2336                                 else
2337                                         newrule->overload_tbl->pfrkt_flags |=
2338                                             PFR_TFLAG_ACTIVE;
2339                         }
2340
2341                         pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2342                         if (((((newrule->action == PF_NAT) ||
2343                             (newrule->action == PF_RDR) ||
2344                             (newrule->action == PF_BINAT) ||
2345                             (newrule->rt > PF_NOPFROUTE)) &&
2346                             !newrule->anchor)) &&
2347                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2348                                 error = EINVAL;
2349
2350                         if (error) {
2351                                 pf_free_rule(newrule);
2352                                 PF_RULES_WUNLOCK();
2353                                 break;
2354                         }
2355
2356                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2357                 }
2358                 pf_empty_kpool(&V_pf_pabuf);
2359
2360                 if (pcr->action == PF_CHANGE_ADD_HEAD)
2361                         oldrule = TAILQ_FIRST(
2362                             ruleset->rules[rs_num].active.ptr);
2363                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
2364                         oldrule = TAILQ_LAST(
2365                             ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2366                 else {
2367                         oldrule = TAILQ_FIRST(
2368                             ruleset->rules[rs_num].active.ptr);
2369                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2370                                 oldrule = TAILQ_NEXT(oldrule, entries);
2371                         if (oldrule == NULL) {
2372                                 if (newrule != NULL)
2373                                         pf_free_rule(newrule);
2374                                 PF_RULES_WUNLOCK();
2375                                 error = EINVAL;
2376                                 break;
2377                         }
2378                 }
2379
2380                 if (pcr->action == PF_CHANGE_REMOVE) {
2381                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2382                             oldrule);
2383                         ruleset->rules[rs_num].active.rcount--;
2384                 } else {
2385                         if (oldrule == NULL)
2386                                 TAILQ_INSERT_TAIL(
2387                                     ruleset->rules[rs_num].active.ptr,
2388                                     newrule, entries);
2389                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2390                             pcr->action == PF_CHANGE_ADD_BEFORE)
2391                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2392                         else
2393                                 TAILQ_INSERT_AFTER(
2394                                     ruleset->rules[rs_num].active.ptr,
2395                                     oldrule, newrule, entries);
2396                         ruleset->rules[rs_num].active.rcount++;
2397                 }
2398
2399                 nr = 0;
2400                 TAILQ_FOREACH(oldrule,
2401                     ruleset->rules[rs_num].active.ptr, entries)
2402                         oldrule->nr = nr++;
2403
2404                 ruleset->rules[rs_num].active.ticket++;
2405
2406                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2407                 pf_remove_if_empty_kruleset(ruleset);
2408
2409                 PF_RULES_WUNLOCK();
2410                 break;
2411
2412 #undef ERROUT
2413 DIOCCHANGERULE_error:
2414                 PF_RULES_WUNLOCK();
2415                 pf_krule_free(newrule);
2416                 pf_kkif_free(kif);
2417                 break;
2418         }
2419
2420         case DIOCCLRSTATES: {
2421                 struct pf_state         *s;
2422                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2423                 u_int                    i, killed = 0;
2424
2425                 for (i = 0; i <= pf_hashmask; i++) {
2426                         struct pf_idhash *ih = &V_pf_idhash[i];
2427
2428 relock_DIOCCLRSTATES:
2429                         PF_HASHROW_LOCK(ih);
2430                         LIST_FOREACH(s, &ih->states, entry)
2431                                 if (!psk->psk_ifname[0] ||
2432                                     !strcmp(psk->psk_ifname,
2433                                     s->kif->pfik_name)) {
2434                                         /*
2435                                          * Don't send out individual
2436                                          * delete messages.
2437                                          */
2438                                         s->state_flags |= PFSTATE_NOSYNC;
2439                                         pf_unlink_state(s, PF_ENTER_LOCKED);
2440                                         killed++;
2441                                         goto relock_DIOCCLRSTATES;
2442                                 }
2443                         PF_HASHROW_UNLOCK(ih);
2444                 }
2445                 psk->psk_killed = killed;
2446                 if (V_pfsync_clear_states_ptr != NULL)
2447                         V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2448                 break;
2449         }
2450
2451         case DIOCKILLSTATES: {
2452                 struct pf_state         *s;
2453                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2454                 u_int                    i, killed = 0;
2455
2456                 if (psk->psk_pfcmp.id) {
2457                         if (psk->psk_pfcmp.creatorid == 0)
2458                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2459                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2460                             psk->psk_pfcmp.creatorid))) {
2461                                 pf_unlink_state(s, PF_ENTER_LOCKED);
2462                                 psk->psk_killed = 1;
2463                         }
2464                         break;
2465                 }
2466
2467                 for (i = 0; i <= pf_hashmask; i++)
2468                         killed += pf_killstates_row(psk, &V_pf_idhash[i]);
2469
2470                 psk->psk_killed = killed;
2471                 break;
2472         }
2473
2474         case DIOCADDSTATE: {
2475                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
2476                 struct pfsync_state     *sp = &ps->state;
2477
2478                 if (sp->timeout >= PFTM_MAX) {
2479                         error = EINVAL;
2480                         break;
2481                 }
2482                 if (V_pfsync_state_import_ptr != NULL) {
2483                         PF_RULES_RLOCK();
2484                         error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2485                         PF_RULES_RUNLOCK();
2486                 } else
2487                         error = EOPNOTSUPP;
2488                 break;
2489         }
2490
2491         case DIOCGETSTATE: {
2492                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
2493                 struct pf_state         *s;
2494
2495                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2496                 if (s == NULL) {
2497                         error = ENOENT;
2498                         break;
2499                 }
2500
2501                 pfsync_state_export(&ps->state, s);
2502                 PF_STATE_UNLOCK(s);
2503                 break;
2504         }
2505
2506         case DIOCGETSTATES: {
2507                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
2508                 struct pf_state         *s;
2509                 struct pfsync_state     *pstore, *p;
2510                 int i, nr;
2511
2512                 if (ps->ps_len <= 0) {
2513                         nr = uma_zone_get_cur(V_pf_state_z);
2514                         ps->ps_len = sizeof(struct pfsync_state) * nr;
2515                         break;
2516                 }
2517
2518                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2519                 nr = 0;
2520
2521                 for (i = 0; i <= pf_hashmask; i++) {
2522                         struct pf_idhash *ih = &V_pf_idhash[i];
2523
2524                         PF_HASHROW_LOCK(ih);
2525                         LIST_FOREACH(s, &ih->states, entry) {
2526
2527                                 if (s->timeout == PFTM_UNLINKED)
2528                                         continue;
2529
2530                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
2531                                         PF_HASHROW_UNLOCK(ih);
2532                                         goto DIOCGETSTATES_full;
2533                                 }
2534                                 pfsync_state_export(p, s);
2535                                 p++;
2536                                 nr++;
2537                         }
2538                         PF_HASHROW_UNLOCK(ih);
2539                 }
2540 DIOCGETSTATES_full:
2541                 error = copyout(pstore, ps->ps_states,
2542                     sizeof(struct pfsync_state) * nr);
2543                 if (error) {
2544                         free(pstore, M_TEMP);
2545                         break;
2546                 }
2547                 ps->ps_len = sizeof(struct pfsync_state) * nr;
2548                 free(pstore, M_TEMP);
2549
2550                 break;
2551         }
2552
2553         case DIOCGETSTATUS: {
2554                 struct pf_status *s = (struct pf_status *)addr;
2555
2556                 PF_RULES_RLOCK();
2557                 s->running = V_pf_status.running;
2558                 s->since   = V_pf_status.since;
2559                 s->debug   = V_pf_status.debug;
2560                 s->hostid  = V_pf_status.hostid;
2561                 s->states  = V_pf_status.states;
2562                 s->src_nodes = V_pf_status.src_nodes;
2563
2564                 for (int i = 0; i < PFRES_MAX; i++)
2565                         s->counters[i] =
2566                             counter_u64_fetch(V_pf_status.counters[i]);
2567                 for (int i = 0; i < LCNT_MAX; i++)
2568                         s->lcounters[i] =
2569                             counter_u64_fetch(V_pf_status.lcounters[i]);
2570                 for (int i = 0; i < FCNT_MAX; i++)
2571                         s->fcounters[i] =
2572                             counter_u64_fetch(V_pf_status.fcounters[i]);
2573                 for (int i = 0; i < SCNT_MAX; i++)
2574                         s->scounters[i] =
2575                             counter_u64_fetch(V_pf_status.scounters[i]);
2576
2577                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2578                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2579                     PF_MD5_DIGEST_LENGTH);
2580
2581                 pfi_update_status(s->ifname, s);
2582                 PF_RULES_RUNLOCK();
2583                 break;
2584         }
2585
2586         case DIOCSETSTATUSIF: {
2587                 struct pfioc_if *pi = (struct pfioc_if *)addr;
2588
2589                 if (pi->ifname[0] == 0) {
2590                         bzero(V_pf_status.ifname, IFNAMSIZ);
2591                         break;
2592                 }
2593                 PF_RULES_WLOCK();
2594                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2595                 PF_RULES_WUNLOCK();
2596                 break;
2597         }
2598
2599         case DIOCCLRSTATUS: {
2600                 PF_RULES_WLOCK();
2601                 for (int i = 0; i < PFRES_MAX; i++)
2602                         counter_u64_zero(V_pf_status.counters[i]);
2603                 for (int i = 0; i < FCNT_MAX; i++)
2604                         counter_u64_zero(V_pf_status.fcounters[i]);
2605                 for (int i = 0; i < SCNT_MAX; i++)
2606                         counter_u64_zero(V_pf_status.scounters[i]);
2607                 for (int i = 0; i < LCNT_MAX; i++)
2608                         counter_u64_zero(V_pf_status.lcounters[i]);
2609                 V_pf_status.since = time_second;
2610                 if (*V_pf_status.ifname)
2611                         pfi_update_status(V_pf_status.ifname, NULL);
2612                 PF_RULES_WUNLOCK();
2613                 break;
2614         }
2615
2616         case DIOCNATLOOK: {
2617                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
2618                 struct pf_state_key     *sk;
2619                 struct pf_state         *state;
2620                 struct pf_state_key_cmp  key;
2621                 int                      m = 0, direction = pnl->direction;
2622                 int                      sidx, didx;
2623
2624                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
2625                 sidx = (direction == PF_IN) ? 1 : 0;
2626                 didx = (direction == PF_IN) ? 0 : 1;
2627
2628                 if (!pnl->proto ||
2629                     PF_AZERO(&pnl->saddr, pnl->af) ||
2630                     PF_AZERO(&pnl->daddr, pnl->af) ||
2631                     ((pnl->proto == IPPROTO_TCP ||
2632                     pnl->proto == IPPROTO_UDP) &&
2633                     (!pnl->dport || !pnl->sport)))
2634                         error = EINVAL;
2635                 else {
2636                         bzero(&key, sizeof(key));
2637                         key.af = pnl->af;
2638                         key.proto = pnl->proto;
2639                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2640                         key.port[sidx] = pnl->sport;
2641                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2642                         key.port[didx] = pnl->dport;
2643
2644                         state = pf_find_state_all(&key, direction, &m);
2645
2646                         if (m > 1)
2647                                 error = E2BIG;  /* more than one state */
2648                         else if (state != NULL) {
2649                                 /* XXXGL: not locked read */
2650                                 sk = state->key[sidx];
2651                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2652                                 pnl->rsport = sk->port[sidx];
2653                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2654                                 pnl->rdport = sk->port[didx];
2655                         } else
2656                                 error = ENOENT;
2657                 }
2658                 break;
2659         }
2660
2661         case DIOCSETTIMEOUT: {
2662                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
2663                 int              old;
2664
2665                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2666                     pt->seconds < 0) {
2667                         error = EINVAL;
2668                         break;
2669                 }
2670                 PF_RULES_WLOCK();
2671                 old = V_pf_default_rule.timeout[pt->timeout];
2672                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2673                         pt->seconds = 1;
2674                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2675                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2676                         wakeup(pf_purge_thread);
2677                 pt->seconds = old;
2678                 PF_RULES_WUNLOCK();
2679                 break;
2680         }
2681
2682         case DIOCGETTIMEOUT: {
2683                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
2684
2685                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2686                         error = EINVAL;
2687                         break;
2688                 }
2689                 PF_RULES_RLOCK();
2690                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2691                 PF_RULES_RUNLOCK();
2692                 break;
2693         }
2694
2695         case DIOCGETLIMIT: {
2696                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
2697
2698                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2699                         error = EINVAL;
2700                         break;
2701                 }
2702                 PF_RULES_RLOCK();
2703                 pl->limit = V_pf_limits[pl->index].limit;
2704                 PF_RULES_RUNLOCK();
2705                 break;
2706         }
2707
2708         case DIOCSETLIMIT: {
2709                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
2710                 int                      old_limit;
2711
2712                 PF_RULES_WLOCK();
2713                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2714                     V_pf_limits[pl->index].zone == NULL) {
2715                         PF_RULES_WUNLOCK();
2716                         error = EINVAL;
2717                         break;
2718                 }
2719                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2720                 old_limit = V_pf_limits[pl->index].limit;
2721                 V_pf_limits[pl->index].limit = pl->limit;
2722                 pl->limit = old_limit;
2723                 PF_RULES_WUNLOCK();
2724                 break;
2725         }
2726
2727         case DIOCSETDEBUG: {
2728                 u_int32_t       *level = (u_int32_t *)addr;
2729
2730                 PF_RULES_WLOCK();
2731                 V_pf_status.debug = *level;
2732                 PF_RULES_WUNLOCK();
2733                 break;
2734         }
2735
2736         case DIOCCLRRULECTRS: {
2737                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2738                 struct pf_kruleset      *ruleset = &pf_main_ruleset;
2739                 struct pf_krule         *rule;
2740
2741                 PF_RULES_WLOCK();
2742                 TAILQ_FOREACH(rule,
2743                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2744                         counter_u64_zero(rule->evaluations);
2745                         for (int i = 0; i < 2; i++) {
2746                                 counter_u64_zero(rule->packets[i]);
2747                                 counter_u64_zero(rule->bytes[i]);
2748                         }
2749                 }
2750                 PF_RULES_WUNLOCK();
2751                 break;
2752         }
2753
2754         case DIOCGIFSPEEDV0:
2755         case DIOCGIFSPEEDV1: {
2756                 struct pf_ifspeed_v1    *psp = (struct pf_ifspeed_v1 *)addr;
2757                 struct pf_ifspeed_v1    ps;
2758                 struct ifnet            *ifp;
2759
2760                 if (psp->ifname[0] != 0) {
2761                         /* Can we completely trust user-land? */
2762                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2763                         ifp = ifunit(ps.ifname);
2764                         if (ifp != NULL) {
2765                                 psp->baudrate32 =
2766                                     (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2767                                 if (cmd == DIOCGIFSPEEDV1)
2768                                         psp->baudrate = ifp->if_baudrate;
2769                         } else
2770                                 error = EINVAL;
2771                 } else
2772                         error = EINVAL;
2773                 break;
2774         }
2775
2776 #ifdef ALTQ
2777         case DIOCSTARTALTQ: {
2778                 struct pf_altq          *altq;
2779
2780                 PF_RULES_WLOCK();
2781                 /* enable all altq interfaces on active list */
2782                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2783                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2784                                 error = pf_enable_altq(altq);
2785                                 if (error != 0)
2786                                         break;
2787                         }
2788                 }
2789                 if (error == 0)
2790                         V_pf_altq_running = 1;
2791                 PF_RULES_WUNLOCK();
2792                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2793                 break;
2794         }
2795
2796         case DIOCSTOPALTQ: {
2797                 struct pf_altq          *altq;
2798
2799                 PF_RULES_WLOCK();
2800                 /* disable all altq interfaces on active list */
2801                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2802                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2803                                 error = pf_disable_altq(altq);
2804                                 if (error != 0)
2805                                         break;
2806                         }
2807                 }
2808                 if (error == 0)
2809                         V_pf_altq_running = 0;
2810                 PF_RULES_WUNLOCK();
2811                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2812                 break;
2813         }
2814
2815         case DIOCADDALTQV0:
2816         case DIOCADDALTQV1: {
2817                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
2818                 struct pf_altq          *altq, *a;
2819                 struct ifnet            *ifp;
2820
2821                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2822                 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2823                 if (error)
2824                         break;
2825                 altq->local_flags = 0;
2826
2827                 PF_RULES_WLOCK();
2828                 if (pa->ticket != V_ticket_altqs_inactive) {
2829                         PF_RULES_WUNLOCK();
2830                         free(altq, M_PFALTQ);
2831                         error = EBUSY;
2832                         break;
2833                 }
2834
2835                 /*
2836                  * if this is for a queue, find the discipline and
2837                  * copy the necessary fields
2838                  */
2839                 if (altq->qname[0] != 0) {
2840                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2841                                 PF_RULES_WUNLOCK();
2842                                 error = EBUSY;
2843                                 free(altq, M_PFALTQ);
2844                                 break;
2845                         }
2846                         altq->altq_disc = NULL;
2847                         TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2848                                 if (strncmp(a->ifname, altq->ifname,
2849                                     IFNAMSIZ) == 0) {
2850                                         altq->altq_disc = a->altq_disc;
2851                                         break;
2852                                 }
2853                         }
2854                 }
2855
2856                 if ((ifp = ifunit(altq->ifname)) == NULL)
2857                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2858                 else
2859                         error = altq_add(ifp, altq);
2860
2861                 if (error) {
2862                         PF_RULES_WUNLOCK();
2863                         free(altq, M_PFALTQ);
2864                         break;
2865                 }
2866
2867                 if (altq->qname[0] != 0)
2868                         TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2869                 else
2870                         TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2871                 /* version error check done on import above */
2872                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2873                 PF_RULES_WUNLOCK();
2874                 break;
2875         }
2876
2877         case DIOCGETALTQSV0:
2878         case DIOCGETALTQSV1: {
2879                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
2880                 struct pf_altq          *altq;
2881
2882                 PF_RULES_RLOCK();
2883                 pa->nr = 0;
2884                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2885                         pa->nr++;
2886                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2887                         pa->nr++;
2888                 pa->ticket = V_ticket_altqs_active;
2889                 PF_RULES_RUNLOCK();
2890                 break;
2891         }
2892
2893         case DIOCGETALTQV0:
2894         case DIOCGETALTQV1: {
2895                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
2896                 struct pf_altq          *altq;
2897
2898                 PF_RULES_RLOCK();
2899                 if (pa->ticket != V_ticket_altqs_active) {
2900                         PF_RULES_RUNLOCK();
2901                         error = EBUSY;
2902                         break;
2903                 }
2904                 altq = pf_altq_get_nth_active(pa->nr);
2905                 if (altq == NULL) {
2906                         PF_RULES_RUNLOCK();
2907                         error = EBUSY;
2908                         break;
2909                 }
2910                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2911                 PF_RULES_RUNLOCK();
2912                 break;
2913         }
2914
2915         case DIOCCHANGEALTQV0:
2916         case DIOCCHANGEALTQV1:
2917                 /* CHANGEALTQ not supported yet! */
2918                 error = ENODEV;
2919                 break;
2920
2921         case DIOCGETQSTATSV0:
2922         case DIOCGETQSTATSV1: {
2923                 struct pfioc_qstats_v1  *pq = (struct pfioc_qstats_v1 *)addr;
2924                 struct pf_altq          *altq;
2925                 int                      nbytes;
2926                 u_int32_t                version;
2927
2928                 PF_RULES_RLOCK();
2929                 if (pq->ticket != V_ticket_altqs_active) {
2930                         PF_RULES_RUNLOCK();
2931                         error = EBUSY;
2932                         break;
2933                 }
2934                 nbytes = pq->nbytes;
2935                 altq = pf_altq_get_nth_active(pq->nr);
2936                 if (altq == NULL) {
2937                         PF_RULES_RUNLOCK();
2938                         error = EBUSY;
2939                         break;
2940                 }
2941
2942                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2943                         PF_RULES_RUNLOCK();
2944                         error = ENXIO;
2945                         break;
2946                 }
2947                 PF_RULES_RUNLOCK();
2948                 if (cmd == DIOCGETQSTATSV0)
2949                         version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2950                 else
2951                         version = pq->version;
2952                 error = altq_getqstats(altq, pq->buf, &nbytes, version);
2953                 if (error == 0) {
2954                         pq->scheduler = altq->scheduler;
2955                         pq->nbytes = nbytes;
2956                 }
2957                 break;
2958         }
2959 #endif /* ALTQ */
2960
2961         case DIOCBEGINADDRS: {
2962                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2963
2964                 PF_RULES_WLOCK();
2965                 pf_empty_kpool(&V_pf_pabuf);
2966                 pp->ticket = ++V_ticket_pabuf;
2967                 PF_RULES_WUNLOCK();
2968                 break;
2969         }
2970
2971         case DIOCADDADDR: {
2972                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2973                 struct pf_kpooladdr     *pa;
2974                 struct pfi_kkif         *kif = NULL;
2975
2976 #ifndef INET
2977                 if (pp->af == AF_INET) {
2978                         error = EAFNOSUPPORT;
2979                         break;
2980                 }
2981 #endif /* INET */
2982 #ifndef INET6
2983                 if (pp->af == AF_INET6) {
2984                         error = EAFNOSUPPORT;
2985                         break;
2986                 }
2987 #endif /* INET6 */
2988                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2989                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2990                     pp->addr.addr.type != PF_ADDR_TABLE) {
2991                         error = EINVAL;
2992                         break;
2993                 }
2994                 if (pp->addr.addr.p.dyn != NULL) {
2995                         error = EINVAL;
2996                         break;
2997                 }
2998                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2999                 pf_pooladdr_to_kpooladdr(&pp->addr, pa);
3000                 if (pa->ifname[0])
3001                         kif = pf_kkif_create(M_WAITOK);
3002                 PF_RULES_WLOCK();
3003                 if (pp->ticket != V_ticket_pabuf) {
3004                         PF_RULES_WUNLOCK();
3005                         if (pa->ifname[0])
3006                                 pf_kkif_free(kif);
3007                         free(pa, M_PFRULE);
3008                         error = EBUSY;
3009                         break;
3010                 }
3011                 if (pa->ifname[0]) {
3012                         pa->kif = pfi_kkif_attach(kif, pa->ifname);
3013                         pfi_kkif_ref(pa->kif);
3014                 } else
3015                         pa->kif = NULL;
3016                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
3017                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
3018                         if (pa->ifname[0])
3019                                 pfi_kkif_unref(pa->kif);
3020                         PF_RULES_WUNLOCK();
3021                         free(pa, M_PFRULE);
3022                         break;
3023                 }
3024                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
3025                 PF_RULES_WUNLOCK();
3026                 break;
3027         }
3028
3029         case DIOCGETADDRS: {
3030                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
3031                 struct pf_kpool         *pool;
3032                 struct pf_kpooladdr     *pa;
3033
3034                 PF_RULES_RLOCK();
3035                 pp->nr = 0;
3036                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3037                     pp->r_num, 0, 1, 0);
3038                 if (pool == NULL) {
3039                         PF_RULES_RUNLOCK();
3040                         error = EBUSY;
3041                         break;
3042                 }
3043                 TAILQ_FOREACH(pa, &pool->list, entries)
3044                         pp->nr++;
3045                 PF_RULES_RUNLOCK();
3046                 break;
3047         }
3048
3049         case DIOCGETADDR: {
3050                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
3051                 struct pf_kpool         *pool;
3052                 struct pf_kpooladdr     *pa;
3053                 u_int32_t                nr = 0;
3054
3055                 PF_RULES_RLOCK();
3056                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3057                     pp->r_num, 0, 1, 1);
3058                 if (pool == NULL) {
3059                         PF_RULES_RUNLOCK();
3060                         error = EBUSY;
3061                         break;
3062                 }
3063                 pa = TAILQ_FIRST(&pool->list);
3064                 while ((pa != NULL) && (nr < pp->nr)) {
3065                         pa = TAILQ_NEXT(pa, entries);
3066                         nr++;
3067                 }
3068                 if (pa == NULL) {
3069                         PF_RULES_RUNLOCK();
3070                         error = EBUSY;
3071                         break;
3072                 }
3073                 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3074                 pf_addr_copyout(&pp->addr.addr);
3075                 PF_RULES_RUNLOCK();
3076                 break;
3077         }
3078
3079         case DIOCCHANGEADDR: {
3080                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
3081                 struct pf_kpool         *pool;
3082                 struct pf_kpooladdr     *oldpa = NULL, *newpa = NULL;
3083                 struct pf_kruleset      *ruleset;
3084                 struct pfi_kkif         *kif = NULL;
3085
3086                 if (pca->action < PF_CHANGE_ADD_HEAD ||
3087                     pca->action > PF_CHANGE_REMOVE) {
3088                         error = EINVAL;
3089                         break;
3090                 }
3091                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3092                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3093                     pca->addr.addr.type != PF_ADDR_TABLE) {
3094                         error = EINVAL;
3095                         break;
3096                 }
3097                 if (pca->addr.addr.p.dyn != NULL) {
3098                         error = EINVAL;
3099                         break;
3100                 }
3101
3102                 if (pca->action != PF_CHANGE_REMOVE) {
3103 #ifndef INET
3104                         if (pca->af == AF_INET) {
3105                                 error = EAFNOSUPPORT;
3106                                 break;
3107                         }
3108 #endif /* INET */
3109 #ifndef INET6
3110                         if (pca->af == AF_INET6) {
3111                                 error = EAFNOSUPPORT;
3112                                 break;
3113                         }
3114 #endif /* INET6 */
3115                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3116                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3117                         if (newpa->ifname[0])
3118                                 kif = pf_kkif_create(M_WAITOK);
3119                         newpa->kif = NULL;
3120                 }
3121
3122 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
3123                 PF_RULES_WLOCK();
3124                 ruleset = pf_find_kruleset(pca->anchor);
3125                 if (ruleset == NULL)
3126                         ERROUT(EBUSY);
3127
3128                 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3129                     pca->r_num, pca->r_last, 1, 1);
3130                 if (pool == NULL)
3131                         ERROUT(EBUSY);
3132
3133                 if (pca->action != PF_CHANGE_REMOVE) {
3134                         if (newpa->ifname[0]) {
3135                                 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3136                                 pfi_kkif_ref(newpa->kif);
3137                                 kif = NULL;
3138                         }
3139
3140                         switch (newpa->addr.type) {
3141                         case PF_ADDR_DYNIFTL:
3142                                 error = pfi_dynaddr_setup(&newpa->addr,
3143                                     pca->af);
3144                                 break;
3145                         case PF_ADDR_TABLE:
3146                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
3147                                     newpa->addr.v.tblname);
3148                                 if (newpa->addr.p.tbl == NULL)
3149                                         error = ENOMEM;
3150                                 break;
3151                         }
3152                         if (error)
3153                                 goto DIOCCHANGEADDR_error;
3154                 }
3155
3156                 switch (pca->action) {
3157                 case PF_CHANGE_ADD_HEAD:
3158                         oldpa = TAILQ_FIRST(&pool->list);
3159                         break;
3160                 case PF_CHANGE_ADD_TAIL:
3161                         oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3162                         break;
3163                 default:
3164                         oldpa = TAILQ_FIRST(&pool->list);
3165                         for (int i = 0; oldpa && i < pca->nr; i++)
3166                                 oldpa = TAILQ_NEXT(oldpa, entries);
3167
3168                         if (oldpa == NULL)
3169                                 ERROUT(EINVAL);
3170                 }
3171
3172                 if (pca->action == PF_CHANGE_REMOVE) {
3173                         TAILQ_REMOVE(&pool->list, oldpa, entries);
3174                         switch (oldpa->addr.type) {
3175                         case PF_ADDR_DYNIFTL:
3176                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
3177                                 break;
3178                         case PF_ADDR_TABLE:
3179                                 pfr_detach_table(oldpa->addr.p.tbl);
3180                                 break;
3181                         }
3182                         if (oldpa->kif)
3183                                 pfi_kkif_unref(oldpa->kif);
3184                         free(oldpa, M_PFRULE);
3185                 } else {
3186                         if (oldpa == NULL)
3187                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3188                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
3189                             pca->action == PF_CHANGE_ADD_BEFORE)
3190                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3191                         else
3192                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
3193                                     newpa, entries);
3194                 }
3195
3196                 pool->cur = TAILQ_FIRST(&pool->list);
3197                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3198                 PF_RULES_WUNLOCK();
3199                 break;
3200
3201 #undef ERROUT
3202 DIOCCHANGEADDR_error:
3203                 if (newpa != NULL) {
3204                         if (newpa->kif)
3205                                 pfi_kkif_unref(newpa->kif);
3206                         free(newpa, M_PFRULE);
3207                 }
3208                 PF_RULES_WUNLOCK();
3209                 pf_kkif_free(kif);
3210                 break;
3211         }
3212
3213         case DIOCGETRULESETS: {
3214                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
3215                 struct pf_kruleset      *ruleset;
3216                 struct pf_kanchor       *anchor;
3217
3218                 PF_RULES_RLOCK();
3219                 pr->path[sizeof(pr->path) - 1] = 0;
3220                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3221                         PF_RULES_RUNLOCK();
3222                         error = ENOENT;
3223                         break;
3224                 }
3225                 pr->nr = 0;
3226                 if (ruleset->anchor == NULL) {
3227                         /* XXX kludge for pf_main_ruleset */
3228                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3229                                 if (anchor->parent == NULL)
3230                                         pr->nr++;
3231                 } else {
3232                         RB_FOREACH(anchor, pf_kanchor_node,
3233                             &ruleset->anchor->children)
3234                                 pr->nr++;
3235                 }
3236                 PF_RULES_RUNLOCK();
3237                 break;
3238         }
3239
3240         case DIOCGETRULESET: {
3241                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
3242                 struct pf_kruleset      *ruleset;
3243                 struct pf_kanchor       *anchor;
3244                 u_int32_t                nr = 0;
3245
3246                 PF_RULES_RLOCK();
3247                 pr->path[sizeof(pr->path) - 1] = 0;
3248                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3249                         PF_RULES_RUNLOCK();
3250                         error = ENOENT;
3251                         break;
3252                 }
3253                 pr->name[0] = 0;
3254                 if (ruleset->anchor == NULL) {
3255                         /* XXX kludge for pf_main_ruleset */
3256                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3257                                 if (anchor->parent == NULL && nr++ == pr->nr) {
3258                                         strlcpy(pr->name, anchor->name,
3259                                             sizeof(pr->name));
3260                                         break;
3261                                 }
3262                 } else {
3263                         RB_FOREACH(anchor, pf_kanchor_node,
3264                             &ruleset->anchor->children)
3265                                 if (nr++ == pr->nr) {
3266                                         strlcpy(pr->name, anchor->name,
3267                                             sizeof(pr->name));
3268                                         break;
3269                                 }
3270                 }
3271                 if (!pr->name[0])
3272                         error = EBUSY;
3273                 PF_RULES_RUNLOCK();
3274                 break;
3275         }
3276
3277         case DIOCRCLRTABLES: {
3278                 struct pfioc_table *io = (struct pfioc_table *)addr;
3279
3280                 if (io->pfrio_esize != 0) {
3281                         error = ENODEV;
3282                         break;
3283                 }
3284                 PF_RULES_WLOCK();
3285                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3286                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
3287                 PF_RULES_WUNLOCK();
3288                 break;
3289         }
3290
3291         case DIOCRADDTABLES: {
3292                 struct pfioc_table *io = (struct pfioc_table *)addr;
3293                 struct pfr_table *pfrts;
3294                 size_t totlen;
3295
3296                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3297                         error = ENODEV;
3298                         break;
3299                 }
3300
3301                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3302                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3303                         error = ENOMEM;
3304                         break;
3305                 }
3306
3307                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3308                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3309                     M_TEMP, M_WAITOK);
3310                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3311                 if (error) {
3312                         free(pfrts, M_TEMP);
3313                         break;
3314                 }
3315                 PF_RULES_WLOCK();
3316                 error = pfr_add_tables(pfrts, io->pfrio_size,
3317                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3318                 PF_RULES_WUNLOCK();
3319                 free(pfrts, M_TEMP);
3320                 break;
3321         }
3322
3323         case DIOCRDELTABLES: {
3324                 struct pfioc_table *io = (struct pfioc_table *)addr;
3325                 struct pfr_table *pfrts;
3326                 size_t totlen;
3327
3328                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3329                         error = ENODEV;
3330                         break;
3331                 }
3332
3333                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3334                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3335                         error = ENOMEM;
3336                         break;
3337                 }
3338
3339                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3340                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3341                     M_TEMP, M_WAITOK);
3342                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3343                 if (error) {
3344                         free(pfrts, M_TEMP);
3345                         break;
3346                 }
3347                 PF_RULES_WLOCK();
3348                 error = pfr_del_tables(pfrts, io->pfrio_size,
3349                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3350                 PF_RULES_WUNLOCK();
3351                 free(pfrts, M_TEMP);
3352                 break;
3353         }
3354
3355         case DIOCRGETTABLES: {
3356                 struct pfioc_table *io = (struct pfioc_table *)addr;
3357                 struct pfr_table *pfrts;
3358                 size_t totlen;
3359                 int n;
3360
3361                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3362                         error = ENODEV;
3363                         break;
3364                 }
3365                 PF_RULES_RLOCK();
3366                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3367                 if (n < 0) {
3368                         PF_RULES_RUNLOCK();
3369                         error = EINVAL;
3370                         break;
3371                 }
3372                 io->pfrio_size = min(io->pfrio_size, n);
3373
3374                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3375
3376                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3377                     M_TEMP, M_NOWAIT);
3378                 if (pfrts == NULL) {
3379                         error = ENOMEM;
3380                         PF_RULES_RUNLOCK();
3381                         break;
3382                 }
3383                 error = pfr_get_tables(&io->pfrio_table, pfrts,
3384                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3385                 PF_RULES_RUNLOCK();
3386                 if (error == 0)
3387                         error = copyout(pfrts, io->pfrio_buffer, totlen);
3388                 free(pfrts, M_TEMP);
3389                 break;
3390         }
3391
3392         case DIOCRGETTSTATS: {
3393                 struct pfioc_table *io = (struct pfioc_table *)addr;
3394                 struct pfr_tstats *pfrtstats;
3395                 size_t totlen;
3396                 int n;
3397
3398                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3399                         error = ENODEV;
3400                         break;
3401                 }
3402                 PF_RULES_WLOCK();
3403                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3404                 if (n < 0) {
3405                         PF_RULES_WUNLOCK();
3406                         error = EINVAL;
3407                         break;
3408                 }
3409                 io->pfrio_size = min(io->pfrio_size, n);
3410
3411                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3412                 pfrtstats = mallocarray(io->pfrio_size,
3413                     sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3414                 if (pfrtstats == NULL) {
3415                         error = ENOMEM;
3416                         PF_RULES_WUNLOCK();
3417                         break;
3418                 }
3419                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3420                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3421                 PF_RULES_WUNLOCK();
3422                 if (error == 0)
3423                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3424                 free(pfrtstats, M_TEMP);
3425                 break;
3426         }
3427
3428         case DIOCRCLRTSTATS: {
3429                 struct pfioc_table *io = (struct pfioc_table *)addr;
3430                 struct pfr_table *pfrts;
3431                 size_t totlen;
3432
3433                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3434                         error = ENODEV;
3435                         break;
3436                 }
3437
3438                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3439                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3440                         /* We used to count tables and use the minimum required
3441                          * size, so we didn't fail on overly large requests.
3442                          * Keep doing so. */
3443                         io->pfrio_size = pf_ioctl_maxcount;
3444                         break;
3445                 }
3446
3447                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3448                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3449                     M_TEMP, M_NOWAIT);
3450                 if (pfrts == NULL) {
3451                         error = ENOMEM;
3452                         break;
3453                 }
3454                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3455                 if (error) {
3456                         free(pfrts, M_TEMP);
3457                         break;
3458                 }
3459
3460                 PF_RULES_WLOCK();
3461                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
3462                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3463                 PF_RULES_WUNLOCK();
3464                 free(pfrts, M_TEMP);
3465                 break;
3466         }
3467
3468         case DIOCRSETTFLAGS: {
3469                 struct pfioc_table *io = (struct pfioc_table *)addr;
3470                 struct pfr_table *pfrts;
3471                 size_t totlen;
3472                 int n;
3473
3474                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
3475                         error = ENODEV;
3476                         break;
3477                 }
3478
3479                 PF_RULES_RLOCK();
3480                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3481                 if (n < 0) {
3482                         PF_RULES_RUNLOCK();
3483                         error = EINVAL;
3484                         break;
3485                 }
3486
3487                 io->pfrio_size = min(io->pfrio_size, n);
3488                 PF_RULES_RUNLOCK();
3489
3490                 totlen = io->pfrio_size * sizeof(struct pfr_table);
3491                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3492                     M_TEMP, M_WAITOK);
3493                 error = copyin(io->pfrio_buffer, pfrts, totlen);
3494                 if (error) {
3495                         free(pfrts, M_TEMP);
3496                         break;
3497                 }
3498                 PF_RULES_WLOCK();
3499                 error = pfr_set_tflags(pfrts, io->pfrio_size,
3500                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3501                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3502                 PF_RULES_WUNLOCK();
3503                 free(pfrts, M_TEMP);
3504                 break;
3505         }
3506
3507         case DIOCRCLRADDRS: {
3508                 struct pfioc_table *io = (struct pfioc_table *)addr;
3509
3510                 if (io->pfrio_esize != 0) {
3511                         error = ENODEV;
3512                         break;
3513                 }
3514                 PF_RULES_WLOCK();
3515                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3516                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
3517                 PF_RULES_WUNLOCK();
3518                 break;
3519         }
3520
3521         case DIOCRADDADDRS: {
3522                 struct pfioc_table *io = (struct pfioc_table *)addr;
3523                 struct pfr_addr *pfras;
3524                 size_t totlen;
3525
3526                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3527                         error = ENODEV;
3528                         break;
3529                 }
3530                 if (io->pfrio_size < 0 ||
3531                     io->pfrio_size > pf_ioctl_maxcount ||
3532                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3533                         error = EINVAL;
3534                         break;
3535                 }
3536                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3537                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3538                     M_TEMP, M_NOWAIT);
3539                 if (! pfras) {
3540                         error = ENOMEM;
3541                         break;
3542                 }
3543                 error = copyin(io->pfrio_buffer, pfras, totlen);
3544                 if (error) {
3545                         free(pfras, M_TEMP);
3546                         break;
3547                 }
3548                 PF_RULES_WLOCK();
3549                 error = pfr_add_addrs(&io->pfrio_table, pfras,
3550                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3551                     PFR_FLAG_USERIOCTL);
3552                 PF_RULES_WUNLOCK();
3553                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3554                         error = copyout(pfras, io->pfrio_buffer, totlen);
3555                 free(pfras, M_TEMP);
3556                 break;
3557         }
3558
3559         case DIOCRDELADDRS: {
3560                 struct pfioc_table *io = (struct pfioc_table *)addr;
3561                 struct pfr_addr *pfras;
3562                 size_t totlen;
3563
3564                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3565                         error = ENODEV;
3566                         break;
3567                 }
3568                 if (io->pfrio_size < 0 ||
3569                     io->pfrio_size > pf_ioctl_maxcount ||
3570                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3571                         error = EINVAL;
3572                         break;
3573                 }
3574                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3575                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3576                     M_TEMP, M_NOWAIT);
3577                 if (! pfras) {
3578                         error = ENOMEM;
3579                         break;
3580                 }
3581                 error = copyin(io->pfrio_buffer, pfras, totlen);
3582                 if (error) {
3583                         free(pfras, M_TEMP);
3584                         break;
3585                 }
3586                 PF_RULES_WLOCK();
3587                 error = pfr_del_addrs(&io->pfrio_table, pfras,
3588                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3589                     PFR_FLAG_USERIOCTL);
3590                 PF_RULES_WUNLOCK();
3591                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3592                         error = copyout(pfras, io->pfrio_buffer, totlen);
3593                 free(pfras, M_TEMP);
3594                 break;
3595         }
3596
3597         case DIOCRSETADDRS: {
3598                 struct pfioc_table *io = (struct pfioc_table *)addr;
3599                 struct pfr_addr *pfras;
3600                 size_t totlen, count;
3601
3602                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3603                         error = ENODEV;
3604                         break;
3605                 }
3606                 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3607                         error = EINVAL;
3608                         break;
3609                 }
3610                 count = max(io->pfrio_size, io->pfrio_size2);
3611                 if (count > pf_ioctl_maxcount ||
3612                     WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3613                         error = EINVAL;
3614                         break;
3615                 }
3616                 totlen = count * sizeof(struct pfr_addr);
3617                 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3618                     M_NOWAIT);
3619                 if (! pfras) {
3620                         error = ENOMEM;
3621                         break;
3622                 }
3623                 error = copyin(io->pfrio_buffer, pfras, totlen);
3624                 if (error) {
3625                         free(pfras, M_TEMP);
3626                         break;
3627                 }
3628                 PF_RULES_WLOCK();
3629                 error = pfr_set_addrs(&io->pfrio_table, pfras,
3630                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3631                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3632                     PFR_FLAG_USERIOCTL, 0);
3633                 PF_RULES_WUNLOCK();
3634                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3635                         error = copyout(pfras, io->pfrio_buffer, totlen);
3636                 free(pfras, M_TEMP);
3637                 break;
3638         }
3639
3640         case DIOCRGETADDRS: {
3641                 struct pfioc_table *io = (struct pfioc_table *)addr;
3642                 struct pfr_addr *pfras;
3643                 size_t totlen;
3644
3645                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3646                         error = ENODEV;
3647                         break;
3648                 }
3649                 if (io->pfrio_size < 0 ||
3650                     io->pfrio_size > pf_ioctl_maxcount ||
3651                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3652                         error = EINVAL;
3653                         break;
3654                 }
3655                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3656                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3657                     M_TEMP, M_NOWAIT);
3658                 if (! pfras) {
3659                         error = ENOMEM;
3660                         break;
3661                 }
3662                 PF_RULES_RLOCK();
3663                 error = pfr_get_addrs(&io->pfrio_table, pfras,
3664                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3665                 PF_RULES_RUNLOCK();
3666                 if (error == 0)
3667                         error = copyout(pfras, io->pfrio_buffer, totlen);
3668                 free(pfras, M_TEMP);
3669                 break;
3670         }
3671
3672         case DIOCRGETASTATS: {
3673                 struct pfioc_table *io = (struct pfioc_table *)addr;
3674                 struct pfr_astats *pfrastats;
3675                 size_t totlen;
3676
3677                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3678                         error = ENODEV;
3679                         break;
3680                 }
3681                 if (io->pfrio_size < 0 ||
3682                     io->pfrio_size > pf_ioctl_maxcount ||
3683                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3684                         error = EINVAL;
3685                         break;
3686                 }
3687                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
3688                 pfrastats = mallocarray(io->pfrio_size,
3689                     sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3690                 if (! pfrastats) {
3691                         error = ENOMEM;
3692                         break;
3693                 }
3694                 PF_RULES_RLOCK();
3695                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
3696                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3697                 PF_RULES_RUNLOCK();
3698                 if (error == 0)
3699                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
3700                 free(pfrastats, M_TEMP);
3701                 break;
3702         }
3703
3704         case DIOCRCLRASTATS: {
3705                 struct pfioc_table *io = (struct pfioc_table *)addr;
3706                 struct pfr_addr *pfras;
3707                 size_t totlen;
3708
3709                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3710                         error = ENODEV;
3711                         break;
3712                 }
3713                 if (io->pfrio_size < 0 ||
3714                     io->pfrio_size > pf_ioctl_maxcount ||
3715                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3716                         error = EINVAL;
3717                         break;
3718                 }
3719                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3720                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3721                     M_TEMP, M_NOWAIT);
3722                 if (! pfras) {
3723                         error = ENOMEM;
3724                         break;
3725                 }
3726                 error = copyin(io->pfrio_buffer, pfras, totlen);
3727                 if (error) {
3728                         free(pfras, M_TEMP);
3729                         break;
3730                 }
3731                 PF_RULES_WLOCK();
3732                 error = pfr_clr_astats(&io->pfrio_table, pfras,
3733                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3734                     PFR_FLAG_USERIOCTL);
3735                 PF_RULES_WUNLOCK();
3736                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3737                         error = copyout(pfras, io->pfrio_buffer, totlen);
3738                 free(pfras, M_TEMP);
3739                 break;
3740         }
3741
3742         case DIOCRTSTADDRS: {
3743                 struct pfioc_table *io = (struct pfioc_table *)addr;
3744                 struct pfr_addr *pfras;
3745                 size_t totlen;
3746
3747                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3748                         error = ENODEV;
3749                         break;
3750                 }
3751                 if (io->pfrio_size < 0 ||
3752                     io->pfrio_size > pf_ioctl_maxcount ||
3753                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3754                         error = EINVAL;
3755                         break;
3756                 }
3757                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3758                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3759                     M_TEMP, M_NOWAIT);
3760                 if (! pfras) {
3761                         error = ENOMEM;
3762                         break;
3763                 }
3764                 error = copyin(io->pfrio_buffer, pfras, totlen);
3765                 if (error) {
3766                         free(pfras, M_TEMP);
3767                         break;
3768                 }
3769                 PF_RULES_RLOCK();
3770                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
3771                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3772                     PFR_FLAG_USERIOCTL);
3773                 PF_RULES_RUNLOCK();
3774                 if (error == 0)
3775                         error = copyout(pfras, io->pfrio_buffer, totlen);
3776                 free(pfras, M_TEMP);
3777                 break;
3778         }
3779
3780         case DIOCRINADEFINE: {
3781                 struct pfioc_table *io = (struct pfioc_table *)addr;
3782                 struct pfr_addr *pfras;
3783                 size_t totlen;
3784
3785                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3786                         error = ENODEV;
3787                         break;
3788                 }
3789                 if (io->pfrio_size < 0 ||
3790                     io->pfrio_size > pf_ioctl_maxcount ||
3791                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3792                         error = EINVAL;
3793                         break;
3794                 }
3795                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
3796                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3797                     M_TEMP, M_NOWAIT);
3798                 if (! pfras) {
3799                         error = ENOMEM;
3800                         break;
3801                 }
3802                 error = copyin(io->pfrio_buffer, pfras, totlen);
3803                 if (error) {
3804                         free(pfras, M_TEMP);
3805                         break;
3806                 }
3807                 PF_RULES_WLOCK();
3808                 error = pfr_ina_define(&io->pfrio_table, pfras,
3809                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3810                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3811                 PF_RULES_WUNLOCK();
3812                 free(pfras, M_TEMP);
3813                 break;
3814         }
3815
3816         case DIOCOSFPADD: {
3817                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3818                 PF_RULES_WLOCK();
3819                 error = pf_osfp_add(io);
3820                 PF_RULES_WUNLOCK();
3821                 break;
3822         }
3823
3824         case DIOCOSFPGET: {
3825                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3826                 PF_RULES_RLOCK();
3827                 error = pf_osfp_get(io);
3828                 PF_RULES_RUNLOCK();
3829                 break;
3830         }
3831
3832         case DIOCXBEGIN: {
3833                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3834                 struct pfioc_trans_e    *ioes, *ioe;
3835                 size_t                   totlen;
3836                 int                      i;
3837
3838                 if (io->esize != sizeof(*ioe)) {
3839                         error = ENODEV;
3840                         break;
3841                 }
3842                 if (io->size < 0 ||
3843                     io->size > pf_ioctl_maxcount ||
3844                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3845                         error = EINVAL;
3846                         break;
3847                 }
3848                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3849                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3850                     M_TEMP, M_NOWAIT);
3851                 if (! ioes) {
3852                         error = ENOMEM;
3853                         break;
3854                 }
3855                 error = copyin(io->array, ioes, totlen);
3856                 if (error) {
3857                         free(ioes, M_TEMP);
3858                         break;
3859                 }
3860                 PF_RULES_WLOCK();
3861                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3862                         switch (ioe->rs_num) {
3863 #ifdef ALTQ
3864                         case PF_RULESET_ALTQ:
3865                                 if (ioe->anchor[0]) {
3866                                         PF_RULES_WUNLOCK();
3867                                         free(ioes, M_TEMP);
3868                                         error = EINVAL;
3869                                         goto fail;
3870                                 }
3871                                 if ((error = pf_begin_altq(&ioe->ticket))) {
3872                                         PF_RULES_WUNLOCK();
3873                                         free(ioes, M_TEMP);
3874                                         goto fail;
3875                                 }
3876                                 break;
3877 #endif /* ALTQ */
3878                         case PF_RULESET_TABLE:
3879                             {
3880                                 struct pfr_table table;
3881
3882                                 bzero(&table, sizeof(table));
3883                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3884                                     sizeof(table.pfrt_anchor));
3885                                 if ((error = pfr_ina_begin(&table,
3886                                     &ioe->ticket, NULL, 0))) {
3887                                         PF_RULES_WUNLOCK();
3888                                         free(ioes, M_TEMP);
3889                                         goto fail;
3890                                 }
3891                                 break;
3892                             }
3893                         default:
3894                                 if ((error = pf_begin_rules(&ioe->ticket,
3895                                     ioe->rs_num, ioe->anchor))) {
3896                                         PF_RULES_WUNLOCK();
3897                                         free(ioes, M_TEMP);
3898                                         goto fail;
3899                                 }
3900                                 break;
3901                         }
3902                 }
3903                 PF_RULES_WUNLOCK();
3904                 error = copyout(ioes, io->array, totlen);
3905                 free(ioes, M_TEMP);
3906                 break;
3907         }
3908
3909         case DIOCXROLLBACK: {
3910                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3911                 struct pfioc_trans_e    *ioe, *ioes;
3912                 size_t                   totlen;
3913                 int                      i;
3914
3915                 if (io->esize != sizeof(*ioe)) {
3916                         error = ENODEV;
3917                         break;
3918                 }
3919                 if (io->size < 0 ||
3920                     io->size > pf_ioctl_maxcount ||
3921                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3922                         error = EINVAL;
3923                         break;
3924                 }
3925                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3926                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3927                     M_TEMP, M_NOWAIT);
3928                 if (! ioes) {
3929                         error = ENOMEM;
3930                         break;
3931                 }
3932                 error = copyin(io->array, ioes, totlen);
3933                 if (error) {
3934                         free(ioes, M_TEMP);
3935                         break;
3936                 }
3937                 PF_RULES_WLOCK();
3938                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3939                         switch (ioe->rs_num) {
3940 #ifdef ALTQ
3941                         case PF_RULESET_ALTQ:
3942                                 if (ioe->anchor[0]) {
3943                                         PF_RULES_WUNLOCK();
3944                                         free(ioes, M_TEMP);
3945                                         error = EINVAL;
3946                                         goto fail;
3947                                 }
3948                                 if ((error = pf_rollback_altq(ioe->ticket))) {
3949                                         PF_RULES_WUNLOCK();
3950                                         free(ioes, M_TEMP);
3951                                         goto fail; /* really bad */
3952                                 }
3953                                 break;
3954 #endif /* ALTQ */
3955                         case PF_RULESET_TABLE:
3956                             {
3957                                 struct pfr_table table;
3958
3959                                 bzero(&table, sizeof(table));
3960                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3961                                     sizeof(table.pfrt_anchor));
3962                                 if ((error = pfr_ina_rollback(&table,
3963                                     ioe->ticket, NULL, 0))) {
3964                                         PF_RULES_WUNLOCK();
3965                                         free(ioes, M_TEMP);
3966                                         goto fail; /* really bad */
3967                                 }
3968                                 break;
3969                             }
3970                         default:
3971                                 if ((error = pf_rollback_rules(ioe->ticket,
3972                                     ioe->rs_num, ioe->anchor))) {
3973                                         PF_RULES_WUNLOCK();
3974                                         free(ioes, M_TEMP);
3975                                         goto fail; /* really bad */
3976                                 }
3977                                 break;
3978                         }
3979                 }
3980                 PF_RULES_WUNLOCK();
3981                 free(ioes, M_TEMP);
3982                 break;
3983         }
3984
3985         case DIOCXCOMMIT: {
3986                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3987                 struct pfioc_trans_e    *ioe, *ioes;
3988                 struct pf_kruleset      *rs;
3989                 size_t                   totlen;
3990                 int                      i;
3991
3992                 if (io->esize != sizeof(*ioe)) {
3993                         error = ENODEV;
3994                         break;
3995                 }
3996
3997                 if (io->size < 0 ||
3998                     io->size > pf_ioctl_maxcount ||
3999                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4000                         error = EINVAL;
4001                         break;
4002                 }
4003
4004                 totlen = sizeof(struct pfioc_trans_e) * io->size;
4005                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4006                     M_TEMP, M_NOWAIT);
4007                 if (ioes == NULL) {
4008                         error = ENOMEM;
4009                         break;
4010                 }
4011                 error = copyin(io->array, ioes, totlen);
4012                 if (error) {
4013                         free(ioes, M_TEMP);
4014                         break;
4015                 }
4016                 PF_RULES_WLOCK();
4017                 /* First makes sure everything will succeed. */
4018                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4019                         switch (ioe->rs_num) {
4020 #ifdef ALTQ
4021                         case PF_RULESET_ALTQ:
4022                                 if (ioe->anchor[0]) {
4023                                         PF_RULES_WUNLOCK();
4024                                         free(ioes, M_TEMP);
4025                                         error = EINVAL;
4026                                         goto fail;
4027                                 }
4028                                 if (!V_altqs_inactive_open || ioe->ticket !=
4029                                     V_ticket_altqs_inactive) {
4030                                         PF_RULES_WUNLOCK();
4031                                         free(ioes, M_TEMP);
4032                                         error = EBUSY;
4033                                         goto fail;
4034                                 }
4035                                 break;
4036 #endif /* ALTQ */
4037                         case PF_RULESET_TABLE:
4038                                 rs = pf_find_kruleset(ioe->anchor);
4039                                 if (rs == NULL || !rs->topen || ioe->ticket !=
4040                                     rs->tticket) {
4041                                         PF_RULES_WUNLOCK();
4042                                         free(ioes, M_TEMP);
4043                                         error = EBUSY;
4044                                         goto fail;
4045                                 }
4046                                 break;
4047                         default:
4048                                 if (ioe->rs_num < 0 || ioe->rs_num >=
4049                                     PF_RULESET_MAX) {
4050                                         PF_RULES_WUNLOCK();
4051                                         free(ioes, M_TEMP);
4052                                         error = EINVAL;
4053                                         goto fail;
4054                                 }
4055                                 rs = pf_find_kruleset(ioe->anchor);
4056                                 if (rs == NULL ||
4057                                     !rs->rules[ioe->rs_num].inactive.open ||
4058                                     rs->rules[ioe->rs_num].inactive.ticket !=
4059                                     ioe->ticket) {
4060                                         PF_RULES_WUNLOCK();
4061                                         free(ioes, M_TEMP);
4062                                         error = EBUSY;
4063                                         goto fail;
4064                                 }
4065                                 break;
4066                         }
4067                 }
4068                 /* Now do the commit - no errors should happen here. */
4069                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4070                         switch (ioe->rs_num) {
4071 #ifdef ALTQ
4072                         case PF_RULESET_ALTQ:
4073                                 if ((error = pf_commit_altq(ioe->ticket))) {
4074                                         PF_RULES_WUNLOCK();
4075                                         free(ioes, M_TEMP);
4076                                         goto fail; /* really bad */
4077                                 }
4078                                 break;
4079 #endif /* ALTQ */
4080                         case PF_RULESET_TABLE:
4081                             {
4082                                 struct pfr_table table;
4083
4084                                 bzero(&table, sizeof(table));
4085                                 strlcpy(table.pfrt_anchor, ioe->anchor,
4086                                     sizeof(table.pfrt_anchor));
4087                                 if ((error = pfr_ina_commit(&table,
4088                                     ioe->ticket, NULL, NULL, 0))) {
4089                                         PF_RULES_WUNLOCK();
4090                                         free(ioes, M_TEMP);
4091                                         goto fail; /* really bad */
4092                                 }
4093                                 break;
4094                             }
4095                         default:
4096                                 if ((error = pf_commit_rules(ioe->ticket,
4097                                     ioe->rs_num, ioe->anchor))) {
4098                                         PF_RULES_WUNLOCK();
4099                                         free(ioes, M_TEMP);
4100                                         goto fail; /* really bad */
4101                                 }
4102                                 break;
4103                         }
4104                 }
4105                 PF_RULES_WUNLOCK();
4106                 free(ioes, M_TEMP);
4107                 break;
4108         }
4109
4110         case DIOCGETSRCNODES: {
4111                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
4112                 struct pf_srchash       *sh;
4113                 struct pf_ksrc_node     *n;
4114                 struct pf_src_node      *p, *pstore;
4115                 uint32_t                 i, nr = 0;
4116
4117                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4118                                 i++, sh++) {
4119                         PF_HASHROW_LOCK(sh);
4120                         LIST_FOREACH(n, &sh->nodes, entry)
4121                                 nr++;
4122                         PF_HASHROW_UNLOCK(sh);
4123                 }
4124
4125                 psn->psn_len = min(psn->psn_len,
4126                     sizeof(struct pf_src_node) * nr);
4127
4128                 if (psn->psn_len == 0) {
4129                         psn->psn_len = sizeof(struct pf_src_node) * nr;
4130                         break;
4131                 }
4132
4133                 nr = 0;
4134
4135                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4136                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4137                     i++, sh++) {
4138                     PF_HASHROW_LOCK(sh);
4139                     LIST_FOREACH(n, &sh->nodes, entry) {
4140
4141                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4142                                 break;
4143
4144                         pf_src_node_copy(n, p);
4145
4146                         p++;
4147                         nr++;
4148                     }
4149                     PF_HASHROW_UNLOCK(sh);
4150                 }
4151                 error = copyout(pstore, psn->psn_src_nodes,
4152                     sizeof(struct pf_src_node) * nr);
4153                 if (error) {
4154                         free(pstore, M_TEMP);
4155                         break;
4156                 }
4157                 psn->psn_len = sizeof(struct pf_src_node) * nr;
4158                 free(pstore, M_TEMP);
4159                 break;
4160         }
4161
4162         case DIOCCLRSRCNODES: {
4163
4164                 pf_clear_srcnodes(NULL);
4165                 pf_purge_expired_src_nodes();
4166                 break;
4167         }
4168
4169         case DIOCKILLSRCNODES:
4170                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4171                 break;
4172
4173         case DIOCSETHOSTID: {
4174                 u_int32_t       *hostid = (u_int32_t *)addr;
4175
4176                 PF_RULES_WLOCK();
4177                 if (*hostid == 0)
4178                         V_pf_status.hostid = arc4random();
4179                 else
4180                         V_pf_status.hostid = *hostid;
4181                 PF_RULES_WUNLOCK();
4182                 break;
4183         }
4184
4185         case DIOCOSFPFLUSH:
4186                 PF_RULES_WLOCK();
4187                 pf_osfp_flush();
4188                 PF_RULES_WUNLOCK();
4189                 break;
4190
4191         case DIOCIGETIFACES: {
4192                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4193                 struct pfi_kif *ifstore;
4194                 size_t bufsiz;
4195
4196                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4197                         error = ENODEV;
4198                         break;
4199                 }
4200
4201                 if (io->pfiio_size < 0 ||
4202                     io->pfiio_size > pf_ioctl_maxcount ||
4203                     WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4204                         error = EINVAL;
4205                         break;
4206                 }
4207
4208                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4209                 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4210                     M_TEMP, M_NOWAIT);
4211                 if (ifstore == NULL) {
4212                         error = ENOMEM;
4213                         break;
4214                 }
4215
4216                 PF_RULES_RLOCK();
4217                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4218                 PF_RULES_RUNLOCK();
4219                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4220                 free(ifstore, M_TEMP);
4221                 break;
4222         }
4223
4224         case DIOCSETIFFLAG: {
4225                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4226
4227                 PF_RULES_WLOCK();
4228                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4229                 PF_RULES_WUNLOCK();
4230                 break;
4231         }
4232
4233         case DIOCCLRIFFLAG: {
4234                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4235
4236                 PF_RULES_WLOCK();
4237                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4238                 PF_RULES_WUNLOCK();
4239                 break;
4240         }
4241
4242         default:
4243                 error = ENODEV;
4244                 break;
4245         }
4246 fail:
4247         if (sx_xlocked(&pf_ioctl_lock))
4248                 sx_xunlock(&pf_ioctl_lock);
4249         CURVNET_RESTORE();
4250
4251         return (error);
4252 }
4253
4254 void
4255 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4256 {
4257         bzero(sp, sizeof(struct pfsync_state));
4258
4259         /* copy from state key */
4260         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4261         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4262         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4263         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4264         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4265         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4266         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4267         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4268         sp->proto = st->key[PF_SK_WIRE]->proto;
4269         sp->af = st->key[PF_SK_WIRE]->af;
4270
4271         /* copy from state */
4272         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4273         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4274         sp->creation = htonl(time_uptime - st->creation);
4275         sp->expire = pf_state_expires(st);
4276         if (sp->expire <= time_uptime)
4277                 sp->expire = htonl(0);
4278         else
4279                 sp->expire = htonl(sp->expire - time_uptime);
4280
4281         sp->direction = st->direction;
4282         sp->log = st->log;
4283         sp->timeout = st->timeout;
4284         sp->state_flags = st->state_flags;
4285         if (st->src_node)
4286                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4287         if (st->nat_src_node)
4288                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4289
4290         sp->id = st->id;
4291         sp->creatorid = st->creatorid;
4292         pf_state_peer_hton(&st->src, &sp->src);
4293         pf_state_peer_hton(&st->dst, &sp->dst);
4294
4295         if (st->rule.ptr == NULL)
4296                 sp->rule = htonl(-1);
4297         else
4298                 sp->rule = htonl(st->rule.ptr->nr);
4299         if (st->anchor.ptr == NULL)
4300                 sp->anchor = htonl(-1);
4301         else
4302                 sp->anchor = htonl(st->anchor.ptr->nr);
4303         if (st->nat_rule.ptr == NULL)
4304                 sp->nat_rule = htonl(-1);
4305         else
4306                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4307
4308         pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4309             sp->packets[0]);
4310         pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4311             sp->packets[1]);
4312         pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4313         pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4314
4315 }
4316
4317 static void
4318 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4319 {
4320         struct pfr_ktable *kt;
4321
4322         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4323
4324         kt = aw->p.tbl;
4325         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4326                 kt = kt->pfrkt_root;
4327         aw->p.tbl = NULL;
4328         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4329                 kt->pfrkt_cnt : -1;
4330 }
4331
4332 /*
4333  * XXX - Check for version missmatch!!!
4334  */
4335 static void
4336 pf_clear_states(void)
4337 {
4338         struct pf_state *s;
4339         u_int i;
4340
4341         for (i = 0; i <= pf_hashmask; i++) {
4342                 struct pf_idhash *ih = &V_pf_idhash[i];
4343 relock:
4344                 PF_HASHROW_LOCK(ih);
4345                 LIST_FOREACH(s, &ih->states, entry) {
4346                         s->timeout = PFTM_PURGE;
4347                         /* Don't send out individual delete messages. */
4348                         s->state_flags |= PFSTATE_NOSYNC;
4349                         pf_unlink_state(s, PF_ENTER_LOCKED);
4350                         goto relock;
4351                 }
4352                 PF_HASHROW_UNLOCK(ih);
4353         }
4354 }
4355
4356 static int
4357 pf_clear_tables(void)
4358 {
4359         struct pfioc_table io;
4360         int error;
4361
4362         bzero(&io, sizeof(io));
4363
4364         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4365             io.pfrio_flags);
4366
4367         return (error);
4368 }
4369
4370 static void
4371 pf_clear_srcnodes(struct pf_ksrc_node *n)
4372 {
4373         struct pf_state *s;
4374         int i;
4375
4376         for (i = 0; i <= pf_hashmask; i++) {
4377                 struct pf_idhash *ih = &V_pf_idhash[i];
4378
4379                 PF_HASHROW_LOCK(ih);
4380                 LIST_FOREACH(s, &ih->states, entry) {
4381                         if (n == NULL || n == s->src_node)
4382                                 s->src_node = NULL;
4383                         if (n == NULL || n == s->nat_src_node)
4384                                 s->nat_src_node = NULL;
4385                 }
4386                 PF_HASHROW_UNLOCK(ih);
4387         }
4388
4389         if (n == NULL) {
4390                 struct pf_srchash *sh;
4391
4392                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4393                     i++, sh++) {
4394                         PF_HASHROW_LOCK(sh);
4395                         LIST_FOREACH(n, &sh->nodes, entry) {
4396                                 n->expire = 1;
4397                                 n->states = 0;
4398                         }
4399                         PF_HASHROW_UNLOCK(sh);
4400                 }
4401         } else {
4402                 /* XXX: hash slot should already be locked here. */
4403                 n->expire = 1;
4404                 n->states = 0;
4405         }
4406 }
4407
4408 static void
4409 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4410 {
4411         struct pf_ksrc_node_list         kill;
4412
4413         LIST_INIT(&kill);
4414         for (int i = 0; i <= pf_srchashmask; i++) {
4415                 struct pf_srchash *sh = &V_pf_srchash[i];
4416                 struct pf_ksrc_node *sn, *tmp;
4417
4418                 PF_HASHROW_LOCK(sh);
4419                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4420                         if (PF_MATCHA(psnk->psnk_src.neg,
4421                               &psnk->psnk_src.addr.v.a.addr,
4422                               &psnk->psnk_src.addr.v.a.mask,
4423                               &sn->addr, sn->af) &&
4424                             PF_MATCHA(psnk->psnk_dst.neg,
4425                               &psnk->psnk_dst.addr.v.a.addr,
4426                               &psnk->psnk_dst.addr.v.a.mask,
4427                               &sn->raddr, sn->af)) {
4428                                 pf_unlink_src_node(sn);
4429                                 LIST_INSERT_HEAD(&kill, sn, entry);
4430                                 sn->expire = 1;
4431                         }
4432                 PF_HASHROW_UNLOCK(sh);
4433         }
4434
4435         for (int i = 0; i <= pf_hashmask; i++) {
4436                 struct pf_idhash *ih = &V_pf_idhash[i];
4437                 struct pf_state *s;
4438
4439                 PF_HASHROW_LOCK(ih);
4440                 LIST_FOREACH(s, &ih->states, entry) {
4441                         if (s->src_node && s->src_node->expire == 1)
4442                                 s->src_node = NULL;
4443                         if (s->nat_src_node && s->nat_src_node->expire == 1)
4444                                 s->nat_src_node = NULL;
4445                 }
4446                 PF_HASHROW_UNLOCK(ih);
4447         }
4448
4449         psnk->psnk_killed = pf_free_src_nodes(&kill);
4450 }
4451
4452 /*
4453  * XXX - Check for version missmatch!!!
4454  */
4455
4456 /*
4457  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4458  */
4459 static int
4460 shutdown_pf(void)
4461 {
4462         int error = 0;
4463         u_int32_t t[5];
4464         char nn = '\0';
4465
4466         do {
4467                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4468                     != 0) {
4469                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4470                         break;
4471                 }
4472                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4473                     != 0) {
4474                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4475                         break;          /* XXX: rollback? */
4476                 }
4477                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4478                     != 0) {
4479                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4480                         break;          /* XXX: rollback? */
4481                 }
4482                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4483                     != 0) {
4484                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4485                         break;          /* XXX: rollback? */
4486                 }
4487                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4488                     != 0) {
4489                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4490                         break;          /* XXX: rollback? */
4491                 }
4492
4493                 /* XXX: these should always succeed here */
4494                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4495                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4496                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4497                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4498                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4499
4500                 if ((error = pf_clear_tables()) != 0)
4501                         break;
4502
4503 #ifdef ALTQ
4504                 if ((error = pf_begin_altq(&t[0])) != 0) {
4505                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4506                         break;
4507                 }
4508                 pf_commit_altq(t[0]);
4509 #endif
4510
4511                 pf_clear_states();
4512
4513                 pf_clear_srcnodes(NULL);
4514
4515                 /* status does not use malloced mem so no need to cleanup */
4516                 /* fingerprints and interfaces have their own cleanup code */
4517         } while(0);
4518
4519         return (error);
4520 }
4521
4522 #ifdef INET
4523 static int
4524 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4525     struct inpcb *inp)
4526 {
4527         int chk;
4528
4529         chk = pf_test(PF_IN, flags, ifp, m, inp);
4530         if (chk && *m) {
4531                 m_freem(*m);
4532                 *m = NULL;
4533         }
4534
4535         if (chk != PF_PASS)
4536                 return (EACCES);
4537         return (0);
4538 }
4539
4540 static int
4541 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4542     struct inpcb *inp)
4543 {
4544         int chk;
4545
4546         chk = pf_test(PF_OUT, flags, ifp, m, inp);
4547         if (chk && *m) {
4548                 m_freem(*m);
4549                 *m = NULL;
4550         }
4551
4552         if (chk != PF_PASS)
4553                 return (EACCES);
4554         return (0);
4555 }
4556 #endif
4557
4558 #ifdef INET6
4559 static int
4560 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4561     struct inpcb *inp)
4562 {
4563         int chk;
4564
4565         /*
4566          * In case of loopback traffic IPv6 uses the real interface in
4567          * order to support scoped addresses. In order to support stateful
4568          * filtering we have change this to lo0 as it is the case in IPv4.
4569          */
4570         CURVNET_SET(ifp->if_vnet);
4571         chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4572         CURVNET_RESTORE();
4573         if (chk && *m) {
4574                 m_freem(*m);
4575                 *m = NULL;
4576         }
4577         if (chk != PF_PASS)
4578                 return (EACCES);
4579         return (0);
4580 }
4581
4582 static int
4583 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4584     struct inpcb *inp)
4585 {
4586         int chk;
4587
4588         CURVNET_SET(ifp->if_vnet);
4589         chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4590         CURVNET_RESTORE();
4591         if (chk && *m) {
4592                 m_freem(*m);
4593                 *m = NULL;
4594         }
4595         if (chk != PF_PASS)
4596                 return (EACCES);
4597         return (0);
4598 }
4599 #endif /* INET6 */
4600
4601 static int
4602 hook_pf(void)
4603 {
4604 #ifdef INET
4605         struct pfil_head *pfh_inet;
4606 #endif
4607 #ifdef INET6
4608         struct pfil_head *pfh_inet6;
4609 #endif
4610
4611         if (V_pf_pfil_hooked)
4612                 return (0);
4613
4614 #ifdef INET
4615         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
4616         if (pfh_inet == NULL)
4617                 return (ESRCH); /* XXX */
4618         pfil_add_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
4619         pfil_add_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
4620 #endif
4621 #ifdef INET6
4622         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
4623         if (pfh_inet6 == NULL) {
4624 #ifdef INET
4625                 pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
4626                     pfh_inet);
4627                 pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
4628                     pfh_inet);
4629 #endif
4630                 return (ESRCH); /* XXX */
4631         }
4632         pfil_add_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
4633         pfil_add_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
4634 #endif
4635
4636         V_pf_pfil_hooked = 1;
4637         return (0);
4638 }
4639
4640 static int
4641 dehook_pf(void)
4642 {
4643 #ifdef INET
4644         struct pfil_head *pfh_inet;
4645 #endif
4646 #ifdef INET6
4647         struct pfil_head *pfh_inet6;
4648 #endif
4649
4650         if (V_pf_pfil_hooked == 0)
4651                 return (0);
4652
4653 #ifdef INET
4654         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
4655         if (pfh_inet == NULL)
4656                 return (ESRCH); /* XXX */
4657         pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
4658             pfh_inet);
4659         pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
4660             pfh_inet);
4661 #endif
4662 #ifdef INET6
4663         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
4664         if (pfh_inet6 == NULL)
4665                 return (ESRCH); /* XXX */
4666         pfil_remove_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
4667             pfh_inet6);
4668         pfil_remove_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
4669             pfh_inet6);
4670 #endif
4671
4672         V_pf_pfil_hooked = 0;
4673         return (0);
4674 }
4675
4676 static void
4677 pf_load_vnet(void)
4678 {
4679         V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4680             NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4681
4682         pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4683             PF_RULE_TAG_HASH_SIZE_DEFAULT);
4684 #ifdef ALTQ
4685         pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4686             PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4687 #endif
4688
4689         pfattach_vnet();
4690         V_pf_vnet_active = 1;
4691 }
4692
4693 static int
4694 pf_load(void)
4695 {
4696         int error;
4697
4698         rm_init(&pf_rules_lock, "pf rulesets");
4699         sx_init(&pf_ioctl_lock, "pf ioctl");
4700         sx_init(&pf_end_lock, "pf end thread");
4701
4702         pf_mtag_initialize();
4703
4704         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
4705         if (pf_dev == NULL)
4706                 return (ENOMEM);
4707
4708         pf_end_threads = 0;
4709         error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4710         if (error != 0)
4711                 return (error);
4712
4713         pfi_initialize();
4714
4715         return (0);
4716 }
4717
4718 static void
4719 pf_unload_vnet(void)
4720 {
4721         int error, ret;
4722
4723         V_pf_vnet_active = 0;
4724         V_pf_status.running = 0;
4725         error = dehook_pf();
4726         if (error) {
4727                 /*
4728                  * Should not happen!
4729                  * XXX Due to error code ESRCH, kldunload will show
4730                  * a message like 'No such process'.
4731                  */
4732                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4733                 return;
4734         }
4735
4736         PF_RULES_WLOCK();
4737         shutdown_pf();
4738         PF_RULES_WUNLOCK();
4739
4740         ret = swi_remove(V_pf_swi_cookie);
4741         MPASS(ret == 0);
4742         ret = intr_event_destroy(V_pf_swi_ie);
4743         MPASS(ret == 0);
4744
4745         pf_unload_vnet_purge();
4746
4747         pf_normalize_cleanup();
4748         PF_RULES_WLOCK();
4749         pfi_cleanup_vnet();
4750         PF_RULES_WUNLOCK();
4751         pfr_cleanup();
4752         pf_osfp_flush();
4753         pf_cleanup();
4754         if (IS_DEFAULT_VNET(curvnet))
4755                 pf_mtag_cleanup();
4756
4757         pf_cleanup_tagset(&V_pf_tags);
4758 #ifdef ALTQ
4759         pf_cleanup_tagset(&V_pf_qids);
4760 #endif
4761         uma_zdestroy(V_pf_tag_z);
4762
4763         /* Free counters last as we updated them during shutdown. */
4764         counter_u64_free(V_pf_default_rule.evaluations);
4765         for (int i = 0; i < 2; i++) {
4766                 counter_u64_free(V_pf_default_rule.packets[i]);
4767                 counter_u64_free(V_pf_default_rule.bytes[i]);
4768         }
4769         counter_u64_free(V_pf_default_rule.states_cur);
4770         counter_u64_free(V_pf_default_rule.states_tot);
4771         counter_u64_free(V_pf_default_rule.src_nodes);
4772
4773         for (int i = 0; i < PFRES_MAX; i++)
4774                 counter_u64_free(V_pf_status.counters[i]);
4775         for (int i = 0; i < LCNT_MAX; i++)
4776                 counter_u64_free(V_pf_status.lcounters[i]);
4777         for (int i = 0; i < FCNT_MAX; i++)
4778                 counter_u64_free(V_pf_status.fcounters[i]);
4779         for (int i = 0; i < SCNT_MAX; i++)
4780                 counter_u64_free(V_pf_status.scounters[i]);
4781 }
4782
4783 static void
4784 pf_unload(void)
4785 {
4786
4787         sx_xlock(&pf_end_lock);
4788         pf_end_threads = 1;
4789         while (pf_end_threads < 2) {
4790                 wakeup_one(pf_purge_thread);
4791                 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4792         }
4793         sx_xunlock(&pf_end_lock);
4794
4795         if (pf_dev != NULL)
4796                 destroy_dev(pf_dev);
4797
4798         pfi_cleanup();
4799
4800         rm_destroy(&pf_rules_lock);
4801         sx_destroy(&pf_ioctl_lock);
4802         sx_destroy(&pf_end_lock);
4803 }
4804
4805 static void
4806 vnet_pf_init(void *unused __unused)
4807 {
4808
4809         pf_load_vnet();
4810 }
4811 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4812     vnet_pf_init, NULL);
4813
4814 static void
4815 vnet_pf_uninit(const void *unused __unused)
4816 {
4817
4818         pf_unload_vnet();
4819 }
4820 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4821 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4822     vnet_pf_uninit, NULL);
4823
4824
4825 static int
4826 pf_modevent(module_t mod, int type, void *data)
4827 {
4828         int error = 0;
4829
4830         switch(type) {
4831         case MOD_LOAD:
4832                 error = pf_load();
4833                 break;
4834         case MOD_UNLOAD:
4835                 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4836                  * the vnet_pf_uninit()s */
4837                 break;
4838         default:
4839                 error = EINVAL;
4840                 break;
4841         }
4842
4843         return (error);
4844 }
4845
4846 static moduledata_t pf_mod = {
4847         "pf",
4848         pf_modevent,
4849         0
4850 };
4851
4852 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4853 MODULE_VERSION(pf, PF_MODVER);