sys/netpfil/pf/pf_ioctl.c

   1 /*-
   2  * Copyright (c) 2001 Daniel Hartmeier
   3  * Copyright (c) 2002,2003 Henning Brauer
   4  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  *    - Redistributions of source code must retain the above copyright
  12  *      notice, this list of conditions and the following disclaimer.
  13  *    - Redistributions in binary form must reproduce the above
  14  *      copyright notice, this list of conditions and the following
  15  *      disclaimer in the documentation and/or other materials provided
  16  *      with the distribution.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  21  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  22  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  26  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  29  * POSSIBILITY OF SUCH DAMAGE.
  30  *
  31  * Effort sponsored in part by the Defense Advanced Research Projects
  32  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  33  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  34  *
  35  *      $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
  36  */
  37
  38 #include <sys/cdefs.h>
  39 __FBSDID("$FreeBSD$");
  40
  41 #include "opt_inet.h"
  42 #include "opt_inet6.h"
  43 #include "opt_bpf.h"
  44 #include "opt_pf.h"
  45
  46 #include <sys/param.h>
  47 #include <sys/bus.h>
  48 #include <sys/conf.h>
  49 #include <sys/endian.h>
  50 #include <sys/fcntl.h>
  51 #include <sys/filio.h>
  52 #include <sys/interrupt.h>
  53 #include <sys/jail.h>
  54 #include <sys/kernel.h>
  55 #include <sys/kthread.h>
  56 #include <sys/mbuf.h>
  57 #include <sys/module.h>
  58 #include <sys/proc.h>
  59 #include <sys/smp.h>
  60 #include <sys/socket.h>
  61 #include <sys/sysctl.h>
  62 #include <sys/md5.h>
  63 #include <sys/ucred.h>
  64
  65 #include <net/if.h>
  66 #include <net/route.h>
  67 #include <net/pfil.h>
  68 #include <net/pfvar.h>
  69 #include <net/if_pfsync.h>
  70 #include <net/if_pflog.h>
  71
  72 #include <netinet/in.h>
  73 #include <netinet/ip.h>
  74 #include <netinet/ip_var.h>
  75 #include <netinet6/ip6_var.h>
  76 #include <netinet/ip_icmp.h>
  77
  78 #ifdef INET6
  79 #include <netinet/ip6.h>
  80 #endif /* INET6 */
  81
  82 #ifdef ALTQ
  83 #include <altq/altq.h>
  84 #endif
  85
  86 static int               pfattach(void);
  87 static struct pf_pool   *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
  88                             u_int8_t, u_int8_t, u_int8_t);
  89
  90 static void              pf_mv_pool(struct pf_palist *, struct pf_palist *);
  91 static void              pf_empty_pool(struct pf_palist *);
  92 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
  93                             struct thread *);
  94 #ifdef ALTQ
  95 static int               pf_begin_altq(u_int32_t *);
  96 static int               pf_rollback_altq(u_int32_t);
  97 static int               pf_commit_altq(u_int32_t);
  98 static int               pf_enable_altq(struct pf_altq *);
  99 static int               pf_disable_altq(struct pf_altq *);
 100 static u_int32_t         pf_qname2qid(char *);
 101 static void              pf_qid_unref(u_int32_t);
 102 #endif /* ALTQ */
 103 static int               pf_begin_rules(u_int32_t *, int, const char *);
 104 static int               pf_rollback_rules(u_int32_t, int, char *);
 105 static int               pf_setup_pfsync_matching(struct pf_ruleset *);
 106 static void              pf_hash_rule(MD5_CTX *, struct pf_rule *);
 107 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 108 static int               pf_commit_rules(u_int32_t, int, char *);
 109 static int               pf_addr_setup(struct pf_ruleset *,
 110                             struct pf_addr_wrap *, sa_family_t);
 111 static void              pf_addr_copyout(struct pf_addr_wrap *);
 112
 113 VNET_DEFINE(struct pf_rule,     pf_default_rule);
 114
 115 #ifdef ALTQ
 116 static VNET_DEFINE(int,         pf_altq_running);
 117 #define V_pf_altq_running       VNET(pf_altq_running)
 118 #endif
 119
 120 #define TAGID_MAX        50000
 121 struct pf_tagname {
 122         TAILQ_ENTRY(pf_tagname) entries;
 123         char                    name[PF_TAG_NAME_SIZE];
 124         uint16_t                tag;
 125         int                     ref;
 126 };
 127
 128 TAILQ_HEAD(pf_tags, pf_tagname);
 129 #define V_pf_tags               VNET(pf_tags)
 130 VNET_DEFINE(struct pf_tags, pf_tags);
 131 #define V_pf_qids               VNET(pf_qids)
 132 VNET_DEFINE(struct pf_tags, pf_qids);
 133 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
 134 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 135 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 136
 137 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 138 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 139 #endif
 140
 141 static u_int16_t         tagname2tag(struct pf_tags *, char *);
 142 static u_int16_t         pf_tagname2tag(char *);
 143 static void              tag_unref(struct pf_tags *, u_int16_t);
 144
 145 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 146
 147 struct cdev *pf_dev;
 148
 149 /*
 150  * XXX - These are new and need to be checked when moveing to a new version
 151  */
 152 static void              pf_clear_states(void);
 153 static int               pf_clear_tables(void);
 154 static void              pf_clear_srcnodes(struct pf_src_node *);
 155 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 156 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 157
 158 /*
 159  * Wrapper functions for pfil(9) hooks
 160  */
 161 #ifdef INET
 162 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 163     int dir, struct inpcb *inp);
 164 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 165     int dir, struct inpcb *inp);
 166 #endif
 167 #ifdef INET6
 168 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 169     int dir, struct inpcb *inp);
 170 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 171     int dir, struct inpcb *inp);
 172 #endif
 173
 174 static int              hook_pf(void);
 175 static int              dehook_pf(void);
 176 static int              shutdown_pf(void);
 177 static int              pf_load(void);
 178 static int              pf_unload(void);
 179
 180 static struct cdevsw pf_cdevsw = {
 181         .d_ioctl =      pfioctl,
 182         .d_name =       PF_NAME,
 183         .d_version =    D_VERSION,
 184 };
 185
 186 static volatile VNET_DEFINE(int, pf_pfil_hooked);
 187 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 188 VNET_DEFINE(int,                pf_end_threads);
 189
 190 struct rwlock                   pf_rules_lock;
 191 struct sx                       pf_ioctl_lock;
 192
 193 /* pfsync */
 194 pfsync_state_import_t           *pfsync_state_import_ptr = NULL;
 195 pfsync_insert_state_t           *pfsync_insert_state_ptr = NULL;
 196 pfsync_update_state_t           *pfsync_update_state_ptr = NULL;
 197 pfsync_delete_state_t           *pfsync_delete_state_ptr = NULL;
 198 pfsync_clear_states_t           *pfsync_clear_states_ptr = NULL;
 199 pfsync_defer_t                  *pfsync_defer_ptr = NULL;
 200 /* pflog */
 201 pflog_packet_t                  *pflog_packet_ptr = NULL;
 202
 203 static int
 204 pfattach(void)
 205 {
 206         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 207         int error;
 208
 209         if (IS_DEFAULT_VNET(curvnet))
 210                 pf_mtag_initialize();
 211         pf_initialize();
 212         pfr_initialize();
 213         pfi_initialize();
 214         pf_normalize_init();
 215
 216         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 217         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 218
 219         RB_INIT(&V_pf_anchors);
 220         pf_init_ruleset(&pf_main_ruleset);
 221
 222         /* default rule should never be garbage collected */
 223         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 224 #ifdef PF_DEFAULT_TO_DROP
 225         V_pf_default_rule.action = PF_DROP;
 226 #else
 227         V_pf_default_rule.action = PF_PASS;
 228 #endif
 229         V_pf_default_rule.nr = -1;
 230         V_pf_default_rule.rtableid = -1;
 231
 232         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 233         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 234         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 235
 236         /* initialize default timeouts */
 237         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 238         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 239         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 240         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 241         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 242         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 243         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 244         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 245         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 246         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 247         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 248         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 249         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 250         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 251         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 252         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 253         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 254         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 255         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 256         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 257
 258         bzero(&V_pf_status, sizeof(V_pf_status));
 259         V_pf_status.debug = PF_DEBUG_URGENT;
 260
 261         V_pf_pfil_hooked = 0;
 262
 263         /* XXX do our best to avoid a conflict */
 264         V_pf_status.hostid = arc4random();
 265
 266         for (int i = 0; i < PFRES_MAX; i++)
 267                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 268         for (int i = 0; i < LCNT_MAX; i++)
 269                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 270         for (int i = 0; i < FCNT_MAX; i++)
 271                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 272         for (int i = 0; i < SCNT_MAX; i++)
 273                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 274
 275         if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
 276             "pf purge")) != 0)
 277                 /* XXXGL: leaked all above. */
 278                 return (error);
 279         if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
 280             INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
 281                 /* XXXGL: leaked all above. */
 282                 return (error);
 283
 284         return (0);
 285 }
 286
 287 static struct pf_pool *
 288 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 289     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 290     u_int8_t check_ticket)
 291 {
 292         struct pf_ruleset       *ruleset;
 293         struct pf_rule          *rule;
 294         int                      rs_num;
 295
 296         ruleset = pf_find_ruleset(anchor);
 297         if (ruleset == NULL)
 298                 return (NULL);
 299         rs_num = pf_get_ruleset_number(rule_action);
 300         if (rs_num >= PF_RULESET_MAX)
 301                 return (NULL);
 302         if (active) {
 303                 if (check_ticket && ticket !=
 304                     ruleset->rules[rs_num].active.ticket)
 305                         return (NULL);
 306                 if (r_last)
 307                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 308                             pf_rulequeue);
 309                 else
 310                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 311         } else {
 312                 if (check_ticket && ticket !=
 313                     ruleset->rules[rs_num].inactive.ticket)
 314                         return (NULL);
 315                 if (r_last)
 316                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 317                             pf_rulequeue);
 318                 else
 319                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 320         }
 321         if (!r_last) {
 322                 while ((rule != NULL) && (rule->nr != rule_number))
 323                         rule = TAILQ_NEXT(rule, entries);
 324         }
 325         if (rule == NULL)
 326                 return (NULL);
 327
 328         return (&rule->rpool);
 329 }
 330
 331 static void
 332 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 333 {
 334         struct pf_pooladdr      *mv_pool_pa;
 335
 336         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 337                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 338                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 339         }
 340 }
 341
 342 static void
 343 pf_empty_pool(struct pf_palist *poola)
 344 {
 345         struct pf_pooladdr *pa;
 346
 347         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 348                 switch (pa->addr.type) {
 349                 case PF_ADDR_DYNIFTL:
 350                         pfi_dynaddr_remove(pa->addr.p.dyn);
 351                         break;
 352                 case PF_ADDR_TABLE:
 353                         pfr_detach_table(pa->addr.p.tbl);
 354                         break;
 355                 }
 356                 if (pa->kif)
 357                         pfi_kif_unref(pa->kif);
 358                 TAILQ_REMOVE(poola, pa, entries);
 359                 free(pa, M_PFRULE);
 360         }
 361 }
 362
 363 static void
 364 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 365 {
 366
 367         PF_RULES_WASSERT();
 368
 369         TAILQ_REMOVE(rulequeue, rule, entries);
 370
 371         PF_UNLNKDRULES_LOCK();
 372         rule->rule_flag |= PFRULE_REFS;
 373         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 374         PF_UNLNKDRULES_UNLOCK();
 375 }
 376
 377 void
 378 pf_free_rule(struct pf_rule *rule)
 379 {
 380
 381         PF_RULES_WASSERT();
 382
 383         if (rule->tag)
 384                 tag_unref(&V_pf_tags, rule->tag);
 385         if (rule->match_tag)
 386                 tag_unref(&V_pf_tags, rule->match_tag);
 387 #ifdef ALTQ
 388         if (rule->pqid != rule->qid)
 389                 pf_qid_unref(rule->pqid);
 390         pf_qid_unref(rule->qid);
 391 #endif
 392         switch (rule->src.addr.type) {
 393         case PF_ADDR_DYNIFTL:
 394                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 395                 break;
 396         case PF_ADDR_TABLE:
 397                 pfr_detach_table(rule->src.addr.p.tbl);
 398                 break;
 399         }
 400         switch (rule->dst.addr.type) {
 401         case PF_ADDR_DYNIFTL:
 402                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 403                 break;
 404         case PF_ADDR_TABLE:
 405                 pfr_detach_table(rule->dst.addr.p.tbl);
 406                 break;
 407         }
 408         if (rule->overload_tbl)
 409                 pfr_detach_table(rule->overload_tbl);
 410         if (rule->kif)
 411                 pfi_kif_unref(rule->kif);
 412         pf_anchor_remove(rule);
 413         pf_empty_pool(&rule->rpool.list);
 414         counter_u64_free(rule->states_cur);
 415         counter_u64_free(rule->states_tot);
 416         counter_u64_free(rule->src_nodes);
 417         free(rule, M_PFRULE);
 418 }
 419
 420 static u_int16_t
 421 tagname2tag(struct pf_tags *head, char *tagname)
 422 {
 423         struct pf_tagname       *tag, *p = NULL;
 424         u_int16_t                new_tagid = 1;
 425
 426         PF_RULES_WASSERT();
 427
 428         TAILQ_FOREACH(tag, head, entries)
 429                 if (strcmp(tagname, tag->name) == 0) {
 430                         tag->ref++;
 431                         return (tag->tag);
 432                 }
 433
 434         /*
 435          * to avoid fragmentation, we do a linear search from the beginning
 436          * and take the first free slot we find. if there is none or the list
 437          * is empty, append a new entry at the end.
 438          */
 439
 440         /* new entry */
 441         if (!TAILQ_EMPTY(head))
 442                 for (p = TAILQ_FIRST(head); p != NULL &&
 443                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 444                         new_tagid = p->tag + 1;
 445
 446         if (new_tagid > TAGID_MAX)
 447                 return (0);
 448
 449         /* allocate and fill new struct pf_tagname */
 450         tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
 451         if (tag == NULL)
 452                 return (0);
 453         strlcpy(tag->name, tagname, sizeof(tag->name));
 454         tag->tag = new_tagid;
 455         tag->ref++;
 456
 457         if (p != NULL)  /* insert new entry before p */
 458                 TAILQ_INSERT_BEFORE(p, tag, entries);
 459         else    /* either list empty or no free slot in between */
 460                 TAILQ_INSERT_TAIL(head, tag, entries);
 461
 462         return (tag->tag);
 463 }
 464
 465 static void
 466 tag_unref(struct pf_tags *head, u_int16_t tag)
 467 {
 468         struct pf_tagname       *p, *next;
 469
 470         PF_RULES_WASSERT();
 471
 472         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 473                 next = TAILQ_NEXT(p, entries);
 474                 if (tag == p->tag) {
 475                         if (--p->ref == 0) {
 476                                 TAILQ_REMOVE(head, p, entries);
 477                                 free(p, M_PFTAG);
 478                         }
 479                         break;
 480                 }
 481         }
 482 }
 483
 484 static u_int16_t
 485 pf_tagname2tag(char *tagname)
 486 {
 487         return (tagname2tag(&V_pf_tags, tagname));
 488 }
 489
 490 #ifdef ALTQ
 491 static u_int32_t
 492 pf_qname2qid(char *qname)
 493 {
 494         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 495 }
 496
 497 static void
 498 pf_qid_unref(u_int32_t qid)
 499 {
 500         tag_unref(&V_pf_qids, (u_int16_t)qid);
 501 }
 502
 503 static int
 504 pf_begin_altq(u_int32_t *ticket)
 505 {
 506         struct pf_altq  *altq;
 507         int              error = 0;
 508
 509         PF_RULES_WASSERT();
 510
 511         /* Purge the old altq list */
 512         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 513                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 514                 if (altq->qname[0] == 0 &&
 515                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 516                         /* detach and destroy the discipline */
 517                         error = altq_remove(altq);
 518                 } else
 519                         pf_qid_unref(altq->qid);
 520                 free(altq, M_PFALTQ);
 521         }
 522         if (error)
 523                 return (error);
 524         *ticket = ++V_ticket_altqs_inactive;
 525         V_altqs_inactive_open = 1;
 526         return (0);
 527 }
 528
 529 static int
 530 pf_rollback_altq(u_int32_t ticket)
 531 {
 532         struct pf_altq  *altq;
 533         int              error = 0;
 534
 535         PF_RULES_WASSERT();
 536
 537         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 538                 return (0);
 539         /* Purge the old altq list */
 540         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 541                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 542                 if (altq->qname[0] == 0 &&
 543                    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 544                         /* detach and destroy the discipline */
 545                         error = altq_remove(altq);
 546                 } else
 547                         pf_qid_unref(altq->qid);
 548                 free(altq, M_PFALTQ);
 549         }
 550         V_altqs_inactive_open = 0;
 551         return (error);
 552 }
 553
 554 static int
 555 pf_commit_altq(u_int32_t ticket)
 556 {
 557         struct pf_altqqueue     *old_altqs;
 558         struct pf_altq          *altq;
 559         int                      err, error = 0;
 560
 561         PF_RULES_WASSERT();
 562
 563         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 564                 return (EBUSY);
 565
 566         /* swap altqs, keep the old. */
 567         old_altqs = V_pf_altqs_active;
 568         V_pf_altqs_active = V_pf_altqs_inactive;
 569         V_pf_altqs_inactive = old_altqs;
 570         V_ticket_altqs_active = V_ticket_altqs_inactive;
 571
 572         /* Attach new disciplines */
 573         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
 574         if (altq->qname[0] == 0 &&
 575            (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 576                         /* attach the discipline */
 577                         error = altq_pfattach(altq);
 578                         if (error == 0 && V_pf_altq_running)
 579                                 error = pf_enable_altq(altq);
 580                         if (error != 0)
 581                                 return (error);
 582                 }
 583         }
 584
 585         /* Purge the old altq list */
 586         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 587                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 588                 if (altq->qname[0] == 0 &&
 589                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 590                         /* detach and destroy the discipline */
 591                         if (V_pf_altq_running)
 592                                 error = pf_disable_altq(altq);
 593                         err = altq_pfdetach(altq);
 594                         if (err != 0 && error == 0)
 595                                 error = err;
 596                         err = altq_remove(altq);
 597                         if (err != 0 && error == 0)
 598                                 error = err;
 599                 } else
 600                         pf_qid_unref(altq->qid);
 601                 free(altq, M_PFALTQ);
 602         }
 603
 604         V_altqs_inactive_open = 0;
 605         return (error);
 606 }
 607
 608 static int
 609 pf_enable_altq(struct pf_altq *altq)
 610 {
 611         struct ifnet            *ifp;
 612         struct tb_profile        tb;
 613         int                      error = 0;
 614
 615         if ((ifp = ifunit(altq->ifname)) == NULL)
 616                 return (EINVAL);
 617
 618         if (ifp->if_snd.altq_type != ALTQT_NONE)
 619                 error = altq_enable(&ifp->if_snd);
 620
 621         /* set tokenbucket regulator */
 622         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 623                 tb.rate = altq->ifbandwidth;
 624                 tb.depth = altq->tbrsize;
 625                 error = tbr_set(&ifp->if_snd, &tb);
 626         }
 627
 628         return (error);
 629 }
 630
 631 static int
 632 pf_disable_altq(struct pf_altq *altq)
 633 {
 634         struct ifnet            *ifp;
 635         struct tb_profile        tb;
 636         int                      error;
 637
 638         if ((ifp = ifunit(altq->ifname)) == NULL)
 639                 return (EINVAL);
 640
 641         /*
 642          * when the discipline is no longer referenced, it was overridden
 643          * by a new one.  if so, just return.
 644          */
 645         if (altq->altq_disc != ifp->if_snd.altq_disc)
 646                 return (0);
 647
 648         error = altq_disable(&ifp->if_snd);
 649
 650         if (error == 0) {
 651                 /* clear tokenbucket regulator */
 652                 tb.rate = 0;
 653                 error = tbr_set(&ifp->if_snd, &tb);
 654         }
 655
 656         return (error);
 657 }
 658
 659 void
 660 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 661 {
 662         struct ifnet    *ifp1;
 663         struct pf_altq  *a1, *a2, *a3;
 664         u_int32_t        ticket;
 665         int              error = 0;
 666
 667         /* Interrupt userland queue modifications */
 668         if (V_altqs_inactive_open)
 669                 pf_rollback_altq(V_ticket_altqs_inactive);
 670
 671         /* Start new altq ruleset */
 672         if (pf_begin_altq(&ticket))
 673                 return;
 674
 675         /* Copy the current active set */
 676         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 677                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 678                 if (a2 == NULL) {
 679                         error = ENOMEM;
 680                         break;
 681                 }
 682                 bcopy(a1, a2, sizeof(struct pf_altq));
 683
 684                 if (a2->qname[0] != 0) {
 685                         if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 686                                 error = EBUSY;
 687                                 free(a2, M_PFALTQ);
 688                                 break;
 689                         }
 690                         a2->altq_disc = NULL;
 691                         TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
 692                                 if (strncmp(a3->ifname, a2->ifname,
 693                                     IFNAMSIZ) == 0 && a3->qname[0] == 0) {
 694                                         a2->altq_disc = a3->altq_disc;
 695                                         break;
 696                                 }
 697                         }
 698                 }
 699                 /* Deactivate the interface in question */
 700                 a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 701                 if ((ifp1 = ifunit(a2->ifname)) == NULL ||
 702                     (remove && ifp1 == ifp)) {
 703                         a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 704                 } else {
 705                         error = altq_add(a2);
 706
 707                         if (ticket != V_ticket_altqs_inactive)
 708                                 error = EBUSY;
 709
 710                         if (error) {
 711                                 free(a2, M_PFALTQ);
 712                                 break;
 713                         }
 714                 }
 715
 716                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 717         }
 718
 719         if (error != 0)
 720                 pf_rollback_altq(ticket);
 721         else
 722                 pf_commit_altq(ticket);
 723 }
 724 #endif /* ALTQ */
 725
 726 static int
 727 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 728 {
 729         struct pf_ruleset       *rs;
 730         struct pf_rule          *rule;
 731
 732         PF_RULES_WASSERT();
 733
 734         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 735                 return (EINVAL);
 736         rs = pf_find_or_create_ruleset(anchor);
 737         if (rs == NULL)
 738                 return (EINVAL);
 739         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 740                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 741                 rs->rules[rs_num].inactive.rcount--;
 742         }
 743         *ticket = ++rs->rules[rs_num].inactive.ticket;
 744         rs->rules[rs_num].inactive.open = 1;
 745         return (0);
 746 }
 747
 748 static int
 749 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 750 {
 751         struct pf_ruleset       *rs;
 752         struct pf_rule          *rule;
 753
 754         PF_RULES_WASSERT();
 755
 756         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 757                 return (EINVAL);
 758         rs = pf_find_ruleset(anchor);
 759         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 760             rs->rules[rs_num].inactive.ticket != ticket)
 761                 return (0);
 762         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 763                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 764                 rs->rules[rs_num].inactive.rcount--;
 765         }
 766         rs->rules[rs_num].inactive.open = 0;
 767         return (0);
 768 }
 769
 770 #define PF_MD5_UPD(st, elm)                                             \
 771                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 772
 773 #define PF_MD5_UPD_STR(st, elm)                                         \
 774                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 775
 776 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 777                 (stor) = htonl((st)->elm);                              \
 778                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 779 } while (0)
 780
 781 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 782                 (stor) = htons((st)->elm);                              \
 783                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 784 } while (0)
 785
 786 static void
 787 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 788 {
 789         PF_MD5_UPD(pfr, addr.type);
 790         switch (pfr->addr.type) {
 791                 case PF_ADDR_DYNIFTL:
 792                         PF_MD5_UPD(pfr, addr.v.ifname);
 793                         PF_MD5_UPD(pfr, addr.iflags);
 794                         break;
 795                 case PF_ADDR_TABLE:
 796                         PF_MD5_UPD(pfr, addr.v.tblname);
 797                         break;
 798                 case PF_ADDR_ADDRMASK:
 799                         /* XXX ignore af? */
 800                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 801                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 802                         break;
 803         }
 804
 805         PF_MD5_UPD(pfr, port[0]);
 806         PF_MD5_UPD(pfr, port[1]);
 807         PF_MD5_UPD(pfr, neg);
 808         PF_MD5_UPD(pfr, port_op);
 809 }
 810
 811 static void
 812 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 813 {
 814         u_int16_t x;
 815         u_int32_t y;
 816
 817         pf_hash_rule_addr(ctx, &rule->src);
 818         pf_hash_rule_addr(ctx, &rule->dst);
 819         PF_MD5_UPD_STR(rule, label);
 820         PF_MD5_UPD_STR(rule, ifname);
 821         PF_MD5_UPD_STR(rule, match_tagname);
 822         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 823         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 824         PF_MD5_UPD_HTONL(rule, prob, y);
 825         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 826         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 827         PF_MD5_UPD(rule, uid.op);
 828         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 829         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 830         PF_MD5_UPD(rule, gid.op);
 831         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 832         PF_MD5_UPD(rule, action);
 833         PF_MD5_UPD(rule, direction);
 834         PF_MD5_UPD(rule, af);
 835         PF_MD5_UPD(rule, quick);
 836         PF_MD5_UPD(rule, ifnot);
 837         PF_MD5_UPD(rule, match_tag_not);
 838         PF_MD5_UPD(rule, natpass);
 839         PF_MD5_UPD(rule, keep_state);
 840         PF_MD5_UPD(rule, proto);
 841         PF_MD5_UPD(rule, type);
 842         PF_MD5_UPD(rule, code);
 843         PF_MD5_UPD(rule, flags);
 844         PF_MD5_UPD(rule, flagset);
 845         PF_MD5_UPD(rule, allow_opts);
 846         PF_MD5_UPD(rule, rt);
 847         PF_MD5_UPD(rule, tos);
 848 }
 849
 850 static int
 851 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 852 {
 853         struct pf_ruleset       *rs;
 854         struct pf_rule          *rule, **old_array;
 855         struct pf_rulequeue     *old_rules;
 856         int                      error;
 857         u_int32_t                old_rcount;
 858
 859         PF_RULES_WASSERT();
 860
 861         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 862                 return (EINVAL);
 863         rs = pf_find_ruleset(anchor);
 864         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 865             ticket != rs->rules[rs_num].inactive.ticket)
 866                 return (EBUSY);
 867
 868         /* Calculate checksum for the main ruleset */
 869         if (rs == &pf_main_ruleset) {
 870                 error = pf_setup_pfsync_matching(rs);
 871                 if (error != 0)
 872                         return (error);
 873         }
 874
 875         /* Swap rules, keep the old. */
 876         old_rules = rs->rules[rs_num].active.ptr;
 877         old_rcount = rs->rules[rs_num].active.rcount;
 878         old_array = rs->rules[rs_num].active.ptr_array;
 879
 880         rs->rules[rs_num].active.ptr =
 881             rs->rules[rs_num].inactive.ptr;
 882         rs->rules[rs_num].active.ptr_array =
 883             rs->rules[rs_num].inactive.ptr_array;
 884         rs->rules[rs_num].active.rcount =
 885             rs->rules[rs_num].inactive.rcount;
 886         rs->rules[rs_num].inactive.ptr = old_rules;
 887         rs->rules[rs_num].inactive.ptr_array = old_array;
 888         rs->rules[rs_num].inactive.rcount = old_rcount;
 889
 890         rs->rules[rs_num].active.ticket =
 891             rs->rules[rs_num].inactive.ticket;
 892         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 893
 894
 895         /* Purge the old rule list. */
 896         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 897                 pf_unlink_rule(old_rules, rule);
 898         if (rs->rules[rs_num].inactive.ptr_array)
 899                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 900         rs->rules[rs_num].inactive.ptr_array = NULL;
 901         rs->rules[rs_num].inactive.rcount = 0;
 902         rs->rules[rs_num].inactive.open = 0;
 903         pf_remove_if_empty_ruleset(rs);
 904
 905         return (0);
 906 }
 907
 908 static int
 909 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 910 {
 911         MD5_CTX                  ctx;
 912         struct pf_rule          *rule;
 913         int                      rs_cnt;
 914         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 915
 916         MD5Init(&ctx);
 917         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 918                 /* XXX PF_RULESET_SCRUB as well? */
 919                 if (rs_cnt == PF_RULESET_SCRUB)
 920                         continue;
 921
 922                 if (rs->rules[rs_cnt].inactive.ptr_array)
 923                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 924                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 925
 926                 if (rs->rules[rs_cnt].inactive.rcount) {
 927                         rs->rules[rs_cnt].inactive.ptr_array =
 928                             malloc(sizeof(caddr_t) *
 929                             rs->rules[rs_cnt].inactive.rcount,
 930                             M_TEMP, M_NOWAIT);
 931
 932                         if (!rs->rules[rs_cnt].inactive.ptr_array)
 933                                 return (ENOMEM);
 934                 }
 935
 936                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 937                     entries) {
 938                         pf_hash_rule(&ctx, rule);
 939                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 940                 }
 941         }
 942
 943         MD5Final(digest, &ctx);
 944         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
 945         return (0);
 946 }
 947
 948 static int
 949 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 950     sa_family_t af)
 951 {
 952         int error = 0;
 953
 954         switch (addr->type) {
 955         case PF_ADDR_TABLE:
 956                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
 957                 if (addr->p.tbl == NULL)
 958                         error = ENOMEM;
 959                 break;
 960         case PF_ADDR_DYNIFTL:
 961                 error = pfi_dynaddr_setup(addr, af);
 962                 break;
 963         }
 964
 965         return (error);
 966 }
 967
 968 static void
 969 pf_addr_copyout(struct pf_addr_wrap *addr)
 970 {
 971
 972         switch (addr->type) {
 973         case PF_ADDR_DYNIFTL:
 974                 pfi_dynaddr_copyout(addr);
 975                 break;
 976         case PF_ADDR_TABLE:
 977                 pf_tbladdr_copyout(addr);
 978                 break;
 979         }
 980 }
 981
 982 static int
 983 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
 984 {
 985         int                      error = 0;
 986
 987         /* XXX keep in sync with switch() below */
 988         if (securelevel_gt(td->td_ucred, 2))
 989                 switch (cmd) {
 990                 case DIOCGETRULES:
 991                 case DIOCGETRULE:
 992                 case DIOCGETADDRS:
 993                 case DIOCGETADDR:
 994                 case DIOCGETSTATE:
 995                 case DIOCSETSTATUSIF:
 996                 case DIOCGETSTATUS:
 997                 case DIOCCLRSTATUS:
 998                 case DIOCNATLOOK:
 999                 case DIOCSETDEBUG:
1000                 case DIOCGETSTATES:
1001                 case DIOCGETTIMEOUT:
1002                 case DIOCCLRRULECTRS:
1003                 case DIOCGETLIMIT:
1004                 case DIOCGETALTQS:
1005                 case DIOCGETALTQ:
1006                 case DIOCGETQSTATS:
1007                 case DIOCGETRULESETS:
1008                 case DIOCGETRULESET:
1009                 case DIOCRGETTABLES:
1010                 case DIOCRGETTSTATS:
1011                 case DIOCRCLRTSTATS:
1012                 case DIOCRCLRADDRS:
1013                 case DIOCRADDADDRS:
1014                 case DIOCRDELADDRS:
1015                 case DIOCRSETADDRS:
1016                 case DIOCRGETADDRS:
1017                 case DIOCRGETASTATS:
1018                 case DIOCRCLRASTATS:
1019                 case DIOCRTSTADDRS:
1020                 case DIOCOSFPGET:
1021                 case DIOCGETSRCNODES:
1022                 case DIOCCLRSRCNODES:
1023                 case DIOCIGETIFACES:
1024                 case DIOCGIFSPEED:
1025                 case DIOCSETIFFLAG:
1026                 case DIOCCLRIFFLAG:
1027                         break;
1028                 case DIOCRCLRTABLES:
1029                 case DIOCRADDTABLES:
1030                 case DIOCRDELTABLES:
1031                 case DIOCRSETTFLAGS:
1032                         if (((struct pfioc_table *)addr)->pfrio_flags &
1033                             PFR_FLAG_DUMMY)
1034                                 break; /* dummy operation ok */
1035                         return (EPERM);
1036                 default:
1037                         return (EPERM);
1038                 }
1039
1040         if (!(flags & FWRITE))
1041                 switch (cmd) {
1042                 case DIOCGETRULES:
1043                 case DIOCGETADDRS:
1044                 case DIOCGETADDR:
1045                 case DIOCGETSTATE:
1046                 case DIOCGETSTATUS:
1047                 case DIOCGETSTATES:
1048                 case DIOCGETTIMEOUT:
1049                 case DIOCGETLIMIT:
1050                 case DIOCGETALTQS:
1051                 case DIOCGETALTQ:
1052                 case DIOCGETQSTATS:
1053                 case DIOCGETRULESETS:
1054                 case DIOCGETRULESET:
1055                 case DIOCNATLOOK:
1056                 case DIOCRGETTABLES:
1057                 case DIOCRGETTSTATS:
1058                 case DIOCRGETADDRS:
1059                 case DIOCRGETASTATS:
1060                 case DIOCRTSTADDRS:
1061                 case DIOCOSFPGET:
1062                 case DIOCGETSRCNODES:
1063                 case DIOCIGETIFACES:
1064                 case DIOCGIFSPEED:
1065                         break;
1066                 case DIOCRCLRTABLES:
1067                 case DIOCRADDTABLES:
1068                 case DIOCRDELTABLES:
1069                 case DIOCRCLRTSTATS:
1070                 case DIOCRCLRADDRS:
1071                 case DIOCRADDADDRS:
1072                 case DIOCRDELADDRS:
1073                 case DIOCRSETADDRS:
1074                 case DIOCRSETTFLAGS:
1075                         if (((struct pfioc_table *)addr)->pfrio_flags &
1076                             PFR_FLAG_DUMMY) {
1077                                 flags |= FWRITE; /* need write lock for dummy */
1078                                 break; /* dummy operation ok */
1079                         }
1080                         return (EACCES);
1081                 case DIOCGETRULE:
1082                         if (((struct pfioc_rule *)addr)->action ==
1083                             PF_GET_CLR_CNTR)
1084                                 return (EACCES);
1085                         break;
1086                 default:
1087                         return (EACCES);
1088                 }
1089
1090         CURVNET_SET(TD_TO_VNET(td));
1091
1092         switch (cmd) {
1093         case DIOCSTART:
1094                 sx_xlock(&pf_ioctl_lock);
1095                 if (V_pf_status.running)
1096                         error = EEXIST;
1097                 else {
1098                         int cpu;
1099
1100                         error = hook_pf();
1101                         if (error) {
1102                                 DPFPRINTF(PF_DEBUG_MISC,
1103                                     ("pf: pfil registration failed\n"));
1104                                 break;
1105                         }
1106                         V_pf_status.running = 1;
1107                         V_pf_status.since = time_second;
1108
1109                         CPU_FOREACH(cpu)
1110                                 V_pf_stateid[cpu] = time_second;
1111
1112                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1113                 }
1114                 break;
1115
1116         case DIOCSTOP:
1117                 sx_xlock(&pf_ioctl_lock);
1118                 if (!V_pf_status.running)
1119                         error = ENOENT;
1120                 else {
1121                         V_pf_status.running = 0;
1122                         error = dehook_pf();
1123                         if (error) {
1124                                 V_pf_status.running = 1;
1125                                 DPFPRINTF(PF_DEBUG_MISC,
1126                                     ("pf: pfil unregistration failed\n"));
1127                         }
1128                         V_pf_status.since = time_second;
1129                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1130                 }
1131                 break;
1132
1133         case DIOCADDRULE: {
1134                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1135                 struct pf_ruleset       *ruleset;
1136                 struct pf_rule          *rule, *tail;
1137                 struct pf_pooladdr      *pa;
1138                 struct pfi_kif          *kif = NULL;
1139                 int                      rs_num;
1140
1141                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1142                         error = EINVAL;
1143                         break;
1144                 }
1145 #ifndef INET
1146                 if (pr->rule.af == AF_INET) {
1147                         error = EAFNOSUPPORT;
1148                         break;
1149                 }
1150 #endif /* INET */
1151 #ifndef INET6
1152                 if (pr->rule.af == AF_INET6) {
1153                         error = EAFNOSUPPORT;
1154                         break;
1155                 }
1156 #endif /* INET6 */
1157
1158                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1159                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1160                 if (rule->ifname[0])
1161                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1162                 rule->states_cur = counter_u64_alloc(M_WAITOK);
1163                 rule->states_tot = counter_u64_alloc(M_WAITOK);
1164                 rule->src_nodes = counter_u64_alloc(M_WAITOK);
1165                 rule->cuid = td->td_ucred->cr_ruid;
1166                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1167                 TAILQ_INIT(&rule->rpool.list);
1168
1169 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1170
1171                 PF_RULES_WLOCK();
1172                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1173                 ruleset = pf_find_ruleset(pr->anchor);
1174                 if (ruleset == NULL)
1175                         ERROUT(EINVAL);
1176                 rs_num = pf_get_ruleset_number(pr->rule.action);
1177                 if (rs_num >= PF_RULESET_MAX)
1178                         ERROUT(EINVAL);
1179                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1180                         DPFPRINTF(PF_DEBUG_MISC,
1181                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1182                             ruleset->rules[rs_num].inactive.ticket));
1183                         ERROUT(EBUSY);
1184                 }
1185                 if (pr->pool_ticket != V_ticket_pabuf) {
1186                         DPFPRINTF(PF_DEBUG_MISC,
1187                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
1188                             V_ticket_pabuf));
1189                         ERROUT(EBUSY);
1190                 }
1191
1192                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1193                     pf_rulequeue);
1194                 if (tail)
1195                         rule->nr = tail->nr + 1;
1196                 else
1197                         rule->nr = 0;
1198                 if (rule->ifname[0]) {
1199                         rule->kif = pfi_kif_attach(kif, rule->ifname);
1200                         pfi_kif_ref(rule->kif);
1201                 } else
1202                         rule->kif = NULL;
1203
1204                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1205                         error = EBUSY;
1206
1207 #ifdef ALTQ
1208                 /* set queue IDs */
1209                 if (rule->qname[0] != 0) {
1210                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1211                                 error = EBUSY;
1212                         else if (rule->pqname[0] != 0) {
1213                                 if ((rule->pqid =
1214                                     pf_qname2qid(rule->pqname)) == 0)
1215                                         error = EBUSY;
1216                         } else
1217                                 rule->pqid = rule->qid;
1218                 }
1219 #endif
1220                 if (rule->tagname[0])
1221                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1222                                 error = EBUSY;
1223                 if (rule->match_tagname[0])
1224                         if ((rule->match_tag =
1225                             pf_tagname2tag(rule->match_tagname)) == 0)
1226                                 error = EBUSY;
1227                 if (rule->rt && !rule->direction)
1228                         error = EINVAL;
1229                 if (!rule->log)
1230                         rule->logif = 0;
1231                 if (rule->logif >= PFLOGIFS_MAX)
1232                         error = EINVAL;
1233                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1234                         error = ENOMEM;
1235                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1236                         error = ENOMEM;
1237                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1238                         error = EINVAL;
1239                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1240                         if (pa->addr.type == PF_ADDR_TABLE) {
1241                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
1242                                     pa->addr.v.tblname);
1243                                 if (pa->addr.p.tbl == NULL)
1244                                         error = ENOMEM;
1245                         }
1246
1247                 if (rule->overload_tblname[0]) {
1248                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1249                             rule->overload_tblname)) == NULL)
1250                                 error = EINVAL;
1251                         else
1252                                 rule->overload_tbl->pfrkt_flags |=
1253                                     PFR_TFLAG_ACTIVE;
1254                 }
1255
1256                 pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1257                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1258                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1259                     (rule->rt > PF_FASTROUTE)) &&
1260                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1261                         error = EINVAL;
1262
1263                 if (error) {
1264                         pf_free_rule(rule);
1265                         PF_RULES_WUNLOCK();
1266                         break;
1267                 }
1268
1269                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1270                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1271                     rule->bytes[0] = rule->bytes[1] = 0;
1272                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1273                     rule, entries);
1274                 ruleset->rules[rs_num].inactive.rcount++;
1275                 PF_RULES_WUNLOCK();
1276                 break;
1277
1278 #undef ERROUT
1279 DIOCADDRULE_error:
1280                 PF_RULES_WUNLOCK();
1281                 counter_u64_free(rule->states_cur);
1282                 counter_u64_free(rule->states_tot);
1283                 counter_u64_free(rule->src_nodes);
1284                 free(rule, M_PFRULE);
1285                 if (kif)
1286                         free(kif, PFI_MTYPE);
1287                 break;
1288         }
1289
1290         case DIOCGETRULES: {
1291                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1292                 struct pf_ruleset       *ruleset;
1293                 struct pf_rule          *tail;
1294                 int                      rs_num;
1295
1296                 PF_RULES_WLOCK();
1297                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1298                 ruleset = pf_find_ruleset(pr->anchor);
1299                 if (ruleset == NULL) {
1300                         PF_RULES_WUNLOCK();
1301                         error = EINVAL;
1302                         break;
1303                 }
1304                 rs_num = pf_get_ruleset_number(pr->rule.action);
1305                 if (rs_num >= PF_RULESET_MAX) {
1306                         PF_RULES_WUNLOCK();
1307                         error = EINVAL;
1308                         break;
1309                 }
1310                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1311                     pf_rulequeue);
1312                 if (tail)
1313                         pr->nr = tail->nr + 1;
1314                 else
1315                         pr->nr = 0;
1316                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1317                 PF_RULES_WUNLOCK();
1318                 break;
1319         }
1320
1321         case DIOCGETRULE: {
1322                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1323                 struct pf_ruleset       *ruleset;
1324                 struct pf_rule          *rule;
1325                 int                      rs_num, i;
1326
1327                 PF_RULES_WLOCK();
1328                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1329                 ruleset = pf_find_ruleset(pr->anchor);
1330                 if (ruleset == NULL) {
1331                         PF_RULES_WUNLOCK();
1332                         error = EINVAL;
1333                         break;
1334                 }
1335                 rs_num = pf_get_ruleset_number(pr->rule.action);
1336                 if (rs_num >= PF_RULESET_MAX) {
1337                         PF_RULES_WUNLOCK();
1338                         error = EINVAL;
1339                         break;
1340                 }
1341                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1342                         PF_RULES_WUNLOCK();
1343                         error = EBUSY;
1344                         break;
1345                 }
1346                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1347                 while ((rule != NULL) && (rule->nr != pr->nr))
1348                         rule = TAILQ_NEXT(rule, entries);
1349                 if (rule == NULL) {
1350                         PF_RULES_WUNLOCK();
1351                         error = EBUSY;
1352                         break;
1353                 }
1354                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1355                 pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1356                 pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1357                 pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1358                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1359                         PF_RULES_WUNLOCK();
1360                         error = EBUSY;
1361                         break;
1362                 }
1363                 pf_addr_copyout(&pr->rule.src.addr);
1364                 pf_addr_copyout(&pr->rule.dst.addr);
1365                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1366                         if (rule->skip[i].ptr == NULL)
1367                                 pr->rule.skip[i].nr = -1;
1368                         else
1369                                 pr->rule.skip[i].nr =
1370                                     rule->skip[i].ptr->nr;
1371
1372                 if (pr->action == PF_GET_CLR_CNTR) {
1373                         rule->evaluations = 0;
1374                         rule->packets[0] = rule->packets[1] = 0;
1375                         rule->bytes[0] = rule->bytes[1] = 0;
1376                         counter_u64_zero(rule->states_tot);
1377                 }
1378                 PF_RULES_WUNLOCK();
1379                 break;
1380         }
1381
1382         case DIOCCHANGERULE: {
1383                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1384                 struct pf_ruleset       *ruleset;
1385                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1386                 struct pfi_kif          *kif = NULL;
1387                 struct pf_pooladdr      *pa;
1388                 u_int32_t                nr = 0;
1389                 int                      rs_num;
1390
1391                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1392                     pcr->action > PF_CHANGE_GET_TICKET) {
1393                         error = EINVAL;
1394                         break;
1395                 }
1396                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1397                         error = EINVAL;
1398                         break;
1399                 }
1400
1401                 if (pcr->action != PF_CHANGE_REMOVE) {
1402 #ifndef INET
1403                         if (pcr->rule.af == AF_INET) {
1404                                 error = EAFNOSUPPORT;
1405                                 break;
1406                         }
1407 #endif /* INET */
1408 #ifndef INET6
1409                         if (pcr->rule.af == AF_INET6) {
1410                                 error = EAFNOSUPPORT;
1411                                 break;
1412                         }
1413 #endif /* INET6 */
1414                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1415                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1416                         if (newrule->ifname[0])
1417                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1418                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
1419                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
1420                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1421                         newrule->cuid = td->td_ucred->cr_ruid;
1422                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1423                         TAILQ_INIT(&newrule->rpool.list);
1424                 }
1425
1426 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
1427
1428                 PF_RULES_WLOCK();
1429                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1430                     pcr->action == PF_CHANGE_GET_TICKET) &&
1431                     pcr->pool_ticket != V_ticket_pabuf)
1432                         ERROUT(EBUSY);
1433
1434                 ruleset = pf_find_ruleset(pcr->anchor);
1435                 if (ruleset == NULL)
1436                         ERROUT(EINVAL);
1437
1438                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1439                 if (rs_num >= PF_RULESET_MAX)
1440                         ERROUT(EINVAL);
1441
1442                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1443                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1444                         ERROUT(0);
1445                 } else if (pcr->ticket !=
1446                             ruleset->rules[rs_num].active.ticket)
1447                                 ERROUT(EINVAL);
1448
1449                 if (pcr->action != PF_CHANGE_REMOVE) {
1450                         if (newrule->ifname[0]) {
1451                                 newrule->kif = pfi_kif_attach(kif,
1452                                     newrule->ifname);
1453                                 pfi_kif_ref(newrule->kif);
1454                         } else
1455                                 newrule->kif = NULL;
1456
1457                         if (newrule->rtableid > 0 &&
1458                             newrule->rtableid >= rt_numfibs)
1459                                 error = EBUSY;
1460
1461 #ifdef ALTQ
1462                         /* set queue IDs */
1463                         if (newrule->qname[0] != 0) {
1464                                 if ((newrule->qid =
1465                                     pf_qname2qid(newrule->qname)) == 0)
1466                                         error = EBUSY;
1467                                 else if (newrule->pqname[0] != 0) {
1468                                         if ((newrule->pqid =
1469                                             pf_qname2qid(newrule->pqname)) == 0)
1470                                                 error = EBUSY;
1471                                 } else
1472                                         newrule->pqid = newrule->qid;
1473                         }
1474 #endif /* ALTQ */
1475                         if (newrule->tagname[0])
1476                                 if ((newrule->tag =
1477                                     pf_tagname2tag(newrule->tagname)) == 0)
1478                                         error = EBUSY;
1479                         if (newrule->match_tagname[0])
1480                                 if ((newrule->match_tag = pf_tagname2tag(
1481                                     newrule->match_tagname)) == 0)
1482                                         error = EBUSY;
1483                         if (newrule->rt && !newrule->direction)
1484                                 error = EINVAL;
1485                         if (!newrule->log)
1486                                 newrule->logif = 0;
1487                         if (newrule->logif >= PFLOGIFS_MAX)
1488                                 error = EINVAL;
1489                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1490                                 error = ENOMEM;
1491                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1492                                 error = ENOMEM;
1493                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1494                                 error = EINVAL;
1495                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1496                                 if (pa->addr.type == PF_ADDR_TABLE) {
1497                                         pa->addr.p.tbl =
1498                                             pfr_attach_table(ruleset,
1499                                             pa->addr.v.tblname);
1500                                         if (pa->addr.p.tbl == NULL)
1501                                                 error = ENOMEM;
1502                                 }
1503
1504                         if (newrule->overload_tblname[0]) {
1505                                 if ((newrule->overload_tbl = pfr_attach_table(
1506                                     ruleset, newrule->overload_tblname)) ==
1507                                     NULL)
1508                                         error = EINVAL;
1509                                 else
1510                                         newrule->overload_tbl->pfrkt_flags |=
1511                                             PFR_TFLAG_ACTIVE;
1512                         }
1513
1514                         pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1515                         if (((((newrule->action == PF_NAT) ||
1516                             (newrule->action == PF_RDR) ||
1517                             (newrule->action == PF_BINAT) ||
1518                             (newrule->rt > PF_FASTROUTE)) &&
1519                             !newrule->anchor)) &&
1520                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1521                                 error = EINVAL;
1522
1523                         if (error) {
1524                                 pf_free_rule(newrule);
1525                                 PF_RULES_WUNLOCK();
1526                                 break;
1527                         }
1528
1529                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1530                         newrule->evaluations = 0;
1531                         newrule->packets[0] = newrule->packets[1] = 0;
1532                         newrule->bytes[0] = newrule->bytes[1] = 0;
1533                 }
1534                 pf_empty_pool(&V_pf_pabuf);
1535
1536                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1537                         oldrule = TAILQ_FIRST(
1538                             ruleset->rules[rs_num].active.ptr);
1539                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1540                         oldrule = TAILQ_LAST(
1541                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1542                 else {
1543                         oldrule = TAILQ_FIRST(
1544                             ruleset->rules[rs_num].active.ptr);
1545                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1546                                 oldrule = TAILQ_NEXT(oldrule, entries);
1547                         if (oldrule == NULL) {
1548                                 if (newrule != NULL)
1549                                         pf_free_rule(newrule);
1550                                 PF_RULES_WUNLOCK();
1551                                 error = EINVAL;
1552                                 break;
1553                         }
1554                 }
1555
1556                 if (pcr->action == PF_CHANGE_REMOVE) {
1557                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1558                             oldrule);
1559                         ruleset->rules[rs_num].active.rcount--;
1560                 } else {
1561                         if (oldrule == NULL)
1562                                 TAILQ_INSERT_TAIL(
1563                                     ruleset->rules[rs_num].active.ptr,
1564                                     newrule, entries);
1565                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1566                             pcr->action == PF_CHANGE_ADD_BEFORE)
1567                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1568                         else
1569                                 TAILQ_INSERT_AFTER(
1570                                     ruleset->rules[rs_num].active.ptr,
1571                                     oldrule, newrule, entries);
1572                         ruleset->rules[rs_num].active.rcount++;
1573                 }
1574
1575                 nr = 0;
1576                 TAILQ_FOREACH(oldrule,
1577                     ruleset->rules[rs_num].active.ptr, entries)
1578                         oldrule->nr = nr++;
1579
1580                 ruleset->rules[rs_num].active.ticket++;
1581
1582                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1583                 pf_remove_if_empty_ruleset(ruleset);
1584
1585                 PF_RULES_WUNLOCK();
1586                 break;
1587
1588 #undef ERROUT
1589 DIOCCHANGERULE_error:
1590                 PF_RULES_WUNLOCK();
1591                 if (newrule != NULL) {
1592                         counter_u64_free(newrule->states_cur);
1593                         counter_u64_free(newrule->states_tot);
1594                         counter_u64_free(newrule->src_nodes);
1595                         free(newrule, M_PFRULE);
1596                 }
1597                 if (kif != NULL)
1598                         free(kif, PFI_MTYPE);
1599                 break;
1600         }
1601
1602         case DIOCCLRSTATES: {
1603                 struct pf_state         *s;
1604                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1605                 u_int                    i, killed = 0;
1606
1607                 for (i = 0; i <= pf_hashmask; i++) {
1608                         struct pf_idhash *ih = &V_pf_idhash[i];
1609
1610 relock_DIOCCLRSTATES:
1611                         PF_HASHROW_LOCK(ih);
1612                         LIST_FOREACH(s, &ih->states, entry)
1613                                 if (!psk->psk_ifname[0] ||
1614                                     !strcmp(psk->psk_ifname,
1615                                     s->kif->pfik_name)) {
1616                                         /*
1617                                          * Don't send out individual
1618                                          * delete messages.
1619                                          */
1620                                         s->state_flags |= PFSTATE_NOSYNC;
1621                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1622                                         killed++;
1623                                         goto relock_DIOCCLRSTATES;
1624                                 }
1625                         PF_HASHROW_UNLOCK(ih);
1626                 }
1627                 psk->psk_killed = killed;
1628                 if (pfsync_clear_states_ptr != NULL)
1629                         pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1630                 break;
1631         }
1632
1633         case DIOCKILLSTATES: {
1634                 struct pf_state         *s;
1635                 struct pf_state_key     *sk;
1636                 struct pf_addr          *srcaddr, *dstaddr;
1637                 u_int16_t                srcport, dstport;
1638                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1639                 u_int                    i, killed = 0;
1640
1641                 if (psk->psk_pfcmp.id) {
1642                         if (psk->psk_pfcmp.creatorid == 0)
1643                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1644                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1645                             psk->psk_pfcmp.creatorid))) {
1646                                 pf_unlink_state(s, PF_ENTER_LOCKED);
1647                                 psk->psk_killed = 1;
1648                         }
1649                         break;
1650                 }
1651
1652                 for (i = 0; i <= pf_hashmask; i++) {
1653                         struct pf_idhash *ih = &V_pf_idhash[i];
1654
1655 relock_DIOCKILLSTATES:
1656                         PF_HASHROW_LOCK(ih);
1657                         LIST_FOREACH(s, &ih->states, entry) {
1658                                 sk = s->key[PF_SK_WIRE];
1659                                 if (s->direction == PF_OUT) {
1660                                         srcaddr = &sk->addr[1];
1661                                         dstaddr = &sk->addr[0];
1662                                         srcport = sk->port[0];
1663                                         dstport = sk->port[0];
1664                                 } else {
1665                                         srcaddr = &sk->addr[0];
1666                                         dstaddr = &sk->addr[1];
1667                                         srcport = sk->port[0];
1668                                         dstport = sk->port[0];
1669                                 }
1670
1671                                 if ((!psk->psk_af || sk->af == psk->psk_af)
1672                                     && (!psk->psk_proto || psk->psk_proto ==
1673                                     sk->proto) &&
1674                                     PF_MATCHA(psk->psk_src.neg,
1675                                     &psk->psk_src.addr.v.a.addr,
1676                                     &psk->psk_src.addr.v.a.mask,
1677                                     srcaddr, sk->af) &&
1678                                     PF_MATCHA(psk->psk_dst.neg,
1679                                     &psk->psk_dst.addr.v.a.addr,
1680                                     &psk->psk_dst.addr.v.a.mask,
1681                                     dstaddr, sk->af) &&
1682                                     (psk->psk_src.port_op == 0 ||
1683                                     pf_match_port(psk->psk_src.port_op,
1684                                     psk->psk_src.port[0], psk->psk_src.port[1],
1685                                     srcport)) &&
1686                                     (psk->psk_dst.port_op == 0 ||
1687                                     pf_match_port(psk->psk_dst.port_op,
1688                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
1689                                     dstport)) &&
1690                                     (!psk->psk_label[0] ||
1691                                     (s->rule.ptr->label[0] &&
1692                                     !strcmp(psk->psk_label,
1693                                     s->rule.ptr->label))) &&
1694                                     (!psk->psk_ifname[0] ||
1695                                     !strcmp(psk->psk_ifname,
1696                                     s->kif->pfik_name))) {
1697                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1698                                         killed++;
1699                                         goto relock_DIOCKILLSTATES;
1700                                 }
1701                         }
1702                         PF_HASHROW_UNLOCK(ih);
1703                 }
1704                 psk->psk_killed = killed;
1705                 break;
1706         }
1707
1708         case DIOCADDSTATE: {
1709                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1710                 struct pfsync_state     *sp = &ps->state;
1711
1712                 if (sp->timeout >= PFTM_MAX) {
1713                         error = EINVAL;
1714                         break;
1715                 }
1716                 if (pfsync_state_import_ptr != NULL) {
1717                         PF_RULES_RLOCK();
1718                         error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1719                         PF_RULES_RUNLOCK();
1720                 } else
1721                         error = EOPNOTSUPP;
1722                 break;
1723         }
1724
1725         case DIOCGETSTATE: {
1726                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1727                 struct pf_state         *s;
1728
1729                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1730                 if (s == NULL) {
1731                         error = ENOENT;
1732                         break;
1733                 }
1734
1735                 pfsync_state_export(&ps->state, s);
1736                 PF_STATE_UNLOCK(s);
1737                 break;
1738         }
1739
1740         case DIOCGETSTATES: {
1741                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1742                 struct pf_state         *s;
1743                 struct pfsync_state     *pstore, *p;
1744                 int i, nr;
1745
1746                 if (ps->ps_len == 0) {
1747                         nr = uma_zone_get_cur(V_pf_state_z);
1748                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1749                         break;
1750                 }
1751
1752                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1753                 nr = 0;
1754
1755                 for (i = 0; i <= pf_hashmask; i++) {
1756                         struct pf_idhash *ih = &V_pf_idhash[i];
1757
1758                         PF_HASHROW_LOCK(ih);
1759                         LIST_FOREACH(s, &ih->states, entry) {
1760
1761                                 if (s->timeout == PFTM_UNLINKED)
1762                                         continue;
1763
1764                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
1765                                         PF_HASHROW_UNLOCK(ih);
1766                                         goto DIOCGETSTATES_full;
1767                                 }
1768                                 pfsync_state_export(p, s);
1769                                 p++;
1770                                 nr++;
1771                         }
1772                         PF_HASHROW_UNLOCK(ih);
1773                 }
1774 DIOCGETSTATES_full:
1775                 error = copyout(pstore, ps->ps_states,
1776                     sizeof(struct pfsync_state) * nr);
1777                 if (error) {
1778                         free(pstore, M_TEMP);
1779                         break;
1780                 }
1781                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1782                 free(pstore, M_TEMP);
1783
1784                 break;
1785         }
1786
1787         case DIOCGETSTATUS: {
1788                 struct pf_status *s = (struct pf_status *)addr;
1789
1790                 PF_RULES_RLOCK();
1791                 s->running = V_pf_status.running;
1792                 s->since   = V_pf_status.since;
1793                 s->debug   = V_pf_status.debug;
1794                 s->hostid  = V_pf_status.hostid;
1795                 s->states  = V_pf_status.states;
1796                 s->src_nodes = V_pf_status.src_nodes;
1797
1798                 for (int i = 0; i < PFRES_MAX; i++)
1799                         s->counters[i] =
1800                             counter_u64_fetch(V_pf_status.counters[i]);
1801                 for (int i = 0; i < LCNT_MAX; i++)
1802                         s->lcounters[i] =
1803                             counter_u64_fetch(V_pf_status.lcounters[i]);
1804                 for (int i = 0; i < FCNT_MAX; i++)
1805                         s->fcounters[i] =
1806                             counter_u64_fetch(V_pf_status.fcounters[i]);
1807                 for (int i = 0; i < SCNT_MAX; i++)
1808                         s->scounters[i] =
1809                             counter_u64_fetch(V_pf_status.scounters[i]);
1810
1811                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
1812                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
1813                     PF_MD5_DIGEST_LENGTH);
1814
1815                 pfi_update_status(s->ifname, s);
1816                 PF_RULES_RUNLOCK();
1817                 break;
1818         }
1819
1820         case DIOCSETSTATUSIF: {
1821                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1822
1823                 if (pi->ifname[0] == 0) {
1824                         bzero(V_pf_status.ifname, IFNAMSIZ);
1825                         break;
1826                 }
1827                 PF_RULES_WLOCK();
1828                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1829                 PF_RULES_WUNLOCK();
1830                 break;
1831         }
1832
1833         case DIOCCLRSTATUS: {
1834                 PF_RULES_WLOCK();
1835                 for (int i = 0; i < PFRES_MAX; i++)
1836                         counter_u64_zero(V_pf_status.counters[i]);
1837                 for (int i = 0; i < FCNT_MAX; i++)
1838                         counter_u64_zero(V_pf_status.fcounters[i]);
1839                 for (int i = 0; i < SCNT_MAX; i++)
1840                         counter_u64_zero(V_pf_status.scounters[i]);
1841                 V_pf_status.since = time_second;
1842                 if (*V_pf_status.ifname)
1843                         pfi_update_status(V_pf_status.ifname, NULL);
1844                 PF_RULES_WUNLOCK();
1845                 break;
1846         }
1847
1848         case DIOCNATLOOK: {
1849                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1850                 struct pf_state_key     *sk;
1851                 struct pf_state         *state;
1852                 struct pf_state_key_cmp  key;
1853                 int                      m = 0, direction = pnl->direction;
1854                 int                      sidx, didx;
1855
1856                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1857                 sidx = (direction == PF_IN) ? 1 : 0;
1858                 didx = (direction == PF_IN) ? 0 : 1;
1859
1860                 if (!pnl->proto ||
1861                     PF_AZERO(&pnl->saddr, pnl->af) ||
1862                     PF_AZERO(&pnl->daddr, pnl->af) ||
1863                     ((pnl->proto == IPPROTO_TCP ||
1864                     pnl->proto == IPPROTO_UDP) &&
1865                     (!pnl->dport || !pnl->sport)))
1866                         error = EINVAL;
1867                 else {
1868                         bzero(&key, sizeof(key));
1869                         key.af = pnl->af;
1870                         key.proto = pnl->proto;
1871                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1872                         key.port[sidx] = pnl->sport;
1873                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1874                         key.port[didx] = pnl->dport;
1875
1876                         state = pf_find_state_all(&key, direction, &m);
1877
1878                         if (m > 1)
1879                                 error = E2BIG;  /* more than one state */
1880                         else if (state != NULL) {
1881                                 /* XXXGL: not locked read */
1882                                 sk = state->key[sidx];
1883                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1884                                 pnl->rsport = sk->port[sidx];
1885                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1886                                 pnl->rdport = sk->port[didx];
1887                         } else
1888                                 error = ENOENT;
1889                 }
1890                 break;
1891         }
1892
1893         case DIOCSETTIMEOUT: {
1894                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1895                 int              old;
1896
1897                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1898                     pt->seconds < 0) {
1899                         error = EINVAL;
1900                         break;
1901                 }
1902                 PF_RULES_WLOCK();
1903                 old = V_pf_default_rule.timeout[pt->timeout];
1904                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1905                         pt->seconds = 1;
1906                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1907                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1908                         wakeup(pf_purge_thread);
1909                 pt->seconds = old;
1910                 PF_RULES_WUNLOCK();
1911                 break;
1912         }
1913
1914         case DIOCGETTIMEOUT: {
1915                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1916
1917                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1918                         error = EINVAL;
1919                         break;
1920                 }
1921                 PF_RULES_RLOCK();
1922                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1923                 PF_RULES_RUNLOCK();
1924                 break;
1925         }
1926
1927         case DIOCGETLIMIT: {
1928                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1929
1930                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1931                         error = EINVAL;
1932                         break;
1933                 }
1934                 PF_RULES_RLOCK();
1935                 pl->limit = V_pf_limits[pl->index].limit;
1936                 PF_RULES_RUNLOCK();
1937                 break;
1938         }
1939
1940         case DIOCSETLIMIT: {
1941                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1942                 int                      old_limit;
1943
1944                 PF_RULES_WLOCK();
1945                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1946                     V_pf_limits[pl->index].zone == NULL) {
1947                         PF_RULES_WUNLOCK();
1948                         error = EINVAL;
1949                         break;
1950                 }
1951                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1952                 old_limit = V_pf_limits[pl->index].limit;
1953                 V_pf_limits[pl->index].limit = pl->limit;
1954                 pl->limit = old_limit;
1955                 PF_RULES_WUNLOCK();
1956                 break;
1957         }
1958
1959         case DIOCSETDEBUG: {
1960                 u_int32_t       *level = (u_int32_t *)addr;
1961
1962                 PF_RULES_WLOCK();
1963                 V_pf_status.debug = *level;
1964                 PF_RULES_WUNLOCK();
1965                 break;
1966         }
1967
1968         case DIOCCLRRULECTRS: {
1969                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1970                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1971                 struct pf_rule          *rule;
1972
1973                 PF_RULES_WLOCK();
1974                 TAILQ_FOREACH(rule,
1975                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1976                         rule->evaluations = 0;
1977                         rule->packets[0] = rule->packets[1] = 0;
1978                         rule->bytes[0] = rule->bytes[1] = 0;
1979                 }
1980                 PF_RULES_WUNLOCK();
1981                 break;
1982         }
1983
1984         case DIOCGIFSPEED: {
1985                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1986                 struct pf_ifspeed       ps;
1987                 struct ifnet            *ifp;
1988
1989                 if (psp->ifname[0] != 0) {
1990                         /* Can we completely trust user-land? */
1991                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1992                         ifp = ifunit(ps.ifname);
1993                         if (ifp != NULL)
1994                                 psp->baudrate = ifp->if_baudrate;
1995                         else
1996                                 error = EINVAL;
1997                 } else
1998                         error = EINVAL;
1999                 break;
2000         }
2001
2002 #ifdef ALTQ
2003         case DIOCSTARTALTQ: {
2004                 struct pf_altq          *altq;
2005
2006                 PF_RULES_WLOCK();
2007                 /* enable all altq interfaces on active list */
2008                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2009                         if (altq->qname[0] == 0 && (altq->local_flags &
2010                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2011                                 error = pf_enable_altq(altq);
2012                                 if (error != 0)
2013                                         break;
2014                         }
2015                 }
2016                 if (error == 0)
2017                         V_pf_altq_running = 1;
2018                 PF_RULES_WUNLOCK();
2019                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2020                 break;
2021         }
2022
2023         case DIOCSTOPALTQ: {
2024                 struct pf_altq          *altq;
2025
2026                 PF_RULES_WLOCK();
2027                 /* disable all altq interfaces on active list */
2028                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2029                         if (altq->qname[0] == 0 && (altq->local_flags &
2030                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2031                                 error = pf_disable_altq(altq);
2032                                 if (error != 0)
2033                                         break;
2034                         }
2035                 }
2036                 if (error == 0)
2037                         V_pf_altq_running = 0;
2038                 PF_RULES_WUNLOCK();
2039                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2040                 break;
2041         }
2042
2043         case DIOCADDALTQ: {
2044                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2045                 struct pf_altq          *altq, *a;
2046                 struct ifnet            *ifp;
2047
2048                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2049                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2050                 altq->local_flags = 0;
2051
2052                 PF_RULES_WLOCK();
2053                 if (pa->ticket != V_ticket_altqs_inactive) {
2054                         PF_RULES_WUNLOCK();
2055                         free(altq, M_PFALTQ);
2056                         error = EBUSY;
2057                         break;
2058                 }
2059
2060                 /*
2061                  * if this is for a queue, find the discipline and
2062                  * copy the necessary fields
2063                  */
2064                 if (altq->qname[0] != 0) {
2065                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2066                                 PF_RULES_WUNLOCK();
2067                                 error = EBUSY;
2068                                 free(altq, M_PFALTQ);
2069                                 break;
2070                         }
2071                         altq->altq_disc = NULL;
2072                         TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2073                                 if (strncmp(a->ifname, altq->ifname,
2074                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
2075                                         altq->altq_disc = a->altq_disc;
2076                                         break;
2077                                 }
2078                         }
2079                 }
2080
2081                 if ((ifp = ifunit(altq->ifname)) == NULL)
2082                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2083                 else
2084                         error = altq_add(altq);
2085
2086                 if (error) {
2087                         PF_RULES_WUNLOCK();
2088                         free(altq, M_PFALTQ);
2089                         break;
2090                 }
2091
2092                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2093                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2094                 PF_RULES_WUNLOCK();
2095                 break;
2096         }
2097
2098         case DIOCGETALTQS: {
2099                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2100                 struct pf_altq          *altq;
2101
2102                 PF_RULES_RLOCK();
2103                 pa->nr = 0;
2104                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2105                         pa->nr++;
2106                 pa->ticket = V_ticket_altqs_active;
2107                 PF_RULES_RUNLOCK();
2108                 break;
2109         }
2110
2111         case DIOCGETALTQ: {
2112                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2113                 struct pf_altq          *altq;
2114                 u_int32_t                nr;
2115
2116                 PF_RULES_RLOCK();
2117                 if (pa->ticket != V_ticket_altqs_active) {
2118                         PF_RULES_RUNLOCK();
2119                         error = EBUSY;
2120                         break;
2121                 }
2122                 nr = 0;
2123                 altq = TAILQ_FIRST(V_pf_altqs_active);
2124                 while ((altq != NULL) && (nr < pa->nr)) {
2125                         altq = TAILQ_NEXT(altq, entries);
2126                         nr++;
2127                 }
2128                 if (altq == NULL) {
2129                         PF_RULES_RUNLOCK();
2130                         error = EBUSY;
2131                         break;
2132                 }
2133                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2134                 PF_RULES_RUNLOCK();
2135                 break;
2136         }
2137
2138         case DIOCCHANGEALTQ:
2139                 /* CHANGEALTQ not supported yet! */
2140                 error = ENODEV;
2141                 break;
2142
2143         case DIOCGETQSTATS: {
2144                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
2145                 struct pf_altq          *altq;
2146                 u_int32_t                nr;
2147                 int                      nbytes;
2148
2149                 PF_RULES_RLOCK();
2150                 if (pq->ticket != V_ticket_altqs_active) {
2151                         PF_RULES_RUNLOCK();
2152                         error = EBUSY;
2153                         break;
2154                 }
2155                 nbytes = pq->nbytes;
2156                 nr = 0;
2157                 altq = TAILQ_FIRST(V_pf_altqs_active);
2158                 while ((altq != NULL) && (nr < pq->nr)) {
2159                         altq = TAILQ_NEXT(altq, entries);
2160                         nr++;
2161                 }
2162                 if (altq == NULL) {
2163                         PF_RULES_RUNLOCK();
2164                         error = EBUSY;
2165                         break;
2166                 }
2167
2168                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2169                         PF_RULES_RUNLOCK();
2170                         error = ENXIO;
2171                         break;
2172                 }
2173                 PF_RULES_RUNLOCK();
2174                 error = altq_getqstats(altq, pq->buf, &nbytes);
2175                 if (error == 0) {
2176                         pq->scheduler = altq->scheduler;
2177                         pq->nbytes = nbytes;
2178                 }
2179                 break;
2180         }
2181 #endif /* ALTQ */
2182
2183         case DIOCBEGINADDRS: {
2184                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2185
2186                 PF_RULES_WLOCK();
2187                 pf_empty_pool(&V_pf_pabuf);
2188                 pp->ticket = ++V_ticket_pabuf;
2189                 PF_RULES_WUNLOCK();
2190                 break;
2191         }
2192
2193         case DIOCADDADDR: {
2194                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2195                 struct pf_pooladdr      *pa;
2196                 struct pfi_kif          *kif = NULL;
2197
2198 #ifndef INET
2199                 if (pp->af == AF_INET) {
2200                         error = EAFNOSUPPORT;
2201                         break;
2202                 }
2203 #endif /* INET */
2204 #ifndef INET6
2205                 if (pp->af == AF_INET6) {
2206                         error = EAFNOSUPPORT;
2207                         break;
2208                 }
2209 #endif /* INET6 */
2210                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2211                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2212                     pp->addr.addr.type != PF_ADDR_TABLE) {
2213                         error = EINVAL;
2214                         break;
2215                 }
2216                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2217                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2218                 if (pa->ifname[0])
2219                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2220                 PF_RULES_WLOCK();
2221                 if (pp->ticket != V_ticket_pabuf) {
2222                         PF_RULES_WUNLOCK();
2223                         if (pa->ifname[0])
2224                                 free(kif, PFI_MTYPE);
2225                         free(pa, M_PFRULE);
2226                         error = EBUSY;
2227                         break;
2228                 }
2229                 if (pa->ifname[0]) {
2230                         pa->kif = pfi_kif_attach(kif, pa->ifname);
2231                         pfi_kif_ref(pa->kif);
2232                 } else
2233                         pa->kif = NULL;
2234                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2235                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2236                         if (pa->ifname[0])
2237                                 pfi_kif_unref(pa->kif);
2238                         PF_RULES_WUNLOCK();
2239                         free(pa, M_PFRULE);
2240                         break;
2241                 }
2242                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2243                 PF_RULES_WUNLOCK();
2244                 break;
2245         }
2246
2247         case DIOCGETADDRS: {
2248                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2249                 struct pf_pool          *pool;
2250                 struct pf_pooladdr      *pa;
2251
2252                 PF_RULES_RLOCK();
2253                 pp->nr = 0;
2254                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2255                     pp->r_num, 0, 1, 0);
2256                 if (pool == NULL) {
2257                         PF_RULES_RUNLOCK();
2258                         error = EBUSY;
2259                         break;
2260                 }
2261                 TAILQ_FOREACH(pa, &pool->list, entries)
2262                         pp->nr++;
2263                 PF_RULES_RUNLOCK();
2264                 break;
2265         }
2266
2267         case DIOCGETADDR: {
2268                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2269                 struct pf_pool          *pool;
2270                 struct pf_pooladdr      *pa;
2271                 u_int32_t                nr = 0;
2272
2273                 PF_RULES_RLOCK();
2274                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2275                     pp->r_num, 0, 1, 1);
2276                 if (pool == NULL) {
2277                         PF_RULES_RUNLOCK();
2278                         error = EBUSY;
2279                         break;
2280                 }
2281                 pa = TAILQ_FIRST(&pool->list);
2282                 while ((pa != NULL) && (nr < pp->nr)) {
2283                         pa = TAILQ_NEXT(pa, entries);
2284                         nr++;
2285                 }
2286                 if (pa == NULL) {
2287                         PF_RULES_RUNLOCK();
2288                         error = EBUSY;
2289                         break;
2290                 }
2291                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2292                 pf_addr_copyout(&pp->addr.addr);
2293                 PF_RULES_RUNLOCK();
2294                 break;
2295         }
2296
2297         case DIOCCHANGEADDR: {
2298                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2299                 struct pf_pool          *pool;
2300                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2301                 struct pf_ruleset       *ruleset;
2302                 struct pfi_kif          *kif = NULL;
2303
2304                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2305                     pca->action > PF_CHANGE_REMOVE) {
2306                         error = EINVAL;
2307                         break;
2308                 }
2309                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2310                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2311                     pca->addr.addr.type != PF_ADDR_TABLE) {
2312                         error = EINVAL;
2313                         break;
2314                 }
2315
2316                 if (pca->action != PF_CHANGE_REMOVE) {
2317 #ifndef INET
2318                         if (pca->af == AF_INET) {
2319                                 error = EAFNOSUPPORT;
2320                                 break;
2321                         }
2322 #endif /* INET */
2323 #ifndef INET6
2324                         if (pca->af == AF_INET6) {
2325                                 error = EAFNOSUPPORT;
2326                                 break;
2327                         }
2328 #endif /* INET6 */
2329                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2330                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2331                         if (newpa->ifname[0])
2332                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2333                         newpa->kif = NULL;
2334                 }
2335
2336 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
2337                 PF_RULES_WLOCK();
2338                 ruleset = pf_find_ruleset(pca->anchor);
2339                 if (ruleset == NULL)
2340                         ERROUT(EBUSY);
2341
2342                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2343                     pca->r_num, pca->r_last, 1, 1);
2344                 if (pool == NULL)
2345                         ERROUT(EBUSY);
2346
2347                 if (pca->action != PF_CHANGE_REMOVE) {
2348                         if (newpa->ifname[0]) {
2349                                 newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2350                                 pfi_kif_ref(newpa->kif);
2351                                 kif = NULL;
2352                         }
2353
2354                         switch (newpa->addr.type) {
2355                         case PF_ADDR_DYNIFTL:
2356                                 error = pfi_dynaddr_setup(&newpa->addr,
2357                                     pca->af);
2358                                 break;
2359                         case PF_ADDR_TABLE:
2360                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
2361                                     newpa->addr.v.tblname);
2362                                 if (newpa->addr.p.tbl == NULL)
2363                                         error = ENOMEM;
2364                                 break;
2365                         }
2366                         if (error)
2367                                 goto DIOCCHANGEADDR_error;
2368                 }
2369
2370                 switch (pca->action) {
2371                 case PF_CHANGE_ADD_HEAD:
2372                         oldpa = TAILQ_FIRST(&pool->list);
2373                         break;
2374                 case PF_CHANGE_ADD_TAIL:
2375                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2376                         break;
2377                 default:
2378                         oldpa = TAILQ_FIRST(&pool->list);
2379                         for (int i = 0; oldpa && i < pca->nr; i++)
2380                                 oldpa = TAILQ_NEXT(oldpa, entries);
2381
2382                         if (oldpa == NULL)
2383                                 ERROUT(EINVAL);
2384                 }
2385
2386                 if (pca->action == PF_CHANGE_REMOVE) {
2387                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2388                         switch (oldpa->addr.type) {
2389                         case PF_ADDR_DYNIFTL:
2390                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
2391                                 break;
2392                         case PF_ADDR_TABLE:
2393                                 pfr_detach_table(oldpa->addr.p.tbl);
2394                                 break;
2395                         }
2396                         if (oldpa->kif)
2397                                 pfi_kif_unref(oldpa->kif);
2398                         free(oldpa, M_PFRULE);
2399                 } else {
2400                         if (oldpa == NULL)
2401                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2402                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2403                             pca->action == PF_CHANGE_ADD_BEFORE)
2404                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2405                         else
2406                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2407                                     newpa, entries);
2408                 }
2409
2410                 pool->cur = TAILQ_FIRST(&pool->list);
2411                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2412                 PF_RULES_WUNLOCK();
2413                 break;
2414
2415 #undef ERROUT
2416 DIOCCHANGEADDR_error:
2417                 if (newpa->kif)
2418                         pfi_kif_unref(newpa->kif);
2419                 PF_RULES_WUNLOCK();
2420                 if (newpa != NULL)
2421                         free(newpa, M_PFRULE);
2422                 if (kif != NULL)
2423                         free(kif, PFI_MTYPE);
2424                 break;
2425         }
2426
2427         case DIOCGETRULESETS: {
2428                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2429                 struct pf_ruleset       *ruleset;
2430                 struct pf_anchor        *anchor;
2431
2432                 PF_RULES_RLOCK();
2433                 pr->path[sizeof(pr->path) - 1] = 0;
2434                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2435                         PF_RULES_RUNLOCK();
2436                         error = ENOENT;
2437                         break;
2438                 }
2439                 pr->nr = 0;
2440                 if (ruleset->anchor == NULL) {
2441                         /* XXX kludge for pf_main_ruleset */
2442                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2443                                 if (anchor->parent == NULL)
2444                                         pr->nr++;
2445                 } else {
2446                         RB_FOREACH(anchor, pf_anchor_node,
2447                             &ruleset->anchor->children)
2448                                 pr->nr++;
2449                 }
2450                 PF_RULES_RUNLOCK();
2451                 break;
2452         }
2453
2454         case DIOCGETRULESET: {
2455                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2456                 struct pf_ruleset       *ruleset;
2457                 struct pf_anchor        *anchor;
2458                 u_int32_t                nr = 0;
2459
2460                 PF_RULES_RLOCK();
2461                 pr->path[sizeof(pr->path) - 1] = 0;
2462                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2463                         PF_RULES_RUNLOCK();
2464                         error = ENOENT;
2465                         break;
2466                 }
2467                 pr->name[0] = 0;
2468                 if (ruleset->anchor == NULL) {
2469                         /* XXX kludge for pf_main_ruleset */
2470                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2471                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2472                                         strlcpy(pr->name, anchor->name,
2473                                             sizeof(pr->name));
2474                                         break;
2475                                 }
2476                 } else {
2477                         RB_FOREACH(anchor, pf_anchor_node,
2478                             &ruleset->anchor->children)
2479                                 if (nr++ == pr->nr) {
2480                                         strlcpy(pr->name, anchor->name,
2481                                             sizeof(pr->name));
2482                                         break;
2483                                 }
2484                 }
2485                 if (!pr->name[0])
2486                         error = EBUSY;
2487                 PF_RULES_RUNLOCK();
2488                 break;
2489         }
2490
2491         case DIOCRCLRTABLES: {
2492                 struct pfioc_table *io = (struct pfioc_table *)addr;
2493
2494                 if (io->pfrio_esize != 0) {
2495                         error = ENODEV;
2496                         break;
2497                 }
2498                 PF_RULES_WLOCK();
2499                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2500                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2501                 PF_RULES_WUNLOCK();
2502                 break;
2503         }
2504
2505         case DIOCRADDTABLES: {
2506                 struct pfioc_table *io = (struct pfioc_table *)addr;
2507                 struct pfr_table *pfrts;
2508                 size_t totlen;
2509
2510                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2511                         error = ENODEV;
2512                         break;
2513                 }
2514                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2515                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2516                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2517                 if (error) {
2518                         free(pfrts, M_TEMP);
2519                         break;
2520                 }
2521                 PF_RULES_WLOCK();
2522                 error = pfr_add_tables(pfrts, io->pfrio_size,
2523                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2524                 PF_RULES_WUNLOCK();
2525                 free(pfrts, M_TEMP);
2526                 break;
2527         }
2528
2529         case DIOCRDELTABLES: {
2530                 struct pfioc_table *io = (struct pfioc_table *)addr;
2531                 struct pfr_table *pfrts;
2532                 size_t totlen;
2533
2534                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2535                         error = ENODEV;
2536                         break;
2537                 }
2538                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2539                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2540                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2541                 if (error) {
2542                         free(pfrts, M_TEMP);
2543                         break;
2544                 }
2545                 PF_RULES_WLOCK();
2546                 error = pfr_del_tables(pfrts, io->pfrio_size,
2547                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2548                 PF_RULES_WUNLOCK();
2549                 free(pfrts, M_TEMP);
2550                 break;
2551         }
2552
2553         case DIOCRGETTABLES: {
2554                 struct pfioc_table *io = (struct pfioc_table *)addr;
2555                 struct pfr_table *pfrts;
2556                 size_t totlen;
2557
2558                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2559                         error = ENODEV;
2560                         break;
2561                 }
2562                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2563                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2564                 PF_RULES_RLOCK();
2565                 error = pfr_get_tables(&io->pfrio_table, pfrts,
2566                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2567                 PF_RULES_RUNLOCK();
2568                 if (error == 0)
2569                         error = copyout(pfrts, io->pfrio_buffer, totlen);
2570                 free(pfrts, M_TEMP);
2571                 break;
2572         }
2573
2574         case DIOCRGETTSTATS: {
2575                 struct pfioc_table *io = (struct pfioc_table *)addr;
2576                 struct pfr_tstats *pfrtstats;
2577                 size_t totlen;
2578
2579                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2580                         error = ENODEV;
2581                         break;
2582                 }
2583                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2584                 pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2585                 PF_RULES_WLOCK();
2586                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2587                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2588                 PF_RULES_WUNLOCK();
2589                 if (error == 0)
2590                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2591                 free(pfrtstats, M_TEMP);
2592                 break;
2593         }
2594
2595         case DIOCRCLRTSTATS: {
2596                 struct pfioc_table *io = (struct pfioc_table *)addr;
2597                 struct pfr_table *pfrts;
2598                 size_t totlen;
2599
2600                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2601                         error = ENODEV;
2602                         break;
2603                 }
2604                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2605                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2606                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2607                 if (error) {
2608                         free(pfrts, M_TEMP);
2609                         break;
2610                 }
2611                 PF_RULES_WLOCK();
2612                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
2613                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2614                 PF_RULES_WUNLOCK();
2615                 free(pfrts, M_TEMP);
2616                 break;
2617         }
2618
2619         case DIOCRSETTFLAGS: {
2620                 struct pfioc_table *io = (struct pfioc_table *)addr;
2621                 struct pfr_table *pfrts;
2622                 size_t totlen;
2623
2624                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2625                         error = ENODEV;
2626                         break;
2627                 }
2628                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2629                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2630                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2631                 if (error) {
2632                         free(pfrts, M_TEMP);
2633                         break;
2634                 }
2635                 PF_RULES_WLOCK();
2636                 error = pfr_set_tflags(pfrts, io->pfrio_size,
2637                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2638                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2639                 PF_RULES_WUNLOCK();
2640                 free(pfrts, M_TEMP);
2641                 break;
2642         }
2643
2644         case DIOCRCLRADDRS: {
2645                 struct pfioc_table *io = (struct pfioc_table *)addr;
2646
2647                 if (io->pfrio_esize != 0) {
2648                         error = ENODEV;
2649                         break;
2650                 }
2651                 PF_RULES_WLOCK();
2652                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2653                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2654                 PF_RULES_WUNLOCK();
2655                 break;
2656         }
2657
2658         case DIOCRADDADDRS: {
2659                 struct pfioc_table *io = (struct pfioc_table *)addr;
2660                 struct pfr_addr *pfras;
2661                 size_t totlen;
2662
2663                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2664                         error = ENODEV;
2665                         break;
2666                 }
2667                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2668                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2669                 error = copyin(io->pfrio_buffer, pfras, totlen);
2670                 if (error) {
2671                         free(pfras, M_TEMP);
2672                         break;
2673                 }
2674                 PF_RULES_WLOCK();
2675                 error = pfr_add_addrs(&io->pfrio_table, pfras,
2676                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2677                     PFR_FLAG_USERIOCTL);
2678                 PF_RULES_WUNLOCK();
2679                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2680                         error = copyout(pfras, io->pfrio_buffer, totlen);
2681                 free(pfras, M_TEMP);
2682                 break;
2683         }
2684
2685         case DIOCRDELADDRS: {
2686                 struct pfioc_table *io = (struct pfioc_table *)addr;
2687                 struct pfr_addr *pfras;
2688                 size_t totlen;
2689
2690                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2691                         error = ENODEV;
2692                         break;
2693                 }
2694                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2695                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2696                 error = copyin(io->pfrio_buffer, pfras, totlen);
2697                 if (error) {
2698                         free(pfras, M_TEMP);
2699                         break;
2700                 }
2701                 PF_RULES_WLOCK();
2702                 error = pfr_del_addrs(&io->pfrio_table, pfras,
2703                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2704                     PFR_FLAG_USERIOCTL);
2705                 PF_RULES_WUNLOCK();
2706                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2707                         error = copyout(pfras, io->pfrio_buffer, totlen);
2708                 free(pfras, M_TEMP);
2709                 break;
2710         }
2711
2712         case DIOCRSETADDRS: {
2713                 struct pfioc_table *io = (struct pfioc_table *)addr;
2714                 struct pfr_addr *pfras;
2715                 size_t totlen;
2716
2717                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2718                         error = ENODEV;
2719                         break;
2720                 }
2721                 totlen = (io->pfrio_size + io->pfrio_size2) *
2722                     sizeof(struct pfr_addr);
2723                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2724                 error = copyin(io->pfrio_buffer, pfras, totlen);
2725                 if (error) {
2726                         free(pfras, M_TEMP);
2727                         break;
2728                 }
2729                 PF_RULES_WLOCK();
2730                 error = pfr_set_addrs(&io->pfrio_table, pfras,
2731                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2732                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2733                     PFR_FLAG_USERIOCTL, 0);
2734                 PF_RULES_WUNLOCK();
2735                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2736                         error = copyout(pfras, io->pfrio_buffer, totlen);
2737                 free(pfras, M_TEMP);
2738                 break;
2739         }
2740
2741         case DIOCRGETADDRS: {
2742                 struct pfioc_table *io = (struct pfioc_table *)addr;
2743                 struct pfr_addr *pfras;
2744                 size_t totlen;
2745
2746                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2747                         error = ENODEV;
2748                         break;
2749                 }
2750                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2751                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2752                 PF_RULES_RLOCK();
2753                 error = pfr_get_addrs(&io->pfrio_table, pfras,
2754                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2755                 PF_RULES_RUNLOCK();
2756                 if (error == 0)
2757                         error = copyout(pfras, io->pfrio_buffer, totlen);
2758                 free(pfras, M_TEMP);
2759                 break;
2760         }
2761
2762         case DIOCRGETASTATS: {
2763                 struct pfioc_table *io = (struct pfioc_table *)addr;
2764                 struct pfr_astats *pfrastats;
2765                 size_t totlen;
2766
2767                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2768                         error = ENODEV;
2769                         break;
2770                 }
2771                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
2772                 pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2773                 PF_RULES_RLOCK();
2774                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
2775                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2776                 PF_RULES_RUNLOCK();
2777                 if (error == 0)
2778                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
2779                 free(pfrastats, M_TEMP);
2780                 break;
2781         }
2782
2783         case DIOCRCLRASTATS: {
2784                 struct pfioc_table *io = (struct pfioc_table *)addr;
2785                 struct pfr_addr *pfras;
2786                 size_t totlen;
2787
2788                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2789                         error = ENODEV;
2790                         break;
2791                 }
2792                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2793                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2794                 error = copyin(io->pfrio_buffer, pfras, totlen);
2795                 if (error) {
2796                         free(pfras, M_TEMP);
2797                         break;
2798                 }
2799                 PF_RULES_WLOCK();
2800                 error = pfr_clr_astats(&io->pfrio_table, pfras,
2801                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2802                     PFR_FLAG_USERIOCTL);
2803                 PF_RULES_WUNLOCK();
2804                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2805                         error = copyout(pfras, io->pfrio_buffer, totlen);
2806                 free(pfras, M_TEMP);
2807                 break;
2808         }
2809
2810         case DIOCRTSTADDRS: {
2811                 struct pfioc_table *io = (struct pfioc_table *)addr;
2812                 struct pfr_addr *pfras;
2813                 size_t totlen;
2814
2815                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2816                         error = ENODEV;
2817                         break;
2818                 }
2819                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2820                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2821                 error = copyin(io->pfrio_buffer, pfras, totlen);
2822                 if (error) {
2823                         free(pfras, M_TEMP);
2824                         break;
2825                 }
2826                 PF_RULES_RLOCK();
2827                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
2828                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2829                     PFR_FLAG_USERIOCTL);
2830                 PF_RULES_RUNLOCK();
2831                 if (error == 0)
2832                         error = copyout(pfras, io->pfrio_buffer, totlen);
2833                 free(pfras, M_TEMP);
2834                 break;
2835         }
2836
2837         case DIOCRINADEFINE: {
2838                 struct pfioc_table *io = (struct pfioc_table *)addr;
2839                 struct pfr_addr *pfras;
2840                 size_t totlen;
2841
2842                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2843                         error = ENODEV;
2844                         break;
2845                 }
2846                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2847                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2848                 error = copyin(io->pfrio_buffer, pfras, totlen);
2849                 if (error) {
2850                         free(pfras, M_TEMP);
2851                         break;
2852                 }
2853                 PF_RULES_WLOCK();
2854                 error = pfr_ina_define(&io->pfrio_table, pfras,
2855                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2856                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2857                 PF_RULES_WUNLOCK();
2858                 free(pfras, M_TEMP);
2859                 break;
2860         }
2861
2862         case DIOCOSFPADD: {
2863                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2864                 PF_RULES_WLOCK();
2865                 error = pf_osfp_add(io);
2866                 PF_RULES_WUNLOCK();
2867                 break;
2868         }
2869
2870         case DIOCOSFPGET: {
2871                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2872                 PF_RULES_RLOCK();
2873                 error = pf_osfp_get(io);
2874                 PF_RULES_RUNLOCK();
2875                 break;
2876         }
2877
2878         case DIOCXBEGIN: {
2879                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2880                 struct pfioc_trans_e    *ioes, *ioe;
2881                 size_t                   totlen;
2882                 int                      i;
2883
2884                 if (io->esize != sizeof(*ioe)) {
2885                         error = ENODEV;
2886                         break;
2887                 }
2888                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2889                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2890                 error = copyin(io->array, ioes, totlen);
2891                 if (error) {
2892                         free(ioes, M_TEMP);
2893                         break;
2894                 }
2895                 PF_RULES_WLOCK();
2896                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2897                         switch (ioe->rs_num) {
2898 #ifdef ALTQ
2899                         case PF_RULESET_ALTQ:
2900                                 if (ioe->anchor[0]) {
2901                                         PF_RULES_WUNLOCK();
2902                                         free(ioes, M_TEMP);
2903                                         error = EINVAL;
2904                                         goto fail;
2905                                 }
2906                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2907                                         PF_RULES_WUNLOCK();
2908                                         free(ioes, M_TEMP);
2909                                         goto fail;
2910                                 }
2911                                 break;
2912 #endif /* ALTQ */
2913                         case PF_RULESET_TABLE:
2914                             {
2915                                 struct pfr_table table;
2916
2917                                 bzero(&table, sizeof(table));
2918                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2919                                     sizeof(table.pfrt_anchor));
2920                                 if ((error = pfr_ina_begin(&table,
2921                                     &ioe->ticket, NULL, 0))) {
2922                                         PF_RULES_WUNLOCK();
2923                                         free(ioes, M_TEMP);
2924                                         goto fail;
2925                                 }
2926                                 break;
2927                             }
2928                         default:
2929                                 if ((error = pf_begin_rules(&ioe->ticket,
2930                                     ioe->rs_num, ioe->anchor))) {
2931                                         PF_RULES_WUNLOCK();
2932                                         free(ioes, M_TEMP);
2933                                         goto fail;
2934                                 }
2935                                 break;
2936                         }
2937                 }
2938                 PF_RULES_WUNLOCK();
2939                 error = copyout(ioes, io->array, totlen);
2940                 free(ioes, M_TEMP);
2941                 break;
2942         }
2943
2944         case DIOCXROLLBACK: {
2945                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2946                 struct pfioc_trans_e    *ioe, *ioes;
2947                 size_t                   totlen;
2948                 int                      i;
2949
2950                 if (io->esize != sizeof(*ioe)) {
2951                         error = ENODEV;
2952                         break;
2953                 }
2954                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2955                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2956                 error = copyin(io->array, ioes, totlen);
2957                 if (error) {
2958                         free(ioes, M_TEMP);
2959                         break;
2960                 }
2961                 PF_RULES_WLOCK();
2962                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2963                         switch (ioe->rs_num) {
2964 #ifdef ALTQ
2965                         case PF_RULESET_ALTQ:
2966                                 if (ioe->anchor[0]) {
2967                                         PF_RULES_WUNLOCK();
2968                                         free(ioes, M_TEMP);
2969                                         error = EINVAL;
2970                                         goto fail;
2971                                 }
2972                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2973                                         PF_RULES_WUNLOCK();
2974                                         free(ioes, M_TEMP);
2975                                         goto fail; /* really bad */
2976                                 }
2977                                 break;
2978 #endif /* ALTQ */
2979                         case PF_RULESET_TABLE:
2980                             {
2981                                 struct pfr_table table;
2982
2983                                 bzero(&table, sizeof(table));
2984                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2985                                     sizeof(table.pfrt_anchor));
2986                                 if ((error = pfr_ina_rollback(&table,
2987                                     ioe->ticket, NULL, 0))) {
2988                                         PF_RULES_WUNLOCK();
2989                                         free(ioes, M_TEMP);
2990                                         goto fail; /* really bad */
2991                                 }
2992                                 break;
2993                             }
2994                         default:
2995                                 if ((error = pf_rollback_rules(ioe->ticket,
2996                                     ioe->rs_num, ioe->anchor))) {
2997                                         PF_RULES_WUNLOCK();
2998                                         free(ioes, M_TEMP);
2999                                         goto fail; /* really bad */
3000                                 }
3001                                 break;
3002                         }
3003                 }
3004                 PF_RULES_WUNLOCK();
3005                 free(ioes, M_TEMP);
3006                 break;
3007         }
3008
3009         case DIOCXCOMMIT: {
3010                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3011                 struct pfioc_trans_e    *ioe, *ioes;
3012                 struct pf_ruleset       *rs;
3013                 size_t                   totlen;
3014                 int                      i;
3015
3016                 if (io->esize != sizeof(*ioe)) {
3017                         error = ENODEV;
3018                         break;
3019                 }
3020                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3021                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
3022                 error = copyin(io->array, ioes, totlen);
3023                 if (error) {
3024                         free(ioes, M_TEMP);
3025                         break;
3026                 }
3027                 PF_RULES_WLOCK();
3028                 /* First makes sure everything will succeed. */
3029                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3030                         switch (ioe->rs_num) {
3031 #ifdef ALTQ
3032                         case PF_RULESET_ALTQ:
3033                                 if (ioe->anchor[0]) {
3034                                         PF_RULES_WUNLOCK();
3035                                         free(ioes, M_TEMP);
3036                                         error = EINVAL;
3037                                         goto fail;
3038                                 }
3039                                 if (!V_altqs_inactive_open || ioe->ticket !=
3040                                     V_ticket_altqs_inactive) {
3041                                         PF_RULES_WUNLOCK();
3042                                         free(ioes, M_TEMP);
3043                                         error = EBUSY;
3044                                         goto fail;
3045                                 }
3046                                 break;
3047 #endif /* ALTQ */
3048                         case PF_RULESET_TABLE:
3049                                 rs = pf_find_ruleset(ioe->anchor);
3050                                 if (rs == NULL || !rs->topen || ioe->ticket !=
3051                                     rs->tticket) {
3052                                         PF_RULES_WUNLOCK();
3053                                         free(ioes, M_TEMP);
3054                                         error = EBUSY;
3055                                         goto fail;
3056                                 }
3057                                 break;
3058                         default:
3059                                 if (ioe->rs_num < 0 || ioe->rs_num >=
3060                                     PF_RULESET_MAX) {
3061                                         PF_RULES_WUNLOCK();
3062                                         free(ioes, M_TEMP);
3063                                         error = EINVAL;
3064                                         goto fail;
3065                                 }
3066                                 rs = pf_find_ruleset(ioe->anchor);
3067                                 if (rs == NULL ||
3068                                     !rs->rules[ioe->rs_num].inactive.open ||
3069                                     rs->rules[ioe->rs_num].inactive.ticket !=
3070                                     ioe->ticket) {
3071                                         PF_RULES_WUNLOCK();
3072                                         free(ioes, M_TEMP);
3073                                         error = EBUSY;
3074                                         goto fail;
3075                                 }
3076                                 break;
3077                         }
3078                 }
3079                 /* Now do the commit - no errors should happen here. */
3080                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3081                         switch (ioe->rs_num) {
3082 #ifdef ALTQ
3083                         case PF_RULESET_ALTQ:
3084                                 if ((error = pf_commit_altq(ioe->ticket))) {
3085                                         PF_RULES_WUNLOCK();
3086                                         free(ioes, M_TEMP);
3087                                         goto fail; /* really bad */
3088                                 }
3089                                 break;
3090 #endif /* ALTQ */
3091                         case PF_RULESET_TABLE:
3092                             {
3093                                 struct pfr_table table;
3094
3095                                 bzero(&table, sizeof(table));
3096                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3097                                     sizeof(table.pfrt_anchor));
3098                                 if ((error = pfr_ina_commit(&table,
3099                                     ioe->ticket, NULL, NULL, 0))) {
3100                                         PF_RULES_WUNLOCK();
3101                                         free(ioes, M_TEMP);
3102                                         goto fail; /* really bad */
3103                                 }
3104                                 break;
3105                             }
3106                         default:
3107                                 if ((error = pf_commit_rules(ioe->ticket,
3108                                     ioe->rs_num, ioe->anchor))) {
3109                                         PF_RULES_WUNLOCK();
3110                                         free(ioes, M_TEMP);
3111                                         goto fail; /* really bad */
3112                                 }
3113                                 break;
3114                         }
3115                 }
3116                 PF_RULES_WUNLOCK();
3117                 free(ioes, M_TEMP);
3118                 break;
3119         }
3120
3121         case DIOCGETSRCNODES: {
3122                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
3123                 struct pf_srchash       *sh;
3124                 struct pf_src_node      *n, *p, *pstore;
3125                 uint32_t                 i, nr = 0;
3126
3127                 if (psn->psn_len == 0) {
3128                         for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3129                             i++, sh++) {
3130                                 PF_HASHROW_LOCK(sh);
3131                                 LIST_FOREACH(n, &sh->nodes, entry)
3132                                         nr++;
3133                                 PF_HASHROW_UNLOCK(sh);
3134                         }
3135                         psn->psn_len = sizeof(struct pf_src_node) * nr;
3136                         break;
3137                 }
3138
3139                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3140                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3141                     i++, sh++) {
3142                     PF_HASHROW_LOCK(sh);
3143                     LIST_FOREACH(n, &sh->nodes, entry) {
3144                         int     secs = time_uptime, diff;
3145
3146                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3147                                 break;
3148
3149                         bcopy(n, p, sizeof(struct pf_src_node));
3150                         if (n->rule.ptr != NULL)
3151                                 p->rule.nr = n->rule.ptr->nr;
3152                         p->creation = secs - p->creation;
3153                         if (p->expire > secs)
3154                                 p->expire -= secs;
3155                         else
3156                                 p->expire = 0;
3157
3158                         /* Adjust the connection rate estimate. */
3159                         diff = secs - n->conn_rate.last;
3160                         if (diff >= n->conn_rate.seconds)
3161                                 p->conn_rate.count = 0;
3162                         else
3163                                 p->conn_rate.count -=
3164                                     n->conn_rate.count * diff /
3165                                     n->conn_rate.seconds;
3166                         p++;
3167                         nr++;
3168                     }
3169                     PF_HASHROW_UNLOCK(sh);
3170                 }
3171                 error = copyout(pstore, psn->psn_src_nodes,
3172                     sizeof(struct pf_src_node) * nr);
3173                 if (error) {
3174                         free(pstore, M_TEMP);
3175                         break;
3176                 }
3177                 psn->psn_len = sizeof(struct pf_src_node) * nr;
3178                 free(pstore, M_TEMP);
3179                 break;
3180         }
3181
3182         case DIOCCLRSRCNODES: {
3183
3184                 pf_clear_srcnodes(NULL);
3185                 pf_purge_expired_src_nodes();
3186                 break;
3187         }
3188
3189         case DIOCKILLSRCNODES:
3190                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3191                 break;
3192
3193         case DIOCSETHOSTID: {
3194                 u_int32_t       *hostid = (u_int32_t *)addr;
3195
3196                 PF_RULES_WLOCK();
3197                 if (*hostid == 0)
3198                         V_pf_status.hostid = arc4random();
3199                 else
3200                         V_pf_status.hostid = *hostid;
3201                 PF_RULES_WUNLOCK();
3202                 break;
3203         }
3204
3205         case DIOCOSFPFLUSH:
3206                 PF_RULES_WLOCK();
3207                 pf_osfp_flush();
3208                 PF_RULES_WUNLOCK();
3209                 break;
3210
3211         case DIOCIGETIFACES: {
3212                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3213                 struct pfi_kif *ifstore;
3214                 size_t bufsiz;
3215
3216                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3217                         error = ENODEV;
3218                         break;
3219                 }
3220
3221                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3222                 ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3223                 PF_RULES_RLOCK();
3224                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3225                 PF_RULES_RUNLOCK();
3226                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3227                 free(ifstore, M_TEMP);
3228                 break;
3229         }
3230
3231         case DIOCSETIFFLAG: {
3232                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3233
3234                 PF_RULES_WLOCK();
3235                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3236                 PF_RULES_WUNLOCK();
3237                 break;
3238         }
3239
3240         case DIOCCLRIFFLAG: {
3241                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3242
3243                 PF_RULES_WLOCK();
3244                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3245                 PF_RULES_WUNLOCK();
3246                 break;
3247         }
3248
3249         default:
3250                 error = ENODEV;
3251                 break;
3252         }
3253 fail:
3254         if (sx_xlocked(&pf_ioctl_lock))
3255                 sx_xunlock(&pf_ioctl_lock);
3256         CURVNET_RESTORE();
3257
3258         return (error);
3259 }
3260
3261 void
3262 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3263 {
3264         bzero(sp, sizeof(struct pfsync_state));
3265
3266         /* copy from state key */
3267         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3268         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3269         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3270         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3271         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3272         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3273         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3274         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3275         sp->proto = st->key[PF_SK_WIRE]->proto;
3276         sp->af = st->key[PF_SK_WIRE]->af;
3277
3278         /* copy from state */
3279         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3280         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3281         sp->creation = htonl(time_uptime - st->creation);
3282         sp->expire = pf_state_expires(st);
3283         if (sp->expire <= time_uptime)
3284                 sp->expire = htonl(0);
3285         else
3286                 sp->expire = htonl(sp->expire - time_uptime);
3287
3288         sp->direction = st->direction;
3289         sp->log = st->log;
3290         sp->timeout = st->timeout;
3291         sp->state_flags = st->state_flags;
3292         if (st->src_node)
3293                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3294         if (st->nat_src_node)
3295                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3296
3297         sp->id = st->id;
3298         sp->creatorid = st->creatorid;
3299         pf_state_peer_hton(&st->src, &sp->src);
3300         pf_state_peer_hton(&st->dst, &sp->dst);
3301
3302         if (st->rule.ptr == NULL)
3303                 sp->rule = htonl(-1);
3304         else
3305                 sp->rule = htonl(st->rule.ptr->nr);
3306         if (st->anchor.ptr == NULL)
3307                 sp->anchor = htonl(-1);
3308         else
3309                 sp->anchor = htonl(st->anchor.ptr->nr);
3310         if (st->nat_rule.ptr == NULL)
3311                 sp->nat_rule = htonl(-1);
3312         else
3313                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3314
3315         pf_state_counter_hton(st->packets[0], sp->packets[0]);
3316         pf_state_counter_hton(st->packets[1], sp->packets[1]);
3317         pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3318         pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3319
3320 }
3321
3322 static void
3323 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3324 {
3325         struct pfr_ktable *kt;
3326
3327         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3328
3329         kt = aw->p.tbl;
3330         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3331                 kt = kt->pfrkt_root;
3332         aw->p.tbl = NULL;
3333         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3334                 kt->pfrkt_cnt : -1;
3335 }
3336
3337 /*
3338  * XXX - Check for version missmatch!!!
3339  */
3340 static void
3341 pf_clear_states(void)
3342 {
3343         struct pf_state *s;
3344         u_int i;
3345
3346         for (i = 0; i <= pf_hashmask; i++) {
3347                 struct pf_idhash *ih = &V_pf_idhash[i];
3348 relock:
3349                 PF_HASHROW_LOCK(ih);
3350                 LIST_FOREACH(s, &ih->states, entry) {
3351                         s->timeout = PFTM_PURGE;
3352                         /* Don't send out individual delete messages. */
3353                         s->sync_state = PFSTATE_NOSYNC;
3354                         pf_unlink_state(s, PF_ENTER_LOCKED);
3355                         goto relock;
3356                 }
3357                 PF_HASHROW_UNLOCK(ih);
3358         }
3359 }
3360
3361 static int
3362 pf_clear_tables(void)
3363 {
3364         struct pfioc_table io;
3365         int error;
3366
3367         bzero(&io, sizeof(io));
3368
3369         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3370             io.pfrio_flags);
3371
3372         return (error);
3373 }
3374
3375 static void
3376 pf_clear_srcnodes(struct pf_src_node *n)
3377 {
3378         struct pf_state *s;
3379         int i;
3380
3381         for (i = 0; i <= pf_hashmask; i++) {
3382                 struct pf_idhash *ih = &V_pf_idhash[i];
3383
3384                 PF_HASHROW_LOCK(ih);
3385                 LIST_FOREACH(s, &ih->states, entry) {
3386                         if (n == NULL || n == s->src_node)
3387                                 s->src_node = NULL;
3388                         if (n == NULL || n == s->nat_src_node)
3389                                 s->nat_src_node = NULL;
3390                 }
3391                 PF_HASHROW_UNLOCK(ih);
3392         }
3393
3394         if (n == NULL) {
3395                 struct pf_srchash *sh;
3396
3397                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3398                     i++, sh++) {
3399                         PF_HASHROW_LOCK(sh);
3400                         LIST_FOREACH(n, &sh->nodes, entry) {
3401                                 n->expire = 1;
3402                                 n->states = 0;
3403                         }
3404                         PF_HASHROW_UNLOCK(sh);
3405                 }
3406         } else {
3407                 /* XXX: hash slot should already be locked here. */
3408                 n->expire = 1;
3409                 n->states = 0;
3410         }
3411 }
3412
3413 static void
3414 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3415 {
3416         struct pf_src_node_list  kill;
3417
3418         LIST_INIT(&kill);
3419         for (int i = 0; i <= pf_srchashmask; i++) {
3420                 struct pf_srchash *sh = &V_pf_srchash[i];
3421                 struct pf_src_node *sn, *tmp;
3422
3423                 PF_HASHROW_LOCK(sh);
3424                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3425                         if (PF_MATCHA(psnk->psnk_src.neg,
3426                               &psnk->psnk_src.addr.v.a.addr,
3427                               &psnk->psnk_src.addr.v.a.mask,
3428                               &sn->addr, sn->af) &&
3429                             PF_MATCHA(psnk->psnk_dst.neg,
3430                               &psnk->psnk_dst.addr.v.a.addr,
3431                               &psnk->psnk_dst.addr.v.a.mask,
3432                               &sn->raddr, sn->af)) {
3433                                 pf_unlink_src_node(sn);
3434                                 LIST_INSERT_HEAD(&kill, sn, entry);
3435                                 sn->expire = 1;
3436                         }
3437                 PF_HASHROW_UNLOCK(sh);
3438         }
3439
3440         for (int i = 0; i <= pf_hashmask; i++) {
3441                 struct pf_idhash *ih = &V_pf_idhash[i];
3442                 struct pf_state *s;
3443
3444                 PF_HASHROW_LOCK(ih);
3445                 LIST_FOREACH(s, &ih->states, entry) {
3446                         if (s->src_node && s->src_node->expire == 1)
3447                                 s->src_node = NULL;
3448                         if (s->nat_src_node && s->nat_src_node->expire == 1)
3449                                 s->nat_src_node = NULL;
3450                 }
3451                 PF_HASHROW_UNLOCK(ih);
3452         }
3453
3454         psnk->psnk_killed = pf_free_src_nodes(&kill);
3455 }
3456
3457 /*
3458  * XXX - Check for version missmatch!!!
3459  */
3460
3461 /*
3462  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3463  */
3464 static int
3465 shutdown_pf(void)
3466 {
3467         int error = 0;
3468         u_int32_t t[5];
3469         char nn = '\0';
3470
3471         V_pf_status.running = 0;
3472
3473         counter_u64_free(V_pf_default_rule.states_cur);
3474         counter_u64_free(V_pf_default_rule.states_tot);
3475         counter_u64_free(V_pf_default_rule.src_nodes);
3476
3477         for (int i = 0; i < PFRES_MAX; i++)
3478                 counter_u64_free(V_pf_status.counters[i]);
3479         for (int i = 0; i < LCNT_MAX; i++)
3480                 counter_u64_free(V_pf_status.lcounters[i]);
3481         for (int i = 0; i < FCNT_MAX; i++)
3482                 counter_u64_free(V_pf_status.fcounters[i]);
3483         for (int i = 0; i < SCNT_MAX; i++)
3484                 counter_u64_free(V_pf_status.scounters[i]);
3485
3486         do {
3487                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3488                     != 0) {
3489                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3490                         break;
3491                 }
3492                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3493                     != 0) {
3494                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3495                         break;          /* XXX: rollback? */
3496                 }
3497                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3498                     != 0) {
3499                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3500                         break;          /* XXX: rollback? */
3501                 }
3502                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3503                     != 0) {
3504                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3505                         break;          /* XXX: rollback? */
3506                 }
3507                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3508                     != 0) {
3509                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3510                         break;          /* XXX: rollback? */
3511                 }
3512
3513                 /* XXX: these should always succeed here */
3514                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3515                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3516                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3517                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3518                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3519
3520                 if ((error = pf_clear_tables()) != 0)
3521                         break;
3522
3523 #ifdef ALTQ
3524                 if ((error = pf_begin_altq(&t[0])) != 0) {
3525                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3526                         break;
3527                 }
3528                 pf_commit_altq(t[0]);
3529 #endif
3530
3531                 pf_clear_states();
3532
3533                 pf_clear_srcnodes(NULL);
3534
3535                 /* status does not use malloced mem so no need to cleanup */
3536                 /* fingerprints and interfaces have thier own cleanup code */
3537         } while(0);
3538
3539         return (error);
3540 }
3541
3542 #ifdef INET
3543 static int
3544 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3545     struct inpcb *inp)
3546 {
3547         int chk;
3548
3549         chk = pf_test(PF_IN, ifp, m, inp);
3550         if (chk && *m) {
3551                 m_freem(*m);
3552                 *m = NULL;
3553         }
3554
3555         return (chk);
3556 }
3557
3558 static int
3559 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3560     struct inpcb *inp)
3561 {
3562         int chk;
3563
3564         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3565         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3566                 in_delayed_cksum(*m);
3567                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3568         }
3569
3570         chk = pf_test(PF_OUT, ifp, m, inp);
3571         if (chk && *m) {
3572                 m_freem(*m);
3573                 *m = NULL;
3574         }
3575
3576         return (chk);
3577 }
3578 #endif
3579
3580 #ifdef INET6
3581 static int
3582 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3583     struct inpcb *inp)
3584 {
3585         int chk;
3586
3587         /*
3588          * In case of loopback traffic IPv6 uses the real interface in
3589          * order to support scoped addresses. In order to support stateful
3590          * filtering we have change this to lo0 as it is the case in IPv4.
3591          */
3592         CURVNET_SET(ifp->if_vnet);
3593         chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3594         CURVNET_RESTORE();
3595         if (chk && *m) {
3596                 m_freem(*m);
3597                 *m = NULL;
3598         }
3599         return chk;
3600 }
3601
3602 static int
3603 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3604     struct inpcb *inp)
3605 {
3606         int chk;
3607
3608         /* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3609         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3610                 in6_delayed_cksum(*m,
3611                     (*m)->m_pkthdr.len - sizeof(struct ip6_hdr),
3612                     sizeof(struct ip6_hdr));
3613                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
3614         }
3615         CURVNET_SET(ifp->if_vnet);
3616         chk = pf_test6(PF_OUT, ifp, m, inp);
3617         CURVNET_RESTORE();
3618         if (chk && *m) {
3619                 m_freem(*m);
3620                 *m = NULL;
3621         }
3622         return chk;
3623 }
3624 #endif /* INET6 */
3625
3626 static int
3627 hook_pf(void)
3628 {
3629 #ifdef INET
3630         struct pfil_head *pfh_inet;
3631 #endif
3632 #ifdef INET6
3633         struct pfil_head *pfh_inet6;
3634 #endif
3635
3636         if (V_pf_pfil_hooked)
3637                 return (0);
3638
3639 #ifdef INET
3640         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3641         if (pfh_inet == NULL)
3642                 return (ESRCH); /* XXX */
3643         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3644         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3645 #endif
3646 #ifdef INET6
3647         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3648         if (pfh_inet6 == NULL) {
3649 #ifdef INET
3650                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3651                     pfh_inet);
3652                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3653                     pfh_inet);
3654 #endif
3655                 return (ESRCH); /* XXX */
3656         }
3657         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3658         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3659 #endif
3660
3661         V_pf_pfil_hooked = 1;
3662         return (0);
3663 }
3664
3665 static int
3666 dehook_pf(void)
3667 {
3668 #ifdef INET
3669         struct pfil_head *pfh_inet;
3670 #endif
3671 #ifdef INET6
3672         struct pfil_head *pfh_inet6;
3673 #endif
3674
3675         if (V_pf_pfil_hooked == 0)
3676                 return (0);
3677
3678 #ifdef INET
3679         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3680         if (pfh_inet == NULL)
3681                 return (ESRCH); /* XXX */
3682         pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3683             pfh_inet);
3684         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3685             pfh_inet);
3686 #endif
3687 #ifdef INET6
3688         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3689         if (pfh_inet6 == NULL)
3690                 return (ESRCH); /* XXX */
3691         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3692             pfh_inet6);
3693         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3694             pfh_inet6);
3695 #endif
3696
3697         V_pf_pfil_hooked = 0;
3698         return (0);
3699 }
3700
3701 static int
3702 pf_load(void)
3703 {
3704         int error;
3705
3706         VNET_ITERATOR_DECL(vnet_iter);
3707
3708         VNET_LIST_RLOCK();
3709         VNET_FOREACH(vnet_iter) {
3710                 CURVNET_SET(vnet_iter);
3711                 V_pf_pfil_hooked = 0;
3712                 V_pf_end_threads = 0;
3713                 TAILQ_INIT(&V_pf_tags);
3714                 TAILQ_INIT(&V_pf_qids);
3715                 CURVNET_RESTORE();
3716         }
3717         VNET_LIST_RUNLOCK();
3718
3719         rw_init(&pf_rules_lock, "pf rulesets");
3720         sx_init(&pf_ioctl_lock, "pf ioctl");
3721
3722         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3723         if ((error = pfattach()) != 0)
3724                 return (error);
3725
3726         return (0);
3727 }
3728
3729 static int
3730 pf_unload(void)
3731 {
3732         int error = 0;
3733
3734         V_pf_status.running = 0;
3735         swi_remove(V_pf_swi_cookie);
3736         error = dehook_pf();
3737         if (error) {
3738                 /*
3739                  * Should not happen!
3740                  * XXX Due to error code ESRCH, kldunload will show
3741                  * a message like 'No such process'.
3742                  */
3743                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3744                 return error;
3745         }
3746         PF_RULES_WLOCK();
3747         shutdown_pf();
3748         V_pf_end_threads = 1;
3749         while (V_pf_end_threads < 2) {
3750                 wakeup_one(pf_purge_thread);
3751                 rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3752         }
3753         PF_RULES_WUNLOCK();
3754         pf_normalize_cleanup();
3755         pfi_cleanup();
3756         pfr_cleanup();
3757         pf_osfp_flush();
3758         pf_cleanup();
3759         if (IS_DEFAULT_VNET(curvnet))
3760                 pf_mtag_cleanup();
3761         destroy_dev(pf_dev);
3762         rw_destroy(&pf_rules_lock);
3763         sx_destroy(&pf_ioctl_lock);
3764
3765         return (error);
3766 }
3767
3768 static int
3769 pf_modevent(module_t mod, int type, void *data)
3770 {
3771         int error = 0;
3772
3773         switch(type) {
3774         case MOD_LOAD:
3775                 error = pf_load();
3776                 break;
3777         case MOD_QUIESCE:
3778                 /*
3779                  * Module should not be unloaded due to race conditions.
3780                  */
3781                 error = EBUSY;
3782                 break;
3783         case MOD_UNLOAD:
3784                 error = pf_unload();
3785                 break;
3786         default:
3787                 error = EINVAL;
3788                 break;
3789         }
3790
3791         return (error);
3792 }
3793
3794 static moduledata_t pf_mod = {
3795         "pf",
3796         pf_modevent,
3797         0
3798 };
3799
3800 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3801 MODULE_VERSION(pf, PF_MODVER);