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 <netinet/ip_icmp.h>
  76
  77 #ifdef INET6
  78 #include <netinet/ip6.h>
  79 #endif /* INET6 */
  80
  81 #ifdef ALTQ
  82 #include <altq/altq.h>
  83 #endif
  84
  85 static int               pfattach(void);
  86 static struct pf_pool   *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
  87                             u_int8_t, u_int8_t, u_int8_t);
  88
  89 static void              pf_mv_pool(struct pf_palist *, struct pf_palist *);
  90 static void              pf_empty_pool(struct pf_palist *);
  91 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
  92                             struct thread *);
  93 #ifdef ALTQ
  94 static int               pf_begin_altq(u_int32_t *);
  95 static int               pf_rollback_altq(u_int32_t);
  96 static int               pf_commit_altq(u_int32_t);
  97 static int               pf_enable_altq(struct pf_altq *);
  98 static int               pf_disable_altq(struct pf_altq *);
  99 static u_int32_t         pf_qname2qid(char *);
 100 static void              pf_qid_unref(u_int32_t);
 101 #endif /* ALTQ */
 102 static int               pf_begin_rules(u_int32_t *, int, const char *);
 103 static int               pf_rollback_rules(u_int32_t, int, char *);
 104 static int               pf_setup_pfsync_matching(struct pf_ruleset *);
 105 static void              pf_hash_rule(MD5_CTX *, struct pf_rule *);
 106 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 107 static int               pf_commit_rules(u_int32_t, int, char *);
 108 static int               pf_addr_setup(struct pf_ruleset *,
 109                             struct pf_addr_wrap *, sa_family_t);
 110 static void              pf_addr_copyout(struct pf_addr_wrap *);
 111
 112 VNET_DEFINE(struct pf_rule,     pf_default_rule);
 113
 114 #ifdef ALTQ
 115 static VNET_DEFINE(int,         pf_altq_running);
 116 #define V_pf_altq_running       VNET(pf_altq_running)
 117 #endif
 118
 119 #define TAGID_MAX        50000
 120 struct pf_tagname {
 121         TAILQ_ENTRY(pf_tagname) entries;
 122         char                    name[PF_TAG_NAME_SIZE];
 123         uint16_t                tag;
 124         int                     ref;
 125 };
 126
 127 TAILQ_HEAD(pf_tags, pf_tagname);
 128 #define V_pf_tags               VNET(pf_tags)
 129 VNET_DEFINE(struct pf_tags, pf_tags);
 130 #define V_pf_qids               VNET(pf_qids)
 131 VNET_DEFINE(struct pf_tags, pf_qids);
 132 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
 133 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 134 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 135
 136 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 137 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 138 #endif
 139
 140 static u_int16_t         tagname2tag(struct pf_tags *, char *);
 141 static u_int16_t         pf_tagname2tag(char *);
 142 static void              tag_unref(struct pf_tags *, u_int16_t);
 143
 144 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 145
 146 struct cdev *pf_dev;
 147
 148 /*
 149  * XXX - These are new and need to be checked when moveing to a new version
 150  */
 151 static void              pf_clear_states(void);
 152 static int               pf_clear_tables(void);
 153 static void              pf_clear_srcnodes(struct pf_src_node *);
 154 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 155 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 156
 157 /*
 158  * Wrapper functions for pfil(9) hooks
 159  */
 160 #ifdef INET
 161 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 162     int dir, struct inpcb *inp);
 163 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 164     int dir, struct inpcb *inp);
 165 #endif
 166 #ifdef INET6
 167 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 168     int dir, struct inpcb *inp);
 169 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 170     int dir, struct inpcb *inp);
 171 #endif
 172
 173 static int              hook_pf(void);
 174 static int              dehook_pf(void);
 175 static int              shutdown_pf(void);
 176 static int              pf_load(void);
 177 static int              pf_unload(void);
 178
 179 static struct cdevsw pf_cdevsw = {
 180         .d_ioctl =      pfioctl,
 181         .d_name =       PF_NAME,
 182         .d_version =    D_VERSION,
 183 };
 184
 185 static volatile VNET_DEFINE(int, pf_pfil_hooked);
 186 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 187 VNET_DEFINE(int,                pf_end_threads);
 188
 189 struct rwlock                   pf_rules_lock;
 190
 191 /* pfsync */
 192 pfsync_state_import_t           *pfsync_state_import_ptr = NULL;
 193 pfsync_insert_state_t           *pfsync_insert_state_ptr = NULL;
 194 pfsync_update_state_t           *pfsync_update_state_ptr = NULL;
 195 pfsync_delete_state_t           *pfsync_delete_state_ptr = NULL;
 196 pfsync_clear_states_t           *pfsync_clear_states_ptr = NULL;
 197 pfsync_defer_t                  *pfsync_defer_ptr = NULL;
 198 /* pflog */
 199 pflog_packet_t                  *pflog_packet_ptr = NULL;
 200
 201 static int
 202 pfattach(void)
 203 {
 204         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 205         int error;
 206
 207         pf_initialize();
 208         pfr_initialize();
 209         pfi_initialize();
 210         pf_normalize_init();
 211
 212         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 213         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 214
 215         RB_INIT(&V_pf_anchors);
 216         pf_init_ruleset(&pf_main_ruleset);
 217
 218         /* default rule should never be garbage collected */
 219         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 220 #ifdef PF_DEFAULT_TO_DROP
 221         V_pf_default_rule.action = PF_DROP;
 222 #else
 223         V_pf_default_rule.action = PF_PASS;
 224 #endif
 225         V_pf_default_rule.nr = -1;
 226         V_pf_default_rule.rtableid = -1;
 227
 228         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 229         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 230         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 231
 232         /* initialize default timeouts */
 233         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 234         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 235         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 236         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 237         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 238         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 239         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 240         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 241         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 242         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 243         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 244         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 245         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 246         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 247         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 248         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 249         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 250         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 251         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 252         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 253
 254         bzero(&V_pf_status, sizeof(V_pf_status));
 255         V_pf_status.debug = PF_DEBUG_URGENT;
 256
 257         V_pf_pfil_hooked = 0;
 258
 259         /* XXX do our best to avoid a conflict */
 260         V_pf_status.hostid = arc4random();
 261
 262         if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
 263             "pf purge")) != 0)
 264                 /* XXXGL: leaked all above. */
 265                 return (error);
 266         if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
 267             INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
 268                 /* XXXGL: leaked all above. */
 269                 return (error);
 270
 271         return (0);
 272 }
 273
 274 static struct pf_pool *
 275 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 276     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 277     u_int8_t check_ticket)
 278 {
 279         struct pf_ruleset       *ruleset;
 280         struct pf_rule          *rule;
 281         int                      rs_num;
 282
 283         ruleset = pf_find_ruleset(anchor);
 284         if (ruleset == NULL)
 285                 return (NULL);
 286         rs_num = pf_get_ruleset_number(rule_action);
 287         if (rs_num >= PF_RULESET_MAX)
 288                 return (NULL);
 289         if (active) {
 290                 if (check_ticket && ticket !=
 291                     ruleset->rules[rs_num].active.ticket)
 292                         return (NULL);
 293                 if (r_last)
 294                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 295                             pf_rulequeue);
 296                 else
 297                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 298         } else {
 299                 if (check_ticket && ticket !=
 300                     ruleset->rules[rs_num].inactive.ticket)
 301                         return (NULL);
 302                 if (r_last)
 303                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 304                             pf_rulequeue);
 305                 else
 306                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 307         }
 308         if (!r_last) {
 309                 while ((rule != NULL) && (rule->nr != rule_number))
 310                         rule = TAILQ_NEXT(rule, entries);
 311         }
 312         if (rule == NULL)
 313                 return (NULL);
 314
 315         return (&rule->rpool);
 316 }
 317
 318 static void
 319 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 320 {
 321         struct pf_pooladdr      *mv_pool_pa;
 322
 323         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 324                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 325                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 326         }
 327 }
 328
 329 static void
 330 pf_empty_pool(struct pf_palist *poola)
 331 {
 332         struct pf_pooladdr *pa;
 333
 334         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 335                 switch (pa->addr.type) {
 336                 case PF_ADDR_DYNIFTL:
 337                         pfi_dynaddr_remove(pa->addr.p.dyn);
 338                         break;
 339                 case PF_ADDR_TABLE:
 340                         pfr_detach_table(pa->addr.p.tbl);
 341                         break;
 342                 }
 343                 if (pa->kif)
 344                         pfi_kif_unref(pa->kif);
 345                 TAILQ_REMOVE(poola, pa, entries);
 346                 free(pa, M_PFRULE);
 347         }
 348 }
 349
 350 static void
 351 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 352 {
 353
 354         PF_RULES_WASSERT();
 355
 356         TAILQ_REMOVE(rulequeue, rule, entries);
 357
 358         PF_UNLNKDRULES_LOCK();
 359         rule->rule_flag |= PFRULE_REFS;
 360         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 361         PF_UNLNKDRULES_UNLOCK();
 362 }
 363
 364 void
 365 pf_free_rule(struct pf_rule *rule)
 366 {
 367
 368         PF_RULES_WASSERT();
 369
 370         if (rule->tag)
 371                 tag_unref(&V_pf_tags, rule->tag);
 372         if (rule->match_tag)
 373                 tag_unref(&V_pf_tags, rule->match_tag);
 374 #ifdef ALTQ
 375         if (rule->pqid != rule->qid)
 376                 pf_qid_unref(rule->pqid);
 377         pf_qid_unref(rule->qid);
 378 #endif
 379         switch (rule->src.addr.type) {
 380         case PF_ADDR_DYNIFTL:
 381                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 382                 break;
 383         case PF_ADDR_TABLE:
 384                 pfr_detach_table(rule->src.addr.p.tbl);
 385                 break;
 386         }
 387         switch (rule->dst.addr.type) {
 388         case PF_ADDR_DYNIFTL:
 389                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 390                 break;
 391         case PF_ADDR_TABLE:
 392                 pfr_detach_table(rule->dst.addr.p.tbl);
 393                 break;
 394         }
 395         if (rule->overload_tbl)
 396                 pfr_detach_table(rule->overload_tbl);
 397         if (rule->kif)
 398                 pfi_kif_unref(rule->kif);
 399         pf_anchor_remove(rule);
 400         pf_empty_pool(&rule->rpool.list);
 401         counter_u64_free(rule->states_cur);
 402         counter_u64_free(rule->states_tot);
 403         counter_u64_free(rule->src_nodes);
 404         free(rule, M_PFRULE);
 405 }
 406
 407 static u_int16_t
 408 tagname2tag(struct pf_tags *head, char *tagname)
 409 {
 410         struct pf_tagname       *tag, *p = NULL;
 411         u_int16_t                new_tagid = 1;
 412
 413         PF_RULES_WASSERT();
 414
 415         TAILQ_FOREACH(tag, head, entries)
 416                 if (strcmp(tagname, tag->name) == 0) {
 417                         tag->ref++;
 418                         return (tag->tag);
 419                 }
 420
 421         /*
 422          * to avoid fragmentation, we do a linear search from the beginning
 423          * and take the first free slot we find. if there is none or the list
 424          * is empty, append a new entry at the end.
 425          */
 426
 427         /* new entry */
 428         if (!TAILQ_EMPTY(head))
 429                 for (p = TAILQ_FIRST(head); p != NULL &&
 430                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 431                         new_tagid = p->tag + 1;
 432
 433         if (new_tagid > TAGID_MAX)
 434                 return (0);
 435
 436         /* allocate and fill new struct pf_tagname */
 437         tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
 438         if (tag == NULL)
 439                 return (0);
 440         strlcpy(tag->name, tagname, sizeof(tag->name));
 441         tag->tag = new_tagid;
 442         tag->ref++;
 443
 444         if (p != NULL)  /* insert new entry before p */
 445                 TAILQ_INSERT_BEFORE(p, tag, entries);
 446         else    /* either list empty or no free slot in between */
 447                 TAILQ_INSERT_TAIL(head, tag, entries);
 448
 449         return (tag->tag);
 450 }
 451
 452 static void
 453 tag_unref(struct pf_tags *head, u_int16_t tag)
 454 {
 455         struct pf_tagname       *p, *next;
 456
 457         PF_RULES_WASSERT();
 458
 459         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 460                 next = TAILQ_NEXT(p, entries);
 461                 if (tag == p->tag) {
 462                         if (--p->ref == 0) {
 463                                 TAILQ_REMOVE(head, p, entries);
 464                                 free(p, M_PFTAG);
 465                         }
 466                         break;
 467                 }
 468         }
 469 }
 470
 471 static u_int16_t
 472 pf_tagname2tag(char *tagname)
 473 {
 474         return (tagname2tag(&V_pf_tags, tagname));
 475 }
 476
 477 #ifdef ALTQ
 478 static u_int32_t
 479 pf_qname2qid(char *qname)
 480 {
 481         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 482 }
 483
 484 static void
 485 pf_qid_unref(u_int32_t qid)
 486 {
 487         tag_unref(&V_pf_qids, (u_int16_t)qid);
 488 }
 489
 490 static int
 491 pf_begin_altq(u_int32_t *ticket)
 492 {
 493         struct pf_altq  *altq;
 494         int              error = 0;
 495
 496         PF_RULES_WASSERT();
 497
 498         /* Purge the old altq list */
 499         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 500                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 501                 if (altq->qname[0] == 0 &&
 502                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 503                         /* detach and destroy the discipline */
 504                         error = altq_remove(altq);
 505                 } else
 506                         pf_qid_unref(altq->qid);
 507                 free(altq, M_PFALTQ);
 508         }
 509         if (error)
 510                 return (error);
 511         *ticket = ++V_ticket_altqs_inactive;
 512         V_altqs_inactive_open = 1;
 513         return (0);
 514 }
 515
 516 static int
 517 pf_rollback_altq(u_int32_t ticket)
 518 {
 519         struct pf_altq  *altq;
 520         int              error = 0;
 521
 522         PF_RULES_WASSERT();
 523
 524         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 525                 return (0);
 526         /* Purge the old altq list */
 527         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 528                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 529                 if (altq->qname[0] == 0 &&
 530                    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 531                         /* detach and destroy the discipline */
 532                         error = altq_remove(altq);
 533                 } else
 534                         pf_qid_unref(altq->qid);
 535                 free(altq, M_PFALTQ);
 536         }
 537         V_altqs_inactive_open = 0;
 538         return (error);
 539 }
 540
 541 static int
 542 pf_commit_altq(u_int32_t ticket)
 543 {
 544         struct pf_altqqueue     *old_altqs;
 545         struct pf_altq          *altq;
 546         int                      err, error = 0;
 547
 548         PF_RULES_WASSERT();
 549
 550         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 551                 return (EBUSY);
 552
 553         /* swap altqs, keep the old. */
 554         old_altqs = V_pf_altqs_active;
 555         V_pf_altqs_active = V_pf_altqs_inactive;
 556         V_pf_altqs_inactive = old_altqs;
 557         V_ticket_altqs_active = V_ticket_altqs_inactive;
 558
 559         /* Attach new disciplines */
 560         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
 561         if (altq->qname[0] == 0 &&
 562            (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 563                         /* attach the discipline */
 564                         error = altq_pfattach(altq);
 565                         if (error == 0 && V_pf_altq_running)
 566                                 error = pf_enable_altq(altq);
 567                         if (error != 0)
 568                                 return (error);
 569                 }
 570         }
 571
 572         /* Purge the old altq list */
 573         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 574                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 575                 if (altq->qname[0] == 0 &&
 576                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 577                         /* detach and destroy the discipline */
 578                         if (V_pf_altq_running)
 579                                 error = pf_disable_altq(altq);
 580                         err = altq_pfdetach(altq);
 581                         if (err != 0 && error == 0)
 582                                 error = err;
 583                         err = altq_remove(altq);
 584                         if (err != 0 && error == 0)
 585                                 error = err;
 586                 } else
 587                         pf_qid_unref(altq->qid);
 588                 free(altq, M_PFALTQ);
 589         }
 590
 591         V_altqs_inactive_open = 0;
 592         return (error);
 593 }
 594
 595 static int
 596 pf_enable_altq(struct pf_altq *altq)
 597 {
 598         struct ifnet            *ifp;
 599         struct tb_profile        tb;
 600         int                      error = 0;
 601
 602         if ((ifp = ifunit(altq->ifname)) == NULL)
 603                 return (EINVAL);
 604
 605         if (ifp->if_snd.altq_type != ALTQT_NONE)
 606                 error = altq_enable(&ifp->if_snd);
 607
 608         /* set tokenbucket regulator */
 609         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 610                 tb.rate = altq->ifbandwidth;
 611                 tb.depth = altq->tbrsize;
 612                 error = tbr_set(&ifp->if_snd, &tb);
 613         }
 614
 615         return (error);
 616 }
 617
 618 static int
 619 pf_disable_altq(struct pf_altq *altq)
 620 {
 621         struct ifnet            *ifp;
 622         struct tb_profile        tb;
 623         int                      error;
 624
 625         if ((ifp = ifunit(altq->ifname)) == NULL)
 626                 return (EINVAL);
 627
 628         /*
 629          * when the discipline is no longer referenced, it was overridden
 630          * by a new one.  if so, just return.
 631          */
 632         if (altq->altq_disc != ifp->if_snd.altq_disc)
 633                 return (0);
 634
 635         error = altq_disable(&ifp->if_snd);
 636
 637         if (error == 0) {
 638                 /* clear tokenbucket regulator */
 639                 tb.rate = 0;
 640                 error = tbr_set(&ifp->if_snd, &tb);
 641         }
 642
 643         return (error);
 644 }
 645
 646 void
 647 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 648 {
 649         struct ifnet    *ifp1;
 650         struct pf_altq  *a1, *a2, *a3;
 651         u_int32_t        ticket;
 652         int              error = 0;
 653
 654         /* Interrupt userland queue modifications */
 655         if (V_altqs_inactive_open)
 656                 pf_rollback_altq(V_ticket_altqs_inactive);
 657
 658         /* Start new altq ruleset */
 659         if (pf_begin_altq(&ticket))
 660                 return;
 661
 662         /* Copy the current active set */
 663         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 664                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 665                 if (a2 == NULL) {
 666                         error = ENOMEM;
 667                         break;
 668                 }
 669                 bcopy(a1, a2, sizeof(struct pf_altq));
 670
 671                 if (a2->qname[0] != 0) {
 672                         if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 673                                 error = EBUSY;
 674                                 free(a2, M_PFALTQ);
 675                                 break;
 676                         }
 677                         a2->altq_disc = NULL;
 678                         TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
 679                                 if (strncmp(a3->ifname, a2->ifname,
 680                                     IFNAMSIZ) == 0 && a3->qname[0] == 0) {
 681                                         a2->altq_disc = a3->altq_disc;
 682                                         break;
 683                                 }
 684                         }
 685                 }
 686                 /* Deactivate the interface in question */
 687                 a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 688                 if ((ifp1 = ifunit(a2->ifname)) == NULL ||
 689                     (remove && ifp1 == ifp)) {
 690                         a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 691                 } else {
 692                         error = altq_add(a2);
 693
 694                         if (ticket != V_ticket_altqs_inactive)
 695                                 error = EBUSY;
 696
 697                         if (error) {
 698                                 free(a2, M_PFALTQ);
 699                                 break;
 700                         }
 701                 }
 702
 703                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 704         }
 705
 706         if (error != 0)
 707                 pf_rollback_altq(ticket);
 708         else
 709                 pf_commit_altq(ticket);
 710 }
 711 #endif /* ALTQ */
 712
 713 static int
 714 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 715 {
 716         struct pf_ruleset       *rs;
 717         struct pf_rule          *rule;
 718
 719         PF_RULES_WASSERT();
 720
 721         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 722                 return (EINVAL);
 723         rs = pf_find_or_create_ruleset(anchor);
 724         if (rs == NULL)
 725                 return (EINVAL);
 726         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 727                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 728                 rs->rules[rs_num].inactive.rcount--;
 729         }
 730         *ticket = ++rs->rules[rs_num].inactive.ticket;
 731         rs->rules[rs_num].inactive.open = 1;
 732         return (0);
 733 }
 734
 735 static int
 736 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 737 {
 738         struct pf_ruleset       *rs;
 739         struct pf_rule          *rule;
 740
 741         PF_RULES_WASSERT();
 742
 743         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 744                 return (EINVAL);
 745         rs = pf_find_ruleset(anchor);
 746         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 747             rs->rules[rs_num].inactive.ticket != ticket)
 748                 return (0);
 749         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 750                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 751                 rs->rules[rs_num].inactive.rcount--;
 752         }
 753         rs->rules[rs_num].inactive.open = 0;
 754         return (0);
 755 }
 756
 757 #define PF_MD5_UPD(st, elm)                                             \
 758                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 759
 760 #define PF_MD5_UPD_STR(st, elm)                                         \
 761                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 762
 763 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 764                 (stor) = htonl((st)->elm);                              \
 765                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 766 } while (0)
 767
 768 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 769                 (stor) = htons((st)->elm);                              \
 770                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 771 } while (0)
 772
 773 static void
 774 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 775 {
 776         PF_MD5_UPD(pfr, addr.type);
 777         switch (pfr->addr.type) {
 778                 case PF_ADDR_DYNIFTL:
 779                         PF_MD5_UPD(pfr, addr.v.ifname);
 780                         PF_MD5_UPD(pfr, addr.iflags);
 781                         break;
 782                 case PF_ADDR_TABLE:
 783                         PF_MD5_UPD(pfr, addr.v.tblname);
 784                         break;
 785                 case PF_ADDR_ADDRMASK:
 786                         /* XXX ignore af? */
 787                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 788                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 789                         break;
 790         }
 791
 792         PF_MD5_UPD(pfr, port[0]);
 793         PF_MD5_UPD(pfr, port[1]);
 794         PF_MD5_UPD(pfr, neg);
 795         PF_MD5_UPD(pfr, port_op);
 796 }
 797
 798 static void
 799 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 800 {
 801         u_int16_t x;
 802         u_int32_t y;
 803
 804         pf_hash_rule_addr(ctx, &rule->src);
 805         pf_hash_rule_addr(ctx, &rule->dst);
 806         PF_MD5_UPD_STR(rule, label);
 807         PF_MD5_UPD_STR(rule, ifname);
 808         PF_MD5_UPD_STR(rule, match_tagname);
 809         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 810         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 811         PF_MD5_UPD_HTONL(rule, prob, y);
 812         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 813         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 814         PF_MD5_UPD(rule, uid.op);
 815         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 816         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 817         PF_MD5_UPD(rule, gid.op);
 818         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 819         PF_MD5_UPD(rule, action);
 820         PF_MD5_UPD(rule, direction);
 821         PF_MD5_UPD(rule, af);
 822         PF_MD5_UPD(rule, quick);
 823         PF_MD5_UPD(rule, ifnot);
 824         PF_MD5_UPD(rule, match_tag_not);
 825         PF_MD5_UPD(rule, natpass);
 826         PF_MD5_UPD(rule, keep_state);
 827         PF_MD5_UPD(rule, proto);
 828         PF_MD5_UPD(rule, type);
 829         PF_MD5_UPD(rule, code);
 830         PF_MD5_UPD(rule, flags);
 831         PF_MD5_UPD(rule, flagset);
 832         PF_MD5_UPD(rule, allow_opts);
 833         PF_MD5_UPD(rule, rt);
 834         PF_MD5_UPD(rule, tos);
 835 }
 836
 837 static int
 838 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 839 {
 840         struct pf_ruleset       *rs;
 841         struct pf_rule          *rule, **old_array;
 842         struct pf_rulequeue     *old_rules;
 843         int                      error;
 844         u_int32_t                old_rcount;
 845
 846         PF_RULES_WASSERT();
 847
 848         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 849                 return (EINVAL);
 850         rs = pf_find_ruleset(anchor);
 851         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 852             ticket != rs->rules[rs_num].inactive.ticket)
 853                 return (EBUSY);
 854
 855         /* Calculate checksum for the main ruleset */
 856         if (rs == &pf_main_ruleset) {
 857                 error = pf_setup_pfsync_matching(rs);
 858                 if (error != 0)
 859                         return (error);
 860         }
 861
 862         /* Swap rules, keep the old. */
 863         old_rules = rs->rules[rs_num].active.ptr;
 864         old_rcount = rs->rules[rs_num].active.rcount;
 865         old_array = rs->rules[rs_num].active.ptr_array;
 866
 867         rs->rules[rs_num].active.ptr =
 868             rs->rules[rs_num].inactive.ptr;
 869         rs->rules[rs_num].active.ptr_array =
 870             rs->rules[rs_num].inactive.ptr_array;
 871         rs->rules[rs_num].active.rcount =
 872             rs->rules[rs_num].inactive.rcount;
 873         rs->rules[rs_num].inactive.ptr = old_rules;
 874         rs->rules[rs_num].inactive.ptr_array = old_array;
 875         rs->rules[rs_num].inactive.rcount = old_rcount;
 876
 877         rs->rules[rs_num].active.ticket =
 878             rs->rules[rs_num].inactive.ticket;
 879         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 880
 881
 882         /* Purge the old rule list. */
 883         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 884                 pf_unlink_rule(old_rules, rule);
 885         if (rs->rules[rs_num].inactive.ptr_array)
 886                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 887         rs->rules[rs_num].inactive.ptr_array = NULL;
 888         rs->rules[rs_num].inactive.rcount = 0;
 889         rs->rules[rs_num].inactive.open = 0;
 890         pf_remove_if_empty_ruleset(rs);
 891
 892         return (0);
 893 }
 894
 895 static int
 896 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 897 {
 898         MD5_CTX                  ctx;
 899         struct pf_rule          *rule;
 900         int                      rs_cnt;
 901         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 902
 903         MD5Init(&ctx);
 904         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 905                 /* XXX PF_RULESET_SCRUB as well? */
 906                 if (rs_cnt == PF_RULESET_SCRUB)
 907                         continue;
 908
 909                 if (rs->rules[rs_cnt].inactive.ptr_array)
 910                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 911                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 912
 913                 if (rs->rules[rs_cnt].inactive.rcount) {
 914                         rs->rules[rs_cnt].inactive.ptr_array =
 915                             malloc(sizeof(caddr_t) *
 916                             rs->rules[rs_cnt].inactive.rcount,
 917                             M_TEMP, M_NOWAIT);
 918
 919                         if (!rs->rules[rs_cnt].inactive.ptr_array)
 920                                 return (ENOMEM);
 921                 }
 922
 923                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 924                     entries) {
 925                         pf_hash_rule(&ctx, rule);
 926                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 927                 }
 928         }
 929
 930         MD5Final(digest, &ctx);
 931         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
 932         return (0);
 933 }
 934
 935 static int
 936 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 937     sa_family_t af)
 938 {
 939         int error = 0;
 940
 941         switch (addr->type) {
 942         case PF_ADDR_TABLE:
 943                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
 944                 if (addr->p.tbl == NULL)
 945                         error = ENOMEM;
 946                 break;
 947         case PF_ADDR_DYNIFTL:
 948                 error = pfi_dynaddr_setup(addr, af);
 949                 break;
 950         }
 951
 952         return (error);
 953 }
 954
 955 static void
 956 pf_addr_copyout(struct pf_addr_wrap *addr)
 957 {
 958
 959         switch (addr->type) {
 960         case PF_ADDR_DYNIFTL:
 961                 pfi_dynaddr_copyout(addr);
 962                 break;
 963         case PF_ADDR_TABLE:
 964                 pf_tbladdr_copyout(addr);
 965                 break;
 966         }
 967 }
 968
 969 static int
 970 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
 971 {
 972         int                      error = 0;
 973
 974         /* XXX keep in sync with switch() below */
 975         if (securelevel_gt(td->td_ucred, 2))
 976                 switch (cmd) {
 977                 case DIOCGETRULES:
 978                 case DIOCGETRULE:
 979                 case DIOCGETADDRS:
 980                 case DIOCGETADDR:
 981                 case DIOCGETSTATE:
 982                 case DIOCSETSTATUSIF:
 983                 case DIOCGETSTATUS:
 984                 case DIOCCLRSTATUS:
 985                 case DIOCNATLOOK:
 986                 case DIOCSETDEBUG:
 987                 case DIOCGETSTATES:
 988                 case DIOCGETTIMEOUT:
 989                 case DIOCCLRRULECTRS:
 990                 case DIOCGETLIMIT:
 991                 case DIOCGETALTQS:
 992                 case DIOCGETALTQ:
 993                 case DIOCGETQSTATS:
 994                 case DIOCGETRULESETS:
 995                 case DIOCGETRULESET:
 996                 case DIOCRGETTABLES:
 997                 case DIOCRGETTSTATS:
 998                 case DIOCRCLRTSTATS:
 999                 case DIOCRCLRADDRS:
1000                 case DIOCRADDADDRS:
1001                 case DIOCRDELADDRS:
1002                 case DIOCRSETADDRS:
1003                 case DIOCRGETADDRS:
1004                 case DIOCRGETASTATS:
1005                 case DIOCRCLRASTATS:
1006                 case DIOCRTSTADDRS:
1007                 case DIOCOSFPGET:
1008                 case DIOCGETSRCNODES:
1009                 case DIOCCLRSRCNODES:
1010                 case DIOCIGETIFACES:
1011                 case DIOCGIFSPEED:
1012                 case DIOCSETIFFLAG:
1013                 case DIOCCLRIFFLAG:
1014                         break;
1015                 case DIOCRCLRTABLES:
1016                 case DIOCRADDTABLES:
1017                 case DIOCRDELTABLES:
1018                 case DIOCRSETTFLAGS:
1019                         if (((struct pfioc_table *)addr)->pfrio_flags &
1020                             PFR_FLAG_DUMMY)
1021                                 break; /* dummy operation ok */
1022                         return (EPERM);
1023                 default:
1024                         return (EPERM);
1025                 }
1026
1027         if (!(flags & FWRITE))
1028                 switch (cmd) {
1029                 case DIOCGETRULES:
1030                 case DIOCGETADDRS:
1031                 case DIOCGETADDR:
1032                 case DIOCGETSTATE:
1033                 case DIOCGETSTATUS:
1034                 case DIOCGETSTATES:
1035                 case DIOCGETTIMEOUT:
1036                 case DIOCGETLIMIT:
1037                 case DIOCGETALTQS:
1038                 case DIOCGETALTQ:
1039                 case DIOCGETQSTATS:
1040                 case DIOCGETRULESETS:
1041                 case DIOCGETRULESET:
1042                 case DIOCNATLOOK:
1043                 case DIOCRGETTABLES:
1044                 case DIOCRGETTSTATS:
1045                 case DIOCRGETADDRS:
1046                 case DIOCRGETASTATS:
1047                 case DIOCRTSTADDRS:
1048                 case DIOCOSFPGET:
1049                 case DIOCGETSRCNODES:
1050                 case DIOCIGETIFACES:
1051                 case DIOCGIFSPEED:
1052                         break;
1053                 case DIOCRCLRTABLES:
1054                 case DIOCRADDTABLES:
1055                 case DIOCRDELTABLES:
1056                 case DIOCRCLRTSTATS:
1057                 case DIOCRCLRADDRS:
1058                 case DIOCRADDADDRS:
1059                 case DIOCRDELADDRS:
1060                 case DIOCRSETADDRS:
1061                 case DIOCRSETTFLAGS:
1062                         if (((struct pfioc_table *)addr)->pfrio_flags &
1063                             PFR_FLAG_DUMMY) {
1064                                 flags |= FWRITE; /* need write lock for dummy */
1065                                 break; /* dummy operation ok */
1066                         }
1067                         return (EACCES);
1068                 case DIOCGETRULE:
1069                         if (((struct pfioc_rule *)addr)->action ==
1070                             PF_GET_CLR_CNTR)
1071                                 return (EACCES);
1072                         break;
1073                 default:
1074                         return (EACCES);
1075                 }
1076
1077         CURVNET_SET(TD_TO_VNET(td));
1078
1079         switch (cmd) {
1080         case DIOCSTART:
1081                 PF_RULES_WLOCK();
1082                 if (V_pf_status.running)
1083                         error = EEXIST;
1084                 else {
1085                         int cpu;
1086
1087                         PF_RULES_WUNLOCK();
1088                         error = hook_pf();
1089                         if (error) {
1090                                 DPFPRINTF(PF_DEBUG_MISC,
1091                                     ("pf: pfil registration failed\n"));
1092                                 break;
1093                         }
1094                         PF_RULES_WLOCK();
1095                         V_pf_status.running = 1;
1096                         V_pf_status.since = time_second;
1097
1098                         CPU_FOREACH(cpu)
1099                                 V_pf_stateid[cpu] = time_second;
1100
1101                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1102                 }
1103                 PF_RULES_WUNLOCK();
1104                 break;
1105
1106         case DIOCSTOP:
1107                 PF_RULES_WLOCK();
1108                 if (!V_pf_status.running)
1109                         error = ENOENT;
1110                 else {
1111                         V_pf_status.running = 0;
1112                         PF_RULES_WUNLOCK();
1113                         error = dehook_pf();
1114                         if (error) {
1115                                 V_pf_status.running = 1;
1116                                 DPFPRINTF(PF_DEBUG_MISC,
1117                                     ("pf: pfil unregistration failed\n"));
1118                         }
1119                         PF_RULES_WLOCK();
1120                         V_pf_status.since = time_second;
1121                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1122                 }
1123                 PF_RULES_WUNLOCK();
1124                 break;
1125
1126         case DIOCADDRULE: {
1127                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1128                 struct pf_ruleset       *ruleset;
1129                 struct pf_rule          *rule, *tail;
1130                 struct pf_pooladdr      *pa;
1131                 struct pfi_kif          *kif = NULL;
1132                 int                      rs_num;
1133
1134                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1135                         error = EINVAL;
1136                         break;
1137                 }
1138 #ifndef INET
1139                 if (pr->rule.af == AF_INET) {
1140                         error = EAFNOSUPPORT;
1141                         break;
1142                 }
1143 #endif /* INET */
1144 #ifndef INET6
1145                 if (pr->rule.af == AF_INET6) {
1146                         error = EAFNOSUPPORT;
1147                         break;
1148                 }
1149 #endif /* INET6 */
1150
1151                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1152                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1153                 if (rule->ifname[0])
1154                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1155                 rule->states_cur = counter_u64_alloc(M_WAITOK);
1156                 rule->states_tot = counter_u64_alloc(M_WAITOK);
1157                 rule->src_nodes = counter_u64_alloc(M_WAITOK);
1158                 rule->cuid = td->td_ucred->cr_ruid;
1159                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1160                 TAILQ_INIT(&rule->rpool.list);
1161
1162 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1163
1164                 PF_RULES_WLOCK();
1165                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1166                 ruleset = pf_find_ruleset(pr->anchor);
1167                 if (ruleset == NULL)
1168                         ERROUT(EINVAL);
1169                 rs_num = pf_get_ruleset_number(pr->rule.action);
1170                 if (rs_num >= PF_RULESET_MAX)
1171                         ERROUT(EINVAL);
1172                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1173                         DPFPRINTF(PF_DEBUG_MISC,
1174                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1175                             ruleset->rules[rs_num].inactive.ticket));
1176                         ERROUT(EBUSY);
1177                 }
1178                 if (pr->pool_ticket != V_ticket_pabuf) {
1179                         DPFPRINTF(PF_DEBUG_MISC,
1180                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
1181                             V_ticket_pabuf));
1182                         ERROUT(EBUSY);
1183                 }
1184
1185                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1186                     pf_rulequeue);
1187                 if (tail)
1188                         rule->nr = tail->nr + 1;
1189                 else
1190                         rule->nr = 0;
1191                 if (rule->ifname[0]) {
1192                         rule->kif = pfi_kif_attach(kif, rule->ifname);
1193                         pfi_kif_ref(rule->kif);
1194                 } else
1195                         rule->kif = NULL;
1196
1197                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1198                         error = EBUSY;
1199
1200 #ifdef ALTQ
1201                 /* set queue IDs */
1202                 if (rule->qname[0] != 0) {
1203                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1204                                 error = EBUSY;
1205                         else if (rule->pqname[0] != 0) {
1206                                 if ((rule->pqid =
1207                                     pf_qname2qid(rule->pqname)) == 0)
1208                                         error = EBUSY;
1209                         } else
1210                                 rule->pqid = rule->qid;
1211                 }
1212 #endif
1213                 if (rule->tagname[0])
1214                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1215                                 error = EBUSY;
1216                 if (rule->match_tagname[0])
1217                         if ((rule->match_tag =
1218                             pf_tagname2tag(rule->match_tagname)) == 0)
1219                                 error = EBUSY;
1220                 if (rule->rt && !rule->direction)
1221                         error = EINVAL;
1222                 if (!rule->log)
1223                         rule->logif = 0;
1224                 if (rule->logif >= PFLOGIFS_MAX)
1225                         error = EINVAL;
1226                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1227                         error = ENOMEM;
1228                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1229                         error = ENOMEM;
1230                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1231                         error = EINVAL;
1232                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1233                         if (pa->addr.type == PF_ADDR_TABLE) {
1234                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
1235                                     pa->addr.v.tblname);
1236                                 if (pa->addr.p.tbl == NULL)
1237                                         error = ENOMEM;
1238                         }
1239
1240                 if (rule->overload_tblname[0]) {
1241                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1242                             rule->overload_tblname)) == NULL)
1243                                 error = EINVAL;
1244                         else
1245                                 rule->overload_tbl->pfrkt_flags |=
1246                                     PFR_TFLAG_ACTIVE;
1247                 }
1248
1249                 pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1250                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1251                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1252                     (rule->rt > PF_FASTROUTE)) &&
1253                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1254                         error = EINVAL;
1255
1256                 if (error) {
1257                         pf_free_rule(rule);
1258                         PF_RULES_WUNLOCK();
1259                         break;
1260                 }
1261
1262                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1263                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1264                     rule->bytes[0] = rule->bytes[1] = 0;
1265                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1266                     rule, entries);
1267                 ruleset->rules[rs_num].inactive.rcount++;
1268                 PF_RULES_WUNLOCK();
1269                 break;
1270
1271 #undef ERROUT
1272 DIOCADDRULE_error:
1273                 PF_RULES_WUNLOCK();
1274                 counter_u64_free(rule->states_cur);
1275                 counter_u64_free(rule->states_tot);
1276                 counter_u64_free(rule->src_nodes);
1277                 free(rule, M_PFRULE);
1278                 if (kif)
1279                         free(kif, PFI_MTYPE);
1280                 break;
1281         }
1282
1283         case DIOCGETRULES: {
1284                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1285                 struct pf_ruleset       *ruleset;
1286                 struct pf_rule          *tail;
1287                 int                      rs_num;
1288
1289                 PF_RULES_WLOCK();
1290                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1291                 ruleset = pf_find_ruleset(pr->anchor);
1292                 if (ruleset == NULL) {
1293                         PF_RULES_WUNLOCK();
1294                         error = EINVAL;
1295                         break;
1296                 }
1297                 rs_num = pf_get_ruleset_number(pr->rule.action);
1298                 if (rs_num >= PF_RULESET_MAX) {
1299                         PF_RULES_WUNLOCK();
1300                         error = EINVAL;
1301                         break;
1302                 }
1303                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1304                     pf_rulequeue);
1305                 if (tail)
1306                         pr->nr = tail->nr + 1;
1307                 else
1308                         pr->nr = 0;
1309                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1310                 PF_RULES_WUNLOCK();
1311                 break;
1312         }
1313
1314         case DIOCGETRULE: {
1315                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1316                 struct pf_ruleset       *ruleset;
1317                 struct pf_rule          *rule;
1318                 int                      rs_num, i;
1319
1320                 PF_RULES_WLOCK();
1321                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1322                 ruleset = pf_find_ruleset(pr->anchor);
1323                 if (ruleset == NULL) {
1324                         PF_RULES_WUNLOCK();
1325                         error = EINVAL;
1326                         break;
1327                 }
1328                 rs_num = pf_get_ruleset_number(pr->rule.action);
1329                 if (rs_num >= PF_RULESET_MAX) {
1330                         PF_RULES_WUNLOCK();
1331                         error = EINVAL;
1332                         break;
1333                 }
1334                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1335                         PF_RULES_WUNLOCK();
1336                         error = EBUSY;
1337                         break;
1338                 }
1339                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1340                 while ((rule != NULL) && (rule->nr != pr->nr))
1341                         rule = TAILQ_NEXT(rule, entries);
1342                 if (rule == NULL) {
1343                         PF_RULES_WUNLOCK();
1344                         error = EBUSY;
1345                         break;
1346                 }
1347                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1348                 pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1349                 pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1350                 pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1351                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1352                         PF_RULES_WUNLOCK();
1353                         error = EBUSY;
1354                         break;
1355                 }
1356                 pf_addr_copyout(&pr->rule.src.addr);
1357                 pf_addr_copyout(&pr->rule.dst.addr);
1358                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1359                         if (rule->skip[i].ptr == NULL)
1360                                 pr->rule.skip[i].nr = -1;
1361                         else
1362                                 pr->rule.skip[i].nr =
1363                                     rule->skip[i].ptr->nr;
1364
1365                 if (pr->action == PF_GET_CLR_CNTR) {
1366                         rule->evaluations = 0;
1367                         rule->packets[0] = rule->packets[1] = 0;
1368                         rule->bytes[0] = rule->bytes[1] = 0;
1369                         counter_u64_zero(rule->states_tot);
1370                 }
1371                 PF_RULES_WUNLOCK();
1372                 break;
1373         }
1374
1375         case DIOCCHANGERULE: {
1376                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1377                 struct pf_ruleset       *ruleset;
1378                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1379                 struct pfi_kif          *kif = NULL;
1380                 struct pf_pooladdr      *pa;
1381                 u_int32_t                nr = 0;
1382                 int                      rs_num;
1383
1384                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1385                     pcr->action > PF_CHANGE_GET_TICKET) {
1386                         error = EINVAL;
1387                         break;
1388                 }
1389                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1390                         error = EINVAL;
1391                         break;
1392                 }
1393
1394                 if (pcr->action != PF_CHANGE_REMOVE) {
1395 #ifndef INET
1396                         if (pcr->rule.af == AF_INET) {
1397                                 error = EAFNOSUPPORT;
1398                                 break;
1399                         }
1400 #endif /* INET */
1401 #ifndef INET6
1402                         if (pcr->rule.af == AF_INET6) {
1403                                 error = EAFNOSUPPORT;
1404                                 break;
1405                         }
1406 #endif /* INET6 */
1407                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1408                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1409                         if (newrule->ifname[0])
1410                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1411                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
1412                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
1413                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1414                         newrule->cuid = td->td_ucred->cr_ruid;
1415                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1416                         TAILQ_INIT(&newrule->rpool.list);
1417                 }
1418
1419 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
1420
1421                 PF_RULES_WLOCK();
1422                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1423                     pcr->action == PF_CHANGE_GET_TICKET) &&
1424                     pcr->pool_ticket != V_ticket_pabuf)
1425                         ERROUT(EBUSY);
1426
1427                 ruleset = pf_find_ruleset(pcr->anchor);
1428                 if (ruleset == NULL)
1429                         ERROUT(EINVAL);
1430
1431                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1432                 if (rs_num >= PF_RULESET_MAX)
1433                         ERROUT(EINVAL);
1434
1435                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1436                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1437                         ERROUT(0);
1438                 } else if (pcr->ticket !=
1439                             ruleset->rules[rs_num].active.ticket)
1440                                 ERROUT(EINVAL);
1441
1442                 if (pcr->action != PF_CHANGE_REMOVE) {
1443                         if (newrule->ifname[0]) {
1444                                 newrule->kif = pfi_kif_attach(kif,
1445                                     newrule->ifname);
1446                                 pfi_kif_ref(newrule->kif);
1447                         } else
1448                                 newrule->kif = NULL;
1449
1450                         if (newrule->rtableid > 0 &&
1451                             newrule->rtableid >= rt_numfibs)
1452                                 error = EBUSY;
1453
1454 #ifdef ALTQ
1455                         /* set queue IDs */
1456                         if (newrule->qname[0] != 0) {
1457                                 if ((newrule->qid =
1458                                     pf_qname2qid(newrule->qname)) == 0)
1459                                         error = EBUSY;
1460                                 else if (newrule->pqname[0] != 0) {
1461                                         if ((newrule->pqid =
1462                                             pf_qname2qid(newrule->pqname)) == 0)
1463                                                 error = EBUSY;
1464                                 } else
1465                                         newrule->pqid = newrule->qid;
1466                         }
1467 #endif /* ALTQ */
1468                         if (newrule->tagname[0])
1469                                 if ((newrule->tag =
1470                                     pf_tagname2tag(newrule->tagname)) == 0)
1471                                         error = EBUSY;
1472                         if (newrule->match_tagname[0])
1473                                 if ((newrule->match_tag = pf_tagname2tag(
1474                                     newrule->match_tagname)) == 0)
1475                                         error = EBUSY;
1476                         if (newrule->rt && !newrule->direction)
1477                                 error = EINVAL;
1478                         if (!newrule->log)
1479                                 newrule->logif = 0;
1480                         if (newrule->logif >= PFLOGIFS_MAX)
1481                                 error = EINVAL;
1482                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1483                                 error = ENOMEM;
1484                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1485                                 error = ENOMEM;
1486                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1487                                 error = EINVAL;
1488                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1489                                 if (pa->addr.type == PF_ADDR_TABLE) {
1490                                         pa->addr.p.tbl =
1491                                             pfr_attach_table(ruleset,
1492                                             pa->addr.v.tblname);
1493                                         if (pa->addr.p.tbl == NULL)
1494                                                 error = ENOMEM;
1495                                 }
1496
1497                         if (newrule->overload_tblname[0]) {
1498                                 if ((newrule->overload_tbl = pfr_attach_table(
1499                                     ruleset, newrule->overload_tblname)) ==
1500                                     NULL)
1501                                         error = EINVAL;
1502                                 else
1503                                         newrule->overload_tbl->pfrkt_flags |=
1504                                             PFR_TFLAG_ACTIVE;
1505                         }
1506
1507                         pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1508                         if (((((newrule->action == PF_NAT) ||
1509                             (newrule->action == PF_RDR) ||
1510                             (newrule->action == PF_BINAT) ||
1511                             (newrule->rt > PF_FASTROUTE)) &&
1512                             !newrule->anchor)) &&
1513                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1514                                 error = EINVAL;
1515
1516                         if (error) {
1517                                 pf_free_rule(newrule);
1518                                 PF_RULES_WUNLOCK();
1519                                 break;
1520                         }
1521
1522                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1523                         newrule->evaluations = 0;
1524                         newrule->packets[0] = newrule->packets[1] = 0;
1525                         newrule->bytes[0] = newrule->bytes[1] = 0;
1526                 }
1527                 pf_empty_pool(&V_pf_pabuf);
1528
1529                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1530                         oldrule = TAILQ_FIRST(
1531                             ruleset->rules[rs_num].active.ptr);
1532                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1533                         oldrule = TAILQ_LAST(
1534                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1535                 else {
1536                         oldrule = TAILQ_FIRST(
1537                             ruleset->rules[rs_num].active.ptr);
1538                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1539                                 oldrule = TAILQ_NEXT(oldrule, entries);
1540                         if (oldrule == NULL) {
1541                                 if (newrule != NULL)
1542                                         pf_free_rule(newrule);
1543                                 PF_RULES_WUNLOCK();
1544                                 error = EINVAL;
1545                                 break;
1546                         }
1547                 }
1548
1549                 if (pcr->action == PF_CHANGE_REMOVE) {
1550                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1551                             oldrule);
1552                         ruleset->rules[rs_num].active.rcount--;
1553                 } else {
1554                         if (oldrule == NULL)
1555                                 TAILQ_INSERT_TAIL(
1556                                     ruleset->rules[rs_num].active.ptr,
1557                                     newrule, entries);
1558                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1559                             pcr->action == PF_CHANGE_ADD_BEFORE)
1560                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1561                         else
1562                                 TAILQ_INSERT_AFTER(
1563                                     ruleset->rules[rs_num].active.ptr,
1564                                     oldrule, newrule, entries);
1565                         ruleset->rules[rs_num].active.rcount++;
1566                 }
1567
1568                 nr = 0;
1569                 TAILQ_FOREACH(oldrule,
1570                     ruleset->rules[rs_num].active.ptr, entries)
1571                         oldrule->nr = nr++;
1572
1573                 ruleset->rules[rs_num].active.ticket++;
1574
1575                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1576                 pf_remove_if_empty_ruleset(ruleset);
1577
1578                 PF_RULES_WUNLOCK();
1579                 break;
1580
1581 #undef ERROUT
1582 DIOCCHANGERULE_error:
1583                 PF_RULES_WUNLOCK();
1584                 if (newrule != NULL) {
1585                         counter_u64_free(newrule->states_cur);
1586                         counter_u64_free(newrule->states_tot);
1587                         counter_u64_free(newrule->src_nodes);
1588                         free(newrule, M_PFRULE);
1589                 }
1590                 if (kif != NULL)
1591                         free(kif, PFI_MTYPE);
1592                 break;
1593         }
1594
1595         case DIOCCLRSTATES: {
1596                 struct pf_state         *s;
1597                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1598                 u_int                    i, killed = 0;
1599
1600                 for (i = 0; i <= V_pf_hashmask; i++) {
1601                         struct pf_idhash *ih = &V_pf_idhash[i];
1602
1603 relock_DIOCCLRSTATES:
1604                         PF_HASHROW_LOCK(ih);
1605                         LIST_FOREACH(s, &ih->states, entry)
1606                                 if (!psk->psk_ifname[0] ||
1607                                     !strcmp(psk->psk_ifname,
1608                                     s->kif->pfik_name)) {
1609                                         /*
1610                                          * Don't send out individual
1611                                          * delete messages.
1612                                          */
1613                                         s->state_flags |= PFSTATE_NOSYNC;
1614                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1615                                         killed++;
1616                                         goto relock_DIOCCLRSTATES;
1617                                 }
1618                         PF_HASHROW_UNLOCK(ih);
1619                 }
1620                 psk->psk_killed = killed;
1621                 if (pfsync_clear_states_ptr != NULL)
1622                         pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1623                 break;
1624         }
1625
1626         case DIOCKILLSTATES: {
1627                 struct pf_state         *s;
1628                 struct pf_state_key     *sk;
1629                 struct pf_addr          *srcaddr, *dstaddr;
1630                 u_int16_t                srcport, dstport;
1631                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1632                 u_int                    i, killed = 0;
1633
1634                 if (psk->psk_pfcmp.id) {
1635                         if (psk->psk_pfcmp.creatorid == 0)
1636                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1637                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1638                             psk->psk_pfcmp.creatorid))) {
1639                                 pf_unlink_state(s, PF_ENTER_LOCKED);
1640                                 psk->psk_killed = 1;
1641                         }
1642                         break;
1643                 }
1644
1645                 for (i = 0; i <= V_pf_hashmask; i++) {
1646                         struct pf_idhash *ih = &V_pf_idhash[i];
1647
1648 relock_DIOCKILLSTATES:
1649                         PF_HASHROW_LOCK(ih);
1650                         LIST_FOREACH(s, &ih->states, entry) {
1651                                 sk = s->key[PF_SK_WIRE];
1652                                 if (s->direction == PF_OUT) {
1653                                         srcaddr = &sk->addr[1];
1654                                         dstaddr = &sk->addr[0];
1655                                         srcport = sk->port[0];
1656                                         dstport = sk->port[0];
1657                                 } else {
1658                                         srcaddr = &sk->addr[0];
1659                                         dstaddr = &sk->addr[1];
1660                                         srcport = sk->port[0];
1661                                         dstport = sk->port[0];
1662                                 }
1663
1664                                 if ((!psk->psk_af || sk->af == psk->psk_af)
1665                                     && (!psk->psk_proto || psk->psk_proto ==
1666                                     sk->proto) &&
1667                                     PF_MATCHA(psk->psk_src.neg,
1668                                     &psk->psk_src.addr.v.a.addr,
1669                                     &psk->psk_src.addr.v.a.mask,
1670                                     srcaddr, sk->af) &&
1671                                     PF_MATCHA(psk->psk_dst.neg,
1672                                     &psk->psk_dst.addr.v.a.addr,
1673                                     &psk->psk_dst.addr.v.a.mask,
1674                                     dstaddr, sk->af) &&
1675                                     (psk->psk_src.port_op == 0 ||
1676                                     pf_match_port(psk->psk_src.port_op,
1677                                     psk->psk_src.port[0], psk->psk_src.port[1],
1678                                     srcport)) &&
1679                                     (psk->psk_dst.port_op == 0 ||
1680                                     pf_match_port(psk->psk_dst.port_op,
1681                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
1682                                     dstport)) &&
1683                                     (!psk->psk_label[0] ||
1684                                     (s->rule.ptr->label[0] &&
1685                                     !strcmp(psk->psk_label,
1686                                     s->rule.ptr->label))) &&
1687                                     (!psk->psk_ifname[0] ||
1688                                     !strcmp(psk->psk_ifname,
1689                                     s->kif->pfik_name))) {
1690                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1691                                         killed++;
1692                                         goto relock_DIOCKILLSTATES;
1693                                 }
1694                         }
1695                         PF_HASHROW_UNLOCK(ih);
1696                 }
1697                 psk->psk_killed = killed;
1698                 break;
1699         }
1700
1701         case DIOCADDSTATE: {
1702                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1703                 struct pfsync_state     *sp = &ps->state;
1704
1705                 if (sp->timeout >= PFTM_MAX) {
1706                         error = EINVAL;
1707                         break;
1708                 }
1709                 if (pfsync_state_import_ptr != NULL) {
1710                         PF_RULES_RLOCK();
1711                         error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1712                         PF_RULES_RUNLOCK();
1713                 } else
1714                         error = EOPNOTSUPP;
1715                 break;
1716         }
1717
1718         case DIOCGETSTATE: {
1719                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1720                 struct pf_state         *s;
1721
1722                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1723                 if (s == NULL) {
1724                         error = ENOENT;
1725                         break;
1726                 }
1727
1728                 pfsync_state_export(&ps->state, s);
1729                 PF_STATE_UNLOCK(s);
1730                 break;
1731         }
1732
1733         case DIOCGETSTATES: {
1734                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1735                 struct pf_state         *s;
1736                 struct pfsync_state     *pstore, *p;
1737                 int i, nr;
1738
1739                 if (ps->ps_len == 0) {
1740                         nr = uma_zone_get_cur(V_pf_state_z);
1741                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1742                         break;
1743                 }
1744
1745                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1746                 nr = 0;
1747
1748                 for (i = 0; i <= V_pf_hashmask; i++) {
1749                         struct pf_idhash *ih = &V_pf_idhash[i];
1750
1751                         PF_HASHROW_LOCK(ih);
1752                         LIST_FOREACH(s, &ih->states, entry) {
1753
1754                                 if (s->timeout == PFTM_UNLINKED)
1755                                         continue;
1756
1757                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
1758                                         PF_HASHROW_UNLOCK(ih);
1759                                         goto DIOCGETSTATES_full;
1760                                 }
1761                                 pfsync_state_export(p, s);
1762                                 p++;
1763                                 nr++;
1764                         }
1765                         PF_HASHROW_UNLOCK(ih);
1766                 }
1767 DIOCGETSTATES_full:
1768                 error = copyout(pstore, ps->ps_states,
1769                     sizeof(struct pfsync_state) * nr);
1770                 if (error) {
1771                         free(pstore, M_TEMP);
1772                         break;
1773                 }
1774                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1775                 free(pstore, M_TEMP);
1776
1777                 break;
1778         }
1779
1780         case DIOCGETSTATUS: {
1781                 struct pf_status *s = (struct pf_status *)addr;
1782                 PF_RULES_RLOCK();
1783                 bcopy(&V_pf_status, s, sizeof(struct pf_status));
1784                 pfi_update_status(s->ifname, s);
1785                 PF_RULES_RUNLOCK();
1786                 break;
1787         }
1788
1789         case DIOCSETSTATUSIF: {
1790                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1791
1792                 if (pi->ifname[0] == 0) {
1793                         bzero(V_pf_status.ifname, IFNAMSIZ);
1794                         break;
1795                 }
1796                 PF_RULES_WLOCK();
1797                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1798                 PF_RULES_WUNLOCK();
1799                 break;
1800         }
1801
1802         case DIOCCLRSTATUS: {
1803                 PF_RULES_WLOCK();
1804                 bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1805                 bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1806                 bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1807                 V_pf_status.since = time_second;
1808                 if (*V_pf_status.ifname)
1809                         pfi_update_status(V_pf_status.ifname, NULL);
1810                 PF_RULES_WUNLOCK();
1811                 break;
1812         }
1813
1814         case DIOCNATLOOK: {
1815                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1816                 struct pf_state_key     *sk;
1817                 struct pf_state         *state;
1818                 struct pf_state_key_cmp  key;
1819                 int                      m = 0, direction = pnl->direction;
1820                 int                      sidx, didx;
1821
1822                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1823                 sidx = (direction == PF_IN) ? 1 : 0;
1824                 didx = (direction == PF_IN) ? 0 : 1;
1825
1826                 if (!pnl->proto ||
1827                     PF_AZERO(&pnl->saddr, pnl->af) ||
1828                     PF_AZERO(&pnl->daddr, pnl->af) ||
1829                     ((pnl->proto == IPPROTO_TCP ||
1830                     pnl->proto == IPPROTO_UDP) &&
1831                     (!pnl->dport || !pnl->sport)))
1832                         error = EINVAL;
1833                 else {
1834                         bzero(&key, sizeof(key));
1835                         key.af = pnl->af;
1836                         key.proto = pnl->proto;
1837                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1838                         key.port[sidx] = pnl->sport;
1839                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1840                         key.port[didx] = pnl->dport;
1841
1842                         state = pf_find_state_all(&key, direction, &m);
1843
1844                         if (m > 1)
1845                                 error = E2BIG;  /* more than one state */
1846                         else if (state != NULL) {
1847                                 /* XXXGL: not locked read */
1848                                 sk = state->key[sidx];
1849                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1850                                 pnl->rsport = sk->port[sidx];
1851                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1852                                 pnl->rdport = sk->port[didx];
1853                         } else
1854                                 error = ENOENT;
1855                 }
1856                 break;
1857         }
1858
1859         case DIOCSETTIMEOUT: {
1860                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1861                 int              old;
1862
1863                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1864                     pt->seconds < 0) {
1865                         error = EINVAL;
1866                         break;
1867                 }
1868                 PF_RULES_WLOCK();
1869                 old = V_pf_default_rule.timeout[pt->timeout];
1870                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1871                         pt->seconds = 1;
1872                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1873                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1874                         wakeup(pf_purge_thread);
1875                 pt->seconds = old;
1876                 PF_RULES_WUNLOCK();
1877                 break;
1878         }
1879
1880         case DIOCGETTIMEOUT: {
1881                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1882
1883                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1884                         error = EINVAL;
1885                         break;
1886                 }
1887                 PF_RULES_RLOCK();
1888                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1889                 PF_RULES_RUNLOCK();
1890                 break;
1891         }
1892
1893         case DIOCGETLIMIT: {
1894                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1895
1896                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1897                         error = EINVAL;
1898                         break;
1899                 }
1900                 PF_RULES_RLOCK();
1901                 pl->limit = V_pf_limits[pl->index].limit;
1902                 PF_RULES_RUNLOCK();
1903                 break;
1904         }
1905
1906         case DIOCSETLIMIT: {
1907                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1908                 int                      old_limit;
1909
1910                 PF_RULES_WLOCK();
1911                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1912                     V_pf_limits[pl->index].zone == NULL) {
1913                         PF_RULES_WUNLOCK();
1914                         error = EINVAL;
1915                         break;
1916                 }
1917                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1918                 old_limit = V_pf_limits[pl->index].limit;
1919                 V_pf_limits[pl->index].limit = pl->limit;
1920                 pl->limit = old_limit;
1921                 PF_RULES_WUNLOCK();
1922                 break;
1923         }
1924
1925         case DIOCSETDEBUG: {
1926                 u_int32_t       *level = (u_int32_t *)addr;
1927
1928                 PF_RULES_WLOCK();
1929                 V_pf_status.debug = *level;
1930                 PF_RULES_WUNLOCK();
1931                 break;
1932         }
1933
1934         case DIOCCLRRULECTRS: {
1935                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1936                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1937                 struct pf_rule          *rule;
1938
1939                 PF_RULES_WLOCK();
1940                 TAILQ_FOREACH(rule,
1941                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1942                         rule->evaluations = 0;
1943                         rule->packets[0] = rule->packets[1] = 0;
1944                         rule->bytes[0] = rule->bytes[1] = 0;
1945                 }
1946                 PF_RULES_WUNLOCK();
1947                 break;
1948         }
1949
1950         case DIOCGIFSPEED: {
1951                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1952                 struct pf_ifspeed       ps;
1953                 struct ifnet            *ifp;
1954
1955                 if (psp->ifname[0] != 0) {
1956                         /* Can we completely trust user-land? */
1957                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1958                         ifp = ifunit(ps.ifname);
1959                         if (ifp != NULL)
1960                                 psp->baudrate = ifp->if_baudrate;
1961                         else
1962                                 error = EINVAL;
1963                 } else
1964                         error = EINVAL;
1965                 break;
1966         }
1967
1968 #ifdef ALTQ
1969         case DIOCSTARTALTQ: {
1970                 struct pf_altq          *altq;
1971
1972                 PF_RULES_WLOCK();
1973                 /* enable all altq interfaces on active list */
1974                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1975                         if (altq->qname[0] == 0 && (altq->local_flags &
1976                             PFALTQ_FLAG_IF_REMOVED) == 0) {
1977                                 error = pf_enable_altq(altq);
1978                                 if (error != 0)
1979                                         break;
1980                         }
1981                 }
1982                 if (error == 0)
1983                         V_pf_altq_running = 1;
1984                 PF_RULES_WUNLOCK();
1985                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1986                 break;
1987         }
1988
1989         case DIOCSTOPALTQ: {
1990                 struct pf_altq          *altq;
1991
1992                 PF_RULES_WLOCK();
1993                 /* disable all altq interfaces on active list */
1994                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1995                         if (altq->qname[0] == 0 && (altq->local_flags &
1996                             PFALTQ_FLAG_IF_REMOVED) == 0) {
1997                                 error = pf_disable_altq(altq);
1998                                 if (error != 0)
1999                                         break;
2000                         }
2001                 }
2002                 if (error == 0)
2003                         V_pf_altq_running = 0;
2004                 PF_RULES_WUNLOCK();
2005                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2006                 break;
2007         }
2008
2009         case DIOCADDALTQ: {
2010                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2011                 struct pf_altq          *altq, *a;
2012                 struct ifnet            *ifp;
2013
2014                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2015                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2016                 altq->local_flags = 0;
2017
2018                 PF_RULES_WLOCK();
2019                 if (pa->ticket != V_ticket_altqs_inactive) {
2020                         PF_RULES_WUNLOCK();
2021                         free(altq, M_PFALTQ);
2022                         error = EBUSY;
2023                         break;
2024                 }
2025
2026                 /*
2027                  * if this is for a queue, find the discipline and
2028                  * copy the necessary fields
2029                  */
2030                 if (altq->qname[0] != 0) {
2031                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2032                                 PF_RULES_WUNLOCK();
2033                                 error = EBUSY;
2034                                 free(altq, M_PFALTQ);
2035                                 break;
2036                         }
2037                         altq->altq_disc = NULL;
2038                         TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2039                                 if (strncmp(a->ifname, altq->ifname,
2040                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
2041                                         altq->altq_disc = a->altq_disc;
2042                                         break;
2043                                 }
2044                         }
2045                 }
2046
2047                 if ((ifp = ifunit(altq->ifname)) == NULL)
2048                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2049                 else
2050                         error = altq_add(altq);
2051
2052                 if (error) {
2053                         PF_RULES_WUNLOCK();
2054                         free(altq, M_PFALTQ);
2055                         break;
2056                 }
2057
2058                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2059                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2060                 PF_RULES_WUNLOCK();
2061                 break;
2062         }
2063
2064         case DIOCGETALTQS: {
2065                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2066                 struct pf_altq          *altq;
2067
2068                 PF_RULES_RLOCK();
2069                 pa->nr = 0;
2070                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2071                         pa->nr++;
2072                 pa->ticket = V_ticket_altqs_active;
2073                 PF_RULES_RUNLOCK();
2074                 break;
2075         }
2076
2077         case DIOCGETALTQ: {
2078                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2079                 struct pf_altq          *altq;
2080                 u_int32_t                nr;
2081
2082                 PF_RULES_RLOCK();
2083                 if (pa->ticket != V_ticket_altqs_active) {
2084                         PF_RULES_RUNLOCK();
2085                         error = EBUSY;
2086                         break;
2087                 }
2088                 nr = 0;
2089                 altq = TAILQ_FIRST(V_pf_altqs_active);
2090                 while ((altq != NULL) && (nr < pa->nr)) {
2091                         altq = TAILQ_NEXT(altq, entries);
2092                         nr++;
2093                 }
2094                 if (altq == NULL) {
2095                         PF_RULES_RUNLOCK();
2096                         error = EBUSY;
2097                         break;
2098                 }
2099                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2100                 PF_RULES_RUNLOCK();
2101                 break;
2102         }
2103
2104         case DIOCCHANGEALTQ:
2105                 /* CHANGEALTQ not supported yet! */
2106                 error = ENODEV;
2107                 break;
2108
2109         case DIOCGETQSTATS: {
2110                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
2111                 struct pf_altq          *altq;
2112                 u_int32_t                nr;
2113                 int                      nbytes;
2114
2115                 PF_RULES_RLOCK();
2116                 if (pq->ticket != V_ticket_altqs_active) {
2117                         PF_RULES_RUNLOCK();
2118                         error = EBUSY;
2119                         break;
2120                 }
2121                 nbytes = pq->nbytes;
2122                 nr = 0;
2123                 altq = TAILQ_FIRST(V_pf_altqs_active);
2124                 while ((altq != NULL) && (nr < pq->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
2134                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2135                         PF_RULES_RUNLOCK();
2136                         error = ENXIO;
2137                         break;
2138                 }
2139                 PF_RULES_RUNLOCK();
2140                 error = altq_getqstats(altq, pq->buf, &nbytes);
2141                 if (error == 0) {
2142                         pq->scheduler = altq->scheduler;
2143                         pq->nbytes = nbytes;
2144                 }
2145                 break;
2146         }
2147 #endif /* ALTQ */
2148
2149         case DIOCBEGINADDRS: {
2150                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2151
2152                 PF_RULES_WLOCK();
2153                 pf_empty_pool(&V_pf_pabuf);
2154                 pp->ticket = ++V_ticket_pabuf;
2155                 PF_RULES_WUNLOCK();
2156                 break;
2157         }
2158
2159         case DIOCADDADDR: {
2160                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2161                 struct pf_pooladdr      *pa;
2162                 struct pfi_kif          *kif = NULL;
2163
2164 #ifndef INET
2165                 if (pp->af == AF_INET) {
2166                         error = EAFNOSUPPORT;
2167                         break;
2168                 }
2169 #endif /* INET */
2170 #ifndef INET6
2171                 if (pp->af == AF_INET6) {
2172                         error = EAFNOSUPPORT;
2173                         break;
2174                 }
2175 #endif /* INET6 */
2176                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2177                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2178                     pp->addr.addr.type != PF_ADDR_TABLE) {
2179                         error = EINVAL;
2180                         break;
2181                 }
2182                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2183                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2184                 if (pa->ifname[0])
2185                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2186                 PF_RULES_WLOCK();
2187                 if (pp->ticket != V_ticket_pabuf) {
2188                         PF_RULES_WUNLOCK();
2189                         if (pa->ifname[0])
2190                                 free(kif, PFI_MTYPE);
2191                         free(pa, M_PFRULE);
2192                         error = EBUSY;
2193                         break;
2194                 }
2195                 if (pa->ifname[0]) {
2196                         pa->kif = pfi_kif_attach(kif, pa->ifname);
2197                         pfi_kif_ref(pa->kif);
2198                 } else
2199                         pa->kif = NULL;
2200                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2201                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2202                         if (pa->ifname[0])
2203                                 pfi_kif_unref(pa->kif);
2204                         PF_RULES_WUNLOCK();
2205                         free(pa, M_PFRULE);
2206                         break;
2207                 }
2208                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2209                 PF_RULES_WUNLOCK();
2210                 break;
2211         }
2212
2213         case DIOCGETADDRS: {
2214                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2215                 struct pf_pool          *pool;
2216                 struct pf_pooladdr      *pa;
2217
2218                 PF_RULES_RLOCK();
2219                 pp->nr = 0;
2220                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2221                     pp->r_num, 0, 1, 0);
2222                 if (pool == NULL) {
2223                         PF_RULES_RUNLOCK();
2224                         error = EBUSY;
2225                         break;
2226                 }
2227                 TAILQ_FOREACH(pa, &pool->list, entries)
2228                         pp->nr++;
2229                 PF_RULES_RUNLOCK();
2230                 break;
2231         }
2232
2233         case DIOCGETADDR: {
2234                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2235                 struct pf_pool          *pool;
2236                 struct pf_pooladdr      *pa;
2237                 u_int32_t                nr = 0;
2238
2239                 PF_RULES_RLOCK();
2240                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2241                     pp->r_num, 0, 1, 1);
2242                 if (pool == NULL) {
2243                         PF_RULES_RUNLOCK();
2244                         error = EBUSY;
2245                         break;
2246                 }
2247                 pa = TAILQ_FIRST(&pool->list);
2248                 while ((pa != NULL) && (nr < pp->nr)) {
2249                         pa = TAILQ_NEXT(pa, entries);
2250                         nr++;
2251                 }
2252                 if (pa == NULL) {
2253                         PF_RULES_RUNLOCK();
2254                         error = EBUSY;
2255                         break;
2256                 }
2257                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2258                 pf_addr_copyout(&pp->addr.addr);
2259                 PF_RULES_RUNLOCK();
2260                 break;
2261         }
2262
2263         case DIOCCHANGEADDR: {
2264                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2265                 struct pf_pool          *pool;
2266                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2267                 struct pf_ruleset       *ruleset;
2268                 struct pfi_kif          *kif = NULL;
2269
2270                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2271                     pca->action > PF_CHANGE_REMOVE) {
2272                         error = EINVAL;
2273                         break;
2274                 }
2275                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2276                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2277                     pca->addr.addr.type != PF_ADDR_TABLE) {
2278                         error = EINVAL;
2279                         break;
2280                 }
2281
2282                 if (pca->action != PF_CHANGE_REMOVE) {
2283 #ifndef INET
2284                         if (pca->af == AF_INET) {
2285                                 error = EAFNOSUPPORT;
2286                                 break;
2287                         }
2288 #endif /* INET */
2289 #ifndef INET6
2290                         if (pca->af == AF_INET6) {
2291                                 error = EAFNOSUPPORT;
2292                                 break;
2293                         }
2294 #endif /* INET6 */
2295                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2296                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2297                         if (newpa->ifname[0])
2298                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2299                         newpa->kif = NULL;
2300                 }
2301
2302 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
2303                 PF_RULES_WLOCK();
2304                 ruleset = pf_find_ruleset(pca->anchor);
2305                 if (ruleset == NULL)
2306                         ERROUT(EBUSY);
2307
2308                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2309                     pca->r_num, pca->r_last, 1, 1);
2310                 if (pool == NULL)
2311                         ERROUT(EBUSY);
2312
2313                 if (pca->action != PF_CHANGE_REMOVE) {
2314                         if (newpa->ifname[0]) {
2315                                 newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2316                                 pfi_kif_ref(newpa->kif);
2317                                 kif = NULL;
2318                         }
2319
2320                         switch (newpa->addr.type) {
2321                         case PF_ADDR_DYNIFTL:
2322                                 error = pfi_dynaddr_setup(&newpa->addr,
2323                                     pca->af);
2324                                 break;
2325                         case PF_ADDR_TABLE:
2326                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
2327                                     newpa->addr.v.tblname);
2328                                 if (newpa->addr.p.tbl == NULL)
2329                                         error = ENOMEM;
2330                                 break;
2331                         }
2332                         if (error)
2333                                 goto DIOCCHANGEADDR_error;
2334                 }
2335
2336                 switch (pca->action) {
2337                 case PF_CHANGE_ADD_HEAD:
2338                         oldpa = TAILQ_FIRST(&pool->list);
2339                         break;
2340                 case PF_CHANGE_ADD_TAIL:
2341                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2342                         break;
2343                 default:
2344                         oldpa = TAILQ_FIRST(&pool->list);
2345                         for (int i = 0; oldpa && i < pca->nr; i++)
2346                                 oldpa = TAILQ_NEXT(oldpa, entries);
2347
2348                         if (oldpa == NULL)
2349                                 ERROUT(EINVAL);
2350                 }
2351
2352                 if (pca->action == PF_CHANGE_REMOVE) {
2353                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2354                         switch (oldpa->addr.type) {
2355                         case PF_ADDR_DYNIFTL:
2356                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
2357                                 break;
2358                         case PF_ADDR_TABLE:
2359                                 pfr_detach_table(oldpa->addr.p.tbl);
2360                                 break;
2361                         }
2362                         if (oldpa->kif)
2363                                 pfi_kif_unref(oldpa->kif);
2364                         free(oldpa, M_PFRULE);
2365                 } else {
2366                         if (oldpa == NULL)
2367                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2368                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2369                             pca->action == PF_CHANGE_ADD_BEFORE)
2370                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2371                         else
2372                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2373                                     newpa, entries);
2374                 }
2375
2376                 pool->cur = TAILQ_FIRST(&pool->list);
2377                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2378                 PF_RULES_WUNLOCK();
2379                 break;
2380
2381 #undef ERROUT
2382 DIOCCHANGEADDR_error:
2383                 if (newpa->kif)
2384                         pfi_kif_unref(newpa->kif);
2385                 PF_RULES_WUNLOCK();
2386                 if (newpa != NULL)
2387                         free(newpa, M_PFRULE);
2388                 if (kif != NULL)
2389                         free(kif, PFI_MTYPE);
2390                 break;
2391         }
2392
2393         case DIOCGETRULESETS: {
2394                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2395                 struct pf_ruleset       *ruleset;
2396                 struct pf_anchor        *anchor;
2397
2398                 PF_RULES_RLOCK();
2399                 pr->path[sizeof(pr->path) - 1] = 0;
2400                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2401                         PF_RULES_RUNLOCK();
2402                         error = ENOENT;
2403                         break;
2404                 }
2405                 pr->nr = 0;
2406                 if (ruleset->anchor == NULL) {
2407                         /* XXX kludge for pf_main_ruleset */
2408                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2409                                 if (anchor->parent == NULL)
2410                                         pr->nr++;
2411                 } else {
2412                         RB_FOREACH(anchor, pf_anchor_node,
2413                             &ruleset->anchor->children)
2414                                 pr->nr++;
2415                 }
2416                 PF_RULES_RUNLOCK();
2417                 break;
2418         }
2419
2420         case DIOCGETRULESET: {
2421                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2422                 struct pf_ruleset       *ruleset;
2423                 struct pf_anchor        *anchor;
2424                 u_int32_t                nr = 0;
2425
2426                 PF_RULES_RLOCK();
2427                 pr->path[sizeof(pr->path) - 1] = 0;
2428                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2429                         PF_RULES_RUNLOCK();
2430                         error = ENOENT;
2431                         break;
2432                 }
2433                 pr->name[0] = 0;
2434                 if (ruleset->anchor == NULL) {
2435                         /* XXX kludge for pf_main_ruleset */
2436                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2437                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2438                                         strlcpy(pr->name, anchor->name,
2439                                             sizeof(pr->name));
2440                                         break;
2441                                 }
2442                 } else {
2443                         RB_FOREACH(anchor, pf_anchor_node,
2444                             &ruleset->anchor->children)
2445                                 if (nr++ == pr->nr) {
2446                                         strlcpy(pr->name, anchor->name,
2447                                             sizeof(pr->name));
2448                                         break;
2449                                 }
2450                 }
2451                 if (!pr->name[0])
2452                         error = EBUSY;
2453                 PF_RULES_RUNLOCK();
2454                 break;
2455         }
2456
2457         case DIOCRCLRTABLES: {
2458                 struct pfioc_table *io = (struct pfioc_table *)addr;
2459
2460                 if (io->pfrio_esize != 0) {
2461                         error = ENODEV;
2462                         break;
2463                 }
2464                 PF_RULES_WLOCK();
2465                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2466                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2467                 PF_RULES_WUNLOCK();
2468                 break;
2469         }
2470
2471         case DIOCRADDTABLES: {
2472                 struct pfioc_table *io = (struct pfioc_table *)addr;
2473                 struct pfr_table *pfrts;
2474                 size_t totlen;
2475
2476                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2477                         error = ENODEV;
2478                         break;
2479                 }
2480                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2481                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2482                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2483                 if (error) {
2484                         free(pfrts, M_TEMP);
2485                         break;
2486                 }
2487                 PF_RULES_WLOCK();
2488                 error = pfr_add_tables(pfrts, io->pfrio_size,
2489                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2490                 PF_RULES_WUNLOCK();
2491                 free(pfrts, M_TEMP);
2492                 break;
2493         }
2494
2495         case DIOCRDELTABLES: {
2496                 struct pfioc_table *io = (struct pfioc_table *)addr;
2497                 struct pfr_table *pfrts;
2498                 size_t totlen;
2499
2500                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2501                         error = ENODEV;
2502                         break;
2503                 }
2504                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2505                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2506                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2507                 if (error) {
2508                         free(pfrts, M_TEMP);
2509                         break;
2510                 }
2511                 PF_RULES_WLOCK();
2512                 error = pfr_del_tables(pfrts, io->pfrio_size,
2513                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2514                 PF_RULES_WUNLOCK();
2515                 free(pfrts, M_TEMP);
2516                 break;
2517         }
2518
2519         case DIOCRGETTABLES: {
2520                 struct pfioc_table *io = (struct pfioc_table *)addr;
2521                 struct pfr_table *pfrts;
2522                 size_t totlen;
2523
2524                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2525                         error = ENODEV;
2526                         break;
2527                 }
2528                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2529                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2530                 PF_RULES_RLOCK();
2531                 error = pfr_get_tables(&io->pfrio_table, pfrts,
2532                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2533                 PF_RULES_RUNLOCK();
2534                 if (error == 0)
2535                         error = copyout(pfrts, io->pfrio_buffer, totlen);
2536                 free(pfrts, M_TEMP);
2537                 break;
2538         }
2539
2540         case DIOCRGETTSTATS: {
2541                 struct pfioc_table *io = (struct pfioc_table *)addr;
2542                 struct pfr_tstats *pfrtstats;
2543                 size_t totlen;
2544
2545                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2546                         error = ENODEV;
2547                         break;
2548                 }
2549                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2550                 pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2551                 PF_RULES_WLOCK();
2552                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2553                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2554                 PF_RULES_WUNLOCK();
2555                 if (error == 0)
2556                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2557                 free(pfrtstats, M_TEMP);
2558                 break;
2559         }
2560
2561         case DIOCRCLRTSTATS: {
2562                 struct pfioc_table *io = (struct pfioc_table *)addr;
2563                 struct pfr_table *pfrts;
2564                 size_t totlen;
2565
2566                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2567                         error = ENODEV;
2568                         break;
2569                 }
2570                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2571                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2572                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2573                 if (error) {
2574                         free(pfrts, M_TEMP);
2575                         break;
2576                 }
2577                 PF_RULES_WLOCK();
2578                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
2579                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2580                 PF_RULES_WUNLOCK();
2581                 free(pfrts, M_TEMP);
2582                 break;
2583         }
2584
2585         case DIOCRSETTFLAGS: {
2586                 struct pfioc_table *io = (struct pfioc_table *)addr;
2587                 struct pfr_table *pfrts;
2588                 size_t totlen;
2589
2590                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2591                         error = ENODEV;
2592                         break;
2593                 }
2594                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2595                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2596                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2597                 if (error) {
2598                         free(pfrts, M_TEMP);
2599                         break;
2600                 }
2601                 PF_RULES_WLOCK();
2602                 error = pfr_set_tflags(pfrts, io->pfrio_size,
2603                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2604                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2605                 PF_RULES_WUNLOCK();
2606                 free(pfrts, M_TEMP);
2607                 break;
2608         }
2609
2610         case DIOCRCLRADDRS: {
2611                 struct pfioc_table *io = (struct pfioc_table *)addr;
2612
2613                 if (io->pfrio_esize != 0) {
2614                         error = ENODEV;
2615                         break;
2616                 }
2617                 PF_RULES_WLOCK();
2618                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2619                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2620                 PF_RULES_WUNLOCK();
2621                 break;
2622         }
2623
2624         case DIOCRADDADDRS: {
2625                 struct pfioc_table *io = (struct pfioc_table *)addr;
2626                 struct pfr_addr *pfras;
2627                 size_t totlen;
2628
2629                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2630                         error = ENODEV;
2631                         break;
2632                 }
2633                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2634                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2635                 error = copyin(io->pfrio_buffer, pfras, totlen);
2636                 if (error) {
2637                         free(pfras, M_TEMP);
2638                         break;
2639                 }
2640                 PF_RULES_WLOCK();
2641                 error = pfr_add_addrs(&io->pfrio_table, pfras,
2642                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2643                     PFR_FLAG_USERIOCTL);
2644                 PF_RULES_WUNLOCK();
2645                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2646                         error = copyout(pfras, io->pfrio_buffer, totlen);
2647                 free(pfras, M_TEMP);
2648                 break;
2649         }
2650
2651         case DIOCRDELADDRS: {
2652                 struct pfioc_table *io = (struct pfioc_table *)addr;
2653                 struct pfr_addr *pfras;
2654                 size_t totlen;
2655
2656                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2657                         error = ENODEV;
2658                         break;
2659                 }
2660                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2661                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2662                 error = copyin(io->pfrio_buffer, pfras, totlen);
2663                 if (error) {
2664                         free(pfras, M_TEMP);
2665                         break;
2666                 }
2667                 PF_RULES_WLOCK();
2668                 error = pfr_del_addrs(&io->pfrio_table, pfras,
2669                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2670                     PFR_FLAG_USERIOCTL);
2671                 PF_RULES_WUNLOCK();
2672                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2673                         error = copyout(pfras, io->pfrio_buffer, totlen);
2674                 free(pfras, M_TEMP);
2675                 break;
2676         }
2677
2678         case DIOCRSETADDRS: {
2679                 struct pfioc_table *io = (struct pfioc_table *)addr;
2680                 struct pfr_addr *pfras;
2681                 size_t totlen;
2682
2683                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2684                         error = ENODEV;
2685                         break;
2686                 }
2687                 totlen = (io->pfrio_size + io->pfrio_size2) *
2688                     sizeof(struct pfr_addr);
2689                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2690                 error = copyin(io->pfrio_buffer, pfras, totlen);
2691                 if (error) {
2692                         free(pfras, M_TEMP);
2693                         break;
2694                 }
2695                 PF_RULES_WLOCK();
2696                 error = pfr_set_addrs(&io->pfrio_table, pfras,
2697                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2698                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2699                     PFR_FLAG_USERIOCTL, 0);
2700                 PF_RULES_WUNLOCK();
2701                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2702                         error = copyout(pfras, io->pfrio_buffer, totlen);
2703                 free(pfras, M_TEMP);
2704                 break;
2705         }
2706
2707         case DIOCRGETADDRS: {
2708                 struct pfioc_table *io = (struct pfioc_table *)addr;
2709                 struct pfr_addr *pfras;
2710                 size_t totlen;
2711
2712                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2713                         error = ENODEV;
2714                         break;
2715                 }
2716                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2717                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2718                 PF_RULES_RLOCK();
2719                 error = pfr_get_addrs(&io->pfrio_table, pfras,
2720                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2721                 PF_RULES_RUNLOCK();
2722                 if (error == 0)
2723                         error = copyout(pfras, io->pfrio_buffer, totlen);
2724                 free(pfras, M_TEMP);
2725                 break;
2726         }
2727
2728         case DIOCRGETASTATS: {
2729                 struct pfioc_table *io = (struct pfioc_table *)addr;
2730                 struct pfr_astats *pfrastats;
2731                 size_t totlen;
2732
2733                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2734                         error = ENODEV;
2735                         break;
2736                 }
2737                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
2738                 pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2739                 PF_RULES_RLOCK();
2740                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
2741                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2742                 PF_RULES_RUNLOCK();
2743                 if (error == 0)
2744                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
2745                 free(pfrastats, M_TEMP);
2746                 break;
2747         }
2748
2749         case DIOCRCLRASTATS: {
2750                 struct pfioc_table *io = (struct pfioc_table *)addr;
2751                 struct pfr_addr *pfras;
2752                 size_t totlen;
2753
2754                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2755                         error = ENODEV;
2756                         break;
2757                 }
2758                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2759                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2760                 error = copyin(io->pfrio_buffer, pfras, totlen);
2761                 if (error) {
2762                         free(pfras, M_TEMP);
2763                         break;
2764                 }
2765                 PF_RULES_WLOCK();
2766                 error = pfr_clr_astats(&io->pfrio_table, pfras,
2767                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2768                     PFR_FLAG_USERIOCTL);
2769                 PF_RULES_WUNLOCK();
2770                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2771                         error = copyout(pfras, io->pfrio_buffer, totlen);
2772                 free(pfras, M_TEMP);
2773                 break;
2774         }
2775
2776         case DIOCRTSTADDRS: {
2777                 struct pfioc_table *io = (struct pfioc_table *)addr;
2778                 struct pfr_addr *pfras;
2779                 size_t totlen;
2780
2781                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2782                         error = ENODEV;
2783                         break;
2784                 }
2785                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2786                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2787                 error = copyin(io->pfrio_buffer, pfras, totlen);
2788                 if (error) {
2789                         free(pfras, M_TEMP);
2790                         break;
2791                 }
2792                 PF_RULES_RLOCK();
2793                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
2794                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2795                     PFR_FLAG_USERIOCTL);
2796                 PF_RULES_RUNLOCK();
2797                 if (error == 0)
2798                         error = copyout(pfras, io->pfrio_buffer, totlen);
2799                 free(pfras, M_TEMP);
2800                 break;
2801         }
2802
2803         case DIOCRINADEFINE: {
2804                 struct pfioc_table *io = (struct pfioc_table *)addr;
2805                 struct pfr_addr *pfras;
2806                 size_t totlen;
2807
2808                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2809                         error = ENODEV;
2810                         break;
2811                 }
2812                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2813                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2814                 error = copyin(io->pfrio_buffer, pfras, totlen);
2815                 if (error) {
2816                         free(pfras, M_TEMP);
2817                         break;
2818                 }
2819                 PF_RULES_WLOCK();
2820                 error = pfr_ina_define(&io->pfrio_table, pfras,
2821                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2822                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2823                 PF_RULES_WUNLOCK();
2824                 free(pfras, M_TEMP);
2825                 break;
2826         }
2827
2828         case DIOCOSFPADD: {
2829                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2830                 PF_RULES_WLOCK();
2831                 error = pf_osfp_add(io);
2832                 PF_RULES_WUNLOCK();
2833                 break;
2834         }
2835
2836         case DIOCOSFPGET: {
2837                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2838                 PF_RULES_RLOCK();
2839                 error = pf_osfp_get(io);
2840                 PF_RULES_RUNLOCK();
2841                 break;
2842         }
2843
2844         case DIOCXBEGIN: {
2845                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2846                 struct pfioc_trans_e    *ioes, *ioe;
2847                 size_t                   totlen;
2848                 int                      i;
2849
2850                 if (io->esize != sizeof(*ioe)) {
2851                         error = ENODEV;
2852                         break;
2853                 }
2854                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2855                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2856                 error = copyin(io->array, ioes, totlen);
2857                 if (error) {
2858                         free(ioes, M_TEMP);
2859                         break;
2860                 }
2861                 PF_RULES_WLOCK();
2862                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2863                         switch (ioe->rs_num) {
2864 #ifdef ALTQ
2865                         case PF_RULESET_ALTQ:
2866                                 if (ioe->anchor[0]) {
2867                                         PF_RULES_WUNLOCK();
2868                                         free(ioes, M_TEMP);
2869                                         error = EINVAL;
2870                                         goto fail;
2871                                 }
2872                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2873                                         PF_RULES_WUNLOCK();
2874                                         free(ioes, M_TEMP);
2875                                         goto fail;
2876                                 }
2877                                 break;
2878 #endif /* ALTQ */
2879                         case PF_RULESET_TABLE:
2880                             {
2881                                 struct pfr_table table;
2882
2883                                 bzero(&table, sizeof(table));
2884                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2885                                     sizeof(table.pfrt_anchor));
2886                                 if ((error = pfr_ina_begin(&table,
2887                                     &ioe->ticket, NULL, 0))) {
2888                                         PF_RULES_WUNLOCK();
2889                                         free(ioes, M_TEMP);
2890                                         goto fail;
2891                                 }
2892                                 break;
2893                             }
2894                         default:
2895                                 if ((error = pf_begin_rules(&ioe->ticket,
2896                                     ioe->rs_num, ioe->anchor))) {
2897                                         PF_RULES_WUNLOCK();
2898                                         free(ioes, M_TEMP);
2899                                         goto fail;
2900                                 }
2901                                 break;
2902                         }
2903                 }
2904                 PF_RULES_WUNLOCK();
2905                 error = copyout(ioes, io->array, totlen);
2906                 free(ioes, M_TEMP);
2907                 break;
2908         }
2909
2910         case DIOCXROLLBACK: {
2911                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2912                 struct pfioc_trans_e    *ioe, *ioes;
2913                 size_t                   totlen;
2914                 int                      i;
2915
2916                 if (io->esize != sizeof(*ioe)) {
2917                         error = ENODEV;
2918                         break;
2919                 }
2920                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2921                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2922                 error = copyin(io->array, ioes, totlen);
2923                 if (error) {
2924                         free(ioes, M_TEMP);
2925                         break;
2926                 }
2927                 PF_RULES_WLOCK();
2928                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2929                         switch (ioe->rs_num) {
2930 #ifdef ALTQ
2931                         case PF_RULESET_ALTQ:
2932                                 if (ioe->anchor[0]) {
2933                                         PF_RULES_WUNLOCK();
2934                                         free(ioes, M_TEMP);
2935                                         error = EINVAL;
2936                                         goto fail;
2937                                 }
2938                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2939                                         PF_RULES_WUNLOCK();
2940                                         free(ioes, M_TEMP);
2941                                         goto fail; /* really bad */
2942                                 }
2943                                 break;
2944 #endif /* ALTQ */
2945                         case PF_RULESET_TABLE:
2946                             {
2947                                 struct pfr_table table;
2948
2949                                 bzero(&table, sizeof(table));
2950                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2951                                     sizeof(table.pfrt_anchor));
2952                                 if ((error = pfr_ina_rollback(&table,
2953                                     ioe->ticket, NULL, 0))) {
2954                                         PF_RULES_WUNLOCK();
2955                                         free(ioes, M_TEMP);
2956                                         goto fail; /* really bad */
2957                                 }
2958                                 break;
2959                             }
2960                         default:
2961                                 if ((error = pf_rollback_rules(ioe->ticket,
2962                                     ioe->rs_num, ioe->anchor))) {
2963                                         PF_RULES_WUNLOCK();
2964                                         free(ioes, M_TEMP);
2965                                         goto fail; /* really bad */
2966                                 }
2967                                 break;
2968                         }
2969                 }
2970                 PF_RULES_WUNLOCK();
2971                 free(ioes, M_TEMP);
2972                 break;
2973         }
2974
2975         case DIOCXCOMMIT: {
2976                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2977                 struct pfioc_trans_e    *ioe, *ioes;
2978                 struct pf_ruleset       *rs;
2979                 size_t                   totlen;
2980                 int                      i;
2981
2982                 if (io->esize != sizeof(*ioe)) {
2983                         error = ENODEV;
2984                         break;
2985                 }
2986                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2987                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2988                 error = copyin(io->array, ioes, totlen);
2989                 if (error) {
2990                         free(ioes, M_TEMP);
2991                         break;
2992                 }
2993                 PF_RULES_WLOCK();
2994                 /* First makes sure everything will succeed. */
2995                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2996                         switch (ioe->rs_num) {
2997 #ifdef ALTQ
2998                         case PF_RULESET_ALTQ:
2999                                 if (ioe->anchor[0]) {
3000                                         PF_RULES_WUNLOCK();
3001                                         free(ioes, M_TEMP);
3002                                         error = EINVAL;
3003                                         goto fail;
3004                                 }
3005                                 if (!V_altqs_inactive_open || ioe->ticket !=
3006                                     V_ticket_altqs_inactive) {
3007                                         PF_RULES_WUNLOCK();
3008                                         free(ioes, M_TEMP);
3009                                         error = EBUSY;
3010                                         goto fail;
3011                                 }
3012                                 break;
3013 #endif /* ALTQ */
3014                         case PF_RULESET_TABLE:
3015                                 rs = pf_find_ruleset(ioe->anchor);
3016                                 if (rs == NULL || !rs->topen || ioe->ticket !=
3017                                     rs->tticket) {
3018                                         PF_RULES_WUNLOCK();
3019                                         free(ioes, M_TEMP);
3020                                         error = EBUSY;
3021                                         goto fail;
3022                                 }
3023                                 break;
3024                         default:
3025                                 if (ioe->rs_num < 0 || ioe->rs_num >=
3026                                     PF_RULESET_MAX) {
3027                                         PF_RULES_WUNLOCK();
3028                                         free(ioes, M_TEMP);
3029                                         error = EINVAL;
3030                                         goto fail;
3031                                 }
3032                                 rs = pf_find_ruleset(ioe->anchor);
3033                                 if (rs == NULL ||
3034                                     !rs->rules[ioe->rs_num].inactive.open ||
3035                                     rs->rules[ioe->rs_num].inactive.ticket !=
3036                                     ioe->ticket) {
3037                                         PF_RULES_WUNLOCK();
3038                                         free(ioes, M_TEMP);
3039                                         error = EBUSY;
3040                                         goto fail;
3041                                 }
3042                                 break;
3043                         }
3044                 }
3045                 /* Now do the commit - no errors should happen here. */
3046                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3047                         switch (ioe->rs_num) {
3048 #ifdef ALTQ
3049                         case PF_RULESET_ALTQ:
3050                                 if ((error = pf_commit_altq(ioe->ticket))) {
3051                                         PF_RULES_WUNLOCK();
3052                                         free(ioes, M_TEMP);
3053                                         goto fail; /* really bad */
3054                                 }
3055                                 break;
3056 #endif /* ALTQ */
3057                         case PF_RULESET_TABLE:
3058                             {
3059                                 struct pfr_table table;
3060
3061                                 bzero(&table, sizeof(table));
3062                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3063                                     sizeof(table.pfrt_anchor));
3064                                 if ((error = pfr_ina_commit(&table,
3065                                     ioe->ticket, NULL, NULL, 0))) {
3066                                         PF_RULES_WUNLOCK();
3067                                         free(ioes, M_TEMP);
3068                                         goto fail; /* really bad */
3069                                 }
3070                                 break;
3071                             }
3072                         default:
3073                                 if ((error = pf_commit_rules(ioe->ticket,
3074                                     ioe->rs_num, ioe->anchor))) {
3075                                         PF_RULES_WUNLOCK();
3076                                         free(ioes, M_TEMP);
3077                                         goto fail; /* really bad */
3078                                 }
3079                                 break;
3080                         }
3081                 }
3082                 PF_RULES_WUNLOCK();
3083                 free(ioes, M_TEMP);
3084                 break;
3085         }
3086
3087         case DIOCGETSRCNODES: {
3088                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
3089                 struct pf_srchash       *sh;
3090                 struct pf_src_node      *n, *p, *pstore;
3091                 uint32_t                 i, nr = 0;
3092
3093                 if (psn->psn_len == 0) {
3094                         for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3095                             i++, sh++) {
3096                                 PF_HASHROW_LOCK(sh);
3097                                 LIST_FOREACH(n, &sh->nodes, entry)
3098                                         nr++;
3099                                 PF_HASHROW_UNLOCK(sh);
3100                         }
3101                         psn->psn_len = sizeof(struct pf_src_node) * nr;
3102                         break;
3103                 }
3104
3105                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3106                 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3107                     i++, sh++) {
3108                     PF_HASHROW_LOCK(sh);
3109                     LIST_FOREACH(n, &sh->nodes, entry) {
3110                         int     secs = time_uptime, diff;
3111
3112                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3113                                 break;
3114
3115                         bcopy(n, p, sizeof(struct pf_src_node));
3116                         if (n->rule.ptr != NULL)
3117                                 p->rule.nr = n->rule.ptr->nr;
3118                         p->creation = secs - p->creation;
3119                         if (p->expire > secs)
3120                                 p->expire -= secs;
3121                         else
3122                                 p->expire = 0;
3123
3124                         /* Adjust the connection rate estimate. */
3125                         diff = secs - n->conn_rate.last;
3126                         if (diff >= n->conn_rate.seconds)
3127                                 p->conn_rate.count = 0;
3128                         else
3129                                 p->conn_rate.count -=
3130                                     n->conn_rate.count * diff /
3131                                     n->conn_rate.seconds;
3132                         p++;
3133                         nr++;
3134                     }
3135                     PF_HASHROW_UNLOCK(sh);
3136                 }
3137                 error = copyout(pstore, psn->psn_src_nodes,
3138                     sizeof(struct pf_src_node) * nr);
3139                 if (error) {
3140                         free(pstore, M_TEMP);
3141                         break;
3142                 }
3143                 psn->psn_len = sizeof(struct pf_src_node) * nr;
3144                 free(pstore, M_TEMP);
3145                 break;
3146         }
3147
3148         case DIOCCLRSRCNODES: {
3149
3150                 pf_clear_srcnodes(NULL);
3151                 pf_purge_expired_src_nodes();
3152                 V_pf_status.src_nodes = 0;
3153                 break;
3154         }
3155
3156         case DIOCKILLSRCNODES:
3157                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3158                 break;
3159
3160         case DIOCSETHOSTID: {
3161                 u_int32_t       *hostid = (u_int32_t *)addr;
3162
3163                 PF_RULES_WLOCK();
3164                 if (*hostid == 0)
3165                         V_pf_status.hostid = arc4random();
3166                 else
3167                         V_pf_status.hostid = *hostid;
3168                 PF_RULES_WUNLOCK();
3169                 break;
3170         }
3171
3172         case DIOCOSFPFLUSH:
3173                 PF_RULES_WLOCK();
3174                 pf_osfp_flush();
3175                 PF_RULES_WUNLOCK();
3176                 break;
3177
3178         case DIOCIGETIFACES: {
3179                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3180                 struct pfi_kif *ifstore;
3181                 size_t bufsiz;
3182
3183                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3184                         error = ENODEV;
3185                         break;
3186                 }
3187
3188                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3189                 ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3190                 PF_RULES_RLOCK();
3191                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3192                 PF_RULES_RUNLOCK();
3193                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3194                 free(ifstore, M_TEMP);
3195                 break;
3196         }
3197
3198         case DIOCSETIFFLAG: {
3199                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3200
3201                 PF_RULES_WLOCK();
3202                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3203                 PF_RULES_WUNLOCK();
3204                 break;
3205         }
3206
3207         case DIOCCLRIFFLAG: {
3208                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3209
3210                 PF_RULES_WLOCK();
3211                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3212                 PF_RULES_WUNLOCK();
3213                 break;
3214         }
3215
3216         default:
3217                 error = ENODEV;
3218                 break;
3219         }
3220 fail:
3221         CURVNET_RESTORE();
3222
3223         return (error);
3224 }
3225
3226 void
3227 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3228 {
3229         bzero(sp, sizeof(struct pfsync_state));
3230
3231         /* copy from state key */
3232         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3233         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3234         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3235         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3236         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3237         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3238         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3239         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3240         sp->proto = st->key[PF_SK_WIRE]->proto;
3241         sp->af = st->key[PF_SK_WIRE]->af;
3242
3243         /* copy from state */
3244         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3245         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3246         sp->creation = htonl(time_uptime - st->creation);
3247         sp->expire = pf_state_expires(st);
3248         if (sp->expire <= time_uptime)
3249                 sp->expire = htonl(0);
3250         else
3251                 sp->expire = htonl(sp->expire - time_uptime);
3252
3253         sp->direction = st->direction;
3254         sp->log = st->log;
3255         sp->timeout = st->timeout;
3256         sp->state_flags = st->state_flags;
3257         if (st->src_node)
3258                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3259         if (st->nat_src_node)
3260                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3261
3262         sp->id = st->id;
3263         sp->creatorid = st->creatorid;
3264         pf_state_peer_hton(&st->src, &sp->src);
3265         pf_state_peer_hton(&st->dst, &sp->dst);
3266
3267         if (st->rule.ptr == NULL)
3268                 sp->rule = htonl(-1);
3269         else
3270                 sp->rule = htonl(st->rule.ptr->nr);
3271         if (st->anchor.ptr == NULL)
3272                 sp->anchor = htonl(-1);
3273         else
3274                 sp->anchor = htonl(st->anchor.ptr->nr);
3275         if (st->nat_rule.ptr == NULL)
3276                 sp->nat_rule = htonl(-1);
3277         else
3278                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3279
3280         pf_state_counter_hton(st->packets[0], sp->packets[0]);
3281         pf_state_counter_hton(st->packets[1], sp->packets[1]);
3282         pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3283         pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3284
3285 }
3286
3287 static void
3288 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3289 {
3290         struct pfr_ktable *kt;
3291
3292         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3293
3294         kt = aw->p.tbl;
3295         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3296                 kt = kt->pfrkt_root;
3297         aw->p.tbl = NULL;
3298         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3299                 kt->pfrkt_cnt : -1;
3300 }
3301
3302 /*
3303  * XXX - Check for version missmatch!!!
3304  */
3305 static void
3306 pf_clear_states(void)
3307 {
3308         struct pf_state *s;
3309         u_int i;
3310
3311         for (i = 0; i <= V_pf_hashmask; i++) {
3312                 struct pf_idhash *ih = &V_pf_idhash[i];
3313 relock:
3314                 PF_HASHROW_LOCK(ih);
3315                 LIST_FOREACH(s, &ih->states, entry) {
3316                         s->timeout = PFTM_PURGE;
3317                         /* Don't send out individual delete messages. */
3318                         s->sync_state = PFSTATE_NOSYNC;
3319                         pf_unlink_state(s, PF_ENTER_LOCKED);
3320                         goto relock;
3321                 }
3322                 PF_HASHROW_UNLOCK(ih);
3323         }
3324 }
3325
3326 static int
3327 pf_clear_tables(void)
3328 {
3329         struct pfioc_table io;
3330         int error;
3331
3332         bzero(&io, sizeof(io));
3333
3334         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3335             io.pfrio_flags);
3336
3337         return (error);
3338 }
3339
3340 static void
3341 pf_clear_srcnodes(struct pf_src_node *n)
3342 {
3343         struct pf_state *s;
3344         int i;
3345
3346         for (i = 0; i <= V_pf_hashmask; i++) {
3347                 struct pf_idhash *ih = &V_pf_idhash[i];
3348
3349                 PF_HASHROW_LOCK(ih);
3350                 LIST_FOREACH(s, &ih->states, entry) {
3351                         if (n == NULL || n == s->src_node)
3352                                 s->src_node = NULL;
3353                         if (n == NULL || n == s->nat_src_node)
3354                                 s->nat_src_node = NULL;
3355                 }
3356                 PF_HASHROW_UNLOCK(ih);
3357         }
3358
3359         if (n == NULL) {
3360                 struct pf_srchash *sh;
3361
3362                 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3363                     i++, sh++) {
3364                         PF_HASHROW_LOCK(sh);
3365                         LIST_FOREACH(n, &sh->nodes, entry) {
3366                                 n->expire = 1;
3367                                 n->states = 0;
3368                         }
3369                         PF_HASHROW_UNLOCK(sh);
3370                 }
3371         } else {
3372                 /* XXX: hash slot should already be locked here. */
3373                 n->expire = 1;
3374                 n->states = 0;
3375         }
3376 }
3377
3378 static void
3379 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3380 {
3381         struct pf_src_node_list  kill;
3382
3383         LIST_INIT(&kill);
3384         for (int i = 0; i <= V_pf_srchashmask; i++) {
3385                 struct pf_srchash *sh = &V_pf_srchash[i];
3386                 struct pf_src_node *sn, *tmp;
3387
3388                 PF_HASHROW_LOCK(sh);
3389                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3390                         if (PF_MATCHA(psnk->psnk_src.neg,
3391                               &psnk->psnk_src.addr.v.a.addr,
3392                               &psnk->psnk_src.addr.v.a.mask,
3393                               &sn->addr, sn->af) &&
3394                             PF_MATCHA(psnk->psnk_dst.neg,
3395                               &psnk->psnk_dst.addr.v.a.addr,
3396                               &psnk->psnk_dst.addr.v.a.mask,
3397                               &sn->raddr, sn->af)) {
3398                                 pf_unlink_src_node_locked(sn);
3399                                 LIST_INSERT_HEAD(&kill, sn, entry);
3400                                 sn->expire = 1;
3401                         }
3402                 PF_HASHROW_UNLOCK(sh);
3403         }
3404
3405         for (int i = 0; i <= V_pf_hashmask; i++) {
3406                 struct pf_idhash *ih = &V_pf_idhash[i];
3407                 struct pf_state *s;
3408
3409                 PF_HASHROW_LOCK(ih);
3410                 LIST_FOREACH(s, &ih->states, entry) {
3411                         if (s->src_node && s->src_node->expire == 1) {
3412 #ifdef INVARIANTS
3413                                 s->src_node->states--;
3414 #endif
3415                                 s->src_node = NULL;
3416                         }
3417                         if (s->nat_src_node && s->nat_src_node->expire == 1) {
3418 #ifdef INVARIANTS
3419                                 s->nat_src_node->states--;
3420 #endif
3421                                 s->nat_src_node = NULL;
3422                         }
3423                 }
3424                 PF_HASHROW_UNLOCK(ih);
3425         }
3426
3427         psnk->psnk_killed = pf_free_src_nodes(&kill);
3428 }
3429
3430 /*
3431  * XXX - Check for version missmatch!!!
3432  */
3433
3434 /*
3435  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3436  */
3437 static int
3438 shutdown_pf(void)
3439 {
3440         int error = 0;
3441         u_int32_t t[5];
3442         char nn = '\0';
3443
3444         V_pf_status.running = 0;
3445
3446         counter_u64_free(V_pf_default_rule.states_cur);
3447         counter_u64_free(V_pf_default_rule.states_tot);
3448         counter_u64_free(V_pf_default_rule.src_nodes);
3449
3450         do {
3451                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3452                     != 0) {
3453                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3454                         break;
3455                 }
3456                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3457                     != 0) {
3458                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3459                         break;          /* XXX: rollback? */
3460                 }
3461                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3462                     != 0) {
3463                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3464                         break;          /* XXX: rollback? */
3465                 }
3466                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3467                     != 0) {
3468                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3469                         break;          /* XXX: rollback? */
3470                 }
3471                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3472                     != 0) {
3473                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3474                         break;          /* XXX: rollback? */
3475                 }
3476
3477                 /* XXX: these should always succeed here */
3478                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3479                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3480                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3481                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3482                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3483
3484                 if ((error = pf_clear_tables()) != 0)
3485                         break;
3486
3487 #ifdef ALTQ
3488                 if ((error = pf_begin_altq(&t[0])) != 0) {
3489                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3490                         break;
3491                 }
3492                 pf_commit_altq(t[0]);
3493 #endif
3494
3495                 pf_clear_states();
3496
3497                 pf_clear_srcnodes(NULL);
3498
3499                 /* status does not use malloced mem so no need to cleanup */
3500                 /* fingerprints and interfaces have thier own cleanup code */
3501         } while(0);
3502
3503         return (error);
3504 }
3505
3506 #ifdef INET
3507 static int
3508 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3509     struct inpcb *inp)
3510 {
3511         int chk;
3512
3513         chk = pf_test(PF_IN, ifp, m, inp);
3514         if (chk && *m) {
3515                 m_freem(*m);
3516                 *m = NULL;
3517         }
3518
3519         return (chk);
3520 }
3521
3522 static int
3523 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3524     struct inpcb *inp)
3525 {
3526         int chk;
3527
3528         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3529         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3530                 in_delayed_cksum(*m);
3531                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3532         }
3533
3534         chk = pf_test(PF_OUT, ifp, m, inp);
3535         if (chk && *m) {
3536                 m_freem(*m);
3537                 *m = NULL;
3538         }
3539
3540         return (chk);
3541 }
3542 #endif
3543
3544 #ifdef INET6
3545 static int
3546 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3547     struct inpcb *inp)
3548 {
3549         int chk;
3550
3551         /*
3552          * In case of loopback traffic IPv6 uses the real interface in
3553          * order to support scoped addresses. In order to support stateful
3554          * filtering we have change this to lo0 as it is the case in IPv4.
3555          */
3556         CURVNET_SET(ifp->if_vnet);
3557         chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3558         CURVNET_RESTORE();
3559         if (chk && *m) {
3560                 m_freem(*m);
3561                 *m = NULL;
3562         }
3563         return chk;
3564 }
3565
3566 static int
3567 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3568     struct inpcb *inp)
3569 {
3570         int chk;
3571
3572         /* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3573         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3574 #ifdef INET
3575                 /* XXX-BZ copy&paste error from r126261? */
3576                 in_delayed_cksum(*m);
3577 #endif
3578                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3579         }
3580         CURVNET_SET(ifp->if_vnet);
3581         chk = pf_test6(PF_OUT, ifp, m, inp);
3582         CURVNET_RESTORE();
3583         if (chk && *m) {
3584                 m_freem(*m);
3585                 *m = NULL;
3586         }
3587         return chk;
3588 }
3589 #endif /* INET6 */
3590
3591 static int
3592 hook_pf(void)
3593 {
3594 #ifdef INET
3595         struct pfil_head *pfh_inet;
3596 #endif
3597 #ifdef INET6
3598         struct pfil_head *pfh_inet6;
3599 #endif
3600
3601         if (V_pf_pfil_hooked)
3602                 return (0);
3603
3604 #ifdef INET
3605         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3606         if (pfh_inet == NULL)
3607                 return (ESRCH); /* XXX */
3608         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3609         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3610 #endif
3611 #ifdef INET6
3612         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3613         if (pfh_inet6 == NULL) {
3614 #ifdef INET
3615                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3616                     pfh_inet);
3617                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3618                     pfh_inet);
3619 #endif
3620                 return (ESRCH); /* XXX */
3621         }
3622         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3623         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3624 #endif
3625
3626         V_pf_pfil_hooked = 1;
3627         return (0);
3628 }
3629
3630 static int
3631 dehook_pf(void)
3632 {
3633 #ifdef INET
3634         struct pfil_head *pfh_inet;
3635 #endif
3636 #ifdef INET6
3637         struct pfil_head *pfh_inet6;
3638 #endif
3639
3640         if (V_pf_pfil_hooked == 0)
3641                 return (0);
3642
3643 #ifdef INET
3644         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3645         if (pfh_inet == NULL)
3646                 return (ESRCH); /* XXX */
3647         pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3648             pfh_inet);
3649         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3650             pfh_inet);
3651 #endif
3652 #ifdef INET6
3653         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3654         if (pfh_inet6 == NULL)
3655                 return (ESRCH); /* XXX */
3656         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3657             pfh_inet6);
3658         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3659             pfh_inet6);
3660 #endif
3661
3662         V_pf_pfil_hooked = 0;
3663         return (0);
3664 }
3665
3666 static int
3667 pf_load(void)
3668 {
3669         int error;
3670
3671         VNET_ITERATOR_DECL(vnet_iter);
3672
3673         VNET_LIST_RLOCK();
3674         VNET_FOREACH(vnet_iter) {
3675                 CURVNET_SET(vnet_iter);
3676                 V_pf_pfil_hooked = 0;
3677                 V_pf_end_threads = 0;
3678                 TAILQ_INIT(&V_pf_tags);
3679                 TAILQ_INIT(&V_pf_qids);
3680                 CURVNET_RESTORE();
3681         }
3682         VNET_LIST_RUNLOCK();
3683
3684         rw_init(&pf_rules_lock, "pf rulesets");
3685
3686         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3687         if ((error = pfattach()) != 0)
3688                 return (error);
3689
3690         return (0);
3691 }
3692
3693 static int
3694 pf_unload(void)
3695 {
3696         int error = 0;
3697
3698         PF_RULES_WLOCK();
3699         V_pf_status.running = 0;
3700         PF_RULES_WUNLOCK();
3701         swi_remove(V_pf_swi_cookie);
3702         error = dehook_pf();
3703         if (error) {
3704                 /*
3705                  * Should not happen!
3706                  * XXX Due to error code ESRCH, kldunload will show
3707                  * a message like 'No such process'.
3708                  */
3709                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3710                 return error;
3711         }
3712         PF_RULES_WLOCK();
3713         shutdown_pf();
3714         V_pf_end_threads = 1;
3715         while (V_pf_end_threads < 2) {
3716                 wakeup_one(pf_purge_thread);
3717                 rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3718         }
3719         pf_normalize_cleanup();
3720         pfi_cleanup();
3721         pfr_cleanup();
3722         pf_osfp_flush();
3723         pf_cleanup();
3724         PF_RULES_WUNLOCK();
3725         destroy_dev(pf_dev);
3726         rw_destroy(&pf_rules_lock);
3727
3728         return (error);
3729 }
3730
3731 static int
3732 pf_modevent(module_t mod, int type, void *data)
3733 {
3734         int error = 0;
3735
3736         switch(type) {
3737         case MOD_LOAD:
3738                 error = pf_load();
3739                 break;
3740         case MOD_QUIESCE:
3741                 /*
3742                  * Module should not be unloaded due to race conditions.
3743                  */
3744                 error = EBUSY;
3745                 break;
3746         case MOD_UNLOAD:
3747                 error = pf_unload();
3748                 break;
3749         default:
3750                 error = EINVAL;
3751                 break;
3752         }
3753
3754         return (error);
3755 }
3756
3757 static moduledata_t pf_mod = {
3758         "pf",
3759         pf_modevent,
3760         0
3761 };
3762
3763 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3764 MODULE_VERSION(pf, PF_MODVER);