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