sys/net/pf/pf_ioctl.c

   1 /*      $OpenBSD: pf_ioctl.c,v 1.209 2008/06/29 08:42:15 mcbride Exp $ */
   2 /*add $OpenBSD: pf_ioctl.c,v 1.212 2009/02/15 20:42:33 mbalmer Exp $ */
   3
   4 /*
   5  * Copyright (c) 2010 The DragonFly Project.  All rights reserved.
   6  *
   7  * Copyright (c) 2001 Daniel Hartmeier
   8  * Copyright (c) 2002,2003 Henning Brauer
   9  * All rights reserved.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  *
  15  *    - Redistributions of source code must retain the above copyright
  16  *      notice, this list of conditions and the following disclaimer.
  17  *    - Redistributions in binary form must reproduce the above
  18  *      copyright notice, this list of conditions and the following
  19  *      disclaimer in the documentation and/or other materials provided
  20  *      with the distribution.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  23  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  24  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  25  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  26  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  27  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  28  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  29  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  30  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  32  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  33  * POSSIBILITY OF SUCH DAMAGE.
  34  *
  35  * Effort sponsored in part by the Defense Advanced Research Projects
  36  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  37  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  38  *
  39  */
  40
  41 #include "opt_inet.h"
  42 #include "opt_inet6.h"
  43
  44 #include <sys/param.h>
  45 #include <sys/systm.h>
  46 #include <sys/conf.h>
  47 #include <sys/device.h>
  48 #include <sys/mbuf.h>
  49 #include <sys/filio.h>
  50 #include <sys/fcntl.h>
  51 #include <sys/socket.h>
  52 #include <sys/socketvar.h>
  53 #include <sys/kernel.h>
  54 #include <sys/kthread.h>
  55 #include <sys/time.h>
  56 #include <sys/proc.h>
  57 #include <sys/malloc.h>
  58 #include <sys/module.h>
  59 #include <sys/lock.h>
  60
  61 #include <sys/thread2.h>
  62
  63 #include <net/if.h>
  64 #include <net/if_types.h>
  65 #include <net/route.h>
  66
  67 #include <netinet/in.h>
  68 #include <netinet/in_var.h>
  69 #include <netinet/in_systm.h>
  70 #include <netinet/ip.h>
  71 #include <netinet/ip_var.h>
  72 #include <netinet/ip_icmp.h>
  73
  74 #include <net/pf/pfvar.h>
  75 #include <sys/md5.h>
  76 #include <net/pf/pfvar.h>
  77
  78 #include <net/pf/if_pfsync.h>
  79
  80 #if NPFLOG > 0
  81 #include <net/if_pflog.h>
  82 #endif /* NPFLOG > 0 */
  83
  84 #ifdef INET6
  85 #include <netinet/ip6.h>
  86 #include <netinet/in_pcb.h>
  87 #endif /* INET6 */
  88
  89 #ifdef ALTQ
  90 #include <net/altq/altq.h>
  91 #endif
  92
  93 #include <machine/limits.h>
  94 #include <net/pfil.h>
  95 #include <sys/mutex.h>
  96
  97 u_int rt_numfibs = RT_NUMFIBS;
  98
  99 void                     pfattach(void);
 100 struct pf_pool          *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
 101                             u_int8_t, u_int8_t, u_int8_t);
 102
 103 void                     pf_mv_pool(struct pf_palist *, struct pf_palist *);
 104 void                     pf_empty_pool(struct pf_palist *);
 105 #ifdef ALTQ
 106 int                      pf_begin_altq(u_int32_t *);
 107 int                      pf_rollback_altq(u_int32_t);
 108 int                      pf_commit_altq(u_int32_t);
 109 int                      pf_enable_altq(struct pf_altq *);
 110 int                      pf_disable_altq(struct pf_altq *);
 111 #endif /* ALTQ */
 112 int                      pf_begin_rules(u_int32_t *, int, const char *);
 113 int                      pf_rollback_rules(u_int32_t, int, char *);
 114 int                      pf_setup_pfsync_matching(struct pf_ruleset *);
 115 void                     pf_hash_rule(MD5_CTX *, struct pf_rule *);
 116 void                     pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 117 int                      pf_commit_rules(u_int32_t, int, char *);
 118 int                      pf_addr_setup(struct pf_ruleset *,
 119                             struct pf_addr_wrap *, sa_family_t);
 120 void                     pf_addr_copyout(struct pf_addr_wrap *);
 121
 122 struct pf_rule           pf_default_rule;
 123 struct lock              pf_consistency_lock;
 124 #ifdef ALTQ
 125 static int               pf_altq_running;
 126 #endif
 127
 128 #define TAGID_MAX        50000
 129 TAILQ_HEAD(pf_tags, pf_tagname) pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
 130                                 pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
 131
 132 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 133 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 134 #endif
 135 u_int16_t                tagname2tag(struct pf_tags *, char *);
 136 void                     tag2tagname(struct pf_tags *, u_int16_t, char *);
 137 void                     tag_unref(struct pf_tags *, u_int16_t);
 138 int                      pf_rtlabel_add(struct pf_addr_wrap *);
 139 void                     pf_rtlabel_remove(struct pf_addr_wrap *);
 140 void                     pf_rtlabel_copyout(struct pf_addr_wrap *);
 141
 142 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 143
 144 static cdev_t   pf_dev;
 145
 146 static MALLOC_DEFINE(M_PFRULEPL, "pfrulepl", "pf rule pool list");
 147 static MALLOC_DEFINE(M_PFALTQPL, "pfaltqpl", "pf altq pool list");
 148 static MALLOC_DEFINE(M_PFPOOLADDRPL, "pfpooladdrpl", "pf pool address pool list");
 149 static MALLOC_DEFINE(M_PFFRENTPL, "pffrent", "pf frent pool list");
 150
 151
 152 /*
 153  * XXX - These are new and need to be checked when moveing to a new version
 154  */
 155 static void              pf_clear_states(void);
 156 static int               pf_clear_tables(void);
 157 static void              pf_clear_srcnodes(void);
 158 /*
 159  * XXX - These are new and need to be checked when moveing to a new version
 160  */
 161
 162 /*
 163  * Wrapper functions for pfil(9) hooks
 164  */
 165 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 166                 int dir);
 167 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 168                 int dir);
 169 #ifdef INET6
 170 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 171                 int dir);
 172 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 173                 int dir);
 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 d_open_t        pfopen;
 183 d_close_t       pfclose;
 184 d_ioctl_t       pfioctl;
 185
 186 static struct dev_ops pf_ops = {            /* XXX convert to port model */
 187         { PF_NAME, 73, 0 },
 188         .d_open =       pfopen,
 189         .d_close =      pfclose,
 190         .d_ioctl =      pfioctl
 191 };
 192
 193 static volatile int pf_pfil_hooked = 0;
 194 int pf_end_threads = 0;
 195
 196 int debug_pfugidhack = 0;
 197 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
 198         "Enable/disable pf user/group rules mpsafe hack");
 199
 200 void
 201 pfattach(void)
 202 {
 203         u_int32_t *my_timeout = pf_default_rule.timeout;
 204
 205
 206         if (!rn_inithead((void **)&pf_maskhead, NULL, 0)) {
 207                 kprintf("pf mask radix tree create failed\n");
 208                 return;
 209         }
 210         kmalloc_create(&pf_state_pl, "pf state pool list");
 211         kmalloc_raise_limit(pf_state_pl, 0);
 212         kmalloc_create(&pf_frent_pl, "pf fragment pool list");
 213         kmalloc_raise_limit(pf_frent_pl, 0);
 214         kmalloc_create(&pf_cent_pl, "pf cent pool list");
 215         kmalloc_raise_limit(pf_cent_pl, 0);
 216
 217         pfr_initialize();
 218         pfi_initialize();
 219         pf_osfp_initialize();
 220
 221         pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
 222         pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 223         pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
 224         pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
 225         if (ctob(physmem) <= 100*1024*1024)
 226                 pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
 227                     PFR_KENTRY_HIWAT_SMALL;
 228
 229         RB_INIT(&tree_src_tracking);
 230         RB_INIT(&pf_anchors);
 231         pf_init_ruleset(&pf_main_ruleset);
 232         TAILQ_INIT(&pf_altqs[0]);
 233         TAILQ_INIT(&pf_altqs[1]);
 234         TAILQ_INIT(&pf_pabuf);
 235         pf_altqs_active = &pf_altqs[0];
 236         pf_altqs_inactive = &pf_altqs[1];
 237         TAILQ_INIT(&state_list);
 238
 239         /* default rule should never be garbage collected */
 240         pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
 241         pf_default_rule.action = PF_PASS;
 242         pf_default_rule.nr = (uint32_t)(-1);
 243         pf_default_rule.rtableid = -1;
 244
 245         /* initialize default timeouts */
 246         my_timeout[PFTM_TCP_FIRST_PACKET] = 120;        /* First TCP packet */
 247         my_timeout[PFTM_TCP_OPENING] = 30;              /* No response yet */
 248         my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;    /* Established */
 249         my_timeout[PFTM_TCP_CLOSING] = 15 * 60;         /* Half closed */
 250         my_timeout[PFTM_TCP_FIN_WAIT] = 45;             /* Got both FINs */
 251         my_timeout[PFTM_TCP_CLOSED] = 90;               /* Got a RST */
 252         my_timeout[PFTM_UDP_FIRST_PACKET] = 60;         /* First UDP packet */
 253         my_timeout[PFTM_UDP_SINGLE] = 30;               /* Unidirectional */
 254         my_timeout[PFTM_UDP_MULTIPLE] = 60;             /* Bidirectional */
 255         my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;        /* First ICMP packet */
 256         my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;         /* Got error response */
 257         my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;       /* First packet */
 258         my_timeout[PFTM_OTHER_SINGLE] = 30;             /* Unidirectional */
 259         my_timeout[PFTM_OTHER_MULTIPLE] = 60;           /* Bidirectional */
 260         my_timeout[PFTM_FRAG] = 30;                     /* Fragment expire */
 261         my_timeout[PFTM_INTERVAL] = 10;                 /* Expire interval */
 262         my_timeout[PFTM_SRC_NODE] = 0;          /* Source Tracking */
 263         my_timeout[PFTM_TS_DIFF] = 30;          /* Allowed TS diff */
 264         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 265         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 266
 267         pf_normalize_init();
 268         bzero(&pf_status, sizeof(pf_status));
 269         pf_status.debug = PF_DEBUG_URGENT;
 270         /* XXX do our best to avoid a conflict */
 271         pf_status.hostid = karc4random();
 272
 273         if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
 274                 panic("pfpurge thread");
 275 }
 276
 277 int
 278 pfopen(struct dev_open_args *ap)
 279 {
 280         lwkt_gettoken(&pf_token);
 281         cdev_t dev = ap->a_head.a_dev;
 282         if (minor(dev) >= 1) {
 283                 lwkt_reltoken(&pf_token);
 284                 return (ENXIO);
 285         }
 286         lwkt_reltoken(&pf_token);
 287         return (0);
 288 }
 289
 290 int
 291 pfclose(struct dev_close_args *ap)
 292 {
 293         lwkt_gettoken(&pf_token);
 294         cdev_t dev = ap->a_head.a_dev;
 295         if (minor(dev) >= 1) {
 296                 lwkt_reltoken(&pf_token);
 297                 return (ENXIO);
 298         }
 299         lwkt_reltoken(&pf_token);
 300         return (0);
 301 }
 302
 303 struct pf_pool *
 304 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 305     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 306     u_int8_t check_ticket)
 307 {
 308         struct pf_ruleset       *ruleset;
 309         struct pf_rule          *rule;
 310         int                      rs_num;
 311
 312         ruleset = pf_find_ruleset(anchor);
 313         if (ruleset == NULL)
 314                 return (NULL);
 315         rs_num = pf_get_ruleset_number(rule_action);
 316         if (rs_num >= PF_RULESET_MAX)
 317                 return (NULL);
 318         if (active) {
 319                 if (check_ticket && ticket !=
 320                     ruleset->rules[rs_num].active.ticket)
 321                         return (NULL);
 322                 if (r_last)
 323                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 324                             pf_rulequeue);
 325                 else
 326                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 327         } else {
 328                 if (check_ticket && ticket !=
 329                     ruleset->rules[rs_num].inactive.ticket)
 330                         return (NULL);
 331                 if (r_last)
 332                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 333                             pf_rulequeue);
 334                 else
 335                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 336         }
 337         if (!r_last) {
 338                 while ((rule != NULL) && (rule->nr != rule_number))
 339                         rule = TAILQ_NEXT(rule, entries);
 340         }
 341         if (rule == NULL)
 342                 return (NULL);
 343
 344         return (&rule->rpool);
 345 }
 346
 347 void
 348 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 349 {
 350         struct pf_pooladdr      *mv_pool_pa;
 351
 352         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 353                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 354                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 355         }
 356 }
 357
 358 void
 359 pf_empty_pool(struct pf_palist *poola)
 360 {
 361         struct pf_pooladdr      *empty_pool_pa;
 362
 363         while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 364                 pfi_dynaddr_remove(&empty_pool_pa->addr);
 365                 pf_tbladdr_remove(&empty_pool_pa->addr);
 366                 pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
 367                 TAILQ_REMOVE(poola, empty_pool_pa, entries);
 368                 kfree(empty_pool_pa, M_PFPOOLADDRPL);
 369         }
 370 }
 371
 372 void
 373 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 374 {
 375         if (rulequeue != NULL) {
 376                 if (rule->states_cur <= 0) {
 377                         /*
 378                          * XXX - we need to remove the table *before* detaching
 379                          * the rule to make sure the table code does not delete
 380                          * the anchor under our feet.
 381                          */
 382                         pf_tbladdr_remove(&rule->src.addr);
 383                         pf_tbladdr_remove(&rule->dst.addr);
 384                         if (rule->overload_tbl)
 385                                 pfr_detach_table(rule->overload_tbl);
 386                 }
 387                 TAILQ_REMOVE(rulequeue, rule, entries);
 388                 rule->entries.tqe_prev = NULL;
 389                 rule->nr = -1;
 390         }
 391
 392         if (rule->states_cur > 0 || rule->src_nodes > 0 ||
 393             rule->entries.tqe_prev != NULL)
 394                 return;
 395         pf_tag_unref(rule->tag);
 396         pf_tag_unref(rule->match_tag);
 397 #ifdef ALTQ
 398         if (rule->pqid != rule->qid)
 399                 pf_qid_unref(rule->pqid);
 400         pf_qid_unref(rule->qid);
 401 #endif
 402         pf_rtlabel_remove(&rule->src.addr);
 403         pf_rtlabel_remove(&rule->dst.addr);
 404         pfi_dynaddr_remove(&rule->src.addr);
 405         pfi_dynaddr_remove(&rule->dst.addr);
 406         if (rulequeue == NULL) {
 407                 pf_tbladdr_remove(&rule->src.addr);
 408                 pf_tbladdr_remove(&rule->dst.addr);
 409                 if (rule->overload_tbl)
 410                         pfr_detach_table(rule->overload_tbl);
 411         }
 412         pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
 413         pf_anchor_remove(rule);
 414         pf_empty_pool(&rule->rpool.list);
 415         kfree(rule, M_PFRULEPL);
 416 }
 417
 418 u_int16_t
 419 tagname2tag(struct pf_tags *head, char *tagname)
 420 {
 421         struct pf_tagname       *tag, *p = NULL;
 422         u_int16_t                new_tagid = 1;
 423
 424         TAILQ_FOREACH(tag, head, entries)
 425                 if (strcmp(tagname, tag->name) == 0) {
 426                         tag->ref++;
 427                         return (tag->tag);
 428                 }
 429
 430         /*
 431          * to avoid fragmentation, we do a linear search from the beginning
 432          * and take the first free slot we find. if there is none or the list
 433          * is empty, append a new entry at the end.
 434          */
 435
 436         /* new entry */
 437         if (!TAILQ_EMPTY(head))
 438                 for (p = TAILQ_FIRST(head); p != NULL &&
 439                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 440                         new_tagid = p->tag + 1;
 441
 442         if (new_tagid > TAGID_MAX)
 443                 return (0);
 444
 445         /* allocate and fill new struct pf_tagname */
 446         tag = kmalloc(sizeof(*tag), M_TEMP, M_WAITOK);
 447         if (tag == NULL)
 448                 return (0);
 449         strlcpy(tag->name, tagname, sizeof(tag->name));
 450         tag->tag = new_tagid;
 451         tag->ref++;
 452
 453         if (p != NULL)  /* insert new entry before p */
 454                 TAILQ_INSERT_BEFORE(p, tag, entries);
 455         else    /* either list empty or no free slot in between */
 456                 TAILQ_INSERT_TAIL(head, tag, entries);
 457
 458         return (tag->tag);
 459 }
 460
 461 void
 462 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
 463 {
 464         struct pf_tagname       *tag;
 465
 466         TAILQ_FOREACH(tag, head, entries)
 467                 if (tag->tag == tagid) {
 468                         strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
 469                         return;
 470                 }
 471 }
 472
 473 void
 474 tag_unref(struct pf_tags *head, u_int16_t tag)
 475 {
 476         struct pf_tagname       *p, *next;
 477
 478         if (tag == 0)
 479                 return;
 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                                 kfree(p, M_TEMP);
 487                         }
 488                         break;
 489                 }
 490         }
 491 }
 492
 493 u_int16_t
 494 pf_tagname2tag(char *tagname)
 495 {
 496         return (tagname2tag(&pf_tags, tagname));
 497 }
 498
 499 void
 500 pf_tag2tagname(u_int16_t tagid, char *p)
 501 {
 502         tag2tagname(&pf_tags, tagid, p);
 503 }
 504
 505 void
 506 pf_tag_ref(u_int16_t tag)
 507 {
 508         struct pf_tagname *t;
 509
 510         TAILQ_FOREACH(t, &pf_tags, entries)
 511                 if (t->tag == tag)
 512                         break;
 513         if (t != NULL)
 514                 t->ref++;
 515 }
 516
 517 void
 518 pf_tag_unref(u_int16_t tag)
 519 {
 520         tag_unref(&pf_tags, tag);
 521 }
 522
 523 int
 524 pf_rtlabel_add(struct pf_addr_wrap *a)
 525 {
 526         return (0);
 527 }
 528
 529 void
 530 pf_rtlabel_remove(struct pf_addr_wrap *a)
 531 {
 532 }
 533
 534 void
 535 pf_rtlabel_copyout(struct pf_addr_wrap *a)
 536 {
 537         if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
 538                 strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
 539 }
 540
 541 #ifdef ALTQ
 542 u_int32_t
 543 pf_qname2qid(char *qname)
 544 {
 545         return ((u_int32_t)tagname2tag(&pf_qids, qname));
 546 }
 547
 548 void
 549 pf_qid2qname(u_int32_t qid, char *p)
 550 {
 551         tag2tagname(&pf_qids, (u_int16_t)qid, p);
 552 }
 553
 554 void
 555 pf_qid_unref(u_int32_t qid)
 556 {
 557         tag_unref(&pf_qids, (u_int16_t)qid);
 558 }
 559
 560 int
 561 pf_begin_altq(u_int32_t *ticket)
 562 {
 563         struct pf_altq  *altq;
 564         int              error = 0;
 565
 566         /* Purge the old altq list */
 567         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 568                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 569                 if (altq->qname[0] == 0) {
 570                         /* detach and destroy the discipline */
 571                         error = altq_remove(altq);
 572                 } else
 573                         pf_qid_unref(altq->qid);
 574                 kfree(altq, M_PFALTQPL);
 575         }
 576         if (error)
 577                 return (error);
 578         *ticket = ++ticket_altqs_inactive;
 579         altqs_inactive_open = 1;
 580         return (0);
 581 }
 582
 583 int
 584 pf_rollback_altq(u_int32_t ticket)
 585 {
 586         struct pf_altq  *altq;
 587         int              error = 0;
 588
 589         if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
 590                 return (0);
 591         /* Purge the old altq list */
 592         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 593                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 594                 if (altq->qname[0] == 0) {
 595                         /* detach and destroy the discipline */
 596                         error = altq_remove(altq);
 597                 } else
 598                         pf_qid_unref(altq->qid);
 599                 kfree(altq, M_PFALTQPL);
 600         }
 601         altqs_inactive_open = 0;
 602         return (error);
 603 }
 604
 605 int
 606 pf_commit_altq(u_int32_t ticket)
 607 {
 608         struct pf_altqqueue     *old_altqs;
 609         struct pf_altq          *altq;
 610         int                      err, error = 0;
 611
 612         if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
 613                 return (EBUSY);
 614
 615         /* swap altqs, keep the old. */
 616         crit_enter();
 617         old_altqs = pf_altqs_active;
 618         pf_altqs_active = pf_altqs_inactive;
 619         pf_altqs_inactive = old_altqs;
 620         ticket_altqs_active = ticket_altqs_inactive;
 621
 622         /* Attach new disciplines */
 623         TAILQ_FOREACH(altq, pf_altqs_active, entries) {
 624                 if (altq->qname[0] == 0) {
 625                         /* attach the discipline */
 626                         error = altq_pfattach(altq);
 627                         if (error) {
 628                                 crit_exit();
 629                                 return (error);
 630                         }
 631                 }
 632         }
 633
 634         /* Purge the old altq list */
 635         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 636                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 637                 if (altq->qname[0] == 0) {
 638                         /* detach and destroy the discipline */
 639                         if (pf_altq_running)
 640                                 error = pf_disable_altq(altq);
 641                         err = altq_pfdetach(altq);
 642                         if (err != 0 && error == 0)
 643                                 error = err;
 644                         err = altq_remove(altq);
 645                         if (err != 0 && error == 0)
 646                                 error = err;
 647                 } else
 648                         pf_qid_unref(altq->qid);
 649                 kfree(altq, M_PFALTQPL);
 650         }
 651         crit_exit();
 652
 653         altqs_inactive_open = 0;
 654         return (error);
 655 }
 656
 657 int
 658 pf_enable_altq(struct pf_altq *altq)
 659 {
 660         struct ifnet            *ifp;
 661         struct tb_profile        tb;
 662         int                      error = 0;
 663
 664         if ((ifp = ifunit(altq->ifname)) == NULL)
 665                 return (EINVAL);
 666
 667         if (ifp->if_snd.altq_type != ALTQT_NONE)
 668                 error = altq_enable(&ifp->if_snd);
 669
 670         /* set tokenbucket regulator */
 671         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 672                 tb.rate = altq->ifbandwidth;
 673                 tb.depth = altq->tbrsize;
 674                 crit_enter();
 675                 error = tbr_set(&ifp->if_snd, &tb);
 676                 crit_exit();
 677         }
 678
 679         return (error);
 680 }
 681
 682 int
 683 pf_disable_altq(struct pf_altq *altq)
 684 {
 685         struct ifnet            *ifp;
 686         struct tb_profile        tb;
 687         int                      error;
 688
 689         if ((ifp = ifunit(altq->ifname)) == NULL)
 690                 return (EINVAL);
 691
 692         /*
 693          * when the discipline is no longer referenced, it was overridden
 694          * by a new one.  if so, just return.
 695          */
 696         if (altq->altq_disc != ifp->if_snd.altq_disc)
 697                 return (0);
 698
 699         error = altq_disable(&ifp->if_snd);
 700
 701         if (error == 0) {
 702                 /* clear tokenbucket regulator */
 703                 tb.rate = 0;
 704                 crit_enter();
 705                 error = tbr_set(&ifp->if_snd, &tb);
 706                 crit_exit();
 707         }
 708
 709         return (error);
 710 }
 711 #endif /* ALTQ */
 712
 713 int
 714 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 715 {
 716         struct pf_ruleset       *rs;
 717         struct pf_rule          *rule;
 718
 719         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 720                 return (EINVAL);
 721         rs = pf_find_or_create_ruleset(anchor);
 722         if (rs == NULL)
 723                 return (EINVAL);
 724         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 725                 pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
 726                 rs->rules[rs_num].inactive.rcount--;
 727         }
 728         *ticket = ++rs->rules[rs_num].inactive.ticket;
 729         rs->rules[rs_num].inactive.open = 1;
 730         return (0);
 731 }
 732
 733 int
 734 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 735 {
 736         struct pf_ruleset       *rs;
 737         struct pf_rule          *rule;
 738
 739         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 740                 return (EINVAL);
 741         rs = pf_find_ruleset(anchor);
 742         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 743             rs->rules[rs_num].inactive.ticket != ticket)
 744                 return (0);
 745         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 746                 pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
 747                 rs->rules[rs_num].inactive.rcount--;
 748         }
 749         rs->rules[rs_num].inactive.open = 0;
 750         return (0);
 751 }
 752
 753 #define PF_MD5_UPD(st, elm)                                             \
 754                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 755
 756 #define PF_MD5_UPD_STR(st, elm)                                         \
 757                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 758
 759 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 760                 (stor) = htonl((st)->elm);                              \
 761                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 762 } while (0)
 763
 764 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 765                 (stor) = htons((st)->elm);                              \
 766                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 767 } while (0)
 768
 769 void
 770 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 771 {
 772         PF_MD5_UPD(pfr, addr.type);
 773         switch (pfr->addr.type) {
 774                 case PF_ADDR_DYNIFTL:
 775                         PF_MD5_UPD(pfr, addr.v.ifname);
 776                         PF_MD5_UPD(pfr, addr.iflags);
 777                         break;
 778                 case PF_ADDR_TABLE:
 779                         PF_MD5_UPD(pfr, addr.v.tblname);
 780                         break;
 781                 case PF_ADDR_ADDRMASK:
 782                         /* XXX ignore af? */
 783                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 784                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 785                         break;
 786                 case PF_ADDR_RTLABEL:
 787                         PF_MD5_UPD(pfr, addr.v.rtlabelname);
 788                         break;
 789         }
 790
 791         PF_MD5_UPD(pfr, port[0]);
 792         PF_MD5_UPD(pfr, port[1]);
 793         PF_MD5_UPD(pfr, neg);
 794         PF_MD5_UPD(pfr, port_op);
 795 }
 796
 797 void
 798 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 799 {
 800         u_int16_t x;
 801         u_int32_t y;
 802
 803         pf_hash_rule_addr(ctx, &rule->src);
 804         pf_hash_rule_addr(ctx, &rule->dst);
 805         PF_MD5_UPD_STR(rule, label);
 806         PF_MD5_UPD_STR(rule, ifname);
 807         PF_MD5_UPD_STR(rule, match_tagname);
 808         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 809         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 810         PF_MD5_UPD_HTONL(rule, prob, y);
 811         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 812         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 813         PF_MD5_UPD(rule, uid.op);
 814         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 815         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 816         PF_MD5_UPD(rule, gid.op);
 817         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 818         PF_MD5_UPD(rule, action);
 819         PF_MD5_UPD(rule, direction);
 820         PF_MD5_UPD(rule, af);
 821         PF_MD5_UPD(rule, quick);
 822         PF_MD5_UPD(rule, ifnot);
 823         PF_MD5_UPD(rule, match_tag_not);
 824         PF_MD5_UPD(rule, natpass);
 825         PF_MD5_UPD(rule, keep_state);
 826         PF_MD5_UPD(rule, proto);
 827         PF_MD5_UPD(rule, type);
 828         PF_MD5_UPD(rule, code);
 829         PF_MD5_UPD(rule, flags);
 830         PF_MD5_UPD(rule, flagset);
 831         PF_MD5_UPD(rule, allow_opts);
 832         PF_MD5_UPD(rule, rt);
 833         PF_MD5_UPD(rule, tos);
 834 }
 835
 836 int
 837 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 838 {
 839         struct pf_ruleset       *rs;
 840         struct pf_rule          *rule, **old_array;
 841         struct pf_rulequeue     *old_rules;
 842         int                      error;
 843         u_int32_t                old_rcount;
 844
 845         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 846                 return (EINVAL);
 847         rs = pf_find_ruleset(anchor);
 848         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 849             ticket != rs->rules[rs_num].inactive.ticket)
 850                 return (EBUSY);
 851
 852         /* Calculate checksum for the main ruleset */
 853         if (rs == &pf_main_ruleset) {
 854                 error = pf_setup_pfsync_matching(rs);
 855                 if (error != 0)
 856                         return (error);
 857         }
 858
 859         /* Swap rules, keep the old. */
 860         crit_enter();
 861         old_rules = rs->rules[rs_num].active.ptr;
 862         old_rcount = rs->rules[rs_num].active.rcount;
 863         old_array = rs->rules[rs_num].active.ptr_array;
 864
 865         rs->rules[rs_num].active.ptr =
 866             rs->rules[rs_num].inactive.ptr;
 867         rs->rules[rs_num].active.ptr_array =
 868             rs->rules[rs_num].inactive.ptr_array;
 869         rs->rules[rs_num].active.rcount =
 870             rs->rules[rs_num].inactive.rcount;
 871         rs->rules[rs_num].inactive.ptr = old_rules;
 872         rs->rules[rs_num].inactive.ptr_array = old_array;
 873         rs->rules[rs_num].inactive.rcount = old_rcount;
 874
 875         rs->rules[rs_num].active.ticket =
 876             rs->rules[rs_num].inactive.ticket;
 877         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 878
 879
 880         /* Purge the old rule list. */
 881         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 882                 pf_rm_rule(old_rules, rule);
 883         if (rs->rules[rs_num].inactive.ptr_array)
 884                 kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 885         rs->rules[rs_num].inactive.ptr_array = NULL;
 886         rs->rules[rs_num].inactive.rcount = 0;
 887         rs->rules[rs_num].inactive.open = 0;
 888         pf_remove_if_empty_ruleset(rs);
 889         crit_exit();
 890         return (0);
 891 }
 892
 893 int
 894 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 895 {
 896         MD5_CTX                  ctx;
 897         struct pf_rule          *rule;
 898         int                      rs_cnt;
 899         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 900
 901         MD5Init(&ctx);
 902         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 903                 /* XXX PF_RULESET_SCRUB as well? */
 904                 if (rs_cnt == PF_RULESET_SCRUB)
 905                         continue;
 906
 907                 if (rs->rules[rs_cnt].inactive.ptr_array)
 908                         kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 909                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 910
 911                 if (rs->rules[rs_cnt].inactive.rcount) {
 912                         rs->rules[rs_cnt].inactive.ptr_array =
 913                             kmalloc(sizeof(caddr_t) *
 914                                     rs->rules[rs_cnt].inactive.rcount,
 915                                     M_TEMP, M_WAITOK);
 916
 917                         if (!rs->rules[rs_cnt].inactive.ptr_array)
 918                                 return (ENOMEM);
 919                 }
 920
 921                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 922                     entries) {
 923                         pf_hash_rule(&ctx, rule);
 924                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 925                 }
 926         }
 927
 928         MD5Final(digest, &ctx);
 929         memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
 930         return (0);
 931 }
 932
 933 int
 934 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 935     sa_family_t af)
 936 {
 937         if (pfi_dynaddr_setup(addr, af) ||
 938             pf_tbladdr_setup(ruleset, addr))
 939                 return (EINVAL);
 940
 941         return (0);
 942 }
 943
 944 void
 945 pf_addr_copyout(struct pf_addr_wrap *addr)
 946 {
 947         pfi_dynaddr_copyout(addr);
 948         pf_tbladdr_copyout(addr);
 949         pf_rtlabel_copyout(addr);
 950 }
 951
 952 int
 953 pfioctl(struct dev_ioctl_args *ap)
 954 {
 955         u_long cmd = ap->a_cmd;
 956         caddr_t addr = ap->a_data;
 957         struct pf_pooladdr      *pa = NULL;
 958         struct pf_pool          *pool = NULL;
 959         int                      error = 0;
 960
 961         lwkt_gettoken(&pf_token);
 962
 963         /* XXX keep in sync with switch() below */
 964         if (securelevel > 1)
 965                 switch (cmd) {
 966                 case DIOCGETRULES:
 967                 case DIOCGETRULE:
 968                 case DIOCGETADDRS:
 969                 case DIOCGETADDR:
 970                 case DIOCGETSTATE:
 971                 case DIOCSETSTATUSIF:
 972                 case DIOCGETSTATUS:
 973                 case DIOCCLRSTATUS:
 974                 case DIOCNATLOOK:
 975                 case DIOCSETDEBUG:
 976                 case DIOCGETSTATES:
 977                 case DIOCGETTIMEOUT:
 978                 case DIOCCLRRULECTRS:
 979                 case DIOCGETLIMIT:
 980                 case DIOCGETALTQS:
 981                 case DIOCGETALTQ:
 982                 case DIOCGETQSTATS:
 983                 case DIOCGETRULESETS:
 984                 case DIOCGETRULESET:
 985                 case DIOCRGETTABLES:
 986                 case DIOCRGETTSTATS:
 987                 case DIOCRCLRTSTATS:
 988                 case DIOCRCLRADDRS:
 989                 case DIOCRADDADDRS:
 990                 case DIOCRDELADDRS:
 991                 case DIOCRSETADDRS:
 992                 case DIOCRGETADDRS:
 993                 case DIOCRGETASTATS:
 994                 case DIOCRCLRASTATS:
 995                 case DIOCRTSTADDRS:
 996                 case DIOCOSFPGET:
 997                 case DIOCGETSRCNODES:
 998                 case DIOCCLRSRCNODES:
 999                 case DIOCIGETIFACES:
1000                 case DIOCSETIFFLAG:
1001                 case DIOCCLRIFFLAG:
1002                 case DIOCGIFSPEED:
1003                         break;
1004                 case DIOCRCLRTABLES:
1005                 case DIOCRADDTABLES:
1006                 case DIOCRDELTABLES:
1007                 case DIOCRSETTFLAGS:
1008                         if (((struct pfioc_table *)addr)->pfrio_flags &
1009                             PFR_FLAG_DUMMY)
1010                                 break; /* dummy operation ok */
1011                         lwkt_reltoken(&pf_token);
1012                         return (EPERM);
1013                 default:
1014                         lwkt_reltoken(&pf_token);
1015                         return (EPERM);
1016                 }
1017
1018         if (!(ap->a_fflag & FWRITE))
1019                 switch (cmd) {
1020                 case DIOCGETRULES:
1021                 case DIOCGETADDRS:
1022                 case DIOCGETADDR:
1023                 case DIOCGETSTATE:
1024                 case DIOCGETSTATUS:
1025                 case DIOCGETSTATES:
1026                 case DIOCGETTIMEOUT:
1027                 case DIOCGETLIMIT:
1028                 case DIOCGETALTQS:
1029                 case DIOCGETALTQ:
1030                 case DIOCGETQSTATS:
1031                 case DIOCGETRULESETS:
1032                 case DIOCGETRULESET:
1033                 case DIOCNATLOOK:
1034                 case DIOCRGETTABLES:
1035                 case DIOCRGETTSTATS:
1036                 case DIOCRGETADDRS:
1037                 case DIOCRGETASTATS:
1038                 case DIOCRTSTADDRS:
1039                 case DIOCOSFPGET:
1040                 case DIOCGETSRCNODES:
1041                 case DIOCIGETIFACES:
1042                 case DIOCGIFSPEED:
1043                         break;
1044                 case DIOCRCLRTABLES:
1045                 case DIOCRADDTABLES:
1046                 case DIOCRDELTABLES:
1047                 case DIOCRCLRTSTATS:
1048                 case DIOCRCLRADDRS:
1049                 case DIOCRADDADDRS:
1050                 case DIOCRDELADDRS:
1051                 case DIOCRSETADDRS:
1052                 case DIOCRSETTFLAGS:
1053                         if (((struct pfioc_table *)addr)->pfrio_flags &
1054                             PFR_FLAG_DUMMY)
1055                                 break; /* dummy operation ok */
1056                         lwkt_reltoken(&pf_token);
1057                         return (EACCES);
1058                 case DIOCGETRULE:
1059                         if (((struct pfioc_rule *)addr)->action ==
1060                             PF_GET_CLR_CNTR) {
1061                                 lwkt_reltoken(&pf_token);
1062                                 return (EACCES);
1063                         }
1064                         break;
1065                 default:
1066                         lwkt_reltoken(&pf_token);
1067                         return (EACCES);
1068                 }
1069
1070         switch (cmd) {
1071
1072         case DIOCSTART:
1073                 if (pf_status.running)
1074                         error = EEXIST;
1075                 else {
1076                         error = hook_pf();
1077                         if (error) {
1078                                 DPFPRINTF(PF_DEBUG_MISC,
1079                                     ("pf: pfil registration fail\n"));
1080                                 break;
1081                         }
1082                         pf_status.running = 1;
1083                         pf_status.since = time_second;
1084                         if (pf_status.stateid == 0) {
1085                                 pf_status.stateid = time_second;
1086                                 pf_status.stateid = pf_status.stateid << 32;
1087                         }
1088                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1089                 }
1090                 break;
1091
1092         case DIOCSTOP:
1093                 if (!pf_status.running)
1094                         error = ENOENT;
1095                 else {
1096                         pf_status.running = 0;
1097                         error = dehook_pf();
1098                         if (error) {
1099                                 pf_status.running = 1;
1100                                 DPFPRINTF(PF_DEBUG_MISC,
1101                                         ("pf: pfil unregistration failed\n"));
1102                         }
1103                         pf_status.since = time_second;
1104                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1105                 }
1106                 break;
1107
1108         case DIOCADDRULE: {
1109                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1110                 struct pf_ruleset       *ruleset;
1111                 struct pf_rule          *rule, *tail;
1112                 struct pf_pooladdr      *pa;
1113                 int                      rs_num;
1114
1115                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1116                 ruleset = pf_find_ruleset(pr->anchor);
1117                 if (ruleset == NULL) {
1118                         error = EINVAL;
1119                         break;
1120                 }
1121                 rs_num = pf_get_ruleset_number(pr->rule.action);
1122                 if (rs_num >= PF_RULESET_MAX) {
1123                         error = EINVAL;
1124                         break;
1125                 }
1126                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1127                         error = EINVAL;
1128                         break;
1129                 }
1130                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1131                         error = EBUSY;
1132                         break;
1133                 }
1134                 if (pr->pool_ticket != ticket_pabuf) {
1135                         error = EBUSY;
1136                         break;
1137                 }
1138                 rule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL,M_WAITOK);
1139                 if (rule == NULL) {
1140                         error = ENOMEM;
1141                         break;
1142                 }
1143                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1144                 rule->cuid = ap->a_cred->cr_ruid;
1145                 rule->cpid = 0;
1146                 rule->anchor = NULL;
1147                 rule->kif = NULL;
1148                 TAILQ_INIT(&rule->rpool.list);
1149                 /* initialize refcounting */
1150                 rule->states_cur = 0;
1151                 rule->src_nodes = 0;
1152                 rule->entries.tqe_prev = NULL;
1153 #ifndef INET
1154                 if (rule->af == AF_INET) {
1155                         kfree(rule, M_PFRULEPL);
1156                         error = EAFNOSUPPORT;
1157                         break;
1158                 }
1159 #endif /* INET */
1160 #ifndef INET6
1161                 if (rule->af == AF_INET6) {
1162                         kfree(rule, M_PFRULEPL);
1163                         error = EAFNOSUPPORT;
1164                         break;
1165                 }
1166 #endif /* INET6 */
1167                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1168                     pf_rulequeue);
1169                 if (tail)
1170                         rule->nr = tail->nr + 1;
1171                 else
1172                         rule->nr = 0;
1173                 if (rule->ifname[0]) {
1174                         rule->kif = pfi_kif_get(rule->ifname);
1175                         if (rule->kif == NULL) {
1176                                 kfree(rule, M_PFRULEPL);
1177                                 error = EINVAL;
1178                                 break;
1179                         }
1180                         pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1181                 }
1182
1183                 if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1184                         error = EBUSY;
1185
1186 #ifdef ALTQ
1187                 /* set queue IDs */
1188                 if (rule->qname[0] != 0) {
1189                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1190                                 error = EBUSY;
1191                         else if (rule->pqname[0] != 0) {
1192                                 if ((rule->pqid =
1193                                     pf_qname2qid(rule->pqname)) == 0)
1194                                         error = EBUSY;
1195                         } else
1196                                 rule->pqid = rule->qid;
1197                 }
1198 #endif
1199                 if (rule->tagname[0])
1200                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1201                                 error = EBUSY;
1202                 if (rule->match_tagname[0])
1203                         if ((rule->match_tag =
1204                             pf_tagname2tag(rule->match_tagname)) == 0)
1205                                 error = EBUSY;
1206                 if (rule->rt && !rule->direction)
1207                         error = EINVAL;
1208 #if NPFLOG > 0
1209                 if (!rule->log)
1210                         rule->logif = 0;
1211                 if (rule->logif >= PFLOGIFS_MAX)
1212                         error = EINVAL;
1213 #endif
1214                 if (pf_rtlabel_add(&rule->src.addr) ||
1215                     pf_rtlabel_add(&rule->dst.addr))
1216                         error = EBUSY;
1217                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1218                         error = EINVAL;
1219                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1220                         error = EINVAL;
1221                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1222                         error = EINVAL;
1223                 TAILQ_FOREACH(pa, &pf_pabuf, entries)
1224                         if (pf_tbladdr_setup(ruleset, &pa->addr))
1225                                 error = EINVAL;
1226
1227                 if (rule->overload_tblname[0]) {
1228                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1229                             rule->overload_tblname)) == NULL)
1230                                 error = EINVAL;
1231                         else
1232                                 rule->overload_tbl->pfrkt_flags |=
1233                                     PFR_TFLAG_ACTIVE;
1234                 }
1235
1236                 pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1237                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1238                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1239                     (rule->rt > PF_FASTROUTE)) &&
1240                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1241                         error = EINVAL;
1242
1243                 if (error) {
1244                         pf_rm_rule(NULL, rule);
1245                         break;
1246                 }
1247                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1248                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1249                     rule->bytes[0] = rule->bytes[1] = 0;
1250                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1251                     rule, entries);
1252                 ruleset->rules[rs_num].inactive.rcount++;
1253                 break;
1254         }
1255
1256         case DIOCGETRULES: {
1257                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1258                 struct pf_ruleset       *ruleset;
1259                 struct pf_rule          *tail;
1260                 int                      rs_num;
1261
1262                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1263                 ruleset = pf_find_ruleset(pr->anchor);
1264                 if (ruleset == NULL) {
1265                         error = EINVAL;
1266                         break;
1267                 }
1268                 rs_num = pf_get_ruleset_number(pr->rule.action);
1269                 if (rs_num >= PF_RULESET_MAX) {
1270                         error = EINVAL;
1271                         break;
1272                 }
1273                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1274                     pf_rulequeue);
1275                 if (tail)
1276                         pr->nr = tail->nr + 1;
1277                 else
1278                         pr->nr = 0;
1279                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1280                 break;
1281         }
1282
1283         case DIOCGETRULE: {
1284                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1285                 struct pf_ruleset       *ruleset;
1286                 struct pf_rule          *rule;
1287                 int                      rs_num, i;
1288
1289                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1290                 ruleset = pf_find_ruleset(pr->anchor);
1291                 if (ruleset == NULL) {
1292                         error = EINVAL;
1293                         break;
1294                 }
1295                 rs_num = pf_get_ruleset_number(pr->rule.action);
1296                 if (rs_num >= PF_RULESET_MAX) {
1297                         error = EINVAL;
1298                         break;
1299                 }
1300                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1301                         error = EBUSY;
1302                         break;
1303                 }
1304                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1305                 while ((rule != NULL) && (rule->nr != pr->nr))
1306                         rule = TAILQ_NEXT(rule, entries);
1307                 if (rule == NULL) {
1308                         error = EBUSY;
1309                         break;
1310                 }
1311                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1312                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1313                         error = EBUSY;
1314                         break;
1315                 }
1316                 pf_addr_copyout(&pr->rule.src.addr);
1317                 pf_addr_copyout(&pr->rule.dst.addr);
1318                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1319                         if (rule->skip[i].ptr == NULL)
1320                                 pr->rule.skip[i].nr = (uint32_t)(-1);
1321                         else
1322                                 pr->rule.skip[i].nr =
1323                                     rule->skip[i].ptr->nr;
1324
1325                 if (pr->action == PF_GET_CLR_CNTR) {
1326                         rule->evaluations = 0;
1327                         rule->packets[0] = rule->packets[1] = 0;
1328                         rule->bytes[0] = rule->bytes[1] = 0;
1329                         rule->states_tot = 0;
1330                 }
1331                 break;
1332         }
1333
1334         case DIOCCHANGERULE: {
1335                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1336                 struct pf_ruleset       *ruleset;
1337                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1338                 u_int32_t                nr = 0;
1339                 int                      rs_num;
1340
1341                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1342                     pcr->action == PF_CHANGE_GET_TICKET) &&
1343                     pcr->pool_ticket != ticket_pabuf) {
1344                         error = EBUSY;
1345                         break;
1346                 }
1347
1348                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1349                     pcr->action > PF_CHANGE_GET_TICKET) {
1350                         error = EINVAL;
1351                         break;
1352                 }
1353                 ruleset = pf_find_ruleset(pcr->anchor);
1354                 if (ruleset == NULL) {
1355                         error = EINVAL;
1356                         break;
1357                 }
1358                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1359                 if (rs_num >= PF_RULESET_MAX) {
1360                         error = EINVAL;
1361                         break;
1362                 }
1363
1364                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1365                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1366                         break;
1367                 } else {
1368                         if (pcr->ticket !=
1369                             ruleset->rules[rs_num].active.ticket) {
1370                                 error = EINVAL;
1371                                 break;
1372                         }
1373                         if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1374                                 error = EINVAL;
1375                                 break;
1376                         }
1377                 }
1378
1379                 if (pcr->action != PF_CHANGE_REMOVE) {
1380                         newrule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK|M_NULLOK);
1381                         if (newrule == NULL) {
1382                                 error = ENOMEM;
1383                                 break;
1384                         }
1385                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1386                         newrule->cuid = ap->a_cred->cr_ruid;
1387                         newrule->cpid = 0;
1388                         TAILQ_INIT(&newrule->rpool.list);
1389                         /* initialize refcounting */
1390                         newrule->states_cur = 0;
1391                         newrule->entries.tqe_prev = NULL;
1392 #ifndef INET
1393                         if (newrule->af == AF_INET) {
1394                                 kfree(newrule, M_PFRULEPL);
1395                                 error = EAFNOSUPPORT;
1396                                 break;
1397                         }
1398 #endif /* INET */
1399 #ifndef INET6
1400                         if (newrule->af == AF_INET6) {
1401                                 kfree(newrule, M_PFRULEPL);
1402                                 error = EAFNOSUPPORT;
1403                                 break;
1404                         }
1405 #endif /* INET6 */
1406                         if (newrule->ifname[0]) {
1407                                 newrule->kif = pfi_kif_get(newrule->ifname);
1408                                 if (newrule->kif == NULL) {
1409                                         kfree(newrule, M_PFRULEPL);
1410                                         error = EINVAL;
1411                                         break;
1412                                 }
1413                                 pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1414                         } else
1415                                 newrule->kif = NULL;
1416
1417                         if (newrule->rtableid > 0 &&
1418                             newrule->rtableid > rt_numfibs)
1419                                 error = EBUSY;
1420
1421 #ifdef ALTQ
1422                         /* set queue IDs */
1423                         if (newrule->qname[0] != 0) {
1424                                 if ((newrule->qid =
1425                                     pf_qname2qid(newrule->qname)) == 0)
1426                                         error = EBUSY;
1427                                 else if (newrule->pqname[0] != 0) {
1428                                         if ((newrule->pqid =
1429                                             pf_qname2qid(newrule->pqname)) == 0)
1430                                                 error = EBUSY;
1431                                 } else
1432                                         newrule->pqid = newrule->qid;
1433                         }
1434 #endif /* ALTQ */
1435                         if (newrule->tagname[0])
1436                                 if ((newrule->tag =
1437                                     pf_tagname2tag(newrule->tagname)) == 0)
1438                                         error = EBUSY;
1439                         if (newrule->match_tagname[0])
1440                                 if ((newrule->match_tag = pf_tagname2tag(
1441                                     newrule->match_tagname)) == 0)
1442                                         error = EBUSY;
1443                         if (newrule->rt && !newrule->direction)
1444                                 error = EINVAL;
1445 #if NPFLOG > 0
1446                         if (!newrule->log)
1447                                 newrule->logif = 0;
1448                         if (newrule->logif >= PFLOGIFS_MAX)
1449                                 error = EINVAL;
1450 #endif
1451                         if (pf_rtlabel_add(&newrule->src.addr) ||
1452                             pf_rtlabel_add(&newrule->dst.addr))
1453                                 error = EBUSY;
1454                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1455                                 error = EINVAL;
1456                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1457                                 error = EINVAL;
1458                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1459                                 error = EINVAL;
1460                         TAILQ_FOREACH(pa, &pf_pabuf, entries)
1461                                 if (pf_tbladdr_setup(ruleset, &pa->addr))
1462                                         error = EINVAL;
1463
1464                         if (newrule->overload_tblname[0]) {
1465                                 if ((newrule->overload_tbl = pfr_attach_table(
1466                                     ruleset, newrule->overload_tblname)) ==
1467                                     NULL)
1468                                         error = EINVAL;
1469                                 else
1470                                         newrule->overload_tbl->pfrkt_flags |=
1471                                             PFR_TFLAG_ACTIVE;
1472                         }
1473
1474                         pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1475                         if (((((newrule->action == PF_NAT) ||
1476                             (newrule->action == PF_RDR) ||
1477                             (newrule->action == PF_BINAT) ||
1478                             (newrule->rt > PF_FASTROUTE)) &&
1479                             !newrule->anchor)) &&
1480                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1481                                 error = EINVAL;
1482
1483                         if (error) {
1484                                 pf_rm_rule(NULL, newrule);
1485                                 break;
1486                         }
1487                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1488                         newrule->evaluations = 0;
1489                         newrule->packets[0] = newrule->packets[1] = 0;
1490                         newrule->bytes[0] = newrule->bytes[1] = 0;
1491                 }
1492                 pf_empty_pool(&pf_pabuf);
1493
1494                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1495                         oldrule = TAILQ_FIRST(
1496                             ruleset->rules[rs_num].active.ptr);
1497                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1498                         oldrule = TAILQ_LAST(
1499                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1500                 else {
1501                         oldrule = TAILQ_FIRST(
1502                             ruleset->rules[rs_num].active.ptr);
1503                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1504                                 oldrule = TAILQ_NEXT(oldrule, entries);
1505                         if (oldrule == NULL) {
1506                                 if (newrule != NULL)
1507                                         pf_rm_rule(NULL, newrule);
1508                                 error = EINVAL;
1509                                 break;
1510                         }
1511                 }
1512
1513                 if (pcr->action == PF_CHANGE_REMOVE) {
1514                         pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1515                         ruleset->rules[rs_num].active.rcount--;
1516                 } else {
1517                         if (oldrule == NULL)
1518                                 TAILQ_INSERT_TAIL(
1519                                     ruleset->rules[rs_num].active.ptr,
1520                                     newrule, entries);
1521                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1522                             pcr->action == PF_CHANGE_ADD_BEFORE)
1523                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1524                         else
1525                                 TAILQ_INSERT_AFTER(
1526                                     ruleset->rules[rs_num].active.ptr,
1527                                     oldrule, newrule, entries);
1528                         ruleset->rules[rs_num].active.rcount++;
1529                 }
1530
1531                 nr = 0;
1532                 TAILQ_FOREACH(oldrule,
1533                     ruleset->rules[rs_num].active.ptr, entries)
1534                         oldrule->nr = nr++;
1535
1536                 ruleset->rules[rs_num].active.ticket++;
1537
1538                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1539                 pf_remove_if_empty_ruleset(ruleset);
1540
1541                 break;
1542         }
1543
1544         case DIOCCLRSTATES: {
1545                 struct pf_state         *s, *nexts;
1546                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1547                 u_int                    killed = 0;
1548
1549                 for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
1550                         nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1551
1552                         if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1553                             s->kif->pfik_name)) {
1554                                 /* don't send out individual delete messages */
1555                                 s->sync_flags = PFSTATE_NOSYNC;
1556                                 pf_unlink_state(s);
1557                                 killed++;
1558                         }
1559                 }
1560                 psk->psk_killed = killed;
1561                 pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1562                 break;
1563         }
1564
1565         case DIOCKILLSTATES: {
1566                 struct pf_state         *s, *nexts;
1567                 struct pf_state_key     *sk;
1568                 struct pf_addr          *srcaddr, *dstaddr;
1569                 u_int16_t                srcport, dstport;
1570                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1571                 u_int                    killed = 0;
1572
1573                 if (psk->psk_pfcmp.id) {
1574                         if (psk->psk_pfcmp.creatorid == 0)
1575                                 psk->psk_pfcmp.creatorid = pf_status.hostid;
1576                         if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1577                                 /* send immediate delete of state */
1578                                 pfsync_delete_state(s);
1579                                 s->sync_flags |= PFSTATE_NOSYNC;
1580                                 pf_unlink_state(s);
1581                                 psk->psk_killed = 1;
1582                         }
1583                         break;
1584                 }
1585
1586                 for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1587                     s = nexts) {
1588                         nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1589                         sk = s->key[PF_SK_WIRE];
1590
1591                         if (s->direction == PF_OUT) {
1592                                 srcaddr = &sk->addr[1];
1593                                 dstaddr = &sk->addr[0];
1594                                 srcport = sk->port[0];
1595                                 dstport = sk->port[0];
1596                         } else {
1597                                 srcaddr = &sk->addr[0];
1598                                 dstaddr = &sk->addr[1];
1599                                 srcport = sk->port[0];
1600                                 dstport = sk->port[0];
1601                         }
1602                         if ((!psk->psk_af || sk->af == psk->psk_af)
1603                             && (!psk->psk_proto || psk->psk_proto ==
1604                             sk->proto) &&
1605                             PF_MATCHA(psk->psk_src.neg,
1606                             &psk->psk_src.addr.v.a.addr,
1607                             &psk->psk_src.addr.v.a.mask,
1608                             srcaddr, sk->af) &&
1609                             PF_MATCHA(psk->psk_dst.neg,
1610                             &psk->psk_dst.addr.v.a.addr,
1611                             &psk->psk_dst.addr.v.a.mask,
1612                             dstaddr, sk->af) &&
1613                             (psk->psk_src.port_op == 0 ||
1614                             pf_match_port(psk->psk_src.port_op,
1615                             psk->psk_src.port[0], psk->psk_src.port[1],
1616                             srcport)) &&
1617                             (psk->psk_dst.port_op == 0 ||
1618                             pf_match_port(psk->psk_dst.port_op,
1619                             psk->psk_dst.port[0], psk->psk_dst.port[1],
1620                             dstport)) &&
1621                             (!psk->psk_label[0] || (s->rule.ptr->label[0] &&
1622                             !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1623                             (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1624                             s->kif->pfik_name))) {
1625                                 /* send immediate delete of state */
1626                                 pfsync_delete_state(s);
1627                                 s->sync_flags |= PFSTATE_NOSYNC;
1628                                 pf_unlink_state(s);
1629                                 killed++;
1630                         }
1631                 }
1632                 psk->psk_killed = killed;
1633                 break;
1634         }
1635
1636         case DIOCADDSTATE: {
1637                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1638                 struct pfsync_state     *sp = &ps->state;
1639
1640                 if (sp->timeout >= PFTM_MAX &&
1641                     sp->timeout != PFTM_UNTIL_PACKET) {
1642                         error = EINVAL;
1643                         break;
1644                 }
1645                 error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1646                 break;
1647         }
1648
1649         case DIOCGETSTATE: {
1650                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1651                 struct pf_state         *s;
1652                 struct pf_state_cmp      id_key;
1653
1654                 bcopy(ps->state.id, &id_key.id, sizeof(id_key.id));
1655                 id_key.creatorid = ps->state.creatorid;
1656
1657                 s = pf_find_state_byid(&id_key);
1658                 if (s == NULL) {
1659                         error = ENOENT;
1660                         break;
1661                 }
1662                 pfsync_state_export(&ps->state, s);
1663                 break;
1664         }
1665
1666         case DIOCGETSTATES: {
1667                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1668                 struct pf_state         *state;
1669                 struct pfsync_state     *p, *pstore;
1670                 u_int32_t                nr = 0;
1671
1672                 if (ps->ps_len == 0) {
1673                         nr = pf_status.states;
1674                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1675                         break;
1676                 }
1677
1678                 pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1679
1680                 p = ps->ps_states;
1681
1682                 state = TAILQ_FIRST(&state_list);
1683                 while (state) {
1684                         if (state->timeout != PFTM_UNLINKED) {
1685                                 if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1686                                         break;
1687                                 pfsync_state_export(pstore, state);
1688                                 error = copyout(pstore, p, sizeof(*p));
1689                                 if (error) {
1690                                         kfree(pstore, M_TEMP);
1691                                         goto fail;
1692                                 }
1693                                 p++;
1694                                 nr++;
1695                         }
1696                         state = TAILQ_NEXT(state, entry_list);
1697                 }
1698
1699                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1700
1701                 kfree(pstore, M_TEMP);
1702                 break;
1703         }
1704
1705         case DIOCGETSTATUS: {
1706                 struct pf_status *s = (struct pf_status *)addr;
1707                 bcopy(&pf_status, s, sizeof(struct pf_status));
1708                 pfi_update_status(s->ifname, s);
1709                 break;
1710         }
1711
1712         case DIOCSETSTATUSIF: {
1713                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1714
1715                 if (pi->ifname[0] == 0) {
1716                         bzero(pf_status.ifname, IFNAMSIZ);
1717                         break;
1718                 }
1719                 strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1720                 break;
1721         }
1722
1723         case DIOCCLRSTATUS: {
1724                 bzero(pf_status.counters, sizeof(pf_status.counters));
1725                 bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1726                 bzero(pf_status.scounters, sizeof(pf_status.scounters));
1727                 pf_status.since = time_second;
1728                 if (*pf_status.ifname)
1729                         pfi_update_status(pf_status.ifname, NULL);
1730                 break;
1731         }
1732
1733         case DIOCNATLOOK: {
1734                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1735                 struct pf_state_key     *sk;
1736                 struct pf_state         *state;
1737                 struct pf_state_key_cmp  key;
1738                 int                      m = 0, direction = pnl->direction;
1739                 int                      sidx, didx;
1740
1741                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1742                 sidx = (direction == PF_IN) ? 1 : 0;
1743                 didx = (direction == PF_IN) ? 0 : 1;
1744
1745                 if (!pnl->proto ||
1746                     PF_AZERO(&pnl->saddr, pnl->af) ||
1747                     PF_AZERO(&pnl->daddr, pnl->af) ||
1748                     ((pnl->proto == IPPROTO_TCP ||
1749                     pnl->proto == IPPROTO_UDP) &&
1750                     (!pnl->dport || !pnl->sport)))
1751                         error = EINVAL;
1752                 else {
1753                         key.af = pnl->af;
1754                         key.proto = pnl->proto;
1755                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1756                         key.port[sidx] = pnl->sport;
1757                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1758                         key.port[didx] = pnl->dport;
1759
1760                         state = pf_find_state_all(&key, direction, &m);
1761
1762                         if (m > 1)
1763                                 error = E2BIG;  /* more than one state */
1764                         else if (state != NULL) {
1765                                 sk = state->key[sidx];
1766                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1767                                 pnl->rsport = sk->port[sidx];
1768                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1769                                 pnl->rdport = sk->port[didx];
1770                         } else
1771                                 error = ENOENT;
1772                 }
1773                 break;
1774         }
1775
1776         case DIOCSETTIMEOUT: {
1777                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1778                 int              old;
1779
1780                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1781                     pt->seconds < 0) {
1782                         error = EINVAL;
1783                         goto fail;
1784                 }
1785                 old = pf_default_rule.timeout[pt->timeout];
1786                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1787                         pt->seconds = 1;
1788                 pf_default_rule.timeout[pt->timeout] = pt->seconds;
1789                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1790                         wakeup(pf_purge_thread);
1791                 pt->seconds = old;
1792                 break;
1793         }
1794
1795         case DIOCGETTIMEOUT: {
1796                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1797
1798                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1799                         error = EINVAL;
1800                         goto fail;
1801                 }
1802                 pt->seconds = pf_default_rule.timeout[pt->timeout];
1803                 break;
1804         }
1805
1806         case DIOCGETLIMIT: {
1807                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1808
1809                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1810                         error = EINVAL;
1811                         goto fail;
1812                 }
1813                 pl->limit = pf_pool_limits[pl->index].limit;
1814                 break;
1815         }
1816
1817         case DIOCSETLIMIT: {
1818                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1819                 int                      old_limit;
1820
1821                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1822                     pf_pool_limits[pl->index].pp == NULL) {
1823                         error = EINVAL;
1824                         goto fail;
1825                 }
1826
1827                 /* XXX Get an API to set limits on the zone/pool */
1828                 old_limit = pf_pool_limits[pl->index].limit;
1829                 pf_pool_limits[pl->index].limit = pl->limit;
1830                 pl->limit = old_limit;
1831                 break;
1832         }
1833
1834         case DIOCSETDEBUG: {
1835                 u_int32_t       *level = (u_int32_t *)addr;
1836
1837                 pf_status.debug = *level;
1838                 break;
1839         }
1840
1841         case DIOCCLRRULECTRS: {
1842                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1843                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1844                 struct pf_rule          *rule;
1845
1846                 TAILQ_FOREACH(rule,
1847                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1848                         rule->evaluations = 0;
1849                         rule->packets[0] = rule->packets[1] = 0;
1850                         rule->bytes[0] = rule->bytes[1] = 0;
1851                 }
1852                 break;
1853         }
1854
1855         case DIOCGIFSPEED: {
1856                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1857                 struct pf_ifspeed       ps;
1858                 struct ifnet            *ifp;
1859
1860                 if (psp->ifname[0] != 0) {
1861                         /* Can we completely trust user-land? */
1862                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1863                         ifp = ifunit(ps.ifname);
1864                         if (ifp )
1865                                 psp->baudrate = ifp->if_baudrate;
1866                         else
1867                                 error = EINVAL;
1868                 } else
1869                         error = EINVAL;
1870                 break;
1871         }
1872 #ifdef ALTQ
1873         case DIOCSTARTALTQ: {
1874                 struct pf_altq          *altq;
1875
1876                 /* enable all altq interfaces on active list */
1877                 TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1878                         if (altq->qname[0] == 0) {
1879                                 error = pf_enable_altq(altq);
1880                                 if (error != 0)
1881                                         break;
1882                         }
1883                 }
1884                 if (error == 0)
1885                         pf_altq_running = 1;
1886                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1887                 break;
1888         }
1889
1890         case DIOCSTOPALTQ: {
1891                 struct pf_altq          *altq;
1892
1893                 /* disable all altq interfaces on active list */
1894                 TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1895                         if (altq->qname[0] == 0) {
1896                                 error = pf_disable_altq(altq);
1897                                 if (error != 0)
1898                                         break;
1899                         }
1900                 }
1901                 if (error == 0)
1902                         pf_altq_running = 0;
1903                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1904                 break;
1905         }
1906
1907         case DIOCADDALTQ: {
1908                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
1909                 struct pf_altq          *altq, *a;
1910
1911                 if (pa->ticket != ticket_altqs_inactive) {
1912                         error = EBUSY;
1913                         break;
1914                 }
1915                 altq = kmalloc(sizeof(struct pf_altq), M_PFALTQPL, M_WAITOK|M_NULLOK);
1916                 if (altq == NULL) {
1917                         error = ENOMEM;
1918                         break;
1919                 }
1920                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1921
1922                 /*
1923                  * if this is for a queue, find the discipline and
1924                  * copy the necessary fields
1925                  */
1926                 if (altq->qname[0] != 0) {
1927                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
1928                                 error = EBUSY;
1929                                 kfree(altq, M_PFALTQPL);
1930                                 break;
1931                         }
1932                         altq->altq_disc = NULL;
1933                         TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
1934                                 if (strncmp(a->ifname, altq->ifname,
1935                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
1936                                         altq->altq_disc = a->altq_disc;
1937                                         break;
1938                                 }
1939                         }
1940                 }
1941
1942                 error = altq_add(altq);
1943                 if (error) {
1944                         kfree(altq, M_PFALTQPL);
1945                         break;
1946                 }
1947
1948                 TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
1949                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1950                 break;
1951         }
1952
1953         case DIOCGETALTQS: {
1954                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
1955                 struct pf_altq          *altq;
1956
1957                 pa->nr = 0;
1958                 TAILQ_FOREACH(altq, pf_altqs_active, entries)
1959                         pa->nr++;
1960                 pa->ticket = ticket_altqs_active;
1961                 break;
1962         }
1963
1964         case DIOCGETALTQ: {
1965                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
1966                 struct pf_altq          *altq;
1967                 u_int32_t                nr;
1968
1969                 if (pa->ticket != ticket_altqs_active) {
1970                         error = EBUSY;
1971                         break;
1972                 }
1973                 nr = 0;
1974                 altq = TAILQ_FIRST(pf_altqs_active);
1975                 while ((altq != NULL) && (nr < pa->nr)) {
1976                         altq = TAILQ_NEXT(altq, entries);
1977                         nr++;
1978                 }
1979                 if (altq == NULL) {
1980                         error = EBUSY;
1981                         break;
1982                 }
1983                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1984                 break;
1985         }
1986
1987         case DIOCCHANGEALTQ:
1988                 /* CHANGEALTQ not supported yet! */
1989                 error = ENODEV;
1990                 break;
1991
1992         case DIOCGETQSTATS: {
1993                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
1994                 struct pf_altq          *altq;
1995                 u_int32_t                nr;
1996                 int                      nbytes;
1997
1998                 if (pq->ticket != ticket_altqs_active) {
1999                         error = EBUSY;
2000                         break;
2001                 }
2002                 nbytes = pq->nbytes;
2003                 nr = 0;
2004                 altq = TAILQ_FIRST(pf_altqs_active);
2005                 while ((altq != NULL) && (nr < pq->nr)) {
2006                         altq = TAILQ_NEXT(altq, entries);
2007                         nr++;
2008                 }
2009                 if (altq == NULL) {
2010                         error = EBUSY;
2011                         break;
2012                 }
2013                 error = altq_getqstats(altq, pq->buf, &nbytes);
2014                 if (error == 0) {
2015                         pq->scheduler = altq->scheduler;
2016                         pq->nbytes = nbytes;
2017                 }
2018                 break;
2019         }
2020 #endif /* ALTQ */
2021
2022         case DIOCBEGINADDRS: {
2023                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2024
2025                 pf_empty_pool(&pf_pabuf);
2026                 pp->ticket = ++ticket_pabuf;
2027                 break;
2028         }
2029
2030         case DIOCADDADDR: {
2031                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2032
2033                 if (pp->ticket != ticket_pabuf) {
2034                         error = EBUSY;
2035                         break;
2036                 }
2037 #ifndef INET
2038                 if (pp->af == AF_INET) {
2039                         error = EAFNOSUPPORT;
2040                         break;
2041                 }
2042 #endif /* INET */
2043 #ifndef INET6
2044                 if (pp->af == AF_INET6) {
2045                         error = EAFNOSUPPORT;
2046                         break;
2047                 }
2048 #endif /* INET6 */
2049                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2050                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2051                     pp->addr.addr.type != PF_ADDR_TABLE) {
2052                         error = EINVAL;
2053                         break;
2054                 }
2055                 pa = kmalloc(sizeof(struct pf_altq), M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2056                 if (pa == NULL) {
2057                         error = ENOMEM;
2058                         break;
2059                 }
2060                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2061                 if (pa->ifname[0]) {
2062                         pa->kif = pfi_kif_get(pa->ifname);
2063                         if (pa->kif == NULL) {
2064                                 kfree(ap, M_PFPOOLADDRPL);
2065                                 error = EINVAL;
2066                                 break;
2067                         }
2068                         pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2069                 }
2070                 if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2071                         pfi_dynaddr_remove(&pa->addr);
2072                         pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2073                         kfree(pa, M_PFPOOLADDRPL);
2074                         error = EINVAL;
2075                         break;
2076                 }
2077                 TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2078                 break;
2079         }
2080
2081         case DIOCGETADDRS: {
2082                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2083
2084                 pp->nr = 0;
2085                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2086                     pp->r_num, 0, 1, 0);
2087                 if (pool == NULL) {
2088                         error = EBUSY;
2089                         break;
2090                 }
2091                 TAILQ_FOREACH(pa, &pool->list, entries)
2092                         pp->nr++;
2093                 break;
2094         }
2095
2096         case DIOCGETADDR: {
2097                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2098                 u_int32_t                nr = 0;
2099
2100                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2101                     pp->r_num, 0, 1, 1);
2102                 if (pool == NULL) {
2103                         error = EBUSY;
2104                         break;
2105                 }
2106                 pa = TAILQ_FIRST(&pool->list);
2107                 while ((pa != NULL) && (nr < pp->nr)) {
2108                         pa = TAILQ_NEXT(pa, entries);
2109                         nr++;
2110                 }
2111                 if (pa == NULL) {
2112                         error = EBUSY;
2113                         break;
2114                 }
2115                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2116                 pf_addr_copyout(&pp->addr.addr);
2117                 break;
2118         }
2119
2120         case DIOCCHANGEADDR: {
2121                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2122                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2123                 struct pf_ruleset       *ruleset;
2124
2125                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2126                     pca->action > PF_CHANGE_REMOVE) {
2127                         error = EINVAL;
2128                         break;
2129                 }
2130                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2131                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2132                     pca->addr.addr.type != PF_ADDR_TABLE) {
2133                         error = EINVAL;
2134                         break;
2135                 }
2136
2137                 ruleset = pf_find_ruleset(pca->anchor);
2138                 if (ruleset == NULL) {
2139                         error = EBUSY;
2140                         break;
2141                 }
2142                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2143                     pca->r_num, pca->r_last, 1, 1);
2144                 if (pool == NULL) {
2145                         error = EBUSY;
2146                         break;
2147                 }
2148                 if (pca->action != PF_CHANGE_REMOVE) {
2149                         newpa = kmalloc(sizeof(struct pf_pooladdr),
2150                                 M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2151                         if (newpa == NULL) {
2152                                 error = ENOMEM;
2153                                 break;
2154                         }
2155                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2156 #ifndef INET
2157                         if (pca->af == AF_INET) {
2158                                 kfree(newpa, M_PFPOOLADDRPL);
2159                                 error = EAFNOSUPPORT;
2160                                 break;
2161                         }
2162 #endif /* INET */
2163 #ifndef INET6
2164                         if (pca->af == AF_INET6) {
2165                                 kfree(newpa, M_PFPOOLADDRPL);
2166                                 error = EAFNOSUPPORT;
2167                                 break;
2168                         }
2169 #endif /* INET6 */
2170                         if (newpa->ifname[0]) {
2171                                 newpa->kif = pfi_kif_get(newpa->ifname);
2172                                 if (newpa->kif == NULL) {
2173                                         kfree(newpa, M_PFPOOLADDRPL);
2174                                         error = EINVAL;
2175                                         break;
2176                                 }
2177                                 pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2178                         } else
2179                                 newpa->kif = NULL;
2180                         if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2181                             pf_tbladdr_setup(ruleset, &newpa->addr)) {
2182                                 pfi_dynaddr_remove(&newpa->addr);
2183                                 pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2184                                 kfree(newpa, M_PFPOOLADDRPL);
2185                                 error = EINVAL;
2186                                 break;
2187                         }
2188                 }
2189
2190                 if (pca->action == PF_CHANGE_ADD_HEAD)
2191                         oldpa = TAILQ_FIRST(&pool->list);
2192                 else if (pca->action == PF_CHANGE_ADD_TAIL)
2193                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2194                 else {
2195                         int     i = 0;
2196
2197                         oldpa = TAILQ_FIRST(&pool->list);
2198                         while ((oldpa != NULL) && (i < pca->nr)) {
2199                                 oldpa = TAILQ_NEXT(oldpa, entries);
2200                                 i++;
2201                         }
2202                         if (oldpa == NULL) {
2203                                 error = EINVAL;
2204                                 break;
2205                         }
2206                 }
2207
2208                 if (pca->action == PF_CHANGE_REMOVE) {
2209                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2210                         pfi_dynaddr_remove(&oldpa->addr);
2211                         pf_tbladdr_remove(&oldpa->addr);
2212                         pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2213                         kfree(oldpa, M_PFPOOLADDRPL);
2214                 } else {
2215                         if (oldpa == NULL)
2216                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2217                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2218                             pca->action == PF_CHANGE_ADD_BEFORE)
2219                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2220                         else
2221                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2222                                     newpa, entries);
2223                 }
2224
2225                 pool->cur = TAILQ_FIRST(&pool->list);
2226                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2227                     pca->af);
2228                 break;
2229         }
2230
2231         case DIOCGETRULESETS: {
2232                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2233                 struct pf_ruleset       *ruleset;
2234                 struct pf_anchor        *anchor;
2235
2236                 pr->path[sizeof(pr->path) - 1] = 0;
2237                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2238                         error = EINVAL;
2239                         break;
2240                 }
2241                 pr->nr = 0;
2242                 if (ruleset->anchor == NULL) {
2243                         /* XXX kludge for pf_main_ruleset */
2244                         RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2245                                 if (anchor->parent == NULL)
2246                                         pr->nr++;
2247                 } else {
2248                         RB_FOREACH(anchor, pf_anchor_node,
2249                             &ruleset->anchor->children)
2250                                 pr->nr++;
2251                 }
2252                 break;
2253         }
2254
2255         case DIOCGETRULESET: {
2256                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2257                 struct pf_ruleset       *ruleset;
2258                 struct pf_anchor        *anchor;
2259                 u_int32_t                nr = 0;
2260
2261                 pr->path[sizeof(pr->path) - 1] = 0;
2262                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2263                         error = EINVAL;
2264                         break;
2265                 }
2266                 pr->name[0] = 0;
2267                 if (ruleset->anchor == NULL) {
2268                         /* XXX kludge for pf_main_ruleset */
2269                         RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2270                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2271                                         strlcpy(pr->name, anchor->name,
2272                                             sizeof(pr->name));
2273                                         break;
2274                                 }
2275                 } else {
2276                         RB_FOREACH(anchor, pf_anchor_node,
2277                             &ruleset->anchor->children)
2278                                 if (nr++ == pr->nr) {
2279                                         strlcpy(pr->name, anchor->name,
2280                                             sizeof(pr->name));
2281                                         break;
2282                                 }
2283                 }
2284                 if (!pr->name[0])
2285                         error = EBUSY;
2286                 break;
2287         }
2288
2289         case DIOCRCLRTABLES: {
2290                 struct pfioc_table *io = (struct pfioc_table *)addr;
2291
2292                 if (io->pfrio_esize != 0) {
2293                         error = ENODEV;
2294                         break;
2295                 }
2296                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2297                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2298                 break;
2299         }
2300
2301         case DIOCRADDTABLES: {
2302                 struct pfioc_table *io = (struct pfioc_table *)addr;
2303
2304                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2305                         error = ENODEV;
2306                         break;
2307                 }
2308                 error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2309                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2310                 break;
2311         }
2312
2313         case DIOCRDELTABLES: {
2314                 struct pfioc_table *io = (struct pfioc_table *)addr;
2315
2316                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2317                         error = ENODEV;
2318                         break;
2319                 }
2320                 error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2321                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2322                 break;
2323         }
2324
2325         case DIOCRGETTABLES: {
2326                 struct pfioc_table *io = (struct pfioc_table *)addr;
2327
2328                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2329                         error = ENODEV;
2330                         break;
2331                 }
2332                 error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2333                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2334                 break;
2335         }
2336
2337         case DIOCRGETTSTATS: {
2338                 struct pfioc_table *io = (struct pfioc_table *)addr;
2339
2340                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2341                         error = ENODEV;
2342                         break;
2343                 }
2344                 error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2345                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2346                 break;
2347         }
2348
2349         case DIOCRCLRTSTATS: {
2350                 struct pfioc_table *io = (struct pfioc_table *)addr;
2351
2352                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2353                         error = ENODEV;
2354                         break;
2355                 }
2356                 error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2357                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2358                 break;
2359         }
2360
2361         case DIOCRSETTFLAGS: {
2362                 struct pfioc_table *io = (struct pfioc_table *)addr;
2363
2364                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2365                         error = ENODEV;
2366                         break;
2367                 }
2368                 error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2369                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2370                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2371                 break;
2372         }
2373
2374         case DIOCRCLRADDRS: {
2375                 struct pfioc_table *io = (struct pfioc_table *)addr;
2376
2377                 if (io->pfrio_esize != 0) {
2378                         error = ENODEV;
2379                         break;
2380                 }
2381                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2382                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2383                 break;
2384         }
2385
2386         case DIOCRADDADDRS: {
2387                 struct pfioc_table *io = (struct pfioc_table *)addr;
2388
2389                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2390                         error = ENODEV;
2391                         break;
2392                 }
2393                 error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2394                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2395                     PFR_FLAG_USERIOCTL);
2396                 break;
2397         }
2398
2399         case DIOCRDELADDRS: {
2400                 struct pfioc_table *io = (struct pfioc_table *)addr;
2401
2402                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2403                         error = ENODEV;
2404                         break;
2405                 }
2406                 error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2407                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2408                     PFR_FLAG_USERIOCTL);
2409                 break;
2410         }
2411
2412         case DIOCRSETADDRS: {
2413                 struct pfioc_table *io = (struct pfioc_table *)addr;
2414
2415                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2416                         error = ENODEV;
2417                         break;
2418                 }
2419                 error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2420                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2421                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2422                     PFR_FLAG_USERIOCTL, 0);
2423                 break;
2424         }
2425
2426         case DIOCRGETADDRS: {
2427                 struct pfioc_table *io = (struct pfioc_table *)addr;
2428
2429                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2430                         error = ENODEV;
2431                         break;
2432                 }
2433                 error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2434                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2435                 break;
2436         }
2437
2438         case DIOCRGETASTATS: {
2439                 struct pfioc_table *io = (struct pfioc_table *)addr;
2440
2441                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2442                         error = ENODEV;
2443                         break;
2444                 }
2445                 error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2446                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2447                 break;
2448         }
2449
2450         case DIOCRCLRASTATS: {
2451                 struct pfioc_table *io = (struct pfioc_table *)addr;
2452
2453                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2454                         error = ENODEV;
2455                         break;
2456                 }
2457                 error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2458                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2459                     PFR_FLAG_USERIOCTL);
2460                 break;
2461         }
2462
2463         case DIOCRTSTADDRS: {
2464                 struct pfioc_table *io = (struct pfioc_table *)addr;
2465
2466                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2467                         error = ENODEV;
2468                         break;
2469                 }
2470                 error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2471                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2472                     PFR_FLAG_USERIOCTL);
2473                 break;
2474         }
2475
2476         case DIOCRINADEFINE: {
2477                 struct pfioc_table *io = (struct pfioc_table *)addr;
2478
2479                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2480                         error = ENODEV;
2481                         break;
2482                 }
2483                 error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2484                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2485                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2486                 break;
2487         }
2488
2489         case DIOCOSFPADD: {
2490                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2491                 error = pf_osfp_add(io);
2492                 break;
2493         }
2494
2495         case DIOCOSFPGET: {
2496                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2497                 error = pf_osfp_get(io);
2498                 break;
2499         }
2500
2501         case DIOCXBEGIN: {
2502                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2503                 struct pfioc_trans_e    *ioe;
2504                 struct pfr_table        *table;
2505                 int                      i;
2506
2507                 if (io->esize != sizeof(*ioe)) {
2508                         error = ENODEV;
2509                         goto fail;
2510                 }
2511                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2512                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2513                 for (i = 0; i < io->size; i++) {
2514                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2515                                 kfree(table, M_TEMP);
2516                                 kfree(ioe, M_TEMP);
2517                                 error = EFAULT;
2518                                 goto fail;
2519                         }
2520                         switch (ioe->rs_num) {
2521 #ifdef ALTQ
2522                         case PF_RULESET_ALTQ:
2523                                 if (ioe->anchor[0]) {
2524                                         kfree(table, M_TEMP);
2525                                         kfree(ioe, M_TEMP);
2526                                         error = EINVAL;
2527                                         goto fail;
2528                                 }
2529                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2530                                         kfree(table, M_TEMP);
2531                                         kfree(ioe, M_TEMP);
2532                                         goto fail;
2533                                 }
2534                                 break;
2535 #endif /* ALTQ */
2536                         case PF_RULESET_TABLE:
2537                                 bzero(table, sizeof(*table));
2538                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2539                                     sizeof(table->pfrt_anchor));
2540                                 if ((error = pfr_ina_begin(table,
2541                                     &ioe->ticket, NULL, 0))) {
2542                                         kfree(table, M_TEMP);
2543                                         kfree(ioe, M_TEMP);
2544                                         goto fail;
2545                                 }
2546                                 break;
2547                         default:
2548                                 if ((error = pf_begin_rules(&ioe->ticket,
2549                                     ioe->rs_num, ioe->anchor))) {
2550                                         kfree(table, M_TEMP);
2551                                         kfree(ioe, M_TEMP);
2552                                         goto fail;
2553                                 }
2554                                 break;
2555                         }
2556                         if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2557                                 kfree(table, M_TEMP);
2558                                 kfree(ioe, M_TEMP);
2559                                 error = EFAULT;
2560                                 goto fail;
2561                         }
2562                 }
2563                 kfree(table, M_TEMP);
2564                 kfree(ioe, M_TEMP);
2565                 break;
2566         }
2567
2568         case DIOCXROLLBACK: {
2569                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2570                 struct pfioc_trans_e    *ioe;
2571                 struct pfr_table        *table;
2572                 int                      i;
2573
2574                 if (io->esize != sizeof(*ioe)) {
2575                         error = ENODEV;
2576                         goto fail;
2577                 }
2578                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2579                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2580                 for (i = 0; i < io->size; i++) {
2581                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2582                                 kfree(table, M_TEMP);
2583                                 kfree(ioe, M_TEMP);
2584                                 error = EFAULT;
2585                                 goto fail;
2586                         }
2587                         switch (ioe->rs_num) {
2588 #ifdef ALTQ
2589                         case PF_RULESET_ALTQ:
2590                                 if (ioe->anchor[0]) {
2591                                         kfree(table, M_TEMP);
2592                                         kfree(ioe, M_TEMP);
2593                                         error = EINVAL;
2594                                         goto fail;
2595                                 }
2596                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2597                                         kfree(table, M_TEMP);
2598                                         kfree(ioe, M_TEMP);
2599                                         goto fail; /* really bad */
2600                                 }
2601                                 break;
2602 #endif /* ALTQ */
2603                         case PF_RULESET_TABLE:
2604                                 bzero(table, sizeof(*table));
2605                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2606                                     sizeof(table->pfrt_anchor));
2607                                 if ((error = pfr_ina_rollback(table,
2608                                     ioe->ticket, NULL, 0))) {
2609                                         kfree(table, M_TEMP);
2610                                         kfree(ioe, M_TEMP);
2611                                         goto fail; /* really bad */
2612                                 }
2613                                 break;
2614                         default:
2615                                 if ((error = pf_rollback_rules(ioe->ticket,
2616                                     ioe->rs_num, ioe->anchor))) {
2617                                         kfree(table, M_TEMP);
2618                                         kfree(ioe, M_TEMP);
2619                                         goto fail; /* really bad */
2620                                 }
2621                                 break;
2622                         }
2623                 }
2624                 kfree(table, M_TEMP);
2625                 kfree(ioe, M_TEMP);
2626                 break;
2627         }
2628
2629         case DIOCXCOMMIT: {
2630                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2631                 struct pfioc_trans_e    *ioe;
2632                 struct pfr_table        *table;
2633                 struct pf_ruleset       *rs;
2634                 int                      i;
2635
2636                 if (io->esize != sizeof(*ioe)) {
2637                         error = ENODEV;
2638                         goto fail;
2639                 }
2640                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2641                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2642                 /* first makes sure everything will succeed */
2643                 for (i = 0; i < io->size; i++) {
2644                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2645                                 kfree(table, M_TEMP);
2646                                 kfree(ioe, M_TEMP);
2647                                 error = EFAULT;
2648                                 goto fail;
2649                         }
2650                         switch (ioe->rs_num) {
2651 #ifdef ALTQ
2652                         case PF_RULESET_ALTQ:
2653                                 if (ioe->anchor[0]) {
2654                                         kfree(table, M_TEMP);
2655                                         kfree(ioe, M_TEMP);
2656                                         error = EINVAL;
2657                                         goto fail;
2658                                 }
2659                                 if (!altqs_inactive_open || ioe->ticket !=
2660                                     ticket_altqs_inactive) {
2661                                         kfree(table, M_TEMP);
2662                                         kfree(ioe, M_TEMP);
2663                                         error = EBUSY;
2664                                         goto fail;
2665                                 }
2666                                 break;
2667 #endif /* ALTQ */
2668                         case PF_RULESET_TABLE:
2669                                 rs = pf_find_ruleset(ioe->anchor);
2670                                 if (rs == NULL || !rs->topen || ioe->ticket !=
2671                                      rs->tticket) {
2672                                         kfree(table, M_TEMP);
2673                                         kfree(ioe, M_TEMP);
2674                                         error = EBUSY;
2675                                         goto fail;
2676                                 }
2677                                 break;
2678                         default:
2679                                 if (ioe->rs_num < 0 || ioe->rs_num >=
2680                                     PF_RULESET_MAX) {
2681                                         kfree(table, M_TEMP);
2682                                         kfree(ioe, M_TEMP);
2683                                         error = EINVAL;
2684                                         goto fail;
2685                                 }
2686                                 rs = pf_find_ruleset(ioe->anchor);
2687                                 if (rs == NULL ||
2688                                     !rs->rules[ioe->rs_num].inactive.open ||
2689                                     rs->rules[ioe->rs_num].inactive.ticket !=
2690                                     ioe->ticket) {
2691                                         kfree(table, M_TEMP);
2692                                         kfree(ioe, M_TEMP);
2693                                         error = EBUSY;
2694                                         goto fail;
2695                                 }
2696                                 break;
2697                         }
2698                 }
2699                 /* now do the commit - no errors should happen here */
2700                 for (i = 0; i < io->size; i++) {
2701                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2702                                 kfree(table, M_TEMP);
2703                                 kfree(ioe, M_TEMP);
2704                                 error = EFAULT;
2705                                 goto fail;
2706                         }
2707                         switch (ioe->rs_num) {
2708 #ifdef ALTQ
2709                         case PF_RULESET_ALTQ:
2710                                 if ((error = pf_commit_altq(ioe->ticket))) {
2711                                         kfree(table, M_TEMP);
2712                                         kfree(ioe, M_TEMP);
2713                                         goto fail; /* really bad */
2714                                 }
2715                                 break;
2716 #endif /* ALTQ */
2717                         case PF_RULESET_TABLE:
2718                                 bzero(table, sizeof(*table));
2719                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2720                                     sizeof(table->pfrt_anchor));
2721                                 if ((error = pfr_ina_commit(table, ioe->ticket,
2722                                     NULL, NULL, 0))) {
2723                                         kfree(table, M_TEMP);
2724                                         kfree(ioe, M_TEMP);
2725                                         goto fail; /* really bad */
2726                                 }
2727                                 break;
2728                         default:
2729                                 if ((error = pf_commit_rules(ioe->ticket,
2730                                     ioe->rs_num, ioe->anchor))) {
2731                                         kfree(table, M_TEMP);
2732                                         kfree(ioe, M_TEMP);
2733                                         goto fail; /* really bad */
2734                                 }
2735                                 break;
2736                         }
2737                 }
2738                 kfree(table, M_TEMP);
2739                 kfree(ioe, M_TEMP);
2740                 break;
2741         }
2742
2743         case DIOCGETSRCNODES: {
2744                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
2745                 struct pf_src_node      *n, *p, *pstore;
2746                 u_int32_t                nr = 0;
2747                 int                      space = psn->psn_len;
2748
2749                 if (space == 0) {
2750                         RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2751                                 nr++;
2752                         psn->psn_len = sizeof(struct pf_src_node) * nr;
2753                         break;
2754                 }
2755
2756                 pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2757
2758                 p = psn->psn_src_nodes;
2759                 RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2760                         int     secs = time_second, diff;
2761
2762                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
2763                                 break;
2764
2765                         bcopy(n, pstore, sizeof(*pstore));
2766                         if (n->rule.ptr != NULL)
2767                                 pstore->rule.nr = n->rule.ptr->nr;
2768                         pstore->creation = secs - pstore->creation;
2769                         if (pstore->expire > secs)
2770                                 pstore->expire -= secs;
2771                         else
2772                                 pstore->expire = 0;
2773
2774                         /* adjust the connection rate estimate */
2775                         diff = secs - n->conn_rate.last;
2776                         if (diff >= n->conn_rate.seconds)
2777                                 pstore->conn_rate.count = 0;
2778                         else
2779                                 pstore->conn_rate.count -=
2780                                     n->conn_rate.count * diff /
2781                                     n->conn_rate.seconds;
2782
2783                         error = copyout(pstore, p, sizeof(*p));
2784                         if (error) {
2785                                 kfree(pstore, M_TEMP);
2786                                 goto fail;
2787                         }
2788                         p++;
2789                         nr++;
2790                 }
2791                 psn->psn_len = sizeof(struct pf_src_node) * nr;
2792
2793                 kfree(pstore, M_TEMP);
2794                 break;
2795         }
2796
2797         case DIOCCLRSRCNODES: {
2798                 struct pf_src_node      *n;
2799                 struct pf_state         *state;
2800
2801                 RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2802                         state->src_node = NULL;
2803                         state->nat_src_node = NULL;
2804                 }
2805                 RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2806                         n->expire = 1;
2807                         n->states = 0;
2808                 }
2809                 pf_purge_expired_src_nodes(1);
2810                 pf_status.src_nodes = 0;
2811                 break;
2812         }
2813
2814         case DIOCKILLSRCNODES: {
2815                 struct pf_src_node      *sn;
2816                 struct pf_state         *s;
2817                 struct pfioc_src_node_kill *psnk =
2818                     (struct pfioc_src_node_kill *)addr;
2819                 u_int                   killed = 0;
2820
2821                 RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2822                         if (PF_MATCHA(psnk->psnk_src.neg,
2823                                 &psnk->psnk_src.addr.v.a.addr,
2824                                 &psnk->psnk_src.addr.v.a.mask,
2825                                 &sn->addr, sn->af) &&
2826                             PF_MATCHA(psnk->psnk_dst.neg,
2827                                 &psnk->psnk_dst.addr.v.a.addr,
2828                                 &psnk->psnk_dst.addr.v.a.mask,
2829                                 &sn->raddr, sn->af)) {
2830                                 /* Handle state to src_node linkage */
2831                                 if (sn->states != 0) {
2832                                         RB_FOREACH(s, pf_state_tree_id,
2833                                             &tree_id) {
2834                                                 if (s->src_node == sn)
2835                                                         s->src_node = NULL;
2836                                                 if (s->nat_src_node == sn)
2837                                                         s->nat_src_node = NULL;
2838                                         }
2839                                         sn->states = 0;
2840                                 }
2841                                 sn->expire = 1;
2842                                 killed++;
2843                         }
2844                 }
2845
2846                 if (killed > 0)
2847                         pf_purge_expired_src_nodes(1);
2848
2849                 psnk->psnk_killed = killed;
2850                 break;
2851         }
2852
2853         case DIOCSETHOSTID: {
2854                 u_int32_t       *hostid = (u_int32_t *)addr;
2855
2856                 if (*hostid == 0)
2857                         pf_status.hostid = karc4random();
2858                 else
2859                         pf_status.hostid = *hostid;
2860                 break;
2861         }
2862
2863         case DIOCOSFPFLUSH:
2864                 crit_enter();
2865                 pf_osfp_flush();
2866                 crit_exit();
2867                 break;
2868
2869         case DIOCIGETIFACES: {
2870                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
2871
2872                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2873                         error = ENODEV;
2874                         break;
2875                 }
2876                 error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2877                     &io->pfiio_size);
2878                 break;
2879         }
2880
2881         case DIOCSETIFFLAG: {
2882                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
2883
2884                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2885                 break;
2886         }
2887
2888         case DIOCCLRIFFLAG: {
2889                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
2890
2891                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2892                 break;
2893         }
2894
2895         default:
2896                 error = ENODEV;
2897                 break;
2898         }
2899 fail:
2900         lwkt_reltoken(&pf_token);
2901         return (error);
2902 }
2903
2904 /*
2905  * XXX - Check for version missmatch!!!
2906  */
2907 static void
2908 pf_clear_states(void)
2909 {
2910         struct pf_state         *s, *nexts;
2911         u_int                   killed = 0;
2912
2913         for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
2914                 nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
2915
2916                 /* don't send out individual delete messages */
2917                 s->sync_flags = PFSTATE_NOSYNC;
2918                 pf_unlink_state(s);
2919                 killed++;
2920
2921         }
2922
2923 #if 0 /* PFSYNC */
2924 /*
2925  * XXX This is called on module unload, we do not want to sync that over? */
2926  */
2927         pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
2928 #endif
2929 }
2930
2931 static int
2932 pf_clear_tables(void)
2933 {
2934         struct pfioc_table io;
2935         int error;
2936
2937         bzero(&io, sizeof(io));
2938
2939         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
2940             io.pfrio_flags);
2941
2942         return (error);
2943 }
2944
2945 static void
2946 pf_clear_srcnodes(void)
2947 {
2948         struct pf_src_node      *n;
2949         struct pf_state         *state;
2950
2951         RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2952                 state->src_node = NULL;
2953                 state->nat_src_node = NULL;
2954         }
2955         RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2956                 n->expire = 1;
2957                 n->states = 0;
2958         }
2959         pf_purge_expired_src_nodes(0);
2960         pf_status.src_nodes = 0;
2961 }
2962
2963 /*
2964  * XXX - Check for version missmatch!!!
2965  */
2966
2967 /*
2968  * Duplicate pfctl -Fa operation to get rid of as much as we can.
2969  */
2970 static int
2971 shutdown_pf(void)
2972 {
2973         int error = 0;
2974         u_int32_t t[5];
2975         char nn = '\0';
2976
2977
2978         pf_status.running = 0;
2979         error = dehook_pf();
2980         if (error) {
2981                 pf_status.running = 1;
2982                 DPFPRINTF(PF_DEBUG_MISC,
2983                     ("pf: pfil unregistration failed\n"));
2984                 return(error);
2985         }
2986         do {
2987                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
2988                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
2989                         break;
2990                 }
2991                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
2992                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
2993                         break;          /* XXX: rollback? */
2994                 }
2995                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
2996                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
2997                         break;          /* XXX: rollback? */
2998                 }
2999                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3000                     != 0) {
3001                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3002                         break;          /* XXX: rollback? */
3003                 }
3004                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3005                     != 0) {
3006                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3007                         break;          /* XXX: rollback? */
3008                 }
3009
3010                 /* XXX: these should always succeed here */
3011                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3012                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3013                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3014                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3015                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3016
3017                 if ((error = pf_clear_tables()) != 0)
3018                         break;
3019 #ifdef ALTQ
3020                 if ((error = pf_begin_altq(&t[0])) != 0) {
3021                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3022                         break;
3023                 }
3024                 pf_commit_altq(t[0]);
3025 #endif
3026                 pf_clear_states();
3027                 pf_clear_srcnodes();
3028
3029                 /* status does not use malloced mem so no need to cleanup */
3030                 /* fingerprints and interfaces have their own cleanup code */
3031         } while(0);
3032         return (error);
3033 }
3034
3035 static int
3036 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3037 {
3038         /*
3039          * DragonFly's version of pf uses FreeBSD's native host byte ordering
3040          * for ip_len/ip_off. This is why we don't have to change byte order
3041          * like the FreeBSD-5 version does.
3042          */
3043         int chk;
3044
3045         lwkt_gettoken(&pf_token);
3046
3047         chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3048         if (chk && *m) {
3049                 m_freem(*m);
3050                 *m = NULL;
3051         }
3052         lwkt_reltoken(&pf_token);
3053         return chk;
3054 }
3055
3056 static int
3057 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3058 {
3059         /*
3060          * DragonFly's version of pf uses FreeBSD's native host byte ordering
3061          * for ip_len/ip_off. This is why we don't have to change byte order
3062          * like the FreeBSD-5 version does.
3063          */
3064         int chk;
3065
3066         lwkt_gettoken(&pf_token);
3067
3068         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3069         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3070                 in_delayed_cksum(*m);
3071                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3072         }
3073         chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3074         if (chk && *m) {
3075                 m_freem(*m);
3076                 *m = NULL;
3077         }
3078         lwkt_reltoken(&pf_token);
3079         return chk;
3080 }
3081
3082 #ifdef INET6
3083 static int
3084 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3085 {
3086         /*
3087          * IPv6 is not affected by ip_len/ip_off byte order changes.
3088          */
3089         int chk;
3090
3091         lwkt_gettoken(&pf_token);
3092
3093         chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3094         if (chk && *m) {
3095                 m_freem(*m);
3096                 *m = NULL;
3097         }
3098         lwkt_reltoken(&pf_token);
3099         return chk;
3100 }
3101
3102 static int
3103 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3104 {
3105         /*
3106          * IPv6 is not affected by ip_len/ip_off byte order changes.
3107          */
3108         int chk;
3109
3110         lwkt_gettoken(&pf_token);
3111
3112         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3113         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3114                 in_delayed_cksum(*m);
3115                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3116         }
3117         chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3118         if (chk && *m) {
3119                 m_freem(*m);
3120                 *m = NULL;
3121         }
3122         lwkt_reltoken(&pf_token);
3123         return chk;
3124 }
3125 #endif /* INET6 */
3126
3127 static int
3128 hook_pf(void)
3129 {
3130         struct pfil_head *pfh_inet;
3131 #ifdef INET6
3132         struct pfil_head *pfh_inet6;
3133 #endif
3134
3135         lwkt_gettoken(&pf_token);
3136
3137         if (pf_pfil_hooked) {
3138                 lwkt_reltoken(&pf_token);
3139                 return (0);
3140         }
3141
3142         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3143         if (pfh_inet == NULL) {
3144                 lwkt_reltoken(&pf_token);
3145                 return (ENODEV);
3146         }
3147         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet);
3148         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet);
3149 #ifdef INET6
3150         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3151         if (pfh_inet6 == NULL) {
3152                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3153                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3154                 lwkt_reltoken(&pf_token);
3155                 return (ENODEV);
3156         }
3157         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet6);
3158         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet6);
3159 #endif
3160
3161         pf_pfil_hooked = 1;
3162         lwkt_reltoken(&pf_token);
3163         return (0);
3164 }
3165
3166 static int
3167 dehook_pf(void)
3168 {
3169         struct pfil_head *pfh_inet;
3170 #ifdef INET6
3171         struct pfil_head *pfh_inet6;
3172 #endif
3173
3174         lwkt_gettoken(&pf_token);
3175
3176         if (pf_pfil_hooked == 0) {
3177                 lwkt_reltoken(&pf_token);
3178                 return (0);
3179         }
3180
3181         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3182         if (pfh_inet == NULL) {
3183                 lwkt_reltoken(&pf_token);
3184                 return (ENODEV);
3185         }
3186         pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3187         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3188 #ifdef INET6
3189         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3190         if (pfh_inet6 == NULL) {
3191                 lwkt_reltoken(&pf_token);
3192                 return (ENODEV);
3193         }
3194         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3195         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3196 #endif
3197
3198         pf_pfil_hooked = 0;
3199         lwkt_reltoken(&pf_token);
3200         return (0);
3201 }
3202
3203 static int
3204 pf_load(void)
3205 {
3206         lwkt_gettoken(&pf_token);
3207
3208         pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3209         pfattach();
3210         lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3211         lwkt_reltoken(&pf_token);
3212         return (0);
3213 }
3214
3215 static int
3216 pf_mask_del(struct radix_node *rn, void *arg)
3217 {
3218         struct radix_node_head *rnh = arg;
3219
3220         rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
3221         Free(rn);
3222         return 0;
3223 }
3224
3225 static int
3226 pf_unload(void)
3227 {
3228         int error;
3229         pf_status.running = 0;
3230
3231         lwkt_gettoken(&pf_token);
3232
3233         error = dehook_pf();
3234         if (error) {
3235                 /*
3236                  * Should not happen!
3237                  * XXX Due to error code ESRCH, kldunload will show
3238                  * a message like 'No such process'.
3239                  */
3240                 kprintf("pfil unregistration fail\n");
3241                 lwkt_reltoken(&pf_token);
3242                 return error;
3243         }
3244         shutdown_pf();
3245         pf_end_threads = 1;
3246         while (pf_end_threads < 2) {
3247                 wakeup_one(pf_purge_thread);
3248                 tsleep(pf_purge_thread, 0, "pftmo", hz);
3249         }
3250         pfi_cleanup();
3251         pf_osfp_flush();
3252         dev_ops_remove_all(&pf_ops);
3253         lockuninit(&pf_consistency_lock);
3254         lwkt_reltoken(&pf_token);
3255
3256         if (pf_maskhead != NULL) {
3257                 pf_maskhead->rnh_walktree(pf_maskhead,
3258                         pf_mask_del, pf_maskhead);
3259                 Free(pf_maskhead);
3260                 pf_maskhead = NULL;
3261         }
3262         kmalloc_destroy(&pf_state_pl);
3263         kmalloc_destroy(&pf_frent_pl);
3264         kmalloc_destroy(&pf_cent_pl);
3265         return 0;
3266 }
3267
3268 static int
3269 pf_modevent(module_t mod, int type, void *data)
3270 {
3271         int error = 0;
3272
3273         lwkt_gettoken(&pf_token);
3274
3275         switch(type) {
3276         case MOD_LOAD:
3277                 error = pf_load();
3278                 break;
3279
3280         case MOD_UNLOAD:
3281                 error = pf_unload();
3282                 break;
3283         default:
3284                 error = EINVAL;
3285                 break;
3286         }
3287         lwkt_reltoken(&pf_token);
3288         return error;
3289 }
3290
3291 static moduledata_t pf_mod = {
3292         "pf",
3293         pf_modevent,
3294         0
3295 };
3296 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3297 MODULE_VERSION(pf, PF_MODVER);