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_pflog.h>
  79 #include <net/pf/if_pfsync.h>
  80
  81 #ifdef INET6
  82 #include <netinet/ip6.h>
  83 #include <netinet/in_pcb.h>
  84 #endif /* INET6 */
  85
  86 #ifdef ALTQ
  87 #include <net/altq/altq.h>
  88 #endif
  89
  90 #include <machine/limits.h>
  91 #include <net/pfil.h>
  92 #include <sys/mutex.h>
  93
  94 u_int rt_numfibs = RT_NUMFIBS;
  95
  96 void                     pfattach(void);
  97 struct pf_pool          *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
  98                             u_int8_t, u_int8_t, u_int8_t);
  99
 100 void                     pf_mv_pool(struct pf_palist *, struct pf_palist *);
 101 void                     pf_empty_pool(struct pf_palist *);
 102 #ifdef ALTQ
 103 int                      pf_begin_altq(u_int32_t *);
 104 int                      pf_rollback_altq(u_int32_t);
 105 int                      pf_commit_altq(u_int32_t);
 106 int                      pf_enable_altq(struct pf_altq *);
 107 int                      pf_disable_altq(struct pf_altq *);
 108 #endif /* ALTQ */
 109 int                      pf_begin_rules(u_int32_t *, int, const char *);
 110 int                      pf_rollback_rules(u_int32_t, int, char *);
 111 int                      pf_setup_pfsync_matching(struct pf_ruleset *);
 112 void                     pf_hash_rule(MD5_CTX *, struct pf_rule *);
 113 void                     pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 114 int                      pf_commit_rules(u_int32_t, int, char *);
 115 int                      pf_addr_setup(struct pf_ruleset *,
 116                             struct pf_addr_wrap *, sa_family_t);
 117 void                     pf_addr_copyout(struct pf_addr_wrap *);
 118
 119 struct pf_rule           pf_default_rule;
 120 struct lock              pf_consistency_lock;
 121 #ifdef ALTQ
 122 static int               pf_altq_running;
 123 #endif
 124
 125 #define TAGID_MAX        50000
 126 TAILQ_HEAD(pf_tags, pf_tagname) pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
 127                                 pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
 128
 129 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 130 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 131 #endif
 132 u_int16_t                tagname2tag(struct pf_tags *, char *);
 133 void                     tag2tagname(struct pf_tags *, u_int16_t, char *);
 134 void                     tag_unref(struct pf_tags *, u_int16_t);
 135 int                      pf_rtlabel_add(struct pf_addr_wrap *);
 136 void                     pf_rtlabel_remove(struct pf_addr_wrap *);
 137 void                     pf_rtlabel_copyout(struct pf_addr_wrap *);
 138
 139 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 140
 141 static cdev_t   pf_dev;
 142
 143 static MALLOC_DEFINE(M_PFRULEPL, "pfrulepl", "pf rule pool list");
 144 static MALLOC_DEFINE(M_PFALTQPL, "pfaltqpl", "pf altq pool list");
 145 static MALLOC_DEFINE(M_PFPOOLADDRPL, "pfpooladdrpl", "pf pool address pool list");
 146 static MALLOC_DEFINE(M_PFFRENTPL, "pffrent", "pf frent pool list");
 147
 148
 149 /*
 150  * XXX - These are new and need to be checked when moveing to a new version
 151  */
 152 static void              pf_clear_states(void);
 153 static int               pf_clear_tables(void);
 154 static void              pf_clear_srcnodes(void);
 155 /*
 156  * XXX - These are new and need to be checked when moveing to a new version
 157  */
 158
 159 /*
 160  * Wrapper functions for pfil(9) hooks
 161  */
 162 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 163                 int dir);
 164 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 165                 int dir);
 166 #ifdef INET6
 167 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 168                 int dir);
 169 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 170                 int dir);
 171 #endif
 172
 173 static int               hook_pf(void);
 174 static int               dehook_pf(void);
 175 static int               shutdown_pf(void);
 176 static int               pf_load(void);
 177 static int               pf_unload(void);
 178
 179 d_open_t        pfopen;
 180 d_close_t       pfclose;
 181 d_ioctl_t       pfioctl;
 182
 183 static struct dev_ops pf_ops = {            /* XXX convert to port model */
 184         { PF_NAME, 73, 0 },
 185         .d_open =       pfopen,
 186         .d_close =      pfclose,
 187         .d_ioctl =      pfioctl
 188 };
 189
 190 static volatile int pf_pfil_hooked = 0;
 191 int pf_end_threads = 0;
 192
 193 int debug_pfugidhack = 0;
 194 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
 195         "Enable/disable pf user/group rules mpsafe hack");
 196
 197 void
 198 pfattach(void)
 199 {
 200         u_int32_t *my_timeout = pf_default_rule.timeout;
 201
 202
 203         if (!rn_inithead((void **)&pf_maskhead, NULL, 0)) {
 204                 kprintf("pf mask radix tree create failed\n");
 205                 return;
 206         }
 207         kmalloc_create(&pf_state_pl, "pf state pool list");
 208         kmalloc_raise_limit(pf_state_pl, 0);
 209         kmalloc_create(&pf_frent_pl, "pf fragment pool list");
 210         kmalloc_raise_limit(pf_frent_pl, 0);
 211         kmalloc_create(&pf_cent_pl, "pf cent pool list");
 212         kmalloc_raise_limit(pf_cent_pl, 0);
 213
 214         pfr_initialize();
 215         pfi_initialize();
 216         pf_osfp_initialize();
 217
 218         pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
 219         pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 220         pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
 221         pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
 222         if (ctob(physmem) <= 100*1024*1024)
 223                 pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
 224                     PFR_KENTRY_HIWAT_SMALL;
 225
 226         RB_INIT(&tree_src_tracking);
 227         RB_INIT(&pf_anchors);
 228         pf_init_ruleset(&pf_main_ruleset);
 229         TAILQ_INIT(&pf_altqs[0]);
 230         TAILQ_INIT(&pf_altqs[1]);
 231         TAILQ_INIT(&pf_pabuf);
 232         pf_altqs_active = &pf_altqs[0];
 233         pf_altqs_inactive = &pf_altqs[1];
 234         TAILQ_INIT(&state_list);
 235
 236         /* default rule should never be garbage collected */
 237         pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
 238         pf_default_rule.action = PF_PASS;
 239         pf_default_rule.nr = (uint32_t)(-1);
 240         pf_default_rule.rtableid = -1;
 241
 242         /* initialize default timeouts */
 243         my_timeout[PFTM_TCP_FIRST_PACKET] = 120;        /* First TCP packet */
 244         my_timeout[PFTM_TCP_OPENING] = 30;              /* No response yet */
 245         my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;    /* Established */
 246         my_timeout[PFTM_TCP_CLOSING] = 15 * 60;         /* Half closed */
 247         my_timeout[PFTM_TCP_FIN_WAIT] = 45;             /* Got both FINs */
 248         my_timeout[PFTM_TCP_CLOSED] = 90;               /* Got a RST */
 249         my_timeout[PFTM_UDP_FIRST_PACKET] = 60;         /* First UDP packet */
 250         my_timeout[PFTM_UDP_SINGLE] = 30;               /* Unidirectional */
 251         my_timeout[PFTM_UDP_MULTIPLE] = 60;             /* Bidirectional */
 252         my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;        /* First ICMP packet */
 253         my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;         /* Got error response */
 254         my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;       /* First packet */
 255         my_timeout[PFTM_OTHER_SINGLE] = 30;             /* Unidirectional */
 256         my_timeout[PFTM_OTHER_MULTIPLE] = 60;           /* Bidirectional */
 257         my_timeout[PFTM_FRAG] = 30;                     /* Fragment expire */
 258         my_timeout[PFTM_INTERVAL] = 10;                 /* Expire interval */
 259         my_timeout[PFTM_SRC_NODE] = 0;          /* Source Tracking */
 260         my_timeout[PFTM_TS_DIFF] = 30;          /* Allowed TS diff */
 261         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 262         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 263
 264         pf_normalize_init();
 265         bzero(&pf_status, sizeof(pf_status));
 266         pf_status.debug = PF_DEBUG_URGENT;
 267         /* XXX do our best to avoid a conflict */
 268         pf_status.hostid = karc4random();
 269
 270         if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
 271                 panic("pfpurge thread");
 272 }
 273
 274 int
 275 pfopen(struct dev_open_args *ap)
 276 {
 277         lwkt_gettoken(&pf_token);
 278         cdev_t dev = ap->a_head.a_dev;
 279         if (minor(dev) >= 1) {
 280                 lwkt_reltoken(&pf_token);
 281                 return (ENXIO);
 282         }
 283         lwkt_reltoken(&pf_token);
 284         return (0);
 285 }
 286
 287 int
 288 pfclose(struct dev_close_args *ap)
 289 {
 290         lwkt_gettoken(&pf_token);
 291         cdev_t dev = ap->a_head.a_dev;
 292         if (minor(dev) >= 1) {
 293                 lwkt_reltoken(&pf_token);
 294                 return (ENXIO);
 295         }
 296         lwkt_reltoken(&pf_token);
 297         return (0);
 298 }
 299
 300 struct pf_pool *
 301 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 302     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 303     u_int8_t check_ticket)
 304 {
 305         struct pf_ruleset       *ruleset;
 306         struct pf_rule          *rule;
 307         int                      rs_num;
 308
 309         ruleset = pf_find_ruleset(anchor);
 310         if (ruleset == NULL)
 311                 return (NULL);
 312         rs_num = pf_get_ruleset_number(rule_action);
 313         if (rs_num >= PF_RULESET_MAX)
 314                 return (NULL);
 315         if (active) {
 316                 if (check_ticket && ticket !=
 317                     ruleset->rules[rs_num].active.ticket)
 318                         return (NULL);
 319                 if (r_last)
 320                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 321                             pf_rulequeue);
 322                 else
 323                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 324         } else {
 325                 if (check_ticket && ticket !=
 326                     ruleset->rules[rs_num].inactive.ticket)
 327                         return (NULL);
 328                 if (r_last)
 329                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 330                             pf_rulequeue);
 331                 else
 332                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 333         }
 334         if (!r_last) {
 335                 while ((rule != NULL) && (rule->nr != rule_number))
 336                         rule = TAILQ_NEXT(rule, entries);
 337         }
 338         if (rule == NULL)
 339                 return (NULL);
 340
 341         return (&rule->rpool);
 342 }
 343
 344 void
 345 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 346 {
 347         struct pf_pooladdr      *mv_pool_pa;
 348
 349         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 350                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 351                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 352         }
 353 }
 354
 355 void
 356 pf_empty_pool(struct pf_palist *poola)
 357 {
 358         struct pf_pooladdr      *empty_pool_pa;
 359
 360         while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 361                 pfi_dynaddr_remove(&empty_pool_pa->addr);
 362                 pf_tbladdr_remove(&empty_pool_pa->addr);
 363                 pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
 364                 TAILQ_REMOVE(poola, empty_pool_pa, entries);
 365                 kfree(empty_pool_pa, M_PFPOOLADDRPL);
 366         }
 367 }
 368
 369 void
 370 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 371 {
 372         if (rulequeue != NULL) {
 373                 if (rule->states_cur <= 0) {
 374                         /*
 375                          * XXX - we need to remove the table *before* detaching
 376                          * the rule to make sure the table code does not delete
 377                          * the anchor under our feet.
 378                          */
 379                         pf_tbladdr_remove(&rule->src.addr);
 380                         pf_tbladdr_remove(&rule->dst.addr);
 381                         if (rule->overload_tbl)
 382                                 pfr_detach_table(rule->overload_tbl);
 383                 }
 384                 TAILQ_REMOVE(rulequeue, rule, entries);
 385                 rule->entries.tqe_prev = NULL;
 386                 rule->nr = -1;
 387         }
 388
 389         if (rule->states_cur > 0 || rule->src_nodes > 0 ||
 390             rule->entries.tqe_prev != NULL)
 391                 return;
 392         pf_tag_unref(rule->tag);
 393         pf_tag_unref(rule->match_tag);
 394 #ifdef ALTQ
 395         if (rule->pqid != rule->qid)
 396                 pf_qid_unref(rule->pqid);
 397         pf_qid_unref(rule->qid);
 398 #endif
 399         pf_rtlabel_remove(&rule->src.addr);
 400         pf_rtlabel_remove(&rule->dst.addr);
 401         pfi_dynaddr_remove(&rule->src.addr);
 402         pfi_dynaddr_remove(&rule->dst.addr);
 403         if (rulequeue == NULL) {
 404                 pf_tbladdr_remove(&rule->src.addr);
 405                 pf_tbladdr_remove(&rule->dst.addr);
 406                 if (rule->overload_tbl)
 407                         pfr_detach_table(rule->overload_tbl);
 408         }
 409         pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
 410         pf_anchor_remove(rule);
 411         pf_empty_pool(&rule->rpool.list);
 412         kfree(rule, M_PFRULEPL);
 413 }
 414
 415 u_int16_t
 416 tagname2tag(struct pf_tags *head, char *tagname)
 417 {
 418         struct pf_tagname       *tag, *p = NULL;
 419         u_int16_t                new_tagid = 1;
 420
 421         TAILQ_FOREACH(tag, head, entries)
 422                 if (strcmp(tagname, tag->name) == 0) {
 423                         tag->ref++;
 424                         return (tag->tag);
 425                 }
 426
 427         /*
 428          * to avoid fragmentation, we do a linear search from the beginning
 429          * and take the first free slot we find. if there is none or the list
 430          * is empty, append a new entry at the end.
 431          */
 432
 433         /* new entry */
 434         if (!TAILQ_EMPTY(head))
 435                 for (p = TAILQ_FIRST(head); p != NULL &&
 436                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 437                         new_tagid = p->tag + 1;
 438
 439         if (new_tagid > TAGID_MAX)
 440                 return (0);
 441
 442         /* allocate and fill new struct pf_tagname */
 443         tag = kmalloc(sizeof(*tag), M_TEMP, M_WAITOK);
 444         strlcpy(tag->name, tagname, sizeof(tag->name));
 445         tag->tag = new_tagid;
 446         tag->ref++;
 447
 448         if (p != NULL)  /* insert new entry before p */
 449                 TAILQ_INSERT_BEFORE(p, tag, entries);
 450         else    /* either list empty or no free slot in between */
 451                 TAILQ_INSERT_TAIL(head, tag, entries);
 452
 453         return (tag->tag);
 454 }
 455
 456 void
 457 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
 458 {
 459         struct pf_tagname       *tag;
 460
 461         TAILQ_FOREACH(tag, head, entries)
 462                 if (tag->tag == tagid) {
 463                         strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
 464                         return;
 465                 }
 466 }
 467
 468 void
 469 tag_unref(struct pf_tags *head, u_int16_t tag)
 470 {
 471         struct pf_tagname       *p, *next;
 472
 473         if (tag == 0)
 474                 return;
 475
 476         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 477                 next = TAILQ_NEXT(p, entries);
 478                 if (tag == p->tag) {
 479                         if (--p->ref == 0) {
 480                                 TAILQ_REMOVE(head, p, entries);
 481                                 kfree(p, M_TEMP);
 482                         }
 483                         break;
 484                 }
 485         }
 486 }
 487
 488 u_int16_t
 489 pf_tagname2tag(char *tagname)
 490 {
 491         return (tagname2tag(&pf_tags, tagname));
 492 }
 493
 494 void
 495 pf_tag2tagname(u_int16_t tagid, char *p)
 496 {
 497         tag2tagname(&pf_tags, tagid, p);
 498 }
 499
 500 void
 501 pf_tag_ref(u_int16_t tag)
 502 {
 503         struct pf_tagname *t;
 504
 505         TAILQ_FOREACH(t, &pf_tags, entries)
 506                 if (t->tag == tag)
 507                         break;
 508         if (t != NULL)
 509                 t->ref++;
 510 }
 511
 512 void
 513 pf_tag_unref(u_int16_t tag)
 514 {
 515         tag_unref(&pf_tags, tag);
 516 }
 517
 518 int
 519 pf_rtlabel_add(struct pf_addr_wrap *a)
 520 {
 521         return (0);
 522 }
 523
 524 void
 525 pf_rtlabel_remove(struct pf_addr_wrap *a)
 526 {
 527 }
 528
 529 void
 530 pf_rtlabel_copyout(struct pf_addr_wrap *a)
 531 {
 532         if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
 533                 strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
 534 }
 535
 536 #ifdef ALTQ
 537 u_int32_t
 538 pf_qname2qid(char *qname)
 539 {
 540         return ((u_int32_t)tagname2tag(&pf_qids, qname));
 541 }
 542
 543 void
 544 pf_qid2qname(u_int32_t qid, char *p)
 545 {
 546         tag2tagname(&pf_qids, (u_int16_t)qid, p);
 547 }
 548
 549 void
 550 pf_qid_unref(u_int32_t qid)
 551 {
 552         tag_unref(&pf_qids, (u_int16_t)qid);
 553 }
 554
 555 int
 556 pf_begin_altq(u_int32_t *ticket)
 557 {
 558         struct pf_altq  *altq;
 559         int              error = 0;
 560
 561         /* Purge the old altq list */
 562         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 563                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 564                 if (altq->qname[0] == 0) {
 565                         /* detach and destroy the discipline */
 566                         error = altq_remove(altq);
 567                 } else
 568                         pf_qid_unref(altq->qid);
 569                 kfree(altq, M_PFALTQPL);
 570         }
 571         if (error)
 572                 return (error);
 573         *ticket = ++ticket_altqs_inactive;
 574         altqs_inactive_open = 1;
 575         return (0);
 576 }
 577
 578 int
 579 pf_rollback_altq(u_int32_t ticket)
 580 {
 581         struct pf_altq  *altq;
 582         int              error = 0;
 583
 584         if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
 585                 return (0);
 586         /* Purge the old altq list */
 587         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 588                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 589                 if (altq->qname[0] == 0) {
 590                         /* detach and destroy the discipline */
 591                         error = altq_remove(altq);
 592                 } else
 593                         pf_qid_unref(altq->qid);
 594                 kfree(altq, M_PFALTQPL);
 595         }
 596         altqs_inactive_open = 0;
 597         return (error);
 598 }
 599
 600 int
 601 pf_commit_altq(u_int32_t ticket)
 602 {
 603         struct pf_altqqueue     *old_altqs;
 604         struct pf_altq          *altq;
 605         int                      err, error = 0;
 606
 607         if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
 608                 return (EBUSY);
 609
 610         /* swap altqs, keep the old. */
 611         crit_enter();
 612         old_altqs = pf_altqs_active;
 613         pf_altqs_active = pf_altqs_inactive;
 614         pf_altqs_inactive = old_altqs;
 615         ticket_altqs_active = ticket_altqs_inactive;
 616
 617         /* Attach new disciplines */
 618         TAILQ_FOREACH(altq, pf_altqs_active, entries) {
 619                 if (altq->qname[0] == 0) {
 620                         /* attach the discipline */
 621                         error = altq_pfattach(altq);
 622                         if (error) {
 623                                 crit_exit();
 624                                 return (error);
 625                         }
 626                 }
 627         }
 628
 629         /* Purge the old altq list */
 630         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 631                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 632                 if (altq->qname[0] == 0) {
 633                         /* detach and destroy the discipline */
 634                         if (pf_altq_running)
 635                                 error = pf_disable_altq(altq);
 636                         err = altq_pfdetach(altq);
 637                         if (err != 0 && error == 0)
 638                                 error = err;
 639                         err = altq_remove(altq);
 640                         if (err != 0 && error == 0)
 641                                 error = err;
 642                 } else
 643                         pf_qid_unref(altq->qid);
 644                 kfree(altq, M_PFALTQPL);
 645         }
 646         crit_exit();
 647
 648         altqs_inactive_open = 0;
 649         return (error);
 650 }
 651
 652 int
 653 pf_enable_altq(struct pf_altq *altq)
 654 {
 655         struct ifnet            *ifp;
 656         struct tb_profile        tb;
 657         int                      error = 0;
 658
 659         if ((ifp = ifunit(altq->ifname)) == NULL)
 660                 return (EINVAL);
 661
 662         if (ifp->if_snd.altq_type != ALTQT_NONE)
 663                 error = altq_enable(&ifp->if_snd);
 664
 665         /* set tokenbucket regulator */
 666         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 667                 tb.rate = altq->ifbandwidth;
 668                 tb.depth = altq->tbrsize;
 669                 crit_enter();
 670                 error = tbr_set(&ifp->if_snd, &tb);
 671                 crit_exit();
 672         }
 673
 674         return (error);
 675 }
 676
 677 int
 678 pf_disable_altq(struct pf_altq *altq)
 679 {
 680         struct ifnet            *ifp;
 681         struct tb_profile        tb;
 682         int                      error;
 683
 684         if ((ifp = ifunit(altq->ifname)) == NULL)
 685                 return (EINVAL);
 686
 687         /*
 688          * when the discipline is no longer referenced, it was overridden
 689          * by a new one.  if so, just return.
 690          */
 691         if (altq->altq_disc != ifp->if_snd.altq_disc)
 692                 return (0);
 693
 694         error = altq_disable(&ifp->if_snd);
 695
 696         if (error == 0) {
 697                 /* clear tokenbucket regulator */
 698                 tb.rate = 0;
 699                 crit_enter();
 700                 error = tbr_set(&ifp->if_snd, &tb);
 701                 crit_exit();
 702         }
 703
 704         return (error);
 705 }
 706 #endif /* ALTQ */
 707
 708 int
 709 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 710 {
 711         struct pf_ruleset       *rs;
 712         struct pf_rule          *rule;
 713
 714         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 715                 return (EINVAL);
 716         rs = pf_find_or_create_ruleset(anchor);
 717         if (rs == NULL)
 718                 return (EINVAL);
 719         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 720                 pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
 721                 rs->rules[rs_num].inactive.rcount--;
 722         }
 723         *ticket = ++rs->rules[rs_num].inactive.ticket;
 724         rs->rules[rs_num].inactive.open = 1;
 725         return (0);
 726 }
 727
 728 int
 729 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 730 {
 731         struct pf_ruleset       *rs;
 732         struct pf_rule          *rule;
 733
 734         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 735                 return (EINVAL);
 736         rs = pf_find_ruleset(anchor);
 737         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 738             rs->rules[rs_num].inactive.ticket != ticket)
 739                 return (0);
 740         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 741                 pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
 742                 rs->rules[rs_num].inactive.rcount--;
 743         }
 744         rs->rules[rs_num].inactive.open = 0;
 745         return (0);
 746 }
 747
 748 #define PF_MD5_UPD(st, elm)                                             \
 749                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 750
 751 #define PF_MD5_UPD_STR(st, elm)                                         \
 752                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 753
 754 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 755                 (stor) = htonl((st)->elm);                              \
 756                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 757 } while (0)
 758
 759 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 760                 (stor) = htons((st)->elm);                              \
 761                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 762 } while (0)
 763
 764 void
 765 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 766 {
 767         PF_MD5_UPD(pfr, addr.type);
 768         switch (pfr->addr.type) {
 769                 case PF_ADDR_DYNIFTL:
 770                         PF_MD5_UPD(pfr, addr.v.ifname);
 771                         PF_MD5_UPD(pfr, addr.iflags);
 772                         break;
 773                 case PF_ADDR_TABLE:
 774                         PF_MD5_UPD(pfr, addr.v.tblname);
 775                         break;
 776                 case PF_ADDR_ADDRMASK:
 777                         /* XXX ignore af? */
 778                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 779                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 780                         break;
 781                 case PF_ADDR_RTLABEL:
 782                         PF_MD5_UPD(pfr, addr.v.rtlabelname);
 783                         break;
 784         }
 785
 786         PF_MD5_UPD(pfr, port[0]);
 787         PF_MD5_UPD(pfr, port[1]);
 788         PF_MD5_UPD(pfr, neg);
 789         PF_MD5_UPD(pfr, port_op);
 790 }
 791
 792 void
 793 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 794 {
 795         u_int16_t x;
 796         u_int32_t y;
 797
 798         pf_hash_rule_addr(ctx, &rule->src);
 799         pf_hash_rule_addr(ctx, &rule->dst);
 800         PF_MD5_UPD_STR(rule, label);
 801         PF_MD5_UPD_STR(rule, ifname);
 802         PF_MD5_UPD_STR(rule, match_tagname);
 803         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 804         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 805         PF_MD5_UPD_HTONL(rule, prob, y);
 806         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 807         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 808         PF_MD5_UPD(rule, uid.op);
 809         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 810         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 811         PF_MD5_UPD(rule, gid.op);
 812         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 813         PF_MD5_UPD(rule, action);
 814         PF_MD5_UPD(rule, direction);
 815         PF_MD5_UPD(rule, af);
 816         PF_MD5_UPD(rule, quick);
 817         PF_MD5_UPD(rule, ifnot);
 818         PF_MD5_UPD(rule, match_tag_not);
 819         PF_MD5_UPD(rule, natpass);
 820         PF_MD5_UPD(rule, keep_state);
 821         PF_MD5_UPD(rule, proto);
 822         PF_MD5_UPD(rule, type);
 823         PF_MD5_UPD(rule, code);
 824         PF_MD5_UPD(rule, flags);
 825         PF_MD5_UPD(rule, flagset);
 826         PF_MD5_UPD(rule, allow_opts);
 827         PF_MD5_UPD(rule, rt);
 828         PF_MD5_UPD(rule, tos);
 829 }
 830
 831 int
 832 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 833 {
 834         struct pf_ruleset       *rs;
 835         struct pf_rule          *rule, **old_array;
 836         struct pf_rulequeue     *old_rules;
 837         int                      error;
 838         u_int32_t                old_rcount;
 839
 840         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 841                 return (EINVAL);
 842         rs = pf_find_ruleset(anchor);
 843         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 844             ticket != rs->rules[rs_num].inactive.ticket)
 845                 return (EBUSY);
 846
 847         /* Calculate checksum for the main ruleset */
 848         if (rs == &pf_main_ruleset) {
 849                 error = pf_setup_pfsync_matching(rs);
 850                 if (error != 0)
 851                         return (error);
 852         }
 853
 854         /* Swap rules, keep the old. */
 855         crit_enter();
 856         old_rules = rs->rules[rs_num].active.ptr;
 857         old_rcount = rs->rules[rs_num].active.rcount;
 858         old_array = rs->rules[rs_num].active.ptr_array;
 859
 860         rs->rules[rs_num].active.ptr =
 861             rs->rules[rs_num].inactive.ptr;
 862         rs->rules[rs_num].active.ptr_array =
 863             rs->rules[rs_num].inactive.ptr_array;
 864         rs->rules[rs_num].active.rcount =
 865             rs->rules[rs_num].inactive.rcount;
 866         rs->rules[rs_num].inactive.ptr = old_rules;
 867         rs->rules[rs_num].inactive.ptr_array = old_array;
 868         rs->rules[rs_num].inactive.rcount = old_rcount;
 869
 870         rs->rules[rs_num].active.ticket =
 871             rs->rules[rs_num].inactive.ticket;
 872         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 873
 874
 875         /* Purge the old rule list. */
 876         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 877                 pf_rm_rule(old_rules, rule);
 878         if (rs->rules[rs_num].inactive.ptr_array)
 879                 kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 880         rs->rules[rs_num].inactive.ptr_array = NULL;
 881         rs->rules[rs_num].inactive.rcount = 0;
 882         rs->rules[rs_num].inactive.open = 0;
 883         pf_remove_if_empty_ruleset(rs);
 884         crit_exit();
 885         return (0);
 886 }
 887
 888 int
 889 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 890 {
 891         MD5_CTX                  ctx;
 892         struct pf_rule          *rule;
 893         int                      rs_cnt;
 894         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 895
 896         MD5Init(&ctx);
 897         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 898                 /* XXX PF_RULESET_SCRUB as well? */
 899                 if (rs_cnt == PF_RULESET_SCRUB)
 900                         continue;
 901
 902                 if (rs->rules[rs_cnt].inactive.ptr_array)
 903                         kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 904                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 905
 906                 if (rs->rules[rs_cnt].inactive.rcount) {
 907                         rs->rules[rs_cnt].inactive.ptr_array =
 908                             kmalloc(sizeof(caddr_t) *
 909                                     rs->rules[rs_cnt].inactive.rcount,
 910                                     M_TEMP, M_WAITOK);
 911                 }
 912
 913                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 914                     entries) {
 915                         pf_hash_rule(&ctx, rule);
 916                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 917                 }
 918         }
 919
 920         MD5Final(digest, &ctx);
 921         memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
 922         return (0);
 923 }
 924
 925 int
 926 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 927     sa_family_t af)
 928 {
 929         if (pfi_dynaddr_setup(addr, af) ||
 930             pf_tbladdr_setup(ruleset, addr))
 931                 return (EINVAL);
 932
 933         return (0);
 934 }
 935
 936 void
 937 pf_addr_copyout(struct pf_addr_wrap *addr)
 938 {
 939         pfi_dynaddr_copyout(addr);
 940         pf_tbladdr_copyout(addr);
 941         pf_rtlabel_copyout(addr);
 942 }
 943
 944 int
 945 pfioctl(struct dev_ioctl_args *ap)
 946 {
 947         u_long cmd = ap->a_cmd;
 948         caddr_t addr = ap->a_data;
 949         struct pf_pooladdr      *pa = NULL;
 950         struct pf_pool          *pool = NULL;
 951         int                      error = 0;
 952
 953         lwkt_gettoken(&pf_token);
 954
 955         /* XXX keep in sync with switch() below */
 956         if (securelevel > 1)
 957                 switch (cmd) {
 958                 case DIOCGETRULES:
 959                 case DIOCGETRULE:
 960                 case DIOCGETADDRS:
 961                 case DIOCGETADDR:
 962                 case DIOCGETSTATE:
 963                 case DIOCSETSTATUSIF:
 964                 case DIOCGETSTATUS:
 965                 case DIOCCLRSTATUS:
 966                 case DIOCNATLOOK:
 967                 case DIOCSETDEBUG:
 968                 case DIOCGETSTATES:
 969                 case DIOCGETTIMEOUT:
 970                 case DIOCCLRRULECTRS:
 971                 case DIOCGETLIMIT:
 972                 case DIOCGETALTQS:
 973                 case DIOCGETALTQ:
 974                 case DIOCGETQSTATS:
 975                 case DIOCGETRULESETS:
 976                 case DIOCGETRULESET:
 977                 case DIOCRGETTABLES:
 978                 case DIOCRGETTSTATS:
 979                 case DIOCRCLRTSTATS:
 980                 case DIOCRCLRADDRS:
 981                 case DIOCRADDADDRS:
 982                 case DIOCRDELADDRS:
 983                 case DIOCRSETADDRS:
 984                 case DIOCRGETADDRS:
 985                 case DIOCRGETASTATS:
 986                 case DIOCRCLRASTATS:
 987                 case DIOCRTSTADDRS:
 988                 case DIOCOSFPGET:
 989                 case DIOCGETSRCNODES:
 990                 case DIOCCLRSRCNODES:
 991                 case DIOCIGETIFACES:
 992                 case DIOCSETIFFLAG:
 993                 case DIOCCLRIFFLAG:
 994                 case DIOCGIFSPEED:
 995                         break;
 996                 case DIOCRCLRTABLES:
 997                 case DIOCRADDTABLES:
 998                 case DIOCRDELTABLES:
 999                 case DIOCRSETTFLAGS:
1000                         if (((struct pfioc_table *)addr)->pfrio_flags &
1001                             PFR_FLAG_DUMMY)
1002                                 break; /* dummy operation ok */
1003                         lwkt_reltoken(&pf_token);
1004                         return (EPERM);
1005                 default:
1006                         lwkt_reltoken(&pf_token);
1007                         return (EPERM);
1008                 }
1009
1010         if (!(ap->a_fflag & FWRITE))
1011                 switch (cmd) {
1012                 case DIOCGETRULES:
1013                 case DIOCGETADDRS:
1014                 case DIOCGETADDR:
1015                 case DIOCGETSTATE:
1016                 case DIOCGETSTATUS:
1017                 case DIOCGETSTATES:
1018                 case DIOCGETTIMEOUT:
1019                 case DIOCGETLIMIT:
1020                 case DIOCGETALTQS:
1021                 case DIOCGETALTQ:
1022                 case DIOCGETQSTATS:
1023                 case DIOCGETRULESETS:
1024                 case DIOCGETRULESET:
1025                 case DIOCNATLOOK:
1026                 case DIOCRGETTABLES:
1027                 case DIOCRGETTSTATS:
1028                 case DIOCRGETADDRS:
1029                 case DIOCRGETASTATS:
1030                 case DIOCRTSTADDRS:
1031                 case DIOCOSFPGET:
1032                 case DIOCGETSRCNODES:
1033                 case DIOCIGETIFACES:
1034                 case DIOCGIFSPEED:
1035                         break;
1036                 case DIOCRCLRTABLES:
1037                 case DIOCRADDTABLES:
1038                 case DIOCRDELTABLES:
1039                 case DIOCRCLRTSTATS:
1040                 case DIOCRCLRADDRS:
1041                 case DIOCRADDADDRS:
1042                 case DIOCRDELADDRS:
1043                 case DIOCRSETADDRS:
1044                 case DIOCRSETTFLAGS:
1045                         if (((struct pfioc_table *)addr)->pfrio_flags &
1046                             PFR_FLAG_DUMMY)
1047                                 break; /* dummy operation ok */
1048                         lwkt_reltoken(&pf_token);
1049                         return (EACCES);
1050                 case DIOCGETRULE:
1051                         if (((struct pfioc_rule *)addr)->action ==
1052                             PF_GET_CLR_CNTR) {
1053                                 lwkt_reltoken(&pf_token);
1054                                 return (EACCES);
1055                         }
1056                         break;
1057                 default:
1058                         lwkt_reltoken(&pf_token);
1059                         return (EACCES);
1060                 }
1061
1062         switch (cmd) {
1063
1064         case DIOCSTART:
1065                 if (pf_status.running)
1066                         error = EEXIST;
1067                 else {
1068                         error = hook_pf();
1069                         if (error) {
1070                                 DPFPRINTF(PF_DEBUG_MISC,
1071                                     ("pf: pfil registration fail\n"));
1072                                 break;
1073                         }
1074                         pf_status.running = 1;
1075                         pf_status.since = time_second;
1076                         if (pf_status.stateid == 0) {
1077                                 pf_status.stateid = time_second;
1078                                 pf_status.stateid = pf_status.stateid << 32;
1079                         }
1080                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1081                 }
1082                 break;
1083
1084         case DIOCSTOP:
1085                 if (!pf_status.running)
1086                         error = ENOENT;
1087                 else {
1088                         pf_status.running = 0;
1089                         error = dehook_pf();
1090                         if (error) {
1091                                 pf_status.running = 1;
1092                                 DPFPRINTF(PF_DEBUG_MISC,
1093                                         ("pf: pfil unregistration failed\n"));
1094                         }
1095                         pf_status.since = time_second;
1096                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1097                 }
1098                 break;
1099
1100         case DIOCADDRULE: {
1101                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1102                 struct pf_ruleset       *ruleset;
1103                 struct pf_rule          *rule, *tail;
1104                 struct pf_pooladdr      *pa;
1105                 int                      rs_num;
1106
1107                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1108                 ruleset = pf_find_ruleset(pr->anchor);
1109                 if (ruleset == NULL) {
1110                         error = EINVAL;
1111                         break;
1112                 }
1113                 rs_num = pf_get_ruleset_number(pr->rule.action);
1114                 if (rs_num >= PF_RULESET_MAX) {
1115                         error = EINVAL;
1116                         break;
1117                 }
1118                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1119                         error = EINVAL;
1120                         break;
1121                 }
1122                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1123                         error = EBUSY;
1124                         break;
1125                 }
1126                 if (pr->pool_ticket != ticket_pabuf) {
1127                         error = EBUSY;
1128                         break;
1129                 }
1130                 rule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK);
1131                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1132                 rule->cuid = ap->a_cred->cr_ruid;
1133                 rule->cpid = 0;
1134                 rule->anchor = NULL;
1135                 rule->kif = NULL;
1136                 TAILQ_INIT(&rule->rpool.list);
1137                 /* initialize refcounting */
1138                 rule->states_cur = 0;
1139                 rule->src_nodes = 0;
1140                 rule->entries.tqe_prev = NULL;
1141 #ifndef INET
1142                 if (rule->af == AF_INET) {
1143                         kfree(rule, M_PFRULEPL);
1144                         error = EAFNOSUPPORT;
1145                         break;
1146                 }
1147 #endif /* INET */
1148 #ifndef INET6
1149                 if (rule->af == AF_INET6) {
1150                         kfree(rule, M_PFRULEPL);
1151                         error = EAFNOSUPPORT;
1152                         break;
1153                 }
1154 #endif /* INET6 */
1155                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1156                     pf_rulequeue);
1157                 if (tail)
1158                         rule->nr = tail->nr + 1;
1159                 else
1160                         rule->nr = 0;
1161                 if (rule->ifname[0]) {
1162                         rule->kif = pfi_kif_get(rule->ifname);
1163                         if (rule->kif == NULL) {
1164                                 kfree(rule, M_PFRULEPL);
1165                                 error = EINVAL;
1166                                 break;
1167                         }
1168                         pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1169                 }
1170
1171                 if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1172                         error = EBUSY;
1173
1174 #ifdef ALTQ
1175                 /* set queue IDs */
1176                 if (rule->qname[0] != 0) {
1177                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1178                                 error = EBUSY;
1179                         else if (rule->pqname[0] != 0) {
1180                                 if ((rule->pqid =
1181                                     pf_qname2qid(rule->pqname)) == 0)
1182                                         error = EBUSY;
1183                         } else
1184                                 rule->pqid = rule->qid;
1185                 }
1186 #endif
1187                 if (rule->tagname[0])
1188                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1189                                 error = EBUSY;
1190                 if (rule->match_tagname[0])
1191                         if ((rule->match_tag =
1192                             pf_tagname2tag(rule->match_tagname)) == 0)
1193                                 error = EBUSY;
1194                 if (rule->rt && !rule->direction)
1195                         error = EINVAL;
1196 #if NPFLOG > 0
1197                 if (!rule->log)
1198                         rule->logif = 0;
1199                 if (rule->logif >= PFLOGIFS_MAX)
1200                         error = EINVAL;
1201 #endif
1202                 if (pf_rtlabel_add(&rule->src.addr) ||
1203                     pf_rtlabel_add(&rule->dst.addr))
1204                         error = EBUSY;
1205                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1206                         error = EINVAL;
1207                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1208                         error = EINVAL;
1209                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1210                         error = EINVAL;
1211                 TAILQ_FOREACH(pa, &pf_pabuf, entries)
1212                         if (pf_tbladdr_setup(ruleset, &pa->addr))
1213                                 error = EINVAL;
1214
1215                 if (rule->overload_tblname[0]) {
1216                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1217                             rule->overload_tblname)) == NULL)
1218                                 error = EINVAL;
1219                         else
1220                                 rule->overload_tbl->pfrkt_flags |=
1221                                     PFR_TFLAG_ACTIVE;
1222                 }
1223
1224                 pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1225                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1226                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1227                     (rule->rt > PF_FASTROUTE)) &&
1228                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1229                         error = EINVAL;
1230
1231                 if (error) {
1232                         pf_rm_rule(NULL, rule);
1233                         break;
1234                 }
1235                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1236                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1237                     rule->bytes[0] = rule->bytes[1] = 0;
1238                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1239                     rule, entries);
1240                 ruleset->rules[rs_num].inactive.rcount++;
1241                 break;
1242         }
1243
1244         case DIOCGETRULES: {
1245                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1246                 struct pf_ruleset       *ruleset;
1247                 struct pf_rule          *tail;
1248                 int                      rs_num;
1249
1250                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1251                 ruleset = pf_find_ruleset(pr->anchor);
1252                 if (ruleset == NULL) {
1253                         error = EINVAL;
1254                         break;
1255                 }
1256                 rs_num = pf_get_ruleset_number(pr->rule.action);
1257                 if (rs_num >= PF_RULESET_MAX) {
1258                         error = EINVAL;
1259                         break;
1260                 }
1261                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1262                     pf_rulequeue);
1263                 if (tail)
1264                         pr->nr = tail->nr + 1;
1265                 else
1266                         pr->nr = 0;
1267                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1268                 break;
1269         }
1270
1271         case DIOCGETRULE: {
1272                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1273                 struct pf_ruleset       *ruleset;
1274                 struct pf_rule          *rule;
1275                 int                      rs_num, i;
1276
1277                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1278                 ruleset = pf_find_ruleset(pr->anchor);
1279                 if (ruleset == NULL) {
1280                         error = EINVAL;
1281                         break;
1282                 }
1283                 rs_num = pf_get_ruleset_number(pr->rule.action);
1284                 if (rs_num >= PF_RULESET_MAX) {
1285                         error = EINVAL;
1286                         break;
1287                 }
1288                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1289                         error = EBUSY;
1290                         break;
1291                 }
1292                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1293                 while ((rule != NULL) && (rule->nr != pr->nr))
1294                         rule = TAILQ_NEXT(rule, entries);
1295                 if (rule == NULL) {
1296                         error = EBUSY;
1297                         break;
1298                 }
1299                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1300                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1301                         error = EBUSY;
1302                         break;
1303                 }
1304                 pf_addr_copyout(&pr->rule.src.addr);
1305                 pf_addr_copyout(&pr->rule.dst.addr);
1306                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1307                         if (rule->skip[i].ptr == NULL)
1308                                 pr->rule.skip[i].nr = (uint32_t)(-1);
1309                         else
1310                                 pr->rule.skip[i].nr =
1311                                     rule->skip[i].ptr->nr;
1312
1313                 if (pr->action == PF_GET_CLR_CNTR) {
1314                         rule->evaluations = 0;
1315                         rule->packets[0] = rule->packets[1] = 0;
1316                         rule->bytes[0] = rule->bytes[1] = 0;
1317                         rule->states_tot = 0;
1318                 }
1319                 break;
1320         }
1321
1322         case DIOCCHANGERULE: {
1323                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1324                 struct pf_ruleset       *ruleset;
1325                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1326                 u_int32_t                nr = 0;
1327                 int                      rs_num;
1328
1329                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1330                     pcr->action == PF_CHANGE_GET_TICKET) &&
1331                     pcr->pool_ticket != ticket_pabuf) {
1332                         error = EBUSY;
1333                         break;
1334                 }
1335
1336                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1337                     pcr->action > PF_CHANGE_GET_TICKET) {
1338                         error = EINVAL;
1339                         break;
1340                 }
1341                 ruleset = pf_find_ruleset(pcr->anchor);
1342                 if (ruleset == NULL) {
1343                         error = EINVAL;
1344                         break;
1345                 }
1346                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1347                 if (rs_num >= PF_RULESET_MAX) {
1348                         error = EINVAL;
1349                         break;
1350                 }
1351
1352                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1353                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1354                         break;
1355                 } else {
1356                         if (pcr->ticket !=
1357                             ruleset->rules[rs_num].active.ticket) {
1358                                 error = EINVAL;
1359                                 break;
1360                         }
1361                         if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1362                                 error = EINVAL;
1363                                 break;
1364                         }
1365                 }
1366
1367                 if (pcr->action != PF_CHANGE_REMOVE) {
1368                         newrule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK|M_NULLOK);
1369                         if (newrule == NULL) {
1370                                 error = ENOMEM;
1371                                 break;
1372                         }
1373                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1374                         newrule->cuid = ap->a_cred->cr_ruid;
1375                         newrule->cpid = 0;
1376                         TAILQ_INIT(&newrule->rpool.list);
1377                         /* initialize refcounting */
1378                         newrule->states_cur = 0;
1379                         newrule->entries.tqe_prev = NULL;
1380 #ifndef INET
1381                         if (newrule->af == AF_INET) {
1382                                 kfree(newrule, M_PFRULEPL);
1383                                 error = EAFNOSUPPORT;
1384                                 break;
1385                         }
1386 #endif /* INET */
1387 #ifndef INET6
1388                         if (newrule->af == AF_INET6) {
1389                                 kfree(newrule, M_PFRULEPL);
1390                                 error = EAFNOSUPPORT;
1391                                 break;
1392                         }
1393 #endif /* INET6 */
1394                         if (newrule->ifname[0]) {
1395                                 newrule->kif = pfi_kif_get(newrule->ifname);
1396                                 if (newrule->kif == NULL) {
1397                                         kfree(newrule, M_PFRULEPL);
1398                                         error = EINVAL;
1399                                         break;
1400                                 }
1401                                 pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1402                         } else
1403                                 newrule->kif = NULL;
1404
1405                         if (newrule->rtableid > 0 &&
1406                             newrule->rtableid > rt_numfibs)
1407                                 error = EBUSY;
1408
1409 #ifdef ALTQ
1410                         /* set queue IDs */
1411                         if (newrule->qname[0] != 0) {
1412                                 if ((newrule->qid =
1413                                     pf_qname2qid(newrule->qname)) == 0)
1414                                         error = EBUSY;
1415                                 else if (newrule->pqname[0] != 0) {
1416                                         if ((newrule->pqid =
1417                                             pf_qname2qid(newrule->pqname)) == 0)
1418                                                 error = EBUSY;
1419                                 } else
1420                                         newrule->pqid = newrule->qid;
1421                         }
1422 #endif /* ALTQ */
1423                         if (newrule->tagname[0])
1424                                 if ((newrule->tag =
1425                                     pf_tagname2tag(newrule->tagname)) == 0)
1426                                         error = EBUSY;
1427                         if (newrule->match_tagname[0])
1428                                 if ((newrule->match_tag = pf_tagname2tag(
1429                                     newrule->match_tagname)) == 0)
1430                                         error = EBUSY;
1431                         if (newrule->rt && !newrule->direction)
1432                                 error = EINVAL;
1433 #if NPFLOG > 0
1434                         if (!newrule->log)
1435                                 newrule->logif = 0;
1436                         if (newrule->logif >= PFLOGIFS_MAX)
1437                                 error = EINVAL;
1438 #endif
1439                         if (pf_rtlabel_add(&newrule->src.addr) ||
1440                             pf_rtlabel_add(&newrule->dst.addr))
1441                                 error = EBUSY;
1442                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1443                                 error = EINVAL;
1444                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1445                                 error = EINVAL;
1446                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1447                                 error = EINVAL;
1448                         TAILQ_FOREACH(pa, &pf_pabuf, entries)
1449                                 if (pf_tbladdr_setup(ruleset, &pa->addr))
1450                                         error = EINVAL;
1451
1452                         if (newrule->overload_tblname[0]) {
1453                                 if ((newrule->overload_tbl = pfr_attach_table(
1454                                     ruleset, newrule->overload_tblname)) ==
1455                                     NULL)
1456                                         error = EINVAL;
1457                                 else
1458                                         newrule->overload_tbl->pfrkt_flags |=
1459                                             PFR_TFLAG_ACTIVE;
1460                         }
1461
1462                         pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1463                         if (((((newrule->action == PF_NAT) ||
1464                             (newrule->action == PF_RDR) ||
1465                             (newrule->action == PF_BINAT) ||
1466                             (newrule->rt > PF_FASTROUTE)) &&
1467                             !newrule->anchor)) &&
1468                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1469                                 error = EINVAL;
1470
1471                         if (error) {
1472                                 pf_rm_rule(NULL, newrule);
1473                                 break;
1474                         }
1475                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1476                         newrule->evaluations = 0;
1477                         newrule->packets[0] = newrule->packets[1] = 0;
1478                         newrule->bytes[0] = newrule->bytes[1] = 0;
1479                 }
1480                 pf_empty_pool(&pf_pabuf);
1481
1482                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1483                         oldrule = TAILQ_FIRST(
1484                             ruleset->rules[rs_num].active.ptr);
1485                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1486                         oldrule = TAILQ_LAST(
1487                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1488                 else {
1489                         oldrule = TAILQ_FIRST(
1490                             ruleset->rules[rs_num].active.ptr);
1491                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1492                                 oldrule = TAILQ_NEXT(oldrule, entries);
1493                         if (oldrule == NULL) {
1494                                 if (newrule != NULL)
1495                                         pf_rm_rule(NULL, newrule);
1496                                 error = EINVAL;
1497                                 break;
1498                         }
1499                 }
1500
1501                 if (pcr->action == PF_CHANGE_REMOVE) {
1502                         pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1503                         ruleset->rules[rs_num].active.rcount--;
1504                 } else {
1505                         if (oldrule == NULL)
1506                                 TAILQ_INSERT_TAIL(
1507                                     ruleset->rules[rs_num].active.ptr,
1508                                     newrule, entries);
1509                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1510                             pcr->action == PF_CHANGE_ADD_BEFORE)
1511                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1512                         else
1513                                 TAILQ_INSERT_AFTER(
1514                                     ruleset->rules[rs_num].active.ptr,
1515                                     oldrule, newrule, entries);
1516                         ruleset->rules[rs_num].active.rcount++;
1517                 }
1518
1519                 nr = 0;
1520                 TAILQ_FOREACH(oldrule,
1521                     ruleset->rules[rs_num].active.ptr, entries)
1522                         oldrule->nr = nr++;
1523
1524                 ruleset->rules[rs_num].active.ticket++;
1525
1526                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1527                 pf_remove_if_empty_ruleset(ruleset);
1528
1529                 break;
1530         }
1531
1532         case DIOCCLRSTATES: {
1533                 struct pf_state         *s, *nexts;
1534                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1535                 u_int                    killed = 0;
1536
1537                 for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
1538                         nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1539
1540                         if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1541                             s->kif->pfik_name)) {
1542                                 /* don't send out individual delete messages */
1543                                 s->sync_flags = PFSTATE_NOSYNC;
1544                                 pf_unlink_state(s);
1545                                 killed++;
1546                         }
1547                 }
1548                 psk->psk_killed = killed;
1549                 pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1550                 break;
1551         }
1552
1553         case DIOCKILLSTATES: {
1554                 struct pf_state         *s, *nexts;
1555                 struct pf_state_key     *sk;
1556                 struct pf_addr          *srcaddr, *dstaddr;
1557                 u_int16_t                srcport, dstport;
1558                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1559                 u_int                    killed = 0;
1560
1561                 if (psk->psk_pfcmp.id) {
1562                         if (psk->psk_pfcmp.creatorid == 0)
1563                                 psk->psk_pfcmp.creatorid = pf_status.hostid;
1564                         if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1565                                 /* send immediate delete of state */
1566                                 pfsync_delete_state(s);
1567                                 s->sync_flags |= PFSTATE_NOSYNC;
1568                                 pf_unlink_state(s);
1569                                 psk->psk_killed = 1;
1570                         }
1571                         break;
1572                 }
1573
1574                 for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1575                     s = nexts) {
1576                         nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1577                         sk = s->key[PF_SK_WIRE];
1578
1579                         if (s->direction == PF_OUT) {
1580                                 srcaddr = &sk->addr[1];
1581                                 dstaddr = &sk->addr[0];
1582                                 srcport = sk->port[0];
1583                                 dstport = sk->port[0];
1584                         } else {
1585                                 srcaddr = &sk->addr[0];
1586                                 dstaddr = &sk->addr[1];
1587                                 srcport = sk->port[0];
1588                                 dstport = sk->port[0];
1589                         }
1590                         if ((!psk->psk_af || sk->af == psk->psk_af)
1591                             && (!psk->psk_proto || psk->psk_proto ==
1592                             sk->proto) &&
1593                             PF_MATCHA(psk->psk_src.neg,
1594                             &psk->psk_src.addr.v.a.addr,
1595                             &psk->psk_src.addr.v.a.mask,
1596                             srcaddr, sk->af) &&
1597                             PF_MATCHA(psk->psk_dst.neg,
1598                             &psk->psk_dst.addr.v.a.addr,
1599                             &psk->psk_dst.addr.v.a.mask,
1600                             dstaddr, sk->af) &&
1601                             (psk->psk_src.port_op == 0 ||
1602                             pf_match_port(psk->psk_src.port_op,
1603                             psk->psk_src.port[0], psk->psk_src.port[1],
1604                             srcport)) &&
1605                             (psk->psk_dst.port_op == 0 ||
1606                             pf_match_port(psk->psk_dst.port_op,
1607                             psk->psk_dst.port[0], psk->psk_dst.port[1],
1608                             dstport)) &&
1609                             (!psk->psk_label[0] || (s->rule.ptr->label[0] &&
1610                             !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1611                             (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1612                             s->kif->pfik_name))) {
1613                                 /* send immediate delete of state */
1614                                 pfsync_delete_state(s);
1615                                 s->sync_flags |= PFSTATE_NOSYNC;
1616                                 pf_unlink_state(s);
1617                                 killed++;
1618                         }
1619                 }
1620                 psk->psk_killed = killed;
1621                 break;
1622         }
1623
1624         case DIOCADDSTATE: {
1625                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1626                 struct pfsync_state     *sp = &ps->state;
1627
1628                 if (sp->timeout >= PFTM_MAX &&
1629                     sp->timeout != PFTM_UNTIL_PACKET) {
1630                         error = EINVAL;
1631                         break;
1632                 }
1633                 error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1634                 break;
1635         }
1636
1637         case DIOCGETSTATE: {
1638                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1639                 struct pf_state         *s;
1640                 struct pf_state_cmp      id_key;
1641
1642                 bcopy(ps->state.id, &id_key.id, sizeof(id_key.id));
1643                 id_key.creatorid = ps->state.creatorid;
1644
1645                 s = pf_find_state_byid(&id_key);
1646                 if (s == NULL) {
1647                         error = ENOENT;
1648                         break;
1649                 }
1650                 pfsync_state_export(&ps->state, s);
1651                 break;
1652         }
1653
1654         case DIOCGETSTATES: {
1655                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1656                 struct pf_state         *state;
1657                 struct pfsync_state     *p, *pstore;
1658                 u_int32_t                nr = 0;
1659
1660                 if (ps->ps_len == 0) {
1661                         nr = pf_status.states;
1662                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1663                         break;
1664                 }
1665
1666                 pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1667
1668                 p = ps->ps_states;
1669
1670                 state = TAILQ_FIRST(&state_list);
1671                 while (state) {
1672                         if (state->timeout != PFTM_UNLINKED) {
1673                                 if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1674                                         break;
1675                                 pfsync_state_export(pstore, state);
1676                                 error = copyout(pstore, p, sizeof(*p));
1677                                 if (error) {
1678                                         kfree(pstore, M_TEMP);
1679                                         goto fail;
1680                                 }
1681                                 p++;
1682                                 nr++;
1683                         }
1684                         state = TAILQ_NEXT(state, entry_list);
1685                 }
1686
1687                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1688
1689                 kfree(pstore, M_TEMP);
1690                 break;
1691         }
1692
1693         case DIOCGETSTATUS: {
1694                 struct pf_status *s = (struct pf_status *)addr;
1695                 bcopy(&pf_status, s, sizeof(struct pf_status));
1696                 pfi_update_status(s->ifname, s);
1697                 break;
1698         }
1699
1700         case DIOCSETSTATUSIF: {
1701                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1702
1703                 if (pi->ifname[0] == 0) {
1704                         bzero(pf_status.ifname, IFNAMSIZ);
1705                         break;
1706                 }
1707                 strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1708                 break;
1709         }
1710
1711         case DIOCCLRSTATUS: {
1712                 bzero(pf_status.counters, sizeof(pf_status.counters));
1713                 bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1714                 bzero(pf_status.scounters, sizeof(pf_status.scounters));
1715                 pf_status.since = time_second;
1716                 if (*pf_status.ifname)
1717                         pfi_update_status(pf_status.ifname, NULL);
1718                 break;
1719         }
1720
1721         case DIOCNATLOOK: {
1722                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1723                 struct pf_state_key     *sk;
1724                 struct pf_state         *state;
1725                 struct pf_state_key_cmp  key;
1726                 int                      m = 0, direction = pnl->direction;
1727                 int                      sidx, didx;
1728
1729                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1730                 sidx = (direction == PF_IN) ? 1 : 0;
1731                 didx = (direction == PF_IN) ? 0 : 1;
1732
1733                 if (!pnl->proto ||
1734                     PF_AZERO(&pnl->saddr, pnl->af) ||
1735                     PF_AZERO(&pnl->daddr, pnl->af) ||
1736                     ((pnl->proto == IPPROTO_TCP ||
1737                     pnl->proto == IPPROTO_UDP) &&
1738                     (!pnl->dport || !pnl->sport)))
1739                         error = EINVAL;
1740                 else {
1741                         key.af = pnl->af;
1742                         key.proto = pnl->proto;
1743                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1744                         key.port[sidx] = pnl->sport;
1745                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1746                         key.port[didx] = pnl->dport;
1747
1748                         state = pf_find_state_all(&key, direction, &m);
1749
1750                         if (m > 1)
1751                                 error = E2BIG;  /* more than one state */
1752                         else if (state != NULL) {
1753                                 sk = state->key[sidx];
1754                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1755                                 pnl->rsport = sk->port[sidx];
1756                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1757                                 pnl->rdport = sk->port[didx];
1758                         } else
1759                                 error = ENOENT;
1760                 }
1761                 break;
1762         }
1763
1764         case DIOCSETTIMEOUT: {
1765                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1766                 int              old;
1767
1768                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1769                     pt->seconds < 0) {
1770                         error = EINVAL;
1771                         goto fail;
1772                 }
1773                 old = pf_default_rule.timeout[pt->timeout];
1774                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1775                         pt->seconds = 1;
1776                 pf_default_rule.timeout[pt->timeout] = pt->seconds;
1777                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1778                         wakeup(pf_purge_thread);
1779                 pt->seconds = old;
1780                 break;
1781         }
1782
1783         case DIOCGETTIMEOUT: {
1784                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1785
1786                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1787                         error = EINVAL;
1788                         goto fail;
1789                 }
1790                 pt->seconds = pf_default_rule.timeout[pt->timeout];
1791                 break;
1792         }
1793
1794         case DIOCGETLIMIT: {
1795                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1796
1797                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1798                         error = EINVAL;
1799                         goto fail;
1800                 }
1801                 pl->limit = pf_pool_limits[pl->index].limit;
1802                 break;
1803         }
1804
1805         case DIOCSETLIMIT: {
1806                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1807                 int                      old_limit;
1808
1809                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1810                     pf_pool_limits[pl->index].pp == NULL) {
1811                         error = EINVAL;
1812                         goto fail;
1813                 }
1814
1815                 /* XXX Get an API to set limits on the zone/pool */
1816                 old_limit = pf_pool_limits[pl->index].limit;
1817                 pf_pool_limits[pl->index].limit = pl->limit;
1818                 pl->limit = old_limit;
1819                 break;
1820         }
1821
1822         case DIOCSETDEBUG: {
1823                 u_int32_t       *level = (u_int32_t *)addr;
1824
1825                 pf_status.debug = *level;
1826                 break;
1827         }
1828
1829         case DIOCCLRRULECTRS: {
1830                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1831                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1832                 struct pf_rule          *rule;
1833
1834                 TAILQ_FOREACH(rule,
1835                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1836                         rule->evaluations = 0;
1837                         rule->packets[0] = rule->packets[1] = 0;
1838                         rule->bytes[0] = rule->bytes[1] = 0;
1839                 }
1840                 break;
1841         }
1842
1843         case DIOCGIFSPEED: {
1844                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1845                 struct pf_ifspeed       ps;
1846                 struct ifnet            *ifp;
1847
1848                 if (psp->ifname[0] != 0) {
1849                         /* Can we completely trust user-land? */
1850                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1851                         ifp = ifunit(ps.ifname);
1852                         if (ifp )
1853                                 psp->baudrate = ifp->if_baudrate;
1854                         else
1855                                 error = EINVAL;
1856                 } else
1857                         error = EINVAL;
1858                 break;
1859         }
1860 #ifdef ALTQ
1861         case DIOCSTARTALTQ: {
1862                 struct pf_altq          *altq;
1863
1864                 /* enable all altq interfaces on active list */
1865                 TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1866                         if (altq->qname[0] == 0) {
1867                                 error = pf_enable_altq(altq);
1868                                 if (error != 0)
1869                                         break;
1870                         }
1871                 }
1872                 if (error == 0)
1873                         pf_altq_running = 1;
1874                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1875                 break;
1876         }
1877
1878         case DIOCSTOPALTQ: {
1879                 struct pf_altq          *altq;
1880
1881                 /* disable all altq interfaces on active list */
1882                 TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1883                         if (altq->qname[0] == 0) {
1884                                 error = pf_disable_altq(altq);
1885                                 if (error != 0)
1886                                         break;
1887                         }
1888                 }
1889                 if (error == 0)
1890                         pf_altq_running = 0;
1891                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1892                 break;
1893         }
1894
1895         case DIOCADDALTQ: {
1896                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
1897                 struct pf_altq          *altq, *a;
1898
1899                 if (pa->ticket != ticket_altqs_inactive) {
1900                         error = EBUSY;
1901                         break;
1902                 }
1903                 altq = kmalloc(sizeof(struct pf_altq), M_PFALTQPL, M_WAITOK|M_NULLOK);
1904                 if (altq == NULL) {
1905                         error = ENOMEM;
1906                         break;
1907                 }
1908                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1909
1910                 /*
1911                  * if this is for a queue, find the discipline and
1912                  * copy the necessary fields
1913                  */
1914                 if (altq->qname[0] != 0) {
1915                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
1916                                 error = EBUSY;
1917                                 kfree(altq, M_PFALTQPL);
1918                                 break;
1919                         }
1920                         altq->altq_disc = NULL;
1921                         TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
1922                                 if (strncmp(a->ifname, altq->ifname,
1923                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
1924                                         altq->altq_disc = a->altq_disc;
1925                                         break;
1926                                 }
1927                         }
1928                 }
1929
1930                 error = altq_add(altq);
1931                 if (error) {
1932                         kfree(altq, M_PFALTQPL);
1933                         break;
1934                 }
1935
1936                 TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
1937                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1938                 break;
1939         }
1940
1941         case DIOCGETALTQS: {
1942                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
1943                 struct pf_altq          *altq;
1944
1945                 pa->nr = 0;
1946                 TAILQ_FOREACH(altq, pf_altqs_active, entries)
1947                         pa->nr++;
1948                 pa->ticket = ticket_altqs_active;
1949                 break;
1950         }
1951
1952         case DIOCGETALTQ: {
1953                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
1954                 struct pf_altq          *altq;
1955                 u_int32_t                nr;
1956
1957                 if (pa->ticket != ticket_altqs_active) {
1958                         error = EBUSY;
1959                         break;
1960                 }
1961                 nr = 0;
1962                 altq = TAILQ_FIRST(pf_altqs_active);
1963                 while ((altq != NULL) && (nr < pa->nr)) {
1964                         altq = TAILQ_NEXT(altq, entries);
1965                         nr++;
1966                 }
1967                 if (altq == NULL) {
1968                         error = EBUSY;
1969                         break;
1970                 }
1971                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1972                 break;
1973         }
1974
1975         case DIOCCHANGEALTQ:
1976                 /* CHANGEALTQ not supported yet! */
1977                 error = ENODEV;
1978                 break;
1979
1980         case DIOCGETQSTATS: {
1981                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
1982                 struct pf_altq          *altq;
1983                 u_int32_t                nr;
1984                 int                      nbytes;
1985
1986                 if (pq->ticket != ticket_altqs_active) {
1987                         error = EBUSY;
1988                         break;
1989                 }
1990                 nbytes = pq->nbytes;
1991                 nr = 0;
1992                 altq = TAILQ_FIRST(pf_altqs_active);
1993                 while ((altq != NULL) && (nr < pq->nr)) {
1994                         altq = TAILQ_NEXT(altq, entries);
1995                         nr++;
1996                 }
1997                 if (altq == NULL) {
1998                         error = EBUSY;
1999                         break;
2000                 }
2001                 error = altq_getqstats(altq, pq->buf, &nbytes);
2002                 if (error == 0) {
2003                         pq->scheduler = altq->scheduler;
2004                         pq->nbytes = nbytes;
2005                 }
2006                 break;
2007         }
2008 #endif /* ALTQ */
2009
2010         case DIOCBEGINADDRS: {
2011                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2012
2013                 pf_empty_pool(&pf_pabuf);
2014                 pp->ticket = ++ticket_pabuf;
2015                 break;
2016         }
2017
2018         case DIOCADDADDR: {
2019                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2020
2021                 if (pp->ticket != ticket_pabuf) {
2022                         error = EBUSY;
2023                         break;
2024                 }
2025 #ifndef INET
2026                 if (pp->af == AF_INET) {
2027                         error = EAFNOSUPPORT;
2028                         break;
2029                 }
2030 #endif /* INET */
2031 #ifndef INET6
2032                 if (pp->af == AF_INET6) {
2033                         error = EAFNOSUPPORT;
2034                         break;
2035                 }
2036 #endif /* INET6 */
2037                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2038                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2039                     pp->addr.addr.type != PF_ADDR_TABLE) {
2040                         error = EINVAL;
2041                         break;
2042                 }
2043                 pa = kmalloc(sizeof(struct pf_altq), M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2044                 if (pa == NULL) {
2045                         error = ENOMEM;
2046                         break;
2047                 }
2048                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2049                 if (pa->ifname[0]) {
2050                         pa->kif = pfi_kif_get(pa->ifname);
2051                         if (pa->kif == NULL) {
2052                                 kfree(ap, M_PFPOOLADDRPL);
2053                                 error = EINVAL;
2054                                 break;
2055                         }
2056                         pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2057                 }
2058                 if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2059                         pfi_dynaddr_remove(&pa->addr);
2060                         pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2061                         kfree(pa, M_PFPOOLADDRPL);
2062                         error = EINVAL;
2063                         break;
2064                 }
2065                 TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2066                 break;
2067         }
2068
2069         case DIOCGETADDRS: {
2070                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2071
2072                 pp->nr = 0;
2073                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2074                     pp->r_num, 0, 1, 0);
2075                 if (pool == NULL) {
2076                         error = EBUSY;
2077                         break;
2078                 }
2079                 TAILQ_FOREACH(pa, &pool->list, entries)
2080                         pp->nr++;
2081                 break;
2082         }
2083
2084         case DIOCGETADDR: {
2085                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2086                 u_int32_t                nr = 0;
2087
2088                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2089                     pp->r_num, 0, 1, 1);
2090                 if (pool == NULL) {
2091                         error = EBUSY;
2092                         break;
2093                 }
2094                 pa = TAILQ_FIRST(&pool->list);
2095                 while ((pa != NULL) && (nr < pp->nr)) {
2096                         pa = TAILQ_NEXT(pa, entries);
2097                         nr++;
2098                 }
2099                 if (pa == NULL) {
2100                         error = EBUSY;
2101                         break;
2102                 }
2103                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2104                 pf_addr_copyout(&pp->addr.addr);
2105                 break;
2106         }
2107
2108         case DIOCCHANGEADDR: {
2109                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2110                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2111                 struct pf_ruleset       *ruleset;
2112
2113                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2114                     pca->action > PF_CHANGE_REMOVE) {
2115                         error = EINVAL;
2116                         break;
2117                 }
2118                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2119                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2120                     pca->addr.addr.type != PF_ADDR_TABLE) {
2121                         error = EINVAL;
2122                         break;
2123                 }
2124
2125                 ruleset = pf_find_ruleset(pca->anchor);
2126                 if (ruleset == NULL) {
2127                         error = EBUSY;
2128                         break;
2129                 }
2130                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2131                     pca->r_num, pca->r_last, 1, 1);
2132                 if (pool == NULL) {
2133                         error = EBUSY;
2134                         break;
2135                 }
2136                 if (pca->action != PF_CHANGE_REMOVE) {
2137                         newpa = kmalloc(sizeof(struct pf_pooladdr),
2138                                 M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2139                         if (newpa == NULL) {
2140                                 error = ENOMEM;
2141                                 break;
2142                         }
2143                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2144 #ifndef INET
2145                         if (pca->af == AF_INET) {
2146                                 kfree(newpa, M_PFPOOLADDRPL);
2147                                 error = EAFNOSUPPORT;
2148                                 break;
2149                         }
2150 #endif /* INET */
2151 #ifndef INET6
2152                         if (pca->af == AF_INET6) {
2153                                 kfree(newpa, M_PFPOOLADDRPL);
2154                                 error = EAFNOSUPPORT;
2155                                 break;
2156                         }
2157 #endif /* INET6 */
2158                         if (newpa->ifname[0]) {
2159                                 newpa->kif = pfi_kif_get(newpa->ifname);
2160                                 if (newpa->kif == NULL) {
2161                                         kfree(newpa, M_PFPOOLADDRPL);
2162                                         error = EINVAL;
2163                                         break;
2164                                 }
2165                                 pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2166                         } else
2167                                 newpa->kif = NULL;
2168                         if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2169                             pf_tbladdr_setup(ruleset, &newpa->addr)) {
2170                                 pfi_dynaddr_remove(&newpa->addr);
2171                                 pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2172                                 kfree(newpa, M_PFPOOLADDRPL);
2173                                 error = EINVAL;
2174                                 break;
2175                         }
2176                 }
2177
2178                 if (pca->action == PF_CHANGE_ADD_HEAD)
2179                         oldpa = TAILQ_FIRST(&pool->list);
2180                 else if (pca->action == PF_CHANGE_ADD_TAIL)
2181                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2182                 else {
2183                         int     i = 0;
2184
2185                         oldpa = TAILQ_FIRST(&pool->list);
2186                         while ((oldpa != NULL) && (i < pca->nr)) {
2187                                 oldpa = TAILQ_NEXT(oldpa, entries);
2188                                 i++;
2189                         }
2190                         if (oldpa == NULL) {
2191                                 error = EINVAL;
2192                                 break;
2193                         }
2194                 }
2195
2196                 if (pca->action == PF_CHANGE_REMOVE) {
2197                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2198                         pfi_dynaddr_remove(&oldpa->addr);
2199                         pf_tbladdr_remove(&oldpa->addr);
2200                         pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2201                         kfree(oldpa, M_PFPOOLADDRPL);
2202                 } else {
2203                         if (oldpa == NULL)
2204                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2205                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2206                             pca->action == PF_CHANGE_ADD_BEFORE)
2207                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2208                         else
2209                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2210                                     newpa, entries);
2211                 }
2212
2213                 pool->cur = TAILQ_FIRST(&pool->list);
2214                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2215                     pca->af);
2216                 break;
2217         }
2218
2219         case DIOCGETRULESETS: {
2220                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2221                 struct pf_ruleset       *ruleset;
2222                 struct pf_anchor        *anchor;
2223
2224                 pr->path[sizeof(pr->path) - 1] = 0;
2225                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2226                         error = EINVAL;
2227                         break;
2228                 }
2229                 pr->nr = 0;
2230                 if (ruleset->anchor == NULL) {
2231                         /* XXX kludge for pf_main_ruleset */
2232                         RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2233                                 if (anchor->parent == NULL)
2234                                         pr->nr++;
2235                 } else {
2236                         RB_FOREACH(anchor, pf_anchor_node,
2237                             &ruleset->anchor->children)
2238                                 pr->nr++;
2239                 }
2240                 break;
2241         }
2242
2243         case DIOCGETRULESET: {
2244                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2245                 struct pf_ruleset       *ruleset;
2246                 struct pf_anchor        *anchor;
2247                 u_int32_t                nr = 0;
2248
2249                 pr->path[sizeof(pr->path) - 1] = 0;
2250                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2251                         error = EINVAL;
2252                         break;
2253                 }
2254                 pr->name[0] = 0;
2255                 if (ruleset->anchor == NULL) {
2256                         /* XXX kludge for pf_main_ruleset */
2257                         RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2258                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2259                                         strlcpy(pr->name, anchor->name,
2260                                             sizeof(pr->name));
2261                                         break;
2262                                 }
2263                 } else {
2264                         RB_FOREACH(anchor, pf_anchor_node,
2265                             &ruleset->anchor->children)
2266                                 if (nr++ == pr->nr) {
2267                                         strlcpy(pr->name, anchor->name,
2268                                             sizeof(pr->name));
2269                                         break;
2270                                 }
2271                 }
2272                 if (!pr->name[0])
2273                         error = EBUSY;
2274                 break;
2275         }
2276
2277         case DIOCRCLRTABLES: {
2278                 struct pfioc_table *io = (struct pfioc_table *)addr;
2279
2280                 if (io->pfrio_esize != 0) {
2281                         error = ENODEV;
2282                         break;
2283                 }
2284                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2285                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2286                 break;
2287         }
2288
2289         case DIOCRADDTABLES: {
2290                 struct pfioc_table *io = (struct pfioc_table *)addr;
2291
2292                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2293                         error = ENODEV;
2294                         break;
2295                 }
2296                 error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2297                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2298                 break;
2299         }
2300
2301         case DIOCRDELTABLES: {
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_del_tables(io->pfrio_buffer, io->pfrio_size,
2309                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2310                 break;
2311         }
2312
2313         case DIOCRGETTABLES: {
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_get_tables(&io->pfrio_table, io->pfrio_buffer,
2321                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2322                 break;
2323         }
2324
2325         case DIOCRGETTSTATS: {
2326                 struct pfioc_table *io = (struct pfioc_table *)addr;
2327
2328                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2329                         error = ENODEV;
2330                         break;
2331                 }
2332                 error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2333                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2334                 break;
2335         }
2336
2337         case DIOCRCLRTSTATS: {
2338                 struct pfioc_table *io = (struct pfioc_table *)addr;
2339
2340                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2341                         error = ENODEV;
2342                         break;
2343                 }
2344                 error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2345                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2346                 break;
2347         }
2348
2349         case DIOCRSETTFLAGS: {
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_set_tflags(io->pfrio_buffer, io->pfrio_size,
2357                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2358                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2359                 break;
2360         }
2361
2362         case DIOCRCLRADDRS: {
2363                 struct pfioc_table *io = (struct pfioc_table *)addr;
2364
2365                 if (io->pfrio_esize != 0) {
2366                         error = ENODEV;
2367                         break;
2368                 }
2369                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2370                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2371                 break;
2372         }
2373
2374         case DIOCRADDADDRS: {
2375                 struct pfioc_table *io = (struct pfioc_table *)addr;
2376
2377                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2378                         error = ENODEV;
2379                         break;
2380                 }
2381                 error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2382                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2383                     PFR_FLAG_USERIOCTL);
2384                 break;
2385         }
2386
2387         case DIOCRDELADDRS: {
2388                 struct pfioc_table *io = (struct pfioc_table *)addr;
2389
2390                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2391                         error = ENODEV;
2392                         break;
2393                 }
2394                 error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2395                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2396                     PFR_FLAG_USERIOCTL);
2397                 break;
2398         }
2399
2400         case DIOCRSETADDRS: {
2401                 struct pfioc_table *io = (struct pfioc_table *)addr;
2402
2403                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2404                         error = ENODEV;
2405                         break;
2406                 }
2407                 error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2408                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2409                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2410                     PFR_FLAG_USERIOCTL, 0);
2411                 break;
2412         }
2413
2414         case DIOCRGETADDRS: {
2415                 struct pfioc_table *io = (struct pfioc_table *)addr;
2416
2417                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2418                         error = ENODEV;
2419                         break;
2420                 }
2421                 error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2422                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2423                 break;
2424         }
2425
2426         case DIOCRGETASTATS: {
2427                 struct pfioc_table *io = (struct pfioc_table *)addr;
2428
2429                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2430                         error = ENODEV;
2431                         break;
2432                 }
2433                 error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2434                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2435                 break;
2436         }
2437
2438         case DIOCRCLRASTATS: {
2439                 struct pfioc_table *io = (struct pfioc_table *)addr;
2440
2441                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2442                         error = ENODEV;
2443                         break;
2444                 }
2445                 error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2446                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2447                     PFR_FLAG_USERIOCTL);
2448                 break;
2449         }
2450
2451         case DIOCRTSTADDRS: {
2452                 struct pfioc_table *io = (struct pfioc_table *)addr;
2453
2454                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2455                         error = ENODEV;
2456                         break;
2457                 }
2458                 error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2459                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2460                     PFR_FLAG_USERIOCTL);
2461                 break;
2462         }
2463
2464         case DIOCRINADEFINE: {
2465                 struct pfioc_table *io = (struct pfioc_table *)addr;
2466
2467                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2468                         error = ENODEV;
2469                         break;
2470                 }
2471                 error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2472                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2473                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2474                 break;
2475         }
2476
2477         case DIOCOSFPADD: {
2478                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2479                 error = pf_osfp_add(io);
2480                 break;
2481         }
2482
2483         case DIOCOSFPGET: {
2484                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2485                 error = pf_osfp_get(io);
2486                 break;
2487         }
2488
2489         case DIOCXBEGIN: {
2490                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2491                 struct pfioc_trans_e    *ioe;
2492                 struct pfr_table        *table;
2493                 int                      i;
2494
2495                 if (io->esize != sizeof(*ioe)) {
2496                         error = ENODEV;
2497                         goto fail;
2498                 }
2499                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2500                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2501                 for (i = 0; i < io->size; i++) {
2502                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2503                                 kfree(table, M_TEMP);
2504                                 kfree(ioe, M_TEMP);
2505                                 error = EFAULT;
2506                                 goto fail;
2507                         }
2508                         switch (ioe->rs_num) {
2509 #ifdef ALTQ
2510                         case PF_RULESET_ALTQ:
2511                                 if (ioe->anchor[0]) {
2512                                         kfree(table, M_TEMP);
2513                                         kfree(ioe, M_TEMP);
2514                                         error = EINVAL;
2515                                         goto fail;
2516                                 }
2517                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2518                                         kfree(table, M_TEMP);
2519                                         kfree(ioe, M_TEMP);
2520                                         goto fail;
2521                                 }
2522                                 break;
2523 #endif /* ALTQ */
2524                         case PF_RULESET_TABLE:
2525                                 bzero(table, sizeof(*table));
2526                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2527                                     sizeof(table->pfrt_anchor));
2528                                 if ((error = pfr_ina_begin(table,
2529                                     &ioe->ticket, NULL, 0))) {
2530                                         kfree(table, M_TEMP);
2531                                         kfree(ioe, M_TEMP);
2532                                         goto fail;
2533                                 }
2534                                 break;
2535                         default:
2536                                 if ((error = pf_begin_rules(&ioe->ticket,
2537                                     ioe->rs_num, ioe->anchor))) {
2538                                         kfree(table, M_TEMP);
2539                                         kfree(ioe, M_TEMP);
2540                                         goto fail;
2541                                 }
2542                                 break;
2543                         }
2544                         if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2545                                 kfree(table, M_TEMP);
2546                                 kfree(ioe, M_TEMP);
2547                                 error = EFAULT;
2548                                 goto fail;
2549                         }
2550                 }
2551                 kfree(table, M_TEMP);
2552                 kfree(ioe, M_TEMP);
2553                 break;
2554         }
2555
2556         case DIOCXROLLBACK: {
2557                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2558                 struct pfioc_trans_e    *ioe;
2559                 struct pfr_table        *table;
2560                 int                      i;
2561
2562                 if (io->esize != sizeof(*ioe)) {
2563                         error = ENODEV;
2564                         goto fail;
2565                 }
2566                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2567                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2568                 for (i = 0; i < io->size; i++) {
2569                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2570                                 kfree(table, M_TEMP);
2571                                 kfree(ioe, M_TEMP);
2572                                 error = EFAULT;
2573                                 goto fail;
2574                         }
2575                         switch (ioe->rs_num) {
2576 #ifdef ALTQ
2577                         case PF_RULESET_ALTQ:
2578                                 if (ioe->anchor[0]) {
2579                                         kfree(table, M_TEMP);
2580                                         kfree(ioe, M_TEMP);
2581                                         error = EINVAL;
2582                                         goto fail;
2583                                 }
2584                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2585                                         kfree(table, M_TEMP);
2586                                         kfree(ioe, M_TEMP);
2587                                         goto fail; /* really bad */
2588                                 }
2589                                 break;
2590 #endif /* ALTQ */
2591                         case PF_RULESET_TABLE:
2592                                 bzero(table, sizeof(*table));
2593                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2594                                     sizeof(table->pfrt_anchor));
2595                                 if ((error = pfr_ina_rollback(table,
2596                                     ioe->ticket, NULL, 0))) {
2597                                         kfree(table, M_TEMP);
2598                                         kfree(ioe, M_TEMP);
2599                                         goto fail; /* really bad */
2600                                 }
2601                                 break;
2602                         default:
2603                                 if ((error = pf_rollback_rules(ioe->ticket,
2604                                     ioe->rs_num, ioe->anchor))) {
2605                                         kfree(table, M_TEMP);
2606                                         kfree(ioe, M_TEMP);
2607                                         goto fail; /* really bad */
2608                                 }
2609                                 break;
2610                         }
2611                 }
2612                 kfree(table, M_TEMP);
2613                 kfree(ioe, M_TEMP);
2614                 break;
2615         }
2616
2617         case DIOCXCOMMIT: {
2618                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2619                 struct pfioc_trans_e    *ioe;
2620                 struct pfr_table        *table;
2621                 struct pf_ruleset       *rs;
2622                 int                      i;
2623
2624                 if (io->esize != sizeof(*ioe)) {
2625                         error = ENODEV;
2626                         goto fail;
2627                 }
2628                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2629                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2630                 /* first makes sure everything will succeed */
2631                 for (i = 0; i < io->size; i++) {
2632                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2633                                 kfree(table, M_TEMP);
2634                                 kfree(ioe, M_TEMP);
2635                                 error = EFAULT;
2636                                 goto fail;
2637                         }
2638                         switch (ioe->rs_num) {
2639 #ifdef ALTQ
2640                         case PF_RULESET_ALTQ:
2641                                 if (ioe->anchor[0]) {
2642                                         kfree(table, M_TEMP);
2643                                         kfree(ioe, M_TEMP);
2644                                         error = EINVAL;
2645                                         goto fail;
2646                                 }
2647                                 if (!altqs_inactive_open || ioe->ticket !=
2648                                     ticket_altqs_inactive) {
2649                                         kfree(table, M_TEMP);
2650                                         kfree(ioe, M_TEMP);
2651                                         error = EBUSY;
2652                                         goto fail;
2653                                 }
2654                                 break;
2655 #endif /* ALTQ */
2656                         case PF_RULESET_TABLE:
2657                                 rs = pf_find_ruleset(ioe->anchor);
2658                                 if (rs == NULL || !rs->topen || ioe->ticket !=
2659                                      rs->tticket) {
2660                                         kfree(table, M_TEMP);
2661                                         kfree(ioe, M_TEMP);
2662                                         error = EBUSY;
2663                                         goto fail;
2664                                 }
2665                                 break;
2666                         default:
2667                                 if (ioe->rs_num < 0 || ioe->rs_num >=
2668                                     PF_RULESET_MAX) {
2669                                         kfree(table, M_TEMP);
2670                                         kfree(ioe, M_TEMP);
2671                                         error = EINVAL;
2672                                         goto fail;
2673                                 }
2674                                 rs = pf_find_ruleset(ioe->anchor);
2675                                 if (rs == NULL ||
2676                                     !rs->rules[ioe->rs_num].inactive.open ||
2677                                     rs->rules[ioe->rs_num].inactive.ticket !=
2678                                     ioe->ticket) {
2679                                         kfree(table, M_TEMP);
2680                                         kfree(ioe, M_TEMP);
2681                                         error = EBUSY;
2682                                         goto fail;
2683                                 }
2684                                 break;
2685                         }
2686                 }
2687                 /* now do the commit - no errors should happen here */
2688                 for (i = 0; i < io->size; i++) {
2689                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2690                                 kfree(table, M_TEMP);
2691                                 kfree(ioe, M_TEMP);
2692                                 error = EFAULT;
2693                                 goto fail;
2694                         }
2695                         switch (ioe->rs_num) {
2696 #ifdef ALTQ
2697                         case PF_RULESET_ALTQ:
2698                                 if ((error = pf_commit_altq(ioe->ticket))) {
2699                                         kfree(table, M_TEMP);
2700                                         kfree(ioe, M_TEMP);
2701                                         goto fail; /* really bad */
2702                                 }
2703                                 break;
2704 #endif /* ALTQ */
2705                         case PF_RULESET_TABLE:
2706                                 bzero(table, sizeof(*table));
2707                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2708                                     sizeof(table->pfrt_anchor));
2709                                 if ((error = pfr_ina_commit(table, ioe->ticket,
2710                                     NULL, NULL, 0))) {
2711                                         kfree(table, M_TEMP);
2712                                         kfree(ioe, M_TEMP);
2713                                         goto fail; /* really bad */
2714                                 }
2715                                 break;
2716                         default:
2717                                 if ((error = pf_commit_rules(ioe->ticket,
2718                                     ioe->rs_num, ioe->anchor))) {
2719                                         kfree(table, M_TEMP);
2720                                         kfree(ioe, M_TEMP);
2721                                         goto fail; /* really bad */
2722                                 }
2723                                 break;
2724                         }
2725                 }
2726                 kfree(table, M_TEMP);
2727                 kfree(ioe, M_TEMP);
2728                 break;
2729         }
2730
2731         case DIOCGETSRCNODES: {
2732                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
2733                 struct pf_src_node      *n, *p, *pstore;
2734                 u_int32_t                nr = 0;
2735                 int                      space = psn->psn_len;
2736
2737                 if (space == 0) {
2738                         RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2739                                 nr++;
2740                         psn->psn_len = sizeof(struct pf_src_node) * nr;
2741                         break;
2742                 }
2743
2744                 pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2745
2746                 p = psn->psn_src_nodes;
2747                 RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2748                         int     secs = time_second, diff;
2749
2750                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
2751                                 break;
2752
2753                         bcopy(n, pstore, sizeof(*pstore));
2754                         if (n->rule.ptr != NULL)
2755                                 pstore->rule.nr = n->rule.ptr->nr;
2756                         pstore->creation = secs - pstore->creation;
2757                         if (pstore->expire > secs)
2758                                 pstore->expire -= secs;
2759                         else
2760                                 pstore->expire = 0;
2761
2762                         /* adjust the connection rate estimate */
2763                         diff = secs - n->conn_rate.last;
2764                         if (diff >= n->conn_rate.seconds)
2765                                 pstore->conn_rate.count = 0;
2766                         else
2767                                 pstore->conn_rate.count -=
2768                                     n->conn_rate.count * diff /
2769                                     n->conn_rate.seconds;
2770
2771                         error = copyout(pstore, p, sizeof(*p));
2772                         if (error) {
2773                                 kfree(pstore, M_TEMP);
2774                                 goto fail;
2775                         }
2776                         p++;
2777                         nr++;
2778                 }
2779                 psn->psn_len = sizeof(struct pf_src_node) * nr;
2780
2781                 kfree(pstore, M_TEMP);
2782                 break;
2783         }
2784
2785         case DIOCCLRSRCNODES: {
2786                 struct pf_src_node      *n;
2787                 struct pf_state         *state;
2788
2789                 RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2790                         state->src_node = NULL;
2791                         state->nat_src_node = NULL;
2792                 }
2793                 RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2794                         n->expire = 1;
2795                         n->states = 0;
2796                 }
2797                 pf_purge_expired_src_nodes(1);
2798                 pf_status.src_nodes = 0;
2799                 break;
2800         }
2801
2802         case DIOCKILLSRCNODES: {
2803                 struct pf_src_node      *sn;
2804                 struct pf_state         *s;
2805                 struct pfioc_src_node_kill *psnk =
2806                     (struct pfioc_src_node_kill *)addr;
2807                 u_int                   killed = 0;
2808
2809                 RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2810                         if (PF_MATCHA(psnk->psnk_src.neg,
2811                                 &psnk->psnk_src.addr.v.a.addr,
2812                                 &psnk->psnk_src.addr.v.a.mask,
2813                                 &sn->addr, sn->af) &&
2814                             PF_MATCHA(psnk->psnk_dst.neg,
2815                                 &psnk->psnk_dst.addr.v.a.addr,
2816                                 &psnk->psnk_dst.addr.v.a.mask,
2817                                 &sn->raddr, sn->af)) {
2818                                 /* Handle state to src_node linkage */
2819                                 if (sn->states != 0) {
2820                                         RB_FOREACH(s, pf_state_tree_id,
2821                                             &tree_id) {
2822                                                 if (s->src_node == sn)
2823                                                         s->src_node = NULL;
2824                                                 if (s->nat_src_node == sn)
2825                                                         s->nat_src_node = NULL;
2826                                         }
2827                                         sn->states = 0;
2828                                 }
2829                                 sn->expire = 1;
2830                                 killed++;
2831                         }
2832                 }
2833
2834                 if (killed > 0)
2835                         pf_purge_expired_src_nodes(1);
2836
2837                 psnk->psnk_killed = killed;
2838                 break;
2839         }
2840
2841         case DIOCSETHOSTID: {
2842                 u_int32_t       *hostid = (u_int32_t *)addr;
2843
2844                 if (*hostid == 0)
2845                         pf_status.hostid = karc4random();
2846                 else
2847                         pf_status.hostid = *hostid;
2848                 break;
2849         }
2850
2851         case DIOCOSFPFLUSH:
2852                 crit_enter();
2853                 pf_osfp_flush();
2854                 crit_exit();
2855                 break;
2856
2857         case DIOCIGETIFACES: {
2858                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
2859
2860                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2861                         error = ENODEV;
2862                         break;
2863                 }
2864                 error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2865                     &io->pfiio_size);
2866                 break;
2867         }
2868
2869         case DIOCSETIFFLAG: {
2870                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
2871
2872                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2873                 break;
2874         }
2875
2876         case DIOCCLRIFFLAG: {
2877                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
2878
2879                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2880                 break;
2881         }
2882
2883         default:
2884                 error = ENODEV;
2885                 break;
2886         }
2887 fail:
2888         lwkt_reltoken(&pf_token);
2889         return (error);
2890 }
2891
2892 /*
2893  * XXX - Check for version missmatch!!!
2894  */
2895 static void
2896 pf_clear_states(void)
2897 {
2898         struct pf_state         *s, *nexts;
2899         u_int                   killed = 0;
2900
2901         for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
2902                 nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
2903
2904                 /* don't send out individual delete messages */
2905                 s->sync_flags = PFSTATE_NOSYNC;
2906                 pf_unlink_state(s);
2907                 killed++;
2908
2909         }
2910
2911 #if 0 /* PFSYNC */
2912 /*
2913  * XXX This is called on module unload, we do not want to sync that over? */
2914  */
2915         pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
2916 #endif
2917 }
2918
2919 static int
2920 pf_clear_tables(void)
2921 {
2922         struct pfioc_table io;
2923         int error;
2924
2925         bzero(&io, sizeof(io));
2926
2927         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
2928             io.pfrio_flags);
2929
2930         return (error);
2931 }
2932
2933 static void
2934 pf_clear_srcnodes(void)
2935 {
2936         struct pf_src_node      *n;
2937         struct pf_state         *state;
2938
2939         RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2940                 state->src_node = NULL;
2941                 state->nat_src_node = NULL;
2942         }
2943         RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2944                 n->expire = 1;
2945                 n->states = 0;
2946         }
2947         pf_purge_expired_src_nodes(0);
2948         pf_status.src_nodes = 0;
2949 }
2950
2951 /*
2952  * XXX - Check for version missmatch!!!
2953  */
2954
2955 /*
2956  * Duplicate pfctl -Fa operation to get rid of as much as we can.
2957  */
2958 static int
2959 shutdown_pf(void)
2960 {
2961         int error = 0;
2962         u_int32_t t[5];
2963         char nn = '\0';
2964
2965
2966         pf_status.running = 0;
2967         error = dehook_pf();
2968         if (error) {
2969                 pf_status.running = 1;
2970                 DPFPRINTF(PF_DEBUG_MISC,
2971                     ("pf: pfil unregistration failed\n"));
2972                 return(error);
2973         }
2974         do {
2975                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
2976                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
2977                         break;
2978                 }
2979                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
2980                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
2981                         break;          /* XXX: rollback? */
2982                 }
2983                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
2984                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
2985                         break;          /* XXX: rollback? */
2986                 }
2987                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
2988                     != 0) {
2989                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
2990                         break;          /* XXX: rollback? */
2991                 }
2992                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
2993                     != 0) {
2994                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
2995                         break;          /* XXX: rollback? */
2996                 }
2997
2998                 /* XXX: these should always succeed here */
2999                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3000                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3001                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3002                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3003                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3004
3005                 if ((error = pf_clear_tables()) != 0)
3006                         break;
3007 #ifdef ALTQ
3008                 if ((error = pf_begin_altq(&t[0])) != 0) {
3009                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3010                         break;
3011                 }
3012                 pf_commit_altq(t[0]);
3013 #endif
3014                 pf_clear_states();
3015                 pf_clear_srcnodes();
3016
3017                 /* status does not use malloced mem so no need to cleanup */
3018                 /* fingerprints and interfaces have their own cleanup code */
3019         } while(0);
3020         return (error);
3021 }
3022
3023 static int
3024 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3025 {
3026         /*
3027          * DragonFly's version of pf uses FreeBSD's native host byte ordering
3028          * for ip_len/ip_off. This is why we don't have to change byte order
3029          * like the FreeBSD-5 version does.
3030          */
3031         int chk;
3032
3033         lwkt_gettoken(&pf_token);
3034
3035         chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3036         if (chk && *m) {
3037                 m_freem(*m);
3038                 *m = NULL;
3039         }
3040         lwkt_reltoken(&pf_token);
3041         return chk;
3042 }
3043
3044 static int
3045 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3046 {
3047         /*
3048          * DragonFly's version of pf uses FreeBSD's native host byte ordering
3049          * for ip_len/ip_off. This is why we don't have to change byte order
3050          * like the FreeBSD-5 version does.
3051          */
3052         int chk;
3053
3054         lwkt_gettoken(&pf_token);
3055
3056         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3057         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3058                 in_delayed_cksum(*m);
3059                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3060         }
3061         chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3062         if (chk && *m) {
3063                 m_freem(*m);
3064                 *m = NULL;
3065         }
3066         lwkt_reltoken(&pf_token);
3067         return chk;
3068 }
3069
3070 #ifdef INET6
3071 static int
3072 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3073 {
3074         /*
3075          * IPv6 is not affected by ip_len/ip_off byte order changes.
3076          */
3077         int chk;
3078
3079         lwkt_gettoken(&pf_token);
3080
3081         chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3082         if (chk && *m) {
3083                 m_freem(*m);
3084                 *m = NULL;
3085         }
3086         lwkt_reltoken(&pf_token);
3087         return chk;
3088 }
3089
3090 static int
3091 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3092 {
3093         /*
3094          * IPv6 is not affected by ip_len/ip_off byte order changes.
3095          */
3096         int chk;
3097
3098         lwkt_gettoken(&pf_token);
3099
3100         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3101         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3102                 in_delayed_cksum(*m);
3103                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3104         }
3105         chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3106         if (chk && *m) {
3107                 m_freem(*m);
3108                 *m = NULL;
3109         }
3110         lwkt_reltoken(&pf_token);
3111         return chk;
3112 }
3113 #endif /* INET6 */
3114
3115 static int
3116 hook_pf(void)
3117 {
3118         struct pfil_head *pfh_inet;
3119 #ifdef INET6
3120         struct pfil_head *pfh_inet6;
3121 #endif
3122
3123         lwkt_gettoken(&pf_token);
3124
3125         if (pf_pfil_hooked) {
3126                 lwkt_reltoken(&pf_token);
3127                 return (0);
3128         }
3129
3130         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3131         if (pfh_inet == NULL) {
3132                 lwkt_reltoken(&pf_token);
3133                 return (ENODEV);
3134         }
3135         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet);
3136         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet);
3137 #ifdef INET6
3138         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3139         if (pfh_inet6 == NULL) {
3140                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3141                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3142                 lwkt_reltoken(&pf_token);
3143                 return (ENODEV);
3144         }
3145         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet6);
3146         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet6);
3147 #endif
3148
3149         pf_pfil_hooked = 1;
3150         lwkt_reltoken(&pf_token);
3151         return (0);
3152 }
3153
3154 static int
3155 dehook_pf(void)
3156 {
3157         struct pfil_head *pfh_inet;
3158 #ifdef INET6
3159         struct pfil_head *pfh_inet6;
3160 #endif
3161
3162         lwkt_gettoken(&pf_token);
3163
3164         if (pf_pfil_hooked == 0) {
3165                 lwkt_reltoken(&pf_token);
3166                 return (0);
3167         }
3168
3169         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3170         if (pfh_inet == NULL) {
3171                 lwkt_reltoken(&pf_token);
3172                 return (ENODEV);
3173         }
3174         pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3175         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3176 #ifdef INET6
3177         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3178         if (pfh_inet6 == NULL) {
3179                 lwkt_reltoken(&pf_token);
3180                 return (ENODEV);
3181         }
3182         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3183         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3184 #endif
3185
3186         pf_pfil_hooked = 0;
3187         lwkt_reltoken(&pf_token);
3188         return (0);
3189 }
3190
3191 static int
3192 pf_load(void)
3193 {
3194         lwkt_gettoken(&pf_token);
3195
3196         pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3197         pfattach();
3198         lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3199         lwkt_reltoken(&pf_token);
3200         return (0);
3201 }
3202
3203 static int
3204 pf_mask_del(struct radix_node *rn, void *arg)
3205 {
3206         struct radix_node_head *rnh = arg;
3207
3208         rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
3209         Free(rn);
3210         return 0;
3211 }
3212
3213 static int
3214 pf_unload(void)
3215 {
3216         int error;
3217         pf_status.running = 0;
3218
3219         lwkt_gettoken(&pf_token);
3220
3221         error = dehook_pf();
3222         if (error) {
3223                 /*
3224                  * Should not happen!
3225                  * XXX Due to error code ESRCH, kldunload will show
3226                  * a message like 'No such process'.
3227                  */
3228                 kprintf("pfil unregistration fail\n");
3229                 lwkt_reltoken(&pf_token);
3230                 return error;
3231         }
3232         shutdown_pf();
3233         pf_end_threads = 1;
3234         while (pf_end_threads < 2) {
3235                 wakeup_one(pf_purge_thread);
3236                 tsleep(pf_purge_thread, 0, "pftmo", hz);
3237         }
3238         pfi_cleanup();
3239         pf_osfp_flush();
3240         dev_ops_remove_all(&pf_ops);
3241         lockuninit(&pf_consistency_lock);
3242         lwkt_reltoken(&pf_token);
3243
3244         if (pf_maskhead != NULL) {
3245                 pf_maskhead->rnh_walktree(pf_maskhead,
3246                         pf_mask_del, pf_maskhead);
3247                 Free(pf_maskhead);
3248                 pf_maskhead = NULL;
3249         }
3250         kmalloc_destroy(&pf_state_pl);
3251         kmalloc_destroy(&pf_frent_pl);
3252         kmalloc_destroy(&pf_cent_pl);
3253         return 0;
3254 }
3255
3256 static int
3257 pf_modevent(module_t mod, int type, void *data)
3258 {
3259         int error = 0;
3260
3261         lwkt_gettoken(&pf_token);
3262
3263         switch(type) {
3264         case MOD_LOAD:
3265                 error = pf_load();
3266                 break;
3267
3268         case MOD_UNLOAD:
3269                 error = pf_unload();
3270                 break;
3271         default:
3272                 error = EINVAL;
3273                 break;
3274         }
3275         lwkt_reltoken(&pf_token);
3276         return error;
3277 }
3278
3279 static moduledata_t pf_mod = {
3280         "pf",
3281         pf_modevent,
3282         0
3283 };
3284 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3285 MODULE_VERSION(pf, PF_MODVER);