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
  77 #include <net/pf/if_pflog.h>
  78 #include <net/pf/if_pfsync.h>
  79
  80 #ifdef INET6
  81 #include <netinet/ip6.h>
  82 #include <netinet/in_pcb.h>
  83 #endif /* INET6 */
  84
  85 #ifdef ALTQ
  86 #include <net/altq/altq.h>
  87 #endif
  88
  89 #include <machine/limits.h>
  90 #include <net/pfil.h>
  91 #include <sys/mutex.h>
  92
  93 u_int rt_numfibs = RT_NUMFIBS;
  94
  95 void                     pfattach(void);
  96 struct pf_pool          *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
  97                             u_int8_t, u_int8_t, u_int8_t);
  98
  99 void                     pf_mv_pool(struct pf_palist *, struct pf_palist *);
 100 void                     pf_empty_pool(struct pf_palist *);
 101 #ifdef ALTQ
 102 int                      pf_begin_altq(u_int32_t *);
 103 int                      pf_rollback_altq(u_int32_t);
 104 int                      pf_commit_altq(u_int32_t);
 105 int                      pf_enable_altq(struct pf_altq *);
 106 int                      pf_disable_altq(struct pf_altq *);
 107 #endif /* ALTQ */
 108 int                      pf_begin_rules(u_int32_t *, int, const char *);
 109 int                      pf_rollback_rules(u_int32_t, int, char *);
 110 int                      pf_setup_pfsync_matching(struct pf_ruleset *);
 111 void                     pf_hash_rule(MD5_CTX *, struct pf_rule *);
 112 void                     pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 113 int                      pf_commit_rules(u_int32_t, int, char *);
 114 int                      pf_addr_setup(struct pf_ruleset *,
 115                             struct pf_addr_wrap *, sa_family_t);
 116 void                     pf_addr_copyout(struct pf_addr_wrap *);
 117
 118 struct pf_rule           pf_default_rule;
 119 struct lock              pf_consistency_lock;
 120 struct lock              pf_global_statetbl_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         int nn;
 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         for (nn = 0; nn < ncpus; ++nn) {
 227                 RB_INIT(&tree_src_tracking[nn]);
 228                 RB_INIT(&tree_id[nn]);
 229         }
 230         RB_INIT(&pf_anchors);
 231         pf_init_ruleset(&pf_main_ruleset);
 232         TAILQ_INIT(&pf_altqs[0]);
 233         TAILQ_INIT(&pf_altqs[1]);
 234         TAILQ_INIT(&pf_pabuf);
 235         pf_altqs_active = &pf_altqs[0];
 236         pf_altqs_inactive = &pf_altqs[1];
 237         for (nn = 0; nn < ncpus; ++nn)
 238                 TAILQ_INIT(&state_list[nn]);
 239
 240         /* default rule should never be garbage collected */
 241         pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
 242         pf_default_rule.action = PF_PASS;
 243         pf_default_rule.nr = (uint32_t)(-1);
 244         pf_default_rule.rtableid = -1;
 245
 246         /* initialize default timeouts */
 247         my_timeout[PFTM_TCP_FIRST_PACKET] = 120;        /* First TCP packet */
 248         my_timeout[PFTM_TCP_OPENING] = 30;              /* No response yet */
 249         my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;    /* Established */
 250         my_timeout[PFTM_TCP_CLOSING] = 15 * 60;         /* Half closed */
 251         my_timeout[PFTM_TCP_FIN_WAIT] = 45;             /* Got both FINs */
 252         my_timeout[PFTM_TCP_CLOSED] = 90;               /* Got a RST */
 253         my_timeout[PFTM_UDP_FIRST_PACKET] = 60;         /* First UDP packet */
 254         my_timeout[PFTM_UDP_SINGLE] = 30;               /* Unidirectional */
 255         my_timeout[PFTM_UDP_MULTIPLE] = 60;             /* Bidirectional */
 256         my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;        /* First ICMP packet */
 257         my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;         /* Got error response */
 258         my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;       /* First packet */
 259         my_timeout[PFTM_OTHER_SINGLE] = 30;             /* Unidirectional */
 260         my_timeout[PFTM_OTHER_MULTIPLE] = 60;           /* Bidirectional */
 261         my_timeout[PFTM_FRAG] = 30;                     /* Fragment expire */
 262         my_timeout[PFTM_INTERVAL] = 10;                 /* Expire interval */
 263         my_timeout[PFTM_SRC_NODE] = 0;                  /* Source Tracking */
 264         my_timeout[PFTM_TS_DIFF] = 30;                  /* Allowed TS diff */
 265         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 266         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 267
 268         pf_normalize_init();
 269         bzero(&pf_status, sizeof(pf_status));
 270         pf_status.debug = PF_DEBUG_URGENT;
 271         /* XXX do our best to avoid a conflict */
 272         pf_status.hostid = karc4random();
 273
 274         if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
 275                 panic("pfpurge thread");
 276 }
 277
 278 int
 279 pfopen(struct dev_open_args *ap)
 280 {
 281         lwkt_gettoken(&pf_token);
 282         cdev_t dev = ap->a_head.a_dev;
 283         if (minor(dev) >= 1) {
 284                 lwkt_reltoken(&pf_token);
 285                 return (ENXIO);
 286         }
 287         lwkt_reltoken(&pf_token);
 288         return (0);
 289 }
 290
 291 int
 292 pfclose(struct dev_close_args *ap)
 293 {
 294         lwkt_gettoken(&pf_token);
 295         cdev_t dev = ap->a_head.a_dev;
 296         if (minor(dev) >= 1) {
 297                 lwkt_reltoken(&pf_token);
 298                 return (ENXIO);
 299         }
 300         lwkt_reltoken(&pf_token);
 301         return (0);
 302 }
 303
 304 struct pf_pool *
 305 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 306     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 307     u_int8_t check_ticket)
 308 {
 309         struct pf_ruleset       *ruleset;
 310         struct pf_rule          *rule;
 311         int                      rs_num;
 312
 313         ruleset = pf_find_ruleset(anchor);
 314         if (ruleset == NULL)
 315                 return (NULL);
 316         rs_num = pf_get_ruleset_number(rule_action);
 317         if (rs_num >= PF_RULESET_MAX)
 318                 return (NULL);
 319         if (active) {
 320                 if (check_ticket && ticket !=
 321                     ruleset->rules[rs_num].active.ticket)
 322                         return (NULL);
 323                 if (r_last)
 324                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 325                             pf_rulequeue);
 326                 else
 327                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 328         } else {
 329                 if (check_ticket && ticket !=
 330                     ruleset->rules[rs_num].inactive.ticket)
 331                         return (NULL);
 332                 if (r_last)
 333                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 334                             pf_rulequeue);
 335                 else
 336                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 337         }
 338         if (!r_last) {
 339                 while ((rule != NULL) && (rule->nr != rule_number))
 340                         rule = TAILQ_NEXT(rule, entries);
 341         }
 342         if (rule == NULL)
 343                 return (NULL);
 344
 345         return (&rule->rpool);
 346 }
 347
 348 void
 349 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 350 {
 351         struct pf_pooladdr      *mv_pool_pa;
 352
 353         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 354                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 355                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 356         }
 357 }
 358
 359 void
 360 pf_empty_pool(struct pf_palist *poola)
 361 {
 362         struct pf_pooladdr      *empty_pool_pa;
 363
 364         while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 365                 pfi_dynaddr_remove(&empty_pool_pa->addr);
 366                 pf_tbladdr_remove(&empty_pool_pa->addr);
 367                 pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
 368                 TAILQ_REMOVE(poola, empty_pool_pa, entries);
 369                 kfree(empty_pool_pa, M_PFPOOLADDRPL);
 370         }
 371 }
 372
 373 void
 374 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 375 {
 376         if (rulequeue != NULL) {
 377                 if (rule->states_cur <= 0) {
 378                         /*
 379                          * XXX - we need to remove the table *before* detaching
 380                          * the rule to make sure the table code does not delete
 381                          * the anchor under our feet.
 382                          */
 383                         pf_tbladdr_remove(&rule->src.addr);
 384                         pf_tbladdr_remove(&rule->dst.addr);
 385                         if (rule->overload_tbl)
 386                                 pfr_detach_table(rule->overload_tbl);
 387                 }
 388                 TAILQ_REMOVE(rulequeue, rule, entries);
 389                 rule->entries.tqe_prev = NULL;
 390                 rule->nr = -1;
 391         }
 392
 393         if (rule->states_cur > 0 || rule->src_nodes > 0 ||
 394             rule->entries.tqe_prev != NULL)
 395                 return;
 396         pf_tag_unref(rule->tag);
 397         pf_tag_unref(rule->match_tag);
 398 #ifdef ALTQ
 399         if (rule->pqid != rule->qid)
 400                 pf_qid_unref(rule->pqid);
 401         pf_qid_unref(rule->qid);
 402 #endif
 403         pf_rtlabel_remove(&rule->src.addr);
 404         pf_rtlabel_remove(&rule->dst.addr);
 405         pfi_dynaddr_remove(&rule->src.addr);
 406         pfi_dynaddr_remove(&rule->dst.addr);
 407         if (rulequeue == NULL) {
 408                 pf_tbladdr_remove(&rule->src.addr);
 409                 pf_tbladdr_remove(&rule->dst.addr);
 410                 if (rule->overload_tbl)
 411                         pfr_detach_table(rule->overload_tbl);
 412         }
 413         pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
 414         pf_anchor_remove(rule);
 415         pf_empty_pool(&rule->rpool.list);
 416         kfree(rule, M_PFRULEPL);
 417 }
 418
 419 u_int16_t
 420 tagname2tag(struct pf_tags *head, char *tagname)
 421 {
 422         struct pf_tagname       *tag, *p = NULL;
 423         u_int16_t                new_tagid = 1;
 424
 425         TAILQ_FOREACH(tag, head, entries)
 426                 if (strcmp(tagname, tag->name) == 0) {
 427                         tag->ref++;
 428                         return (tag->tag);
 429                 }
 430
 431         /*
 432          * to avoid fragmentation, we do a linear search from the beginning
 433          * and take the first free slot we find. if there is none or the list
 434          * is empty, append a new entry at the end.
 435          */
 436
 437         /* new entry */
 438         if (!TAILQ_EMPTY(head))
 439                 for (p = TAILQ_FIRST(head); p != NULL &&
 440                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 441                         new_tagid = p->tag + 1;
 442
 443         if (new_tagid > TAGID_MAX)
 444                 return (0);
 445
 446         /* allocate and fill new struct pf_tagname */
 447         tag = kmalloc(sizeof(*tag), M_TEMP, M_WAITOK);
 448         strlcpy(tag->name, tagname, sizeof(tag->name));
 449         tag->tag = new_tagid;
 450         tag->ref++;
 451
 452         if (p != NULL)  /* insert new entry before p */
 453                 TAILQ_INSERT_BEFORE(p, tag, entries);
 454         else    /* either list empty or no free slot in between */
 455                 TAILQ_INSERT_TAIL(head, tag, entries);
 456
 457         return (tag->tag);
 458 }
 459
 460 void
 461 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
 462 {
 463         struct pf_tagname       *tag;
 464
 465         TAILQ_FOREACH(tag, head, entries)
 466                 if (tag->tag == tagid) {
 467                         strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
 468                         return;
 469                 }
 470 }
 471
 472 void
 473 tag_unref(struct pf_tags *head, u_int16_t tag)
 474 {
 475         struct pf_tagname       *p, *next;
 476
 477         if (tag == 0)
 478                 return;
 479
 480         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 481                 next = TAILQ_NEXT(p, entries);
 482                 if (tag == p->tag) {
 483                         if (--p->ref == 0) {
 484                                 TAILQ_REMOVE(head, p, entries);
 485                                 kfree(p, M_TEMP);
 486                         }
 487                         break;
 488                 }
 489         }
 490 }
 491
 492 u_int16_t
 493 pf_tagname2tag(char *tagname)
 494 {
 495         return (tagname2tag(&pf_tags, tagname));
 496 }
 497
 498 void
 499 pf_tag2tagname(u_int16_t tagid, char *p)
 500 {
 501         tag2tagname(&pf_tags, tagid, p);
 502 }
 503
 504 void
 505 pf_tag_ref(u_int16_t tag)
 506 {
 507         struct pf_tagname *t;
 508
 509         TAILQ_FOREACH(t, &pf_tags, entries)
 510                 if (t->tag == tag)
 511                         break;
 512         if (t != NULL)
 513                 t->ref++;
 514 }
 515
 516 void
 517 pf_tag_unref(u_int16_t tag)
 518 {
 519         tag_unref(&pf_tags, tag);
 520 }
 521
 522 int
 523 pf_rtlabel_add(struct pf_addr_wrap *a)
 524 {
 525         return (0);
 526 }
 527
 528 void
 529 pf_rtlabel_remove(struct pf_addr_wrap *a)
 530 {
 531 }
 532
 533 void
 534 pf_rtlabel_copyout(struct pf_addr_wrap *a)
 535 {
 536         if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
 537                 strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
 538 }
 539
 540 #ifdef ALTQ
 541 u_int32_t
 542 pf_qname2qid(char *qname)
 543 {
 544         return ((u_int32_t)tagname2tag(&pf_qids, qname));
 545 }
 546
 547 void
 548 pf_qid2qname(u_int32_t qid, char *p)
 549 {
 550         tag2tagname(&pf_qids, (u_int16_t)qid, p);
 551 }
 552
 553 void
 554 pf_qid_unref(u_int32_t qid)
 555 {
 556         tag_unref(&pf_qids, (u_int16_t)qid);
 557 }
 558
 559 int
 560 pf_begin_altq(u_int32_t *ticket)
 561 {
 562         struct pf_altq  *altq;
 563         int              error = 0;
 564
 565         /* Purge the old altq list */
 566         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 567                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 568                 if (altq->qname[0] == 0 &&
 569                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 570                         /* detach and destroy the discipline */
 571                         error = altq_remove(altq);
 572                 } else
 573                         pf_qid_unref(altq->qid);
 574                 kfree(altq, M_PFALTQPL);
 575         }
 576         if (error)
 577                 return (error);
 578         *ticket = ++ticket_altqs_inactive;
 579         altqs_inactive_open = 1;
 580         return (0);
 581 }
 582
 583 int
 584 pf_rollback_altq(u_int32_t ticket)
 585 {
 586         struct pf_altq  *altq;
 587         int              error = 0;
 588
 589         if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
 590                 return (0);
 591         /* Purge the old altq list */
 592         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 593                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 594                 if (altq->qname[0] == 0 &&
 595                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 596                         /* detach and destroy the discipline */
 597                         error = altq_remove(altq);
 598                 } else
 599                         pf_qid_unref(altq->qid);
 600                 kfree(altq, M_PFALTQPL);
 601         }
 602         altqs_inactive_open = 0;
 603         return (error);
 604 }
 605
 606 int
 607 pf_commit_altq(u_int32_t ticket)
 608 {
 609         struct pf_altqqueue     *old_altqs;
 610         struct pf_altq          *altq;
 611         int                      err, error = 0;
 612
 613         if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
 614                 return (EBUSY);
 615
 616         /* swap altqs, keep the old. */
 617         crit_enter();
 618         old_altqs = pf_altqs_active;
 619         pf_altqs_active = pf_altqs_inactive;
 620         pf_altqs_inactive = old_altqs;
 621         ticket_altqs_active = ticket_altqs_inactive;
 622
 623         /* Attach new disciplines */
 624         TAILQ_FOREACH(altq, pf_altqs_active, entries) {
 625                 if (altq->qname[0] == 0 &&
 626                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 627                         /* attach the discipline */
 628                         error = altq_pfattach(altq);
 629                         if (error) {
 630                                 crit_exit();
 631                                 return (error);
 632                         }
 633                 }
 634         }
 635
 636         /* Purge the old altq list */
 637         while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
 638                 TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
 639                 if (altq->qname[0] == 0 &&
 640                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 641                         /* detach and destroy the discipline */
 642                         if (pf_altq_running)
 643                                 error = pf_disable_altq(altq);
 644                         err = altq_pfdetach(altq);
 645                         if (err != 0 && error == 0)
 646                                 error = err;
 647                         err = altq_remove(altq);
 648                         if (err != 0 && error == 0)
 649                                 error = err;
 650                 } else
 651                         pf_qid_unref(altq->qid);
 652                 kfree(altq, M_PFALTQPL);
 653         }
 654         crit_exit();
 655
 656         altqs_inactive_open = 0;
 657         return (error);
 658 }
 659
 660 int
 661 pf_enable_altq(struct pf_altq *altq)
 662 {
 663         struct ifnet            *ifp;
 664         struct tb_profile        tb;
 665         int                      error = 0;
 666
 667         ifnet_lock();
 668         ifp = ifunit(altq->ifname);
 669         ifnet_unlock();
 670
 671         if (ifp == NULL)
 672                 return (EINVAL);
 673
 674         if (ifp->if_snd.altq_type != ALTQT_NONE)
 675                 error = altq_enable(&ifp->if_snd);
 676
 677         /* set tokenbucket regulator */
 678         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 679                 tb.rate = altq->ifbandwidth;
 680                 tb.depth = altq->tbrsize;
 681                 crit_enter();
 682                 error = tbr_set(&ifp->if_snd, &tb);
 683                 crit_exit();
 684         }
 685
 686         return (error);
 687 }
 688
 689 int
 690 pf_disable_altq(struct pf_altq *altq)
 691 {
 692         struct ifnet            *ifp;
 693         struct tb_profile        tb;
 694         int                      error;
 695
 696         ifnet_lock();
 697         ifp = ifunit(altq->ifname);
 698         ifnet_unlock();
 699
 700         if (ifp == NULL)
 701                 return (EINVAL);
 702
 703         /*
 704          * when the discipline is no longer referenced, it was overridden
 705          * by a new one.  if so, just return.
 706          */
 707         if (altq->altq_disc != ifp->if_snd.altq_disc)
 708                 return (0);
 709
 710         error = altq_disable(&ifp->if_snd);
 711
 712         if (error == 0) {
 713                 /* clear tokenbucket regulator */
 714                 tb.rate = 0;
 715                 crit_enter();
 716                 error = tbr_set(&ifp->if_snd, &tb);
 717                 crit_exit();
 718         }
 719
 720         return (error);
 721 }
 722
 723 void
 724 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 725 {
 726         struct ifnet    *ifp1;
 727         struct pf_altq  *a1, *a2, *a3;
 728         u_int32_t        ticket;
 729         int              error = 0;
 730
 731         /* Interrupt userland queue modifications */
 732         if (altqs_inactive_open)
 733                 pf_rollback_altq(ticket_altqs_inactive);
 734
 735         /* Start new altq ruleset */
 736         if (pf_begin_altq(&ticket))
 737                 return;
 738
 739         /* Copy the current active set */
 740         TAILQ_FOREACH(a1, pf_altqs_active, entries) {
 741                 a2 = kmalloc(sizeof(*a2), M_PFALTQPL, M_INTWAIT);
 742                 if (a2 == NULL) {
 743                         error = ENOMEM;
 744                         break;
 745                 }
 746                 bcopy(a1, a2, sizeof(struct pf_altq));
 747
 748                 if (a2->qname[0] != 0) {
 749                         if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 750                                 error = EBUSY;
 751                                 kfree(a2, M_PFALTQPL);
 752                                 break;
 753                         }
 754                         a2->altq_disc = NULL;
 755                         TAILQ_FOREACH(a3, pf_altqs_inactive, entries) {
 756                                 if (strncmp(a3->ifname, a2->ifname,
 757                                     IFNAMSIZ) == 0 && a3->qname[0] == 0) {
 758                                         a2->altq_disc = a3->altq_disc;
 759                                         break;
 760                                 }
 761                         }
 762                 }
 763                 /* Deactivate the interface in question */
 764                 a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 765                 ifnet_lock();
 766                 ifp1 = ifunit(a2->ifname);
 767                 ifnet_unlock();
 768                 if ((ifp1 == NULL) || (remove && ifp1 == ifp)) {
 769                         a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 770                 } else {
 771                         error = altq_add(a2);
 772
 773                         if (ticket != ticket_altqs_inactive)
 774                                 error = EBUSY;
 775
 776                         if (error) {
 777                                 kfree(a2, M_PFALTQPL);
 778                                 break;
 779                         }
 780                 }
 781
 782                 TAILQ_INSERT_TAIL(pf_altqs_inactive, a2, entries);
 783         }
 784
 785         if (error != 0)
 786                 pf_rollback_altq(ticket);
 787         else
 788                 pf_commit_altq(ticket);
 789 }
 790 #endif /* ALTQ */
 791
 792 int
 793 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 794 {
 795         struct pf_ruleset       *rs;
 796         struct pf_rule          *rule;
 797
 798         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 799                 return (EINVAL);
 800         rs = pf_find_or_create_ruleset(anchor);
 801         if (rs == NULL)
 802                 return (EINVAL);
 803         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 804                 pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
 805                 rs->rules[rs_num].inactive.rcount--;
 806         }
 807         *ticket = ++rs->rules[rs_num].inactive.ticket;
 808         rs->rules[rs_num].inactive.open = 1;
 809         return (0);
 810 }
 811
 812 int
 813 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 814 {
 815         struct pf_ruleset       *rs;
 816         struct pf_rule          *rule;
 817
 818         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 819                 return (EINVAL);
 820         rs = pf_find_ruleset(anchor);
 821         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 822             rs->rules[rs_num].inactive.ticket != ticket)
 823                 return (0);
 824         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 825                 pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
 826                 rs->rules[rs_num].inactive.rcount--;
 827         }
 828         rs->rules[rs_num].inactive.open = 0;
 829         return (0);
 830 }
 831
 832 #define PF_MD5_UPD(st, elm)                                             \
 833                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 834
 835 #define PF_MD5_UPD_STR(st, elm)                                         \
 836                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 837
 838 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 839                 (stor) = htonl((st)->elm);                              \
 840                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 841 } while (0)
 842
 843 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 844                 (stor) = htons((st)->elm);                              \
 845                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 846 } while (0)
 847
 848 void
 849 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 850 {
 851         PF_MD5_UPD(pfr, addr.type);
 852         switch (pfr->addr.type) {
 853                 case PF_ADDR_DYNIFTL:
 854                         PF_MD5_UPD(pfr, addr.v.ifname);
 855                         PF_MD5_UPD(pfr, addr.iflags);
 856                         break;
 857                 case PF_ADDR_TABLE:
 858                         PF_MD5_UPD(pfr, addr.v.tblname);
 859                         break;
 860                 case PF_ADDR_ADDRMASK:
 861                         /* XXX ignore af? */
 862                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 863                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 864                         break;
 865                 case PF_ADDR_RTLABEL:
 866                         PF_MD5_UPD(pfr, addr.v.rtlabelname);
 867                         break;
 868         }
 869
 870         PF_MD5_UPD(pfr, port[0]);
 871         PF_MD5_UPD(pfr, port[1]);
 872         PF_MD5_UPD(pfr, neg);
 873         PF_MD5_UPD(pfr, port_op);
 874 }
 875
 876 void
 877 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 878 {
 879         u_int16_t x;
 880         u_int32_t y;
 881
 882         pf_hash_rule_addr(ctx, &rule->src);
 883         pf_hash_rule_addr(ctx, &rule->dst);
 884         PF_MD5_UPD_STR(rule, label);
 885         PF_MD5_UPD_STR(rule, ifname);
 886         PF_MD5_UPD_STR(rule, match_tagname);
 887         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 888         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 889         PF_MD5_UPD_HTONL(rule, prob, y);
 890         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 891         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 892         PF_MD5_UPD(rule, uid.op);
 893         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 894         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 895         PF_MD5_UPD(rule, gid.op);
 896         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 897         PF_MD5_UPD(rule, action);
 898         PF_MD5_UPD(rule, direction);
 899         PF_MD5_UPD(rule, af);
 900         PF_MD5_UPD(rule, quick);
 901         PF_MD5_UPD(rule, ifnot);
 902         PF_MD5_UPD(rule, match_tag_not);
 903         PF_MD5_UPD(rule, natpass);
 904         PF_MD5_UPD(rule, keep_state);
 905         PF_MD5_UPD(rule, proto);
 906         PF_MD5_UPD(rule, type);
 907         PF_MD5_UPD(rule, code);
 908         PF_MD5_UPD(rule, flags);
 909         PF_MD5_UPD(rule, flagset);
 910         PF_MD5_UPD(rule, allow_opts);
 911         PF_MD5_UPD(rule, rt);
 912         PF_MD5_UPD(rule, tos);
 913 }
 914
 915 int
 916 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 917 {
 918         struct pf_ruleset       *rs;
 919         struct pf_rule          *rule, **old_array;
 920         struct pf_rulequeue     *old_rules;
 921         int                      error;
 922         u_int32_t                old_rcount;
 923
 924         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 925                 return (EINVAL);
 926         rs = pf_find_ruleset(anchor);
 927         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 928             ticket != rs->rules[rs_num].inactive.ticket)
 929                 return (EBUSY);
 930
 931         /* Calculate checksum for the main ruleset */
 932         if (rs == &pf_main_ruleset) {
 933                 error = pf_setup_pfsync_matching(rs);
 934                 if (error != 0)
 935                         return (error);
 936         }
 937
 938         /* Swap rules, keep the old. */
 939         crit_enter();
 940         old_rules = rs->rules[rs_num].active.ptr;
 941         old_rcount = rs->rules[rs_num].active.rcount;
 942         old_array = rs->rules[rs_num].active.ptr_array;
 943
 944         rs->rules[rs_num].active.ptr =
 945             rs->rules[rs_num].inactive.ptr;
 946         rs->rules[rs_num].active.ptr_array =
 947             rs->rules[rs_num].inactive.ptr_array;
 948         rs->rules[rs_num].active.rcount =
 949             rs->rules[rs_num].inactive.rcount;
 950         rs->rules[rs_num].inactive.ptr = old_rules;
 951         rs->rules[rs_num].inactive.ptr_array = old_array;
 952         rs->rules[rs_num].inactive.rcount = old_rcount;
 953
 954         rs->rules[rs_num].active.ticket =
 955             rs->rules[rs_num].inactive.ticket;
 956         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 957
 958
 959         /* Purge the old rule list. */
 960         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 961                 pf_rm_rule(old_rules, rule);
 962         if (rs->rules[rs_num].inactive.ptr_array)
 963                 kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 964         rs->rules[rs_num].inactive.ptr_array = NULL;
 965         rs->rules[rs_num].inactive.rcount = 0;
 966         rs->rules[rs_num].inactive.open = 0;
 967         pf_remove_if_empty_ruleset(rs);
 968         crit_exit();
 969         return (0);
 970 }
 971
 972 int
 973 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 974 {
 975         MD5_CTX                  ctx;
 976         struct pf_rule          *rule;
 977         int                      rs_cnt;
 978         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 979
 980         MD5Init(&ctx);
 981         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 982                 /* XXX PF_RULESET_SCRUB as well? */
 983                 if (rs_cnt == PF_RULESET_SCRUB)
 984                         continue;
 985
 986                 if (rs->rules[rs_cnt].inactive.ptr_array)
 987                         kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 988                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 989
 990                 if (rs->rules[rs_cnt].inactive.rcount) {
 991                         rs->rules[rs_cnt].inactive.ptr_array =
 992                             kmalloc(sizeof(caddr_t) *
 993                                     rs->rules[rs_cnt].inactive.rcount,
 994                                     M_TEMP, M_WAITOK);
 995                 }
 996
 997                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 998                     entries) {
 999                         pf_hash_rule(&ctx, rule);
1000                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1001                 }
1002         }
1003
1004         MD5Final(digest, &ctx);
1005         memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
1006         return (0);
1007 }
1008
1009 int
1010 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
1011     sa_family_t af)
1012 {
1013         if (pfi_dynaddr_setup(addr, af) ||
1014             pf_tbladdr_setup(ruleset, addr))
1015                 return (EINVAL);
1016
1017         return (0);
1018 }
1019
1020 void
1021 pf_addr_copyout(struct pf_addr_wrap *addr)
1022 {
1023         pfi_dynaddr_copyout(addr);
1024         pf_tbladdr_copyout(addr);
1025         pf_rtlabel_copyout(addr);
1026 }
1027
1028 int
1029 pfioctl(struct dev_ioctl_args *ap)
1030 {
1031         u_long cmd = ap->a_cmd;
1032         caddr_t addr = ap->a_data;
1033         struct pf_pooladdr      *pa = NULL;
1034         struct pf_pool          *pool = NULL;
1035         int                      error = 0;
1036
1037         lwkt_gettoken(&pf_token);
1038
1039         /* XXX keep in sync with switch() below */
1040         if (securelevel > 1) {
1041                 switch (cmd) {
1042                 case DIOCGETRULES:
1043                 case DIOCGETRULE:
1044                 case DIOCGETADDRS:
1045                 case DIOCGETADDR:
1046                 case DIOCGETSTATE:
1047                 case DIOCSETSTATUSIF:
1048                 case DIOCGETSTATUS:
1049                 case DIOCCLRSTATUS:
1050                 case DIOCNATLOOK:
1051                 case DIOCSETDEBUG:
1052                 case DIOCGETSTATES:
1053                 case DIOCGETTIMEOUT:
1054                 case DIOCCLRRULECTRS:
1055                 case DIOCGETLIMIT:
1056                 case DIOCGETALTQS:
1057                 case DIOCGETALTQ:
1058                 case DIOCGETQSTATS:
1059                 case DIOCGETRULESETS:
1060                 case DIOCGETRULESET:
1061                 case DIOCRGETTABLES:
1062                 case DIOCRGETTSTATS:
1063                 case DIOCRCLRTSTATS:
1064                 case DIOCRCLRADDRS:
1065                 case DIOCRADDADDRS:
1066                 case DIOCRDELADDRS:
1067                 case DIOCRSETADDRS:
1068                 case DIOCRGETADDRS:
1069                 case DIOCRGETASTATS:
1070                 case DIOCRCLRASTATS:
1071                 case DIOCRTSTADDRS:
1072                 case DIOCOSFPGET:
1073                 case DIOCGETSRCNODES:
1074                 case DIOCCLRSRCNODES:
1075                 case DIOCIGETIFACES:
1076                 case DIOCSETIFFLAG:
1077                 case DIOCCLRIFFLAG:
1078                 case DIOCGIFSPEED:
1079                         break;
1080                 case DIOCRCLRTABLES:
1081                 case DIOCRADDTABLES:
1082                 case DIOCRDELTABLES:
1083                 case DIOCRSETTFLAGS:
1084                         if (((struct pfioc_table *)addr)->pfrio_flags &
1085                             PFR_FLAG_DUMMY)
1086                                 break; /* dummy operation ok */
1087                         lwkt_reltoken(&pf_token);
1088                         return (EPERM);
1089                 default:
1090                         lwkt_reltoken(&pf_token);
1091                         return (EPERM);
1092                 }
1093         }
1094
1095         if (!(ap->a_fflag & FWRITE)) {
1096                 switch (cmd) {
1097                 case DIOCGETRULES:
1098                 case DIOCGETADDRS:
1099                 case DIOCGETADDR:
1100                 case DIOCGETSTATE:
1101                 case DIOCGETSTATUS:
1102                 case DIOCGETSTATES:
1103                 case DIOCGETTIMEOUT:
1104                 case DIOCGETLIMIT:
1105                 case DIOCGETALTQS:
1106                 case DIOCGETALTQ:
1107                 case DIOCGETQSTATS:
1108                 case DIOCGETRULESETS:
1109                 case DIOCGETRULESET:
1110                 case DIOCNATLOOK:
1111                 case DIOCRGETTABLES:
1112                 case DIOCRGETTSTATS:
1113                 case DIOCRGETADDRS:
1114                 case DIOCRGETASTATS:
1115                 case DIOCRTSTADDRS:
1116                 case DIOCOSFPGET:
1117                 case DIOCGETSRCNODES:
1118                 case DIOCIGETIFACES:
1119                 case DIOCGIFSPEED:
1120                         break;
1121                 case DIOCRCLRTABLES:
1122                 case DIOCRADDTABLES:
1123                 case DIOCRDELTABLES:
1124                 case DIOCRCLRTSTATS:
1125                 case DIOCRCLRADDRS:
1126                 case DIOCRADDADDRS:
1127                 case DIOCRDELADDRS:
1128                 case DIOCRSETADDRS:
1129                 case DIOCRSETTFLAGS:
1130                         if (((struct pfioc_table *)addr)->pfrio_flags &
1131                             PFR_FLAG_DUMMY)
1132                                 break; /* dummy operation ok */
1133                         lwkt_reltoken(&pf_token);
1134                         return (EACCES);
1135                 case DIOCGETRULE:
1136                         if (((struct pfioc_rule *)addr)->action ==
1137                             PF_GET_CLR_CNTR) {
1138                                 lwkt_reltoken(&pf_token);
1139                                 return (EACCES);
1140                         }
1141                         break;
1142                 default:
1143                         lwkt_reltoken(&pf_token);
1144                         return (EACCES);
1145                 }
1146         }
1147
1148         switch (cmd) {
1149         case DIOCSTART:
1150                 if (pf_status.running)
1151                         error = EEXIST;
1152                 else {
1153                         error = hook_pf();
1154                         if (error) {
1155                                 DPFPRINTF(PF_DEBUG_MISC,
1156                                     ("pf: pfil registration fail\n"));
1157                                 break;
1158                         }
1159                         pf_status.running = 1;
1160                         pf_status.since = time_second;
1161                         if (pf_status.stateid == 0) {
1162                                 pf_status.stateid = time_second;
1163                                 pf_status.stateid = pf_status.stateid << 32;
1164                         }
1165                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1166                 }
1167                 break;
1168
1169         case DIOCSTOP:
1170                 if (!pf_status.running)
1171                         error = ENOENT;
1172                 else {
1173                         pf_status.running = 0;
1174                         error = dehook_pf();
1175                         if (error) {
1176                                 pf_status.running = 1;
1177                                 DPFPRINTF(PF_DEBUG_MISC,
1178                                         ("pf: pfil unregistration failed\n"));
1179                         }
1180                         pf_status.since = time_second;
1181                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1182                 }
1183                 break;
1184
1185         case DIOCADDRULE: {
1186                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1187                 struct pf_ruleset       *ruleset;
1188                 struct pf_rule          *rule, *tail;
1189                 struct pf_pooladdr      *pa;
1190                 int                      rs_num;
1191
1192                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1193                 ruleset = pf_find_ruleset(pr->anchor);
1194                 if (ruleset == NULL) {
1195                         error = EINVAL;
1196                         break;
1197                 }
1198                 rs_num = pf_get_ruleset_number(pr->rule.action);
1199                 if (rs_num >= PF_RULESET_MAX) {
1200                         error = EINVAL;
1201                         break;
1202                 }
1203                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1204                         error = EINVAL;
1205                         break;
1206                 }
1207                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1208                         error = EBUSY;
1209                         break;
1210                 }
1211                 if (pr->pool_ticket != ticket_pabuf) {
1212                         error = EBUSY;
1213                         break;
1214                 }
1215                 rule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK);
1216                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1217                 rule->cuid = ap->a_cred->cr_ruid;
1218                 rule->cpid = 0;
1219                 rule->anchor = NULL;
1220                 rule->kif = NULL;
1221                 TAILQ_INIT(&rule->rpool.list);
1222                 /* initialize refcounting */
1223                 rule->states_cur = 0;
1224                 rule->src_nodes = 0;
1225                 rule->entries.tqe_prev = NULL;
1226 #ifndef INET
1227                 if (rule->af == AF_INET) {
1228                         kfree(rule, M_PFRULEPL);
1229                         error = EAFNOSUPPORT;
1230                         break;
1231                 }
1232 #endif /* INET */
1233 #ifndef INET6
1234                 if (rule->af == AF_INET6) {
1235                         kfree(rule, M_PFRULEPL);
1236                         error = EAFNOSUPPORT;
1237                         break;
1238                 }
1239 #endif /* INET6 */
1240                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1241                     pf_rulequeue);
1242                 if (tail)
1243                         rule->nr = tail->nr + 1;
1244                 else
1245                         rule->nr = 0;
1246                 if (rule->ifname[0]) {
1247                         rule->kif = pfi_kif_get(rule->ifname);
1248                         if (rule->kif == NULL) {
1249                                 kfree(rule, M_PFRULEPL);
1250                                 error = EINVAL;
1251                                 break;
1252                         }
1253                         pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1254                 }
1255
1256                 if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1257                         error = EBUSY;
1258
1259 #ifdef ALTQ
1260                 /* set queue IDs */
1261                 if (rule->qname[0] != 0) {
1262                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1263                                 error = EBUSY;
1264                         else if (rule->pqname[0] != 0) {
1265                                 if ((rule->pqid =
1266                                     pf_qname2qid(rule->pqname)) == 0)
1267                                         error = EBUSY;
1268                         } else
1269                                 rule->pqid = rule->qid;
1270                 }
1271 #endif
1272                 if (rule->tagname[0])
1273                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1274                                 error = EBUSY;
1275                 if (rule->match_tagname[0])
1276                         if ((rule->match_tag =
1277                             pf_tagname2tag(rule->match_tagname)) == 0)
1278                                 error = EBUSY;
1279                 if (rule->rt && !rule->direction)
1280                         error = EINVAL;
1281 #if NPFLOG > 0
1282                 if (!rule->log)
1283                         rule->logif = 0;
1284                 if (rule->logif >= PFLOGIFS_MAX)
1285                         error = EINVAL;
1286 #endif
1287                 if (pf_rtlabel_add(&rule->src.addr) ||
1288                     pf_rtlabel_add(&rule->dst.addr))
1289                         error = EBUSY;
1290                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1291                         error = EINVAL;
1292                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1293                         error = EINVAL;
1294                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1295                         error = EINVAL;
1296                 TAILQ_FOREACH(pa, &pf_pabuf, entries)
1297                         if (pf_tbladdr_setup(ruleset, &pa->addr))
1298                                 error = EINVAL;
1299
1300                 if (rule->overload_tblname[0]) {
1301                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1302                             rule->overload_tblname)) == NULL)
1303                                 error = EINVAL;
1304                         else
1305                                 rule->overload_tbl->pfrkt_flags |=
1306                                     PFR_TFLAG_ACTIVE;
1307                 }
1308
1309                 pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1310                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1311                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1312                     (rule->rt > PF_FASTROUTE)) &&
1313                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1314                         error = EINVAL;
1315
1316                 if (error) {
1317                         pf_rm_rule(NULL, rule);
1318                         break;
1319                 }
1320                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1321                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1322                     rule->bytes[0] = rule->bytes[1] = 0;
1323                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1324                     rule, entries);
1325                 ruleset->rules[rs_num].inactive.rcount++;
1326                 break;
1327         }
1328
1329         case DIOCGETRULES: {
1330                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1331                 struct pf_ruleset       *ruleset;
1332                 struct pf_rule          *tail;
1333                 int                      rs_num;
1334
1335                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1336                 ruleset = pf_find_ruleset(pr->anchor);
1337                 if (ruleset == NULL) {
1338                         error = EINVAL;
1339                         break;
1340                 }
1341                 rs_num = pf_get_ruleset_number(pr->rule.action);
1342                 if (rs_num >= PF_RULESET_MAX) {
1343                         error = EINVAL;
1344                         break;
1345                 }
1346                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1347                     pf_rulequeue);
1348                 if (tail)
1349                         pr->nr = tail->nr + 1;
1350                 else
1351                         pr->nr = 0;
1352                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1353                 break;
1354         }
1355
1356         case DIOCGETRULE: {
1357                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1358                 struct pf_ruleset       *ruleset;
1359                 struct pf_rule          *rule;
1360                 int                      rs_num, i;
1361
1362                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1363                 ruleset = pf_find_ruleset(pr->anchor);
1364                 if (ruleset == NULL) {
1365                         error = EINVAL;
1366                         break;
1367                 }
1368                 rs_num = pf_get_ruleset_number(pr->rule.action);
1369                 if (rs_num >= PF_RULESET_MAX) {
1370                         error = EINVAL;
1371                         break;
1372                 }
1373                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1374                         error = EBUSY;
1375                         break;
1376                 }
1377                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1378                 while ((rule != NULL) && (rule->nr != pr->nr))
1379                         rule = TAILQ_NEXT(rule, entries);
1380                 if (rule == NULL) {
1381                         error = EBUSY;
1382                         break;
1383                 }
1384                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1385                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1386                         error = EBUSY;
1387                         break;
1388                 }
1389                 pf_addr_copyout(&pr->rule.src.addr);
1390                 pf_addr_copyout(&pr->rule.dst.addr);
1391                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1392                         if (rule->skip[i].ptr == NULL)
1393                                 pr->rule.skip[i].nr = (uint32_t)(-1);
1394                         else
1395                                 pr->rule.skip[i].nr =
1396                                     rule->skip[i].ptr->nr;
1397
1398                 if (pr->action == PF_GET_CLR_CNTR) {
1399                         rule->evaluations = 0;
1400                         rule->packets[0] = rule->packets[1] = 0;
1401                         rule->bytes[0] = rule->bytes[1] = 0;
1402                         rule->states_tot = 0;
1403                 }
1404                 break;
1405         }
1406
1407         case DIOCCHANGERULE: {
1408                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1409                 struct pf_ruleset       *ruleset;
1410                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1411                 u_int32_t                nr = 0;
1412                 int                      rs_num;
1413
1414                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1415                     pcr->action == PF_CHANGE_GET_TICKET) &&
1416                     pcr->pool_ticket != ticket_pabuf) {
1417                         error = EBUSY;
1418                         break;
1419                 }
1420
1421                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1422                     pcr->action > PF_CHANGE_GET_TICKET) {
1423                         error = EINVAL;
1424                         break;
1425                 }
1426                 ruleset = pf_find_ruleset(pcr->anchor);
1427                 if (ruleset == NULL) {
1428                         error = EINVAL;
1429                         break;
1430                 }
1431                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1432                 if (rs_num >= PF_RULESET_MAX) {
1433                         error = EINVAL;
1434                         break;
1435                 }
1436
1437                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1438                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1439                         break;
1440                 } else {
1441                         if (pcr->ticket !=
1442                             ruleset->rules[rs_num].active.ticket) {
1443                                 error = EINVAL;
1444                                 break;
1445                         }
1446                         if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1447                                 error = EINVAL;
1448                                 break;
1449                         }
1450                 }
1451
1452                 if (pcr->action != PF_CHANGE_REMOVE) {
1453                         newrule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK|M_NULLOK);
1454                         if (newrule == NULL) {
1455                                 error = ENOMEM;
1456                                 break;
1457                         }
1458                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1459                         newrule->cuid = ap->a_cred->cr_ruid;
1460                         newrule->cpid = 0;
1461                         TAILQ_INIT(&newrule->rpool.list);
1462                         /* initialize refcounting */
1463                         newrule->states_cur = 0;
1464                         newrule->entries.tqe_prev = NULL;
1465 #ifndef INET
1466                         if (newrule->af == AF_INET) {
1467                                 kfree(newrule, M_PFRULEPL);
1468                                 error = EAFNOSUPPORT;
1469                                 break;
1470                         }
1471 #endif /* INET */
1472 #ifndef INET6
1473                         if (newrule->af == AF_INET6) {
1474                                 kfree(newrule, M_PFRULEPL);
1475                                 error = EAFNOSUPPORT;
1476                                 break;
1477                         }
1478 #endif /* INET6 */
1479                         if (newrule->ifname[0]) {
1480                                 newrule->kif = pfi_kif_get(newrule->ifname);
1481                                 if (newrule->kif == NULL) {
1482                                         kfree(newrule, M_PFRULEPL);
1483                                         error = EINVAL;
1484                                         break;
1485                                 }
1486                                 pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1487                         } else
1488                                 newrule->kif = NULL;
1489
1490                         if (newrule->rtableid > 0 &&
1491                             newrule->rtableid > rt_numfibs)
1492                                 error = EBUSY;
1493
1494 #ifdef ALTQ
1495                         /* set queue IDs */
1496                         if (newrule->qname[0] != 0) {
1497                                 if ((newrule->qid =
1498                                     pf_qname2qid(newrule->qname)) == 0)
1499                                         error = EBUSY;
1500                                 else if (newrule->pqname[0] != 0) {
1501                                         if ((newrule->pqid =
1502                                             pf_qname2qid(newrule->pqname)) == 0)
1503                                                 error = EBUSY;
1504                                 } else
1505                                         newrule->pqid = newrule->qid;
1506                         }
1507 #endif /* ALTQ */
1508                         if (newrule->tagname[0])
1509                                 if ((newrule->tag =
1510                                     pf_tagname2tag(newrule->tagname)) == 0)
1511                                         error = EBUSY;
1512                         if (newrule->match_tagname[0])
1513                                 if ((newrule->match_tag = pf_tagname2tag(
1514                                     newrule->match_tagname)) == 0)
1515                                         error = EBUSY;
1516                         if (newrule->rt && !newrule->direction)
1517                                 error = EINVAL;
1518 #if NPFLOG > 0
1519                         if (!newrule->log)
1520                                 newrule->logif = 0;
1521                         if (newrule->logif >= PFLOGIFS_MAX)
1522                                 error = EINVAL;
1523 #endif
1524                         if (pf_rtlabel_add(&newrule->src.addr) ||
1525                             pf_rtlabel_add(&newrule->dst.addr))
1526                                 error = EBUSY;
1527                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1528                                 error = EINVAL;
1529                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1530                                 error = EINVAL;
1531                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1532                                 error = EINVAL;
1533                         TAILQ_FOREACH(pa, &pf_pabuf, entries)
1534                                 if (pf_tbladdr_setup(ruleset, &pa->addr))
1535                                         error = EINVAL;
1536
1537                         if (newrule->overload_tblname[0]) {
1538                                 if ((newrule->overload_tbl = pfr_attach_table(
1539                                     ruleset, newrule->overload_tblname)) ==
1540                                     NULL)
1541                                         error = EINVAL;
1542                                 else
1543                                         newrule->overload_tbl->pfrkt_flags |=
1544                                             PFR_TFLAG_ACTIVE;
1545                         }
1546
1547                         pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1548                         if (((((newrule->action == PF_NAT) ||
1549                             (newrule->action == PF_RDR) ||
1550                             (newrule->action == PF_BINAT) ||
1551                             (newrule->rt > PF_FASTROUTE)) &&
1552                             !newrule->anchor)) &&
1553                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1554                                 error = EINVAL;
1555
1556                         if (error) {
1557                                 pf_rm_rule(NULL, newrule);
1558                                 break;
1559                         }
1560                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1561                         newrule->evaluations = 0;
1562                         newrule->packets[0] = newrule->packets[1] = 0;
1563                         newrule->bytes[0] = newrule->bytes[1] = 0;
1564                 }
1565                 pf_empty_pool(&pf_pabuf);
1566
1567                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1568                         oldrule = TAILQ_FIRST(
1569                             ruleset->rules[rs_num].active.ptr);
1570                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1571                         oldrule = TAILQ_LAST(
1572                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1573                 else {
1574                         oldrule = TAILQ_FIRST(
1575                             ruleset->rules[rs_num].active.ptr);
1576                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1577                                 oldrule = TAILQ_NEXT(oldrule, entries);
1578                         if (oldrule == NULL) {
1579                                 if (newrule != NULL)
1580                                         pf_rm_rule(NULL, newrule);
1581                                 error = EINVAL;
1582                                 break;
1583                         }
1584                 }
1585
1586                 if (pcr->action == PF_CHANGE_REMOVE) {
1587                         pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1588                         ruleset->rules[rs_num].active.rcount--;
1589                 } else {
1590                         if (oldrule == NULL)
1591                                 TAILQ_INSERT_TAIL(
1592                                     ruleset->rules[rs_num].active.ptr,
1593                                     newrule, entries);
1594                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1595                             pcr->action == PF_CHANGE_ADD_BEFORE)
1596                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1597                         else
1598                                 TAILQ_INSERT_AFTER(
1599                                     ruleset->rules[rs_num].active.ptr,
1600                                     oldrule, newrule, entries);
1601                         ruleset->rules[rs_num].active.rcount++;
1602                 }
1603
1604                 nr = 0;
1605                 TAILQ_FOREACH(oldrule,
1606                     ruleset->rules[rs_num].active.ptr, entries)
1607                         oldrule->nr = nr++;
1608
1609                 ruleset->rules[rs_num].active.ticket++;
1610
1611                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1612                 pf_remove_if_empty_ruleset(ruleset);
1613
1614                 break;
1615         }
1616
1617         case DIOCCLRSTATES: {
1618                 struct pf_state         *s, *nexts;
1619                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1620                 u_int                    killed = 0;
1621                 globaldata_t save_gd = mycpu;
1622                 int nn;
1623
1624                 for (nn = 0; nn < ncpus; ++nn) {
1625                         lwkt_setcpu_self(globaldata_find(nn));
1626                         for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]);
1627                              s; s = nexts) {
1628                                 nexts = RB_NEXT(pf_state_tree_id,
1629                                                 &tree_id[nn], s);
1630
1631                                 if (!psk->psk_ifname[0] ||
1632                                     !strcmp(psk->psk_ifname,
1633                                             s->kif->pfik_name)) {
1634                                         /*
1635                                          * don't send out individual
1636                                          * delete messages
1637                                          */
1638                                         s->sync_flags = PFSTATE_NOSYNC;
1639                                         pf_unlink_state(s);
1640                                         killed++;
1641                                 }
1642                         }
1643                 }
1644                 lwkt_setcpu_self(save_gd);
1645                 psk->psk_killed = killed;
1646                 pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1647                 break;
1648         }
1649
1650         case DIOCKILLSTATES: {
1651                 struct pf_state         *s, *nexts;
1652                 struct pf_state_key     *sk;
1653                 struct pf_addr          *srcaddr, *dstaddr;
1654                 u_int16_t                srcport, dstport;
1655                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1656                 u_int                    killed = 0;
1657                 globaldata_t save_gd = mycpu;
1658                 int nn;
1659
1660                 if (psk->psk_pfcmp.id) {
1661                         if (psk->psk_pfcmp.creatorid == 0)
1662                                 psk->psk_pfcmp.creatorid = pf_status.hostid;
1663                         for (nn = 0; nn < ncpus; ++nn) {
1664                                 lwkt_setcpu_self(globaldata_find(nn));
1665                                 if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1666                                         /* send immediate delete of state */
1667                                         pfsync_delete_state(s);
1668                                         s->sync_flags |= PFSTATE_NOSYNC;
1669                                         pf_unlink_state(s);
1670                                         ++psk->psk_killed;
1671                                 }
1672                         }
1673                         lwkt_setcpu_self(save_gd);
1674                         break;
1675                 }
1676
1677                 for (nn = 0; nn < ncpus; ++nn) {
1678                     lwkt_setcpu_self(globaldata_find(nn));
1679                     for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]);
1680                          s; s = nexts) {
1681                             nexts = RB_NEXT(pf_state_tree_id, &tree_id[nn], s);
1682                             sk = s->key[PF_SK_WIRE];
1683
1684                             if (s->direction == PF_OUT) {
1685                                     srcaddr = &sk->addr[1];
1686                                     dstaddr = &sk->addr[0];
1687                                     srcport = sk->port[0];
1688                                     dstport = sk->port[0];
1689                             } else {
1690                                     srcaddr = &sk->addr[0];
1691                                     dstaddr = &sk->addr[1];
1692                                     srcport = sk->port[0];
1693                                     dstport = sk->port[0];
1694                             }
1695                             if ((!psk->psk_af || sk->af == psk->psk_af)
1696                                 && (!psk->psk_proto || psk->psk_proto ==
1697                                                        sk->proto) &&
1698                                 PF_MATCHA(psk->psk_src.neg,
1699                                           &psk->psk_src.addr.v.a.addr,
1700                                           &psk->psk_src.addr.v.a.mask,
1701                                           srcaddr, sk->af) &&
1702                                 PF_MATCHA(psk->psk_dst.neg,
1703                                           &psk->psk_dst.addr.v.a.addr,
1704                                           &psk->psk_dst.addr.v.a.mask,
1705                                           dstaddr, sk->af) &&
1706                                 (psk->psk_src.port_op == 0 ||
1707                                  pf_match_port(psk->psk_src.port_op,
1708                                                psk->psk_src.port[0],
1709                                                psk->psk_src.port[1],
1710                                                srcport)) &&
1711                                 (psk->psk_dst.port_op == 0 ||
1712                                  pf_match_port(psk->psk_dst.port_op,
1713                                                psk->psk_dst.port[0],
1714                                                psk->psk_dst.port[1],
1715                                                dstport)) &&
1716                                 (!psk->psk_label[0] ||
1717                                  (s->rule.ptr->label[0] &&
1718                                   !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1719                                 (!psk->psk_ifname[0] ||
1720                                  !strcmp(psk->psk_ifname, s->kif->pfik_name))) {
1721                                     /* send immediate delete of state */
1722                                     pfsync_delete_state(s);
1723                                     s->sync_flags |= PFSTATE_NOSYNC;
1724                                     pf_unlink_state(s);
1725                                     killed++;
1726                             }
1727                     }
1728                 }
1729                 lwkt_setcpu_self(save_gd);
1730                 psk->psk_killed = killed;
1731                 break;
1732         }
1733
1734         case DIOCADDSTATE: {
1735                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1736                 struct pfsync_state     *sp = &ps->state;
1737
1738                 if (sp->timeout >= PFTM_MAX &&
1739                     sp->timeout != PFTM_UNTIL_PACKET) {
1740                         error = EINVAL;
1741                         break;
1742                 }
1743                 error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1744                 break;
1745         }
1746
1747         case DIOCGETSTATE: {
1748                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1749                 struct pf_state         *s;
1750                 struct pf_state_cmp      id_key;
1751                 globaldata_t save_gd = mycpu;
1752                 int nn;
1753
1754                 bcopy(ps->state.id, &id_key.id, sizeof(id_key.id));
1755                 id_key.creatorid = ps->state.creatorid;
1756                 s = NULL;
1757                 for (nn = 0; nn < ncpus; ++nn) {
1758                         lwkt_setcpu_self(globaldata_find(nn));
1759                         s = pf_find_state_byid(&id_key);
1760                         if (s)
1761                                 break;
1762                 }
1763                 if (s) {
1764                         pfsync_state_export(&ps->state, s);
1765                 } else {
1766                         error = ENOENT;
1767                 }
1768                 lwkt_setcpu_self(save_gd);
1769                 break;
1770         }
1771
1772         case DIOCGETSTATES: {
1773                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1774                 struct pf_state         *state;
1775                 struct pfsync_state     *p, *pstore;
1776                 u_int32_t                nr = 0;
1777                 globaldata_t save_gd = mycpu;
1778                 int nn;
1779
1780                 if (ps->ps_len == 0) {
1781                         nr = pf_status.states;
1782                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1783                         break;
1784                 }
1785
1786                 pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1787
1788                 p = ps->ps_states;
1789
1790                 for (nn = 0; nn < ncpus; ++nn) {
1791                         lwkt_setcpu_self(globaldata_find(nn));
1792                         state = TAILQ_FIRST(&state_list[nn]);
1793                         while (state) {
1794                                 if (state->timeout != PFTM_UNLINKED) {
1795                                         if ((nr + 1) * sizeof(*p) >
1796                                             (unsigned)ps->ps_len) {
1797                                                 break;
1798                                         }
1799                                         pfsync_state_export(pstore, state);
1800                                         error = copyout(pstore, p, sizeof(*p));
1801                                         if (error) {
1802                                                 kfree(pstore, M_TEMP);
1803                                                 lwkt_setcpu_self(save_gd);
1804                                                 goto fail;
1805                                         }
1806                                         p++;
1807                                         nr++;
1808                                 }
1809                                 state = TAILQ_NEXT(state, entry_list);
1810                         }
1811                 }
1812                 lwkt_setcpu_self(save_gd);
1813                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1814                 kfree(pstore, M_TEMP);
1815                 break;
1816         }
1817
1818         case DIOCGETSTATUS: {
1819                 struct pf_status *s = (struct pf_status *)addr;
1820                 bcopy(&pf_status, s, sizeof(struct pf_status));
1821                 pfi_update_status(s->ifname, s);
1822                 break;
1823         }
1824
1825         case DIOCSETSTATUSIF: {
1826                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1827
1828                 if (pi->ifname[0] == 0) {
1829                         bzero(pf_status.ifname, IFNAMSIZ);
1830                         break;
1831                 }
1832                 strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1833                 break;
1834         }
1835
1836         case DIOCCLRSTATUS: {
1837                 bzero(pf_status.counters, sizeof(pf_status.counters));
1838                 bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1839                 bzero(pf_status.scounters, sizeof(pf_status.scounters));
1840                 pf_status.since = time_second;
1841                 if (*pf_status.ifname)
1842                         pfi_update_status(pf_status.ifname, NULL);
1843                 break;
1844         }
1845
1846         case DIOCNATLOOK: {
1847                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1848                 struct pf_state_key     *sk;
1849                 struct pf_state         *state;
1850                 struct pf_state_key_cmp  key;
1851                 int                      m = 0, direction = pnl->direction;
1852                 int                      sidx, didx;
1853                 globaldata_t save_gd = mycpu;
1854                 int nn;
1855
1856                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1857                 sidx = (direction == PF_IN) ? 1 : 0;
1858                 didx = (direction == PF_IN) ? 0 : 1;
1859
1860                 if (!pnl->proto ||
1861                     PF_AZERO(&pnl->saddr, pnl->af) ||
1862                     PF_AZERO(&pnl->daddr, pnl->af) ||
1863                     ((pnl->proto == IPPROTO_TCP ||
1864                     pnl->proto == IPPROTO_UDP) &&
1865                     (!pnl->dport || !pnl->sport)))
1866                         error = EINVAL;
1867                 else {
1868                         key.af = pnl->af;
1869                         key.proto = pnl->proto;
1870                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1871                         key.port[sidx] = pnl->sport;
1872                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1873                         key.port[didx] = pnl->dport;
1874
1875                         state = NULL;
1876                         for (nn = 0; nn < ncpus; ++nn) {
1877                                 lwkt_setcpu_self(globaldata_find(nn));
1878                                 state = pf_find_state_all(&key, direction, &m);
1879                                 if (state || m > 1)
1880                                         break;
1881                                 m = 0;
1882                         }
1883
1884                         if (m > 1) {
1885                                 error = E2BIG;  /* more than one state */
1886                         } else if (state != NULL) {
1887                                 sk = state->key[sidx];
1888                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1889                                 pnl->rsport = sk->port[sidx];
1890                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1891                                 pnl->rdport = sk->port[didx];
1892                         } else {
1893                                 error = ENOENT;
1894                         }
1895                         lwkt_setcpu_self(save_gd);
1896                 }
1897                 break;
1898         }
1899
1900         case DIOCSETTIMEOUT: {
1901                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1902                 int              old;
1903
1904                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1905                     pt->seconds < 0) {
1906                         error = EINVAL;
1907                         goto fail;
1908                 }
1909                 old = pf_default_rule.timeout[pt->timeout];
1910                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1911                         pt->seconds = 1;
1912                 pf_default_rule.timeout[pt->timeout] = pt->seconds;
1913                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1914                         wakeup(pf_purge_thread);
1915                 pt->seconds = old;
1916                 break;
1917         }
1918
1919         case DIOCGETTIMEOUT: {
1920                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1921
1922                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1923                         error = EINVAL;
1924                         goto fail;
1925                 }
1926                 pt->seconds = pf_default_rule.timeout[pt->timeout];
1927                 break;
1928         }
1929
1930         case DIOCGETLIMIT: {
1931                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1932
1933                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1934                         error = EINVAL;
1935                         goto fail;
1936                 }
1937                 pl->limit = pf_pool_limits[pl->index].limit;
1938                 break;
1939         }
1940
1941         case DIOCSETLIMIT: {
1942                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1943                 int                      old_limit;
1944
1945                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1946                     pf_pool_limits[pl->index].pp == NULL) {
1947                         error = EINVAL;
1948                         goto fail;
1949                 }
1950
1951                 /* XXX Get an API to set limits on the zone/pool */
1952                 old_limit = pf_pool_limits[pl->index].limit;
1953                 pf_pool_limits[pl->index].limit = pl->limit;
1954                 pl->limit = old_limit;
1955                 break;
1956         }
1957
1958         case DIOCSETDEBUG: {
1959                 u_int32_t       *level = (u_int32_t *)addr;
1960
1961                 pf_status.debug = *level;
1962                 break;
1963         }
1964
1965         case DIOCCLRRULECTRS: {
1966                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1967                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1968                 struct pf_rule          *rule;
1969
1970                 TAILQ_FOREACH(rule,
1971                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1972                         rule->evaluations = 0;
1973                         rule->packets[0] = rule->packets[1] = 0;
1974                         rule->bytes[0] = rule->bytes[1] = 0;
1975                 }
1976                 break;
1977         }
1978
1979         case DIOCGIFSPEED: {
1980                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1981                 struct pf_ifspeed       ps;
1982                 struct ifnet            *ifp;
1983
1984                 if (psp->ifname[0] != 0) {
1985                         /* Can we completely trust user-land? */
1986                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1987                         ifnet_lock();
1988                         ifp = ifunit(ps.ifname);
1989                         ifnet_unlock();
1990                         if (ifp)
1991                                 psp->baudrate = ifp->if_baudrate;
1992                         else
1993                                 error = EINVAL;
1994                 } else
1995                         error = EINVAL;
1996                 break;
1997         }
1998 #ifdef ALTQ
1999         case DIOCSTARTALTQ: {
2000                 struct pf_altq          *altq;
2001
2002                 /* enable all altq interfaces on active list */
2003                 /* XXX: locking? */
2004                 TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2005                         if (altq->qname[0] == 0 && (altq->local_flags &
2006                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2007                                 error = pf_enable_altq(altq);
2008                                 if (error != 0)
2009                                         break;
2010                         }
2011                 }
2012                 if (error == 0)
2013                         pf_altq_running = 1;
2014                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2015                 break;
2016         }
2017
2018         case DIOCSTOPALTQ: {
2019                 struct pf_altq          *altq;
2020
2021                 /* disable all altq interfaces on active list */
2022                 TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2023                         if (altq->qname[0] == 0 && (altq->local_flags &
2024                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2025                                 error = pf_disable_altq(altq);
2026                                 if (error != 0)
2027                                         break;
2028                         }
2029                 }
2030                 if (error == 0)
2031                         pf_altq_running = 0;
2032                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2033                 break;
2034         }
2035
2036         case DIOCADDALTQ: {
2037                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2038                 struct pf_altq          *altq, *a;
2039                 struct ifnet            *ifp;
2040
2041                 if (pa->ticket != ticket_altqs_inactive) {
2042                         error = EBUSY;
2043                         break;
2044                 }
2045                 altq = kmalloc(sizeof(struct pf_altq), M_PFALTQPL, M_WAITOK|M_NULLOK);
2046                 if (altq == NULL) {
2047                         error = ENOMEM;
2048                         break;
2049                 }
2050                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2051
2052                 /*
2053                  * if this is for a queue, find the discipline and
2054                  * copy the necessary fields
2055                  */
2056                 if (altq->qname[0] != 0) {
2057                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2058                                 error = EBUSY;
2059                                 kfree(altq, M_PFALTQPL);
2060                                 break;
2061                         }
2062                         altq->altq_disc = NULL;
2063                         TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2064                                 if (strncmp(a->ifname, altq->ifname,
2065                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
2066                                         altq->altq_disc = a->altq_disc;
2067                                         break;
2068                                 }
2069                         }
2070                 }
2071
2072                 ifnet_lock();
2073                 ifp = ifunit(altq->ifname);
2074                 ifnet_unlock();
2075                 if (ifp == NULL)
2076                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2077                 else
2078                         error = altq_add(altq);
2079
2080                 if (error) {
2081                         kfree(altq, M_PFALTQPL);
2082                         break;
2083                 }
2084
2085                 TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2086                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2087                 break;
2088         }
2089
2090         case DIOCGETALTQS: {
2091                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2092                 struct pf_altq          *altq;
2093
2094                 pa->nr = 0;
2095                 TAILQ_FOREACH(altq, pf_altqs_active, entries)
2096                         pa->nr++;
2097                 pa->ticket = ticket_altqs_active;
2098                 break;
2099         }
2100
2101         case DIOCGETALTQ: {
2102                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2103                 struct pf_altq          *altq;
2104                 u_int32_t                nr;
2105
2106                 if (pa->ticket != ticket_altqs_active) {
2107                         error = EBUSY;
2108                         break;
2109                 }
2110                 nr = 0;
2111                 altq = TAILQ_FIRST(pf_altqs_active);
2112                 while ((altq != NULL) && (nr < pa->nr)) {
2113                         altq = TAILQ_NEXT(altq, entries);
2114                         nr++;
2115                 }
2116                 if (altq == NULL) {
2117                         error = EBUSY;
2118                         break;
2119                 }
2120                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2121                 break;
2122         }
2123
2124         case DIOCCHANGEALTQ:
2125                 /* CHANGEALTQ not supported yet! */
2126                 error = ENODEV;
2127                 break;
2128
2129         case DIOCGETQSTATS: {
2130                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
2131                 struct pf_altq          *altq;
2132                 u_int32_t                nr;
2133                 int                      nbytes;
2134
2135                 if (pq->ticket != ticket_altqs_active) {
2136                         error = EBUSY;
2137                         break;
2138                 }
2139                 nbytes = pq->nbytes;
2140                 nr = 0;
2141                 altq = TAILQ_FIRST(pf_altqs_active);
2142                 while ((altq != NULL) && (nr < pq->nr)) {
2143                         altq = TAILQ_NEXT(altq, entries);
2144                         nr++;
2145                 }
2146                 if (altq == NULL) {
2147                         error = EBUSY;
2148                         break;
2149                 }
2150                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2151                         error = ENXIO;
2152                         break;
2153                 }
2154                 error = altq_getqstats(altq, pq->buf, &nbytes);
2155                 if (error == 0) {
2156                         pq->scheduler = altq->scheduler;
2157                         pq->nbytes = nbytes;
2158                 }
2159                 break;
2160         }
2161 #endif /* ALTQ */
2162
2163         case DIOCBEGINADDRS: {
2164                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2165
2166                 pf_empty_pool(&pf_pabuf);
2167                 pp->ticket = ++ticket_pabuf;
2168                 break;
2169         }
2170
2171         case DIOCADDADDR: {
2172                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2173
2174                 if (pp->ticket != ticket_pabuf) {
2175                         error = EBUSY;
2176                         break;
2177                 }
2178 #ifndef INET
2179                 if (pp->af == AF_INET) {
2180                         error = EAFNOSUPPORT;
2181                         break;
2182                 }
2183 #endif /* INET */
2184 #ifndef INET6
2185                 if (pp->af == AF_INET6) {
2186                         error = EAFNOSUPPORT;
2187                         break;
2188                 }
2189 #endif /* INET6 */
2190                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2191                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2192                     pp->addr.addr.type != PF_ADDR_TABLE) {
2193                         error = EINVAL;
2194                         break;
2195                 }
2196                 pa = kmalloc(sizeof(struct pf_altq), M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2197                 if (pa == NULL) {
2198                         error = ENOMEM;
2199                         break;
2200                 }
2201                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2202                 if (pa->ifname[0]) {
2203                         pa->kif = pfi_kif_get(pa->ifname);
2204                         if (pa->kif == NULL) {
2205                                 kfree(ap, M_PFPOOLADDRPL);
2206                                 error = EINVAL;
2207                                 break;
2208                         }
2209                         pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2210                 }
2211                 if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2212                         pfi_dynaddr_remove(&pa->addr);
2213                         pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2214                         kfree(pa, M_PFPOOLADDRPL);
2215                         error = EINVAL;
2216                         break;
2217                 }
2218                 TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2219                 break;
2220         }
2221
2222         case DIOCGETADDRS: {
2223                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2224
2225                 pp->nr = 0;
2226                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2227                     pp->r_num, 0, 1, 0);
2228                 if (pool == NULL) {
2229                         error = EBUSY;
2230                         break;
2231                 }
2232                 TAILQ_FOREACH(pa, &pool->list, entries)
2233                         pp->nr++;
2234                 break;
2235         }
2236
2237         case DIOCGETADDR: {
2238                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2239                 u_int32_t                nr = 0;
2240
2241                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2242                     pp->r_num, 0, 1, 1);
2243                 if (pool == NULL) {
2244                         error = EBUSY;
2245                         break;
2246                 }
2247                 pa = TAILQ_FIRST(&pool->list);
2248                 while ((pa != NULL) && (nr < pp->nr)) {
2249                         pa = TAILQ_NEXT(pa, entries);
2250                         nr++;
2251                 }
2252                 if (pa == NULL) {
2253                         error = EBUSY;
2254                         break;
2255                 }
2256                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2257                 pf_addr_copyout(&pp->addr.addr);
2258                 break;
2259         }
2260
2261         case DIOCCHANGEADDR: {
2262                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2263                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2264                 struct pf_ruleset       *ruleset;
2265
2266                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2267                     pca->action > PF_CHANGE_REMOVE) {
2268                         error = EINVAL;
2269                         break;
2270                 }
2271                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2272                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2273                     pca->addr.addr.type != PF_ADDR_TABLE) {
2274                         error = EINVAL;
2275                         break;
2276                 }
2277
2278                 ruleset = pf_find_ruleset(pca->anchor);
2279                 if (ruleset == NULL) {
2280                         error = EBUSY;
2281                         break;
2282                 }
2283                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2284                     pca->r_num, pca->r_last, 1, 1);
2285                 if (pool == NULL) {
2286                         error = EBUSY;
2287                         break;
2288                 }
2289                 if (pca->action != PF_CHANGE_REMOVE) {
2290                         newpa = kmalloc(sizeof(struct pf_pooladdr),
2291                                 M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2292                         if (newpa == NULL) {
2293                                 error = ENOMEM;
2294                                 break;
2295                         }
2296                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2297 #ifndef INET
2298                         if (pca->af == AF_INET) {
2299                                 kfree(newpa, M_PFPOOLADDRPL);
2300                                 error = EAFNOSUPPORT;
2301                                 break;
2302                         }
2303 #endif /* INET */
2304 #ifndef INET6
2305                         if (pca->af == AF_INET6) {
2306                                 kfree(newpa, M_PFPOOLADDRPL);
2307                                 error = EAFNOSUPPORT;
2308                                 break;
2309                         }
2310 #endif /* INET6 */
2311                         if (newpa->ifname[0]) {
2312                                 newpa->kif = pfi_kif_get(newpa->ifname);
2313                                 if (newpa->kif == NULL) {
2314                                         kfree(newpa, M_PFPOOLADDRPL);
2315                                         error = EINVAL;
2316                                         break;
2317                                 }
2318                                 pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2319                         } else
2320                                 newpa->kif = NULL;
2321                         if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2322                             pf_tbladdr_setup(ruleset, &newpa->addr)) {
2323                                 pfi_dynaddr_remove(&newpa->addr);
2324                                 pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2325                                 kfree(newpa, M_PFPOOLADDRPL);
2326                                 error = EINVAL;
2327                                 break;
2328                         }
2329                 }
2330
2331                 if (pca->action == PF_CHANGE_ADD_HEAD)
2332                         oldpa = TAILQ_FIRST(&pool->list);
2333                 else if (pca->action == PF_CHANGE_ADD_TAIL)
2334                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2335                 else {
2336                         int     i = 0;
2337
2338                         oldpa = TAILQ_FIRST(&pool->list);
2339                         while ((oldpa != NULL) && (i < pca->nr)) {
2340                                 oldpa = TAILQ_NEXT(oldpa, entries);
2341                                 i++;
2342                         }
2343                         if (oldpa == NULL) {
2344                                 error = EINVAL;
2345                                 break;
2346                         }
2347                 }
2348
2349                 if (pca->action == PF_CHANGE_REMOVE) {
2350                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2351                         pfi_dynaddr_remove(&oldpa->addr);
2352                         pf_tbladdr_remove(&oldpa->addr);
2353                         pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2354                         kfree(oldpa, M_PFPOOLADDRPL);
2355                 } else {
2356                         if (oldpa == NULL)
2357                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2358                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2359                             pca->action == PF_CHANGE_ADD_BEFORE)
2360                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2361                         else
2362                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2363                                     newpa, entries);
2364                 }
2365
2366                 pool->cur = TAILQ_FIRST(&pool->list);
2367                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2368                     pca->af);
2369                 break;
2370         }
2371
2372         case DIOCGETRULESETS: {
2373                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2374                 struct pf_ruleset       *ruleset;
2375                 struct pf_anchor        *anchor;
2376
2377                 pr->path[sizeof(pr->path) - 1] = 0;
2378                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2379                         error = EINVAL;
2380                         break;
2381                 }
2382                 pr->nr = 0;
2383                 if (ruleset->anchor == NULL) {
2384                         /* XXX kludge for pf_main_ruleset */
2385                         RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2386                                 if (anchor->parent == NULL)
2387                                         pr->nr++;
2388                 } else {
2389                         RB_FOREACH(anchor, pf_anchor_node,
2390                             &ruleset->anchor->children)
2391                                 pr->nr++;
2392                 }
2393                 break;
2394         }
2395
2396         case DIOCGETRULESET: {
2397                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2398                 struct pf_ruleset       *ruleset;
2399                 struct pf_anchor        *anchor;
2400                 u_int32_t                nr = 0;
2401
2402                 pr->path[sizeof(pr->path) - 1] = 0;
2403                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2404                         error = EINVAL;
2405                         break;
2406                 }
2407                 pr->name[0] = 0;
2408                 if (ruleset->anchor == NULL) {
2409                         /* XXX kludge for pf_main_ruleset */
2410                         RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2411                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2412                                         strlcpy(pr->name, anchor->name,
2413                                             sizeof(pr->name));
2414                                         break;
2415                                 }
2416                 } else {
2417                         RB_FOREACH(anchor, pf_anchor_node,
2418                             &ruleset->anchor->children)
2419                                 if (nr++ == pr->nr) {
2420                                         strlcpy(pr->name, anchor->name,
2421                                             sizeof(pr->name));
2422                                         break;
2423                                 }
2424                 }
2425                 if (!pr->name[0])
2426                         error = EBUSY;
2427                 break;
2428         }
2429
2430         case DIOCRCLRTABLES: {
2431                 struct pfioc_table *io = (struct pfioc_table *)addr;
2432
2433                 if (io->pfrio_esize != 0) {
2434                         error = ENODEV;
2435                         break;
2436                 }
2437                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2438                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2439                 break;
2440         }
2441
2442         case DIOCRADDTABLES: {
2443                 struct pfioc_table *io = (struct pfioc_table *)addr;
2444
2445                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2446                         error = ENODEV;
2447                         break;
2448                 }
2449                 error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2450                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2451                 break;
2452         }
2453
2454         case DIOCRDELTABLES: {
2455                 struct pfioc_table *io = (struct pfioc_table *)addr;
2456
2457                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2458                         error = ENODEV;
2459                         break;
2460                 }
2461                 error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2462                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2463                 break;
2464         }
2465
2466         case DIOCRGETTABLES: {
2467                 struct pfioc_table *io = (struct pfioc_table *)addr;
2468
2469                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2470                         error = ENODEV;
2471                         break;
2472                 }
2473                 error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2474                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2475                 break;
2476         }
2477
2478         case DIOCRGETTSTATS: {
2479                 struct pfioc_table *io = (struct pfioc_table *)addr;
2480
2481                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2482                         error = ENODEV;
2483                         break;
2484                 }
2485                 error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2486                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2487                 break;
2488         }
2489
2490         case DIOCRCLRTSTATS: {
2491                 struct pfioc_table *io = (struct pfioc_table *)addr;
2492
2493                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2494                         error = ENODEV;
2495                         break;
2496                 }
2497                 error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2498                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2499                 break;
2500         }
2501
2502         case DIOCRSETTFLAGS: {
2503                 struct pfioc_table *io = (struct pfioc_table *)addr;
2504
2505                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2506                         error = ENODEV;
2507                         break;
2508                 }
2509                 error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2510                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2511                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2512                 break;
2513         }
2514
2515         case DIOCRCLRADDRS: {
2516                 struct pfioc_table *io = (struct pfioc_table *)addr;
2517
2518                 if (io->pfrio_esize != 0) {
2519                         error = ENODEV;
2520                         break;
2521                 }
2522                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2523                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2524                 break;
2525         }
2526
2527         case DIOCRADDADDRS: {
2528                 struct pfioc_table *io = (struct pfioc_table *)addr;
2529
2530                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2531                         error = ENODEV;
2532                         break;
2533                 }
2534                 error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2535                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2536                     PFR_FLAG_USERIOCTL);
2537                 break;
2538         }
2539
2540         case DIOCRDELADDRS: {
2541                 struct pfioc_table *io = (struct pfioc_table *)addr;
2542
2543                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2544                         error = ENODEV;
2545                         break;
2546                 }
2547                 error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2548                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2549                     PFR_FLAG_USERIOCTL);
2550                 break;
2551         }
2552
2553         case DIOCRSETADDRS: {
2554                 struct pfioc_table *io = (struct pfioc_table *)addr;
2555
2556                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2557                         error = ENODEV;
2558                         break;
2559                 }
2560                 error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2561                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2562                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2563                     PFR_FLAG_USERIOCTL, 0);
2564                 break;
2565         }
2566
2567         case DIOCRGETADDRS: {
2568                 struct pfioc_table *io = (struct pfioc_table *)addr;
2569
2570                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2571                         error = ENODEV;
2572                         break;
2573                 }
2574                 error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2575                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2576                 break;
2577         }
2578
2579         case DIOCRGETASTATS: {
2580                 struct pfioc_table *io = (struct pfioc_table *)addr;
2581
2582                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2583                         error = ENODEV;
2584                         break;
2585                 }
2586                 error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2587                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2588                 break;
2589         }
2590
2591         case DIOCRCLRASTATS: {
2592                 struct pfioc_table *io = (struct pfioc_table *)addr;
2593
2594                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2595                         error = ENODEV;
2596                         break;
2597                 }
2598                 error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2599                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2600                     PFR_FLAG_USERIOCTL);
2601                 break;
2602         }
2603
2604         case DIOCRTSTADDRS: {
2605                 struct pfioc_table *io = (struct pfioc_table *)addr;
2606
2607                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2608                         error = ENODEV;
2609                         break;
2610                 }
2611                 error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2612                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2613                     PFR_FLAG_USERIOCTL);
2614                 break;
2615         }
2616
2617         case DIOCRINADEFINE: {
2618                 struct pfioc_table *io = (struct pfioc_table *)addr;
2619
2620                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2621                         error = ENODEV;
2622                         break;
2623                 }
2624                 error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2625                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2626                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2627                 break;
2628         }
2629
2630         case DIOCOSFPADD: {
2631                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2632                 error = pf_osfp_add(io);
2633                 break;
2634         }
2635
2636         case DIOCOSFPGET: {
2637                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2638                 error = pf_osfp_get(io);
2639                 break;
2640         }
2641
2642         case DIOCXBEGIN: {
2643                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2644                 struct pfioc_trans_e    *ioe;
2645                 struct pfr_table        *table;
2646                 int                      i;
2647
2648                 if (io->esize != sizeof(*ioe)) {
2649                         error = ENODEV;
2650                         goto fail;
2651                 }
2652                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2653                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2654                 for (i = 0; i < io->size; i++) {
2655                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2656                                 kfree(table, M_TEMP);
2657                                 kfree(ioe, M_TEMP);
2658                                 error = EFAULT;
2659                                 goto fail;
2660                         }
2661                         switch (ioe->rs_num) {
2662 #ifdef ALTQ
2663                         case PF_RULESET_ALTQ:
2664                                 if (ioe->anchor[0]) {
2665                                         kfree(table, M_TEMP);
2666                                         kfree(ioe, M_TEMP);
2667                                         error = EINVAL;
2668                                         goto fail;
2669                                 }
2670                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2671                                         kfree(table, M_TEMP);
2672                                         kfree(ioe, M_TEMP);
2673                                         goto fail;
2674                                 }
2675                                 break;
2676 #endif /* ALTQ */
2677                         case PF_RULESET_TABLE:
2678                                 bzero(table, sizeof(*table));
2679                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2680                                     sizeof(table->pfrt_anchor));
2681                                 if ((error = pfr_ina_begin(table,
2682                                     &ioe->ticket, NULL, 0))) {
2683                                         kfree(table, M_TEMP);
2684                                         kfree(ioe, M_TEMP);
2685                                         goto fail;
2686                                 }
2687                                 break;
2688                         default:
2689                                 if ((error = pf_begin_rules(&ioe->ticket,
2690                                     ioe->rs_num, ioe->anchor))) {
2691                                         kfree(table, M_TEMP);
2692                                         kfree(ioe, M_TEMP);
2693                                         goto fail;
2694                                 }
2695                                 break;
2696                         }
2697                         if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2698                                 kfree(table, M_TEMP);
2699                                 kfree(ioe, M_TEMP);
2700                                 error = EFAULT;
2701                                 goto fail;
2702                         }
2703                 }
2704                 kfree(table, M_TEMP);
2705                 kfree(ioe, M_TEMP);
2706                 break;
2707         }
2708
2709         case DIOCXROLLBACK: {
2710                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2711                 struct pfioc_trans_e    *ioe;
2712                 struct pfr_table        *table;
2713                 int                      i;
2714
2715                 if (io->esize != sizeof(*ioe)) {
2716                         error = ENODEV;
2717                         goto fail;
2718                 }
2719                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2720                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2721                 for (i = 0; i < io->size; i++) {
2722                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2723                                 kfree(table, M_TEMP);
2724                                 kfree(ioe, M_TEMP);
2725                                 error = EFAULT;
2726                                 goto fail;
2727                         }
2728                         switch (ioe->rs_num) {
2729 #ifdef ALTQ
2730                         case PF_RULESET_ALTQ:
2731                                 if (ioe->anchor[0]) {
2732                                         kfree(table, M_TEMP);
2733                                         kfree(ioe, M_TEMP);
2734                                         error = EINVAL;
2735                                         goto fail;
2736                                 }
2737                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2738                                         kfree(table, M_TEMP);
2739                                         kfree(ioe, M_TEMP);
2740                                         goto fail; /* really bad */
2741                                 }
2742                                 break;
2743 #endif /* ALTQ */
2744                         case PF_RULESET_TABLE:
2745                                 bzero(table, sizeof(*table));
2746                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2747                                     sizeof(table->pfrt_anchor));
2748                                 if ((error = pfr_ina_rollback(table,
2749                                     ioe->ticket, NULL, 0))) {
2750                                         kfree(table, M_TEMP);
2751                                         kfree(ioe, M_TEMP);
2752                                         goto fail; /* really bad */
2753                                 }
2754                                 break;
2755                         default:
2756                                 if ((error = pf_rollback_rules(ioe->ticket,
2757                                     ioe->rs_num, ioe->anchor))) {
2758                                         kfree(table, M_TEMP);
2759                                         kfree(ioe, M_TEMP);
2760                                         goto fail; /* really bad */
2761                                 }
2762                                 break;
2763                         }
2764                 }
2765                 kfree(table, M_TEMP);
2766                 kfree(ioe, M_TEMP);
2767                 break;
2768         }
2769
2770         case DIOCXCOMMIT: {
2771                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2772                 struct pfioc_trans_e    *ioe;
2773                 struct pfr_table        *table;
2774                 struct pf_ruleset       *rs;
2775                 int                      i;
2776
2777                 if (io->esize != sizeof(*ioe)) {
2778                         error = ENODEV;
2779                         goto fail;
2780                 }
2781                 ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2782                 table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2783                 /* first makes sure everything will succeed */
2784                 for (i = 0; i < io->size; i++) {
2785                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2786                                 kfree(table, M_TEMP);
2787                                 kfree(ioe, M_TEMP);
2788                                 error = EFAULT;
2789                                 goto fail;
2790                         }
2791                         switch (ioe->rs_num) {
2792 #ifdef ALTQ
2793                         case PF_RULESET_ALTQ:
2794                                 if (ioe->anchor[0]) {
2795                                         kfree(table, M_TEMP);
2796                                         kfree(ioe, M_TEMP);
2797                                         error = EINVAL;
2798                                         goto fail;
2799                                 }
2800                                 if (!altqs_inactive_open || ioe->ticket !=
2801                                     ticket_altqs_inactive) {
2802                                         kfree(table, M_TEMP);
2803                                         kfree(ioe, M_TEMP);
2804                                         error = EBUSY;
2805                                         goto fail;
2806                                 }
2807                                 break;
2808 #endif /* ALTQ */
2809                         case PF_RULESET_TABLE:
2810                                 rs = pf_find_ruleset(ioe->anchor);
2811                                 if (rs == NULL || !rs->topen || ioe->ticket !=
2812                                      rs->tticket) {
2813                                         kfree(table, M_TEMP);
2814                                         kfree(ioe, M_TEMP);
2815                                         error = EBUSY;
2816                                         goto fail;
2817                                 }
2818                                 break;
2819                         default:
2820                                 if (ioe->rs_num < 0 || ioe->rs_num >=
2821                                     PF_RULESET_MAX) {
2822                                         kfree(table, M_TEMP);
2823                                         kfree(ioe, M_TEMP);
2824                                         error = EINVAL;
2825                                         goto fail;
2826                                 }
2827                                 rs = pf_find_ruleset(ioe->anchor);
2828                                 if (rs == NULL ||
2829                                     !rs->rules[ioe->rs_num].inactive.open ||
2830                                     rs->rules[ioe->rs_num].inactive.ticket !=
2831                                     ioe->ticket) {
2832                                         kfree(table, M_TEMP);
2833                                         kfree(ioe, M_TEMP);
2834                                         error = EBUSY;
2835                                         goto fail;
2836                                 }
2837                                 break;
2838                         }
2839                 }
2840                 /* now do the commit - no errors should happen here */
2841                 for (i = 0; i < io->size; i++) {
2842                         if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2843                                 kfree(table, M_TEMP);
2844                                 kfree(ioe, M_TEMP);
2845                                 error = EFAULT;
2846                                 goto fail;
2847                         }
2848                         switch (ioe->rs_num) {
2849 #ifdef ALTQ
2850                         case PF_RULESET_ALTQ:
2851                                 if ((error = pf_commit_altq(ioe->ticket))) {
2852                                         kfree(table, M_TEMP);
2853                                         kfree(ioe, M_TEMP);
2854                                         goto fail; /* really bad */
2855                                 }
2856                                 break;
2857 #endif /* ALTQ */
2858                         case PF_RULESET_TABLE:
2859                                 bzero(table, sizeof(*table));
2860                                 strlcpy(table->pfrt_anchor, ioe->anchor,
2861                                     sizeof(table->pfrt_anchor));
2862                                 if ((error = pfr_ina_commit(table, ioe->ticket,
2863                                     NULL, NULL, 0))) {
2864                                         kfree(table, M_TEMP);
2865                                         kfree(ioe, M_TEMP);
2866                                         goto fail; /* really bad */
2867                                 }
2868                                 break;
2869                         default:
2870                                 if ((error = pf_commit_rules(ioe->ticket,
2871                                     ioe->rs_num, ioe->anchor))) {
2872                                         kfree(table, M_TEMP);
2873                                         kfree(ioe, M_TEMP);
2874                                         goto fail; /* really bad */
2875                                 }
2876                                 break;
2877                         }
2878                 }
2879                 kfree(table, M_TEMP);
2880                 kfree(ioe, M_TEMP);
2881                 break;
2882         }
2883
2884         case DIOCGETSRCNODES: {
2885                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
2886                 struct pf_src_node      *n, *p, *pstore;
2887                 u_int32_t nr = 0;
2888                 int     space = psn->psn_len;
2889                 int     nn;
2890
2891                 if (space == 0) {
2892                         for (nn = 0; nn < ncpus; ++nn) {
2893                                 RB_FOREACH(n, pf_src_tree,
2894                                            &tree_src_tracking[nn]) {
2895                                         nr++;
2896                                 }
2897                         }
2898                         psn->psn_len = sizeof(struct pf_src_node) * nr;
2899                         break;
2900                 }
2901
2902                 pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2903
2904                 p = psn->psn_src_nodes;
2905
2906                 /*
2907                  * WARNING: We are not switching cpus so we cannot call
2908                  *          nominal pf.c support routines for cpu-specific
2909                  *          data.
2910                  */
2911                 for (nn = 0; nn < ncpus; ++nn) {
2912                         RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
2913                                 int     secs = time_second, diff;
2914
2915                                 if ((nr + 1) * sizeof(*p) >
2916                                     (unsigned)psn->psn_len) {
2917                                         break;
2918                                 }
2919
2920                                 bcopy(n, pstore, sizeof(*pstore));
2921                                 if (n->rule.ptr != NULL)
2922                                         pstore->rule.nr = n->rule.ptr->nr;
2923                                 pstore->creation = secs - pstore->creation;
2924                                 if (pstore->expire > secs)
2925                                         pstore->expire -= secs;
2926                                 else
2927                                         pstore->expire = 0;
2928
2929                                 /* adjust the connection rate estimate */
2930                                 diff = secs - n->conn_rate.last;
2931                                 if (diff >= n->conn_rate.seconds)
2932                                         pstore->conn_rate.count = 0;
2933                                 else
2934                                         pstore->conn_rate.count -=
2935                                             n->conn_rate.count * diff /
2936                                             n->conn_rate.seconds;
2937
2938                                 error = copyout(pstore, p, sizeof(*p));
2939                                 if (error) {
2940                                         kfree(pstore, M_TEMP);
2941                                         goto fail;
2942                                 }
2943                                 p++;
2944                                 nr++;
2945                         }
2946                 }
2947                 psn->psn_len = sizeof(struct pf_src_node) * nr;
2948                 kfree(pstore, M_TEMP);
2949                 break;
2950         }
2951
2952         case DIOCCLRSRCNODES: {
2953                 struct pf_src_node      *n;
2954                 struct pf_state         *state;
2955                 globaldata_t save_gd = mycpu;
2956                 int nn;
2957
2958                 /*
2959                  * WARNING: We are not switching cpus so we cannot call
2960                  *          nominal pf.c support routines for cpu-specific
2961                  *          data.
2962                  */
2963                 for (nn = 0; nn < ncpus; ++nn) {
2964                         RB_FOREACH(state, pf_state_tree_id, &tree_id[nn]) {
2965                                 state->src_node = NULL;
2966                                 state->nat_src_node = NULL;
2967                         }
2968                         RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
2969                                 n->expire = 1;
2970                                 n->states = 0;
2971                         }
2972                 }
2973
2974                 /*
2975                  * WARNING: Must move to the target cpu for nominal calls
2976                  *          into pf.c
2977                  */
2978                 for (nn = 0; nn < ncpus; ++nn) {
2979                         lwkt_setcpu_self(globaldata_find(nn));
2980                         pf_purge_expired_src_nodes(1);
2981                 }
2982                 lwkt_setcpu_self(save_gd);
2983                 pf_status.src_nodes = 0;
2984                 break;
2985         }
2986
2987         case DIOCKILLSRCNODES: {
2988                 struct pf_src_node      *sn;
2989                 struct pf_state         *s;
2990                 struct pfioc_src_node_kill *psnk =
2991                     (struct pfioc_src_node_kill *)addr;
2992                 u_int                   killed = 0;
2993                 globaldata_t save_gd = mycpu;
2994                 int nn;
2995
2996                 /*
2997                  * WARNING: We are not switching cpus so we cannot call
2998                  *          nominal pf.c support routines for cpu-specific
2999                  *          data.
3000                  */
3001                 for (nn = 0; nn < ncpus; ++nn) {
3002                     RB_FOREACH(sn, pf_src_tree, &tree_src_tracking[nn]) {
3003                         if (PF_MATCHA(psnk->psnk_src.neg,
3004                                 &psnk->psnk_src.addr.v.a.addr,
3005                                 &psnk->psnk_src.addr.v.a.mask,
3006                                 &sn->addr, sn->af) &&
3007                             PF_MATCHA(psnk->psnk_dst.neg,
3008                                 &psnk->psnk_dst.addr.v.a.addr,
3009                                 &psnk->psnk_dst.addr.v.a.mask,
3010                                 &sn->raddr, sn->af)) {
3011                                 /* Handle state to src_node linkage */
3012                                 if (sn->states != 0) {
3013                                         RB_FOREACH(s, pf_state_tree_id,
3014                                             &tree_id[nn]) {
3015                                                 if (s->src_node == sn)
3016                                                         s->src_node = NULL;
3017                                                 if (s->nat_src_node == sn)
3018                                                         s->nat_src_node = NULL;
3019                                         }
3020                                         sn->states = 0;
3021                                 }
3022                                 sn->expire = 1;
3023                                 killed++;
3024                         }
3025                     }
3026                 }
3027                 if (killed > 0) {
3028                         for (nn = 0; nn < ncpus; ++nn) {
3029                                 lwkt_setcpu_self(globaldata_find(nn));
3030                                 pf_purge_expired_src_nodes(1);
3031                         }
3032                         lwkt_setcpu_self(save_gd);
3033                 }
3034
3035                 psnk->psnk_killed = killed;
3036                 break;
3037         }
3038
3039         case DIOCSETHOSTID: {
3040                 u_int32_t       *hostid = (u_int32_t *)addr;
3041
3042                 if (*hostid == 0)
3043                         pf_status.hostid = karc4random();
3044                 else
3045                         pf_status.hostid = *hostid;
3046                 break;
3047         }
3048
3049         case DIOCOSFPFLUSH:
3050                 crit_enter();
3051                 pf_osfp_flush();
3052                 crit_exit();
3053                 break;
3054
3055         case DIOCIGETIFACES: {
3056                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3057
3058                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3059                         error = ENODEV;
3060                         break;
3061                 }
3062                 error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3063                     &io->pfiio_size);
3064                 break;
3065         }
3066
3067         case DIOCSETIFFLAG: {
3068                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3069
3070                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3071                 break;
3072         }
3073
3074         case DIOCCLRIFFLAG: {
3075                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3076
3077                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3078                 break;
3079         }
3080
3081         default:
3082                 error = ENODEV;
3083                 break;
3084         }
3085 fail:
3086         lwkt_reltoken(&pf_token);
3087         return (error);
3088 }
3089
3090 /*
3091  * XXX - Check for version missmatch!!!
3092  */
3093 static void
3094 pf_clear_states(void)
3095 {
3096         struct pf_state         *s, *nexts;
3097         globaldata_t save_gd = mycpu;
3098         int nn;
3099
3100         for (nn = 0; nn < ncpus; ++nn) {
3101                 lwkt_setcpu_self(globaldata_find(nn));
3102                 for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]); s; s = nexts) {
3103                         nexts = RB_NEXT(pf_state_tree_id, &tree_id[nn], s);
3104
3105                         /* don't send out individual delete messages */
3106                         s->sync_flags = PFSTATE_NOSYNC;
3107                         pf_unlink_state(s);
3108                 }
3109
3110         }
3111         lwkt_setcpu_self(save_gd);
3112
3113 #if 0 /* PFSYNC */
3114 /*
3115  * XXX This is called on module unload, we do not want to sync that over? */
3116  */
3117         pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3118 #endif
3119 }
3120
3121 static int
3122 pf_clear_tables(void)
3123 {
3124         struct pfioc_table io;
3125         int error;
3126
3127         bzero(&io, sizeof(io));
3128
3129         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3130             io.pfrio_flags);
3131
3132         return (error);
3133 }
3134
3135 static void
3136 pf_clear_srcnodes(void)
3137 {
3138         struct pf_src_node      *n;
3139         struct pf_state         *state;
3140         globaldata_t save_gd = mycpu;
3141         int nn;
3142
3143         for (nn = 0; nn < ncpus; ++nn) {
3144                 lwkt_setcpu_self(globaldata_find(nn));
3145                 RB_FOREACH(state, pf_state_tree_id, &tree_id[nn]) {
3146                         state->src_node = NULL;
3147                         state->nat_src_node = NULL;
3148                 }
3149                 RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
3150                         n->expire = 1;
3151                         n->states = 0;
3152                 }
3153                 pf_purge_expired_src_nodes(0);
3154         }
3155         lwkt_setcpu_self(save_gd);
3156
3157         pf_status.src_nodes = 0;
3158 }
3159
3160 /*
3161  * XXX - Check for version missmatch!!!
3162  */
3163
3164 /*
3165  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3166  */
3167 static int
3168 shutdown_pf(void)
3169 {
3170         int error = 0;
3171         u_int32_t t[5];
3172         char nn = '\0';
3173
3174
3175         pf_status.running = 0;
3176         error = dehook_pf();
3177         if (error) {
3178                 pf_status.running = 1;
3179                 DPFPRINTF(PF_DEBUG_MISC,
3180                     ("pf: pfil unregistration failed\n"));
3181                 return(error);
3182         }
3183         do {
3184                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
3185                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3186                         break;
3187                 }
3188                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
3189                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3190                         break;          /* XXX: rollback? */
3191                 }
3192                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
3193                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3194                         break;          /* XXX: rollback? */
3195                 }
3196                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3197                     != 0) {
3198                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3199                         break;          /* XXX: rollback? */
3200                 }
3201                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3202                     != 0) {
3203                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3204                         break;          /* XXX: rollback? */
3205                 }
3206
3207                 /* XXX: these should always succeed here */
3208                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3209                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3210                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3211                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3212                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3213
3214                 if ((error = pf_clear_tables()) != 0)
3215                         break;
3216 #ifdef ALTQ
3217                 if ((error = pf_begin_altq(&t[0])) != 0) {
3218                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3219                         break;
3220                 }
3221                 pf_commit_altq(t[0]);
3222 #endif
3223                 pf_clear_states();
3224                 pf_clear_srcnodes();
3225
3226                 /* status does not use malloced mem so no need to cleanup */
3227                 /* fingerprints and interfaces have their own cleanup code */
3228         } while(0);
3229         return (error);
3230 }
3231
3232 static int
3233 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3234 {
3235         /*
3236          * DragonFly's version of pf uses FreeBSD's native host byte ordering
3237          * for ip_len/ip_off. This is why we don't have to change byte order
3238          * like the FreeBSD-5 version does.
3239          */
3240         int chk;
3241
3242         lwkt_gettoken_shared(&pf_token);
3243
3244         chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3245         if (chk && *m) {
3246                 m_freem(*m);
3247                 *m = NULL;
3248         }
3249         lwkt_reltoken(&pf_token);
3250         return chk;
3251 }
3252
3253 static int
3254 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3255 {
3256         /*
3257          * DragonFly's version of pf uses FreeBSD's native host byte ordering
3258          * for ip_len/ip_off. This is why we don't have to change byte order
3259          * like the FreeBSD-5 version does.
3260          */
3261         int chk;
3262
3263         lwkt_gettoken_shared(&pf_token);
3264
3265         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3266         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3267                 in_delayed_cksum(*m);
3268                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3269         }
3270         chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3271         if (chk && *m) {
3272                 m_freem(*m);
3273                 *m = NULL;
3274         }
3275         lwkt_reltoken(&pf_token);
3276         return chk;
3277 }
3278
3279 #ifdef INET6
3280 static int
3281 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3282 {
3283         /*
3284          * IPv6 is not affected by ip_len/ip_off byte order changes.
3285          */
3286         int chk;
3287
3288         lwkt_gettoken_shared(&pf_token);
3289
3290         chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3291         if (chk && *m) {
3292                 m_freem(*m);
3293                 *m = NULL;
3294         }
3295         lwkt_reltoken(&pf_token);
3296         return chk;
3297 }
3298
3299 static int
3300 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3301 {
3302         /*
3303          * IPv6 is not affected by ip_len/ip_off byte order changes.
3304          */
3305         int chk;
3306
3307         lwkt_gettoken_shared(&pf_token);
3308
3309         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3310         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3311                 in_delayed_cksum(*m);
3312                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3313         }
3314         chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3315         if (chk && *m) {
3316                 m_freem(*m);
3317                 *m = NULL;
3318         }
3319         lwkt_reltoken(&pf_token);
3320         return chk;
3321 }
3322 #endif /* INET6 */
3323
3324 static int
3325 hook_pf(void)
3326 {
3327         struct pfil_head *pfh_inet;
3328 #ifdef INET6
3329         struct pfil_head *pfh_inet6;
3330 #endif
3331
3332         lwkt_gettoken(&pf_token);
3333
3334         if (pf_pfil_hooked) {
3335                 lwkt_reltoken(&pf_token);
3336                 return (0);
3337         }
3338
3339         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3340         if (pfh_inet == NULL) {
3341                 lwkt_reltoken(&pf_token);
3342                 return (ENODEV);
3343         }
3344         pfil_add_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3345         pfil_add_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3346 #ifdef INET6
3347         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3348         if (pfh_inet6 == NULL) {
3349                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3350                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3351                 lwkt_reltoken(&pf_token);
3352                 return (ENODEV);
3353         }
3354         pfil_add_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3355         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3356 #endif
3357
3358         pf_pfil_hooked = 1;
3359         lwkt_reltoken(&pf_token);
3360         return (0);
3361 }
3362
3363 static int
3364 dehook_pf(void)
3365 {
3366         struct pfil_head *pfh_inet;
3367 #ifdef INET6
3368         struct pfil_head *pfh_inet6;
3369 #endif
3370
3371         lwkt_gettoken(&pf_token);
3372
3373         if (pf_pfil_hooked == 0) {
3374                 lwkt_reltoken(&pf_token);
3375                 return (0);
3376         }
3377
3378         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3379         if (pfh_inet == NULL) {
3380                 lwkt_reltoken(&pf_token);
3381                 return (ENODEV);
3382         }
3383         pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3384         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3385 #ifdef INET6
3386         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3387         if (pfh_inet6 == NULL) {
3388                 lwkt_reltoken(&pf_token);
3389                 return (ENODEV);
3390         }
3391         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3392         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3393 #endif
3394
3395         pf_pfil_hooked = 0;
3396         lwkt_reltoken(&pf_token);
3397         return (0);
3398 }
3399
3400 static int
3401 pf_load(void)
3402 {
3403         lwkt_gettoken(&pf_token);
3404
3405         pf_dev = make_dev(&pf_ops, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
3406         pfattach();
3407         lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3408         lockinit(&pf_global_statetbl_lock, "pfglstlk", 0, 0);
3409         lwkt_reltoken(&pf_token);
3410         return (0);
3411 }
3412
3413 static int
3414 pf_mask_del(struct radix_node *rn, void *arg)
3415 {
3416         struct radix_node_head *rnh = arg;
3417
3418         rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
3419         Free(rn);
3420         return 0;
3421 }
3422
3423 static int
3424 pf_unload(void)
3425 {
3426         int error;
3427         pf_status.running = 0;
3428
3429         lwkt_gettoken(&pf_token);
3430
3431         error = dehook_pf();
3432         if (error) {
3433                 /*
3434                  * Should not happen!
3435                  * XXX Due to error code ESRCH, kldunload will show
3436                  * a message like 'No such process'.
3437                  */
3438                 kprintf("pfil unregistration fail\n");
3439                 lwkt_reltoken(&pf_token);
3440                 return error;
3441         }
3442         shutdown_pf();
3443         pf_end_threads = 1;
3444         while (pf_end_threads < 2) {
3445                 wakeup_one(pf_purge_thread);
3446                 tsleep(pf_purge_thread, 0, "pftmo", hz);
3447         }
3448         pfi_cleanup();
3449         pf_osfp_flush();
3450         dev_ops_remove_all(&pf_ops);
3451         lockuninit(&pf_consistency_lock);
3452         lwkt_reltoken(&pf_token);
3453
3454         if (pf_maskhead != NULL) {
3455                 pf_maskhead->rnh_walktree(pf_maskhead,
3456                         pf_mask_del, pf_maskhead);
3457                 Free(pf_maskhead);
3458                 pf_maskhead = NULL;
3459         }
3460         kmalloc_destroy(&pf_state_pl);
3461         kmalloc_destroy(&pf_frent_pl);
3462         kmalloc_destroy(&pf_cent_pl);
3463         return 0;
3464 }
3465
3466 static int
3467 pf_modevent(module_t mod, int type, void *data __unused)
3468 {
3469         int error = 0;
3470
3471         lwkt_gettoken(&pf_token);
3472
3473         switch(type) {
3474         case MOD_LOAD:
3475                 error = pf_load();
3476                 break;
3477
3478         case MOD_UNLOAD:
3479                 error = pf_unload();
3480                 break;
3481         default:
3482                 error = EINVAL;
3483                 break;
3484         }
3485         lwkt_reltoken(&pf_token);
3486         return error;
3487 }
3488
3489 static moduledata_t pf_mod = {
3490         "pf",
3491         pf_modevent,
3492         0
3493 };
3494
3495 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3496 MODULE_VERSION(pf, PF_MODVER);