sys/netpfil/pf/pf_ioctl.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-2-Clause
   3  *
   4  * Copyright (c) 2001 Daniel Hartmeier
   5  * Copyright (c) 2002,2003 Henning Brauer
   6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  *
  13  *    - Redistributions of source code must retain the above copyright
  14  *      notice, this list of conditions and the following disclaimer.
  15  *    - Redistributions in binary form must reproduce the above
  16  *      copyright notice, this list of conditions and the following
  17  *      disclaimer in the documentation and/or other materials provided
  18  *      with the distribution.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  31  * POSSIBILITY OF SUCH DAMAGE.
  32  *
  33  * Effort sponsored in part by the Defense Advanced Research Projects
  34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  36  *
  37  *      $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
  38  */
  39
  40 #include <sys/cdefs.h>
  41 __FBSDID("$FreeBSD$");
  42
  43 #include "opt_inet.h"
  44 #include "opt_inet6.h"
  45 #include "opt_bpf.h"
  46 #include "opt_pf.h"
  47
  48 #include <sys/param.h>
  49 #include <sys/_bitset.h>
  50 #include <sys/bitset.h>
  51 #include <sys/bus.h>
  52 #include <sys/conf.h>
  53 #include <sys/endian.h>
  54 #include <sys/fcntl.h>
  55 #include <sys/filio.h>
  56 #include <sys/hash.h>
  57 #include <sys/interrupt.h>
  58 #include <sys/jail.h>
  59 #include <sys/kernel.h>
  60 #include <sys/kthread.h>
  61 #include <sys/lock.h>
  62 #include <sys/mbuf.h>
  63 #include <sys/module.h>
  64 #include <sys/nv.h>
  65 #include <sys/proc.h>
  66 #include <sys/smp.h>
  67 #include <sys/socket.h>
  68 #include <sys/sysctl.h>
  69 #include <sys/md5.h>
  70 #include <sys/ucred.h>
  71
  72 #include <net/if.h>
  73 #include <net/if_var.h>
  74 #include <net/vnet.h>
  75 #include <net/route.h>
  76 #include <net/pfil.h>
  77 #include <net/pfvar.h>
  78 #include <net/if_pfsync.h>
  79 #include <net/if_pflog.h>
  80
  81 #include <netinet/in.h>
  82 #include <netinet/ip.h>
  83 #include <netinet/ip_var.h>
  84 #include <netinet6/ip6_var.h>
  85 #include <netinet/ip_icmp.h>
  86 #include <netpfil/pf/pf_nv.h>
  87
  88 #ifdef INET6
  89 #include <netinet/ip6.h>
  90 #endif /* INET6 */
  91
  92 #ifdef ALTQ
  93 #include <net/altq/altq.h>
  94 #endif
  95
  96 static struct pf_kpool  *pf_get_kpool(char *, u_int32_t, u_int8_t, u_int32_t,
  97                             u_int8_t, u_int8_t, u_int8_t);
  98
  99 static void              pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
 100 static void              pf_empty_kpool(struct pf_kpalist *);
 101 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
 102                             struct thread *);
 103 #ifdef ALTQ
 104 static int               pf_begin_altq(u_int32_t *);
 105 static int               pf_rollback_altq(u_int32_t);
 106 static int               pf_commit_altq(u_int32_t);
 107 static int               pf_enable_altq(struct pf_altq *);
 108 static int               pf_disable_altq(struct pf_altq *);
 109 static u_int32_t         pf_qname2qid(char *);
 110 static void              pf_qid_unref(u_int32_t);
 111 #endif /* ALTQ */
 112 static int               pf_begin_rules(u_int32_t *, int, const char *);
 113 static int               pf_rollback_rules(u_int32_t, int, char *);
 114 static int               pf_setup_pfsync_matching(struct pf_kruleset *);
 115 static void              pf_hash_rule(MD5_CTX *, struct pf_krule *);
 116 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 117 static int               pf_commit_rules(u_int32_t, int, char *);
 118 static int               pf_addr_setup(struct pf_kruleset *,
 119                             struct pf_addr_wrap *, sa_family_t);
 120 static void              pf_addr_copyout(struct pf_addr_wrap *);
 121 static void              pf_src_node_copy(const struct pf_ksrc_node *,
 122                             struct pf_src_node *);
 123 #ifdef ALTQ
 124 static int               pf_export_kaltq(struct pf_altq *,
 125                             struct pfioc_altq_v1 *, size_t);
 126 static int               pf_import_kaltq(struct pfioc_altq_v1 *,
 127                             struct pf_altq *, size_t);
 128 #endif /* ALTQ */
 129
 130 VNET_DEFINE(struct pf_krule,    pf_default_rule);
 131
 132 #ifdef ALTQ
 133 VNET_DEFINE_STATIC(int,         pf_altq_running);
 134 #define V_pf_altq_running       VNET(pf_altq_running)
 135 #endif
 136
 137 #define TAGID_MAX        50000
 138 struct pf_tagname {
 139         TAILQ_ENTRY(pf_tagname) namehash_entries;
 140         TAILQ_ENTRY(pf_tagname) taghash_entries;
 141         char                    name[PF_TAG_NAME_SIZE];
 142         uint16_t                tag;
 143         int                     ref;
 144 };
 145
 146 struct pf_tagset {
 147         TAILQ_HEAD(, pf_tagname)        *namehash;
 148         TAILQ_HEAD(, pf_tagname)        *taghash;
 149         unsigned int                     mask;
 150         uint32_t                         seed;
 151         BITSET_DEFINE(, TAGID_MAX)       avail;
 152 };
 153
 154 VNET_DEFINE(struct pf_tagset, pf_tags);
 155 #define V_pf_tags       VNET(pf_tags)
 156 static unsigned int     pf_rule_tag_hashsize;
 157 #define PF_RULE_TAG_HASH_SIZE_DEFAULT   128
 158 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
 159     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
 160     "Size of pf(4) rule tag hashtable");
 161
 162 #ifdef ALTQ
 163 VNET_DEFINE(struct pf_tagset, pf_qids);
 164 #define V_pf_qids       VNET(pf_qids)
 165 static unsigned int     pf_queue_tag_hashsize;
 166 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT  128
 167 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
 168     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
 169     "Size of pf(4) queue tag hashtable");
 170 #endif
 171 VNET_DEFINE(uma_zone_t,  pf_tag_z);
 172 #define V_pf_tag_z               VNET(pf_tag_z)
 173 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 174 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 175
 176 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 177 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 178 #endif
 179
 180 static void              pf_init_tagset(struct pf_tagset *, unsigned int *,
 181                             unsigned int);
 182 static void              pf_cleanup_tagset(struct pf_tagset *);
 183 static uint16_t          tagname2hashindex(const struct pf_tagset *, const char *);
 184 static uint16_t          tag2hashindex(const struct pf_tagset *, uint16_t);
 185 static u_int16_t         tagname2tag(struct pf_tagset *, char *);
 186 static u_int16_t         pf_tagname2tag(char *);
 187 static void              tag_unref(struct pf_tagset *, u_int16_t);
 188
 189 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 190
 191 struct cdev *pf_dev;
 192
 193 /*
 194  * XXX - These are new and need to be checked when moveing to a new version
 195  */
 196 static void              pf_clear_states(void);
 197 static int               pf_clear_tables(void);
 198 static void              pf_clear_srcnodes(struct pf_ksrc_node *);
 199 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 200 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 201
 202 /*
 203  * Wrapper functions for pfil(9) hooks
 204  */
 205 #ifdef INET
 206 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
 207     int flags, void *ruleset __unused, struct inpcb *inp);
 208 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
 209     int flags, void *ruleset __unused, struct inpcb *inp);
 210 #endif
 211 #ifdef INET6
 212 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
 213     int flags, void *ruleset __unused, struct inpcb *inp);
 214 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
 215     int flags, void *ruleset __unused, struct inpcb *inp);
 216 #endif
 217
 218 static void             hook_pf(void);
 219 static void             dehook_pf(void);
 220 static int              shutdown_pf(void);
 221 static int              pf_load(void);
 222 static void             pf_unload(void);
 223
 224 static struct cdevsw pf_cdevsw = {
 225         .d_ioctl =      pfioctl,
 226         .d_name =       PF_NAME,
 227         .d_version =    D_VERSION,
 228 };
 229
 230 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
 231 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 232
 233 /*
 234  * We need a flag that is neither hooked nor running to know when
 235  * the VNET is "valid".  We primarily need this to control (global)
 236  * external event, e.g., eventhandlers.
 237  */
 238 VNET_DEFINE(int, pf_vnet_active);
 239 #define V_pf_vnet_active        VNET(pf_vnet_active)
 240
 241 int pf_end_threads;
 242 struct proc *pf_purge_proc;
 243
 244 struct rmlock                   pf_rules_lock;
 245 struct sx                       pf_ioctl_lock;
 246 struct sx                       pf_end_lock;
 247
 248 /* pfsync */
 249 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
 250 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
 251 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
 252 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
 253 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
 254 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
 255 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
 256
 257 /* pflog */
 258 pflog_packet_t                  *pflog_packet_ptr = NULL;
 259
 260 extern u_long   pf_ioctl_maxcount;
 261
 262 static void
 263 pfattach_vnet(void)
 264 {
 265         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 266
 267         pf_initialize();
 268         pfr_initialize();
 269         pfi_initialize_vnet();
 270         pf_normalize_init();
 271
 272         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 273         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 274
 275         RB_INIT(&V_pf_anchors);
 276         pf_init_kruleset(&pf_main_ruleset);
 277
 278         /* default rule should never be garbage collected */
 279         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 280 #ifdef PF_DEFAULT_TO_DROP
 281         V_pf_default_rule.action = PF_DROP;
 282 #else
 283         V_pf_default_rule.action = PF_PASS;
 284 #endif
 285         V_pf_default_rule.nr = -1;
 286         V_pf_default_rule.rtableid = -1;
 287
 288         V_pf_default_rule.evaluations = counter_u64_alloc(M_WAITOK);
 289         for (int i = 0; i < 2; i++) {
 290                 V_pf_default_rule.packets[i] = counter_u64_alloc(M_WAITOK);
 291                 V_pf_default_rule.bytes[i] = counter_u64_alloc(M_WAITOK);
 292         }
 293         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 294         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 295         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 296
 297         /* initialize default timeouts */
 298         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 299         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 300         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 301         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 302         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 303         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 304         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 305         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 306         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 307         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 308         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 309         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 310         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 311         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 312         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 313         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 314         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 315         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 316         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 317         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 318
 319         bzero(&V_pf_status, sizeof(V_pf_status));
 320         V_pf_status.debug = PF_DEBUG_URGENT;
 321
 322         V_pf_pfil_hooked = 0;
 323
 324         /* XXX do our best to avoid a conflict */
 325         V_pf_status.hostid = arc4random();
 326
 327         for (int i = 0; i < PFRES_MAX; i++)
 328                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 329         for (int i = 0; i < LCNT_MAX; i++)
 330                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 331         for (int i = 0; i < FCNT_MAX; i++)
 332                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 333         for (int i = 0; i < SCNT_MAX; i++)
 334                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 335
 336         if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
 337             INTR_MPSAFE, &V_pf_swi_cookie) != 0)
 338                 /* XXXGL: leaked all above. */
 339                 return;
 340 }
 341
 342 static struct pf_kpool *
 343 pf_get_kpool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 344     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 345     u_int8_t check_ticket)
 346 {
 347         struct pf_kruleset      *ruleset;
 348         struct pf_krule         *rule;
 349         int                      rs_num;
 350
 351         ruleset = pf_find_kruleset(anchor);
 352         if (ruleset == NULL)
 353                 return (NULL);
 354         rs_num = pf_get_ruleset_number(rule_action);
 355         if (rs_num >= PF_RULESET_MAX)
 356                 return (NULL);
 357         if (active) {
 358                 if (check_ticket && ticket !=
 359                     ruleset->rules[rs_num].active.ticket)
 360                         return (NULL);
 361                 if (r_last)
 362                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 363                             pf_krulequeue);
 364                 else
 365                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 366         } else {
 367                 if (check_ticket && ticket !=
 368                     ruleset->rules[rs_num].inactive.ticket)
 369                         return (NULL);
 370                 if (r_last)
 371                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 372                             pf_krulequeue);
 373                 else
 374                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 375         }
 376         if (!r_last) {
 377                 while ((rule != NULL) && (rule->nr != rule_number))
 378                         rule = TAILQ_NEXT(rule, entries);
 379         }
 380         if (rule == NULL)
 381                 return (NULL);
 382
 383         return (&rule->rpool);
 384 }
 385
 386 static void
 387 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
 388 {
 389         struct pf_kpooladdr     *mv_pool_pa;
 390
 391         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 392                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 393                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 394         }
 395 }
 396
 397 static void
 398 pf_empty_kpool(struct pf_kpalist *poola)
 399 {
 400         struct pf_kpooladdr *pa;
 401
 402         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 403                 switch (pa->addr.type) {
 404                 case PF_ADDR_DYNIFTL:
 405                         pfi_dynaddr_remove(pa->addr.p.dyn);
 406                         break;
 407                 case PF_ADDR_TABLE:
 408                         /* XXX: this could be unfinished pooladdr on pabuf */
 409                         if (pa->addr.p.tbl != NULL)
 410                                 pfr_detach_table(pa->addr.p.tbl);
 411                         break;
 412                 }
 413                 if (pa->kif)
 414                         pfi_kkif_unref(pa->kif);
 415                 TAILQ_REMOVE(poola, pa, entries);
 416                 free(pa, M_PFRULE);
 417         }
 418 }
 419
 420 static void
 421 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
 422 {
 423
 424         PF_RULES_WASSERT();
 425
 426         TAILQ_REMOVE(rulequeue, rule, entries);
 427
 428         PF_UNLNKDRULES_LOCK();
 429         rule->rule_flag |= PFRULE_REFS;
 430         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 431         PF_UNLNKDRULES_UNLOCK();
 432 }
 433
 434 void
 435 pf_free_rule(struct pf_krule *rule)
 436 {
 437
 438         PF_RULES_WASSERT();
 439
 440         if (rule->tag)
 441                 tag_unref(&V_pf_tags, rule->tag);
 442         if (rule->match_tag)
 443                 tag_unref(&V_pf_tags, rule->match_tag);
 444 #ifdef ALTQ
 445         if (rule->pqid != rule->qid)
 446                 pf_qid_unref(rule->pqid);
 447         pf_qid_unref(rule->qid);
 448 #endif
 449         switch (rule->src.addr.type) {
 450         case PF_ADDR_DYNIFTL:
 451                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 452                 break;
 453         case PF_ADDR_TABLE:
 454                 pfr_detach_table(rule->src.addr.p.tbl);
 455                 break;
 456         }
 457         switch (rule->dst.addr.type) {
 458         case PF_ADDR_DYNIFTL:
 459                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 460                 break;
 461         case PF_ADDR_TABLE:
 462                 pfr_detach_table(rule->dst.addr.p.tbl);
 463                 break;
 464         }
 465         if (rule->overload_tbl)
 466                 pfr_detach_table(rule->overload_tbl);
 467         if (rule->kif)
 468                 pfi_kkif_unref(rule->kif);
 469         pf_kanchor_remove(rule);
 470         pf_empty_kpool(&rule->rpool.list);
 471
 472         pf_krule_free(rule);
 473 }
 474
 475 static void
 476 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
 477     unsigned int default_size)
 478 {
 479         unsigned int i;
 480         unsigned int hashsize;
 481
 482         if (*tunable_size == 0 || !powerof2(*tunable_size))
 483                 *tunable_size = default_size;
 484
 485         hashsize = *tunable_size;
 486         ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
 487             M_WAITOK);
 488         ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
 489             M_WAITOK);
 490         ts->mask = hashsize - 1;
 491         ts->seed = arc4random();
 492         for (i = 0; i < hashsize; i++) {
 493                 TAILQ_INIT(&ts->namehash[i]);
 494                 TAILQ_INIT(&ts->taghash[i]);
 495         }
 496         BIT_FILL(TAGID_MAX, &ts->avail);
 497 }
 498
 499 static void
 500 pf_cleanup_tagset(struct pf_tagset *ts)
 501 {
 502         unsigned int i;
 503         unsigned int hashsize;
 504         struct pf_tagname *t, *tmp;
 505
 506         /*
 507          * Only need to clean up one of the hashes as each tag is hashed
 508          * into each table.
 509          */
 510         hashsize = ts->mask + 1;
 511         for (i = 0; i < hashsize; i++)
 512                 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
 513                         uma_zfree(V_pf_tag_z, t);
 514
 515         free(ts->namehash, M_PFHASH);
 516         free(ts->taghash, M_PFHASH);
 517 }
 518
 519 static uint16_t
 520 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
 521 {
 522         size_t len;
 523
 524         len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
 525         return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
 526 }
 527
 528 static uint16_t
 529 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
 530 {
 531
 532         return (tag & ts->mask);
 533 }
 534
 535 static u_int16_t
 536 tagname2tag(struct pf_tagset *ts, char *tagname)
 537 {
 538         struct pf_tagname       *tag;
 539         u_int32_t                index;
 540         u_int16_t                new_tagid;
 541
 542         PF_RULES_WASSERT();
 543
 544         index = tagname2hashindex(ts, tagname);
 545         TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
 546                 if (strcmp(tagname, tag->name) == 0) {
 547                         tag->ref++;
 548                         return (tag->tag);
 549                 }
 550
 551         /*
 552          * new entry
 553          *
 554          * to avoid fragmentation, we do a linear search from the beginning
 555          * and take the first free slot we find.
 556          */
 557         new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
 558         /*
 559          * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
 560          * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
 561          * set.  It may also return a bit number greater than TAGID_MAX due
 562          * to rounding of the number of bits in the vector up to a multiple
 563          * of the vector word size at declaration/allocation time.
 564          */
 565         if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
 566                 return (0);
 567
 568         /* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
 569         BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
 570
 571         /* allocate and fill new struct pf_tagname */
 572         tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
 573         if (tag == NULL)
 574                 return (0);
 575         strlcpy(tag->name, tagname, sizeof(tag->name));
 576         tag->tag = new_tagid;
 577         tag->ref = 1;
 578
 579         /* Insert into namehash */
 580         TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
 581
 582         /* Insert into taghash */
 583         index = tag2hashindex(ts, new_tagid);
 584         TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
 585
 586         return (tag->tag);
 587 }
 588
 589 static void
 590 tag_unref(struct pf_tagset *ts, u_int16_t tag)
 591 {
 592         struct pf_tagname       *t;
 593         uint16_t                 index;
 594
 595         PF_RULES_WASSERT();
 596
 597         index = tag2hashindex(ts, tag);
 598         TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
 599                 if (tag == t->tag) {
 600                         if (--t->ref == 0) {
 601                                 TAILQ_REMOVE(&ts->taghash[index], t,
 602                                     taghash_entries);
 603                                 index = tagname2hashindex(ts, t->name);
 604                                 TAILQ_REMOVE(&ts->namehash[index], t,
 605                                     namehash_entries);
 606                                 /* Bits are 0-based for BIT_SET() */
 607                                 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
 608                                 uma_zfree(V_pf_tag_z, t);
 609                         }
 610                         break;
 611                 }
 612 }
 613
 614 static u_int16_t
 615 pf_tagname2tag(char *tagname)
 616 {
 617         return (tagname2tag(&V_pf_tags, tagname));
 618 }
 619
 620 #ifdef ALTQ
 621 static u_int32_t
 622 pf_qname2qid(char *qname)
 623 {
 624         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 625 }
 626
 627 static void
 628 pf_qid_unref(u_int32_t qid)
 629 {
 630         tag_unref(&V_pf_qids, (u_int16_t)qid);
 631 }
 632
 633 static int
 634 pf_begin_altq(u_int32_t *ticket)
 635 {
 636         struct pf_altq  *altq, *tmp;
 637         int              error = 0;
 638
 639         PF_RULES_WASSERT();
 640
 641         /* Purge the old altq lists */
 642         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 643                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 644                         /* detach and destroy the discipline */
 645                         error = altq_remove(altq);
 646                 }
 647                 free(altq, M_PFALTQ);
 648         }
 649         TAILQ_INIT(V_pf_altq_ifs_inactive);
 650         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 651                 pf_qid_unref(altq->qid);
 652                 free(altq, M_PFALTQ);
 653         }
 654         TAILQ_INIT(V_pf_altqs_inactive);
 655         if (error)
 656                 return (error);
 657         *ticket = ++V_ticket_altqs_inactive;
 658         V_altqs_inactive_open = 1;
 659         return (0);
 660 }
 661
 662 static int
 663 pf_rollback_altq(u_int32_t ticket)
 664 {
 665         struct pf_altq  *altq, *tmp;
 666         int              error = 0;
 667
 668         PF_RULES_WASSERT();
 669
 670         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 671                 return (0);
 672         /* Purge the old altq lists */
 673         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 674                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 675                         /* detach and destroy the discipline */
 676                         error = altq_remove(altq);
 677                 }
 678                 free(altq, M_PFALTQ);
 679         }
 680         TAILQ_INIT(V_pf_altq_ifs_inactive);
 681         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 682                 pf_qid_unref(altq->qid);
 683                 free(altq, M_PFALTQ);
 684         }
 685         TAILQ_INIT(V_pf_altqs_inactive);
 686         V_altqs_inactive_open = 0;
 687         return (error);
 688 }
 689
 690 static int
 691 pf_commit_altq(u_int32_t ticket)
 692 {
 693         struct pf_altqqueue     *old_altqs, *old_altq_ifs;
 694         struct pf_altq          *altq, *tmp;
 695         int                      err, error = 0;
 696
 697         PF_RULES_WASSERT();
 698
 699         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 700                 return (EBUSY);
 701
 702         /* swap altqs, keep the old. */
 703         old_altqs = V_pf_altqs_active;
 704         old_altq_ifs = V_pf_altq_ifs_active;
 705         V_pf_altqs_active = V_pf_altqs_inactive;
 706         V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
 707         V_pf_altqs_inactive = old_altqs;
 708         V_pf_altq_ifs_inactive = old_altq_ifs;
 709         V_ticket_altqs_active = V_ticket_altqs_inactive;
 710
 711         /* Attach new disciplines */
 712         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
 713                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 714                         /* attach the discipline */
 715                         error = altq_pfattach(altq);
 716                         if (error == 0 && V_pf_altq_running)
 717                                 error = pf_enable_altq(altq);
 718                         if (error != 0)
 719                                 return (error);
 720                 }
 721         }
 722
 723         /* Purge the old altq lists */
 724         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 725                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 726                         /* detach and destroy the discipline */
 727                         if (V_pf_altq_running)
 728                                 error = pf_disable_altq(altq);
 729                         err = altq_pfdetach(altq);
 730                         if (err != 0 && error == 0)
 731                                 error = err;
 732                         err = altq_remove(altq);
 733                         if (err != 0 && error == 0)
 734                                 error = err;
 735                 }
 736                 free(altq, M_PFALTQ);
 737         }
 738         TAILQ_INIT(V_pf_altq_ifs_inactive);
 739         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 740                 pf_qid_unref(altq->qid);
 741                 free(altq, M_PFALTQ);
 742         }
 743         TAILQ_INIT(V_pf_altqs_inactive);
 744
 745         V_altqs_inactive_open = 0;
 746         return (error);
 747 }
 748
 749 static int
 750 pf_enable_altq(struct pf_altq *altq)
 751 {
 752         struct ifnet            *ifp;
 753         struct tb_profile        tb;
 754         int                      error = 0;
 755
 756         if ((ifp = ifunit(altq->ifname)) == NULL)
 757                 return (EINVAL);
 758
 759         if (ifp->if_snd.altq_type != ALTQT_NONE)
 760                 error = altq_enable(&ifp->if_snd);
 761
 762         /* set tokenbucket regulator */
 763         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 764                 tb.rate = altq->ifbandwidth;
 765                 tb.depth = altq->tbrsize;
 766                 error = tbr_set(&ifp->if_snd, &tb);
 767         }
 768
 769         return (error);
 770 }
 771
 772 static int
 773 pf_disable_altq(struct pf_altq *altq)
 774 {
 775         struct ifnet            *ifp;
 776         struct tb_profile        tb;
 777         int                      error;
 778
 779         if ((ifp = ifunit(altq->ifname)) == NULL)
 780                 return (EINVAL);
 781
 782         /*
 783          * when the discipline is no longer referenced, it was overridden
 784          * by a new one.  if so, just return.
 785          */
 786         if (altq->altq_disc != ifp->if_snd.altq_disc)
 787                 return (0);
 788
 789         error = altq_disable(&ifp->if_snd);
 790
 791         if (error == 0) {
 792                 /* clear tokenbucket regulator */
 793                 tb.rate = 0;
 794                 error = tbr_set(&ifp->if_snd, &tb);
 795         }
 796
 797         return (error);
 798 }
 799
 800 static int
 801 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
 802     struct pf_altq *altq)
 803 {
 804         struct ifnet    *ifp1;
 805         int              error = 0;
 806
 807         /* Deactivate the interface in question */
 808         altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 809         if ((ifp1 = ifunit(altq->ifname)) == NULL ||
 810             (remove && ifp1 == ifp)) {
 811                 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 812         } else {
 813                 error = altq_add(ifp1, altq);
 814
 815                 if (ticket != V_ticket_altqs_inactive)
 816                         error = EBUSY;
 817
 818                 if (error)
 819                         free(altq, M_PFALTQ);
 820         }
 821
 822         return (error);
 823 }
 824
 825 void
 826 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 827 {
 828         struct pf_altq  *a1, *a2, *a3;
 829         u_int32_t        ticket;
 830         int              error = 0;
 831
 832         /*
 833          * No need to re-evaluate the configuration for events on interfaces
 834          * that do not support ALTQ, as it's not possible for such
 835          * interfaces to be part of the configuration.
 836          */
 837         if (!ALTQ_IS_READY(&ifp->if_snd))
 838                 return;
 839
 840         /* Interrupt userland queue modifications */
 841         if (V_altqs_inactive_open)
 842                 pf_rollback_altq(V_ticket_altqs_inactive);
 843
 844         /* Start new altq ruleset */
 845         if (pf_begin_altq(&ticket))
 846                 return;
 847
 848         /* Copy the current active set */
 849         TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
 850                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 851                 if (a2 == NULL) {
 852                         error = ENOMEM;
 853                         break;
 854                 }
 855                 bcopy(a1, a2, sizeof(struct pf_altq));
 856
 857                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 858                 if (error)
 859                         break;
 860
 861                 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
 862         }
 863         if (error)
 864                 goto out;
 865         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 866                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 867                 if (a2 == NULL) {
 868                         error = ENOMEM;
 869                         break;
 870                 }
 871                 bcopy(a1, a2, sizeof(struct pf_altq));
 872
 873                 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 874                         error = EBUSY;
 875                         free(a2, M_PFALTQ);
 876                         break;
 877                 }
 878                 a2->altq_disc = NULL;
 879                 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
 880                         if (strncmp(a3->ifname, a2->ifname,
 881                                 IFNAMSIZ) == 0) {
 882                                 a2->altq_disc = a3->altq_disc;
 883                                 break;
 884                         }
 885                 }
 886                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 887                 if (error)
 888                         break;
 889
 890                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 891         }
 892
 893 out:
 894         if (error != 0)
 895                 pf_rollback_altq(ticket);
 896         else
 897                 pf_commit_altq(ticket);
 898 }
 899 #endif /* ALTQ */
 900
 901 static int
 902 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 903 {
 904         struct pf_kruleset      *rs;
 905         struct pf_krule         *rule;
 906
 907         PF_RULES_WASSERT();
 908
 909         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 910                 return (EINVAL);
 911         rs = pf_find_or_create_kruleset(anchor);
 912         if (rs == NULL)
 913                 return (EINVAL);
 914         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 915                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 916                 rs->rules[rs_num].inactive.rcount--;
 917         }
 918         *ticket = ++rs->rules[rs_num].inactive.ticket;
 919         rs->rules[rs_num].inactive.open = 1;
 920         return (0);
 921 }
 922
 923 static int
 924 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 925 {
 926         struct pf_kruleset      *rs;
 927         struct pf_krule         *rule;
 928
 929         PF_RULES_WASSERT();
 930
 931         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 932                 return (EINVAL);
 933         rs = pf_find_kruleset(anchor);
 934         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 935             rs->rules[rs_num].inactive.ticket != ticket)
 936                 return (0);
 937         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 938                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 939                 rs->rules[rs_num].inactive.rcount--;
 940         }
 941         rs->rules[rs_num].inactive.open = 0;
 942         return (0);
 943 }
 944
 945 #define PF_MD5_UPD(st, elm)                                             \
 946                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 947
 948 #define PF_MD5_UPD_STR(st, elm)                                         \
 949                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 950
 951 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 952                 (stor) = htonl((st)->elm);                              \
 953                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 954 } while (0)
 955
 956 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 957                 (stor) = htons((st)->elm);                              \
 958                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 959 } while (0)
 960
 961 static void
 962 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 963 {
 964         PF_MD5_UPD(pfr, addr.type);
 965         switch (pfr->addr.type) {
 966                 case PF_ADDR_DYNIFTL:
 967                         PF_MD5_UPD(pfr, addr.v.ifname);
 968                         PF_MD5_UPD(pfr, addr.iflags);
 969                         break;
 970                 case PF_ADDR_TABLE:
 971                         PF_MD5_UPD(pfr, addr.v.tblname);
 972                         break;
 973                 case PF_ADDR_ADDRMASK:
 974                         /* XXX ignore af? */
 975                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 976                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 977                         break;
 978         }
 979
 980         PF_MD5_UPD(pfr, port[0]);
 981         PF_MD5_UPD(pfr, port[1]);
 982         PF_MD5_UPD(pfr, neg);
 983         PF_MD5_UPD(pfr, port_op);
 984 }
 985
 986 static void
 987 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
 988 {
 989         u_int16_t x;
 990         u_int32_t y;
 991
 992         pf_hash_rule_addr(ctx, &rule->src);
 993         pf_hash_rule_addr(ctx, &rule->dst);
 994         PF_MD5_UPD_STR(rule, label);
 995         PF_MD5_UPD_STR(rule, ifname);
 996         PF_MD5_UPD_STR(rule, match_tagname);
 997         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 998         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 999         PF_MD5_UPD_HTONL(rule, prob, y);
1000         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1001         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1002         PF_MD5_UPD(rule, uid.op);
1003         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1004         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1005         PF_MD5_UPD(rule, gid.op);
1006         PF_MD5_UPD_HTONL(rule, rule_flag, y);
1007         PF_MD5_UPD(rule, action);
1008         PF_MD5_UPD(rule, direction);
1009         PF_MD5_UPD(rule, af);
1010         PF_MD5_UPD(rule, quick);
1011         PF_MD5_UPD(rule, ifnot);
1012         PF_MD5_UPD(rule, match_tag_not);
1013         PF_MD5_UPD(rule, natpass);
1014         PF_MD5_UPD(rule, keep_state);
1015         PF_MD5_UPD(rule, proto);
1016         PF_MD5_UPD(rule, type);
1017         PF_MD5_UPD(rule, code);
1018         PF_MD5_UPD(rule, flags);
1019         PF_MD5_UPD(rule, flagset);
1020         PF_MD5_UPD(rule, allow_opts);
1021         PF_MD5_UPD(rule, rt);
1022         PF_MD5_UPD(rule, tos);
1023 }
1024
1025 static int
1026 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1027 {
1028         struct pf_kruleset      *rs;
1029         struct pf_krule         *rule, **old_array;
1030         struct pf_krulequeue    *old_rules;
1031         int                      error;
1032         u_int32_t                old_rcount;
1033
1034         PF_RULES_WASSERT();
1035
1036         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1037                 return (EINVAL);
1038         rs = pf_find_kruleset(anchor);
1039         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1040             ticket != rs->rules[rs_num].inactive.ticket)
1041                 return (EBUSY);
1042
1043         /* Calculate checksum for the main ruleset */
1044         if (rs == &pf_main_ruleset) {
1045                 error = pf_setup_pfsync_matching(rs);
1046                 if (error != 0)
1047                         return (error);
1048         }
1049
1050         /* Swap rules, keep the old. */
1051         old_rules = rs->rules[rs_num].active.ptr;
1052         old_rcount = rs->rules[rs_num].active.rcount;
1053         old_array = rs->rules[rs_num].active.ptr_array;
1054
1055         rs->rules[rs_num].active.ptr =
1056             rs->rules[rs_num].inactive.ptr;
1057         rs->rules[rs_num].active.ptr_array =
1058             rs->rules[rs_num].inactive.ptr_array;
1059         rs->rules[rs_num].active.rcount =
1060             rs->rules[rs_num].inactive.rcount;
1061         rs->rules[rs_num].inactive.ptr = old_rules;
1062         rs->rules[rs_num].inactive.ptr_array = old_array;
1063         rs->rules[rs_num].inactive.rcount = old_rcount;
1064
1065         rs->rules[rs_num].active.ticket =
1066             rs->rules[rs_num].inactive.ticket;
1067         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1068
1069         /* Purge the old rule list. */
1070         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1071                 pf_unlink_rule(old_rules, rule);
1072         if (rs->rules[rs_num].inactive.ptr_array)
1073                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1074         rs->rules[rs_num].inactive.ptr_array = NULL;
1075         rs->rules[rs_num].inactive.rcount = 0;
1076         rs->rules[rs_num].inactive.open = 0;
1077         pf_remove_if_empty_kruleset(rs);
1078
1079         return (0);
1080 }
1081
1082 static int
1083 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1084 {
1085         MD5_CTX                  ctx;
1086         struct pf_krule         *rule;
1087         int                      rs_cnt;
1088         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
1089
1090         MD5Init(&ctx);
1091         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1092                 /* XXX PF_RULESET_SCRUB as well? */
1093                 if (rs_cnt == PF_RULESET_SCRUB)
1094                         continue;
1095
1096                 if (rs->rules[rs_cnt].inactive.ptr_array)
1097                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1098                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1099
1100                 if (rs->rules[rs_cnt].inactive.rcount) {
1101                         rs->rules[rs_cnt].inactive.ptr_array =
1102                             malloc(sizeof(caddr_t) *
1103                             rs->rules[rs_cnt].inactive.rcount,
1104                             M_TEMP, M_NOWAIT);
1105
1106                         if (!rs->rules[rs_cnt].inactive.ptr_array)
1107                                 return (ENOMEM);
1108                 }
1109
1110                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1111                     entries) {
1112                         pf_hash_rule(&ctx, rule);
1113                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1114                 }
1115         }
1116
1117         MD5Final(digest, &ctx);
1118         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1119         return (0);
1120 }
1121
1122 static int
1123 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1124     sa_family_t af)
1125 {
1126         int error = 0;
1127
1128         switch (addr->type) {
1129         case PF_ADDR_TABLE:
1130                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1131                 if (addr->p.tbl == NULL)
1132                         error = ENOMEM;
1133                 break;
1134         case PF_ADDR_DYNIFTL:
1135                 error = pfi_dynaddr_setup(addr, af);
1136                 break;
1137         }
1138
1139         return (error);
1140 }
1141
1142 static void
1143 pf_addr_copyout(struct pf_addr_wrap *addr)
1144 {
1145
1146         switch (addr->type) {
1147         case PF_ADDR_DYNIFTL:
1148                 pfi_dynaddr_copyout(addr);
1149                 break;
1150         case PF_ADDR_TABLE:
1151                 pf_tbladdr_copyout(addr);
1152                 break;
1153         }
1154 }
1155
1156 static void
1157 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1158 {
1159         int     secs = time_uptime, diff;
1160
1161         bzero(out, sizeof(struct pf_src_node));
1162
1163         bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1164         bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1165
1166         if (in->rule.ptr != NULL)
1167                 out->rule.nr = in->rule.ptr->nr;
1168
1169         for (int i = 0; i < 2; i++) {
1170                 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1171                 out->packets[i] = counter_u64_fetch(in->packets[i]);
1172         }
1173
1174         out->states = in->states;
1175         out->conn = in->conn;
1176         out->af = in->af;
1177         out->ruletype = in->ruletype;
1178
1179         out->creation = secs - in->creation;
1180         if (out->expire > secs)
1181                 out->expire -= secs;
1182         else
1183                 out->expire = 0;
1184
1185         /* Adjust the connection rate estimate. */
1186         diff = secs - in->conn_rate.last;
1187         if (diff >= in->conn_rate.seconds)
1188                 out->conn_rate.count = 0;
1189         else
1190                 out->conn_rate.count -=
1191                     in->conn_rate.count * diff /
1192                     in->conn_rate.seconds;
1193 }
1194
1195 #ifdef ALTQ
1196 /*
1197  * Handle export of struct pf_kaltq to user binaries that may be using any
1198  * version of struct pf_altq.
1199  */
1200 static int
1201 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1202 {
1203         u_int32_t version;
1204
1205         if (ioc_size == sizeof(struct pfioc_altq_v0))
1206                 version = 0;
1207         else
1208                 version = pa->version;
1209
1210         if (version > PFIOC_ALTQ_VERSION)
1211                 return (EINVAL);
1212
1213 #define ASSIGN(x) exported_q->x = q->x
1214 #define COPY(x) \
1215         bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1216 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1217 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1218
1219         switch (version) {
1220         case 0: {
1221                 struct pf_altq_v0 *exported_q =
1222                     &((struct pfioc_altq_v0 *)pa)->altq;
1223
1224                 COPY(ifname);
1225
1226                 ASSIGN(scheduler);
1227                 ASSIGN(tbrsize);
1228                 exported_q->tbrsize = SATU16(q->tbrsize);
1229                 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1230
1231                 COPY(qname);
1232                 COPY(parent);
1233                 ASSIGN(parent_qid);
1234                 exported_q->bandwidth = SATU32(q->bandwidth);
1235                 ASSIGN(priority);
1236                 ASSIGN(local_flags);
1237
1238                 ASSIGN(qlimit);
1239                 ASSIGN(flags);
1240
1241                 if (q->scheduler == ALTQT_HFSC) {
1242 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1243 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1244                             SATU32(q->pq_u.hfsc_opts.x)
1245
1246                         ASSIGN_OPT_SATU32(rtsc_m1);
1247                         ASSIGN_OPT(rtsc_d);
1248                         ASSIGN_OPT_SATU32(rtsc_m2);
1249
1250                         ASSIGN_OPT_SATU32(lssc_m1);
1251                         ASSIGN_OPT(lssc_d);
1252                         ASSIGN_OPT_SATU32(lssc_m2);
1253
1254                         ASSIGN_OPT_SATU32(ulsc_m1);
1255                         ASSIGN_OPT(ulsc_d);
1256                         ASSIGN_OPT_SATU32(ulsc_m2);
1257
1258                         ASSIGN_OPT(flags);
1259
1260 #undef ASSIGN_OPT
1261 #undef ASSIGN_OPT_SATU32
1262                 } else
1263                         COPY(pq_u);
1264
1265                 ASSIGN(qid);
1266                 break;
1267         }
1268         case 1: {
1269                 struct pf_altq_v1 *exported_q =
1270                     &((struct pfioc_altq_v1 *)pa)->altq;
1271
1272                 COPY(ifname);
1273
1274                 ASSIGN(scheduler);
1275                 ASSIGN(tbrsize);
1276                 ASSIGN(ifbandwidth);
1277
1278                 COPY(qname);
1279                 COPY(parent);
1280                 ASSIGN(parent_qid);
1281                 ASSIGN(bandwidth);
1282                 ASSIGN(priority);
1283                 ASSIGN(local_flags);
1284
1285                 ASSIGN(qlimit);
1286                 ASSIGN(flags);
1287                 COPY(pq_u);
1288
1289                 ASSIGN(qid);
1290                 break;
1291         }
1292         default:
1293                 panic("%s: unhandled struct pfioc_altq version", __func__);
1294                 break;
1295         }
1296
1297 #undef ASSIGN
1298 #undef COPY
1299 #undef SATU16
1300 #undef SATU32
1301
1302         return (0);
1303 }
1304
1305 /*
1306  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1307  * that may be using any version of it.
1308  */
1309 static int
1310 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1311 {
1312         u_int32_t version;
1313
1314         if (ioc_size == sizeof(struct pfioc_altq_v0))
1315                 version = 0;
1316         else
1317                 version = pa->version;
1318
1319         if (version > PFIOC_ALTQ_VERSION)
1320                 return (EINVAL);
1321
1322 #define ASSIGN(x) q->x = imported_q->x
1323 #define COPY(x) \
1324         bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1325
1326         switch (version) {
1327         case 0: {
1328                 struct pf_altq_v0 *imported_q =
1329                     &((struct pfioc_altq_v0 *)pa)->altq;
1330
1331                 COPY(ifname);
1332
1333                 ASSIGN(scheduler);
1334                 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1335                 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1336
1337                 COPY(qname);
1338                 COPY(parent);
1339                 ASSIGN(parent_qid);
1340                 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1341                 ASSIGN(priority);
1342                 ASSIGN(local_flags);
1343
1344                 ASSIGN(qlimit);
1345                 ASSIGN(flags);
1346
1347                 if (imported_q->scheduler == ALTQT_HFSC) {
1348 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1349
1350                         /*
1351                          * The m1 and m2 parameters are being copied from
1352                          * 32-bit to 64-bit.
1353                          */
1354                         ASSIGN_OPT(rtsc_m1);
1355                         ASSIGN_OPT(rtsc_d);
1356                         ASSIGN_OPT(rtsc_m2);
1357
1358                         ASSIGN_OPT(lssc_m1);
1359                         ASSIGN_OPT(lssc_d);
1360                         ASSIGN_OPT(lssc_m2);
1361
1362                         ASSIGN_OPT(ulsc_m1);
1363                         ASSIGN_OPT(ulsc_d);
1364                         ASSIGN_OPT(ulsc_m2);
1365
1366                         ASSIGN_OPT(flags);
1367
1368 #undef ASSIGN_OPT
1369                 } else
1370                         COPY(pq_u);
1371
1372                 ASSIGN(qid);
1373                 break;
1374         }
1375         case 1: {
1376                 struct pf_altq_v1 *imported_q =
1377                     &((struct pfioc_altq_v1 *)pa)->altq;
1378
1379                 COPY(ifname);
1380
1381                 ASSIGN(scheduler);
1382                 ASSIGN(tbrsize);
1383                 ASSIGN(ifbandwidth);
1384
1385                 COPY(qname);
1386                 COPY(parent);
1387                 ASSIGN(parent_qid);
1388                 ASSIGN(bandwidth);
1389                 ASSIGN(priority);
1390                 ASSIGN(local_flags);
1391
1392                 ASSIGN(qlimit);
1393                 ASSIGN(flags);
1394                 COPY(pq_u);
1395
1396                 ASSIGN(qid);
1397                 break;
1398         }
1399         default:
1400                 panic("%s: unhandled struct pfioc_altq version", __func__);
1401                 break;
1402         }
1403
1404 #undef ASSIGN
1405 #undef COPY
1406
1407         return (0);
1408 }
1409
1410 static struct pf_altq *
1411 pf_altq_get_nth_active(u_int32_t n)
1412 {
1413         struct pf_altq          *altq;
1414         u_int32_t                nr;
1415
1416         nr = 0;
1417         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1418                 if (nr == n)
1419                         return (altq);
1420                 nr++;
1421         }
1422
1423         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1424                 if (nr == n)
1425                         return (altq);
1426                 nr++;
1427         }
1428
1429         return (NULL);
1430 }
1431 #endif /* ALTQ */
1432
1433 void
1434 pf_krule_free(struct pf_krule *rule)
1435 {
1436         if (rule == NULL)
1437                 return;
1438
1439         counter_u64_free(rule->evaluations);
1440         for (int i = 0; i < 2; i++) {
1441                 counter_u64_free(rule->packets[i]);
1442                 counter_u64_free(rule->bytes[i]);
1443         }
1444         counter_u64_free(rule->states_cur);
1445         counter_u64_free(rule->states_tot);
1446         counter_u64_free(rule->src_nodes);
1447         free(rule, M_PFRULE);
1448 }
1449
1450 static void
1451 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1452     struct pf_pooladdr *pool)
1453 {
1454
1455         bzero(pool, sizeof(*pool));
1456         bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1457         strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1458 }
1459
1460 static void
1461 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1462     struct pf_kpooladdr *kpool)
1463 {
1464
1465         bzero(kpool, sizeof(*kpool));
1466         bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1467         strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1468 }
1469
1470 static void
1471 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1472 {
1473         bzero(pool, sizeof(*pool));
1474
1475         bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1476         bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1477
1478         pool->tblidx = kpool->tblidx;
1479         pool->proxy_port[0] = kpool->proxy_port[0];
1480         pool->proxy_port[1] = kpool->proxy_port[1];
1481         pool->opts = kpool->opts;
1482 }
1483
1484 static int
1485 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1486 {
1487         _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1488         _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1489
1490         bzero(kpool, sizeof(*kpool));
1491
1492         bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1493         bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1494
1495         kpool->tblidx = pool->tblidx;
1496         kpool->proxy_port[0] = pool->proxy_port[0];
1497         kpool->proxy_port[1] = pool->proxy_port[1];
1498         kpool->opts = pool->opts;
1499
1500         return (0);
1501 }
1502
1503 static void
1504 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1505 {
1506
1507         bzero(rule, sizeof(*rule));
1508
1509         bcopy(&krule->src, &rule->src, sizeof(rule->src));
1510         bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1511
1512         for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1513                 if (rule->skip[i].ptr == NULL)
1514                         rule->skip[i].nr = -1;
1515                 else
1516                         rule->skip[i].nr = krule->skip[i].ptr->nr;
1517         }
1518
1519         strlcpy(rule->label, krule->label, sizeof(rule->label));
1520         strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1521         strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1522         strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1523         strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1524         strlcpy(rule->match_tagname, krule->match_tagname,
1525             sizeof(rule->match_tagname));
1526         strlcpy(rule->overload_tblname, krule->overload_tblname,
1527             sizeof(rule->overload_tblname));
1528
1529         pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1530
1531         rule->evaluations = counter_u64_fetch(krule->evaluations);
1532         for (int i = 0; i < 2; i++) {
1533                 rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1534                 rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1535         }
1536
1537         /* kif, anchor, overload_tbl are not copied over. */
1538
1539         rule->os_fingerprint = krule->os_fingerprint;
1540
1541         rule->rtableid = krule->rtableid;
1542         bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1543         rule->max_states = krule->max_states;
1544         rule->max_src_nodes = krule->max_src_nodes;
1545         rule->max_src_states = krule->max_src_states;
1546         rule->max_src_conn = krule->max_src_conn;
1547         rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1548         rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1549         rule->qid = krule->qid;
1550         rule->pqid = krule->pqid;
1551         rule->nr = krule->nr;
1552         rule->prob = krule->prob;
1553         rule->cuid = krule->cuid;
1554         rule->cpid = krule->cpid;
1555
1556         rule->return_icmp = krule->return_icmp;
1557         rule->return_icmp6 = krule->return_icmp6;
1558         rule->max_mss = krule->max_mss;
1559         rule->tag = krule->tag;
1560         rule->match_tag = krule->match_tag;
1561         rule->scrub_flags = krule->scrub_flags;
1562
1563         bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1564         bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1565
1566         rule->rule_flag = krule->rule_flag;
1567         rule->action = krule->action;
1568         rule->direction = krule->direction;
1569         rule->log = krule->log;
1570         rule->logif = krule->logif;
1571         rule->quick = krule->quick;
1572         rule->ifnot = krule->ifnot;
1573         rule->match_tag_not = krule->match_tag_not;
1574         rule->natpass = krule->natpass;
1575
1576         rule->keep_state = krule->keep_state;
1577         rule->af = krule->af;
1578         rule->proto = krule->proto;
1579         rule->type = krule->type;
1580         rule->code = krule->code;
1581         rule->flags = krule->flags;
1582         rule->flagset = krule->flagset;
1583         rule->min_ttl = krule->min_ttl;
1584         rule->allow_opts = krule->allow_opts;
1585         rule->rt = krule->rt;
1586         rule->return_ttl = krule->return_ttl;
1587         rule->tos = krule->tos;
1588         rule->set_tos = krule->set_tos;
1589         rule->anchor_relative = krule->anchor_relative;
1590         rule->anchor_wildcard = krule->anchor_wildcard;
1591
1592         rule->flush = krule->flush;
1593         rule->prio = krule->prio;
1594         rule->set_prio[0] = krule->set_prio[0];
1595         rule->set_prio[1] = krule->set_prio[1];
1596
1597         bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1598
1599         rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1600         rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1601         rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1602 }
1603
1604 static int
1605 pf_check_rule_addr(const struct pf_rule_addr *addr)
1606 {
1607
1608         switch (addr->addr.type) {
1609         case PF_ADDR_ADDRMASK:
1610         case PF_ADDR_NOROUTE:
1611         case PF_ADDR_DYNIFTL:
1612         case PF_ADDR_TABLE:
1613         case PF_ADDR_URPFFAILED:
1614         case PF_ADDR_RANGE:
1615                 break;
1616         default:
1617                 return (EINVAL);
1618         }
1619
1620         if (addr->addr.p.dyn != NULL) {
1621                 return (EINVAL);
1622         }
1623
1624         return (0);
1625 }
1626
1627 static int
1628 pf_nvaddr_to_addr(const nvlist_t *nvl, struct pf_addr *paddr)
1629 {
1630         return (pf_nvbinary(nvl, "addr", paddr, sizeof(*paddr)));
1631 }
1632
1633 static nvlist_t *
1634 pf_addr_to_nvaddr(const struct pf_addr *paddr)
1635 {
1636         nvlist_t *nvl;
1637
1638         nvl = nvlist_create(0);
1639         if (nvl == NULL)
1640                 return (NULL);
1641
1642         nvlist_add_binary(nvl, "addr", paddr, sizeof(*paddr));
1643
1644         return (nvl);
1645 }
1646
1647 static int
1648 pf_nvpool_to_pool(const nvlist_t *nvl, struct pf_kpool *kpool)
1649 {
1650         int error = 0;
1651
1652         bzero(kpool, sizeof(*kpool));
1653
1654         PFNV_CHK(pf_nvbinary(nvl, "key", &kpool->key, sizeof(kpool->key)));
1655
1656         if (nvlist_exists_nvlist(nvl, "counter")) {
1657                 PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvl, "counter"),
1658                     &kpool->counter));
1659         }
1660
1661         PFNV_CHK(pf_nvint(nvl, "tblidx", &kpool->tblidx));
1662         PFNV_CHK(pf_nvuint16_array(nvl, "proxy_port", kpool->proxy_port, 2,
1663             NULL));
1664         PFNV_CHK(pf_nvuint8(nvl, "opts", &kpool->opts));
1665
1666 errout:
1667         return (error);
1668 }
1669
1670 static nvlist_t *
1671 pf_pool_to_nvpool(const struct pf_kpool *pool)
1672 {
1673         nvlist_t *nvl;
1674         nvlist_t *tmp;
1675
1676         nvl = nvlist_create(0);
1677         if (nvl == NULL)
1678                 return (NULL);
1679
1680         nvlist_add_binary(nvl, "key", &pool->key, sizeof(pool->key));
1681         tmp = pf_addr_to_nvaddr(&pool->counter);
1682         if (tmp == NULL)
1683                 goto error;
1684         nvlist_add_nvlist(nvl, "counter", tmp);
1685
1686         nvlist_add_number(nvl, "tblidx", pool->tblidx);
1687         pf_uint16_array_nv(nvl, "proxy_port", pool->proxy_port, 2);
1688         nvlist_add_number(nvl, "opts", pool->opts);
1689
1690         return (nvl);
1691
1692 error:
1693         nvlist_destroy(nvl);
1694         return (NULL);
1695 }
1696
1697 static int
1698 pf_nvaddr_wrap_to_addr_wrap(const nvlist_t *nvl, struct pf_addr_wrap *addr)
1699 {
1700         int error = 0;
1701
1702         bzero(addr, sizeof(*addr));
1703
1704         PFNV_CHK(pf_nvuint8(nvl, "type", &addr->type));
1705         PFNV_CHK(pf_nvuint8(nvl, "iflags", &addr->iflags));
1706         PFNV_CHK(pf_nvstring(nvl, "ifname", addr->v.ifname,
1707             sizeof(addr->v.ifname)));
1708         PFNV_CHK(pf_nvstring(nvl, "tblname", addr->v.tblname,
1709             sizeof(addr->v.tblname)));
1710
1711         if (! nvlist_exists_nvlist(nvl, "addr"))
1712                 return (EINVAL);
1713         PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvl, "addr"),
1714             &addr->v.a.addr));
1715
1716         if (! nvlist_exists_nvlist(nvl, "mask"))
1717                 return (EINVAL);
1718         PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvl, "mask"),
1719             &addr->v.a.mask));
1720
1721         switch (addr->type) {
1722         case PF_ADDR_DYNIFTL:
1723         case PF_ADDR_TABLE:
1724         case PF_ADDR_RANGE:
1725         case PF_ADDR_ADDRMASK:
1726         case PF_ADDR_NOROUTE:
1727         case PF_ADDR_URPFFAILED:
1728                 break;
1729         default:
1730                 return (EINVAL);
1731         }
1732
1733 errout:
1734         return (error);
1735 }
1736
1737 static nvlist_t *
1738 pf_addr_wrap_to_nvaddr_wrap(const struct pf_addr_wrap *addr)
1739 {
1740         nvlist_t *nvl;
1741         nvlist_t *tmp;
1742
1743         nvl = nvlist_create(0);
1744         if (nvl == NULL)
1745                 return (NULL);
1746
1747         nvlist_add_number(nvl, "type", addr->type);
1748         nvlist_add_number(nvl, "iflags", addr->iflags);
1749         nvlist_add_string(nvl, "ifname", addr->v.ifname);
1750         nvlist_add_string(nvl, "tblname", addr->v.tblname);
1751
1752         tmp = pf_addr_to_nvaddr(&addr->v.a.addr);
1753         if (tmp == NULL)
1754                 goto error;
1755         nvlist_add_nvlist(nvl, "addr", tmp);
1756         tmp = pf_addr_to_nvaddr(&addr->v.a.mask);
1757         if (tmp == NULL)
1758                 goto error;
1759         nvlist_add_nvlist(nvl, "mask", tmp);
1760
1761         return (nvl);
1762
1763 error:
1764         nvlist_destroy(nvl);
1765         return (NULL);
1766 }
1767
1768 static int
1769 pf_validate_op(uint8_t op)
1770 {
1771         switch (op) {
1772         case PF_OP_NONE:
1773         case PF_OP_IRG:
1774         case PF_OP_EQ:
1775         case PF_OP_NE:
1776         case PF_OP_LT:
1777         case PF_OP_LE:
1778         case PF_OP_GT:
1779         case PF_OP_GE:
1780         case PF_OP_XRG:
1781         case PF_OP_RRG:
1782                 break;
1783         default:
1784                 return (EINVAL);
1785         }
1786
1787         return (0);
1788 }
1789
1790 static int
1791 pf_nvrule_addr_to_rule_addr(const nvlist_t *nvl, struct pf_rule_addr *addr)
1792 {
1793         int error = 0;
1794
1795         if (! nvlist_exists_nvlist(nvl, "addr"))
1796                 return (EINVAL);
1797
1798         PFNV_CHK(pf_nvaddr_wrap_to_addr_wrap(nvlist_get_nvlist(nvl, "addr"),
1799             &addr->addr));
1800         PFNV_CHK(pf_nvuint16_array(nvl, "port", addr->port, 2, NULL));
1801         PFNV_CHK(pf_nvuint8(nvl, "neg", &addr->neg));
1802         PFNV_CHK(pf_nvuint8(nvl, "port_op", &addr->port_op));
1803
1804         PFNV_CHK(pf_validate_op(addr->port_op));
1805
1806 errout:
1807         return (error);
1808 }
1809
1810 static nvlist_t *
1811 pf_rule_addr_to_nvrule_addr(const struct pf_rule_addr *addr)
1812 {
1813         nvlist_t *nvl;
1814         nvlist_t *tmp;
1815
1816         nvl = nvlist_create(0);
1817         if (nvl == NULL)
1818                 return (NULL);
1819
1820         tmp = pf_addr_wrap_to_nvaddr_wrap(&addr->addr);
1821         if (tmp == NULL)
1822                 goto error;
1823         nvlist_add_nvlist(nvl, "addr", tmp);
1824         pf_uint16_array_nv(nvl, "port", addr->port, 2);
1825         nvlist_add_number(nvl, "neg", addr->neg);
1826         nvlist_add_number(nvl, "port_op", addr->port_op);
1827
1828         return (nvl);
1829
1830 error:
1831         nvlist_destroy(nvl);
1832         return (NULL);
1833 }
1834
1835 static int
1836 pf_nvrule_uid_to_rule_uid(const nvlist_t *nvl, struct pf_rule_uid *uid)
1837 {
1838         int error = 0;
1839
1840         bzero(uid, sizeof(*uid));
1841
1842         PFNV_CHK(pf_nvuint32_array(nvl, "uid", uid->uid, 2, NULL));
1843         PFNV_CHK(pf_nvuint8(nvl, "op", &uid->op));
1844
1845         PFNV_CHK(pf_validate_op(uid->op));
1846
1847 errout:
1848         return (error);
1849 }
1850
1851 static nvlist_t *
1852 pf_rule_uid_to_nvrule_uid(const struct pf_rule_uid *uid)
1853 {
1854         nvlist_t *nvl;
1855
1856         nvl = nvlist_create(0);
1857         if (nvl == NULL)
1858                 return (NULL);
1859
1860         pf_uint32_array_nv(nvl, "uid", uid->uid, 2);
1861         nvlist_add_number(nvl, "op", uid->op);
1862
1863         return (nvl);
1864 }
1865
1866 static int
1867 pf_nvrule_gid_to_rule_gid(const nvlist_t *nvl, struct pf_rule_gid *gid)
1868 {
1869         /* Cheat a little. These stucts are the same, other than the name of
1870          * the first field. */
1871         return (pf_nvrule_uid_to_rule_uid(nvl, (struct pf_rule_uid *)gid));
1872 }
1873
1874 static int
1875 pf_nvrule_to_krule(const nvlist_t *nvl, struct pf_krule **prule)
1876 {
1877         struct pf_krule *rule;
1878         int error = 0;
1879
1880         rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK | M_ZERO);
1881
1882         PFNV_CHK(pf_nvuint32(nvl, "nr", &rule->nr));
1883
1884         if (! nvlist_exists_nvlist(nvl, "src")) {
1885                 error = EINVAL;
1886                 goto errout;
1887         }
1888         error = pf_nvrule_addr_to_rule_addr(nvlist_get_nvlist(nvl, "src"),
1889             &rule->src);
1890         if (error != 0)
1891                 goto errout;
1892
1893         if (! nvlist_exists_nvlist(nvl, "dst")) {
1894                 error = EINVAL;
1895                 goto errout;
1896         }
1897         PFNV_CHK(pf_nvrule_addr_to_rule_addr(nvlist_get_nvlist(nvl, "dst"),
1898             &rule->dst));
1899
1900         PFNV_CHK(pf_nvstring(nvl, "label", rule->label, sizeof(rule->label)));
1901         PFNV_CHK(pf_nvstring(nvl, "ifname", rule->ifname,
1902             sizeof(rule->ifname)));
1903         PFNV_CHK(pf_nvstring(nvl, "qname", rule->qname, sizeof(rule->qname)));
1904         PFNV_CHK(pf_nvstring(nvl, "pqname", rule->pqname,
1905             sizeof(rule->pqname)));
1906         PFNV_CHK(pf_nvstring(nvl, "tagname", rule->tagname,
1907             sizeof(rule->tagname)));
1908         PFNV_CHK(pf_nvstring(nvl, "match_tagname", rule->match_tagname,
1909             sizeof(rule->match_tagname)));
1910         PFNV_CHK(pf_nvstring(nvl, "overload_tblname", rule->overload_tblname,
1911             sizeof(rule->overload_tblname)));
1912
1913         if (! nvlist_exists_nvlist(nvl, "rpool")) {
1914                 error = EINVAL;
1915                 goto errout;
1916         }
1917         PFNV_CHK(pf_nvpool_to_pool(nvlist_get_nvlist(nvl, "rpool"),
1918             &rule->rpool));
1919
1920         PFNV_CHK(pf_nvuint32(nvl, "os_fingerprint", &rule->os_fingerprint));
1921
1922         PFNV_CHK(pf_nvint(nvl, "rtableid", &rule->rtableid));
1923         PFNV_CHK(pf_nvuint32_array(nvl, "timeout", rule->timeout, PFTM_MAX, NULL));
1924         PFNV_CHK(pf_nvuint32(nvl, "max_states", &rule->max_states));
1925         PFNV_CHK(pf_nvuint32(nvl, "max_src_nodes", &rule->max_src_nodes));
1926         PFNV_CHK(pf_nvuint32(nvl, "max_src_states", &rule->max_src_states));
1927         PFNV_CHK(pf_nvuint32(nvl, "max_src_conn", &rule->max_src_conn));
1928         PFNV_CHK(pf_nvuint32(nvl, "max_src_conn_rate.limit",
1929             &rule->max_src_conn_rate.limit));
1930         PFNV_CHK(pf_nvuint32(nvl, "max_src_conn_rate.seconds",
1931             &rule->max_src_conn_rate.seconds));
1932         PFNV_CHK(pf_nvuint32(nvl, "prob", &rule->prob));
1933         PFNV_CHK(pf_nvuint32(nvl, "cuid", &rule->cuid));
1934         PFNV_CHK(pf_nvuint32(nvl, "cpid", &rule->cpid));
1935
1936         PFNV_CHK(pf_nvuint16(nvl, "return_icmp", &rule->return_icmp));
1937         PFNV_CHK(pf_nvuint16(nvl, "return_icmp6", &rule->return_icmp6));
1938
1939         PFNV_CHK(pf_nvuint16(nvl, "max_mss", &rule->max_mss));
1940         PFNV_CHK(pf_nvuint16(nvl, "scrub_flags", &rule->scrub_flags));
1941
1942         if (! nvlist_exists_nvlist(nvl, "uid")) {
1943                 error = EINVAL;
1944                 goto errout;
1945         }
1946         PFNV_CHK(pf_nvrule_uid_to_rule_uid(nvlist_get_nvlist(nvl, "uid"),
1947             &rule->uid));
1948
1949         if (! nvlist_exists_nvlist(nvl, "gid")) {
1950                 error = EINVAL;
1951                 goto errout;
1952         }
1953         PFNV_CHK(pf_nvrule_gid_to_rule_gid(nvlist_get_nvlist(nvl, "gid"),
1954             &rule->gid));
1955
1956         PFNV_CHK(pf_nvuint32(nvl, "rule_flag", &rule->rule_flag));
1957         PFNV_CHK(pf_nvuint8(nvl, "action", &rule->action));
1958         PFNV_CHK(pf_nvuint8(nvl, "direction", &rule->direction));
1959         PFNV_CHK(pf_nvuint8(nvl, "log", &rule->log));
1960         PFNV_CHK(pf_nvuint8(nvl, "logif", &rule->logif));
1961         PFNV_CHK(pf_nvuint8(nvl, "quick", &rule->quick));
1962         PFNV_CHK(pf_nvuint8(nvl, "ifnot", &rule->ifnot));
1963         PFNV_CHK(pf_nvuint8(nvl, "match_tag_not", &rule->match_tag_not));
1964         PFNV_CHK(pf_nvuint8(nvl, "natpass", &rule->natpass));
1965
1966         PFNV_CHK(pf_nvuint8(nvl, "keep_state", &rule->keep_state));
1967         PFNV_CHK(pf_nvuint8(nvl, "af", &rule->af));
1968         PFNV_CHK(pf_nvuint8(nvl, "proto", &rule->proto));
1969         PFNV_CHK(pf_nvuint8(nvl, "type", &rule->type));
1970         PFNV_CHK(pf_nvuint8(nvl, "code", &rule->code));
1971         PFNV_CHK(pf_nvuint8(nvl, "flags", &rule->flags));
1972         PFNV_CHK(pf_nvuint8(nvl, "flagset", &rule->flagset));
1973         PFNV_CHK(pf_nvuint8(nvl, "min_ttl", &rule->min_ttl));
1974         PFNV_CHK(pf_nvuint8(nvl, "allow_opts", &rule->allow_opts));
1975         PFNV_CHK(pf_nvuint8(nvl, "rt", &rule->rt));
1976         PFNV_CHK(pf_nvuint8(nvl, "return_ttl", &rule->return_ttl));
1977         PFNV_CHK(pf_nvuint8(nvl, "tos", &rule->tos));
1978         PFNV_CHK(pf_nvuint8(nvl, "set_tos", &rule->set_tos));
1979         PFNV_CHK(pf_nvuint8(nvl, "anchor_relative", &rule->anchor_relative));
1980         PFNV_CHK(pf_nvuint8(nvl, "anchor_wildcard", &rule->anchor_wildcard));
1981
1982         PFNV_CHK(pf_nvuint8(nvl, "flush", &rule->flush));
1983         PFNV_CHK(pf_nvuint8(nvl, "prio", &rule->prio));
1984
1985         PFNV_CHK(pf_nvuint8_array(nvl, "set_prio", &rule->prio, 2, NULL));
1986
1987         if (nvlist_exists_nvlist(nvl, "divert")) {
1988                 const nvlist_t *nvldivert = nvlist_get_nvlist(nvl, "divert");
1989
1990                 if (! nvlist_exists_nvlist(nvldivert, "addr")) {
1991                         error = EINVAL;
1992                         goto errout;
1993                 }
1994                 PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvldivert, "addr"),
1995                     &rule->divert.addr));
1996                 PFNV_CHK(pf_nvuint16(nvldivert, "port", &rule->divert.port));
1997         }
1998
1999         /* Validation */
2000 #ifndef INET
2001         if (rule->af == AF_INET) {
2002                 error = EAFNOSUPPORT;
2003                 goto errout;
2004         }
2005 #endif /* INET */
2006 #ifndef INET6
2007         if (rule->af == AF_INET6) {
2008                 error = EAFNOSUPPORT;
2009                 goto errout;
2010         }
2011 #endif /* INET6 */
2012
2013         PFNV_CHK(pf_check_rule_addr(&rule->src));
2014         PFNV_CHK(pf_check_rule_addr(&rule->dst));
2015
2016         *prule = rule;
2017
2018         return (0);
2019
2020 errout:
2021         pf_krule_free(rule);
2022         *prule = NULL;
2023
2024         return (error);
2025 }
2026
2027 static nvlist_t *
2028 pf_divert_to_nvdivert(const struct pf_krule *rule)
2029 {
2030         nvlist_t *nvl;
2031         nvlist_t *tmp;
2032
2033         nvl = nvlist_create(0);
2034         if (nvl == NULL)
2035                 return (NULL);
2036
2037         tmp = pf_addr_to_nvaddr(&rule->divert.addr);
2038         if (tmp == NULL)
2039                 goto error;
2040         nvlist_add_nvlist(nvl, "addr", tmp);
2041         nvlist_add_number(nvl, "port", rule->divert.port);
2042
2043         return (nvl);
2044
2045 error:
2046         nvlist_destroy(nvl);
2047         return (NULL);
2048 }
2049
2050 static nvlist_t *
2051 pf_krule_to_nvrule(const struct pf_krule *rule)
2052 {
2053         nvlist_t *nvl, *tmp;
2054
2055         nvl = nvlist_create(0);
2056         if (nvl == NULL)
2057                 return (nvl);
2058
2059         nvlist_add_number(nvl, "nr", rule->nr);
2060         tmp = pf_rule_addr_to_nvrule_addr(&rule->src);
2061         if (tmp == NULL)
2062                 goto error;
2063         nvlist_add_nvlist(nvl, "src", tmp);
2064         tmp = pf_rule_addr_to_nvrule_addr(&rule->dst);
2065         if (tmp == NULL)
2066                 goto error;
2067         nvlist_add_nvlist(nvl, "dst", tmp);
2068
2069         for (int i = 0; i < PF_SKIP_COUNT; i++) {
2070                 nvlist_append_number_array(nvl, "skip",
2071                     rule->skip[i].ptr ? rule->skip[i].ptr->nr : -1);
2072         }
2073
2074         nvlist_add_string(nvl, "label", rule->label);
2075         nvlist_add_string(nvl, "ifname", rule->ifname);
2076         nvlist_add_string(nvl, "qname", rule->qname);
2077         nvlist_add_string(nvl, "pqname", rule->pqname);
2078         nvlist_add_string(nvl, "tagname", rule->tagname);
2079         nvlist_add_string(nvl, "match_tagname", rule->match_tagname);
2080         nvlist_add_string(nvl, "overload_tblname", rule->overload_tblname);
2081
2082         tmp = pf_pool_to_nvpool(&rule->rpool);
2083         if (tmp == NULL)
2084                 goto error;
2085         nvlist_add_nvlist(nvl, "rpool", tmp);
2086
2087         nvlist_add_number(nvl, "evaluations",
2088             counter_u64_fetch(rule->evaluations));
2089         for (int i = 0; i < 2; i++) {
2090                 nvlist_append_number_array(nvl, "packets",
2091                     counter_u64_fetch(rule->packets[i]));
2092                 nvlist_append_number_array(nvl, "bytes",
2093                     counter_u64_fetch(rule->bytes[i]));
2094         }
2095
2096         nvlist_add_number(nvl, "os_fingerprint", rule->os_fingerprint);
2097
2098         nvlist_add_number(nvl, "rtableid", rule->rtableid);
2099         pf_uint32_array_nv(nvl, "timeout", rule->timeout, PFTM_MAX);
2100         nvlist_add_number(nvl, "max_states", rule->max_states);
2101         nvlist_add_number(nvl, "max_src_nodes", rule->max_src_nodes);
2102         nvlist_add_number(nvl, "max_src_states", rule->max_src_states);
2103         nvlist_add_number(nvl, "max_src_conn", rule->max_src_conn);
2104         nvlist_add_number(nvl, "max_src_conn_rate.limit",
2105             rule->max_src_conn_rate.limit);
2106         nvlist_add_number(nvl, "max_src_conn_rate.seconds",
2107             rule->max_src_conn_rate.seconds);
2108         nvlist_add_number(nvl, "qid", rule->qid);
2109         nvlist_add_number(nvl, "pqid", rule->pqid);
2110         nvlist_add_number(nvl, "prob", rule->prob);
2111         nvlist_add_number(nvl, "cuid", rule->cuid);
2112         nvlist_add_number(nvl, "cpid", rule->cpid);
2113
2114         nvlist_add_number(nvl, "states_cur",
2115             counter_u64_fetch(rule->states_cur));
2116         nvlist_add_number(nvl, "states_tot",
2117             counter_u64_fetch(rule->states_tot));
2118         nvlist_add_number(nvl, "src_nodes",
2119             counter_u64_fetch(rule->src_nodes));
2120
2121         nvlist_add_number(nvl, "return_icmp", rule->return_icmp);
2122         nvlist_add_number(nvl, "return_icmp6", rule->return_icmp6);
2123
2124         nvlist_add_number(nvl, "max_mss", rule->max_mss);
2125         nvlist_add_number(nvl, "scrub_flags", rule->scrub_flags);
2126
2127         tmp = pf_rule_uid_to_nvrule_uid(&rule->uid);
2128         if (tmp == NULL)
2129                 goto error;
2130         nvlist_add_nvlist(nvl, "uid", tmp);
2131         tmp = pf_rule_uid_to_nvrule_uid((const struct pf_rule_uid *)&rule->gid);
2132         if (tmp == NULL)
2133                 goto error;
2134         nvlist_add_nvlist(nvl, "gid", tmp);
2135
2136         nvlist_add_number(nvl, "rule_flag", rule->rule_flag);
2137         nvlist_add_number(nvl, "action", rule->action);
2138         nvlist_add_number(nvl, "direction", rule->direction);
2139         nvlist_add_number(nvl, "log", rule->log);
2140         nvlist_add_number(nvl, "logif", rule->logif);
2141         nvlist_add_number(nvl, "quick", rule->quick);
2142         nvlist_add_number(nvl, "ifnot", rule->ifnot);
2143         nvlist_add_number(nvl, "match_tag_not", rule->match_tag_not);
2144         nvlist_add_number(nvl, "natpass", rule->natpass);
2145
2146         nvlist_add_number(nvl, "keep_state", rule->keep_state);
2147         nvlist_add_number(nvl, "af", rule->af);
2148         nvlist_add_number(nvl, "proto", rule->proto);
2149         nvlist_add_number(nvl, "type", rule->type);
2150         nvlist_add_number(nvl, "code", rule->code);
2151         nvlist_add_number(nvl, "flags", rule->flags);
2152         nvlist_add_number(nvl, "flagset", rule->flagset);
2153         nvlist_add_number(nvl, "min_ttl", rule->min_ttl);
2154         nvlist_add_number(nvl, "allow_opts", rule->allow_opts);
2155         nvlist_add_number(nvl, "rt", rule->rt);
2156         nvlist_add_number(nvl, "return_ttl", rule->return_ttl);
2157         nvlist_add_number(nvl, "tos", rule->tos);
2158         nvlist_add_number(nvl, "set_tos", rule->set_tos);
2159         nvlist_add_number(nvl, "anchor_relative", rule->anchor_relative);
2160         nvlist_add_number(nvl, "anchor_wildcard", rule->anchor_wildcard);
2161
2162         nvlist_add_number(nvl, "flush", rule->flush);
2163         nvlist_add_number(nvl, "prio", rule->prio);
2164
2165         pf_uint8_array_nv(nvl, "set_prio", &rule->prio, 2);
2166
2167         tmp = pf_divert_to_nvdivert(rule);
2168         if (tmp == NULL)
2169                 goto error;
2170         nvlist_add_nvlist(nvl, "divert", tmp);
2171
2172         return (nvl);
2173
2174 error:
2175         nvlist_destroy(nvl);
2176         return (NULL);
2177 }
2178
2179 static int
2180 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
2181 {
2182         int ret;
2183
2184 #ifndef INET
2185         if (rule->af == AF_INET) {
2186                 return (EAFNOSUPPORT);
2187         }
2188 #endif /* INET */
2189 #ifndef INET6
2190         if (rule->af == AF_INET6) {
2191                 return (EAFNOSUPPORT);
2192         }
2193 #endif /* INET6 */
2194
2195         ret = pf_check_rule_addr(&rule->src);
2196         if (ret != 0)
2197                 return (ret);
2198         ret = pf_check_rule_addr(&rule->dst);
2199         if (ret != 0)
2200                 return (ret);
2201
2202         bzero(krule, sizeof(*krule));
2203
2204         bcopy(&rule->src, &krule->src, sizeof(rule->src));
2205         bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
2206
2207         strlcpy(krule->label, rule->label, sizeof(rule->label));
2208         strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
2209         strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
2210         strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
2211         strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
2212         strlcpy(krule->match_tagname, rule->match_tagname,
2213             sizeof(rule->match_tagname));
2214         strlcpy(krule->overload_tblname, rule->overload_tblname,
2215             sizeof(rule->overload_tblname));
2216
2217         ret = pf_pool_to_kpool(&rule->rpool, &krule->rpool);
2218         if (ret != 0)
2219                 return (ret);
2220
2221         /* Don't allow userspace to set evaulations, packets or bytes. */
2222         /* kif, anchor, overload_tbl are not copied over. */
2223
2224         krule->os_fingerprint = rule->os_fingerprint;
2225
2226         krule->rtableid = rule->rtableid;
2227         bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
2228         krule->max_states = rule->max_states;
2229         krule->max_src_nodes = rule->max_src_nodes;
2230         krule->max_src_states = rule->max_src_states;
2231         krule->max_src_conn = rule->max_src_conn;
2232         krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2233         krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2234         krule->qid = rule->qid;
2235         krule->pqid = rule->pqid;
2236         krule->nr = rule->nr;
2237         krule->prob = rule->prob;
2238         krule->cuid = rule->cuid;
2239         krule->cpid = rule->cpid;
2240
2241         krule->return_icmp = rule->return_icmp;
2242         krule->return_icmp6 = rule->return_icmp6;
2243         krule->max_mss = rule->max_mss;
2244         krule->tag = rule->tag;
2245         krule->match_tag = rule->match_tag;
2246         krule->scrub_flags = rule->scrub_flags;
2247
2248         bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2249         bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2250
2251         krule->rule_flag = rule->rule_flag;
2252         krule->action = rule->action;
2253         krule->direction = rule->direction;
2254         krule->log = rule->log;
2255         krule->logif = rule->logif;
2256         krule->quick = rule->quick;
2257         krule->ifnot = rule->ifnot;
2258         krule->match_tag_not = rule->match_tag_not;
2259         krule->natpass = rule->natpass;
2260
2261         krule->keep_state = rule->keep_state;
2262         krule->af = rule->af;
2263         krule->proto = rule->proto;
2264         krule->type = rule->type;
2265         krule->code = rule->code;
2266         krule->flags = rule->flags;
2267         krule->flagset = rule->flagset;
2268         krule->min_ttl = rule->min_ttl;
2269         krule->allow_opts = rule->allow_opts;
2270         krule->rt = rule->rt;
2271         krule->return_ttl = rule->return_ttl;
2272         krule->tos = rule->tos;
2273         krule->set_tos = rule->set_tos;
2274         krule->anchor_relative = rule->anchor_relative;
2275         krule->anchor_wildcard = rule->anchor_wildcard;
2276
2277         krule->flush = rule->flush;
2278         krule->prio = rule->prio;
2279         krule->set_prio[0] = rule->set_prio[0];
2280         krule->set_prio[1] = rule->set_prio[1];
2281
2282         bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2283
2284         return (0);
2285 }
2286
2287 static int
2288 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2289     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2290     struct thread *td)
2291 {
2292         struct pf_kruleset      *ruleset;
2293         struct pf_krule         *tail;
2294         struct pf_kpooladdr     *pa;
2295         struct pfi_kkif         *kif = NULL;
2296         int                      rs_num;
2297         int                      error = 0;
2298
2299         if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2300                 error = EINVAL;
2301                 goto errout_unlocked;
2302         }
2303
2304 #define ERROUT(x)       { error = (x); goto errout; }
2305
2306         if (rule->ifname[0])
2307                 kif = pf_kkif_create(M_WAITOK);
2308         rule->evaluations = counter_u64_alloc(M_WAITOK);
2309         for (int i = 0; i < 2; i++) {
2310                 rule->packets[i] = counter_u64_alloc(M_WAITOK);
2311                 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2312         }
2313         rule->states_cur = counter_u64_alloc(M_WAITOK);
2314         rule->states_tot = counter_u64_alloc(M_WAITOK);
2315         rule->src_nodes = counter_u64_alloc(M_WAITOK);
2316         rule->cuid = td->td_ucred->cr_ruid;
2317         rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2318         TAILQ_INIT(&rule->rpool.list);
2319
2320         PF_RULES_WLOCK();
2321         ruleset = pf_find_kruleset(anchor);
2322         if (ruleset == NULL)
2323                 ERROUT(EINVAL);
2324         rs_num = pf_get_ruleset_number(rule->action);
2325         if (rs_num >= PF_RULESET_MAX)
2326                 ERROUT(EINVAL);
2327         if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2328                 DPFPRINTF(PF_DEBUG_MISC,
2329                     ("ticket: %d != [%d]%d\n", ticket, rs_num,
2330                     ruleset->rules[rs_num].inactive.ticket));
2331                 ERROUT(EBUSY);
2332         }
2333         if (pool_ticket != V_ticket_pabuf) {
2334                 DPFPRINTF(PF_DEBUG_MISC,
2335                     ("pool_ticket: %d != %d\n", pool_ticket,
2336                     V_ticket_pabuf));
2337                 ERROUT(EBUSY);
2338         }
2339
2340         tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2341             pf_krulequeue);
2342         if (tail)
2343                 rule->nr = tail->nr + 1;
2344         else
2345                 rule->nr = 0;
2346         if (rule->ifname[0]) {
2347                 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2348                 kif = NULL;
2349                 pfi_kkif_ref(rule->kif);
2350         } else
2351                 rule->kif = NULL;
2352
2353         if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2354                 error = EBUSY;
2355
2356 #ifdef ALTQ
2357         /* set queue IDs */
2358         if (rule->qname[0] != 0) {
2359                 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2360                         error = EBUSY;
2361                 else if (rule->pqname[0] != 0) {
2362                         if ((rule->pqid =
2363                             pf_qname2qid(rule->pqname)) == 0)
2364                                 error = EBUSY;
2365                 } else
2366                         rule->pqid = rule->qid;
2367         }
2368 #endif
2369         if (rule->tagname[0])
2370                 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2371                         error = EBUSY;
2372         if (rule->match_tagname[0])
2373                 if ((rule->match_tag =
2374                     pf_tagname2tag(rule->match_tagname)) == 0)
2375                         error = EBUSY;
2376         if (rule->rt && !rule->direction)
2377                 error = EINVAL;
2378         if (!rule->log)
2379                 rule->logif = 0;
2380         if (rule->logif >= PFLOGIFS_MAX)
2381                 error = EINVAL;
2382         if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2383                 error = ENOMEM;
2384         if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2385                 error = ENOMEM;
2386         if (pf_kanchor_setup(rule, ruleset, anchor_call))
2387                 error = EINVAL;
2388         if (rule->scrub_flags & PFSTATE_SETPRIO &&
2389             (rule->set_prio[0] > PF_PRIO_MAX ||
2390             rule->set_prio[1] > PF_PRIO_MAX))
2391                 error = EINVAL;
2392         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2393                 if (pa->addr.type == PF_ADDR_TABLE) {
2394                         pa->addr.p.tbl = pfr_attach_table(ruleset,
2395                             pa->addr.v.tblname);
2396                         if (pa->addr.p.tbl == NULL)
2397                                 error = ENOMEM;
2398                 }
2399
2400         rule->overload_tbl = NULL;
2401         if (rule->overload_tblname[0]) {
2402                 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2403                     rule->overload_tblname)) == NULL)
2404                         error = EINVAL;
2405                 else
2406                         rule->overload_tbl->pfrkt_flags |=
2407                             PFR_TFLAG_ACTIVE;
2408         }
2409
2410         pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2411         if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2412             (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2413             (rule->rt > PF_NOPFROUTE)) &&
2414             (TAILQ_FIRST(&rule->rpool.list) == NULL))
2415                 error = EINVAL;
2416
2417         if (error) {
2418                 pf_free_rule(rule);
2419                 rule = NULL;
2420                 ERROUT(error);
2421         }
2422
2423         rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2424         counter_u64_zero(rule->evaluations);
2425         for (int i = 0; i < 2; i++) {
2426                 counter_u64_zero(rule->packets[i]);
2427                 counter_u64_zero(rule->bytes[i]);
2428         }
2429         TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2430             rule, entries);
2431         ruleset->rules[rs_num].inactive.rcount++;
2432         PF_RULES_WUNLOCK();
2433
2434         return (0);
2435
2436 #undef ERROUT
2437 errout:
2438         PF_RULES_WUNLOCK();
2439 errout_unlocked:
2440         pf_kkif_free(kif);
2441         pf_krule_free(rule);
2442         return (error);
2443 }
2444
2445 static int
2446 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2447 {
2448         int                      error = 0;
2449         PF_RULES_RLOCK_TRACKER;
2450
2451         /* XXX keep in sync with switch() below */
2452         if (securelevel_gt(td->td_ucred, 2))
2453                 switch (cmd) {
2454                 case DIOCGETRULES:
2455                 case DIOCGETRULE:
2456                 case DIOCGETRULENV:
2457                 case DIOCGETADDRS:
2458                 case DIOCGETADDR:
2459                 case DIOCGETSTATE:
2460                 case DIOCSETSTATUSIF:
2461                 case DIOCGETSTATUS:
2462                 case DIOCCLRSTATUS:
2463                 case DIOCNATLOOK:
2464                 case DIOCSETDEBUG:
2465                 case DIOCGETSTATES:
2466                 case DIOCGETTIMEOUT:
2467                 case DIOCCLRRULECTRS:
2468                 case DIOCGETLIMIT:
2469                 case DIOCGETALTQSV0:
2470                 case DIOCGETALTQSV1:
2471                 case DIOCGETALTQV0:
2472                 case DIOCGETALTQV1:
2473                 case DIOCGETQSTATSV0:
2474                 case DIOCGETQSTATSV1:
2475                 case DIOCGETRULESETS:
2476                 case DIOCGETRULESET:
2477                 case DIOCRGETTABLES:
2478                 case DIOCRGETTSTATS:
2479                 case DIOCRCLRTSTATS:
2480                 case DIOCRCLRADDRS:
2481                 case DIOCRADDADDRS:
2482                 case DIOCRDELADDRS:
2483                 case DIOCRSETADDRS:
2484                 case DIOCRGETADDRS:
2485                 case DIOCRGETASTATS:
2486                 case DIOCRCLRASTATS:
2487                 case DIOCRTSTADDRS:
2488                 case DIOCOSFPGET:
2489                 case DIOCGETSRCNODES:
2490                 case DIOCCLRSRCNODES:
2491                 case DIOCIGETIFACES:
2492                 case DIOCGIFSPEEDV0:
2493                 case DIOCGIFSPEEDV1:
2494                 case DIOCSETIFFLAG:
2495                 case DIOCCLRIFFLAG:
2496                         break;
2497                 case DIOCRCLRTABLES:
2498                 case DIOCRADDTABLES:
2499                 case DIOCRDELTABLES:
2500                 case DIOCRSETTFLAGS:
2501                         if (((struct pfioc_table *)addr)->pfrio_flags &
2502                             PFR_FLAG_DUMMY)
2503                                 break; /* dummy operation ok */
2504                         return (EPERM);
2505                 default:
2506                         return (EPERM);
2507                 }
2508
2509         if (!(flags & FWRITE))
2510                 switch (cmd) {
2511                 case DIOCGETRULES:
2512                 case DIOCGETADDRS:
2513                 case DIOCGETADDR:
2514                 case DIOCGETSTATE:
2515                 case DIOCGETSTATUS:
2516                 case DIOCGETSTATES:
2517                 case DIOCGETTIMEOUT:
2518                 case DIOCGETLIMIT:
2519                 case DIOCGETALTQSV0:
2520                 case DIOCGETALTQSV1:
2521                 case DIOCGETALTQV0:
2522                 case DIOCGETALTQV1:
2523                 case DIOCGETQSTATSV0:
2524                 case DIOCGETQSTATSV1:
2525                 case DIOCGETRULESETS:
2526                 case DIOCGETRULESET:
2527                 case DIOCNATLOOK:
2528                 case DIOCRGETTABLES:
2529                 case DIOCRGETTSTATS:
2530                 case DIOCRGETADDRS:
2531                 case DIOCRGETASTATS:
2532                 case DIOCRTSTADDRS:
2533                 case DIOCOSFPGET:
2534                 case DIOCGETSRCNODES:
2535                 case DIOCIGETIFACES:
2536                 case DIOCGIFSPEEDV1:
2537                 case DIOCGIFSPEEDV0:
2538                 case DIOCGETRULENV:
2539                         break;
2540                 case DIOCRCLRTABLES:
2541                 case DIOCRADDTABLES:
2542                 case DIOCRDELTABLES:
2543                 case DIOCRCLRTSTATS:
2544                 case DIOCRCLRADDRS:
2545                 case DIOCRADDADDRS:
2546                 case DIOCRDELADDRS:
2547                 case DIOCRSETADDRS:
2548                 case DIOCRSETTFLAGS:
2549                         if (((struct pfioc_table *)addr)->pfrio_flags &
2550                             PFR_FLAG_DUMMY) {
2551                                 flags |= FWRITE; /* need write lock for dummy */
2552                                 break; /* dummy operation ok */
2553                         }
2554                         return (EACCES);
2555                 case DIOCGETRULE:
2556                         if (((struct pfioc_rule *)addr)->action ==
2557                             PF_GET_CLR_CNTR)
2558                                 return (EACCES);
2559                         break;
2560                 default:
2561                         return (EACCES);
2562                 }
2563
2564         CURVNET_SET(TD_TO_VNET(td));
2565
2566         switch (cmd) {
2567         case DIOCSTART:
2568                 sx_xlock(&pf_ioctl_lock);
2569                 if (V_pf_status.running)
2570                         error = EEXIST;
2571                 else {
2572                         int cpu;
2573
2574                         hook_pf();
2575                         V_pf_status.running = 1;
2576                         V_pf_status.since = time_second;
2577
2578                         CPU_FOREACH(cpu)
2579                                 V_pf_stateid[cpu] = time_second;
2580
2581                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2582                 }
2583                 break;
2584
2585         case DIOCSTOP:
2586                 sx_xlock(&pf_ioctl_lock);
2587                 if (!V_pf_status.running)
2588                         error = ENOENT;
2589                 else {
2590                         V_pf_status.running = 0;
2591                         dehook_pf();
2592                         V_pf_status.since = time_second;
2593                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2594                 }
2595                 break;
2596
2597         case DIOCADDRULENV: {
2598                 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2599                 nvlist_t        *nvl = NULL;
2600                 void            *nvlpacked = NULL;
2601                 struct pf_krule *rule = NULL;
2602                 const char      *anchor = "", *anchor_call = "";
2603                 uint32_t         ticket = 0, pool_ticket = 0;
2604
2605 #define ERROUT(x)       do { error = (x); goto DIOCADDRULENV_error; } while (0)
2606
2607                 if (nv->len > pf_ioctl_maxcount)
2608                         ERROUT(ENOMEM);
2609
2610                 nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2611                 error = copyin(nv->data, nvlpacked, nv->len);
2612                 if (error)
2613                         ERROUT(error);
2614
2615                 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2616                 if (nvl == NULL)
2617                         ERROUT(EBADMSG);
2618
2619                 if (! nvlist_exists_number(nvl, "ticket"))
2620                         ERROUT(EINVAL);
2621                 ticket = nvlist_get_number(nvl, "ticket");
2622
2623                 if (! nvlist_exists_number(nvl, "pool_ticket"))
2624                         ERROUT(EINVAL);
2625                 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
2626
2627                 if (! nvlist_exists_nvlist(nvl, "rule"))
2628                         ERROUT(EINVAL);
2629
2630                 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
2631                     &rule);
2632                 if (error)
2633                         ERROUT(error);
2634
2635                 if (nvlist_exists_string(nvl, "anchor"))
2636                         anchor = nvlist_get_string(nvl, "anchor");
2637                 if (nvlist_exists_string(nvl, "anchor_call"))
2638                         anchor_call = nvlist_get_string(nvl, "anchor_call");
2639
2640                 if ((error = nvlist_error(nvl)))
2641                         ERROUT(error);
2642
2643                 /* Frees rule on error */
2644                 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
2645                     anchor_call, td);
2646
2647                 nvlist_destroy(nvl);
2648                 free(nvlpacked, M_TEMP);
2649                 break;
2650 #undef ERROUT
2651 DIOCADDRULENV_error:
2652                 pf_krule_free(rule);
2653                 nvlist_destroy(nvl);
2654                 free(nvlpacked, M_TEMP);
2655
2656                 break;
2657         }
2658         case DIOCADDRULE: {
2659                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
2660                 struct pf_krule         *rule;
2661
2662                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2663                 error = pf_rule_to_krule(&pr->rule, rule);
2664                 if (error != 0) {
2665                         free(rule, M_PFRULE);
2666                         break;
2667                 }
2668
2669                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
2670
2671                 /* Frees rule on error */
2672                 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
2673                     pr->anchor, pr->anchor_call, td);
2674                 break;
2675         }
2676
2677         case DIOCGETRULES: {
2678                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
2679                 struct pf_kruleset      *ruleset;
2680                 struct pf_krule         *tail;
2681                 int                      rs_num;
2682
2683                 PF_RULES_WLOCK();
2684                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
2685                 ruleset = pf_find_kruleset(pr->anchor);
2686                 if (ruleset == NULL) {
2687                         PF_RULES_WUNLOCK();
2688                         error = EINVAL;
2689                         break;
2690                 }
2691                 rs_num = pf_get_ruleset_number(pr->rule.action);
2692                 if (rs_num >= PF_RULESET_MAX) {
2693                         PF_RULES_WUNLOCK();
2694                         error = EINVAL;
2695                         break;
2696                 }
2697                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2698                     pf_krulequeue);
2699                 if (tail)
2700                         pr->nr = tail->nr + 1;
2701                 else
2702                         pr->nr = 0;
2703                 pr->ticket = ruleset->rules[rs_num].active.ticket;
2704                 PF_RULES_WUNLOCK();
2705                 break;
2706         }
2707
2708         case DIOCGETRULE: {
2709                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
2710                 struct pf_kruleset      *ruleset;
2711                 struct pf_krule         *rule;
2712                 int                      rs_num;
2713
2714                 PF_RULES_WLOCK();
2715                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
2716                 ruleset = pf_find_kruleset(pr->anchor);
2717                 if (ruleset == NULL) {
2718                         PF_RULES_WUNLOCK();
2719                         error = EINVAL;
2720                         break;
2721                 }
2722                 rs_num = pf_get_ruleset_number(pr->rule.action);
2723                 if (rs_num >= PF_RULESET_MAX) {
2724                         PF_RULES_WUNLOCK();
2725                         error = EINVAL;
2726                         break;
2727                 }
2728                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2729                         PF_RULES_WUNLOCK();
2730                         error = EBUSY;
2731                         break;
2732                 }
2733                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2734                 while ((rule != NULL) && (rule->nr != pr->nr))
2735                         rule = TAILQ_NEXT(rule, entries);
2736                 if (rule == NULL) {
2737                         PF_RULES_WUNLOCK();
2738                         error = EBUSY;
2739                         break;
2740                 }
2741
2742                 pf_krule_to_rule(rule, &pr->rule);
2743
2744                 if (pf_kanchor_copyout(ruleset, rule, pr)) {
2745                         PF_RULES_WUNLOCK();
2746                         error = EBUSY;
2747                         break;
2748                 }
2749                 pf_addr_copyout(&pr->rule.src.addr);
2750                 pf_addr_copyout(&pr->rule.dst.addr);
2751
2752                 if (pr->action == PF_GET_CLR_CNTR) {
2753                         counter_u64_zero(rule->evaluations);
2754                         for (int i = 0; i < 2; i++) {
2755                                 counter_u64_zero(rule->packets[i]);
2756                                 counter_u64_zero(rule->bytes[i]);
2757                         }
2758                         counter_u64_zero(rule->states_tot);
2759                 }
2760                 PF_RULES_WUNLOCK();
2761                 break;
2762         }
2763
2764         case DIOCGETRULENV: {
2765                 struct pfioc_nv         *nv = (struct pfioc_nv *)addr;
2766                 nvlist_t                *nvrule = NULL;
2767                 nvlist_t                *nvl = NULL;
2768                 struct pf_kruleset      *ruleset;
2769                 struct pf_krule         *rule;
2770                 void                    *nvlpacked = NULL;
2771                 int                      rs_num, nr;
2772                 bool                     clear_counter = false;
2773
2774 #define ERROUT(x)       do { error = (x); goto DIOCGETRULENV_error; } while (0)
2775
2776                 if (nv->len > pf_ioctl_maxcount)
2777                         ERROUT(ENOMEM);
2778
2779                 /* Copy the request in */
2780                 nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2781                 if (nvlpacked == NULL)
2782                         ERROUT(ENOMEM);
2783
2784                 error = copyin(nv->data, nvlpacked, nv->len);
2785                 if (error)
2786                         ERROUT(error);
2787
2788                 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2789                 if (nvl == NULL)
2790                         ERROUT(EBADMSG);
2791
2792                 if (! nvlist_exists_string(nvl, "anchor"))
2793                         ERROUT(EBADMSG);
2794                 if (! nvlist_exists_number(nvl, "ruleset"))
2795                         ERROUT(EBADMSG);
2796                 if (! nvlist_exists_number(nvl, "ticket"))
2797                         ERROUT(EBADMSG);
2798                 if (! nvlist_exists_number(nvl, "nr"))
2799                         ERROUT(EBADMSG);
2800
2801                 if (nvlist_exists_bool(nvl, "clear_counter"))
2802                         clear_counter = nvlist_get_bool(nvl, "clear_counter");
2803
2804                 if (clear_counter && !(flags & FWRITE))
2805                         ERROUT(EACCES);
2806
2807                 nr = nvlist_get_number(nvl, "nr");
2808
2809                 PF_RULES_WLOCK();
2810                 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
2811                 if (ruleset == NULL) {
2812                         PF_RULES_WUNLOCK();
2813                         ERROUT(ENOENT);
2814                 }
2815
2816                 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
2817                 if (rs_num >= PF_RULESET_MAX) {
2818                         PF_RULES_WUNLOCK();
2819                         ERROUT(EINVAL);
2820                 }
2821
2822                 if (nvlist_get_number(nvl, "ticket") !=
2823                     ruleset->rules[rs_num].active.ticket) {
2824                         PF_RULES_WUNLOCK();
2825                         ERROUT(EBUSY);
2826                         break;
2827                 }
2828
2829                 if ((error = nvlist_error(nvl))) {
2830                         PF_RULES_WUNLOCK();
2831                         ERROUT(error);
2832                 }
2833
2834                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2835                 while ((rule != NULL) && (rule->nr != nr))
2836                         rule = TAILQ_NEXT(rule, entries);
2837                 if (rule == NULL) {
2838                         PF_RULES_WUNLOCK();
2839                         ERROUT(EBUSY);
2840                         break;
2841                 }
2842
2843                 nvrule = pf_krule_to_nvrule(rule);
2844
2845                 nvlist_destroy(nvl);
2846                 nvl = nvlist_create(0);
2847                 if (nvl == NULL) {
2848                         PF_RULES_WUNLOCK();
2849                         ERROUT(ENOMEM);
2850                 }
2851                 nvlist_add_number(nvl, "nr", nr);
2852                 nvlist_add_nvlist(nvl, "rule", nvrule);
2853                 nvrule = NULL;
2854                 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
2855                         PF_RULES_WUNLOCK();
2856                         ERROUT(EBUSY);
2857                 }
2858
2859                 free(nvlpacked, M_TEMP);
2860                 nvlpacked = nvlist_pack(nvl, &nv->len);
2861                 if (nvlpacked == NULL) {
2862                         PF_RULES_WUNLOCK();
2863                         ERROUT(ENOMEM);
2864                 }
2865
2866                 if (nv->size == 0) {
2867                         PF_RULES_WUNLOCK();
2868                         ERROUT(0);
2869                 }
2870                 else if (nv->size < nv->len) {
2871                         PF_RULES_WUNLOCK();
2872                         ERROUT(ENOSPC);
2873                 }
2874
2875                 error = copyout(nvlpacked, nv->data, nv->len);
2876
2877                 if (clear_counter) {
2878                         counter_u64_zero(rule->evaluations);
2879                         for (int i = 0; i < 2; i++) {
2880                                 counter_u64_zero(rule->packets[i]);
2881                                 counter_u64_zero(rule->bytes[i]);
2882                         }
2883                         counter_u64_zero(rule->states_tot);
2884                 }
2885                 PF_RULES_WUNLOCK();
2886
2887 #undef ERROUT
2888 DIOCGETRULENV_error:
2889                 free(nvlpacked, M_TEMP);
2890                 nvlist_destroy(nvrule);
2891                 nvlist_destroy(nvl);
2892
2893                 break;
2894         }
2895
2896         case DIOCCHANGERULE: {
2897                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
2898                 struct pf_kruleset      *ruleset;
2899                 struct pf_krule         *oldrule = NULL, *newrule = NULL;
2900                 struct pfi_kkif         *kif = NULL;
2901                 struct pf_kpooladdr     *pa;
2902                 u_int32_t                nr = 0;
2903                 int                      rs_num;
2904
2905                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
2906                     pcr->action > PF_CHANGE_GET_TICKET) {
2907                         error = EINVAL;
2908                         break;
2909                 }
2910                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2911                         error = EINVAL;
2912                         break;
2913                 }
2914
2915                 if (pcr->action != PF_CHANGE_REMOVE) {
2916                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2917                         error = pf_rule_to_krule(&pcr->rule, newrule);
2918                         if (error != 0) {
2919                                 free(newrule, M_PFRULE);
2920                                 break;
2921                         }
2922
2923                         if (newrule->ifname[0])
2924                                 kif = pf_kkif_create(M_WAITOK);
2925                         newrule->evaluations = counter_u64_alloc(M_WAITOK);
2926                         for (int i = 0; i < 2; i++) {
2927                                 newrule->packets[i] =
2928                                     counter_u64_alloc(M_WAITOK);
2929                                 newrule->bytes[i] =
2930                                     counter_u64_alloc(M_WAITOK);
2931                         }
2932                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
2933                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
2934                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2935                         newrule->cuid = td->td_ucred->cr_ruid;
2936                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2937                         TAILQ_INIT(&newrule->rpool.list);
2938                 }
2939 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
2940
2941                 PF_RULES_WLOCK();
2942                 if (!(pcr->action == PF_CHANGE_REMOVE ||
2943                     pcr->action == PF_CHANGE_GET_TICKET) &&
2944                     pcr->pool_ticket != V_ticket_pabuf)
2945                         ERROUT(EBUSY);
2946
2947                 ruleset = pf_find_kruleset(pcr->anchor);
2948                 if (ruleset == NULL)
2949                         ERROUT(EINVAL);
2950
2951                 rs_num = pf_get_ruleset_number(pcr->rule.action);
2952                 if (rs_num >= PF_RULESET_MAX)
2953                         ERROUT(EINVAL);
2954
2955                 if (pcr->action == PF_CHANGE_GET_TICKET) {
2956                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2957                         ERROUT(0);
2958                 } else if (pcr->ticket !=
2959                             ruleset->rules[rs_num].active.ticket)
2960                                 ERROUT(EINVAL);
2961
2962                 if (pcr->action != PF_CHANGE_REMOVE) {
2963                         if (newrule->ifname[0]) {
2964                                 newrule->kif = pfi_kkif_attach(kif,
2965                                     newrule->ifname);
2966                                 kif = NULL;
2967                                 pfi_kkif_ref(newrule->kif);
2968                         } else
2969                                 newrule->kif = NULL;
2970
2971                         if (newrule->rtableid > 0 &&
2972                             newrule->rtableid >= rt_numfibs)
2973                                 error = EBUSY;
2974
2975 #ifdef ALTQ
2976                         /* set queue IDs */
2977                         if (newrule->qname[0] != 0) {
2978                                 if ((newrule->qid =
2979                                     pf_qname2qid(newrule->qname)) == 0)
2980                                         error = EBUSY;
2981                                 else if (newrule->pqname[0] != 0) {
2982                                         if ((newrule->pqid =
2983                                             pf_qname2qid(newrule->pqname)) == 0)
2984                                                 error = EBUSY;
2985                                 } else
2986                                         newrule->pqid = newrule->qid;
2987                         }
2988 #endif /* ALTQ */
2989                         if (newrule->tagname[0])
2990                                 if ((newrule->tag =
2991                                     pf_tagname2tag(newrule->tagname)) == 0)
2992                                         error = EBUSY;
2993                         if (newrule->match_tagname[0])
2994                                 if ((newrule->match_tag = pf_tagname2tag(
2995                                     newrule->match_tagname)) == 0)
2996                                         error = EBUSY;
2997                         if (newrule->rt && !newrule->direction)
2998                                 error = EINVAL;
2999                         if (!newrule->log)
3000                                 newrule->logif = 0;
3001                         if (newrule->logif >= PFLOGIFS_MAX)
3002                                 error = EINVAL;
3003                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3004                                 error = ENOMEM;
3005                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3006                                 error = ENOMEM;
3007                         if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3008                                 error = EINVAL;
3009                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3010                                 if (pa->addr.type == PF_ADDR_TABLE) {
3011                                         pa->addr.p.tbl =
3012                                             pfr_attach_table(ruleset,
3013                                             pa->addr.v.tblname);
3014                                         if (pa->addr.p.tbl == NULL)
3015                                                 error = ENOMEM;
3016                                 }
3017
3018                         newrule->overload_tbl = NULL;
3019                         if (newrule->overload_tblname[0]) {
3020                                 if ((newrule->overload_tbl = pfr_attach_table(
3021                                     ruleset, newrule->overload_tblname)) ==
3022                                     NULL)
3023                                         error = EINVAL;
3024                                 else
3025                                         newrule->overload_tbl->pfrkt_flags |=
3026                                             PFR_TFLAG_ACTIVE;
3027                         }
3028
3029                         pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3030                         if (((((newrule->action == PF_NAT) ||
3031                             (newrule->action == PF_RDR) ||
3032                             (newrule->action == PF_BINAT) ||
3033                             (newrule->rt > PF_NOPFROUTE)) &&
3034                             !newrule->anchor)) &&
3035                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3036                                 error = EINVAL;
3037
3038                         if (error) {
3039                                 pf_free_rule(newrule);
3040                                 PF_RULES_WUNLOCK();
3041                                 break;
3042                         }
3043
3044                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3045                 }
3046                 pf_empty_kpool(&V_pf_pabuf);
3047
3048                 if (pcr->action == PF_CHANGE_ADD_HEAD)
3049                         oldrule = TAILQ_FIRST(
3050                             ruleset->rules[rs_num].active.ptr);
3051                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3052                         oldrule = TAILQ_LAST(
3053                             ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3054                 else {
3055                         oldrule = TAILQ_FIRST(
3056                             ruleset->rules[rs_num].active.ptr);
3057                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3058                                 oldrule = TAILQ_NEXT(oldrule, entries);
3059                         if (oldrule == NULL) {
3060                                 if (newrule != NULL)
3061                                         pf_free_rule(newrule);
3062                                 PF_RULES_WUNLOCK();
3063                                 error = EINVAL;
3064                                 break;
3065                         }
3066                 }
3067
3068                 if (pcr->action == PF_CHANGE_REMOVE) {
3069                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3070                             oldrule);
3071                         ruleset->rules[rs_num].active.rcount--;
3072                 } else {
3073                         if (oldrule == NULL)
3074                                 TAILQ_INSERT_TAIL(
3075                                     ruleset->rules[rs_num].active.ptr,
3076                                     newrule, entries);
3077                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3078                             pcr->action == PF_CHANGE_ADD_BEFORE)
3079                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3080                         else
3081                                 TAILQ_INSERT_AFTER(
3082                                     ruleset->rules[rs_num].active.ptr,
3083                                     oldrule, newrule, entries);
3084                         ruleset->rules[rs_num].active.rcount++;
3085                 }
3086
3087                 nr = 0;
3088                 TAILQ_FOREACH(oldrule,
3089                     ruleset->rules[rs_num].active.ptr, entries)
3090                         oldrule->nr = nr++;
3091
3092                 ruleset->rules[rs_num].active.ticket++;
3093
3094                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3095                 pf_remove_if_empty_kruleset(ruleset);
3096
3097                 PF_RULES_WUNLOCK();
3098                 break;
3099
3100 #undef ERROUT
3101 DIOCCHANGERULE_error:
3102                 PF_RULES_WUNLOCK();
3103                 pf_krule_free(newrule);
3104                 pf_kkif_free(kif);
3105                 break;
3106         }
3107
3108         case DIOCCLRSTATES: {
3109                 struct pf_state         *s;
3110                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3111                 u_int                    i, killed = 0;
3112
3113                 for (i = 0; i <= pf_hashmask; i++) {
3114                         struct pf_idhash *ih = &V_pf_idhash[i];
3115
3116 relock_DIOCCLRSTATES:
3117                         PF_HASHROW_LOCK(ih);
3118                         LIST_FOREACH(s, &ih->states, entry)
3119                                 if (!psk->psk_ifname[0] ||
3120                                     !strcmp(psk->psk_ifname,
3121                                     s->kif->pfik_name)) {
3122                                         /*
3123                                          * Don't send out individual
3124                                          * delete messages.
3125                                          */
3126                                         s->state_flags |= PFSTATE_NOSYNC;
3127                                         pf_unlink_state(s, PF_ENTER_LOCKED);
3128                                         killed++;
3129                                         goto relock_DIOCCLRSTATES;
3130                                 }
3131                         PF_HASHROW_UNLOCK(ih);
3132                 }
3133                 psk->psk_killed = killed;
3134                 if (V_pfsync_clear_states_ptr != NULL)
3135                         V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
3136                 break;
3137         }
3138
3139         case DIOCKILLSTATES: {
3140                 struct pf_state         *s;
3141                 struct pf_state_key     *sk;
3142                 struct pf_addr          *srcaddr, *dstaddr;
3143                 u_int16_t                srcport, dstport;
3144                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3145                 u_int                    i, killed = 0;
3146
3147                 if (psk->psk_pfcmp.id) {
3148                         if (psk->psk_pfcmp.creatorid == 0)
3149                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
3150                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
3151                             psk->psk_pfcmp.creatorid))) {
3152                                 pf_unlink_state(s, PF_ENTER_LOCKED);
3153                                 psk->psk_killed = 1;
3154                         }
3155                         break;
3156                 }
3157
3158                 for (i = 0; i <= pf_hashmask; i++) {
3159                         struct pf_idhash *ih = &V_pf_idhash[i];
3160
3161 relock_DIOCKILLSTATES:
3162                         PF_HASHROW_LOCK(ih);
3163                         LIST_FOREACH(s, &ih->states, entry) {
3164                                 sk = s->key[PF_SK_WIRE];
3165                                 if (s->direction == PF_OUT) {
3166                                         srcaddr = &sk->addr[1];
3167                                         dstaddr = &sk->addr[0];
3168                                         srcport = sk->port[1];
3169                                         dstport = sk->port[0];
3170                                 } else {
3171                                         srcaddr = &sk->addr[0];
3172                                         dstaddr = &sk->addr[1];
3173                                         srcport = sk->port[0];
3174                                         dstport = sk->port[1];
3175                                 }
3176
3177                                 if ((!psk->psk_af || sk->af == psk->psk_af)
3178                                     && (!psk->psk_proto || psk->psk_proto ==
3179                                     sk->proto) &&
3180                                     PF_MATCHA(psk->psk_src.neg,
3181                                     &psk->psk_src.addr.v.a.addr,
3182                                     &psk->psk_src.addr.v.a.mask,
3183                                     srcaddr, sk->af) &&
3184                                     PF_MATCHA(psk->psk_dst.neg,
3185                                     &psk->psk_dst.addr.v.a.addr,
3186                                     &psk->psk_dst.addr.v.a.mask,
3187                                     dstaddr, sk->af) &&
3188                                     (psk->psk_src.port_op == 0 ||
3189                                     pf_match_port(psk->psk_src.port_op,
3190                                     psk->psk_src.port[0], psk->psk_src.port[1],
3191                                     srcport)) &&
3192                                     (psk->psk_dst.port_op == 0 ||
3193                                     pf_match_port(psk->psk_dst.port_op,
3194                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
3195                                     dstport)) &&
3196                                     (!psk->psk_label[0] ||
3197                                     (s->rule.ptr->label[0] &&
3198                                     !strcmp(psk->psk_label,
3199                                     s->rule.ptr->label))) &&
3200                                     (!psk->psk_ifname[0] ||
3201                                     !strcmp(psk->psk_ifname,
3202                                     s->kif->pfik_name))) {
3203                                         pf_unlink_state(s, PF_ENTER_LOCKED);
3204                                         killed++;
3205                                         goto relock_DIOCKILLSTATES;
3206                                 }
3207                         }
3208                         PF_HASHROW_UNLOCK(ih);
3209                 }
3210                 psk->psk_killed = killed;
3211                 break;
3212         }
3213
3214         case DIOCADDSTATE: {
3215                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
3216                 struct pfsync_state     *sp = &ps->state;
3217
3218                 if (sp->timeout >= PFTM_MAX) {
3219                         error = EINVAL;
3220                         break;
3221                 }
3222                 if (V_pfsync_state_import_ptr != NULL) {
3223                         PF_RULES_RLOCK();
3224                         error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
3225                         PF_RULES_RUNLOCK();
3226                 } else
3227                         error = EOPNOTSUPP;
3228                 break;
3229         }
3230
3231         case DIOCGETSTATE: {
3232                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
3233                 struct pf_state         *s;
3234
3235                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3236                 if (s == NULL) {
3237                         error = ENOENT;
3238                         break;
3239                 }
3240
3241                 pfsync_state_export(&ps->state, s);
3242                 PF_STATE_UNLOCK(s);
3243                 break;
3244         }
3245
3246         case DIOCGETSTATES: {
3247                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
3248                 struct pf_state         *s;
3249                 struct pfsync_state     *pstore, *p;
3250                 int i, nr;
3251
3252                 if (ps->ps_len <= 0) {
3253                         nr = uma_zone_get_cur(V_pf_state_z);
3254                         ps->ps_len = sizeof(struct pfsync_state) * nr;
3255                         break;
3256                 }
3257
3258                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
3259                 nr = 0;
3260
3261                 for (i = 0; i <= pf_hashmask; i++) {
3262                         struct pf_idhash *ih = &V_pf_idhash[i];
3263
3264                         PF_HASHROW_LOCK(ih);
3265                         LIST_FOREACH(s, &ih->states, entry) {
3266                                 if (s->timeout == PFTM_UNLINKED)
3267                                         continue;
3268
3269                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
3270                                         PF_HASHROW_UNLOCK(ih);
3271                                         goto DIOCGETSTATES_full;
3272                                 }
3273                                 pfsync_state_export(p, s);
3274                                 p++;
3275                                 nr++;
3276                         }
3277                         PF_HASHROW_UNLOCK(ih);
3278                 }
3279 DIOCGETSTATES_full:
3280                 error = copyout(pstore, ps->ps_states,
3281                     sizeof(struct pfsync_state) * nr);
3282                 if (error) {
3283                         free(pstore, M_TEMP);
3284                         break;
3285                 }
3286                 ps->ps_len = sizeof(struct pfsync_state) * nr;
3287                 free(pstore, M_TEMP);
3288
3289                 break;
3290         }
3291
3292         case DIOCGETSTATUS: {
3293                 struct pf_status *s = (struct pf_status *)addr;
3294
3295                 PF_RULES_RLOCK();
3296                 s->running = V_pf_status.running;
3297                 s->since   = V_pf_status.since;
3298                 s->debug   = V_pf_status.debug;
3299                 s->hostid  = V_pf_status.hostid;
3300                 s->states  = V_pf_status.states;
3301                 s->src_nodes = V_pf_status.src_nodes;
3302
3303                 for (int i = 0; i < PFRES_MAX; i++)
3304                         s->counters[i] =
3305                             counter_u64_fetch(V_pf_status.counters[i]);
3306                 for (int i = 0; i < LCNT_MAX; i++)
3307                         s->lcounters[i] =
3308                             counter_u64_fetch(V_pf_status.lcounters[i]);
3309                 for (int i = 0; i < FCNT_MAX; i++)
3310                         s->fcounters[i] =
3311                             counter_u64_fetch(V_pf_status.fcounters[i]);
3312                 for (int i = 0; i < SCNT_MAX; i++)
3313                         s->scounters[i] =
3314                             counter_u64_fetch(V_pf_status.scounters[i]);
3315
3316                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3317                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3318                     PF_MD5_DIGEST_LENGTH);
3319
3320                 pfi_update_status(s->ifname, s);
3321                 PF_RULES_RUNLOCK();
3322                 break;
3323         }
3324
3325         case DIOCSETSTATUSIF: {
3326                 struct pfioc_if *pi = (struct pfioc_if *)addr;
3327
3328                 if (pi->ifname[0] == 0) {
3329                         bzero(V_pf_status.ifname, IFNAMSIZ);
3330                         break;
3331                 }
3332                 PF_RULES_WLOCK();
3333                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3334                 PF_RULES_WUNLOCK();
3335                 break;
3336         }
3337
3338         case DIOCCLRSTATUS: {
3339                 PF_RULES_WLOCK();
3340                 for (int i = 0; i < PFRES_MAX; i++)
3341                         counter_u64_zero(V_pf_status.counters[i]);
3342                 for (int i = 0; i < FCNT_MAX; i++)
3343                         counter_u64_zero(V_pf_status.fcounters[i]);
3344                 for (int i = 0; i < SCNT_MAX; i++)
3345                         counter_u64_zero(V_pf_status.scounters[i]);
3346                 for (int i = 0; i < LCNT_MAX; i++)
3347                         counter_u64_zero(V_pf_status.lcounters[i]);
3348                 V_pf_status.since = time_second;
3349                 if (*V_pf_status.ifname)
3350                         pfi_update_status(V_pf_status.ifname, NULL);
3351                 PF_RULES_WUNLOCK();
3352                 break;
3353         }
3354
3355         case DIOCNATLOOK: {
3356                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
3357                 struct pf_state_key     *sk;
3358                 struct pf_state         *state;
3359                 struct pf_state_key_cmp  key;
3360                 int                      m = 0, direction = pnl->direction;
3361                 int                      sidx, didx;
3362
3363                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
3364                 sidx = (direction == PF_IN) ? 1 : 0;
3365                 didx = (direction == PF_IN) ? 0 : 1;
3366
3367                 if (!pnl->proto ||
3368                     PF_AZERO(&pnl->saddr, pnl->af) ||
3369                     PF_AZERO(&pnl->daddr, pnl->af) ||
3370                     ((pnl->proto == IPPROTO_TCP ||
3371                     pnl->proto == IPPROTO_UDP) &&
3372                     (!pnl->dport || !pnl->sport)))
3373                         error = EINVAL;
3374                 else {
3375                         bzero(&key, sizeof(key));
3376                         key.af = pnl->af;
3377                         key.proto = pnl->proto;
3378                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3379                         key.port[sidx] = pnl->sport;
3380                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3381                         key.port[didx] = pnl->dport;
3382
3383                         state = pf_find_state_all(&key, direction, &m);
3384
3385                         if (m > 1)
3386                                 error = E2BIG;  /* more than one state */
3387                         else if (state != NULL) {
3388                                 /* XXXGL: not locked read */
3389                                 sk = state->key[sidx];
3390                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3391                                 pnl->rsport = sk->port[sidx];
3392                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3393                                 pnl->rdport = sk->port[didx];
3394                         } else
3395                                 error = ENOENT;
3396                 }
3397                 break;
3398         }
3399
3400         case DIOCSETTIMEOUT: {
3401                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
3402                 int              old;
3403
3404                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3405                     pt->seconds < 0) {
3406                         error = EINVAL;
3407                         break;
3408                 }
3409                 PF_RULES_WLOCK();
3410                 old = V_pf_default_rule.timeout[pt->timeout];
3411                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3412                         pt->seconds = 1;
3413                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3414                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3415                         wakeup(pf_purge_thread);
3416                 pt->seconds = old;
3417                 PF_RULES_WUNLOCK();
3418                 break;
3419         }
3420
3421         case DIOCGETTIMEOUT: {
3422                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
3423
3424                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3425                         error = EINVAL;
3426                         break;
3427                 }
3428                 PF_RULES_RLOCK();
3429                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3430                 PF_RULES_RUNLOCK();
3431                 break;
3432         }
3433
3434         case DIOCGETLIMIT: {
3435                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
3436
3437                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3438                         error = EINVAL;
3439                         break;
3440                 }
3441                 PF_RULES_RLOCK();
3442                 pl->limit = V_pf_limits[pl->index].limit;
3443                 PF_RULES_RUNLOCK();
3444                 break;
3445         }
3446
3447         case DIOCSETLIMIT: {
3448                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
3449                 int                      old_limit;
3450
3451                 PF_RULES_WLOCK();
3452                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3453                     V_pf_limits[pl->index].zone == NULL) {
3454                         PF_RULES_WUNLOCK();
3455                         error = EINVAL;
3456                         break;
3457                 }
3458                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3459                 old_limit = V_pf_limits[pl->index].limit;
3460                 V_pf_limits[pl->index].limit = pl->limit;
3461                 pl->limit = old_limit;
3462                 PF_RULES_WUNLOCK();
3463                 break;
3464         }
3465
3466         case DIOCSETDEBUG: {
3467                 u_int32_t       *level = (u_int32_t *)addr;
3468
3469                 PF_RULES_WLOCK();
3470                 V_pf_status.debug = *level;
3471                 PF_RULES_WUNLOCK();
3472                 break;
3473         }
3474
3475         case DIOCCLRRULECTRS: {
3476                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3477                 struct pf_kruleset      *ruleset = &pf_main_ruleset;
3478                 struct pf_krule         *rule;
3479
3480                 PF_RULES_WLOCK();
3481                 TAILQ_FOREACH(rule,
3482                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3483                         counter_u64_zero(rule->evaluations);
3484                         for (int i = 0; i < 2; i++) {
3485                                 counter_u64_zero(rule->packets[i]);
3486                                 counter_u64_zero(rule->bytes[i]);
3487                         }
3488                 }
3489                 PF_RULES_WUNLOCK();
3490                 break;
3491         }
3492
3493         case DIOCGIFSPEEDV0:
3494         case DIOCGIFSPEEDV1: {
3495                 struct pf_ifspeed_v1    *psp = (struct pf_ifspeed_v1 *)addr;
3496                 struct pf_ifspeed_v1    ps;
3497                 struct ifnet            *ifp;
3498
3499                 if (psp->ifname[0] != 0) {
3500                         /* Can we completely trust user-land? */
3501                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3502                         ifp = ifunit(ps.ifname);
3503                         if (ifp != NULL) {
3504                                 psp->baudrate32 =
3505                                     (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3506                                 if (cmd == DIOCGIFSPEEDV1)
3507                                         psp->baudrate = ifp->if_baudrate;
3508                         } else
3509                                 error = EINVAL;
3510                 } else
3511                         error = EINVAL;
3512                 break;
3513         }
3514
3515 #ifdef ALTQ
3516         case DIOCSTARTALTQ: {
3517                 struct pf_altq          *altq;
3518
3519                 PF_RULES_WLOCK();
3520                 /* enable all altq interfaces on active list */
3521                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3522                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3523                                 error = pf_enable_altq(altq);
3524                                 if (error != 0)
3525                                         break;
3526                         }
3527                 }
3528                 if (error == 0)
3529                         V_pf_altq_running = 1;
3530                 PF_RULES_WUNLOCK();
3531                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3532                 break;
3533         }
3534
3535         case DIOCSTOPALTQ: {
3536                 struct pf_altq          *altq;
3537
3538                 PF_RULES_WLOCK();
3539                 /* disable all altq interfaces on active list */
3540                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3541                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3542                                 error = pf_disable_altq(altq);
3543                                 if (error != 0)
3544                                         break;
3545                         }
3546                 }
3547                 if (error == 0)
3548                         V_pf_altq_running = 0;
3549                 PF_RULES_WUNLOCK();
3550                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3551                 break;
3552         }
3553
3554         case DIOCADDALTQV0:
3555         case DIOCADDALTQV1: {
3556                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
3557                 struct pf_altq          *altq, *a;
3558                 struct ifnet            *ifp;
3559
3560                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
3561                 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
3562                 if (error)
3563                         break;
3564                 altq->local_flags = 0;
3565
3566                 PF_RULES_WLOCK();
3567                 if (pa->ticket != V_ticket_altqs_inactive) {
3568                         PF_RULES_WUNLOCK();
3569                         free(altq, M_PFALTQ);
3570                         error = EBUSY;
3571                         break;
3572                 }
3573
3574                 /*
3575                  * if this is for a queue, find the discipline and
3576                  * copy the necessary fields
3577                  */
3578                 if (altq->qname[0] != 0) {
3579                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
3580                                 PF_RULES_WUNLOCK();
3581                                 error = EBUSY;
3582                                 free(altq, M_PFALTQ);
3583                                 break;
3584                         }
3585                         altq->altq_disc = NULL;
3586                         TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
3587                                 if (strncmp(a->ifname, altq->ifname,
3588                                     IFNAMSIZ) == 0) {
3589                                         altq->altq_disc = a->altq_disc;
3590                                         break;
3591                                 }
3592                         }
3593                 }
3594
3595                 if ((ifp = ifunit(altq->ifname)) == NULL)
3596                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
3597                 else
3598                         error = altq_add(ifp, altq);
3599
3600                 if (error) {
3601                         PF_RULES_WUNLOCK();
3602                         free(altq, M_PFALTQ);
3603                         break;
3604                 }
3605
3606                 if (altq->qname[0] != 0)
3607                         TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
3608                 else
3609                         TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
3610                 /* version error check done on import above */
3611                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3612                 PF_RULES_WUNLOCK();
3613                 break;
3614         }
3615
3616         case DIOCGETALTQSV0:
3617         case DIOCGETALTQSV1: {
3618                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
3619                 struct pf_altq          *altq;
3620
3621                 PF_RULES_RLOCK();
3622                 pa->nr = 0;
3623                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
3624                         pa->nr++;
3625                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
3626                         pa->nr++;
3627                 pa->ticket = V_ticket_altqs_active;
3628                 PF_RULES_RUNLOCK();
3629                 break;
3630         }
3631
3632         case DIOCGETALTQV0:
3633         case DIOCGETALTQV1: {
3634                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
3635                 struct pf_altq          *altq;
3636
3637                 PF_RULES_RLOCK();
3638                 if (pa->ticket != V_ticket_altqs_active) {
3639                         PF_RULES_RUNLOCK();
3640                         error = EBUSY;
3641                         break;
3642                 }
3643                 altq = pf_altq_get_nth_active(pa->nr);
3644                 if (altq == NULL) {
3645                         PF_RULES_RUNLOCK();
3646                         error = EBUSY;
3647                         break;
3648                 }
3649                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3650                 PF_RULES_RUNLOCK();
3651                 break;
3652         }
3653
3654         case DIOCCHANGEALTQV0:
3655         case DIOCCHANGEALTQV1:
3656                 /* CHANGEALTQ not supported yet! */
3657                 error = ENODEV;
3658                 break;
3659
3660         case DIOCGETQSTATSV0:
3661         case DIOCGETQSTATSV1: {
3662                 struct pfioc_qstats_v1  *pq = (struct pfioc_qstats_v1 *)addr;
3663                 struct pf_altq          *altq;
3664                 int                      nbytes;
3665                 u_int32_t                version;
3666
3667                 PF_RULES_RLOCK();
3668                 if (pq->ticket != V_ticket_altqs_active) {
3669                         PF_RULES_RUNLOCK();
3670                         error = EBUSY;
3671                         break;
3672                 }
3673                 nbytes = pq->nbytes;
3674                 altq = pf_altq_get_nth_active(pq->nr);
3675                 if (altq == NULL) {
3676                         PF_RULES_RUNLOCK();
3677                         error = EBUSY;
3678                         break;
3679                 }
3680
3681                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
3682                         PF_RULES_RUNLOCK();
3683                         error = ENXIO;
3684                         break;
3685                 }
3686                 PF_RULES_RUNLOCK();
3687                 if (cmd == DIOCGETQSTATSV0)
3688                         version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
3689                 else
3690                         version = pq->version;
3691                 error = altq_getqstats(altq, pq->buf, &nbytes, version);
3692                 if (error == 0) {
3693                         pq->scheduler = altq->scheduler;
3694                         pq->nbytes = nbytes;
3695                 }
3696                 break;
3697         }
3698 #endif /* ALTQ */
3699
3700         case DIOCBEGINADDRS: {
3701                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
3702
3703                 PF_RULES_WLOCK();
3704                 pf_empty_kpool(&V_pf_pabuf);
3705                 pp->ticket = ++V_ticket_pabuf;
3706                 PF_RULES_WUNLOCK();
3707                 break;
3708         }
3709
3710         case DIOCADDADDR: {
3711                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
3712                 struct pf_kpooladdr     *pa;
3713                 struct pfi_kkif         *kif = NULL;
3714
3715 #ifndef INET
3716                 if (pp->af == AF_INET) {
3717                         error = EAFNOSUPPORT;
3718                         break;
3719                 }
3720 #endif /* INET */
3721 #ifndef INET6
3722                 if (pp->af == AF_INET6) {
3723                         error = EAFNOSUPPORT;
3724                         break;
3725                 }
3726 #endif /* INET6 */
3727                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
3728                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
3729                     pp->addr.addr.type != PF_ADDR_TABLE) {
3730                         error = EINVAL;
3731                         break;
3732                 }
3733                 if (pp->addr.addr.p.dyn != NULL) {
3734                         error = EINVAL;
3735                         break;
3736                 }
3737                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
3738                 pf_pooladdr_to_kpooladdr(&pp->addr, pa);
3739                 if (pa->ifname[0])
3740                         kif = pf_kkif_create(M_WAITOK);
3741                 PF_RULES_WLOCK();
3742                 if (pp->ticket != V_ticket_pabuf) {
3743                         PF_RULES_WUNLOCK();
3744                         if (pa->ifname[0])
3745                                 pf_kkif_free(kif);
3746                         free(pa, M_PFRULE);
3747                         error = EBUSY;
3748                         break;
3749                 }
3750                 if (pa->ifname[0]) {
3751                         pa->kif = pfi_kkif_attach(kif, pa->ifname);
3752                         kif = NULL;
3753                         pfi_kkif_ref(pa->kif);
3754                 } else
3755                         pa->kif = NULL;
3756                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
3757                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
3758                         if (pa->ifname[0])
3759                                 pfi_kkif_unref(pa->kif);
3760                         PF_RULES_WUNLOCK();
3761                         free(pa, M_PFRULE);
3762                         break;
3763                 }
3764                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
3765                 PF_RULES_WUNLOCK();
3766                 break;
3767         }
3768
3769         case DIOCGETADDRS: {
3770                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
3771                 struct pf_kpool         *pool;
3772                 struct pf_kpooladdr     *pa;
3773
3774                 PF_RULES_RLOCK();
3775                 pp->nr = 0;
3776                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3777                     pp->r_num, 0, 1, 0);
3778                 if (pool == NULL) {
3779                         PF_RULES_RUNLOCK();
3780                         error = EBUSY;
3781                         break;
3782                 }
3783                 TAILQ_FOREACH(pa, &pool->list, entries)
3784                         pp->nr++;
3785                 PF_RULES_RUNLOCK();
3786                 break;
3787         }
3788
3789         case DIOCGETADDR: {
3790                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
3791                 struct pf_kpool         *pool;
3792                 struct pf_kpooladdr     *pa;
3793                 u_int32_t                nr = 0;
3794
3795                 PF_RULES_RLOCK();
3796                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3797                     pp->r_num, 0, 1, 1);
3798                 if (pool == NULL) {
3799                         PF_RULES_RUNLOCK();
3800                         error = EBUSY;
3801                         break;
3802                 }
3803                 pa = TAILQ_FIRST(&pool->list);
3804                 while ((pa != NULL) && (nr < pp->nr)) {
3805                         pa = TAILQ_NEXT(pa, entries);
3806                         nr++;
3807                 }
3808                 if (pa == NULL) {
3809                         PF_RULES_RUNLOCK();
3810                         error = EBUSY;
3811                         break;
3812                 }
3813                 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3814                 pf_addr_copyout(&pp->addr.addr);
3815                 PF_RULES_RUNLOCK();
3816                 break;
3817         }
3818
3819         case DIOCCHANGEADDR: {
3820                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
3821                 struct pf_kpool         *pool;
3822                 struct pf_kpooladdr     *oldpa = NULL, *newpa = NULL;
3823                 struct pf_kruleset      *ruleset;
3824                 struct pfi_kkif         *kif = NULL;
3825
3826                 if (pca->action < PF_CHANGE_ADD_HEAD ||
3827                     pca->action > PF_CHANGE_REMOVE) {
3828                         error = EINVAL;
3829                         break;
3830                 }
3831                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3832                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3833                     pca->addr.addr.type != PF_ADDR_TABLE) {
3834                         error = EINVAL;
3835                         break;
3836                 }
3837                 if (pca->addr.addr.p.dyn != NULL) {
3838                         error = EINVAL;
3839                         break;
3840                 }
3841
3842                 if (pca->action != PF_CHANGE_REMOVE) {
3843 #ifndef INET
3844                         if (pca->af == AF_INET) {
3845                                 error = EAFNOSUPPORT;
3846                                 break;
3847                         }
3848 #endif /* INET */
3849 #ifndef INET6
3850                         if (pca->af == AF_INET6) {
3851                                 error = EAFNOSUPPORT;
3852                                 break;
3853                         }
3854 #endif /* INET6 */
3855                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3856                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3857                         if (newpa->ifname[0])
3858                                 kif = pf_kkif_create(M_WAITOK);
3859                         newpa->kif = NULL;
3860                 }
3861 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
3862                 PF_RULES_WLOCK();
3863                 ruleset = pf_find_kruleset(pca->anchor);
3864                 if (ruleset == NULL)
3865                         ERROUT(EBUSY);
3866
3867                 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3868                     pca->r_num, pca->r_last, 1, 1);
3869                 if (pool == NULL)
3870                         ERROUT(EBUSY);
3871
3872                 if (pca->action != PF_CHANGE_REMOVE) {
3873                         if (newpa->ifname[0]) {
3874                                 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3875                                 pfi_kkif_ref(newpa->kif);
3876                                 kif = NULL;
3877                         }
3878
3879                         switch (newpa->addr.type) {
3880                         case PF_ADDR_DYNIFTL:
3881                                 error = pfi_dynaddr_setup(&newpa->addr,
3882                                     pca->af);
3883                                 break;
3884                         case PF_ADDR_TABLE:
3885                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
3886                                     newpa->addr.v.tblname);
3887                                 if (newpa->addr.p.tbl == NULL)
3888                                         error = ENOMEM;
3889                                 break;
3890                         }
3891                         if (error)
3892                                 goto DIOCCHANGEADDR_error;
3893                 }
3894
3895                 switch (pca->action) {
3896                 case PF_CHANGE_ADD_HEAD:
3897                         oldpa = TAILQ_FIRST(&pool->list);
3898                         break;
3899                 case PF_CHANGE_ADD_TAIL:
3900                         oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3901                         break;
3902                 default:
3903                         oldpa = TAILQ_FIRST(&pool->list);
3904                         for (int i = 0; oldpa && i < pca->nr; i++)
3905                                 oldpa = TAILQ_NEXT(oldpa, entries);
3906
3907                         if (oldpa == NULL)
3908                                 ERROUT(EINVAL);
3909                 }
3910
3911                 if (pca->action == PF_CHANGE_REMOVE) {
3912                         TAILQ_REMOVE(&pool->list, oldpa, entries);
3913                         switch (oldpa->addr.type) {
3914                         case PF_ADDR_DYNIFTL:
3915                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
3916                                 break;
3917                         case PF_ADDR_TABLE:
3918                                 pfr_detach_table(oldpa->addr.p.tbl);
3919                                 break;
3920                         }
3921                         if (oldpa->kif)
3922                                 pfi_kkif_unref(oldpa->kif);
3923                         free(oldpa, M_PFRULE);
3924                 } else {
3925                         if (oldpa == NULL)
3926                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3927                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
3928                             pca->action == PF_CHANGE_ADD_BEFORE)
3929                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3930                         else
3931                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
3932                                     newpa, entries);
3933                 }
3934
3935                 pool->cur = TAILQ_FIRST(&pool->list);
3936                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3937                 PF_RULES_WUNLOCK();
3938                 break;
3939
3940 #undef ERROUT
3941 DIOCCHANGEADDR_error:
3942                 if (newpa != NULL) {
3943                         if (newpa->kif)
3944                                 pfi_kkif_unref(newpa->kif);
3945                         free(newpa, M_PFRULE);
3946                 }
3947                 PF_RULES_WUNLOCK();
3948                 pf_kkif_free(kif);
3949                 break;
3950         }
3951
3952         case DIOCGETRULESETS: {
3953                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
3954                 struct pf_kruleset      *ruleset;
3955                 struct pf_kanchor       *anchor;
3956
3957                 PF_RULES_RLOCK();
3958                 pr->path[sizeof(pr->path) - 1] = 0;
3959                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3960                         PF_RULES_RUNLOCK();
3961                         error = ENOENT;
3962                         break;
3963                 }
3964                 pr->nr = 0;
3965                 if (ruleset->anchor == NULL) {
3966                         /* XXX kludge for pf_main_ruleset */
3967                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3968                                 if (anchor->parent == NULL)
3969                                         pr->nr++;
3970                 } else {
3971                         RB_FOREACH(anchor, pf_kanchor_node,
3972                             &ruleset->anchor->children)
3973                                 pr->nr++;
3974                 }
3975                 PF_RULES_RUNLOCK();
3976                 break;
3977         }
3978
3979         case DIOCGETRULESET: {
3980                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
3981                 struct pf_kruleset      *ruleset;
3982                 struct pf_kanchor       *anchor;
3983                 u_int32_t                nr = 0;
3984
3985                 PF_RULES_RLOCK();
3986                 pr->path[sizeof(pr->path) - 1] = 0;
3987                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3988                         PF_RULES_RUNLOCK();
3989                         error = ENOENT;
3990                         break;
3991                 }
3992                 pr->name[0] = 0;
3993                 if (ruleset->anchor == NULL) {
3994                         /* XXX kludge for pf_main_ruleset */
3995                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3996                                 if (anchor->parent == NULL && nr++ == pr->nr) {
3997                                         strlcpy(pr->name, anchor->name,
3998                                             sizeof(pr->name));
3999                                         break;
4000                                 }
4001                 } else {
4002                         RB_FOREACH(anchor, pf_kanchor_node,
4003                             &ruleset->anchor->children)
4004                                 if (nr++ == pr->nr) {
4005                                         strlcpy(pr->name, anchor->name,
4006                                             sizeof(pr->name));
4007                                         break;
4008                                 }
4009                 }
4010                 if (!pr->name[0])
4011                         error = EBUSY;
4012                 PF_RULES_RUNLOCK();
4013                 break;
4014         }
4015
4016         case DIOCRCLRTABLES: {
4017                 struct pfioc_table *io = (struct pfioc_table *)addr;
4018
4019                 if (io->pfrio_esize != 0) {
4020                         error = ENODEV;
4021                         break;
4022                 }
4023                 PF_RULES_WLOCK();
4024                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4025                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
4026                 PF_RULES_WUNLOCK();
4027                 break;
4028         }
4029
4030         case DIOCRADDTABLES: {
4031                 struct pfioc_table *io = (struct pfioc_table *)addr;
4032                 struct pfr_table *pfrts;
4033                 size_t totlen;
4034
4035                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4036                         error = ENODEV;
4037                         break;
4038                 }
4039
4040                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4041                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4042                         error = ENOMEM;
4043                         break;
4044                 }
4045
4046                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4047                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4048                     M_TEMP, M_WAITOK);
4049                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4050                 if (error) {
4051                         free(pfrts, M_TEMP);
4052                         break;
4053                 }
4054                 PF_RULES_WLOCK();
4055                 error = pfr_add_tables(pfrts, io->pfrio_size,
4056                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4057                 PF_RULES_WUNLOCK();
4058                 free(pfrts, M_TEMP);
4059                 break;
4060         }
4061
4062         case DIOCRDELTABLES: {
4063                 struct pfioc_table *io = (struct pfioc_table *)addr;
4064                 struct pfr_table *pfrts;
4065                 size_t totlen;
4066
4067                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4068                         error = ENODEV;
4069                         break;
4070                 }
4071
4072                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4073                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4074                         error = ENOMEM;
4075                         break;
4076                 }
4077
4078                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4079                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4080                     M_TEMP, M_WAITOK);
4081                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4082                 if (error) {
4083                         free(pfrts, M_TEMP);
4084                         break;
4085                 }
4086                 PF_RULES_WLOCK();
4087                 error = pfr_del_tables(pfrts, io->pfrio_size,
4088                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4089                 PF_RULES_WUNLOCK();
4090                 free(pfrts, M_TEMP);
4091                 break;
4092         }
4093
4094         case DIOCRGETTABLES: {
4095                 struct pfioc_table *io = (struct pfioc_table *)addr;
4096                 struct pfr_table *pfrts;
4097                 size_t totlen;
4098                 int n;
4099
4100                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4101                         error = ENODEV;
4102                         break;
4103                 }
4104                 PF_RULES_RLOCK();
4105                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4106                 if (n < 0) {
4107                         PF_RULES_RUNLOCK();
4108                         error = EINVAL;
4109                         break;
4110                 }
4111                 io->pfrio_size = min(io->pfrio_size, n);
4112
4113                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4114
4115                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4116                     M_TEMP, M_NOWAIT);
4117                 if (pfrts == NULL) {
4118                         error = ENOMEM;
4119                         PF_RULES_RUNLOCK();
4120                         break;
4121                 }
4122                 error = pfr_get_tables(&io->pfrio_table, pfrts,
4123                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4124                 PF_RULES_RUNLOCK();
4125                 if (error == 0)
4126                         error = copyout(pfrts, io->pfrio_buffer, totlen);
4127                 free(pfrts, M_TEMP);
4128                 break;
4129         }
4130
4131         case DIOCRGETTSTATS: {
4132                 struct pfioc_table *io = (struct pfioc_table *)addr;
4133                 struct pfr_tstats *pfrtstats;
4134                 size_t totlen;
4135                 int n;
4136
4137                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4138                         error = ENODEV;
4139                         break;
4140                 }
4141                 PF_RULES_WLOCK();
4142                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4143                 if (n < 0) {
4144                         PF_RULES_WUNLOCK();
4145                         error = EINVAL;
4146                         break;
4147                 }
4148                 io->pfrio_size = min(io->pfrio_size, n);
4149
4150                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4151                 pfrtstats = mallocarray(io->pfrio_size,
4152                     sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
4153                 if (pfrtstats == NULL) {
4154                         error = ENOMEM;
4155                         PF_RULES_WUNLOCK();
4156                         break;
4157                 }
4158                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4159                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4160                 PF_RULES_WUNLOCK();
4161                 if (error == 0)
4162                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4163                 free(pfrtstats, M_TEMP);
4164                 break;
4165         }
4166
4167         case DIOCRCLRTSTATS: {
4168                 struct pfioc_table *io = (struct pfioc_table *)addr;
4169                 struct pfr_table *pfrts;
4170                 size_t totlen;
4171
4172                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4173                         error = ENODEV;
4174                         break;
4175                 }
4176
4177                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4178                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4179                         /* We used to count tables and use the minimum required
4180                          * size, so we didn't fail on overly large requests.
4181                          * Keep doing so. */
4182                         io->pfrio_size = pf_ioctl_maxcount;
4183                         break;
4184                 }
4185
4186                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4187                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4188                     M_TEMP, M_NOWAIT);
4189                 if (pfrts == NULL) {
4190                         error = ENOMEM;
4191                         break;
4192                 }
4193                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4194                 if (error) {
4195                         free(pfrts, M_TEMP);
4196                         break;
4197                 }
4198
4199                 PF_RULES_WLOCK();
4200                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4201                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4202                 PF_RULES_WUNLOCK();
4203                 free(pfrts, M_TEMP);
4204                 break;
4205         }
4206
4207         case DIOCRSETTFLAGS: {
4208                 struct pfioc_table *io = (struct pfioc_table *)addr;
4209                 struct pfr_table *pfrts;
4210                 size_t totlen;
4211                 int n;
4212
4213                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4214                         error = ENODEV;
4215                         break;
4216                 }
4217
4218                 PF_RULES_RLOCK();
4219                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4220                 if (n < 0) {
4221                         PF_RULES_RUNLOCK();
4222                         error = EINVAL;
4223                         break;
4224                 }
4225
4226                 io->pfrio_size = min(io->pfrio_size, n);
4227                 PF_RULES_RUNLOCK();
4228
4229                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4230                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4231                     M_TEMP, M_WAITOK);
4232                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4233                 if (error) {
4234                         free(pfrts, M_TEMP);
4235                         break;
4236                 }
4237                 PF_RULES_WLOCK();
4238                 error = pfr_set_tflags(pfrts, io->pfrio_size,
4239                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4240                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4241                 PF_RULES_WUNLOCK();
4242                 free(pfrts, M_TEMP);
4243                 break;
4244         }
4245
4246         case DIOCRCLRADDRS: {
4247                 struct pfioc_table *io = (struct pfioc_table *)addr;
4248
4249                 if (io->pfrio_esize != 0) {
4250                         error = ENODEV;
4251                         break;
4252                 }
4253                 PF_RULES_WLOCK();
4254                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4255                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
4256                 PF_RULES_WUNLOCK();
4257                 break;
4258         }
4259
4260         case DIOCRADDADDRS: {
4261                 struct pfioc_table *io = (struct pfioc_table *)addr;
4262                 struct pfr_addr *pfras;
4263                 size_t totlen;
4264
4265                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4266                         error = ENODEV;
4267                         break;
4268                 }
4269                 if (io->pfrio_size < 0 ||
4270                     io->pfrio_size > pf_ioctl_maxcount ||
4271                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4272                         error = EINVAL;
4273                         break;
4274                 }
4275                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4276                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4277                     M_TEMP, M_NOWAIT);
4278                 if (! pfras) {
4279                         error = ENOMEM;
4280                         break;
4281                 }
4282                 error = copyin(io->pfrio_buffer, pfras, totlen);
4283                 if (error) {
4284                         free(pfras, M_TEMP);
4285                         break;
4286                 }
4287                 PF_RULES_WLOCK();
4288                 error = pfr_add_addrs(&io->pfrio_table, pfras,
4289                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4290                     PFR_FLAG_USERIOCTL);
4291                 PF_RULES_WUNLOCK();
4292                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4293                         error = copyout(pfras, io->pfrio_buffer, totlen);
4294                 free(pfras, M_TEMP);
4295                 break;
4296         }
4297
4298         case DIOCRDELADDRS: {
4299                 struct pfioc_table *io = (struct pfioc_table *)addr;
4300                 struct pfr_addr *pfras;
4301                 size_t totlen;
4302
4303                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4304                         error = ENODEV;
4305                         break;
4306                 }
4307                 if (io->pfrio_size < 0 ||
4308                     io->pfrio_size > pf_ioctl_maxcount ||
4309                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4310                         error = EINVAL;
4311                         break;
4312                 }
4313                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4314                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4315                     M_TEMP, M_NOWAIT);
4316                 if (! pfras) {
4317                         error = ENOMEM;
4318                         break;
4319                 }
4320                 error = copyin(io->pfrio_buffer, pfras, totlen);
4321                 if (error) {
4322                         free(pfras, M_TEMP);
4323                         break;
4324                 }
4325                 PF_RULES_WLOCK();
4326                 error = pfr_del_addrs(&io->pfrio_table, pfras,
4327                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4328                     PFR_FLAG_USERIOCTL);
4329                 PF_RULES_WUNLOCK();
4330                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4331                         error = copyout(pfras, io->pfrio_buffer, totlen);
4332                 free(pfras, M_TEMP);
4333                 break;
4334         }
4335
4336         case DIOCRSETADDRS: {
4337                 struct pfioc_table *io = (struct pfioc_table *)addr;
4338                 struct pfr_addr *pfras;
4339                 size_t totlen, count;
4340
4341                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4342                         error = ENODEV;
4343                         break;
4344                 }
4345                 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4346                         error = EINVAL;
4347                         break;
4348                 }
4349                 count = max(io->pfrio_size, io->pfrio_size2);
4350                 if (count > pf_ioctl_maxcount ||
4351                     WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4352                         error = EINVAL;
4353                         break;
4354                 }
4355                 totlen = count * sizeof(struct pfr_addr);
4356                 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4357                     M_NOWAIT);
4358                 if (! pfras) {
4359                         error = ENOMEM;
4360                         break;
4361                 }
4362                 error = copyin(io->pfrio_buffer, pfras, totlen);
4363                 if (error) {
4364                         free(pfras, M_TEMP);
4365                         break;
4366                 }
4367                 PF_RULES_WLOCK();
4368                 error = pfr_set_addrs(&io->pfrio_table, pfras,
4369                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4370                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4371                     PFR_FLAG_USERIOCTL, 0);
4372                 PF_RULES_WUNLOCK();
4373                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4374                         error = copyout(pfras, io->pfrio_buffer, totlen);
4375                 free(pfras, M_TEMP);
4376                 break;
4377         }
4378
4379         case DIOCRGETADDRS: {
4380                 struct pfioc_table *io = (struct pfioc_table *)addr;
4381                 struct pfr_addr *pfras;
4382                 size_t totlen;
4383
4384                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4385                         error = ENODEV;
4386                         break;
4387                 }
4388                 if (io->pfrio_size < 0 ||
4389                     io->pfrio_size > pf_ioctl_maxcount ||
4390                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4391                         error = EINVAL;
4392                         break;
4393                 }
4394                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4395                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4396                     M_TEMP, M_NOWAIT);
4397                 if (! pfras) {
4398                         error = ENOMEM;
4399                         break;
4400                 }
4401                 PF_RULES_RLOCK();
4402                 error = pfr_get_addrs(&io->pfrio_table, pfras,
4403                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4404                 PF_RULES_RUNLOCK();
4405                 if (error == 0)
4406                         error = copyout(pfras, io->pfrio_buffer, totlen);
4407                 free(pfras, M_TEMP);
4408                 break;
4409         }
4410
4411         case DIOCRGETASTATS: {
4412                 struct pfioc_table *io = (struct pfioc_table *)addr;
4413                 struct pfr_astats *pfrastats;
4414                 size_t totlen;
4415
4416                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4417                         error = ENODEV;
4418                         break;
4419                 }
4420                 if (io->pfrio_size < 0 ||
4421                     io->pfrio_size > pf_ioctl_maxcount ||
4422                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4423                         error = EINVAL;
4424                         break;
4425                 }
4426                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
4427                 pfrastats = mallocarray(io->pfrio_size,
4428                     sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
4429                 if (! pfrastats) {
4430                         error = ENOMEM;
4431                         break;
4432                 }
4433                 PF_RULES_RLOCK();
4434                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
4435                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4436                 PF_RULES_RUNLOCK();
4437                 if (error == 0)
4438                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
4439                 free(pfrastats, M_TEMP);
4440                 break;
4441         }
4442
4443         case DIOCRCLRASTATS: {
4444                 struct pfioc_table *io = (struct pfioc_table *)addr;
4445                 struct pfr_addr *pfras;
4446                 size_t totlen;
4447
4448                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4449                         error = ENODEV;
4450                         break;
4451                 }
4452                 if (io->pfrio_size < 0 ||
4453                     io->pfrio_size > pf_ioctl_maxcount ||
4454                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4455                         error = EINVAL;
4456                         break;
4457                 }
4458                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4459                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4460                     M_TEMP, M_NOWAIT);
4461                 if (! pfras) {
4462                         error = ENOMEM;
4463                         break;
4464                 }
4465                 error = copyin(io->pfrio_buffer, pfras, totlen);
4466                 if (error) {
4467                         free(pfras, M_TEMP);
4468                         break;
4469                 }
4470                 PF_RULES_WLOCK();
4471                 error = pfr_clr_astats(&io->pfrio_table, pfras,
4472                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4473                     PFR_FLAG_USERIOCTL);
4474                 PF_RULES_WUNLOCK();
4475                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4476                         error = copyout(pfras, io->pfrio_buffer, totlen);
4477                 free(pfras, M_TEMP);
4478                 break;
4479         }
4480
4481         case DIOCRTSTADDRS: {
4482                 struct pfioc_table *io = (struct pfioc_table *)addr;
4483                 struct pfr_addr *pfras;
4484                 size_t totlen;
4485
4486                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4487                         error = ENODEV;
4488                         break;
4489                 }
4490                 if (io->pfrio_size < 0 ||
4491                     io->pfrio_size > pf_ioctl_maxcount ||
4492                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4493                         error = EINVAL;
4494                         break;
4495                 }
4496                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4497                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4498                     M_TEMP, M_NOWAIT);
4499                 if (! pfras) {
4500                         error = ENOMEM;
4501                         break;
4502                 }
4503                 error = copyin(io->pfrio_buffer, pfras, totlen);
4504                 if (error) {
4505                         free(pfras, M_TEMP);
4506                         break;
4507                 }
4508                 PF_RULES_RLOCK();
4509                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
4510                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4511                     PFR_FLAG_USERIOCTL);
4512                 PF_RULES_RUNLOCK();
4513                 if (error == 0)
4514                         error = copyout(pfras, io->pfrio_buffer, totlen);
4515                 free(pfras, M_TEMP);
4516                 break;
4517         }
4518
4519         case DIOCRINADEFINE: {
4520                 struct pfioc_table *io = (struct pfioc_table *)addr;
4521                 struct pfr_addr *pfras;
4522                 size_t totlen;
4523
4524                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4525                         error = ENODEV;
4526                         break;
4527                 }
4528                 if (io->pfrio_size < 0 ||
4529                     io->pfrio_size > pf_ioctl_maxcount ||
4530                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4531                         error = EINVAL;
4532                         break;
4533                 }
4534                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4535                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4536                     M_TEMP, M_NOWAIT);
4537                 if (! pfras) {
4538                         error = ENOMEM;
4539                         break;
4540                 }
4541                 error = copyin(io->pfrio_buffer, pfras, totlen);
4542                 if (error) {
4543                         free(pfras, M_TEMP);
4544                         break;
4545                 }
4546                 PF_RULES_WLOCK();
4547                 error = pfr_ina_define(&io->pfrio_table, pfras,
4548                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4549                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4550                 PF_RULES_WUNLOCK();
4551                 free(pfras, M_TEMP);
4552                 break;
4553         }
4554
4555         case DIOCOSFPADD: {
4556                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4557                 PF_RULES_WLOCK();
4558                 error = pf_osfp_add(io);
4559                 PF_RULES_WUNLOCK();
4560                 break;
4561         }
4562
4563         case DIOCOSFPGET: {
4564                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4565                 PF_RULES_RLOCK();
4566                 error = pf_osfp_get(io);
4567                 PF_RULES_RUNLOCK();
4568                 break;
4569         }
4570
4571         case DIOCXBEGIN: {
4572                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
4573                 struct pfioc_trans_e    *ioes, *ioe;
4574                 size_t                   totlen;
4575                 int                      i;
4576
4577                 if (io->esize != sizeof(*ioe)) {
4578                         error = ENODEV;
4579                         break;
4580                 }
4581                 if (io->size < 0 ||
4582                     io->size > pf_ioctl_maxcount ||
4583                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4584                         error = EINVAL;
4585                         break;
4586                 }
4587                 totlen = sizeof(struct pfioc_trans_e) * io->size;
4588                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4589                     M_TEMP, M_NOWAIT);
4590                 if (! ioes) {
4591                         error = ENOMEM;
4592                         break;
4593                 }
4594                 error = copyin(io->array, ioes, totlen);
4595                 if (error) {
4596                         free(ioes, M_TEMP);
4597                         break;
4598                 }
4599                 PF_RULES_WLOCK();
4600                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4601                         switch (ioe->rs_num) {
4602 #ifdef ALTQ
4603                         case PF_RULESET_ALTQ:
4604                                 if (ioe->anchor[0]) {
4605                                         PF_RULES_WUNLOCK();
4606                                         free(ioes, M_TEMP);
4607                                         error = EINVAL;
4608                                         goto fail;
4609                                 }
4610                                 if ((error = pf_begin_altq(&ioe->ticket))) {
4611                                         PF_RULES_WUNLOCK();
4612                                         free(ioes, M_TEMP);
4613                                         goto fail;
4614                                 }
4615                                 break;
4616 #endif /* ALTQ */
4617                         case PF_RULESET_TABLE:
4618                             {
4619                                 struct pfr_table table;
4620
4621                                 bzero(&table, sizeof(table));
4622                                 strlcpy(table.pfrt_anchor, ioe->anchor,
4623                                     sizeof(table.pfrt_anchor));
4624                                 if ((error = pfr_ina_begin(&table,
4625                                     &ioe->ticket, NULL, 0))) {
4626                                         PF_RULES_WUNLOCK();
4627                                         free(ioes, M_TEMP);
4628                                         goto fail;
4629                                 }
4630                                 break;
4631                             }
4632                         default:
4633                                 if ((error = pf_begin_rules(&ioe->ticket,
4634                                     ioe->rs_num, ioe->anchor))) {
4635                                         PF_RULES_WUNLOCK();
4636                                         free(ioes, M_TEMP);
4637                                         goto fail;
4638                                 }
4639                                 break;
4640                         }
4641                 }
4642                 PF_RULES_WUNLOCK();
4643                 error = copyout(ioes, io->array, totlen);
4644                 free(ioes, M_TEMP);
4645                 break;
4646         }
4647
4648         case DIOCXROLLBACK: {
4649                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
4650                 struct pfioc_trans_e    *ioe, *ioes;
4651                 size_t                   totlen;
4652                 int                      i;
4653
4654                 if (io->esize != sizeof(*ioe)) {
4655                         error = ENODEV;
4656                         break;
4657                 }
4658                 if (io->size < 0 ||
4659                     io->size > pf_ioctl_maxcount ||
4660                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4661                         error = EINVAL;
4662                         break;
4663                 }
4664                 totlen = sizeof(struct pfioc_trans_e) * io->size;
4665                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4666                     M_TEMP, M_NOWAIT);
4667                 if (! ioes) {
4668                         error = ENOMEM;
4669                         break;
4670                 }
4671                 error = copyin(io->array, ioes, totlen);
4672                 if (error) {
4673                         free(ioes, M_TEMP);
4674                         break;
4675                 }
4676                 PF_RULES_WLOCK();
4677                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4678                         switch (ioe->rs_num) {
4679 #ifdef ALTQ
4680                         case PF_RULESET_ALTQ:
4681                                 if (ioe->anchor[0]) {
4682                                         PF_RULES_WUNLOCK();
4683                                         free(ioes, M_TEMP);
4684                                         error = EINVAL;
4685                                         goto fail;
4686                                 }
4687                                 if ((error = pf_rollback_altq(ioe->ticket))) {
4688                                         PF_RULES_WUNLOCK();
4689                                         free(ioes, M_TEMP);
4690                                         goto fail; /* really bad */
4691                                 }
4692                                 break;
4693 #endif /* ALTQ */
4694                         case PF_RULESET_TABLE:
4695                             {
4696                                 struct pfr_table table;
4697
4698                                 bzero(&table, sizeof(table));
4699                                 strlcpy(table.pfrt_anchor, ioe->anchor,
4700                                     sizeof(table.pfrt_anchor));
4701                                 if ((error = pfr_ina_rollback(&table,
4702                                     ioe->ticket, NULL, 0))) {
4703                                         PF_RULES_WUNLOCK();
4704                                         free(ioes, M_TEMP);
4705                                         goto fail; /* really bad */
4706                                 }
4707                                 break;
4708                             }
4709                         default:
4710                                 if ((error = pf_rollback_rules(ioe->ticket,
4711                                     ioe->rs_num, ioe->anchor))) {
4712                                         PF_RULES_WUNLOCK();
4713                                         free(ioes, M_TEMP);
4714                                         goto fail; /* really bad */
4715                                 }
4716                                 break;
4717                         }
4718                 }
4719                 PF_RULES_WUNLOCK();
4720                 free(ioes, M_TEMP);
4721                 break;
4722         }
4723
4724         case DIOCXCOMMIT: {
4725                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
4726                 struct pfioc_trans_e    *ioe, *ioes;
4727                 struct pf_kruleset      *rs;
4728                 size_t                   totlen;
4729                 int                      i;
4730
4731                 if (io->esize != sizeof(*ioe)) {
4732                         error = ENODEV;
4733                         break;
4734                 }
4735
4736                 if (io->size < 0 ||
4737                     io->size > pf_ioctl_maxcount ||
4738                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4739                         error = EINVAL;
4740                         break;
4741                 }
4742
4743                 totlen = sizeof(struct pfioc_trans_e) * io->size;
4744                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4745                     M_TEMP, M_NOWAIT);
4746                 if (ioes == NULL) {
4747                         error = ENOMEM;
4748                         break;
4749                 }
4750                 error = copyin(io->array, ioes, totlen);
4751                 if (error) {
4752                         free(ioes, M_TEMP);
4753                         break;
4754                 }
4755                 PF_RULES_WLOCK();
4756                 /* First makes sure everything will succeed. */
4757                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4758                         switch (ioe->rs_num) {
4759 #ifdef ALTQ
4760                         case PF_RULESET_ALTQ:
4761                                 if (ioe->anchor[0]) {
4762                                         PF_RULES_WUNLOCK();
4763                                         free(ioes, M_TEMP);
4764                                         error = EINVAL;
4765                                         goto fail;
4766                                 }
4767                                 if (!V_altqs_inactive_open || ioe->ticket !=
4768                                     V_ticket_altqs_inactive) {
4769                                         PF_RULES_WUNLOCK();
4770                                         free(ioes, M_TEMP);
4771                                         error = EBUSY;
4772                                         goto fail;
4773                                 }
4774                                 break;
4775 #endif /* ALTQ */
4776                         case PF_RULESET_TABLE:
4777                                 rs = pf_find_kruleset(ioe->anchor);
4778                                 if (rs == NULL || !rs->topen || ioe->ticket !=
4779                                     rs->tticket) {
4780                                         PF_RULES_WUNLOCK();
4781                                         free(ioes, M_TEMP);
4782                                         error = EBUSY;
4783                                         goto fail;
4784                                 }
4785                                 break;
4786                         default:
4787                                 if (ioe->rs_num < 0 || ioe->rs_num >=
4788                                     PF_RULESET_MAX) {
4789                                         PF_RULES_WUNLOCK();
4790                                         free(ioes, M_TEMP);
4791                                         error = EINVAL;
4792                                         goto fail;
4793                                 }
4794                                 rs = pf_find_kruleset(ioe->anchor);
4795                                 if (rs == NULL ||
4796                                     !rs->rules[ioe->rs_num].inactive.open ||
4797                                     rs->rules[ioe->rs_num].inactive.ticket !=
4798                                     ioe->ticket) {
4799                                         PF_RULES_WUNLOCK();
4800                                         free(ioes, M_TEMP);
4801                                         error = EBUSY;
4802                                         goto fail;
4803                                 }
4804                                 break;
4805                         }
4806                 }
4807                 /* Now do the commit - no errors should happen here. */
4808                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4809                         switch (ioe->rs_num) {
4810 #ifdef ALTQ
4811                         case PF_RULESET_ALTQ:
4812                                 if ((error = pf_commit_altq(ioe->ticket))) {
4813                                         PF_RULES_WUNLOCK();
4814                                         free(ioes, M_TEMP);
4815                                         goto fail; /* really bad */
4816                                 }
4817                                 break;
4818 #endif /* ALTQ */
4819                         case PF_RULESET_TABLE:
4820                             {
4821                                 struct pfr_table table;
4822
4823                                 bzero(&table, sizeof(table));
4824                                 strlcpy(table.pfrt_anchor, ioe->anchor,
4825                                     sizeof(table.pfrt_anchor));
4826                                 if ((error = pfr_ina_commit(&table,
4827                                     ioe->ticket, NULL, NULL, 0))) {
4828                                         PF_RULES_WUNLOCK();
4829                                         free(ioes, M_TEMP);
4830                                         goto fail; /* really bad */
4831                                 }
4832                                 break;
4833                             }
4834                         default:
4835                                 if ((error = pf_commit_rules(ioe->ticket,
4836                                     ioe->rs_num, ioe->anchor))) {
4837                                         PF_RULES_WUNLOCK();
4838                                         free(ioes, M_TEMP);
4839                                         goto fail; /* really bad */
4840                                 }
4841                                 break;
4842                         }
4843                 }
4844                 PF_RULES_WUNLOCK();
4845                 free(ioes, M_TEMP);
4846                 break;
4847         }
4848
4849         case DIOCGETSRCNODES: {
4850                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
4851                 struct pf_srchash       *sh;
4852                 struct pf_ksrc_node     *n;
4853                 struct pf_src_node      *p, *pstore;
4854                 uint32_t                 i, nr = 0;
4855
4856                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4857                                 i++, sh++) {
4858                         PF_HASHROW_LOCK(sh);
4859                         LIST_FOREACH(n, &sh->nodes, entry)
4860                                 nr++;
4861                         PF_HASHROW_UNLOCK(sh);
4862                 }
4863
4864                 psn->psn_len = min(psn->psn_len,
4865                     sizeof(struct pf_src_node) * nr);
4866
4867                 if (psn->psn_len == 0) {
4868                         psn->psn_len = sizeof(struct pf_src_node) * nr;
4869                         break;
4870                 }
4871
4872                 nr = 0;
4873
4874                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4875                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4876                     i++, sh++) {
4877                     PF_HASHROW_LOCK(sh);
4878                     LIST_FOREACH(n, &sh->nodes, entry) {
4879
4880                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4881                                 break;
4882
4883                         pf_src_node_copy(n, p);
4884
4885                         p++;
4886                         nr++;
4887                     }
4888                     PF_HASHROW_UNLOCK(sh);
4889                 }
4890                 error = copyout(pstore, psn->psn_src_nodes,
4891                     sizeof(struct pf_src_node) * nr);
4892                 if (error) {
4893                         free(pstore, M_TEMP);
4894                         break;
4895                 }
4896                 psn->psn_len = sizeof(struct pf_src_node) * nr;
4897                 free(pstore, M_TEMP);
4898                 break;
4899         }
4900
4901         case DIOCCLRSRCNODES: {
4902                 pf_clear_srcnodes(NULL);
4903                 pf_purge_expired_src_nodes();
4904                 break;
4905         }
4906
4907         case DIOCKILLSRCNODES:
4908                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4909                 break;
4910
4911         case DIOCSETHOSTID: {
4912                 u_int32_t       *hostid = (u_int32_t *)addr;
4913
4914                 PF_RULES_WLOCK();
4915                 if (*hostid == 0)
4916                         V_pf_status.hostid = arc4random();
4917                 else
4918                         V_pf_status.hostid = *hostid;
4919                 PF_RULES_WUNLOCK();
4920                 break;
4921         }
4922
4923         case DIOCOSFPFLUSH:
4924                 PF_RULES_WLOCK();
4925                 pf_osfp_flush();
4926                 PF_RULES_WUNLOCK();
4927                 break;
4928
4929         case DIOCIGETIFACES: {
4930                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4931                 struct pfi_kif *ifstore;
4932                 size_t bufsiz;
4933
4934                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4935                         error = ENODEV;
4936                         break;
4937                 }
4938
4939                 if (io->pfiio_size < 0 ||
4940                     io->pfiio_size > pf_ioctl_maxcount ||
4941                     WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4942                         error = EINVAL;
4943                         break;
4944                 }
4945
4946                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4947                 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4948                     M_TEMP, M_NOWAIT);
4949                 if (ifstore == NULL) {
4950                         error = ENOMEM;
4951                         break;
4952                 }
4953
4954                 PF_RULES_RLOCK();
4955                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4956                 PF_RULES_RUNLOCK();
4957                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4958                 free(ifstore, M_TEMP);
4959                 break;
4960         }
4961
4962         case DIOCSETIFFLAG: {
4963                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4964
4965                 PF_RULES_WLOCK();
4966                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4967                 PF_RULES_WUNLOCK();
4968                 break;
4969         }
4970
4971         case DIOCCLRIFFLAG: {
4972                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
4973
4974                 PF_RULES_WLOCK();
4975                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4976                 PF_RULES_WUNLOCK();
4977                 break;
4978         }
4979
4980         default:
4981                 error = ENODEV;
4982                 break;
4983         }
4984 fail:
4985         if (sx_xlocked(&pf_ioctl_lock))
4986                 sx_xunlock(&pf_ioctl_lock);
4987         CURVNET_RESTORE();
4988
4989         return (error);
4990 }
4991
4992 void
4993 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4994 {
4995         bzero(sp, sizeof(struct pfsync_state));
4996
4997         /* copy from state key */
4998         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4999         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5000         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5001         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5002         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5003         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5004         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5005         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5006         sp->proto = st->key[PF_SK_WIRE]->proto;
5007         sp->af = st->key[PF_SK_WIRE]->af;
5008
5009         /* copy from state */
5010         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5011         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5012         sp->creation = htonl(time_uptime - st->creation);
5013         sp->expire = pf_state_expires(st);
5014         if (sp->expire <= time_uptime)
5015                 sp->expire = htonl(0);
5016         else
5017                 sp->expire = htonl(sp->expire - time_uptime);
5018
5019         sp->direction = st->direction;
5020         sp->log = st->log;
5021         sp->timeout = st->timeout;
5022         sp->state_flags = st->state_flags;
5023         if (st->src_node)
5024                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5025         if (st->nat_src_node)
5026                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5027
5028         sp->id = st->id;
5029         sp->creatorid = st->creatorid;
5030         pf_state_peer_hton(&st->src, &sp->src);
5031         pf_state_peer_hton(&st->dst, &sp->dst);
5032
5033         if (st->rule.ptr == NULL)
5034                 sp->rule = htonl(-1);
5035         else
5036                 sp->rule = htonl(st->rule.ptr->nr);
5037         if (st->anchor.ptr == NULL)
5038                 sp->anchor = htonl(-1);
5039         else
5040                 sp->anchor = htonl(st->anchor.ptr->nr);
5041         if (st->nat_rule.ptr == NULL)
5042                 sp->nat_rule = htonl(-1);
5043         else
5044                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5045
5046         pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
5047             sp->packets[0]);
5048         pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
5049             sp->packets[1]);
5050         pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
5051         pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
5052
5053 }
5054
5055 static void
5056 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5057 {
5058         struct pfr_ktable *kt;
5059
5060         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5061
5062         kt = aw->p.tbl;
5063         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5064                 kt = kt->pfrkt_root;
5065         aw->p.tbl = NULL;
5066         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5067                 kt->pfrkt_cnt : -1;
5068 }
5069
5070 /*
5071  * XXX - Check for version missmatch!!!
5072  */
5073 static void
5074 pf_clear_states(void)
5075 {
5076         struct pf_state *s;
5077         u_int i;
5078
5079         for (i = 0; i <= pf_hashmask; i++) {
5080                 struct pf_idhash *ih = &V_pf_idhash[i];
5081 relock:
5082                 PF_HASHROW_LOCK(ih);
5083                 LIST_FOREACH(s, &ih->states, entry) {
5084                         s->timeout = PFTM_PURGE;
5085                         /* Don't send out individual delete messages. */
5086                         s->state_flags |= PFSTATE_NOSYNC;
5087                         pf_unlink_state(s, PF_ENTER_LOCKED);
5088                         goto relock;
5089                 }
5090                 PF_HASHROW_UNLOCK(ih);
5091         }
5092 }
5093
5094 static int
5095 pf_clear_tables(void)
5096 {
5097         struct pfioc_table io;
5098         int error;
5099
5100         bzero(&io, sizeof(io));
5101
5102         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5103             io.pfrio_flags);
5104
5105         return (error);
5106 }
5107
5108 static void
5109 pf_clear_srcnodes(struct pf_ksrc_node *n)
5110 {
5111         struct pf_state *s;
5112         int i;
5113
5114         for (i = 0; i <= pf_hashmask; i++) {
5115                 struct pf_idhash *ih = &V_pf_idhash[i];
5116
5117                 PF_HASHROW_LOCK(ih);
5118                 LIST_FOREACH(s, &ih->states, entry) {
5119                         if (n == NULL || n == s->src_node)
5120                                 s->src_node = NULL;
5121                         if (n == NULL || n == s->nat_src_node)
5122                                 s->nat_src_node = NULL;
5123                 }
5124                 PF_HASHROW_UNLOCK(ih);
5125         }
5126
5127         if (n == NULL) {
5128                 struct pf_srchash *sh;
5129
5130                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5131                     i++, sh++) {
5132                         PF_HASHROW_LOCK(sh);
5133                         LIST_FOREACH(n, &sh->nodes, entry) {
5134                                 n->expire = 1;
5135                                 n->states = 0;
5136                         }
5137                         PF_HASHROW_UNLOCK(sh);
5138                 }
5139         } else {
5140                 /* XXX: hash slot should already be locked here. */
5141                 n->expire = 1;
5142                 n->states = 0;
5143         }
5144 }
5145
5146 static void
5147 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5148 {
5149         struct pf_ksrc_node_list         kill;
5150
5151         LIST_INIT(&kill);
5152         for (int i = 0; i <= pf_srchashmask; i++) {
5153                 struct pf_srchash *sh = &V_pf_srchash[i];
5154                 struct pf_ksrc_node *sn, *tmp;
5155
5156                 PF_HASHROW_LOCK(sh);
5157                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5158                         if (PF_MATCHA(psnk->psnk_src.neg,
5159                               &psnk->psnk_src.addr.v.a.addr,
5160                               &psnk->psnk_src.addr.v.a.mask,
5161                               &sn->addr, sn->af) &&
5162                             PF_MATCHA(psnk->psnk_dst.neg,
5163                               &psnk->psnk_dst.addr.v.a.addr,
5164                               &psnk->psnk_dst.addr.v.a.mask,
5165                               &sn->raddr, sn->af)) {
5166                                 pf_unlink_src_node(sn);
5167                                 LIST_INSERT_HEAD(&kill, sn, entry);
5168                                 sn->expire = 1;
5169                         }
5170                 PF_HASHROW_UNLOCK(sh);
5171         }
5172
5173         for (int i = 0; i <= pf_hashmask; i++) {
5174                 struct pf_idhash *ih = &V_pf_idhash[i];
5175                 struct pf_state *s;
5176
5177                 PF_HASHROW_LOCK(ih);
5178                 LIST_FOREACH(s, &ih->states, entry) {
5179                         if (s->src_node && s->src_node->expire == 1)
5180                                 s->src_node = NULL;
5181                         if (s->nat_src_node && s->nat_src_node->expire == 1)
5182                                 s->nat_src_node = NULL;
5183                 }
5184                 PF_HASHROW_UNLOCK(ih);
5185         }
5186
5187         psnk->psnk_killed = pf_free_src_nodes(&kill);
5188 }
5189
5190 /*
5191  * XXX - Check for version missmatch!!!
5192  */
5193
5194 /*
5195  * Duplicate pfctl -Fa operation to get rid of as much as we can.
5196  */
5197 static int
5198 shutdown_pf(void)
5199 {
5200         int error = 0;
5201         u_int32_t t[5];
5202         char nn = '\0';
5203
5204         do {
5205                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
5206                     != 0) {
5207                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
5208                         break;
5209                 }
5210                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
5211                     != 0) {
5212                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
5213                         break;          /* XXX: rollback? */
5214                 }
5215                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
5216                     != 0) {
5217                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
5218                         break;          /* XXX: rollback? */
5219                 }
5220                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
5221                     != 0) {
5222                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
5223                         break;          /* XXX: rollback? */
5224                 }
5225                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
5226                     != 0) {
5227                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
5228                         break;          /* XXX: rollback? */
5229                 }
5230
5231                 /* XXX: these should always succeed here */
5232                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
5233                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
5234                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
5235                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
5236                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
5237
5238                 if ((error = pf_clear_tables()) != 0)
5239                         break;
5240
5241 #ifdef ALTQ
5242                 if ((error = pf_begin_altq(&t[0])) != 0) {
5243                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
5244                         break;
5245                 }
5246                 pf_commit_altq(t[0]);
5247 #endif
5248
5249                 pf_clear_states();
5250
5251                 pf_clear_srcnodes(NULL);
5252
5253                 /* status does not use malloced mem so no need to cleanup */
5254                 /* fingerprints and interfaces have their own cleanup code */
5255         } while(0);
5256
5257         return (error);
5258 }
5259
5260 static pfil_return_t
5261 pf_check_return(int chk, struct mbuf **m)
5262 {
5263
5264         switch (chk) {
5265         case PF_PASS:
5266                 if (*m == NULL)
5267                         return (PFIL_CONSUMED);
5268                 else
5269                         return (PFIL_PASS);
5270                 break;
5271         default:
5272                 if (*m != NULL) {
5273                         m_freem(*m);
5274                         *m = NULL;
5275                 }
5276                 return (PFIL_DROPPED);
5277         }
5278 }
5279
5280 #ifdef INET
5281 static pfil_return_t
5282 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
5283     void *ruleset __unused, struct inpcb *inp)
5284 {
5285         int chk;
5286
5287         chk = pf_test(PF_IN, flags, ifp, m, inp);
5288
5289         return (pf_check_return(chk, m));
5290 }
5291
5292 static pfil_return_t
5293 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
5294     void *ruleset __unused,  struct inpcb *inp)
5295 {
5296         int chk;
5297
5298         chk = pf_test(PF_OUT, flags, ifp, m, inp);
5299
5300         return (pf_check_return(chk, m));
5301 }
5302 #endif
5303
5304 #ifdef INET6
5305 static pfil_return_t
5306 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
5307     void *ruleset __unused,  struct inpcb *inp)
5308 {
5309         int chk;
5310
5311         /*
5312          * In case of loopback traffic IPv6 uses the real interface in
5313          * order to support scoped addresses. In order to support stateful
5314          * filtering we have change this to lo0 as it is the case in IPv4.
5315          */
5316         CURVNET_SET(ifp->if_vnet);
5317         chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
5318         CURVNET_RESTORE();
5319
5320         return (pf_check_return(chk, m));
5321 }
5322
5323 static pfil_return_t
5324 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
5325     void *ruleset __unused,  struct inpcb *inp)
5326 {
5327         int chk;
5328
5329         CURVNET_SET(ifp->if_vnet);
5330         chk = pf_test6(PF_OUT, flags, ifp, m, inp);
5331         CURVNET_RESTORE();
5332
5333         return (pf_check_return(chk, m));
5334 }
5335 #endif /* INET6 */
5336
5337 #ifdef INET
5338 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
5339 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
5340 #define V_pf_ip4_in_hook        VNET(pf_ip4_in_hook)
5341 #define V_pf_ip4_out_hook       VNET(pf_ip4_out_hook)
5342 #endif
5343 #ifdef INET6
5344 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
5345 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
5346 #define V_pf_ip6_in_hook        VNET(pf_ip6_in_hook)
5347 #define V_pf_ip6_out_hook       VNET(pf_ip6_out_hook)
5348 #endif
5349
5350 static void
5351 hook_pf(void)
5352 {
5353         struct pfil_hook_args pha;
5354         struct pfil_link_args pla;
5355         int ret;
5356
5357         if (V_pf_pfil_hooked)
5358                 return;
5359
5360         pha.pa_version = PFIL_VERSION;
5361         pha.pa_modname = "pf";
5362         pha.pa_ruleset = NULL;
5363
5364         pla.pa_version = PFIL_VERSION;
5365
5366 #ifdef INET
5367         pha.pa_type = PFIL_TYPE_IP4;
5368         pha.pa_func = pf_check_in;
5369         pha.pa_flags = PFIL_IN;
5370         pha.pa_rulname = "default-in";
5371         V_pf_ip4_in_hook = pfil_add_hook(&pha);
5372         pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
5373         pla.pa_head = V_inet_pfil_head;
5374         pla.pa_hook = V_pf_ip4_in_hook;
5375         ret = pfil_link(&pla);
5376         MPASS(ret == 0);
5377         pha.pa_func = pf_check_out;
5378         pha.pa_flags = PFIL_OUT;
5379         pha.pa_rulname = "default-out";
5380         V_pf_ip4_out_hook = pfil_add_hook(&pha);
5381         pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
5382         pla.pa_head = V_inet_pfil_head;
5383         pla.pa_hook = V_pf_ip4_out_hook;
5384         ret = pfil_link(&pla);
5385         MPASS(ret == 0);
5386 #endif
5387 #ifdef INET6
5388         pha.pa_type = PFIL_TYPE_IP6;
5389         pha.pa_func = pf_check6_in;
5390         pha.pa_flags = PFIL_IN;
5391         pha.pa_rulname = "default-in6";
5392         V_pf_ip6_in_hook = pfil_add_hook(&pha);
5393         pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
5394         pla.pa_head = V_inet6_pfil_head;
5395         pla.pa_hook = V_pf_ip6_in_hook;
5396         ret = pfil_link(&pla);
5397         MPASS(ret == 0);
5398         pha.pa_func = pf_check6_out;
5399         pha.pa_rulname = "default-out6";
5400         pha.pa_flags = PFIL_OUT;
5401         V_pf_ip6_out_hook = pfil_add_hook(&pha);
5402         pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
5403         pla.pa_head = V_inet6_pfil_head;
5404         pla.pa_hook = V_pf_ip6_out_hook;
5405         ret = pfil_link(&pla);
5406         MPASS(ret == 0);
5407 #endif
5408
5409         V_pf_pfil_hooked = 1;
5410 }
5411
5412 static void
5413 dehook_pf(void)
5414 {
5415
5416         if (V_pf_pfil_hooked == 0)
5417                 return;
5418
5419 #ifdef INET
5420         pfil_remove_hook(V_pf_ip4_in_hook);
5421         pfil_remove_hook(V_pf_ip4_out_hook);
5422 #endif
5423 #ifdef INET6
5424         pfil_remove_hook(V_pf_ip6_in_hook);
5425         pfil_remove_hook(V_pf_ip6_out_hook);
5426 #endif
5427
5428         V_pf_pfil_hooked = 0;
5429 }
5430
5431 static void
5432 pf_load_vnet(void)
5433 {
5434         V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
5435             NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
5436
5437         pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
5438             PF_RULE_TAG_HASH_SIZE_DEFAULT);
5439 #ifdef ALTQ
5440         pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
5441             PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
5442 #endif
5443
5444         pfattach_vnet();
5445         V_pf_vnet_active = 1;
5446 }
5447
5448 static int
5449 pf_load(void)
5450 {
5451         int error;
5452
5453         rm_init(&pf_rules_lock, "pf rulesets");
5454         sx_init(&pf_ioctl_lock, "pf ioctl");
5455         sx_init(&pf_end_lock, "pf end thread");
5456
5457         pf_mtag_initialize();
5458
5459         pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
5460         if (pf_dev == NULL)
5461                 return (ENOMEM);
5462
5463         pf_end_threads = 0;
5464         error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
5465         if (error != 0)
5466                 return (error);
5467
5468         pfi_initialize();
5469
5470         return (0);
5471 }
5472
5473 static void
5474 pf_unload_vnet(void)
5475 {
5476         int ret;
5477
5478         V_pf_vnet_active = 0;
5479         V_pf_status.running = 0;
5480         dehook_pf();
5481
5482         PF_RULES_WLOCK();
5483         shutdown_pf();
5484         PF_RULES_WUNLOCK();
5485
5486         ret = swi_remove(V_pf_swi_cookie);
5487         MPASS(ret == 0);
5488         ret = intr_event_destroy(V_pf_swi_ie);
5489         MPASS(ret == 0);
5490
5491         pf_unload_vnet_purge();
5492
5493         pf_normalize_cleanup();
5494         PF_RULES_WLOCK();
5495         pfi_cleanup_vnet();
5496         PF_RULES_WUNLOCK();
5497         pfr_cleanup();
5498         pf_osfp_flush();
5499         pf_cleanup();
5500         if (IS_DEFAULT_VNET(curvnet))
5501                 pf_mtag_cleanup();
5502
5503         pf_cleanup_tagset(&V_pf_tags);
5504 #ifdef ALTQ
5505         pf_cleanup_tagset(&V_pf_qids);
5506 #endif
5507         uma_zdestroy(V_pf_tag_z);
5508
5509         /* Free counters last as we updated them during shutdown. */
5510         counter_u64_free(V_pf_default_rule.evaluations);
5511         for (int i = 0; i < 2; i++) {
5512                 counter_u64_free(V_pf_default_rule.packets[i]);
5513                 counter_u64_free(V_pf_default_rule.bytes[i]);
5514         }
5515         counter_u64_free(V_pf_default_rule.states_cur);
5516         counter_u64_free(V_pf_default_rule.states_tot);
5517         counter_u64_free(V_pf_default_rule.src_nodes);
5518
5519         for (int i = 0; i < PFRES_MAX; i++)
5520                 counter_u64_free(V_pf_status.counters[i]);
5521         for (int i = 0; i < LCNT_MAX; i++)
5522                 counter_u64_free(V_pf_status.lcounters[i]);
5523         for (int i = 0; i < FCNT_MAX; i++)
5524                 counter_u64_free(V_pf_status.fcounters[i]);
5525         for (int i = 0; i < SCNT_MAX; i++)
5526                 counter_u64_free(V_pf_status.scounters[i]);
5527 }
5528
5529 static void
5530 pf_unload(void)
5531 {
5532
5533         sx_xlock(&pf_end_lock);
5534         pf_end_threads = 1;
5535         while (pf_end_threads < 2) {
5536                 wakeup_one(pf_purge_thread);
5537                 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
5538         }
5539         sx_xunlock(&pf_end_lock);
5540
5541         if (pf_dev != NULL)
5542                 destroy_dev(pf_dev);
5543
5544         pfi_cleanup();
5545
5546         rm_destroy(&pf_rules_lock);
5547         sx_destroy(&pf_ioctl_lock);
5548         sx_destroy(&pf_end_lock);
5549 }
5550
5551 static void
5552 vnet_pf_init(void *unused __unused)
5553 {
5554
5555         pf_load_vnet();
5556 }
5557 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
5558     vnet_pf_init, NULL);
5559
5560 static void
5561 vnet_pf_uninit(const void *unused __unused)
5562 {
5563
5564         pf_unload_vnet();
5565 }
5566 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
5567 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
5568     vnet_pf_uninit, NULL);
5569
5570 static int
5571 pf_modevent(module_t mod, int type, void *data)
5572 {
5573         int error = 0;
5574
5575         switch(type) {
5576         case MOD_LOAD:
5577                 error = pf_load();
5578                 break;
5579         case MOD_UNLOAD:
5580                 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
5581                  * the vnet_pf_uninit()s */
5582                 break;
5583         default:
5584                 error = EINVAL;
5585                 break;
5586         }
5587
5588         return (error);
5589 }
5590
5591 static moduledata_t pf_mod = {
5592         "pf",
5593         pf_modevent,
5594         0
5595 };
5596
5597 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
5598 MODULE_VERSION(pf, PF_MODVER);