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/sdt.h>
  67 #include <sys/smp.h>
  68 #include <sys/socket.h>
  69 #include <sys/sysctl.h>
  70 #include <sys/md5.h>
  71 #include <sys/ucred.h>
  72
  73 #include <net/if.h>
  74 #include <net/if_var.h>
  75 #include <net/vnet.h>
  76 #include <net/route.h>
  77 #include <net/pfil.h>
  78 #include <net/pfvar.h>
  79 #include <net/if_pfsync.h>
  80 #include <net/if_pflog.h>
  81
  82 #include <netinet/in.h>
  83 #include <netinet/ip.h>
  84 #include <netinet/ip_var.h>
  85 #include <netinet6/ip6_var.h>
  86 #include <netinet/ip_icmp.h>
  87 #include <netpfil/pf/pf_nv.h>
  88
  89 #ifdef INET6
  90 #include <netinet/ip6.h>
  91 #endif /* INET6 */
  92
  93 #ifdef ALTQ
  94 #include <net/altq/altq.h>
  95 #endif
  96
  97 SDT_PROVIDER_DECLARE(pf);
  98 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
  99 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
 100 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
 101 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
 102
 103 static struct pf_kpool  *pf_get_kpool(char *, u_int32_t, u_int8_t, u_int32_t,
 104                             u_int8_t, u_int8_t, u_int8_t);
 105
 106 static void              pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
 107 static void              pf_empty_kpool(struct pf_kpalist *);
 108 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
 109                             struct thread *);
 110 #ifdef ALTQ
 111 static int               pf_begin_altq(u_int32_t *);
 112 static int               pf_rollback_altq(u_int32_t);
 113 static int               pf_commit_altq(u_int32_t);
 114 static int               pf_enable_altq(struct pf_altq *);
 115 static int               pf_disable_altq(struct pf_altq *);
 116 static u_int32_t         pf_qname2qid(char *);
 117 static void              pf_qid_unref(u_int32_t);
 118 #endif /* ALTQ */
 119 static int               pf_begin_rules(u_int32_t *, int, const char *);
 120 static int               pf_rollback_rules(u_int32_t, int, char *);
 121 static int               pf_setup_pfsync_matching(struct pf_kruleset *);
 122 static void              pf_hash_rule(MD5_CTX *, struct pf_krule *);
 123 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 124 static int               pf_commit_rules(u_int32_t, int, char *);
 125 static int               pf_addr_setup(struct pf_kruleset *,
 126                             struct pf_addr_wrap *, sa_family_t);
 127 static void              pf_addr_copyout(struct pf_addr_wrap *);
 128 static void              pf_src_node_copy(const struct pf_ksrc_node *,
 129                             struct pf_src_node *);
 130 #ifdef ALTQ
 131 static int               pf_export_kaltq(struct pf_altq *,
 132                             struct pfioc_altq_v1 *, size_t);
 133 static int               pf_import_kaltq(struct pfioc_altq_v1 *,
 134                             struct pf_altq *, size_t);
 135 #endif /* ALTQ */
 136
 137 VNET_DEFINE(struct pf_krule,    pf_default_rule);
 138
 139 #ifdef ALTQ
 140 VNET_DEFINE_STATIC(int,         pf_altq_running);
 141 #define V_pf_altq_running       VNET(pf_altq_running)
 142 #endif
 143
 144 #define TAGID_MAX        50000
 145 struct pf_tagname {
 146         TAILQ_ENTRY(pf_tagname) namehash_entries;
 147         TAILQ_ENTRY(pf_tagname) taghash_entries;
 148         char                    name[PF_TAG_NAME_SIZE];
 149         uint16_t                tag;
 150         int                     ref;
 151 };
 152
 153 struct pf_tagset {
 154         TAILQ_HEAD(, pf_tagname)        *namehash;
 155         TAILQ_HEAD(, pf_tagname)        *taghash;
 156         unsigned int                     mask;
 157         uint32_t                         seed;
 158         BITSET_DEFINE(, TAGID_MAX)       avail;
 159 };
 160
 161 VNET_DEFINE(struct pf_tagset, pf_tags);
 162 #define V_pf_tags       VNET(pf_tags)
 163 static unsigned int     pf_rule_tag_hashsize;
 164 #define PF_RULE_TAG_HASH_SIZE_DEFAULT   128
 165 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
 166     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
 167     "Size of pf(4) rule tag hashtable");
 168
 169 #ifdef ALTQ
 170 VNET_DEFINE(struct pf_tagset, pf_qids);
 171 #define V_pf_qids       VNET(pf_qids)
 172 static unsigned int     pf_queue_tag_hashsize;
 173 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT  128
 174 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
 175     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
 176     "Size of pf(4) queue tag hashtable");
 177 #endif
 178 VNET_DEFINE(uma_zone_t,  pf_tag_z);
 179 #define V_pf_tag_z               VNET(pf_tag_z)
 180 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 181 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 182
 183 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 184 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 185 #endif
 186
 187 static void              pf_init_tagset(struct pf_tagset *, unsigned int *,
 188                             unsigned int);
 189 static void              pf_cleanup_tagset(struct pf_tagset *);
 190 static uint16_t          tagname2hashindex(const struct pf_tagset *, const char *);
 191 static uint16_t          tag2hashindex(const struct pf_tagset *, uint16_t);
 192 static u_int16_t         tagname2tag(struct pf_tagset *, char *);
 193 static u_int16_t         pf_tagname2tag(char *);
 194 static void              tag_unref(struct pf_tagset *, u_int16_t);
 195
 196 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 197
 198 struct cdev *pf_dev;
 199
 200 /*
 201  * XXX - These are new and need to be checked when moveing to a new version
 202  */
 203 static void              pf_clear_all_states(void);
 204 static unsigned int      pf_clear_states(const struct pf_kstate_kill *);
 205 static int               pf_killstates(struct pf_kstate_kill *,
 206                             unsigned int *);
 207 static int               pf_killstates_row(struct pf_kstate_kill *,
 208                             struct pf_idhash *);
 209 static int               pf_killstates_nv(struct pfioc_nv *);
 210 static int               pf_clearstates_nv(struct pfioc_nv *);
 211 static int               pf_getstate(struct pfioc_nv *);
 212 static int               pf_clear_tables(void);
 213 static void              pf_clear_srcnodes(struct pf_ksrc_node *);
 214 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 215 static int               pf_keepcounters(struct pfioc_nv *);
 216 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 217
 218 /*
 219  * Wrapper functions for pfil(9) hooks
 220  */
 221 #ifdef INET
 222 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
 223     int flags, void *ruleset __unused, struct inpcb *inp);
 224 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
 225     int flags, void *ruleset __unused, struct inpcb *inp);
 226 #endif
 227 #ifdef INET6
 228 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
 229     int flags, void *ruleset __unused, struct inpcb *inp);
 230 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
 231     int flags, void *ruleset __unused, struct inpcb *inp);
 232 #endif
 233
 234 static void             hook_pf(void);
 235 static void             dehook_pf(void);
 236 static int              shutdown_pf(void);
 237 static int              pf_load(void);
 238 static void             pf_unload(void);
 239
 240 static struct cdevsw pf_cdevsw = {
 241         .d_ioctl =      pfioctl,
 242         .d_name =       PF_NAME,
 243         .d_version =    D_VERSION,
 244 };
 245
 246 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
 247 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 248
 249 /*
 250  * We need a flag that is neither hooked nor running to know when
 251  * the VNET is "valid".  We primarily need this to control (global)
 252  * external event, e.g., eventhandlers.
 253  */
 254 VNET_DEFINE(int, pf_vnet_active);
 255 #define V_pf_vnet_active        VNET(pf_vnet_active)
 256
 257 int pf_end_threads;
 258 struct proc *pf_purge_proc;
 259
 260 struct rmlock                   pf_rules_lock;
 261 struct sx                       pf_ioctl_lock;
 262 struct sx                       pf_end_lock;
 263
 264 /* pfsync */
 265 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
 266 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
 267 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
 268 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
 269 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
 270 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
 271 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
 272
 273 /* pflog */
 274 pflog_packet_t                  *pflog_packet_ptr = NULL;
 275
 276 extern u_long   pf_ioctl_maxcount;
 277
 278 #define ERROUT_FUNCTION(target, x)                                      \
 279         do {                                                            \
 280                 error = (x);                                            \
 281                 SDT_PROBE3(pf, ioctl, function, error, __func__, error, \
 282                     __LINE__);                                          \
 283                 goto target;                                            \
 284         } while (0)
 285
 286 static void
 287 pfattach_vnet(void)
 288 {
 289         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 290
 291         pf_initialize();
 292         pfr_initialize();
 293         pfi_initialize_vnet();
 294         pf_normalize_init();
 295
 296         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 297         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 298
 299         RB_INIT(&V_pf_anchors);
 300         pf_init_kruleset(&pf_main_ruleset);
 301
 302         /* default rule should never be garbage collected */
 303         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 304 #ifdef PF_DEFAULT_TO_DROP
 305         V_pf_default_rule.action = PF_DROP;
 306 #else
 307         V_pf_default_rule.action = PF_PASS;
 308 #endif
 309         V_pf_default_rule.nr = -1;
 310         V_pf_default_rule.rtableid = -1;
 311
 312         V_pf_default_rule.evaluations = counter_u64_alloc(M_WAITOK);
 313         for (int i = 0; i < 2; i++) {
 314                 V_pf_default_rule.packets[i] = counter_u64_alloc(M_WAITOK);
 315                 V_pf_default_rule.bytes[i] = counter_u64_alloc(M_WAITOK);
 316         }
 317         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 318         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 319         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 320
 321         /* initialize default timeouts */
 322         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 323         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 324         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 325         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 326         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 327         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 328         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 329         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 330         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 331         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 332         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 333         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 334         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 335         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 336         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 337         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 338         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 339         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 340         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 341         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 342
 343         bzero(&V_pf_status, sizeof(V_pf_status));
 344         V_pf_status.debug = PF_DEBUG_URGENT;
 345
 346         V_pf_pfil_hooked = 0;
 347
 348         /* XXX do our best to avoid a conflict */
 349         V_pf_status.hostid = arc4random();
 350
 351         for (int i = 0; i < PFRES_MAX; i++)
 352                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 353         for (int i = 0; i < LCNT_MAX; i++)
 354                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 355         for (int i = 0; i < FCNT_MAX; i++)
 356                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 357         for (int i = 0; i < SCNT_MAX; i++)
 358                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 359
 360         if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
 361             INTR_MPSAFE, &V_pf_swi_cookie) != 0)
 362                 /* XXXGL: leaked all above. */
 363                 return;
 364 }
 365
 366 static struct pf_kpool *
 367 pf_get_kpool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 368     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 369     u_int8_t check_ticket)
 370 {
 371         struct pf_kruleset      *ruleset;
 372         struct pf_krule         *rule;
 373         int                      rs_num;
 374
 375         ruleset = pf_find_kruleset(anchor);
 376         if (ruleset == NULL)
 377                 return (NULL);
 378         rs_num = pf_get_ruleset_number(rule_action);
 379         if (rs_num >= PF_RULESET_MAX)
 380                 return (NULL);
 381         if (active) {
 382                 if (check_ticket && ticket !=
 383                     ruleset->rules[rs_num].active.ticket)
 384                         return (NULL);
 385                 if (r_last)
 386                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 387                             pf_krulequeue);
 388                 else
 389                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 390         } else {
 391                 if (check_ticket && ticket !=
 392                     ruleset->rules[rs_num].inactive.ticket)
 393                         return (NULL);
 394                 if (r_last)
 395                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 396                             pf_krulequeue);
 397                 else
 398                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 399         }
 400         if (!r_last) {
 401                 while ((rule != NULL) && (rule->nr != rule_number))
 402                         rule = TAILQ_NEXT(rule, entries);
 403         }
 404         if (rule == NULL)
 405                 return (NULL);
 406
 407         return (&rule->rpool);
 408 }
 409
 410 static void
 411 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
 412 {
 413         struct pf_kpooladdr     *mv_pool_pa;
 414
 415         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 416                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 417                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 418         }
 419 }
 420
 421 static void
 422 pf_empty_kpool(struct pf_kpalist *poola)
 423 {
 424         struct pf_kpooladdr *pa;
 425
 426         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 427                 switch (pa->addr.type) {
 428                 case PF_ADDR_DYNIFTL:
 429                         pfi_dynaddr_remove(pa->addr.p.dyn);
 430                         break;
 431                 case PF_ADDR_TABLE:
 432                         /* XXX: this could be unfinished pooladdr on pabuf */
 433                         if (pa->addr.p.tbl != NULL)
 434                                 pfr_detach_table(pa->addr.p.tbl);
 435                         break;
 436                 }
 437                 if (pa->kif)
 438                         pfi_kkif_unref(pa->kif);
 439                 TAILQ_REMOVE(poola, pa, entries);
 440                 free(pa, M_PFRULE);
 441         }
 442 }
 443
 444 static void
 445 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
 446 {
 447
 448         PF_RULES_WASSERT();
 449
 450         TAILQ_REMOVE(rulequeue, rule, entries);
 451
 452         PF_UNLNKDRULES_LOCK();
 453         rule->rule_ref |= PFRULE_REFS;
 454         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 455         PF_UNLNKDRULES_UNLOCK();
 456 }
 457
 458 void
 459 pf_free_rule(struct pf_krule *rule)
 460 {
 461
 462         PF_RULES_WASSERT();
 463
 464         if (rule->tag)
 465                 tag_unref(&V_pf_tags, rule->tag);
 466         if (rule->match_tag)
 467                 tag_unref(&V_pf_tags, rule->match_tag);
 468 #ifdef ALTQ
 469         if (rule->pqid != rule->qid)
 470                 pf_qid_unref(rule->pqid);
 471         pf_qid_unref(rule->qid);
 472 #endif
 473         switch (rule->src.addr.type) {
 474         case PF_ADDR_DYNIFTL:
 475                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 476                 break;
 477         case PF_ADDR_TABLE:
 478                 pfr_detach_table(rule->src.addr.p.tbl);
 479                 break;
 480         }
 481         switch (rule->dst.addr.type) {
 482         case PF_ADDR_DYNIFTL:
 483                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 484                 break;
 485         case PF_ADDR_TABLE:
 486                 pfr_detach_table(rule->dst.addr.p.tbl);
 487                 break;
 488         }
 489         if (rule->overload_tbl)
 490                 pfr_detach_table(rule->overload_tbl);
 491         if (rule->kif)
 492                 pfi_kkif_unref(rule->kif);
 493         pf_kanchor_remove(rule);
 494         pf_empty_kpool(&rule->rpool.list);
 495
 496         pf_krule_free(rule);
 497 }
 498
 499 static void
 500 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
 501     unsigned int default_size)
 502 {
 503         unsigned int i;
 504         unsigned int hashsize;
 505
 506         if (*tunable_size == 0 || !powerof2(*tunable_size))
 507                 *tunable_size = default_size;
 508
 509         hashsize = *tunable_size;
 510         ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
 511             M_WAITOK);
 512         ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
 513             M_WAITOK);
 514         ts->mask = hashsize - 1;
 515         ts->seed = arc4random();
 516         for (i = 0; i < hashsize; i++) {
 517                 TAILQ_INIT(&ts->namehash[i]);
 518                 TAILQ_INIT(&ts->taghash[i]);
 519         }
 520         BIT_FILL(TAGID_MAX, &ts->avail);
 521 }
 522
 523 static void
 524 pf_cleanup_tagset(struct pf_tagset *ts)
 525 {
 526         unsigned int i;
 527         unsigned int hashsize;
 528         struct pf_tagname *t, *tmp;
 529
 530         /*
 531          * Only need to clean up one of the hashes as each tag is hashed
 532          * into each table.
 533          */
 534         hashsize = ts->mask + 1;
 535         for (i = 0; i < hashsize; i++)
 536                 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
 537                         uma_zfree(V_pf_tag_z, t);
 538
 539         free(ts->namehash, M_PFHASH);
 540         free(ts->taghash, M_PFHASH);
 541 }
 542
 543 static uint16_t
 544 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
 545 {
 546         size_t len;
 547
 548         len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
 549         return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
 550 }
 551
 552 static uint16_t
 553 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
 554 {
 555
 556         return (tag & ts->mask);
 557 }
 558
 559 static u_int16_t
 560 tagname2tag(struct pf_tagset *ts, char *tagname)
 561 {
 562         struct pf_tagname       *tag;
 563         u_int32_t                index;
 564         u_int16_t                new_tagid;
 565
 566         PF_RULES_WASSERT();
 567
 568         index = tagname2hashindex(ts, tagname);
 569         TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
 570                 if (strcmp(tagname, tag->name) == 0) {
 571                         tag->ref++;
 572                         return (tag->tag);
 573                 }
 574
 575         /*
 576          * new entry
 577          *
 578          * to avoid fragmentation, we do a linear search from the beginning
 579          * and take the first free slot we find.
 580          */
 581         new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
 582         /*
 583          * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
 584          * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
 585          * set.  It may also return a bit number greater than TAGID_MAX due
 586          * to rounding of the number of bits in the vector up to a multiple
 587          * of the vector word size at declaration/allocation time.
 588          */
 589         if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
 590                 return (0);
 591
 592         /* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
 593         BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
 594
 595         /* allocate and fill new struct pf_tagname */
 596         tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
 597         if (tag == NULL)
 598                 return (0);
 599         strlcpy(tag->name, tagname, sizeof(tag->name));
 600         tag->tag = new_tagid;
 601         tag->ref = 1;
 602
 603         /* Insert into namehash */
 604         TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
 605
 606         /* Insert into taghash */
 607         index = tag2hashindex(ts, new_tagid);
 608         TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
 609
 610         return (tag->tag);
 611 }
 612
 613 static void
 614 tag_unref(struct pf_tagset *ts, u_int16_t tag)
 615 {
 616         struct pf_tagname       *t;
 617         uint16_t                 index;
 618
 619         PF_RULES_WASSERT();
 620
 621         index = tag2hashindex(ts, tag);
 622         TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
 623                 if (tag == t->tag) {
 624                         if (--t->ref == 0) {
 625                                 TAILQ_REMOVE(&ts->taghash[index], t,
 626                                     taghash_entries);
 627                                 index = tagname2hashindex(ts, t->name);
 628                                 TAILQ_REMOVE(&ts->namehash[index], t,
 629                                     namehash_entries);
 630                                 /* Bits are 0-based for BIT_SET() */
 631                                 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
 632                                 uma_zfree(V_pf_tag_z, t);
 633                         }
 634                         break;
 635                 }
 636 }
 637
 638 static u_int16_t
 639 pf_tagname2tag(char *tagname)
 640 {
 641         return (tagname2tag(&V_pf_tags, tagname));
 642 }
 643
 644 #ifdef ALTQ
 645 static u_int32_t
 646 pf_qname2qid(char *qname)
 647 {
 648         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 649 }
 650
 651 static void
 652 pf_qid_unref(u_int32_t qid)
 653 {
 654         tag_unref(&V_pf_qids, (u_int16_t)qid);
 655 }
 656
 657 static int
 658 pf_begin_altq(u_int32_t *ticket)
 659 {
 660         struct pf_altq  *altq, *tmp;
 661         int              error = 0;
 662
 663         PF_RULES_WASSERT();
 664
 665         /* Purge the old altq lists */
 666         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 667                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 668                         /* detach and destroy the discipline */
 669                         error = altq_remove(altq);
 670                 }
 671                 free(altq, M_PFALTQ);
 672         }
 673         TAILQ_INIT(V_pf_altq_ifs_inactive);
 674         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 675                 pf_qid_unref(altq->qid);
 676                 free(altq, M_PFALTQ);
 677         }
 678         TAILQ_INIT(V_pf_altqs_inactive);
 679         if (error)
 680                 return (error);
 681         *ticket = ++V_ticket_altqs_inactive;
 682         V_altqs_inactive_open = 1;
 683         return (0);
 684 }
 685
 686 static int
 687 pf_rollback_altq(u_int32_t ticket)
 688 {
 689         struct pf_altq  *altq, *tmp;
 690         int              error = 0;
 691
 692         PF_RULES_WASSERT();
 693
 694         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 695                 return (0);
 696         /* Purge the old altq lists */
 697         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 698                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 699                         /* detach and destroy the discipline */
 700                         error = altq_remove(altq);
 701                 }
 702                 free(altq, M_PFALTQ);
 703         }
 704         TAILQ_INIT(V_pf_altq_ifs_inactive);
 705         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 706                 pf_qid_unref(altq->qid);
 707                 free(altq, M_PFALTQ);
 708         }
 709         TAILQ_INIT(V_pf_altqs_inactive);
 710         V_altqs_inactive_open = 0;
 711         return (error);
 712 }
 713
 714 static int
 715 pf_commit_altq(u_int32_t ticket)
 716 {
 717         struct pf_altqqueue     *old_altqs, *old_altq_ifs;
 718         struct pf_altq          *altq, *tmp;
 719         int                      err, error = 0;
 720
 721         PF_RULES_WASSERT();
 722
 723         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 724                 return (EBUSY);
 725
 726         /* swap altqs, keep the old. */
 727         old_altqs = V_pf_altqs_active;
 728         old_altq_ifs = V_pf_altq_ifs_active;
 729         V_pf_altqs_active = V_pf_altqs_inactive;
 730         V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
 731         V_pf_altqs_inactive = old_altqs;
 732         V_pf_altq_ifs_inactive = old_altq_ifs;
 733         V_ticket_altqs_active = V_ticket_altqs_inactive;
 734
 735         /* Attach new disciplines */
 736         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
 737                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 738                         /* attach the discipline */
 739                         error = altq_pfattach(altq);
 740                         if (error == 0 && V_pf_altq_running)
 741                                 error = pf_enable_altq(altq);
 742                         if (error != 0)
 743                                 return (error);
 744                 }
 745         }
 746
 747         /* Purge the old altq lists */
 748         TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
 749                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 750                         /* detach and destroy the discipline */
 751                         if (V_pf_altq_running)
 752                                 error = pf_disable_altq(altq);
 753                         err = altq_pfdetach(altq);
 754                         if (err != 0 && error == 0)
 755                                 error = err;
 756                         err = altq_remove(altq);
 757                         if (err != 0 && error == 0)
 758                                 error = err;
 759                 }
 760                 free(altq, M_PFALTQ);
 761         }
 762         TAILQ_INIT(V_pf_altq_ifs_inactive);
 763         TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
 764                 pf_qid_unref(altq->qid);
 765                 free(altq, M_PFALTQ);
 766         }
 767         TAILQ_INIT(V_pf_altqs_inactive);
 768
 769         V_altqs_inactive_open = 0;
 770         return (error);
 771 }
 772
 773 static int
 774 pf_enable_altq(struct pf_altq *altq)
 775 {
 776         struct ifnet            *ifp;
 777         struct tb_profile        tb;
 778         int                      error = 0;
 779
 780         if ((ifp = ifunit(altq->ifname)) == NULL)
 781                 return (EINVAL);
 782
 783         if (ifp->if_snd.altq_type != ALTQT_NONE)
 784                 error = altq_enable(&ifp->if_snd);
 785
 786         /* set tokenbucket regulator */
 787         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 788                 tb.rate = altq->ifbandwidth;
 789                 tb.depth = altq->tbrsize;
 790                 error = tbr_set(&ifp->if_snd, &tb);
 791         }
 792
 793         return (error);
 794 }
 795
 796 static int
 797 pf_disable_altq(struct pf_altq *altq)
 798 {
 799         struct ifnet            *ifp;
 800         struct tb_profile        tb;
 801         int                      error;
 802
 803         if ((ifp = ifunit(altq->ifname)) == NULL)
 804                 return (EINVAL);
 805
 806         /*
 807          * when the discipline is no longer referenced, it was overridden
 808          * by a new one.  if so, just return.
 809          */
 810         if (altq->altq_disc != ifp->if_snd.altq_disc)
 811                 return (0);
 812
 813         error = altq_disable(&ifp->if_snd);
 814
 815         if (error == 0) {
 816                 /* clear tokenbucket regulator */
 817                 tb.rate = 0;
 818                 error = tbr_set(&ifp->if_snd, &tb);
 819         }
 820
 821         return (error);
 822 }
 823
 824 static int
 825 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
 826     struct pf_altq *altq)
 827 {
 828         struct ifnet    *ifp1;
 829         int              error = 0;
 830
 831         /* Deactivate the interface in question */
 832         altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 833         if ((ifp1 = ifunit(altq->ifname)) == NULL ||
 834             (remove && ifp1 == ifp)) {
 835                 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 836         } else {
 837                 error = altq_add(ifp1, altq);
 838
 839                 if (ticket != V_ticket_altqs_inactive)
 840                         error = EBUSY;
 841
 842                 if (error)
 843                         free(altq, M_PFALTQ);
 844         }
 845
 846         return (error);
 847 }
 848
 849 void
 850 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 851 {
 852         struct pf_altq  *a1, *a2, *a3;
 853         u_int32_t        ticket;
 854         int              error = 0;
 855
 856         /*
 857          * No need to re-evaluate the configuration for events on interfaces
 858          * that do not support ALTQ, as it's not possible for such
 859          * interfaces to be part of the configuration.
 860          */
 861         if (!ALTQ_IS_READY(&ifp->if_snd))
 862                 return;
 863
 864         /* Interrupt userland queue modifications */
 865         if (V_altqs_inactive_open)
 866                 pf_rollback_altq(V_ticket_altqs_inactive);
 867
 868         /* Start new altq ruleset */
 869         if (pf_begin_altq(&ticket))
 870                 return;
 871
 872         /* Copy the current active set */
 873         TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
 874                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 875                 if (a2 == NULL) {
 876                         error = ENOMEM;
 877                         break;
 878                 }
 879                 bcopy(a1, a2, sizeof(struct pf_altq));
 880
 881                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 882                 if (error)
 883                         break;
 884
 885                 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
 886         }
 887         if (error)
 888                 goto out;
 889         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 890                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 891                 if (a2 == NULL) {
 892                         error = ENOMEM;
 893                         break;
 894                 }
 895                 bcopy(a1, a2, sizeof(struct pf_altq));
 896
 897                 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 898                         error = EBUSY;
 899                         free(a2, M_PFALTQ);
 900                         break;
 901                 }
 902                 a2->altq_disc = NULL;
 903                 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
 904                         if (strncmp(a3->ifname, a2->ifname,
 905                                 IFNAMSIZ) == 0) {
 906                                 a2->altq_disc = a3->altq_disc;
 907                                 break;
 908                         }
 909                 }
 910                 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
 911                 if (error)
 912                         break;
 913
 914                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 915         }
 916
 917 out:
 918         if (error != 0)
 919                 pf_rollback_altq(ticket);
 920         else
 921                 pf_commit_altq(ticket);
 922 }
 923 #endif /* ALTQ */
 924
 925 static int
 926 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 927 {
 928         struct pf_kruleset      *rs;
 929         struct pf_krule         *rule;
 930
 931         PF_RULES_WASSERT();
 932
 933         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 934                 return (EINVAL);
 935         rs = pf_find_or_create_kruleset(anchor);
 936         if (rs == NULL)
 937                 return (EINVAL);
 938         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 939                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 940                 rs->rules[rs_num].inactive.rcount--;
 941         }
 942         *ticket = ++rs->rules[rs_num].inactive.ticket;
 943         rs->rules[rs_num].inactive.open = 1;
 944         return (0);
 945 }
 946
 947 static int
 948 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 949 {
 950         struct pf_kruleset      *rs;
 951         struct pf_krule         *rule;
 952
 953         PF_RULES_WASSERT();
 954
 955         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 956                 return (EINVAL);
 957         rs = pf_find_kruleset(anchor);
 958         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 959             rs->rules[rs_num].inactive.ticket != ticket)
 960                 return (0);
 961         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 962                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 963                 rs->rules[rs_num].inactive.rcount--;
 964         }
 965         rs->rules[rs_num].inactive.open = 0;
 966         return (0);
 967 }
 968
 969 #define PF_MD5_UPD(st, elm)                                             \
 970                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 971
 972 #define PF_MD5_UPD_STR(st, elm)                                         \
 973                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 974
 975 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 976                 (stor) = htonl((st)->elm);                              \
 977                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 978 } while (0)
 979
 980 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 981                 (stor) = htons((st)->elm);                              \
 982                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 983 } while (0)
 984
 985 static void
 986 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 987 {
 988         PF_MD5_UPD(pfr, addr.type);
 989         switch (pfr->addr.type) {
 990                 case PF_ADDR_DYNIFTL:
 991                         PF_MD5_UPD(pfr, addr.v.ifname);
 992                         PF_MD5_UPD(pfr, addr.iflags);
 993                         break;
 994                 case PF_ADDR_TABLE:
 995                         PF_MD5_UPD(pfr, addr.v.tblname);
 996                         break;
 997                 case PF_ADDR_ADDRMASK:
 998                         /* XXX ignore af? */
 999                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1000                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1001                         break;
1002         }
1003
1004         PF_MD5_UPD(pfr, port[0]);
1005         PF_MD5_UPD(pfr, port[1]);
1006         PF_MD5_UPD(pfr, neg);
1007         PF_MD5_UPD(pfr, port_op);
1008 }
1009
1010 static void
1011 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
1012 {
1013         u_int16_t x;
1014         u_int32_t y;
1015
1016         pf_hash_rule_addr(ctx, &rule->src);
1017         pf_hash_rule_addr(ctx, &rule->dst);
1018         for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1019                 PF_MD5_UPD_STR(rule, label[i]);
1020         PF_MD5_UPD_STR(rule, ifname);
1021         PF_MD5_UPD_STR(rule, match_tagname);
1022         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1023         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1024         PF_MD5_UPD_HTONL(rule, prob, y);
1025         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1026         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1027         PF_MD5_UPD(rule, uid.op);
1028         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1029         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1030         PF_MD5_UPD(rule, gid.op);
1031         PF_MD5_UPD_HTONL(rule, rule_flag, y);
1032         PF_MD5_UPD(rule, action);
1033         PF_MD5_UPD(rule, direction);
1034         PF_MD5_UPD(rule, af);
1035         PF_MD5_UPD(rule, quick);
1036         PF_MD5_UPD(rule, ifnot);
1037         PF_MD5_UPD(rule, match_tag_not);
1038         PF_MD5_UPD(rule, natpass);
1039         PF_MD5_UPD(rule, keep_state);
1040         PF_MD5_UPD(rule, proto);
1041         PF_MD5_UPD(rule, type);
1042         PF_MD5_UPD(rule, code);
1043         PF_MD5_UPD(rule, flags);
1044         PF_MD5_UPD(rule, flagset);
1045         PF_MD5_UPD(rule, allow_opts);
1046         PF_MD5_UPD(rule, rt);
1047         PF_MD5_UPD(rule, tos);
1048 }
1049
1050 static bool
1051 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1052 {
1053         MD5_CTX         ctx[2];
1054         u_int8_t        digest[2][PF_MD5_DIGEST_LENGTH];
1055
1056         MD5Init(&ctx[0]);
1057         MD5Init(&ctx[1]);
1058         pf_hash_rule(&ctx[0], a);
1059         pf_hash_rule(&ctx[1], b);
1060         MD5Final(digest[0], &ctx[0]);
1061         MD5Final(digest[1], &ctx[1]);
1062
1063         return (memcmp(digest[0], digest[1], PF_MD5_DIGEST_LENGTH) == 0);
1064 }
1065
1066 static int
1067 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1068 {
1069         struct pf_kruleset      *rs;
1070         struct pf_krule         *rule, **old_array, *tail;
1071         struct pf_krulequeue    *old_rules;
1072         int                      error;
1073         u_int32_t                old_rcount;
1074
1075         PF_RULES_WASSERT();
1076
1077         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1078                 return (EINVAL);
1079         rs = pf_find_kruleset(anchor);
1080         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1081             ticket != rs->rules[rs_num].inactive.ticket)
1082                 return (EBUSY);
1083
1084         /* Calculate checksum for the main ruleset */
1085         if (rs == &pf_main_ruleset) {
1086                 error = pf_setup_pfsync_matching(rs);
1087                 if (error != 0)
1088                         return (error);
1089         }
1090
1091         /* Swap rules, keep the old. */
1092         old_rules = rs->rules[rs_num].active.ptr;
1093         old_rcount = rs->rules[rs_num].active.rcount;
1094         old_array = rs->rules[rs_num].active.ptr_array;
1095
1096         rs->rules[rs_num].active.ptr =
1097             rs->rules[rs_num].inactive.ptr;
1098         rs->rules[rs_num].active.ptr_array =
1099             rs->rules[rs_num].inactive.ptr_array;
1100         rs->rules[rs_num].active.rcount =
1101             rs->rules[rs_num].inactive.rcount;
1102
1103         /* Attempt to preserve counter information. */
1104         if (V_pf_status.keep_counters) {
1105                 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1106                     entries) {
1107                         tail = TAILQ_FIRST(old_rules);
1108                         while ((tail != NULL) && ! pf_krule_compare(tail, rule))
1109                                 tail = TAILQ_NEXT(tail, entries);
1110                         if (tail != NULL) {
1111                                 counter_u64_add(rule->evaluations,
1112                                     counter_u64_fetch(tail->evaluations));
1113                                 counter_u64_add(rule->packets[0],
1114                                     counter_u64_fetch(tail->packets[0]));
1115                                 counter_u64_add(rule->packets[1],
1116                                     counter_u64_fetch(tail->packets[1]));
1117                                 counter_u64_add(rule->bytes[0],
1118                                     counter_u64_fetch(tail->bytes[0]));
1119                                 counter_u64_add(rule->bytes[1],
1120                                     counter_u64_fetch(tail->bytes[1]));
1121                         }
1122                 }
1123         }
1124
1125         rs->rules[rs_num].inactive.ptr = old_rules;
1126         rs->rules[rs_num].inactive.ptr_array = old_array;
1127         rs->rules[rs_num].inactive.rcount = old_rcount;
1128
1129         rs->rules[rs_num].active.ticket =
1130             rs->rules[rs_num].inactive.ticket;
1131         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1132
1133         /* Purge the old rule list. */
1134         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1135                 pf_unlink_rule(old_rules, rule);
1136         if (rs->rules[rs_num].inactive.ptr_array)
1137                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1138         rs->rules[rs_num].inactive.ptr_array = NULL;
1139         rs->rules[rs_num].inactive.rcount = 0;
1140         rs->rules[rs_num].inactive.open = 0;
1141         pf_remove_if_empty_kruleset(rs);
1142
1143         return (0);
1144 }
1145
1146 static int
1147 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1148 {
1149         MD5_CTX                  ctx;
1150         struct pf_krule         *rule;
1151         int                      rs_cnt;
1152         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
1153
1154         MD5Init(&ctx);
1155         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1156                 /* XXX PF_RULESET_SCRUB as well? */
1157                 if (rs_cnt == PF_RULESET_SCRUB)
1158                         continue;
1159
1160                 if (rs->rules[rs_cnt].inactive.ptr_array)
1161                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1162                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1163
1164                 if (rs->rules[rs_cnt].inactive.rcount) {
1165                         rs->rules[rs_cnt].inactive.ptr_array =
1166                             malloc(sizeof(caddr_t) *
1167                             rs->rules[rs_cnt].inactive.rcount,
1168                             M_TEMP, M_NOWAIT);
1169
1170                         if (!rs->rules[rs_cnt].inactive.ptr_array)
1171                                 return (ENOMEM);
1172                 }
1173
1174                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1175                     entries) {
1176                         pf_hash_rule(&ctx, rule);
1177                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1178                 }
1179         }
1180
1181         MD5Final(digest, &ctx);
1182         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1183         return (0);
1184 }
1185
1186 static int
1187 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1188     sa_family_t af)
1189 {
1190         int error = 0;
1191
1192         switch (addr->type) {
1193         case PF_ADDR_TABLE:
1194                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1195                 if (addr->p.tbl == NULL)
1196                         error = ENOMEM;
1197                 break;
1198         case PF_ADDR_DYNIFTL:
1199                 error = pfi_dynaddr_setup(addr, af);
1200                 break;
1201         }
1202
1203         return (error);
1204 }
1205
1206 static void
1207 pf_addr_copyout(struct pf_addr_wrap *addr)
1208 {
1209
1210         switch (addr->type) {
1211         case PF_ADDR_DYNIFTL:
1212                 pfi_dynaddr_copyout(addr);
1213                 break;
1214         case PF_ADDR_TABLE:
1215                 pf_tbladdr_copyout(addr);
1216                 break;
1217         }
1218 }
1219
1220 static void
1221 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1222 {
1223         int     secs = time_uptime, diff;
1224
1225         bzero(out, sizeof(struct pf_src_node));
1226
1227         bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1228         bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1229
1230         if (in->rule.ptr != NULL)
1231                 out->rule.nr = in->rule.ptr->nr;
1232
1233         for (int i = 0; i < 2; i++) {
1234                 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1235                 out->packets[i] = counter_u64_fetch(in->packets[i]);
1236         }
1237
1238         out->states = in->states;
1239         out->conn = in->conn;
1240         out->af = in->af;
1241         out->ruletype = in->ruletype;
1242
1243         out->creation = secs - in->creation;
1244         if (out->expire > secs)
1245                 out->expire -= secs;
1246         else
1247                 out->expire = 0;
1248
1249         /* Adjust the connection rate estimate. */
1250         diff = secs - in->conn_rate.last;
1251         if (diff >= in->conn_rate.seconds)
1252                 out->conn_rate.count = 0;
1253         else
1254                 out->conn_rate.count -=
1255                     in->conn_rate.count * diff /
1256                     in->conn_rate.seconds;
1257 }
1258
1259 #ifdef ALTQ
1260 /*
1261  * Handle export of struct pf_kaltq to user binaries that may be using any
1262  * version of struct pf_altq.
1263  */
1264 static int
1265 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1266 {
1267         u_int32_t version;
1268
1269         if (ioc_size == sizeof(struct pfioc_altq_v0))
1270                 version = 0;
1271         else
1272                 version = pa->version;
1273
1274         if (version > PFIOC_ALTQ_VERSION)
1275                 return (EINVAL);
1276
1277 #define ASSIGN(x) exported_q->x = q->x
1278 #define COPY(x) \
1279         bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1280 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1281 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1282
1283         switch (version) {
1284         case 0: {
1285                 struct pf_altq_v0 *exported_q =
1286                     &((struct pfioc_altq_v0 *)pa)->altq;
1287
1288                 COPY(ifname);
1289
1290                 ASSIGN(scheduler);
1291                 ASSIGN(tbrsize);
1292                 exported_q->tbrsize = SATU16(q->tbrsize);
1293                 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1294
1295                 COPY(qname);
1296                 COPY(parent);
1297                 ASSIGN(parent_qid);
1298                 exported_q->bandwidth = SATU32(q->bandwidth);
1299                 ASSIGN(priority);
1300                 ASSIGN(local_flags);
1301
1302                 ASSIGN(qlimit);
1303                 ASSIGN(flags);
1304
1305                 if (q->scheduler == ALTQT_HFSC) {
1306 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1307 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1308                             SATU32(q->pq_u.hfsc_opts.x)
1309
1310                         ASSIGN_OPT_SATU32(rtsc_m1);
1311                         ASSIGN_OPT(rtsc_d);
1312                         ASSIGN_OPT_SATU32(rtsc_m2);
1313
1314                         ASSIGN_OPT_SATU32(lssc_m1);
1315                         ASSIGN_OPT(lssc_d);
1316                         ASSIGN_OPT_SATU32(lssc_m2);
1317
1318                         ASSIGN_OPT_SATU32(ulsc_m1);
1319                         ASSIGN_OPT(ulsc_d);
1320                         ASSIGN_OPT_SATU32(ulsc_m2);
1321
1322                         ASSIGN_OPT(flags);
1323
1324 #undef ASSIGN_OPT
1325 #undef ASSIGN_OPT_SATU32
1326                 } else
1327                         COPY(pq_u);
1328
1329                 ASSIGN(qid);
1330                 break;
1331         }
1332         case 1: {
1333                 struct pf_altq_v1 *exported_q =
1334                     &((struct pfioc_altq_v1 *)pa)->altq;
1335
1336                 COPY(ifname);
1337
1338                 ASSIGN(scheduler);
1339                 ASSIGN(tbrsize);
1340                 ASSIGN(ifbandwidth);
1341
1342                 COPY(qname);
1343                 COPY(parent);
1344                 ASSIGN(parent_qid);
1345                 ASSIGN(bandwidth);
1346                 ASSIGN(priority);
1347                 ASSIGN(local_flags);
1348
1349                 ASSIGN(qlimit);
1350                 ASSIGN(flags);
1351                 COPY(pq_u);
1352
1353                 ASSIGN(qid);
1354                 break;
1355         }
1356         default:
1357                 panic("%s: unhandled struct pfioc_altq version", __func__);
1358                 break;
1359         }
1360
1361 #undef ASSIGN
1362 #undef COPY
1363 #undef SATU16
1364 #undef SATU32
1365
1366         return (0);
1367 }
1368
1369 /*
1370  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1371  * that may be using any version of it.
1372  */
1373 static int
1374 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1375 {
1376         u_int32_t version;
1377
1378         if (ioc_size == sizeof(struct pfioc_altq_v0))
1379                 version = 0;
1380         else
1381                 version = pa->version;
1382
1383         if (version > PFIOC_ALTQ_VERSION)
1384                 return (EINVAL);
1385
1386 #define ASSIGN(x) q->x = imported_q->x
1387 #define COPY(x) \
1388         bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1389
1390         switch (version) {
1391         case 0: {
1392                 struct pf_altq_v0 *imported_q =
1393                     &((struct pfioc_altq_v0 *)pa)->altq;
1394
1395                 COPY(ifname);
1396
1397                 ASSIGN(scheduler);
1398                 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1399                 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1400
1401                 COPY(qname);
1402                 COPY(parent);
1403                 ASSIGN(parent_qid);
1404                 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1405                 ASSIGN(priority);
1406                 ASSIGN(local_flags);
1407
1408                 ASSIGN(qlimit);
1409                 ASSIGN(flags);
1410
1411                 if (imported_q->scheduler == ALTQT_HFSC) {
1412 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1413
1414                         /*
1415                          * The m1 and m2 parameters are being copied from
1416                          * 32-bit to 64-bit.
1417                          */
1418                         ASSIGN_OPT(rtsc_m1);
1419                         ASSIGN_OPT(rtsc_d);
1420                         ASSIGN_OPT(rtsc_m2);
1421
1422                         ASSIGN_OPT(lssc_m1);
1423                         ASSIGN_OPT(lssc_d);
1424                         ASSIGN_OPT(lssc_m2);
1425
1426                         ASSIGN_OPT(ulsc_m1);
1427                         ASSIGN_OPT(ulsc_d);
1428                         ASSIGN_OPT(ulsc_m2);
1429
1430                         ASSIGN_OPT(flags);
1431
1432 #undef ASSIGN_OPT
1433                 } else
1434                         COPY(pq_u);
1435
1436                 ASSIGN(qid);
1437                 break;
1438         }
1439         case 1: {
1440                 struct pf_altq_v1 *imported_q =
1441                     &((struct pfioc_altq_v1 *)pa)->altq;
1442
1443                 COPY(ifname);
1444
1445                 ASSIGN(scheduler);
1446                 ASSIGN(tbrsize);
1447                 ASSIGN(ifbandwidth);
1448
1449                 COPY(qname);
1450                 COPY(parent);
1451                 ASSIGN(parent_qid);
1452                 ASSIGN(bandwidth);
1453                 ASSIGN(priority);
1454                 ASSIGN(local_flags);
1455
1456                 ASSIGN(qlimit);
1457                 ASSIGN(flags);
1458                 COPY(pq_u);
1459
1460                 ASSIGN(qid);
1461                 break;
1462         }
1463         default:
1464                 panic("%s: unhandled struct pfioc_altq version", __func__);
1465                 break;
1466         }
1467
1468 #undef ASSIGN
1469 #undef COPY
1470
1471         return (0);
1472 }
1473
1474 static struct pf_altq *
1475 pf_altq_get_nth_active(u_int32_t n)
1476 {
1477         struct pf_altq          *altq;
1478         u_int32_t                nr;
1479
1480         nr = 0;
1481         TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1482                 if (nr == n)
1483                         return (altq);
1484                 nr++;
1485         }
1486
1487         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1488                 if (nr == n)
1489                         return (altq);
1490                 nr++;
1491         }
1492
1493         return (NULL);
1494 }
1495 #endif /* ALTQ */
1496
1497 void
1498 pf_krule_free(struct pf_krule *rule)
1499 {
1500         if (rule == NULL)
1501                 return;
1502
1503         counter_u64_free(rule->evaluations);
1504         for (int i = 0; i < 2; i++) {
1505                 counter_u64_free(rule->packets[i]);
1506                 counter_u64_free(rule->bytes[i]);
1507         }
1508         counter_u64_free(rule->states_cur);
1509         counter_u64_free(rule->states_tot);
1510         counter_u64_free(rule->src_nodes);
1511         free(rule, M_PFRULE);
1512 }
1513
1514 static void
1515 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1516     struct pf_pooladdr *pool)
1517 {
1518
1519         bzero(pool, sizeof(*pool));
1520         bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1521         strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1522 }
1523
1524 static void
1525 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1526     struct pf_kpooladdr *kpool)
1527 {
1528
1529         bzero(kpool, sizeof(*kpool));
1530         bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1531         strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1532 }
1533
1534 static void
1535 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1536 {
1537         bzero(pool, sizeof(*pool));
1538
1539         bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1540         bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1541
1542         pool->tblidx = kpool->tblidx;
1543         pool->proxy_port[0] = kpool->proxy_port[0];
1544         pool->proxy_port[1] = kpool->proxy_port[1];
1545         pool->opts = kpool->opts;
1546 }
1547
1548 static int
1549 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1550 {
1551         _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1552         _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1553
1554         bzero(kpool, sizeof(*kpool));
1555
1556         bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1557         bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1558
1559         kpool->tblidx = pool->tblidx;
1560         kpool->proxy_port[0] = pool->proxy_port[0];
1561         kpool->proxy_port[1] = pool->proxy_port[1];
1562         kpool->opts = pool->opts;
1563
1564         return (0);
1565 }
1566
1567 static void
1568 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1569 {
1570
1571         bzero(rule, sizeof(*rule));
1572
1573         bcopy(&krule->src, &rule->src, sizeof(rule->src));
1574         bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1575
1576         for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1577                 if (rule->skip[i].ptr == NULL)
1578                         rule->skip[i].nr = -1;
1579                 else
1580                         rule->skip[i].nr = krule->skip[i].ptr->nr;
1581         }
1582
1583         strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1584         strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1585         strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1586         strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1587         strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1588         strlcpy(rule->match_tagname, krule->match_tagname,
1589             sizeof(rule->match_tagname));
1590         strlcpy(rule->overload_tblname, krule->overload_tblname,
1591             sizeof(rule->overload_tblname));
1592
1593         pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1594
1595         rule->evaluations = counter_u64_fetch(krule->evaluations);
1596         for (int i = 0; i < 2; i++) {
1597                 rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1598                 rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1599         }
1600
1601         /* kif, anchor, overload_tbl are not copied over. */
1602
1603         rule->os_fingerprint = krule->os_fingerprint;
1604
1605         rule->rtableid = krule->rtableid;
1606         bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1607         rule->max_states = krule->max_states;
1608         rule->max_src_nodes = krule->max_src_nodes;
1609         rule->max_src_states = krule->max_src_states;
1610         rule->max_src_conn = krule->max_src_conn;
1611         rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1612         rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1613         rule->qid = krule->qid;
1614         rule->pqid = krule->pqid;
1615         rule->nr = krule->nr;
1616         rule->prob = krule->prob;
1617         rule->cuid = krule->cuid;
1618         rule->cpid = krule->cpid;
1619
1620         rule->return_icmp = krule->return_icmp;
1621         rule->return_icmp6 = krule->return_icmp6;
1622         rule->max_mss = krule->max_mss;
1623         rule->tag = krule->tag;
1624         rule->match_tag = krule->match_tag;
1625         rule->scrub_flags = krule->scrub_flags;
1626
1627         bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1628         bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1629
1630         rule->rule_flag = krule->rule_flag;
1631         rule->action = krule->action;
1632         rule->direction = krule->direction;
1633         rule->log = krule->log;
1634         rule->logif = krule->logif;
1635         rule->quick = krule->quick;
1636         rule->ifnot = krule->ifnot;
1637         rule->match_tag_not = krule->match_tag_not;
1638         rule->natpass = krule->natpass;
1639
1640         rule->keep_state = krule->keep_state;
1641         rule->af = krule->af;
1642         rule->proto = krule->proto;
1643         rule->type = krule->type;
1644         rule->code = krule->code;
1645         rule->flags = krule->flags;
1646         rule->flagset = krule->flagset;
1647         rule->min_ttl = krule->min_ttl;
1648         rule->allow_opts = krule->allow_opts;
1649         rule->rt = krule->rt;
1650         rule->return_ttl = krule->return_ttl;
1651         rule->tos = krule->tos;
1652         rule->set_tos = krule->set_tos;
1653         rule->anchor_relative = krule->anchor_relative;
1654         rule->anchor_wildcard = krule->anchor_wildcard;
1655
1656         rule->flush = krule->flush;
1657         rule->prio = krule->prio;
1658         rule->set_prio[0] = krule->set_prio[0];
1659         rule->set_prio[1] = krule->set_prio[1];
1660
1661         bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1662
1663         rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1664         rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1665         rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1666 }
1667
1668 static int
1669 pf_check_rule_addr(const struct pf_rule_addr *addr)
1670 {
1671
1672         switch (addr->addr.type) {
1673         case PF_ADDR_ADDRMASK:
1674         case PF_ADDR_NOROUTE:
1675         case PF_ADDR_DYNIFTL:
1676         case PF_ADDR_TABLE:
1677         case PF_ADDR_URPFFAILED:
1678         case PF_ADDR_RANGE:
1679                 break;
1680         default:
1681                 return (EINVAL);
1682         }
1683
1684         if (addr->addr.p.dyn != NULL) {
1685                 return (EINVAL);
1686         }
1687
1688         return (0);
1689 }
1690
1691 static int
1692 pf_nvaddr_to_addr(const nvlist_t *nvl, struct pf_addr *paddr)
1693 {
1694         return (pf_nvbinary(nvl, "addr", paddr, sizeof(*paddr)));
1695 }
1696
1697 static nvlist_t *
1698 pf_addr_to_nvaddr(const struct pf_addr *paddr)
1699 {
1700         nvlist_t *nvl;
1701
1702         nvl = nvlist_create(0);
1703         if (nvl == NULL)
1704                 return (NULL);
1705
1706         nvlist_add_binary(nvl, "addr", paddr, sizeof(*paddr));
1707
1708         return (nvl);
1709 }
1710
1711 static int
1712 pf_nvmape_to_mape(const nvlist_t *nvl, struct pf_mape_portset *mape)
1713 {
1714         int error = 0;
1715
1716         bzero(mape, sizeof(*mape));
1717         PFNV_CHK(pf_nvuint8(nvl, "offset", &mape->offset));
1718         PFNV_CHK(pf_nvuint8(nvl, "psidlen", &mape->psidlen));
1719         PFNV_CHK(pf_nvuint16(nvl, "psid", &mape->psid));
1720
1721 errout:
1722         return (error);
1723 }
1724
1725 static nvlist_t *
1726 pf_mape_to_nvmape(const struct pf_mape_portset *mape)
1727 {
1728         nvlist_t *nvl;
1729
1730         nvl = nvlist_create(0);
1731         if (nvl == NULL)
1732                 return (NULL);
1733
1734         nvlist_add_number(nvl, "offset", mape->offset);
1735         nvlist_add_number(nvl, "psidlen", mape->psidlen);
1736         nvlist_add_number(nvl, "psid", mape->psid);
1737
1738         return (nvl);
1739 }
1740
1741 static int
1742 pf_nvpool_to_pool(const nvlist_t *nvl, struct pf_kpool *kpool)
1743 {
1744         int error = 0;
1745
1746         bzero(kpool, sizeof(*kpool));
1747
1748         PFNV_CHK(pf_nvbinary(nvl, "key", &kpool->key, sizeof(kpool->key)));
1749
1750         if (nvlist_exists_nvlist(nvl, "counter")) {
1751                 PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvl, "counter"),
1752                     &kpool->counter));
1753         }
1754
1755         PFNV_CHK(pf_nvint(nvl, "tblidx", &kpool->tblidx));
1756         PFNV_CHK(pf_nvuint16_array(nvl, "proxy_port", kpool->proxy_port, 2,
1757             NULL));
1758         PFNV_CHK(pf_nvuint8(nvl, "opts", &kpool->opts));
1759
1760         if (nvlist_exists_nvlist(nvl, "mape")) {
1761                 PFNV_CHK(pf_nvmape_to_mape(nvlist_get_nvlist(nvl, "mape"),
1762                     &kpool->mape));
1763         }
1764
1765 errout:
1766         return (error);
1767 }
1768
1769 static nvlist_t *
1770 pf_pool_to_nvpool(const struct pf_kpool *pool)
1771 {
1772         nvlist_t *nvl;
1773         nvlist_t *tmp;
1774
1775         nvl = nvlist_create(0);
1776         if (nvl == NULL)
1777                 return (NULL);
1778
1779         nvlist_add_binary(nvl, "key", &pool->key, sizeof(pool->key));
1780         tmp = pf_addr_to_nvaddr(&pool->counter);
1781         if (tmp == NULL)
1782                 goto error;
1783         nvlist_add_nvlist(nvl, "counter", tmp);
1784
1785         nvlist_add_number(nvl, "tblidx", pool->tblidx);
1786         pf_uint16_array_nv(nvl, "proxy_port", pool->proxy_port, 2);
1787         nvlist_add_number(nvl, "opts", pool->opts);
1788
1789         tmp = pf_mape_to_nvmape(&pool->mape);
1790         if (tmp == NULL)
1791                 goto error;
1792         nvlist_add_nvlist(nvl, "mape", tmp);
1793
1794         return (nvl);
1795
1796 error:
1797         nvlist_destroy(nvl);
1798         return (NULL);
1799 }
1800
1801 static int
1802 pf_nvaddr_wrap_to_addr_wrap(const nvlist_t *nvl, struct pf_addr_wrap *addr)
1803 {
1804         int error = 0;
1805
1806         bzero(addr, sizeof(*addr));
1807
1808         PFNV_CHK(pf_nvuint8(nvl, "type", &addr->type));
1809         PFNV_CHK(pf_nvuint8(nvl, "iflags", &addr->iflags));
1810         if (addr->type == PF_ADDR_DYNIFTL)
1811                 PFNV_CHK(pf_nvstring(nvl, "ifname", addr->v.ifname,
1812                     sizeof(addr->v.ifname)));
1813         if (addr->type == PF_ADDR_TABLE)
1814                 PFNV_CHK(pf_nvstring(nvl, "tblname", addr->v.tblname,
1815                     sizeof(addr->v.tblname)));
1816
1817         if (! nvlist_exists_nvlist(nvl, "addr"))
1818                 return (EINVAL);
1819         PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvl, "addr"),
1820             &addr->v.a.addr));
1821
1822         if (! nvlist_exists_nvlist(nvl, "mask"))
1823                 return (EINVAL);
1824         PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvl, "mask"),
1825             &addr->v.a.mask));
1826
1827         switch (addr->type) {
1828         case PF_ADDR_DYNIFTL:
1829         case PF_ADDR_TABLE:
1830         case PF_ADDR_RANGE:
1831         case PF_ADDR_ADDRMASK:
1832         case PF_ADDR_NOROUTE:
1833         case PF_ADDR_URPFFAILED:
1834                 break;
1835         default:
1836                 return (EINVAL);
1837         }
1838
1839 errout:
1840         return (error);
1841 }
1842
1843 static nvlist_t *
1844 pf_addr_wrap_to_nvaddr_wrap(const struct pf_addr_wrap *addr)
1845 {
1846         nvlist_t *nvl;
1847         nvlist_t *tmp;
1848
1849         nvl = nvlist_create(0);
1850         if (nvl == NULL)
1851                 return (NULL);
1852
1853         nvlist_add_number(nvl, "type", addr->type);
1854         nvlist_add_number(nvl, "iflags", addr->iflags);
1855         if (addr->type == PF_ADDR_DYNIFTL)
1856                 nvlist_add_string(nvl, "ifname", addr->v.ifname);
1857         if (addr->type == PF_ADDR_TABLE)
1858                 nvlist_add_string(nvl, "tblname", addr->v.tblname);
1859
1860         tmp = pf_addr_to_nvaddr(&addr->v.a.addr);
1861         if (tmp == NULL)
1862                 goto error;
1863         nvlist_add_nvlist(nvl, "addr", tmp);
1864         tmp = pf_addr_to_nvaddr(&addr->v.a.mask);
1865         if (tmp == NULL)
1866                 goto error;
1867         nvlist_add_nvlist(nvl, "mask", tmp);
1868
1869         return (nvl);
1870
1871 error:
1872         nvlist_destroy(nvl);
1873         return (NULL);
1874 }
1875
1876 static int
1877 pf_validate_op(uint8_t op)
1878 {
1879         switch (op) {
1880         case PF_OP_NONE:
1881         case PF_OP_IRG:
1882         case PF_OP_EQ:
1883         case PF_OP_NE:
1884         case PF_OP_LT:
1885         case PF_OP_LE:
1886         case PF_OP_GT:
1887         case PF_OP_GE:
1888         case PF_OP_XRG:
1889         case PF_OP_RRG:
1890                 break;
1891         default:
1892                 return (EINVAL);
1893         }
1894
1895         return (0);
1896 }
1897
1898 static int
1899 pf_nvrule_addr_to_rule_addr(const nvlist_t *nvl, struct pf_rule_addr *addr)
1900 {
1901         int error = 0;
1902
1903         if (! nvlist_exists_nvlist(nvl, "addr"))
1904                 return (EINVAL);
1905
1906         PFNV_CHK(pf_nvaddr_wrap_to_addr_wrap(nvlist_get_nvlist(nvl, "addr"),
1907             &addr->addr));
1908         PFNV_CHK(pf_nvuint16_array(nvl, "port", addr->port, 2, NULL));
1909         PFNV_CHK(pf_nvuint8(nvl, "neg", &addr->neg));
1910         PFNV_CHK(pf_nvuint8(nvl, "port_op", &addr->port_op));
1911
1912         PFNV_CHK(pf_validate_op(addr->port_op));
1913
1914 errout:
1915         return (error);
1916 }
1917
1918 static nvlist_t *
1919 pf_rule_addr_to_nvrule_addr(const struct pf_rule_addr *addr)
1920 {
1921         nvlist_t *nvl;
1922         nvlist_t *tmp;
1923
1924         nvl = nvlist_create(0);
1925         if (nvl == NULL)
1926                 return (NULL);
1927
1928         tmp = pf_addr_wrap_to_nvaddr_wrap(&addr->addr);
1929         if (tmp == NULL)
1930                 goto error;
1931         nvlist_add_nvlist(nvl, "addr", tmp);
1932         pf_uint16_array_nv(nvl, "port", addr->port, 2);
1933         nvlist_add_number(nvl, "neg", addr->neg);
1934         nvlist_add_number(nvl, "port_op", addr->port_op);
1935
1936         return (nvl);
1937
1938 error:
1939         nvlist_destroy(nvl);
1940         return (NULL);
1941 }
1942
1943 static int
1944 pf_nvrule_uid_to_rule_uid(const nvlist_t *nvl, struct pf_rule_uid *uid)
1945 {
1946         int error = 0;
1947
1948         bzero(uid, sizeof(*uid));
1949
1950         PFNV_CHK(pf_nvuint32_array(nvl, "uid", uid->uid, 2, NULL));
1951         PFNV_CHK(pf_nvuint8(nvl, "op", &uid->op));
1952
1953         PFNV_CHK(pf_validate_op(uid->op));
1954
1955 errout:
1956         return (error);
1957 }
1958
1959 static nvlist_t *
1960 pf_rule_uid_to_nvrule_uid(const struct pf_rule_uid *uid)
1961 {
1962         nvlist_t *nvl;
1963
1964         nvl = nvlist_create(0);
1965         if (nvl == NULL)
1966                 return (NULL);
1967
1968         pf_uint32_array_nv(nvl, "uid", uid->uid, 2);
1969         nvlist_add_number(nvl, "op", uid->op);
1970
1971         return (nvl);
1972 }
1973
1974 static int
1975 pf_nvrule_gid_to_rule_gid(const nvlist_t *nvl, struct pf_rule_gid *gid)
1976 {
1977         /* Cheat a little. These stucts are the same, other than the name of
1978          * the first field. */
1979         return (pf_nvrule_uid_to_rule_uid(nvl, (struct pf_rule_uid *)gid));
1980 }
1981
1982 static int
1983 pf_nvrule_to_krule(const nvlist_t *nvl, struct pf_krule **prule)
1984 {
1985         struct pf_krule *rule;
1986         int error = 0;
1987
1988 #define ERROUT(x)       ERROUT_FUNCTION(errout, x)
1989
1990         rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK | M_ZERO);
1991
1992         PFNV_CHK(pf_nvuint32(nvl, "nr", &rule->nr));
1993
1994         if (! nvlist_exists_nvlist(nvl, "src"))
1995                 ERROUT(EINVAL);
1996
1997         error = pf_nvrule_addr_to_rule_addr(nvlist_get_nvlist(nvl, "src"),
1998             &rule->src);
1999         if (error != 0)
2000                 ERROUT(error);
2001
2002         if (! nvlist_exists_nvlist(nvl, "dst"))
2003                 ERROUT(EINVAL);
2004
2005         PFNV_CHK(pf_nvrule_addr_to_rule_addr(nvlist_get_nvlist(nvl, "dst"),
2006             &rule->dst));
2007
2008         if (nvlist_exists_string(nvl, "label")) {
2009                 PFNV_CHK(pf_nvstring(nvl, "label", rule->label[0],
2010                     sizeof(rule->label[0])));
2011         } else if (nvlist_exists_string_array(nvl, "labels")) {
2012                 const char *const *strs;
2013                 size_t items;
2014                 int ret;
2015
2016                 strs = nvlist_get_string_array(nvl, "labels", &items);
2017                 if (items > PF_RULE_MAX_LABEL_COUNT)
2018                         ERROUT(E2BIG);
2019
2020                 for (size_t i = 0; i < items; i++) {
2021                         ret = strlcpy(rule->label[i], strs[i],
2022                             sizeof(rule->label[0]));
2023                         if (ret >= sizeof(rule->label[0]))
2024                                 ERROUT(E2BIG);
2025                 }
2026         }
2027
2028         PFNV_CHK(pf_nvstring(nvl, "ifname", rule->ifname,
2029             sizeof(rule->ifname)));
2030         PFNV_CHK(pf_nvstring(nvl, "qname", rule->qname, sizeof(rule->qname)));
2031         PFNV_CHK(pf_nvstring(nvl, "pqname", rule->pqname,
2032             sizeof(rule->pqname)));
2033         PFNV_CHK(pf_nvstring(nvl, "tagname", rule->tagname,
2034             sizeof(rule->tagname)));
2035         PFNV_CHK(pf_nvstring(nvl, "match_tagname", rule->match_tagname,
2036             sizeof(rule->match_tagname)));
2037         PFNV_CHK(pf_nvstring(nvl, "overload_tblname", rule->overload_tblname,
2038             sizeof(rule->overload_tblname)));
2039
2040         if (! nvlist_exists_nvlist(nvl, "rpool"))
2041                 ERROUT(EINVAL);
2042         PFNV_CHK(pf_nvpool_to_pool(nvlist_get_nvlist(nvl, "rpool"),
2043             &rule->rpool));
2044
2045         PFNV_CHK(pf_nvuint32(nvl, "os_fingerprint", &rule->os_fingerprint));
2046
2047         PFNV_CHK(pf_nvint(nvl, "rtableid", &rule->rtableid));
2048         PFNV_CHK(pf_nvuint32_array(nvl, "timeout", rule->timeout, PFTM_MAX, NULL));
2049         PFNV_CHK(pf_nvuint32(nvl, "max_states", &rule->max_states));
2050         PFNV_CHK(pf_nvuint32(nvl, "max_src_nodes", &rule->max_src_nodes));
2051         PFNV_CHK(pf_nvuint32(nvl, "max_src_states", &rule->max_src_states));
2052         PFNV_CHK(pf_nvuint32(nvl, "max_src_conn", &rule->max_src_conn));
2053         PFNV_CHK(pf_nvuint32(nvl, "max_src_conn_rate.limit",
2054             &rule->max_src_conn_rate.limit));
2055         PFNV_CHK(pf_nvuint32(nvl, "max_src_conn_rate.seconds",
2056             &rule->max_src_conn_rate.seconds));
2057         PFNV_CHK(pf_nvuint32(nvl, "prob", &rule->prob));
2058         PFNV_CHK(pf_nvuint32(nvl, "cuid", &rule->cuid));
2059         PFNV_CHK(pf_nvuint32(nvl, "cpid", &rule->cpid));
2060
2061         PFNV_CHK(pf_nvuint16(nvl, "return_icmp", &rule->return_icmp));
2062         PFNV_CHK(pf_nvuint16(nvl, "return_icmp6", &rule->return_icmp6));
2063
2064         PFNV_CHK(pf_nvuint16(nvl, "max_mss", &rule->max_mss));
2065         PFNV_CHK(pf_nvuint16(nvl, "scrub_flags", &rule->scrub_flags));
2066
2067         if (! nvlist_exists_nvlist(nvl, "uid"))
2068                 ERROUT(EINVAL);
2069         PFNV_CHK(pf_nvrule_uid_to_rule_uid(nvlist_get_nvlist(nvl, "uid"),
2070             &rule->uid));
2071
2072         if (! nvlist_exists_nvlist(nvl, "gid"))
2073                 ERROUT(EINVAL);
2074         PFNV_CHK(pf_nvrule_gid_to_rule_gid(nvlist_get_nvlist(nvl, "gid"),
2075             &rule->gid));
2076
2077         PFNV_CHK(pf_nvuint32(nvl, "rule_flag", &rule->rule_flag));
2078         PFNV_CHK(pf_nvuint8(nvl, "action", &rule->action));
2079         PFNV_CHK(pf_nvuint8(nvl, "direction", &rule->direction));
2080         PFNV_CHK(pf_nvuint8(nvl, "log", &rule->log));
2081         PFNV_CHK(pf_nvuint8(nvl, "logif", &rule->logif));
2082         PFNV_CHK(pf_nvuint8(nvl, "quick", &rule->quick));
2083         PFNV_CHK(pf_nvuint8(nvl, "ifnot", &rule->ifnot));
2084         PFNV_CHK(pf_nvuint8(nvl, "match_tag_not", &rule->match_tag_not));
2085         PFNV_CHK(pf_nvuint8(nvl, "natpass", &rule->natpass));
2086
2087         PFNV_CHK(pf_nvuint8(nvl, "keep_state", &rule->keep_state));
2088         PFNV_CHK(pf_nvuint8(nvl, "af", &rule->af));
2089         PFNV_CHK(pf_nvuint8(nvl, "proto", &rule->proto));
2090         PFNV_CHK(pf_nvuint8(nvl, "type", &rule->type));
2091         PFNV_CHK(pf_nvuint8(nvl, "code", &rule->code));
2092         PFNV_CHK(pf_nvuint8(nvl, "flags", &rule->flags));
2093         PFNV_CHK(pf_nvuint8(nvl, "flagset", &rule->flagset));
2094         PFNV_CHK(pf_nvuint8(nvl, "min_ttl", &rule->min_ttl));
2095         PFNV_CHK(pf_nvuint8(nvl, "allow_opts", &rule->allow_opts));
2096         PFNV_CHK(pf_nvuint8(nvl, "rt", &rule->rt));
2097         PFNV_CHK(pf_nvuint8(nvl, "return_ttl", &rule->return_ttl));
2098         PFNV_CHK(pf_nvuint8(nvl, "tos", &rule->tos));
2099         PFNV_CHK(pf_nvuint8(nvl, "set_tos", &rule->set_tos));
2100         PFNV_CHK(pf_nvuint8(nvl, "anchor_relative", &rule->anchor_relative));
2101         PFNV_CHK(pf_nvuint8(nvl, "anchor_wildcard", &rule->anchor_wildcard));
2102
2103         PFNV_CHK(pf_nvuint8(nvl, "flush", &rule->flush));
2104         PFNV_CHK(pf_nvuint8(nvl, "prio", &rule->prio));
2105
2106         PFNV_CHK(pf_nvuint8_array(nvl, "set_prio", &rule->prio, 2, NULL));
2107
2108         if (nvlist_exists_nvlist(nvl, "divert")) {
2109                 const nvlist_t *nvldivert = nvlist_get_nvlist(nvl, "divert");
2110
2111                 if (! nvlist_exists_nvlist(nvldivert, "addr"))
2112                         ERROUT(EINVAL);
2113                 PFNV_CHK(pf_nvaddr_to_addr(nvlist_get_nvlist(nvldivert, "addr"),
2114                     &rule->divert.addr));
2115                 PFNV_CHK(pf_nvuint16(nvldivert, "port", &rule->divert.port));
2116         }
2117
2118         /* Validation */
2119 #ifndef INET
2120         if (rule->af == AF_INET)
2121                 ERROUT(EAFNOSUPPORT);
2122 #endif /* INET */
2123 #ifndef INET6
2124         if (rule->af == AF_INET6)
2125                 ERROUT(EAFNOSUPPORT);
2126 #endif /* INET6 */
2127
2128         PFNV_CHK(pf_check_rule_addr(&rule->src));
2129         PFNV_CHK(pf_check_rule_addr(&rule->dst));
2130
2131         *prule = rule;
2132
2133         return (0);
2134
2135 #undef ERROUT
2136 errout:
2137         pf_krule_free(rule);
2138         *prule = NULL;
2139
2140         return (error);
2141 }
2142
2143 static nvlist_t *
2144 pf_divert_to_nvdivert(const struct pf_krule *rule)
2145 {
2146         nvlist_t *nvl;
2147         nvlist_t *tmp;
2148
2149         nvl = nvlist_create(0);
2150         if (nvl == NULL)
2151                 return (NULL);
2152
2153         tmp = pf_addr_to_nvaddr(&rule->divert.addr);
2154         if (tmp == NULL)
2155                 goto error;
2156         nvlist_add_nvlist(nvl, "addr", tmp);
2157         nvlist_add_number(nvl, "port", rule->divert.port);
2158
2159         return (nvl);
2160
2161 error:
2162         nvlist_destroy(nvl);
2163         return (NULL);
2164 }
2165
2166 static nvlist_t *
2167 pf_krule_to_nvrule(const struct pf_krule *rule)
2168 {
2169         nvlist_t *nvl, *tmp;
2170
2171         nvl = nvlist_create(0);
2172         if (nvl == NULL)
2173                 return (nvl);
2174
2175         nvlist_add_number(nvl, "nr", rule->nr);
2176         tmp = pf_rule_addr_to_nvrule_addr(&rule->src);
2177         if (tmp == NULL)
2178                 goto error;
2179         nvlist_add_nvlist(nvl, "src", tmp);
2180         tmp = pf_rule_addr_to_nvrule_addr(&rule->dst);
2181         if (tmp == NULL)
2182                 goto error;
2183         nvlist_add_nvlist(nvl, "dst", tmp);
2184
2185         for (int i = 0; i < PF_SKIP_COUNT; i++) {
2186                 nvlist_append_number_array(nvl, "skip",
2187                     rule->skip[i].ptr ? rule->skip[i].ptr->nr : -1);
2188         }
2189
2190         for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) {
2191                 nvlist_append_string_array(nvl, "labels", rule->label[i]);
2192         }
2193         nvlist_add_string(nvl, "label", rule->label[0]);
2194         nvlist_add_string(nvl, "ifname", rule->ifname);
2195         nvlist_add_string(nvl, "qname", rule->qname);
2196         nvlist_add_string(nvl, "pqname", rule->pqname);
2197         nvlist_add_string(nvl, "tagname", rule->tagname);
2198         nvlist_add_string(nvl, "match_tagname", rule->match_tagname);
2199         nvlist_add_string(nvl, "overload_tblname", rule->overload_tblname);
2200
2201         tmp = pf_pool_to_nvpool(&rule->rpool);
2202         if (tmp == NULL)
2203                 goto error;
2204         nvlist_add_nvlist(nvl, "rpool", tmp);
2205
2206         nvlist_add_number(nvl, "evaluations",
2207             counter_u64_fetch(rule->evaluations));
2208         for (int i = 0; i < 2; i++) {
2209                 nvlist_append_number_array(nvl, "packets",
2210                     counter_u64_fetch(rule->packets[i]));
2211                 nvlist_append_number_array(nvl, "bytes",
2212                     counter_u64_fetch(rule->bytes[i]));
2213         }
2214
2215         nvlist_add_number(nvl, "os_fingerprint", rule->os_fingerprint);
2216
2217         nvlist_add_number(nvl, "rtableid", rule->rtableid);
2218         pf_uint32_array_nv(nvl, "timeout", rule->timeout, PFTM_MAX);
2219         nvlist_add_number(nvl, "max_states", rule->max_states);
2220         nvlist_add_number(nvl, "max_src_nodes", rule->max_src_nodes);
2221         nvlist_add_number(nvl, "max_src_states", rule->max_src_states);
2222         nvlist_add_number(nvl, "max_src_conn", rule->max_src_conn);
2223         nvlist_add_number(nvl, "max_src_conn_rate.limit",
2224             rule->max_src_conn_rate.limit);
2225         nvlist_add_number(nvl, "max_src_conn_rate.seconds",
2226             rule->max_src_conn_rate.seconds);
2227         nvlist_add_number(nvl, "qid", rule->qid);
2228         nvlist_add_number(nvl, "pqid", rule->pqid);
2229         nvlist_add_number(nvl, "prob", rule->prob);
2230         nvlist_add_number(nvl, "cuid", rule->cuid);
2231         nvlist_add_number(nvl, "cpid", rule->cpid);
2232
2233         nvlist_add_number(nvl, "states_cur",
2234             counter_u64_fetch(rule->states_cur));
2235         nvlist_add_number(nvl, "states_tot",
2236             counter_u64_fetch(rule->states_tot));
2237         nvlist_add_number(nvl, "src_nodes",
2238             counter_u64_fetch(rule->src_nodes));
2239
2240         nvlist_add_number(nvl, "return_icmp", rule->return_icmp);
2241         nvlist_add_number(nvl, "return_icmp6", rule->return_icmp6);
2242
2243         nvlist_add_number(nvl, "max_mss", rule->max_mss);
2244         nvlist_add_number(nvl, "scrub_flags", rule->scrub_flags);
2245
2246         tmp = pf_rule_uid_to_nvrule_uid(&rule->uid);
2247         if (tmp == NULL)
2248                 goto error;
2249         nvlist_add_nvlist(nvl, "uid", tmp);
2250         tmp = pf_rule_uid_to_nvrule_uid((const struct pf_rule_uid *)&rule->gid);
2251         if (tmp == NULL)
2252                 goto error;
2253         nvlist_add_nvlist(nvl, "gid", tmp);
2254
2255         nvlist_add_number(nvl, "rule_flag", rule->rule_flag);
2256         nvlist_add_number(nvl, "action", rule->action);
2257         nvlist_add_number(nvl, "direction", rule->direction);
2258         nvlist_add_number(nvl, "log", rule->log);
2259         nvlist_add_number(nvl, "logif", rule->logif);
2260         nvlist_add_number(nvl, "quick", rule->quick);
2261         nvlist_add_number(nvl, "ifnot", rule->ifnot);
2262         nvlist_add_number(nvl, "match_tag_not", rule->match_tag_not);
2263         nvlist_add_number(nvl, "natpass", rule->natpass);
2264
2265         nvlist_add_number(nvl, "keep_state", rule->keep_state);
2266         nvlist_add_number(nvl, "af", rule->af);
2267         nvlist_add_number(nvl, "proto", rule->proto);
2268         nvlist_add_number(nvl, "type", rule->type);
2269         nvlist_add_number(nvl, "code", rule->code);
2270         nvlist_add_number(nvl, "flags", rule->flags);
2271         nvlist_add_number(nvl, "flagset", rule->flagset);
2272         nvlist_add_number(nvl, "min_ttl", rule->min_ttl);
2273         nvlist_add_number(nvl, "allow_opts", rule->allow_opts);
2274         nvlist_add_number(nvl, "rt", rule->rt);
2275         nvlist_add_number(nvl, "return_ttl", rule->return_ttl);
2276         nvlist_add_number(nvl, "tos", rule->tos);
2277         nvlist_add_number(nvl, "set_tos", rule->set_tos);
2278         nvlist_add_number(nvl, "anchor_relative", rule->anchor_relative);
2279         nvlist_add_number(nvl, "anchor_wildcard", rule->anchor_wildcard);
2280
2281         nvlist_add_number(nvl, "flush", rule->flush);
2282         nvlist_add_number(nvl, "prio", rule->prio);
2283
2284         pf_uint8_array_nv(nvl, "set_prio", &rule->prio, 2);
2285
2286         tmp = pf_divert_to_nvdivert(rule);
2287         if (tmp == NULL)
2288                 goto error;
2289         nvlist_add_nvlist(nvl, "divert", tmp);
2290
2291         return (nvl);
2292
2293 error:
2294         nvlist_destroy(nvl);
2295         return (NULL);
2296 }
2297
2298 static int
2299 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
2300 {
2301         int ret;
2302
2303 #ifndef INET
2304         if (rule->af == AF_INET) {
2305                 return (EAFNOSUPPORT);
2306         }
2307 #endif /* INET */
2308 #ifndef INET6
2309         if (rule->af == AF_INET6) {
2310                 return (EAFNOSUPPORT);
2311         }
2312 #endif /* INET6 */
2313
2314         ret = pf_check_rule_addr(&rule->src);
2315         if (ret != 0)
2316                 return (ret);
2317         ret = pf_check_rule_addr(&rule->dst);
2318         if (ret != 0)
2319                 return (ret);
2320
2321         bzero(krule, sizeof(*krule));
2322
2323         bcopy(&rule->src, &krule->src, sizeof(rule->src));
2324         bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
2325
2326         strlcpy(krule->label[0], rule->label, sizeof(rule->label));
2327         strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
2328         strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
2329         strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
2330         strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
2331         strlcpy(krule->match_tagname, rule->match_tagname,
2332             sizeof(rule->match_tagname));
2333         strlcpy(krule->overload_tblname, rule->overload_tblname,
2334             sizeof(rule->overload_tblname));
2335
2336         ret = pf_pool_to_kpool(&rule->rpool, &krule->rpool);
2337         if (ret != 0)
2338                 return (ret);
2339
2340         /* Don't allow userspace to set evaulations, packets or bytes. */
2341         /* kif, anchor, overload_tbl are not copied over. */
2342
2343         krule->os_fingerprint = rule->os_fingerprint;
2344
2345         krule->rtableid = rule->rtableid;
2346         bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
2347         krule->max_states = rule->max_states;
2348         krule->max_src_nodes = rule->max_src_nodes;
2349         krule->max_src_states = rule->max_src_states;
2350         krule->max_src_conn = rule->max_src_conn;
2351         krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2352         krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2353         krule->qid = rule->qid;
2354         krule->pqid = rule->pqid;
2355         krule->nr = rule->nr;
2356         krule->prob = rule->prob;
2357         krule->cuid = rule->cuid;
2358         krule->cpid = rule->cpid;
2359
2360         krule->return_icmp = rule->return_icmp;
2361         krule->return_icmp6 = rule->return_icmp6;
2362         krule->max_mss = rule->max_mss;
2363         krule->tag = rule->tag;
2364         krule->match_tag = rule->match_tag;
2365         krule->scrub_flags = rule->scrub_flags;
2366
2367         bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2368         bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2369
2370         krule->rule_flag = rule->rule_flag;
2371         krule->action = rule->action;
2372         krule->direction = rule->direction;
2373         krule->log = rule->log;
2374         krule->logif = rule->logif;
2375         krule->quick = rule->quick;
2376         krule->ifnot = rule->ifnot;
2377         krule->match_tag_not = rule->match_tag_not;
2378         krule->natpass = rule->natpass;
2379
2380         krule->keep_state = rule->keep_state;
2381         krule->af = rule->af;
2382         krule->proto = rule->proto;
2383         krule->type = rule->type;
2384         krule->code = rule->code;
2385         krule->flags = rule->flags;
2386         krule->flagset = rule->flagset;
2387         krule->min_ttl = rule->min_ttl;
2388         krule->allow_opts = rule->allow_opts;
2389         krule->rt = rule->rt;
2390         krule->return_ttl = rule->return_ttl;
2391         krule->tos = rule->tos;
2392         krule->set_tos = rule->set_tos;
2393         krule->anchor_relative = rule->anchor_relative;
2394         krule->anchor_wildcard = rule->anchor_wildcard;
2395
2396         krule->flush = rule->flush;
2397         krule->prio = rule->prio;
2398         krule->set_prio[0] = rule->set_prio[0];
2399         krule->set_prio[1] = rule->set_prio[1];
2400
2401         bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2402
2403         return (0);
2404 }
2405
2406 static int
2407 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk,
2408     struct pf_kstate_kill *kill)
2409 {
2410         bzero(kill, sizeof(*kill));
2411
2412         bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp));
2413         kill->psk_af = psk->psk_af;
2414         kill->psk_proto = psk->psk_proto;
2415         bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src));
2416         bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst));
2417         strlcpy(kill->psk_ifname, psk->psk_ifname, sizeof(kill->psk_ifname));
2418         strlcpy(kill->psk_label, psk->psk_label, sizeof(kill->psk_label));
2419
2420         return (0);
2421 }
2422
2423 static int
2424 pf_nvstate_cmp_to_state_cmp(const nvlist_t *nvl, struct pf_state_cmp *cmp)
2425 {
2426         int error = 0;
2427
2428         bzero(cmp, sizeof(*cmp));
2429
2430         PFNV_CHK(pf_nvuint64(nvl, "id", &cmp->id));
2431         PFNV_CHK(pf_nvuint32(nvl, "creatorid", &cmp->creatorid));
2432         PFNV_CHK(pf_nvuint8(nvl, "direction", &cmp->direction));
2433
2434 errout:
2435         return (error);
2436 }
2437
2438 static int
2439 pf_nvstate_kill_to_kstate_kill(const nvlist_t *nvl,
2440     struct pf_kstate_kill *kill)
2441 {
2442         int error = 0;
2443
2444         bzero(kill, sizeof(*kill));
2445
2446         if (! nvlist_exists_nvlist(nvl, "cmp"))
2447                 return (EINVAL);
2448
2449         PFNV_CHK(pf_nvstate_cmp_to_state_cmp(nvlist_get_nvlist(nvl, "cmp"),
2450             &kill->psk_pfcmp));
2451         PFNV_CHK(pf_nvuint8(nvl, "af", &kill->psk_af));
2452         PFNV_CHK(pf_nvint(nvl, "proto", &kill->psk_proto));
2453
2454         if (! nvlist_exists_nvlist(nvl, "src"))
2455                 return (EINVAL);
2456         PFNV_CHK(pf_nvrule_addr_to_rule_addr(nvlist_get_nvlist(nvl, "src"),
2457             &kill->psk_src));
2458         if (! nvlist_exists_nvlist(nvl, "dst"))
2459                 return (EINVAL);
2460         PFNV_CHK(pf_nvrule_addr_to_rule_addr(nvlist_get_nvlist(nvl, "dst"),
2461             &kill->psk_dst));
2462         if (nvlist_exists_nvlist(nvl, "rt_addr")) {
2463                 PFNV_CHK(pf_nvrule_addr_to_rule_addr(
2464                     nvlist_get_nvlist(nvl, "rt_addr"), &kill->psk_rt_addr));
2465         }
2466
2467         PFNV_CHK(pf_nvstring(nvl, "ifname", kill->psk_ifname,
2468             sizeof(kill->psk_ifname)));
2469         PFNV_CHK(pf_nvstring(nvl, "label", kill->psk_label,
2470             sizeof(kill->psk_label)));
2471         if (nvlist_exists_bool(nvl, "kill_match"))
2472                 kill->psk_kill_match = nvlist_get_bool(nvl, "kill_match");
2473
2474 errout:
2475         return (error);
2476 }
2477
2478 static nvlist_t *
2479 pf_state_key_to_nvstate_key(const struct pf_state_key *key)
2480 {
2481         nvlist_t        *nvl, *tmp;
2482
2483         nvl = nvlist_create(0);
2484         if (nvl == NULL)
2485                 return (NULL);
2486
2487         for (int i = 0; i < 2; i++) {
2488                 tmp = pf_addr_to_nvaddr(&key->addr[i]);
2489                 if (tmp == NULL)
2490                         goto errout;
2491                 nvlist_append_nvlist_array(nvl, "addr", tmp);
2492                 nvlist_append_number_array(nvl, "port", key->port[i]);
2493         }
2494         nvlist_add_number(nvl, "af", key->af);
2495         nvlist_add_number(nvl, "proto", key->proto);
2496
2497         return (nvl);
2498
2499 errout:
2500         nvlist_destroy(nvl);
2501         return (NULL);
2502 }
2503
2504 static nvlist_t *
2505 pf_state_scrub_to_nvstate_scrub(const struct pf_state_scrub *scrub)
2506 {
2507         nvlist_t *nvl;
2508
2509         nvl = nvlist_create(0);
2510         if (nvl == NULL)
2511                 return (NULL);
2512
2513         nvlist_add_bool(nvl, "timestamp", scrub->pfss_flags & PFSS_TIMESTAMP);
2514         nvlist_add_number(nvl, "ttl", scrub->pfss_ttl);
2515         nvlist_add_number(nvl, "ts_mod", scrub->pfss_ts_mod);
2516
2517         return (nvl);
2518 }
2519
2520 static nvlist_t *
2521 pf_state_peer_to_nvstate_peer(const struct pf_state_peer *peer)
2522 {
2523         nvlist_t *nvl, *tmp;
2524
2525         nvl = nvlist_create(0);
2526         if (nvl == NULL)
2527                 return (NULL);
2528
2529         if (peer->scrub) {
2530                 tmp = pf_state_scrub_to_nvstate_scrub(peer->scrub);
2531                 if (tmp == NULL)
2532                         goto errout;
2533                 nvlist_add_nvlist(nvl, "scrub", tmp);
2534         }
2535
2536         nvlist_add_number(nvl, "seqlo", peer->seqlo);
2537         nvlist_add_number(nvl, "seqhi", peer->seqhi);
2538         nvlist_add_number(nvl, "seqdiff", peer->seqdiff);
2539         nvlist_add_number(nvl, "max_win", peer->max_win);
2540         nvlist_add_number(nvl, "mss", peer->mss);
2541         nvlist_add_number(nvl, "state", peer->state);
2542         nvlist_add_number(nvl, "wscale", peer->wscale);
2543
2544         return (nvl);
2545
2546 errout:
2547         nvlist_destroy(nvl);
2548         return (NULL);
2549 }
2550
2551
2552 static nvlist_t *
2553 pf_state_to_nvstate(const struct pf_state *s)
2554 {
2555         nvlist_t        *nvl, *tmp;
2556         uint32_t         expire, flags = 0;
2557
2558         nvl = nvlist_create(0);
2559         if (nvl == NULL)
2560                 return (NULL);
2561
2562         nvlist_add_number(nvl, "id", s->id);
2563         nvlist_add_string(nvl, "ifname", s->kif->pfik_name);
2564
2565         tmp = pf_state_key_to_nvstate_key(s->key[PF_SK_STACK]);
2566         if (tmp == NULL)
2567                 goto errout;
2568         nvlist_add_nvlist(nvl, "stack_key", tmp);
2569
2570         tmp = pf_state_key_to_nvstate_key(s->key[PF_SK_WIRE]);
2571         if (tmp == NULL)
2572                 goto errout;
2573         nvlist_add_nvlist(nvl, "wire_key", tmp);
2574
2575         tmp = pf_state_peer_to_nvstate_peer(&s->src);
2576         if (tmp == NULL)
2577                 goto errout;
2578         nvlist_add_nvlist(nvl, "src", tmp);
2579
2580         tmp = pf_state_peer_to_nvstate_peer(&s->dst);
2581         if (tmp == NULL)
2582                 goto errout;
2583         nvlist_add_nvlist(nvl, "dst", tmp);
2584
2585         tmp = pf_addr_to_nvaddr(&s->rt_addr);
2586         if (tmp == NULL)
2587                 goto errout;
2588         nvlist_add_nvlist(nvl, "rt_addr", tmp);
2589
2590         nvlist_add_number(nvl, "rule", s->rule.ptr ? s->rule.ptr->nr : -1);
2591         nvlist_add_number(nvl, "anchor",
2592             s->anchor.ptr ? s->anchor.ptr->nr : -1);
2593         nvlist_add_number(nvl, "nat_rule",
2594             s->nat_rule.ptr ? s->nat_rule.ptr->nr : -1);
2595         nvlist_add_number(nvl, "creation", s->creation);
2596
2597         expire = pf_state_expires(s);
2598         if (expire <= time_uptime)
2599                 expire = 0;
2600         else
2601                 expire = expire - time_uptime;
2602         nvlist_add_number(nvl, "expire", expire);
2603
2604         for (int i = 0; i < 2; i++) {
2605                 nvlist_append_number_array(nvl, "packets",
2606                     counter_u64_fetch(s->packets[i]));
2607                 nvlist_append_number_array(nvl, "bytes",
2608                     counter_u64_fetch(s->bytes[i]));
2609         }
2610
2611         nvlist_add_number(nvl, "creatorid", s->creatorid);
2612         nvlist_add_number(nvl, "direction", s->direction);
2613         nvlist_add_number(nvl, "log", s->log);
2614         nvlist_add_number(nvl, "state_flags", s->state_flags);
2615         nvlist_add_number(nvl, "timeout", s->timeout);
2616         if (s->src_node)
2617                 flags |= PFSYNC_FLAG_SRCNODE;
2618         if (s->nat_src_node)
2619                 flags |= PFSYNC_FLAG_NATSRCNODE;
2620         nvlist_add_number(nvl, "sync_flags", flags);
2621
2622         return (nvl);
2623
2624 errout:
2625         nvlist_destroy(nvl);
2626         return (NULL);
2627 }
2628
2629 static int
2630 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2631     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2632     struct thread *td)
2633 {
2634         struct pf_kruleset      *ruleset;
2635         struct pf_krule         *tail;
2636         struct pf_kpooladdr     *pa;
2637         struct pfi_kkif         *kif = NULL;
2638         int                      rs_num;
2639         int                      error = 0;
2640
2641         if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2642                 error = EINVAL;
2643                 goto errout_unlocked;
2644         }
2645
2646 #define ERROUT(x)       ERROUT_FUNCTION(errout, x)
2647
2648         if (rule->ifname[0])
2649                 kif = pf_kkif_create(M_WAITOK);
2650         rule->evaluations = counter_u64_alloc(M_WAITOK);
2651         for (int i = 0; i < 2; i++) {
2652                 rule->packets[i] = counter_u64_alloc(M_WAITOK);
2653                 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2654         }
2655         rule->states_cur = counter_u64_alloc(M_WAITOK);
2656         rule->states_tot = counter_u64_alloc(M_WAITOK);
2657         rule->src_nodes = counter_u64_alloc(M_WAITOK);
2658         rule->cuid = td->td_ucred->cr_ruid;
2659         rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2660         TAILQ_INIT(&rule->rpool.list);
2661
2662         PF_RULES_WLOCK();
2663         ruleset = pf_find_kruleset(anchor);
2664         if (ruleset == NULL)
2665                 ERROUT(EINVAL);
2666         rs_num = pf_get_ruleset_number(rule->action);
2667         if (rs_num >= PF_RULESET_MAX)
2668                 ERROUT(EINVAL);
2669         if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2670                 DPFPRINTF(PF_DEBUG_MISC,
2671                     ("ticket: %d != [%d]%d\n", ticket, rs_num,
2672                     ruleset->rules[rs_num].inactive.ticket));
2673                 ERROUT(EBUSY);
2674         }
2675         if (pool_ticket != V_ticket_pabuf) {
2676                 DPFPRINTF(PF_DEBUG_MISC,
2677                     ("pool_ticket: %d != %d\n", pool_ticket,
2678                     V_ticket_pabuf));
2679                 ERROUT(EBUSY);
2680         }
2681
2682         tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2683             pf_krulequeue);
2684         if (tail)
2685                 rule->nr = tail->nr + 1;
2686         else
2687                 rule->nr = 0;
2688         if (rule->ifname[0]) {
2689                 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2690                 kif = NULL;
2691                 pfi_kkif_ref(rule->kif);
2692         } else
2693                 rule->kif = NULL;
2694
2695         if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2696                 error = EBUSY;
2697
2698 #ifdef ALTQ
2699         /* set queue IDs */
2700         if (rule->qname[0] != 0) {
2701                 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2702                         error = EBUSY;
2703                 else if (rule->pqname[0] != 0) {
2704                         if ((rule->pqid =
2705                             pf_qname2qid(rule->pqname)) == 0)
2706                                 error = EBUSY;
2707                 } else
2708                         rule->pqid = rule->qid;
2709         }
2710 #endif
2711         if (rule->tagname[0])
2712                 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2713                         error = EBUSY;
2714         if (rule->match_tagname[0])
2715                 if ((rule->match_tag =
2716                     pf_tagname2tag(rule->match_tagname)) == 0)
2717                         error = EBUSY;
2718         if (rule->rt && !rule->direction)
2719                 error = EINVAL;
2720         if (!rule->log)
2721                 rule->logif = 0;
2722         if (rule->logif >= PFLOGIFS_MAX)
2723                 error = EINVAL;
2724         if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2725                 error = ENOMEM;
2726         if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2727                 error = ENOMEM;
2728         if (pf_kanchor_setup(rule, ruleset, anchor_call))
2729                 error = EINVAL;
2730         if (rule->scrub_flags & PFSTATE_SETPRIO &&
2731             (rule->set_prio[0] > PF_PRIO_MAX ||
2732             rule->set_prio[1] > PF_PRIO_MAX))
2733                 error = EINVAL;
2734         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2735                 if (pa->addr.type == PF_ADDR_TABLE) {
2736                         pa->addr.p.tbl = pfr_attach_table(ruleset,
2737                             pa->addr.v.tblname);
2738                         if (pa->addr.p.tbl == NULL)
2739                                 error = ENOMEM;
2740                 }
2741
2742         rule->overload_tbl = NULL;
2743         if (rule->overload_tblname[0]) {
2744                 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2745                     rule->overload_tblname)) == NULL)
2746                         error = EINVAL;
2747                 else
2748                         rule->overload_tbl->pfrkt_flags |=
2749                             PFR_TFLAG_ACTIVE;
2750         }
2751
2752         pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2753         if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2754             (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2755             (rule->rt > PF_NOPFROUTE)) &&
2756             (TAILQ_FIRST(&rule->rpool.list) == NULL))
2757                 error = EINVAL;
2758
2759         if (error) {
2760                 pf_free_rule(rule);
2761                 rule = NULL;
2762                 ERROUT(error);
2763         }
2764
2765         rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2766         counter_u64_zero(rule->evaluations);
2767         for (int i = 0; i < 2; i++) {
2768                 counter_u64_zero(rule->packets[i]);
2769                 counter_u64_zero(rule->bytes[i]);
2770         }
2771         TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2772             rule, entries);
2773         ruleset->rules[rs_num].inactive.rcount++;
2774         PF_RULES_WUNLOCK();
2775
2776         return (0);
2777
2778 #undef ERROUT
2779 errout:
2780         PF_RULES_WUNLOCK();
2781 errout_unlocked:
2782         pf_kkif_free(kif);
2783         pf_krule_free(rule);
2784         return (error);
2785 }
2786
2787 static bool
2788 pf_label_match(const struct pf_krule *rule, const char *label)
2789 {
2790         int i = 0;
2791
2792         while (*rule->label[i]) {
2793                 if (strcmp(rule->label[i], label) == 0)
2794                         return (true);
2795                 i++;
2796         }
2797
2798         return (false);
2799 }
2800
2801 static unsigned int
2802 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2803 {
2804         struct pf_state *match;
2805         int more = 0;
2806         unsigned int killed = 0;
2807
2808         /* Call with unlocked hashrow */
2809
2810         match = pf_find_state_all(key, dir, &more);
2811         if (match && !more) {
2812                 pf_unlink_state(match, 0);
2813                 killed++;
2814         }
2815
2816         return (killed);
2817 }
2818
2819 static int
2820 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2821 {
2822         struct pf_state         *s;
2823         struct pf_state_key     *sk;
2824         struct pf_addr          *srcaddr, *dstaddr;
2825         struct pf_state_key_cmp  match_key;
2826         int                      idx, killed = 0;
2827         unsigned int             dir;
2828         u_int16_t                srcport, dstport;
2829
2830 relock_DIOCKILLSTATES:
2831         PF_HASHROW_LOCK(ih);
2832         LIST_FOREACH(s, &ih->states, entry) {
2833                 sk = s->key[PF_SK_WIRE];
2834                 if (s->direction == PF_OUT) {
2835                         srcaddr = &sk->addr[1];
2836                         dstaddr = &sk->addr[0];
2837                         srcport = sk->port[1];
2838                         dstport = sk->port[0];
2839                 } else {
2840                         srcaddr = &sk->addr[0];
2841                         dstaddr = &sk->addr[1];
2842                         srcport = sk->port[0];
2843                         dstport = sk->port[1];
2844                 }
2845
2846                 if (psk->psk_af && sk->af != psk->psk_af)
2847                         continue;
2848
2849                 if (psk->psk_proto && psk->psk_proto != sk->proto)
2850                         continue;
2851
2852                 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2853                     &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2854                         continue;
2855
2856                 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2857                     &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2858                         continue;
2859
2860                 if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2861                     &psk->psk_rt_addr.addr.v.a.addr,
2862                     &psk->psk_rt_addr.addr.v.a.mask,
2863                     &s->rt_addr, sk->af))
2864                         continue;
2865
2866                 if (psk->psk_src.port_op != 0 &&
2867                     ! pf_match_port(psk->psk_src.port_op,
2868                     psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2869                         continue;
2870
2871                 if (psk->psk_dst.port_op != 0 &&
2872                     ! pf_match_port(psk->psk_dst.port_op,
2873                     psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2874                         continue;
2875
2876                 if (psk->psk_label[0] &&
2877                     ! pf_label_match(s->rule.ptr, psk->psk_label))
2878                         continue;
2879
2880                 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2881                     s->kif->pfik_name))
2882                         continue;
2883
2884                 if (psk->psk_kill_match) {
2885                         /* Create the key to find matching states, with lock
2886                          * held. */
2887
2888                         bzero(&match_key, sizeof(match_key));
2889
2890                         if (s->direction == PF_OUT) {
2891                                 dir = PF_IN;
2892                                 idx = PF_SK_STACK;
2893                         } else {
2894                                 dir = PF_OUT;
2895                                 idx = PF_SK_WIRE;
2896                         }
2897
2898                         match_key.af = s->key[idx]->af;
2899                         match_key.proto = s->key[idx]->proto;
2900                         PF_ACPY(&match_key.addr[0],
2901                             &s->key[idx]->addr[1], match_key.af);
2902                         match_key.port[0] = s->key[idx]->port[1];
2903                         PF_ACPY(&match_key.addr[1],
2904                             &s->key[idx]->addr[0], match_key.af);
2905                         match_key.port[1] = s->key[idx]->port[0];
2906                 }
2907
2908                 pf_unlink_state(s, PF_ENTER_LOCKED);
2909                 killed++;
2910
2911                 if (psk->psk_kill_match)
2912                         killed += pf_kill_matching_state(&match_key, dir);
2913
2914                 goto relock_DIOCKILLSTATES;
2915         }
2916         PF_HASHROW_UNLOCK(ih);
2917
2918         return (killed);
2919 }
2920
2921 static int
2922 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2923 {
2924         int                      error = 0;
2925         PF_RULES_RLOCK_TRACKER;
2926
2927 #define ERROUT_IOCTL(target, x)                                 \
2928     do {                                                                \
2929             error = (x);                                                \
2930             SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);  \
2931             goto target;                                                \
2932     } while (0)
2933
2934
2935         /* XXX keep in sync with switch() below */
2936         if (securelevel_gt(td->td_ucred, 2))
2937                 switch (cmd) {
2938                 case DIOCGETRULES:
2939                 case DIOCGETRULE:
2940                 case DIOCGETRULENV:
2941                 case DIOCGETADDRS:
2942                 case DIOCGETADDR:
2943                 case DIOCGETSTATE:
2944                 case DIOCGETSTATENV:
2945                 case DIOCSETSTATUSIF:
2946                 case DIOCGETSTATUS:
2947                 case DIOCCLRSTATUS:
2948                 case DIOCNATLOOK:
2949                 case DIOCSETDEBUG:
2950                 case DIOCGETSTATES:
2951                 case DIOCGETTIMEOUT:
2952                 case DIOCCLRRULECTRS:
2953                 case DIOCGETLIMIT:
2954                 case DIOCGETALTQSV0:
2955                 case DIOCGETALTQSV1:
2956                 case DIOCGETALTQV0:
2957                 case DIOCGETALTQV1:
2958                 case DIOCGETQSTATSV0:
2959                 case DIOCGETQSTATSV1:
2960                 case DIOCGETRULESETS:
2961                 case DIOCGETRULESET:
2962                 case DIOCRGETTABLES:
2963                 case DIOCRGETTSTATS:
2964                 case DIOCRCLRTSTATS:
2965                 case DIOCRCLRADDRS:
2966                 case DIOCRADDADDRS:
2967                 case DIOCRDELADDRS:
2968                 case DIOCRSETADDRS:
2969                 case DIOCRGETADDRS:
2970                 case DIOCRGETASTATS:
2971                 case DIOCRCLRASTATS:
2972                 case DIOCRTSTADDRS:
2973                 case DIOCOSFPGET:
2974                 case DIOCGETSRCNODES:
2975                 case DIOCCLRSRCNODES:
2976                 case DIOCIGETIFACES:
2977                 case DIOCGIFSPEEDV0:
2978                 case DIOCGIFSPEEDV1:
2979                 case DIOCSETIFFLAG:
2980                 case DIOCCLRIFFLAG:
2981                         break;
2982                 case DIOCRCLRTABLES:
2983                 case DIOCRADDTABLES:
2984                 case DIOCRDELTABLES:
2985                 case DIOCRSETTFLAGS:
2986                         if (((struct pfioc_table *)addr)->pfrio_flags &
2987                             PFR_FLAG_DUMMY)
2988                                 break; /* dummy operation ok */
2989                         return (EPERM);
2990                 default:
2991                         return (EPERM);
2992                 }
2993
2994         if (!(flags & FWRITE))
2995                 switch (cmd) {
2996                 case DIOCGETRULES:
2997                 case DIOCGETADDRS:
2998                 case DIOCGETADDR:
2999                 case DIOCGETSTATE:
3000                 case DIOCGETSTATENV:
3001                 case DIOCGETSTATUS:
3002                 case DIOCGETSTATES:
3003                 case DIOCGETTIMEOUT:
3004                 case DIOCGETLIMIT:
3005                 case DIOCGETALTQSV0:
3006                 case DIOCGETALTQSV1:
3007                 case DIOCGETALTQV0:
3008                 case DIOCGETALTQV1:
3009                 case DIOCGETQSTATSV0:
3010                 case DIOCGETQSTATSV1:
3011                 case DIOCGETRULESETS:
3012                 case DIOCGETRULESET:
3013                 case DIOCNATLOOK:
3014                 case DIOCRGETTABLES:
3015                 case DIOCRGETTSTATS:
3016                 case DIOCRGETADDRS:
3017                 case DIOCRGETASTATS:
3018                 case DIOCRTSTADDRS:
3019                 case DIOCOSFPGET:
3020                 case DIOCGETSRCNODES:
3021                 case DIOCIGETIFACES:
3022                 case DIOCGIFSPEEDV1:
3023                 case DIOCGIFSPEEDV0:
3024                 case DIOCGETRULENV:
3025                         break;
3026                 case DIOCRCLRTABLES:
3027                 case DIOCRADDTABLES:
3028                 case DIOCRDELTABLES:
3029                 case DIOCRCLRTSTATS:
3030                 case DIOCRCLRADDRS:
3031                 case DIOCRADDADDRS:
3032                 case DIOCRDELADDRS:
3033                 case DIOCRSETADDRS:
3034                 case DIOCRSETTFLAGS:
3035                         if (((struct pfioc_table *)addr)->pfrio_flags &
3036                             PFR_FLAG_DUMMY) {
3037                                 flags |= FWRITE; /* need write lock for dummy */
3038                                 break; /* dummy operation ok */
3039                         }
3040                         return (EACCES);
3041                 case DIOCGETRULE:
3042                         if (((struct pfioc_rule *)addr)->action ==
3043                             PF_GET_CLR_CNTR)
3044                                 return (EACCES);
3045                         break;
3046                 default:
3047                         return (EACCES);
3048                 }
3049
3050         CURVNET_SET(TD_TO_VNET(td));
3051
3052         switch (cmd) {
3053         case DIOCSTART:
3054                 sx_xlock(&pf_ioctl_lock);
3055                 if (V_pf_status.running)
3056                         error = EEXIST;
3057                 else {
3058                         int cpu;
3059
3060                         hook_pf();
3061                         V_pf_status.running = 1;
3062                         V_pf_status.since = time_second;
3063
3064                         CPU_FOREACH(cpu)
3065                                 V_pf_stateid[cpu] = time_second;
3066
3067                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
3068                 }
3069                 break;
3070
3071         case DIOCSTOP:
3072                 sx_xlock(&pf_ioctl_lock);
3073                 if (!V_pf_status.running)
3074                         error = ENOENT;
3075                 else {
3076                         V_pf_status.running = 0;
3077                         dehook_pf();
3078                         V_pf_status.since = time_second;
3079                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
3080                 }
3081                 break;
3082
3083         case DIOCADDRULENV: {
3084                 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3085                 nvlist_t        *nvl = NULL;
3086                 void            *nvlpacked = NULL;
3087                 struct pf_krule *rule = NULL;
3088                 const char      *anchor = "", *anchor_call = "";
3089                 uint32_t         ticket = 0, pool_ticket = 0;
3090
3091 #define ERROUT(x)       ERROUT_IOCTL(DIOCADDRULENV_error, x)
3092
3093                 if (nv->len > pf_ioctl_maxcount)
3094                         ERROUT(ENOMEM);
3095
3096                 nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
3097                 error = copyin(nv->data, nvlpacked, nv->len);
3098                 if (error)
3099                         ERROUT(error);
3100
3101                 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3102                 if (nvl == NULL)
3103                         ERROUT(EBADMSG);
3104
3105                 if (! nvlist_exists_number(nvl, "ticket"))
3106                         ERROUT(EINVAL);
3107                 ticket = nvlist_get_number(nvl, "ticket");
3108
3109                 if (! nvlist_exists_number(nvl, "pool_ticket"))
3110                         ERROUT(EINVAL);
3111                 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3112
3113                 if (! nvlist_exists_nvlist(nvl, "rule"))
3114                         ERROUT(EINVAL);
3115
3116                 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3117                     &rule);
3118                 if (error)
3119                         ERROUT(error);
3120
3121                 if (nvlist_exists_string(nvl, "anchor"))
3122                         anchor = nvlist_get_string(nvl, "anchor");
3123                 if (nvlist_exists_string(nvl, "anchor_call"))
3124                         anchor_call = nvlist_get_string(nvl, "anchor_call");
3125
3126                 if ((error = nvlist_error(nvl)))
3127                         ERROUT(error);
3128
3129                 /* Frees rule on error */
3130                 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3131                     anchor_call, td);
3132
3133                 nvlist_destroy(nvl);
3134                 free(nvlpacked, M_TEMP);
3135                 break;
3136 #undef ERROUT
3137 DIOCADDRULENV_error:
3138                 pf_krule_free(rule);
3139                 nvlist_destroy(nvl);
3140                 free(nvlpacked, M_TEMP);
3141
3142                 break;
3143         }
3144         case DIOCADDRULE: {
3145                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
3146                 struct pf_krule         *rule;
3147
3148                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3149                 error = pf_rule_to_krule(&pr->rule, rule);
3150                 if (error != 0) {
3151                         free(rule, M_PFRULE);
3152                         break;
3153                 }
3154
3155                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3156
3157                 /* Frees rule on error */
3158                 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3159                     pr->anchor, pr->anchor_call, td);
3160                 break;
3161         }
3162
3163         case DIOCGETRULES: {
3164                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
3165                 struct pf_kruleset      *ruleset;
3166                 struct pf_krule         *tail;
3167                 int                      rs_num;
3168
3169                 PF_RULES_WLOCK();
3170                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3171                 ruleset = pf_find_kruleset(pr->anchor);
3172                 if (ruleset == NULL) {
3173                         PF_RULES_WUNLOCK();
3174                         error = EINVAL;
3175                         break;
3176                 }
3177                 rs_num = pf_get_ruleset_number(pr->rule.action);
3178                 if (rs_num >= PF_RULESET_MAX) {
3179                         PF_RULES_WUNLOCK();
3180                         error = EINVAL;
3181                         break;
3182                 }
3183                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
3184                     pf_krulequeue);
3185                 if (tail)
3186                         pr->nr = tail->nr + 1;
3187                 else
3188                         pr->nr = 0;
3189                 pr->ticket = ruleset->rules[rs_num].active.ticket;
3190                 PF_RULES_WUNLOCK();
3191                 break;
3192         }
3193
3194         case DIOCGETRULE: {
3195                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
3196                 struct pf_kruleset      *ruleset;
3197                 struct pf_krule         *rule;
3198                 int                      rs_num;
3199
3200                 PF_RULES_WLOCK();
3201                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3202                 ruleset = pf_find_kruleset(pr->anchor);
3203                 if (ruleset == NULL) {
3204                         PF_RULES_WUNLOCK();
3205                         error = EINVAL;
3206                         break;
3207                 }
3208                 rs_num = pf_get_ruleset_number(pr->rule.action);
3209                 if (rs_num >= PF_RULESET_MAX) {
3210                         PF_RULES_WUNLOCK();
3211                         error = EINVAL;
3212                         break;
3213                 }
3214                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
3215                         PF_RULES_WUNLOCK();
3216                         error = EBUSY;
3217                         break;
3218                 }
3219                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3220                 while ((rule != NULL) && (rule->nr != pr->nr))
3221                         rule = TAILQ_NEXT(rule, entries);
3222                 if (rule == NULL) {
3223                         PF_RULES_WUNLOCK();
3224                         error = EBUSY;
3225                         break;
3226                 }
3227
3228                 pf_krule_to_rule(rule, &pr->rule);
3229
3230                 if (pf_kanchor_copyout(ruleset, rule, pr)) {
3231                         PF_RULES_WUNLOCK();
3232                         error = EBUSY;
3233                         break;
3234                 }
3235                 pf_addr_copyout(&pr->rule.src.addr);
3236                 pf_addr_copyout(&pr->rule.dst.addr);
3237
3238                 if (pr->action == PF_GET_CLR_CNTR) {
3239                         counter_u64_zero(rule->evaluations);
3240                         for (int i = 0; i < 2; i++) {
3241                                 counter_u64_zero(rule->packets[i]);
3242                                 counter_u64_zero(rule->bytes[i]);
3243                         }
3244                         counter_u64_zero(rule->states_tot);
3245                 }
3246                 PF_RULES_WUNLOCK();
3247                 break;
3248         }
3249
3250         case DIOCGETRULENV: {
3251                 struct pfioc_nv         *nv = (struct pfioc_nv *)addr;
3252                 nvlist_t                *nvrule = NULL;
3253                 nvlist_t                *nvl = NULL;
3254                 struct pf_kruleset      *ruleset;
3255                 struct pf_krule         *rule;
3256                 void                    *nvlpacked = NULL;
3257                 int                      rs_num, nr;
3258                 bool                     clear_counter = false;
3259
3260 #define ERROUT(x)       ERROUT_IOCTL(DIOCGETRULENV_error, x)
3261
3262                 if (nv->len > pf_ioctl_maxcount)
3263                         ERROUT(ENOMEM);
3264
3265                 /* Copy the request in */
3266                 nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
3267                 if (nvlpacked == NULL)
3268                         ERROUT(ENOMEM);
3269
3270                 error = copyin(nv->data, nvlpacked, nv->len);
3271                 if (error)
3272                         ERROUT(error);
3273
3274                 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3275                 if (nvl == NULL)
3276                         ERROUT(EBADMSG);
3277
3278                 if (! nvlist_exists_string(nvl, "anchor"))
3279                         ERROUT(EBADMSG);
3280                 if (! nvlist_exists_number(nvl, "ruleset"))
3281                         ERROUT(EBADMSG);
3282                 if (! nvlist_exists_number(nvl, "ticket"))
3283                         ERROUT(EBADMSG);
3284                 if (! nvlist_exists_number(nvl, "nr"))
3285                         ERROUT(EBADMSG);
3286
3287                 if (nvlist_exists_bool(nvl, "clear_counter"))
3288                         clear_counter = nvlist_get_bool(nvl, "clear_counter");
3289
3290                 if (clear_counter && !(flags & FWRITE))
3291                         ERROUT(EACCES);
3292
3293                 nr = nvlist_get_number(nvl, "nr");
3294
3295                 PF_RULES_WLOCK();
3296                 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3297                 if (ruleset == NULL) {
3298                         PF_RULES_WUNLOCK();
3299                         ERROUT(ENOENT);
3300                 }
3301
3302                 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3303                 if (rs_num >= PF_RULESET_MAX) {
3304                         PF_RULES_WUNLOCK();
3305                         ERROUT(EINVAL);
3306                 }
3307
3308                 if (nvlist_get_number(nvl, "ticket") !=
3309                     ruleset->rules[rs_num].active.ticket) {
3310                         PF_RULES_WUNLOCK();
3311                         ERROUT(EBUSY);
3312                         break;
3313                 }
3314
3315                 if ((error = nvlist_error(nvl))) {
3316                         PF_RULES_WUNLOCK();
3317                         ERROUT(error);
3318                 }
3319
3320                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3321                 while ((rule != NULL) && (rule->nr != nr))
3322                         rule = TAILQ_NEXT(rule, entries);
3323                 if (rule == NULL) {
3324                         PF_RULES_WUNLOCK();
3325                         ERROUT(EBUSY);
3326                         break;
3327                 }
3328
3329                 nvrule = pf_krule_to_nvrule(rule);
3330
3331                 nvlist_destroy(nvl);
3332                 nvl = nvlist_create(0);
3333                 if (nvl == NULL) {
3334                         PF_RULES_WUNLOCK();
3335                         ERROUT(ENOMEM);
3336                 }
3337                 nvlist_add_number(nvl, "nr", nr);
3338                 nvlist_add_nvlist(nvl, "rule", nvrule);
3339                 nvrule = NULL;
3340                 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3341                         PF_RULES_WUNLOCK();
3342                         ERROUT(EBUSY);
3343                 }
3344
3345                 free(nvlpacked, M_TEMP);
3346                 nvlpacked = nvlist_pack(nvl, &nv->len);
3347                 if (nvlpacked == NULL) {
3348                         PF_RULES_WUNLOCK();
3349                         ERROUT(ENOMEM);
3350                 }
3351
3352                 if (nv->size == 0) {
3353                         PF_RULES_WUNLOCK();
3354                         ERROUT(0);
3355                 }
3356                 else if (nv->size < nv->len) {
3357                         PF_RULES_WUNLOCK();
3358                         ERROUT(ENOSPC);
3359                 }
3360
3361                 error = copyout(nvlpacked, nv->data, nv->len);
3362
3363                 if (clear_counter) {
3364                         counter_u64_zero(rule->evaluations);
3365                         for (int i = 0; i < 2; i++) {
3366                                 counter_u64_zero(rule->packets[i]);
3367                                 counter_u64_zero(rule->bytes[i]);
3368                         }
3369                         counter_u64_zero(rule->states_tot);
3370                 }
3371                 PF_RULES_WUNLOCK();
3372
3373 #undef ERROUT
3374 DIOCGETRULENV_error:
3375                 free(nvlpacked, M_TEMP);
3376                 nvlist_destroy(nvrule);
3377                 nvlist_destroy(nvl);
3378
3379                 break;
3380         }
3381
3382         case DIOCCHANGERULE: {
3383                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
3384                 struct pf_kruleset      *ruleset;
3385                 struct pf_krule         *oldrule = NULL, *newrule = NULL;
3386                 struct pfi_kkif         *kif = NULL;
3387                 struct pf_kpooladdr     *pa;
3388                 u_int32_t                nr = 0;
3389                 int                      rs_num;
3390
3391                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3392                     pcr->action > PF_CHANGE_GET_TICKET) {
3393                         error = EINVAL;
3394                         break;
3395                 }
3396                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3397                         error = EINVAL;
3398                         break;
3399                 }
3400
3401                 if (pcr->action != PF_CHANGE_REMOVE) {
3402                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
3403                         error = pf_rule_to_krule(&pcr->rule, newrule);
3404                         if (error != 0) {
3405                                 free(newrule, M_PFRULE);
3406                                 break;
3407                         }
3408
3409                         if (newrule->ifname[0])
3410                                 kif = pf_kkif_create(M_WAITOK);
3411                         newrule->evaluations = counter_u64_alloc(M_WAITOK);
3412                         for (int i = 0; i < 2; i++) {
3413                                 newrule->packets[i] =
3414                                     counter_u64_alloc(M_WAITOK);
3415                                 newrule->bytes[i] =
3416                                     counter_u64_alloc(M_WAITOK);
3417                         }
3418                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
3419                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
3420                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3421                         newrule->cuid = td->td_ucred->cr_ruid;
3422                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3423                         TAILQ_INIT(&newrule->rpool.list);
3424                 }
3425 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
3426
3427                 PF_RULES_WLOCK();
3428                 if (!(pcr->action == PF_CHANGE_REMOVE ||
3429                     pcr->action == PF_CHANGE_GET_TICKET) &&
3430                     pcr->pool_ticket != V_ticket_pabuf)
3431                         ERROUT(EBUSY);
3432
3433                 ruleset = pf_find_kruleset(pcr->anchor);
3434                 if (ruleset == NULL)
3435                         ERROUT(EINVAL);
3436
3437                 rs_num = pf_get_ruleset_number(pcr->rule.action);
3438                 if (rs_num >= PF_RULESET_MAX)
3439                         ERROUT(EINVAL);
3440
3441                 if (pcr->action == PF_CHANGE_GET_TICKET) {
3442                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3443                         ERROUT(0);
3444                 } else if (pcr->ticket !=
3445                             ruleset->rules[rs_num].active.ticket)
3446                                 ERROUT(EINVAL);
3447
3448                 if (pcr->action != PF_CHANGE_REMOVE) {
3449                         if (newrule->ifname[0]) {
3450                                 newrule->kif = pfi_kkif_attach(kif,
3451                                     newrule->ifname);
3452                                 kif = NULL;
3453                                 pfi_kkif_ref(newrule->kif);
3454                         } else
3455                                 newrule->kif = NULL;
3456
3457                         if (newrule->rtableid > 0 &&
3458                             newrule->rtableid >= rt_numfibs)
3459                                 error = EBUSY;
3460
3461 #ifdef ALTQ
3462                         /* set queue IDs */
3463                         if (newrule->qname[0] != 0) {
3464                                 if ((newrule->qid =
3465                                     pf_qname2qid(newrule->qname)) == 0)
3466                                         error = EBUSY;
3467                                 else if (newrule->pqname[0] != 0) {
3468                                         if ((newrule->pqid =
3469                                             pf_qname2qid(newrule->pqname)) == 0)
3470                                                 error = EBUSY;
3471                                 } else
3472                                         newrule->pqid = newrule->qid;
3473                         }
3474 #endif /* ALTQ */
3475                         if (newrule->tagname[0])
3476                                 if ((newrule->tag =
3477                                     pf_tagname2tag(newrule->tagname)) == 0)
3478                                         error = EBUSY;
3479                         if (newrule->match_tagname[0])
3480                                 if ((newrule->match_tag = pf_tagname2tag(
3481                                     newrule->match_tagname)) == 0)
3482                                         error = EBUSY;
3483                         if (newrule->rt && !newrule->direction)
3484                                 error = EINVAL;
3485                         if (!newrule->log)
3486                                 newrule->logif = 0;
3487                         if (newrule->logif >= PFLOGIFS_MAX)
3488                                 error = EINVAL;
3489                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3490                                 error = ENOMEM;
3491                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3492                                 error = ENOMEM;
3493                         if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3494                                 error = EINVAL;
3495                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3496                                 if (pa->addr.type == PF_ADDR_TABLE) {
3497                                         pa->addr.p.tbl =
3498                                             pfr_attach_table(ruleset,
3499                                             pa->addr.v.tblname);
3500                                         if (pa->addr.p.tbl == NULL)
3501                                                 error = ENOMEM;
3502                                 }
3503
3504                         newrule->overload_tbl = NULL;
3505                         if (newrule->overload_tblname[0]) {
3506                                 if ((newrule->overload_tbl = pfr_attach_table(
3507                                     ruleset, newrule->overload_tblname)) ==
3508                                     NULL)
3509                                         error = EINVAL;
3510                                 else
3511                                         newrule->overload_tbl->pfrkt_flags |=
3512                                             PFR_TFLAG_ACTIVE;
3513                         }
3514
3515                         pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3516                         if (((((newrule->action == PF_NAT) ||
3517                             (newrule->action == PF_RDR) ||
3518                             (newrule->action == PF_BINAT) ||
3519                             (newrule->rt > PF_NOPFROUTE)) &&
3520                             !newrule->anchor)) &&
3521                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3522                                 error = EINVAL;
3523
3524                         if (error) {
3525                                 pf_free_rule(newrule);
3526                                 PF_RULES_WUNLOCK();
3527                                 break;
3528                         }
3529
3530                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3531                 }
3532                 pf_empty_kpool(&V_pf_pabuf);
3533
3534                 if (pcr->action == PF_CHANGE_ADD_HEAD)
3535                         oldrule = TAILQ_FIRST(
3536                             ruleset->rules[rs_num].active.ptr);
3537                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3538                         oldrule = TAILQ_LAST(
3539                             ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3540                 else {
3541                         oldrule = TAILQ_FIRST(
3542                             ruleset->rules[rs_num].active.ptr);
3543                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3544                                 oldrule = TAILQ_NEXT(oldrule, entries);
3545                         if (oldrule == NULL) {
3546                                 if (newrule != NULL)
3547                                         pf_free_rule(newrule);
3548                                 PF_RULES_WUNLOCK();
3549                                 error = EINVAL;
3550                                 break;
3551                         }
3552                 }
3553
3554                 if (pcr->action == PF_CHANGE_REMOVE) {
3555                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3556                             oldrule);
3557                         ruleset->rules[rs_num].active.rcount--;
3558                 } else {
3559                         if (oldrule == NULL)
3560                                 TAILQ_INSERT_TAIL(
3561                                     ruleset->rules[rs_num].active.ptr,
3562                                     newrule, entries);
3563                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3564                             pcr->action == PF_CHANGE_ADD_BEFORE)
3565                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3566                         else
3567                                 TAILQ_INSERT_AFTER(
3568                                     ruleset->rules[rs_num].active.ptr,
3569                                     oldrule, newrule, entries);
3570                         ruleset->rules[rs_num].active.rcount++;
3571                 }
3572
3573                 nr = 0;
3574                 TAILQ_FOREACH(oldrule,
3575                     ruleset->rules[rs_num].active.ptr, entries)
3576                         oldrule->nr = nr++;
3577
3578                 ruleset->rules[rs_num].active.ticket++;
3579
3580                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3581                 pf_remove_if_empty_kruleset(ruleset);
3582
3583                 PF_RULES_WUNLOCK();
3584                 break;
3585
3586 #undef ERROUT
3587 DIOCCHANGERULE_error:
3588                 PF_RULES_WUNLOCK();
3589                 pf_krule_free(newrule);
3590                 pf_kkif_free(kif);
3591                 break;
3592         }
3593
3594         case DIOCCLRSTATES: {
3595                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3596                 struct pf_kstate_kill    kill;
3597
3598                 error = pf_state_kill_to_kstate_kill(psk, &kill);
3599                 if (error)
3600                         break;
3601
3602                 psk->psk_killed = pf_clear_states(&kill);
3603                 break;
3604         }
3605
3606         case DIOCCLRSTATESNV: {
3607                 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3608                 break;
3609         }
3610
3611         case DIOCKILLSTATES: {
3612                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3613                 struct pf_kstate_kill    kill;
3614
3615                 error = pf_state_kill_to_kstate_kill(psk, &kill);
3616                 if (error)
3617                         break;
3618
3619                 psk->psk_killed = 0;
3620                 error = pf_killstates(&kill, &psk->psk_killed);
3621                 break;
3622         }
3623
3624         case DIOCKILLSTATESNV: {
3625                 error = pf_killstates_nv((struct pfioc_nv *)addr);
3626                 break;
3627         }
3628
3629         case DIOCADDSTATE: {
3630                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
3631                 struct pfsync_state     *sp = &ps->state;
3632
3633                 if (sp->timeout >= PFTM_MAX) {
3634                         error = EINVAL;
3635                         break;
3636                 }
3637                 if (V_pfsync_state_import_ptr != NULL) {
3638                         PF_RULES_RLOCK();
3639                         error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
3640                         PF_RULES_RUNLOCK();
3641                 } else
3642                         error = EOPNOTSUPP;
3643                 break;
3644         }
3645
3646         case DIOCGETSTATE: {
3647                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
3648                 struct pf_state         *s;
3649
3650                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3651                 if (s == NULL) {
3652                         error = ENOENT;
3653                         break;
3654                 }
3655
3656                 pfsync_state_export(&ps->state, s);
3657                 PF_STATE_UNLOCK(s);
3658                 break;
3659         }
3660
3661         case DIOCGETSTATENV: {
3662                 error = pf_getstate((struct pfioc_nv *)addr);
3663                 break;
3664         }
3665
3666         case DIOCGETSTATES: {
3667                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
3668                 struct pf_state         *s;
3669                 struct pfsync_state     *pstore, *p;
3670                 int i, nr;
3671
3672                 if (ps->ps_len <= 0) {
3673                         nr = uma_zone_get_cur(V_pf_state_z);
3674                         ps->ps_len = sizeof(struct pfsync_state) * nr;
3675                         break;
3676                 }
3677
3678                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
3679                 nr = 0;
3680
3681                 for (i = 0; i <= pf_hashmask; i++) {
3682                         struct pf_idhash *ih = &V_pf_idhash[i];
3683
3684                         PF_HASHROW_LOCK(ih);
3685                         LIST_FOREACH(s, &ih->states, entry) {
3686                                 if (s->timeout == PFTM_UNLINKED)
3687                                         continue;
3688
3689                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
3690                                         PF_HASHROW_UNLOCK(ih);
3691                                         goto DIOCGETSTATES_full;
3692                                 }
3693                                 pfsync_state_export(p, s);
3694                                 p++;
3695                                 nr++;
3696                         }
3697                         PF_HASHROW_UNLOCK(ih);
3698                 }
3699 DIOCGETSTATES_full:
3700                 error = copyout(pstore, ps->ps_states,
3701                     sizeof(struct pfsync_state) * nr);
3702                 if (error) {
3703                         free(pstore, M_TEMP);
3704                         break;
3705                 }
3706                 ps->ps_len = sizeof(struct pfsync_state) * nr;
3707                 free(pstore, M_TEMP);
3708
3709                 break;
3710         }
3711
3712         case DIOCGETSTATUS: {
3713                 struct pf_status *s = (struct pf_status *)addr;
3714
3715                 PF_RULES_RLOCK();
3716                 s->running = V_pf_status.running;
3717                 s->since   = V_pf_status.since;
3718                 s->debug   = V_pf_status.debug;
3719                 s->hostid  = V_pf_status.hostid;
3720                 s->states  = V_pf_status.states;
3721                 s->src_nodes = V_pf_status.src_nodes;
3722
3723                 for (int i = 0; i < PFRES_MAX; i++)
3724                         s->counters[i] =
3725                             counter_u64_fetch(V_pf_status.counters[i]);
3726                 for (int i = 0; i < LCNT_MAX; i++)
3727                         s->lcounters[i] =
3728                             counter_u64_fetch(V_pf_status.lcounters[i]);
3729                 for (int i = 0; i < FCNT_MAX; i++)
3730                         s->fcounters[i] =
3731                             counter_u64_fetch(V_pf_status.fcounters[i]);
3732                 for (int i = 0; i < SCNT_MAX; i++)
3733                         s->scounters[i] =
3734                             counter_u64_fetch(V_pf_status.scounters[i]);
3735
3736                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3737                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3738                     PF_MD5_DIGEST_LENGTH);
3739
3740                 pfi_update_status(s->ifname, s);
3741                 PF_RULES_RUNLOCK();
3742                 break;
3743         }
3744
3745         case DIOCSETSTATUSIF: {
3746                 struct pfioc_if *pi = (struct pfioc_if *)addr;
3747
3748                 if (pi->ifname[0] == 0) {
3749                         bzero(V_pf_status.ifname, IFNAMSIZ);
3750                         break;
3751                 }
3752                 PF_RULES_WLOCK();
3753                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3754                 PF_RULES_WUNLOCK();
3755                 break;
3756         }
3757
3758         case DIOCCLRSTATUS: {
3759                 PF_RULES_WLOCK();
3760                 for (int i = 0; i < PFRES_MAX; i++)
3761                         counter_u64_zero(V_pf_status.counters[i]);
3762                 for (int i = 0; i < FCNT_MAX; i++)
3763                         counter_u64_zero(V_pf_status.fcounters[i]);
3764                 for (int i = 0; i < SCNT_MAX; i++)
3765                         counter_u64_zero(V_pf_status.scounters[i]);
3766                 for (int i = 0; i < LCNT_MAX; i++)
3767                         counter_u64_zero(V_pf_status.lcounters[i]);
3768                 V_pf_status.since = time_second;
3769                 if (*V_pf_status.ifname)
3770                         pfi_update_status(V_pf_status.ifname, NULL);
3771                 PF_RULES_WUNLOCK();
3772                 break;
3773         }
3774
3775         case DIOCNATLOOK: {
3776                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
3777                 struct pf_state_key     *sk;
3778                 struct pf_state         *state;
3779                 struct pf_state_key_cmp  key;
3780                 int                      m = 0, direction = pnl->direction;
3781                 int                      sidx, didx;
3782
3783                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
3784                 sidx = (direction == PF_IN) ? 1 : 0;
3785                 didx = (direction == PF_IN) ? 0 : 1;
3786
3787                 if (!pnl->proto ||
3788                     PF_AZERO(&pnl->saddr, pnl->af) ||
3789                     PF_AZERO(&pnl->daddr, pnl->af) ||
3790                     ((pnl->proto == IPPROTO_TCP ||
3791                     pnl->proto == IPPROTO_UDP) &&
3792                     (!pnl->dport || !pnl->sport)))
3793                         error = EINVAL;
3794                 else {
3795                         bzero(&key, sizeof(key));
3796                         key.af = pnl->af;
3797                         key.proto = pnl->proto;
3798                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3799                         key.port[sidx] = pnl->sport;
3800                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3801                         key.port[didx] = pnl->dport;
3802
3803                         state = pf_find_state_all(&key, direction, &m);
3804
3805                         if (m > 1)
3806                                 error = E2BIG;  /* more than one state */
3807                         else if (state != NULL) {
3808                                 /* XXXGL: not locked read */
3809                                 sk = state->key[sidx];
3810                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3811                                 pnl->rsport = sk->port[sidx];
3812                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3813                                 pnl->rdport = sk->port[didx];
3814                         } else
3815                                 error = ENOENT;
3816                 }
3817                 break;
3818         }
3819
3820         case DIOCSETTIMEOUT: {
3821                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
3822                 int              old;
3823
3824                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3825                     pt->seconds < 0) {
3826                         error = EINVAL;
3827                         break;
3828                 }
3829                 PF_RULES_WLOCK();
3830                 old = V_pf_default_rule.timeout[pt->timeout];
3831                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3832                         pt->seconds = 1;
3833                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3834                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3835                         wakeup(pf_purge_thread);
3836                 pt->seconds = old;
3837                 PF_RULES_WUNLOCK();
3838                 break;
3839         }
3840
3841         case DIOCGETTIMEOUT: {
3842                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
3843
3844                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3845                         error = EINVAL;
3846                         break;
3847                 }
3848                 PF_RULES_RLOCK();
3849                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3850                 PF_RULES_RUNLOCK();
3851                 break;
3852         }
3853
3854         case DIOCGETLIMIT: {
3855                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
3856
3857                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3858                         error = EINVAL;
3859                         break;
3860                 }
3861                 PF_RULES_RLOCK();
3862                 pl->limit = V_pf_limits[pl->index].limit;
3863                 PF_RULES_RUNLOCK();
3864                 break;
3865         }
3866
3867         case DIOCSETLIMIT: {
3868                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
3869                 int                      old_limit;
3870
3871                 PF_RULES_WLOCK();
3872                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3873                     V_pf_limits[pl->index].zone == NULL) {
3874                         PF_RULES_WUNLOCK();
3875                         error = EINVAL;
3876                         break;
3877                 }
3878                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3879                 old_limit = V_pf_limits[pl->index].limit;
3880                 V_pf_limits[pl->index].limit = pl->limit;
3881                 pl->limit = old_limit;
3882                 PF_RULES_WUNLOCK();
3883                 break;
3884         }
3885
3886         case DIOCSETDEBUG: {
3887                 u_int32_t       *level = (u_int32_t *)addr;
3888
3889                 PF_RULES_WLOCK();
3890                 V_pf_status.debug = *level;
3891                 PF_RULES_WUNLOCK();
3892                 break;
3893         }
3894
3895         case DIOCCLRRULECTRS: {
3896                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3897                 struct pf_kruleset      *ruleset = &pf_main_ruleset;
3898                 struct pf_krule         *rule;
3899
3900                 PF_RULES_WLOCK();
3901                 TAILQ_FOREACH(rule,
3902                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3903                         counter_u64_zero(rule->evaluations);
3904                         for (int i = 0; i < 2; i++) {
3905                                 counter_u64_zero(rule->packets[i]);
3906                                 counter_u64_zero(rule->bytes[i]);
3907                         }
3908                 }
3909                 PF_RULES_WUNLOCK();
3910                 break;
3911         }
3912
3913         case DIOCGIFSPEEDV0:
3914         case DIOCGIFSPEEDV1: {
3915                 struct pf_ifspeed_v1    *psp = (struct pf_ifspeed_v1 *)addr;
3916                 struct pf_ifspeed_v1    ps;
3917                 struct ifnet            *ifp;
3918
3919                 if (psp->ifname[0] != 0) {
3920                         /* Can we completely trust user-land? */
3921                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3922                         ifp = ifunit(ps.ifname);
3923                         if (ifp != NULL) {
3924                                 psp->baudrate32 =
3925                                     (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3926                                 if (cmd == DIOCGIFSPEEDV1)
3927                                         psp->baudrate = ifp->if_baudrate;
3928                         } else
3929                                 error = EINVAL;
3930                 } else
3931                         error = EINVAL;
3932                 break;
3933         }
3934
3935 #ifdef ALTQ
3936         case DIOCSTARTALTQ: {
3937                 struct pf_altq          *altq;
3938
3939                 PF_RULES_WLOCK();
3940                 /* enable all altq interfaces on active list */
3941                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3942                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3943                                 error = pf_enable_altq(altq);
3944                                 if (error != 0)
3945                                         break;
3946                         }
3947                 }
3948                 if (error == 0)
3949                         V_pf_altq_running = 1;
3950                 PF_RULES_WUNLOCK();
3951                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3952                 break;
3953         }
3954
3955         case DIOCSTOPALTQ: {
3956                 struct pf_altq          *altq;
3957
3958                 PF_RULES_WLOCK();
3959                 /* disable all altq interfaces on active list */
3960                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3961                         if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3962                                 error = pf_disable_altq(altq);
3963                                 if (error != 0)
3964                                         break;
3965                         }
3966                 }
3967                 if (error == 0)
3968                         V_pf_altq_running = 0;
3969                 PF_RULES_WUNLOCK();
3970                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3971                 break;
3972         }
3973
3974         case DIOCADDALTQV0:
3975         case DIOCADDALTQV1: {
3976                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
3977                 struct pf_altq          *altq, *a;
3978                 struct ifnet            *ifp;
3979
3980                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
3981                 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
3982                 if (error)
3983                         break;
3984                 altq->local_flags = 0;
3985
3986                 PF_RULES_WLOCK();
3987                 if (pa->ticket != V_ticket_altqs_inactive) {
3988                         PF_RULES_WUNLOCK();
3989                         free(altq, M_PFALTQ);
3990                         error = EBUSY;
3991                         break;
3992                 }
3993
3994                 /*
3995                  * if this is for a queue, find the discipline and
3996                  * copy the necessary fields
3997                  */
3998                 if (altq->qname[0] != 0) {
3999                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4000                                 PF_RULES_WUNLOCK();
4001                                 error = EBUSY;
4002                                 free(altq, M_PFALTQ);
4003                                 break;
4004                         }
4005                         altq->altq_disc = NULL;
4006                         TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4007                                 if (strncmp(a->ifname, altq->ifname,
4008                                     IFNAMSIZ) == 0) {
4009                                         altq->altq_disc = a->altq_disc;
4010                                         break;
4011                                 }
4012                         }
4013                 }
4014
4015                 if ((ifp = ifunit(altq->ifname)) == NULL)
4016                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4017                 else
4018                         error = altq_add(ifp, altq);
4019
4020                 if (error) {
4021                         PF_RULES_WUNLOCK();
4022                         free(altq, M_PFALTQ);
4023                         break;
4024                 }
4025
4026                 if (altq->qname[0] != 0)
4027                         TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4028                 else
4029                         TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4030                 /* version error check done on import above */
4031                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4032                 PF_RULES_WUNLOCK();
4033                 break;
4034         }
4035
4036         case DIOCGETALTQSV0:
4037         case DIOCGETALTQSV1: {
4038                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
4039                 struct pf_altq          *altq;
4040
4041                 PF_RULES_RLOCK();
4042                 pa->nr = 0;
4043                 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4044                         pa->nr++;
4045                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4046                         pa->nr++;
4047                 pa->ticket = V_ticket_altqs_active;
4048                 PF_RULES_RUNLOCK();
4049                 break;
4050         }
4051
4052         case DIOCGETALTQV0:
4053         case DIOCGETALTQV1: {
4054                 struct pfioc_altq_v1    *pa = (struct pfioc_altq_v1 *)addr;
4055                 struct pf_altq          *altq;
4056
4057                 PF_RULES_RLOCK();
4058                 if (pa->ticket != V_ticket_altqs_active) {
4059                         PF_RULES_RUNLOCK();
4060                         error = EBUSY;
4061                         break;
4062                 }
4063                 altq = pf_altq_get_nth_active(pa->nr);
4064                 if (altq == NULL) {
4065                         PF_RULES_RUNLOCK();
4066                         error = EBUSY;
4067                         break;
4068                 }
4069                 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4070                 PF_RULES_RUNLOCK();
4071                 break;
4072         }
4073
4074         case DIOCCHANGEALTQV0:
4075         case DIOCCHANGEALTQV1:
4076                 /* CHANGEALTQ not supported yet! */
4077                 error = ENODEV;
4078                 break;
4079
4080         case DIOCGETQSTATSV0:
4081         case DIOCGETQSTATSV1: {
4082                 struct pfioc_qstats_v1  *pq = (struct pfioc_qstats_v1 *)addr;
4083                 struct pf_altq          *altq;
4084                 int                      nbytes;
4085                 u_int32_t                version;
4086
4087                 PF_RULES_RLOCK();
4088                 if (pq->ticket != V_ticket_altqs_active) {
4089                         PF_RULES_RUNLOCK();
4090                         error = EBUSY;
4091                         break;
4092                 }
4093                 nbytes = pq->nbytes;
4094                 altq = pf_altq_get_nth_active(pq->nr);
4095                 if (altq == NULL) {
4096                         PF_RULES_RUNLOCK();
4097                         error = EBUSY;
4098                         break;
4099                 }
4100
4101                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4102                         PF_RULES_RUNLOCK();
4103                         error = ENXIO;
4104                         break;
4105                 }
4106                 PF_RULES_RUNLOCK();
4107                 if (cmd == DIOCGETQSTATSV0)
4108                         version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4109                 else
4110                         version = pq->version;
4111                 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4112                 if (error == 0) {
4113                         pq->scheduler = altq->scheduler;
4114                         pq->nbytes = nbytes;
4115                 }
4116                 break;
4117         }
4118 #endif /* ALTQ */
4119
4120         case DIOCBEGINADDRS: {
4121                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
4122
4123                 PF_RULES_WLOCK();
4124                 pf_empty_kpool(&V_pf_pabuf);
4125                 pp->ticket = ++V_ticket_pabuf;
4126                 PF_RULES_WUNLOCK();
4127                 break;
4128         }
4129
4130         case DIOCADDADDR: {
4131                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
4132                 struct pf_kpooladdr     *pa;
4133                 struct pfi_kkif         *kif = NULL;
4134
4135 #ifndef INET
4136                 if (pp->af == AF_INET) {
4137                         error = EAFNOSUPPORT;
4138                         break;
4139                 }
4140 #endif /* INET */
4141 #ifndef INET6
4142                 if (pp->af == AF_INET6) {
4143                         error = EAFNOSUPPORT;
4144                         break;
4145                 }
4146 #endif /* INET6 */
4147                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4148                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4149                     pp->addr.addr.type != PF_ADDR_TABLE) {
4150                         error = EINVAL;
4151                         break;
4152                 }
4153                 if (pp->addr.addr.p.dyn != NULL) {
4154                         error = EINVAL;
4155                         break;
4156                 }
4157                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4158                 pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4159                 if (pa->ifname[0])
4160                         kif = pf_kkif_create(M_WAITOK);
4161                 PF_RULES_WLOCK();
4162                 if (pp->ticket != V_ticket_pabuf) {
4163                         PF_RULES_WUNLOCK();
4164                         if (pa->ifname[0])
4165                                 pf_kkif_free(kif);
4166                         free(pa, M_PFRULE);
4167                         error = EBUSY;
4168                         break;
4169                 }
4170                 if (pa->ifname[0]) {
4171                         pa->kif = pfi_kkif_attach(kif, pa->ifname);
4172                         kif = NULL;
4173                         pfi_kkif_ref(pa->kif);
4174                 } else
4175                         pa->kif = NULL;
4176                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4177                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4178                         if (pa->ifname[0])
4179                                 pfi_kkif_unref(pa->kif);
4180                         PF_RULES_WUNLOCK();
4181                         free(pa, M_PFRULE);
4182                         break;
4183                 }
4184                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4185                 PF_RULES_WUNLOCK();
4186                 break;
4187         }
4188
4189         case DIOCGETADDRS: {
4190                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
4191                 struct pf_kpool         *pool;
4192                 struct pf_kpooladdr     *pa;
4193
4194                 PF_RULES_RLOCK();
4195                 pp->nr = 0;
4196                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4197                     pp->r_num, 0, 1, 0);
4198                 if (pool == NULL) {
4199                         PF_RULES_RUNLOCK();
4200                         error = EBUSY;
4201                         break;
4202                 }
4203                 TAILQ_FOREACH(pa, &pool->list, entries)
4204                         pp->nr++;
4205                 PF_RULES_RUNLOCK();
4206                 break;
4207         }
4208
4209         case DIOCGETADDR: {
4210                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
4211                 struct pf_kpool         *pool;
4212                 struct pf_kpooladdr     *pa;
4213                 u_int32_t                nr = 0;
4214
4215                 PF_RULES_RLOCK();
4216                 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4217                     pp->r_num, 0, 1, 1);
4218                 if (pool == NULL) {
4219                         PF_RULES_RUNLOCK();
4220                         error = EBUSY;
4221                         break;
4222                 }
4223                 pa = TAILQ_FIRST(&pool->list);
4224                 while ((pa != NULL) && (nr < pp->nr)) {
4225                         pa = TAILQ_NEXT(pa, entries);
4226                         nr++;
4227                 }
4228                 if (pa == NULL) {
4229                         PF_RULES_RUNLOCK();
4230                         error = EBUSY;
4231                         break;
4232                 }
4233                 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4234                 pf_addr_copyout(&pp->addr.addr);
4235                 PF_RULES_RUNLOCK();
4236                 break;
4237         }
4238
4239         case DIOCCHANGEADDR: {
4240                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
4241                 struct pf_kpool         *pool;
4242                 struct pf_kpooladdr     *oldpa = NULL, *newpa = NULL;
4243                 struct pf_kruleset      *ruleset;
4244                 struct pfi_kkif         *kif = NULL;
4245
4246                 if (pca->action < PF_CHANGE_ADD_HEAD ||
4247                     pca->action > PF_CHANGE_REMOVE) {
4248                         error = EINVAL;
4249                         break;
4250                 }
4251                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4252                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4253                     pca->addr.addr.type != PF_ADDR_TABLE) {
4254                         error = EINVAL;
4255                         break;
4256                 }
4257                 if (pca->addr.addr.p.dyn != NULL) {
4258                         error = EINVAL;
4259                         break;
4260                 }
4261
4262                 if (pca->action != PF_CHANGE_REMOVE) {
4263 #ifndef INET
4264                         if (pca->af == AF_INET) {
4265                                 error = EAFNOSUPPORT;
4266                                 break;
4267                         }
4268 #endif /* INET */
4269 #ifndef INET6
4270                         if (pca->af == AF_INET6) {
4271                                 error = EAFNOSUPPORT;
4272                                 break;
4273                         }
4274 #endif /* INET6 */
4275                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4276                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4277                         if (newpa->ifname[0])
4278                                 kif = pf_kkif_create(M_WAITOK);
4279                         newpa->kif = NULL;
4280                 }
4281 #define ERROUT(x)       ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4282                 PF_RULES_WLOCK();
4283                 ruleset = pf_find_kruleset(pca->anchor);
4284                 if (ruleset == NULL)
4285                         ERROUT(EBUSY);
4286
4287                 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4288                     pca->r_num, pca->r_last, 1, 1);
4289                 if (pool == NULL)
4290                         ERROUT(EBUSY);
4291
4292                 if (pca->action != PF_CHANGE_REMOVE) {
4293                         if (newpa->ifname[0]) {
4294                                 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4295                                 pfi_kkif_ref(newpa->kif);
4296                                 kif = NULL;
4297                         }
4298
4299                         switch (newpa->addr.type) {
4300                         case PF_ADDR_DYNIFTL:
4301                                 error = pfi_dynaddr_setup(&newpa->addr,
4302                                     pca->af);
4303                                 break;
4304                         case PF_ADDR_TABLE:
4305                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4306                                     newpa->addr.v.tblname);
4307                                 if (newpa->addr.p.tbl == NULL)
4308                                         error = ENOMEM;
4309                                 break;
4310                         }
4311                         if (error)
4312                                 goto DIOCCHANGEADDR_error;
4313                 }
4314
4315                 switch (pca->action) {
4316                 case PF_CHANGE_ADD_HEAD:
4317                         oldpa = TAILQ_FIRST(&pool->list);
4318                         break;
4319                 case PF_CHANGE_ADD_TAIL:
4320                         oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4321                         break;
4322                 default:
4323                         oldpa = TAILQ_FIRST(&pool->list);
4324                         for (int i = 0; oldpa && i < pca->nr; i++)
4325                                 oldpa = TAILQ_NEXT(oldpa, entries);
4326
4327                         if (oldpa == NULL)
4328                                 ERROUT(EINVAL);
4329                 }
4330
4331                 if (pca->action == PF_CHANGE_REMOVE) {
4332                         TAILQ_REMOVE(&pool->list, oldpa, entries);
4333                         switch (oldpa->addr.type) {
4334                         case PF_ADDR_DYNIFTL:
4335                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4336                                 break;
4337                         case PF_ADDR_TABLE:
4338                                 pfr_detach_table(oldpa->addr.p.tbl);
4339                                 break;
4340                         }
4341                         if (oldpa->kif)
4342                                 pfi_kkif_unref(oldpa->kif);
4343                         free(oldpa, M_PFRULE);
4344                 } else {
4345                         if (oldpa == NULL)
4346                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4347                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
4348                             pca->action == PF_CHANGE_ADD_BEFORE)
4349                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4350                         else
4351                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4352                                     newpa, entries);
4353                 }
4354
4355                 pool->cur = TAILQ_FIRST(&pool->list);
4356                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4357                 PF_RULES_WUNLOCK();
4358                 break;
4359
4360 #undef ERROUT
4361 DIOCCHANGEADDR_error:
4362                 if (newpa != NULL) {
4363                         if (newpa->kif)
4364                                 pfi_kkif_unref(newpa->kif);
4365                         free(newpa, M_PFRULE);
4366                 }
4367                 PF_RULES_WUNLOCK();
4368                 pf_kkif_free(kif);
4369                 break;
4370         }
4371
4372         case DIOCGETRULESETS: {
4373                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
4374                 struct pf_kruleset      *ruleset;
4375                 struct pf_kanchor       *anchor;
4376
4377                 PF_RULES_RLOCK();
4378                 pr->path[sizeof(pr->path) - 1] = 0;
4379                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4380                         PF_RULES_RUNLOCK();
4381                         error = ENOENT;
4382                         break;
4383                 }
4384                 pr->nr = 0;
4385                 if (ruleset->anchor == NULL) {
4386                         /* XXX kludge for pf_main_ruleset */
4387                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4388                                 if (anchor->parent == NULL)
4389                                         pr->nr++;
4390                 } else {
4391                         RB_FOREACH(anchor, pf_kanchor_node,
4392                             &ruleset->anchor->children)
4393                                 pr->nr++;
4394                 }
4395                 PF_RULES_RUNLOCK();
4396                 break;
4397         }
4398
4399         case DIOCGETRULESET: {
4400                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
4401                 struct pf_kruleset      *ruleset;
4402                 struct pf_kanchor       *anchor;
4403                 u_int32_t                nr = 0;
4404
4405                 PF_RULES_RLOCK();
4406                 pr->path[sizeof(pr->path) - 1] = 0;
4407                 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4408                         PF_RULES_RUNLOCK();
4409                         error = ENOENT;
4410                         break;
4411                 }
4412                 pr->name[0] = 0;
4413                 if (ruleset->anchor == NULL) {
4414                         /* XXX kludge for pf_main_ruleset */
4415                         RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4416                                 if (anchor->parent == NULL && nr++ == pr->nr) {
4417                                         strlcpy(pr->name, anchor->name,
4418                                             sizeof(pr->name));
4419                                         break;
4420                                 }
4421                 } else {
4422                         RB_FOREACH(anchor, pf_kanchor_node,
4423                             &ruleset->anchor->children)
4424                                 if (nr++ == pr->nr) {
4425                                         strlcpy(pr->name, anchor->name,
4426                                             sizeof(pr->name));
4427                                         break;
4428                                 }
4429                 }
4430                 if (!pr->name[0])
4431                         error = EBUSY;
4432                 PF_RULES_RUNLOCK();
4433                 break;
4434         }
4435
4436         case DIOCRCLRTABLES: {
4437                 struct pfioc_table *io = (struct pfioc_table *)addr;
4438
4439                 if (io->pfrio_esize != 0) {
4440                         error = ENODEV;
4441                         break;
4442                 }
4443                 PF_RULES_WLOCK();
4444                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4445                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
4446                 PF_RULES_WUNLOCK();
4447                 break;
4448         }
4449
4450         case DIOCRADDTABLES: {
4451                 struct pfioc_table *io = (struct pfioc_table *)addr;
4452                 struct pfr_table *pfrts;
4453                 size_t totlen;
4454
4455                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4456                         error = ENODEV;
4457                         break;
4458                 }
4459
4460                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4461                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4462                         error = ENOMEM;
4463                         break;
4464                 }
4465
4466                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4467                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4468                     M_TEMP, M_WAITOK);
4469                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4470                 if (error) {
4471                         free(pfrts, M_TEMP);
4472                         break;
4473                 }
4474                 PF_RULES_WLOCK();
4475                 error = pfr_add_tables(pfrts, io->pfrio_size,
4476                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4477                 PF_RULES_WUNLOCK();
4478                 free(pfrts, M_TEMP);
4479                 break;
4480         }
4481
4482         case DIOCRDELTABLES: {
4483                 struct pfioc_table *io = (struct pfioc_table *)addr;
4484                 struct pfr_table *pfrts;
4485                 size_t totlen;
4486
4487                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4488                         error = ENODEV;
4489                         break;
4490                 }
4491
4492                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4493                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4494                         error = ENOMEM;
4495                         break;
4496                 }
4497
4498                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4499                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4500                     M_TEMP, M_WAITOK);
4501                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4502                 if (error) {
4503                         free(pfrts, M_TEMP);
4504                         break;
4505                 }
4506                 PF_RULES_WLOCK();
4507                 error = pfr_del_tables(pfrts, io->pfrio_size,
4508                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4509                 PF_RULES_WUNLOCK();
4510                 free(pfrts, M_TEMP);
4511                 break;
4512         }
4513
4514         case DIOCRGETTABLES: {
4515                 struct pfioc_table *io = (struct pfioc_table *)addr;
4516                 struct pfr_table *pfrts;
4517                 size_t totlen;
4518                 int n;
4519
4520                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4521                         error = ENODEV;
4522                         break;
4523                 }
4524                 PF_RULES_RLOCK();
4525                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4526                 if (n < 0) {
4527                         PF_RULES_RUNLOCK();
4528                         error = EINVAL;
4529                         break;
4530                 }
4531                 io->pfrio_size = min(io->pfrio_size, n);
4532
4533                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4534
4535                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4536                     M_TEMP, M_NOWAIT);
4537                 if (pfrts == NULL) {
4538                         error = ENOMEM;
4539                         PF_RULES_RUNLOCK();
4540                         break;
4541                 }
4542                 error = pfr_get_tables(&io->pfrio_table, pfrts,
4543                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4544                 PF_RULES_RUNLOCK();
4545                 if (error == 0)
4546                         error = copyout(pfrts, io->pfrio_buffer, totlen);
4547                 free(pfrts, M_TEMP);
4548                 break;
4549         }
4550
4551         case DIOCRGETTSTATS: {
4552                 struct pfioc_table *io = (struct pfioc_table *)addr;
4553                 struct pfr_tstats *pfrtstats;
4554                 size_t totlen;
4555                 int n;
4556
4557                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4558                         error = ENODEV;
4559                         break;
4560                 }
4561                 PF_RULES_WLOCK();
4562                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4563                 if (n < 0) {
4564                         PF_RULES_WUNLOCK();
4565                         error = EINVAL;
4566                         break;
4567                 }
4568                 io->pfrio_size = min(io->pfrio_size, n);
4569
4570                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4571                 pfrtstats = mallocarray(io->pfrio_size,
4572                     sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
4573                 if (pfrtstats == NULL) {
4574                         error = ENOMEM;
4575                         PF_RULES_WUNLOCK();
4576                         break;
4577                 }
4578                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4579                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4580                 PF_RULES_WUNLOCK();
4581                 if (error == 0)
4582                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4583                 free(pfrtstats, M_TEMP);
4584                 break;
4585         }
4586
4587         case DIOCRCLRTSTATS: {
4588                 struct pfioc_table *io = (struct pfioc_table *)addr;
4589                 struct pfr_table *pfrts;
4590                 size_t totlen;
4591
4592                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4593                         error = ENODEV;
4594                         break;
4595                 }
4596
4597                 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4598                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4599                         /* We used to count tables and use the minimum required
4600                          * size, so we didn't fail on overly large requests.
4601                          * Keep doing so. */
4602                         io->pfrio_size = pf_ioctl_maxcount;
4603                         break;
4604                 }
4605
4606                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4607                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4608                     M_TEMP, M_NOWAIT);
4609                 if (pfrts == NULL) {
4610                         error = ENOMEM;
4611                         break;
4612                 }
4613                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4614                 if (error) {
4615                         free(pfrts, M_TEMP);
4616                         break;
4617                 }
4618
4619                 PF_RULES_WLOCK();
4620                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4621                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4622                 PF_RULES_WUNLOCK();
4623                 free(pfrts, M_TEMP);
4624                 break;
4625         }
4626
4627         case DIOCRSETTFLAGS: {
4628                 struct pfioc_table *io = (struct pfioc_table *)addr;
4629                 struct pfr_table *pfrts;
4630                 size_t totlen;
4631                 int n;
4632
4633                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4634                         error = ENODEV;
4635                         break;
4636                 }
4637
4638                 PF_RULES_RLOCK();
4639                 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4640                 if (n < 0) {
4641                         PF_RULES_RUNLOCK();
4642                         error = EINVAL;
4643                         break;
4644                 }
4645
4646                 io->pfrio_size = min(io->pfrio_size, n);
4647                 PF_RULES_RUNLOCK();
4648
4649                 totlen = io->pfrio_size * sizeof(struct pfr_table);
4650                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4651                     M_TEMP, M_WAITOK);
4652                 error = copyin(io->pfrio_buffer, pfrts, totlen);
4653                 if (error) {
4654                         free(pfrts, M_TEMP);
4655                         break;
4656                 }
4657                 PF_RULES_WLOCK();
4658                 error = pfr_set_tflags(pfrts, io->pfrio_size,
4659                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4660                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4661                 PF_RULES_WUNLOCK();
4662                 free(pfrts, M_TEMP);
4663                 break;
4664         }
4665
4666         case DIOCRCLRADDRS: {
4667                 struct pfioc_table *io = (struct pfioc_table *)addr;
4668
4669                 if (io->pfrio_esize != 0) {
4670                         error = ENODEV;
4671                         break;
4672                 }
4673                 PF_RULES_WLOCK();
4674                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4675                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
4676                 PF_RULES_WUNLOCK();
4677                 break;
4678         }
4679
4680         case DIOCRADDADDRS: {
4681                 struct pfioc_table *io = (struct pfioc_table *)addr;
4682                 struct pfr_addr *pfras;
4683                 size_t totlen;
4684
4685                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4686                         error = ENODEV;
4687                         break;
4688                 }
4689                 if (io->pfrio_size < 0 ||
4690                     io->pfrio_size > pf_ioctl_maxcount ||
4691                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4692                         error = EINVAL;
4693                         break;
4694                 }
4695                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4696                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4697                     M_TEMP, M_NOWAIT);
4698                 if (! pfras) {
4699                         error = ENOMEM;
4700                         break;
4701                 }
4702                 error = copyin(io->pfrio_buffer, pfras, totlen);
4703                 if (error) {
4704                         free(pfras, M_TEMP);
4705                         break;
4706                 }
4707                 PF_RULES_WLOCK();
4708                 error = pfr_add_addrs(&io->pfrio_table, pfras,
4709                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4710                     PFR_FLAG_USERIOCTL);
4711                 PF_RULES_WUNLOCK();
4712                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4713                         error = copyout(pfras, io->pfrio_buffer, totlen);
4714                 free(pfras, M_TEMP);
4715                 break;
4716         }
4717
4718         case DIOCRDELADDRS: {
4719                 struct pfioc_table *io = (struct pfioc_table *)addr;
4720                 struct pfr_addr *pfras;
4721                 size_t totlen;
4722
4723                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4724                         error = ENODEV;
4725                         break;
4726                 }
4727                 if (io->pfrio_size < 0 ||
4728                     io->pfrio_size > pf_ioctl_maxcount ||
4729                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4730                         error = EINVAL;
4731                         break;
4732                 }
4733                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4734                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4735                     M_TEMP, M_NOWAIT);
4736                 if (! pfras) {
4737                         error = ENOMEM;
4738                         break;
4739                 }
4740                 error = copyin(io->pfrio_buffer, pfras, totlen);
4741                 if (error) {
4742                         free(pfras, M_TEMP);
4743                         break;
4744                 }
4745                 PF_RULES_WLOCK();
4746                 error = pfr_del_addrs(&io->pfrio_table, pfras,
4747                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4748                     PFR_FLAG_USERIOCTL);
4749                 PF_RULES_WUNLOCK();
4750                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4751                         error = copyout(pfras, io->pfrio_buffer, totlen);
4752                 free(pfras, M_TEMP);
4753                 break;
4754         }
4755
4756         case DIOCRSETADDRS: {
4757                 struct pfioc_table *io = (struct pfioc_table *)addr;
4758                 struct pfr_addr *pfras;
4759                 size_t totlen, count;
4760
4761                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4762                         error = ENODEV;
4763                         break;
4764                 }
4765                 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4766                         error = EINVAL;
4767                         break;
4768                 }
4769                 count = max(io->pfrio_size, io->pfrio_size2);
4770                 if (count > pf_ioctl_maxcount ||
4771                     WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4772                         error = EINVAL;
4773                         break;
4774                 }
4775                 totlen = count * sizeof(struct pfr_addr);
4776                 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4777                     M_NOWAIT);
4778                 if (! pfras) {
4779                         error = ENOMEM;
4780                         break;
4781                 }
4782                 error = copyin(io->pfrio_buffer, pfras, totlen);
4783                 if (error) {
4784                         free(pfras, M_TEMP);
4785                         break;
4786                 }
4787                 PF_RULES_WLOCK();
4788                 error = pfr_set_addrs(&io->pfrio_table, pfras,
4789                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4790                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4791                     PFR_FLAG_USERIOCTL, 0);
4792                 PF_RULES_WUNLOCK();
4793                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4794                         error = copyout(pfras, io->pfrio_buffer, totlen);
4795                 free(pfras, M_TEMP);
4796                 break;
4797         }
4798
4799         case DIOCRGETADDRS: {
4800                 struct pfioc_table *io = (struct pfioc_table *)addr;
4801                 struct pfr_addr *pfras;
4802                 size_t totlen;
4803
4804                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4805                         error = ENODEV;
4806                         break;
4807                 }
4808                 if (io->pfrio_size < 0 ||
4809                     io->pfrio_size > pf_ioctl_maxcount ||
4810                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4811                         error = EINVAL;
4812                         break;
4813                 }
4814                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4815                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4816                     M_TEMP, M_NOWAIT);
4817                 if (! pfras) {
4818                         error = ENOMEM;
4819                         break;
4820                 }
4821                 PF_RULES_RLOCK();
4822                 error = pfr_get_addrs(&io->pfrio_table, pfras,
4823                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4824                 PF_RULES_RUNLOCK();
4825                 if (error == 0)
4826                         error = copyout(pfras, io->pfrio_buffer, totlen);
4827                 free(pfras, M_TEMP);
4828                 break;
4829         }
4830
4831         case DIOCRGETASTATS: {
4832                 struct pfioc_table *io = (struct pfioc_table *)addr;
4833                 struct pfr_astats *pfrastats;
4834                 size_t totlen;
4835
4836                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4837                         error = ENODEV;
4838                         break;
4839                 }
4840                 if (io->pfrio_size < 0 ||
4841                     io->pfrio_size > pf_ioctl_maxcount ||
4842                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4843                         error = EINVAL;
4844                         break;
4845                 }
4846                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
4847                 pfrastats = mallocarray(io->pfrio_size,
4848                     sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
4849                 if (! pfrastats) {
4850                         error = ENOMEM;
4851                         break;
4852                 }
4853                 PF_RULES_RLOCK();
4854                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
4855                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4856                 PF_RULES_RUNLOCK();
4857                 if (error == 0)
4858                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
4859                 free(pfrastats, M_TEMP);
4860                 break;
4861         }
4862
4863         case DIOCRCLRASTATS: {
4864                 struct pfioc_table *io = (struct pfioc_table *)addr;
4865                 struct pfr_addr *pfras;
4866                 size_t totlen;
4867
4868                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4869                         error = ENODEV;
4870                         break;
4871                 }
4872                 if (io->pfrio_size < 0 ||
4873                     io->pfrio_size > pf_ioctl_maxcount ||
4874                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4875                         error = EINVAL;
4876                         break;
4877                 }
4878                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4879                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4880                     M_TEMP, M_NOWAIT);
4881                 if (! pfras) {
4882                         error = ENOMEM;
4883                         break;
4884                 }
4885                 error = copyin(io->pfrio_buffer, pfras, totlen);
4886                 if (error) {
4887                         free(pfras, M_TEMP);
4888                         break;
4889                 }
4890                 PF_RULES_WLOCK();
4891                 error = pfr_clr_astats(&io->pfrio_table, pfras,
4892                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4893                     PFR_FLAG_USERIOCTL);
4894                 PF_RULES_WUNLOCK();
4895                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4896                         error = copyout(pfras, io->pfrio_buffer, totlen);
4897                 free(pfras, M_TEMP);
4898                 break;
4899         }
4900
4901         case DIOCRTSTADDRS: {
4902                 struct pfioc_table *io = (struct pfioc_table *)addr;
4903                 struct pfr_addr *pfras;
4904                 size_t totlen;
4905
4906                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4907                         error = ENODEV;
4908                         break;
4909                 }
4910                 if (io->pfrio_size < 0 ||
4911                     io->pfrio_size > pf_ioctl_maxcount ||
4912                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4913                         error = EINVAL;
4914                         break;
4915                 }
4916                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4917                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4918                     M_TEMP, M_NOWAIT);
4919                 if (! pfras) {
4920                         error = ENOMEM;
4921                         break;
4922                 }
4923                 error = copyin(io->pfrio_buffer, pfras, totlen);
4924                 if (error) {
4925                         free(pfras, M_TEMP);
4926                         break;
4927                 }
4928                 PF_RULES_RLOCK();
4929                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
4930                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4931                     PFR_FLAG_USERIOCTL);
4932                 PF_RULES_RUNLOCK();
4933                 if (error == 0)
4934                         error = copyout(pfras, io->pfrio_buffer, totlen);
4935                 free(pfras, M_TEMP);
4936                 break;
4937         }
4938
4939         case DIOCRINADEFINE: {
4940                 struct pfioc_table *io = (struct pfioc_table *)addr;
4941                 struct pfr_addr *pfras;
4942                 size_t totlen;
4943
4944                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4945                         error = ENODEV;
4946                         break;
4947                 }
4948                 if (io->pfrio_size < 0 ||
4949                     io->pfrio_size > pf_ioctl_maxcount ||
4950                     WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4951                         error = EINVAL;
4952                         break;
4953                 }
4954                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4955                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4956                     M_TEMP, M_NOWAIT);
4957                 if (! pfras) {
4958                         error = ENOMEM;
4959                         break;
4960                 }
4961                 error = copyin(io->pfrio_buffer, pfras, totlen);
4962                 if (error) {
4963                         free(pfras, M_TEMP);
4964                         break;
4965                 }
4966                 PF_RULES_WLOCK();
4967                 error = pfr_ina_define(&io->pfrio_table, pfras,
4968                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4969                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4970                 PF_RULES_WUNLOCK();
4971                 free(pfras, M_TEMP);
4972                 break;
4973         }
4974
4975         case DIOCOSFPADD: {
4976                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4977                 PF_RULES_WLOCK();
4978                 error = pf_osfp_add(io);
4979                 PF_RULES_WUNLOCK();
4980                 break;
4981         }
4982
4983         case DIOCOSFPGET: {
4984                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4985                 PF_RULES_RLOCK();
4986                 error = pf_osfp_get(io);
4987                 PF_RULES_RUNLOCK();
4988                 break;
4989         }
4990
4991         case DIOCXBEGIN: {
4992                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
4993                 struct pfioc_trans_e    *ioes, *ioe;
4994                 size_t                   totlen;
4995                 int                      i;
4996
4997                 if (io->esize != sizeof(*ioe)) {
4998                         error = ENODEV;
4999                         break;
5000                 }
5001                 if (io->size < 0 ||
5002                     io->size > pf_ioctl_maxcount ||
5003                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5004                         error = EINVAL;
5005                         break;
5006                 }
5007                 totlen = sizeof(struct pfioc_trans_e) * io->size;
5008                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5009                     M_TEMP, M_NOWAIT);
5010                 if (! ioes) {
5011                         error = ENOMEM;
5012                         break;
5013                 }
5014                 error = copyin(io->array, ioes, totlen);
5015                 if (error) {
5016                         free(ioes, M_TEMP);
5017                         break;
5018                 }
5019                 PF_RULES_WLOCK();
5020                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5021                         switch (ioe->rs_num) {
5022 #ifdef ALTQ
5023                         case PF_RULESET_ALTQ:
5024                                 if (ioe->anchor[0]) {
5025                                         PF_RULES_WUNLOCK();
5026                                         free(ioes, M_TEMP);
5027                                         error = EINVAL;
5028                                         goto fail;
5029                                 }
5030                                 if ((error = pf_begin_altq(&ioe->ticket))) {
5031                                         PF_RULES_WUNLOCK();
5032                                         free(ioes, M_TEMP);
5033                                         goto fail;
5034                                 }
5035                                 break;
5036 #endif /* ALTQ */
5037                         case PF_RULESET_TABLE:
5038                             {
5039                                 struct pfr_table table;
5040
5041                                 bzero(&table, sizeof(table));
5042                                 strlcpy(table.pfrt_anchor, ioe->anchor,
5043                                     sizeof(table.pfrt_anchor));
5044                                 if ((error = pfr_ina_begin(&table,
5045                                     &ioe->ticket, NULL, 0))) {
5046                                         PF_RULES_WUNLOCK();
5047                                         free(ioes, M_TEMP);
5048                                         goto fail;
5049                                 }
5050                                 break;
5051                             }
5052                         default:
5053                                 if ((error = pf_begin_rules(&ioe->ticket,
5054                                     ioe->rs_num, ioe->anchor))) {
5055                                         PF_RULES_WUNLOCK();
5056                                         free(ioes, M_TEMP);
5057                                         goto fail;
5058                                 }
5059                                 break;
5060                         }
5061                 }
5062                 PF_RULES_WUNLOCK();
5063                 error = copyout(ioes, io->array, totlen);
5064                 free(ioes, M_TEMP);
5065                 break;
5066         }
5067
5068         case DIOCXROLLBACK: {
5069                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
5070                 struct pfioc_trans_e    *ioe, *ioes;
5071                 size_t                   totlen;
5072                 int                      i;
5073
5074                 if (io->esize != sizeof(*ioe)) {
5075                         error = ENODEV;
5076                         break;
5077                 }
5078                 if (io->size < 0 ||
5079                     io->size > pf_ioctl_maxcount ||
5080                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5081                         error = EINVAL;
5082                         break;
5083                 }
5084                 totlen = sizeof(struct pfioc_trans_e) * io->size;
5085                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5086                     M_TEMP, M_NOWAIT);
5087                 if (! ioes) {
5088                         error = ENOMEM;
5089                         break;
5090                 }
5091                 error = copyin(io->array, ioes, totlen);
5092                 if (error) {
5093                         free(ioes, M_TEMP);
5094                         break;
5095                 }
5096                 PF_RULES_WLOCK();
5097                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5098                         switch (ioe->rs_num) {
5099 #ifdef ALTQ
5100                         case PF_RULESET_ALTQ:
5101                                 if (ioe->anchor[0]) {
5102                                         PF_RULES_WUNLOCK();
5103                                         free(ioes, M_TEMP);
5104                                         error = EINVAL;
5105                                         goto fail;
5106                                 }
5107                                 if ((error = pf_rollback_altq(ioe->ticket))) {
5108                                         PF_RULES_WUNLOCK();
5109                                         free(ioes, M_TEMP);
5110                                         goto fail; /* really bad */
5111                                 }
5112                                 break;
5113 #endif /* ALTQ */
5114                         case PF_RULESET_TABLE:
5115                             {
5116                                 struct pfr_table table;
5117
5118                                 bzero(&table, sizeof(table));
5119                                 strlcpy(table.pfrt_anchor, ioe->anchor,
5120                                     sizeof(table.pfrt_anchor));
5121                                 if ((error = pfr_ina_rollback(&table,
5122                                     ioe->ticket, NULL, 0))) {
5123                                         PF_RULES_WUNLOCK();
5124                                         free(ioes, M_TEMP);
5125                                         goto fail; /* really bad */
5126                                 }
5127                                 break;
5128                             }
5129                         default:
5130                                 if ((error = pf_rollback_rules(ioe->ticket,
5131                                     ioe->rs_num, ioe->anchor))) {
5132                                         PF_RULES_WUNLOCK();
5133                                         free(ioes, M_TEMP);
5134                                         goto fail; /* really bad */
5135                                 }
5136                                 break;
5137                         }
5138                 }
5139                 PF_RULES_WUNLOCK();
5140                 free(ioes, M_TEMP);
5141                 break;
5142         }
5143
5144         case DIOCXCOMMIT: {
5145                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
5146                 struct pfioc_trans_e    *ioe, *ioes;
5147                 struct pf_kruleset      *rs;
5148                 size_t                   totlen;
5149                 int                      i;
5150
5151                 if (io->esize != sizeof(*ioe)) {
5152                         error = ENODEV;
5153                         break;
5154                 }
5155
5156                 if (io->size < 0 ||
5157                     io->size > pf_ioctl_maxcount ||
5158                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5159                         error = EINVAL;
5160                         break;
5161                 }
5162
5163                 totlen = sizeof(struct pfioc_trans_e) * io->size;
5164                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5165                     M_TEMP, M_NOWAIT);
5166                 if (ioes == NULL) {
5167                         error = ENOMEM;
5168                         break;
5169                 }
5170                 error = copyin(io->array, ioes, totlen);
5171                 if (error) {
5172                         free(ioes, M_TEMP);
5173                         break;
5174                 }
5175                 PF_RULES_WLOCK();
5176                 /* First makes sure everything will succeed. */
5177                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5178                         switch (ioe->rs_num) {
5179 #ifdef ALTQ
5180                         case PF_RULESET_ALTQ:
5181                                 if (ioe->anchor[0]) {
5182                                         PF_RULES_WUNLOCK();
5183                                         free(ioes, M_TEMP);
5184                                         error = EINVAL;
5185                                         goto fail;
5186                                 }
5187                                 if (!V_altqs_inactive_open || ioe->ticket !=
5188                                     V_ticket_altqs_inactive) {
5189                                         PF_RULES_WUNLOCK();
5190                                         free(ioes, M_TEMP);
5191                                         error = EBUSY;
5192                                         goto fail;
5193                                 }
5194                                 break;
5195 #endif /* ALTQ */
5196                         case PF_RULESET_TABLE:
5197                                 rs = pf_find_kruleset(ioe->anchor);
5198                                 if (rs == NULL || !rs->topen || ioe->ticket !=
5199                                     rs->tticket) {
5200                                         PF_RULES_WUNLOCK();
5201                                         free(ioes, M_TEMP);
5202                                         error = EBUSY;
5203                                         goto fail;
5204                                 }
5205                                 break;
5206                         default:
5207                                 if (ioe->rs_num < 0 || ioe->rs_num >=
5208                                     PF_RULESET_MAX) {
5209                                         PF_RULES_WUNLOCK();
5210                                         free(ioes, M_TEMP);
5211                                         error = EINVAL;
5212                                         goto fail;
5213                                 }
5214                                 rs = pf_find_kruleset(ioe->anchor);
5215                                 if (rs == NULL ||
5216                                     !rs->rules[ioe->rs_num].inactive.open ||
5217                                     rs->rules[ioe->rs_num].inactive.ticket !=
5218                                     ioe->ticket) {
5219                                         PF_RULES_WUNLOCK();
5220                                         free(ioes, M_TEMP);
5221                                         error = EBUSY;
5222                                         goto fail;
5223                                 }
5224                                 break;
5225                         }
5226                 }
5227                 /* Now do the commit - no errors should happen here. */
5228                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5229                         switch (ioe->rs_num) {
5230 #ifdef ALTQ
5231                         case PF_RULESET_ALTQ:
5232                                 if ((error = pf_commit_altq(ioe->ticket))) {
5233                                         PF_RULES_WUNLOCK();
5234                                         free(ioes, M_TEMP);
5235                                         goto fail; /* really bad */
5236                                 }
5237                                 break;
5238 #endif /* ALTQ */
5239                         case PF_RULESET_TABLE:
5240                             {
5241                                 struct pfr_table table;
5242
5243                                 bzero(&table, sizeof(table));
5244                                 strlcpy(table.pfrt_anchor, ioe->anchor,
5245                                     sizeof(table.pfrt_anchor));
5246                                 if ((error = pfr_ina_commit(&table,
5247                                     ioe->ticket, NULL, NULL, 0))) {
5248                                         PF_RULES_WUNLOCK();
5249                                         free(ioes, M_TEMP);
5250                                         goto fail; /* really bad */
5251                                 }
5252                                 break;
5253                             }
5254                         default:
5255                                 if ((error = pf_commit_rules(ioe->ticket,
5256                                     ioe->rs_num, ioe->anchor))) {
5257                                         PF_RULES_WUNLOCK();
5258                                         free(ioes, M_TEMP);
5259                                         goto fail; /* really bad */
5260                                 }
5261                                 break;
5262                         }
5263                 }
5264                 PF_RULES_WUNLOCK();
5265                 free(ioes, M_TEMP);
5266                 break;
5267         }
5268
5269         case DIOCGETSRCNODES: {
5270                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
5271                 struct pf_srchash       *sh;
5272                 struct pf_ksrc_node     *n;
5273                 struct pf_src_node      *p, *pstore;
5274                 uint32_t                 i, nr = 0;
5275
5276                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5277                                 i++, sh++) {
5278                         PF_HASHROW_LOCK(sh);
5279                         LIST_FOREACH(n, &sh->nodes, entry)
5280                                 nr++;
5281                         PF_HASHROW_UNLOCK(sh);
5282                 }
5283
5284                 psn->psn_len = min(psn->psn_len,
5285                     sizeof(struct pf_src_node) * nr);
5286
5287                 if (psn->psn_len == 0) {
5288                         psn->psn_len = sizeof(struct pf_src_node) * nr;
5289                         break;
5290                 }
5291
5292                 nr = 0;
5293
5294                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5295                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5296                     i++, sh++) {
5297                     PF_HASHROW_LOCK(sh);
5298                     LIST_FOREACH(n, &sh->nodes, entry) {
5299
5300                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5301                                 break;
5302
5303                         pf_src_node_copy(n, p);
5304
5305                         p++;
5306                         nr++;
5307                     }
5308                     PF_HASHROW_UNLOCK(sh);
5309                 }
5310                 error = copyout(pstore, psn->psn_src_nodes,
5311                     sizeof(struct pf_src_node) * nr);
5312                 if (error) {
5313                         free(pstore, M_TEMP);
5314                         break;
5315                 }
5316                 psn->psn_len = sizeof(struct pf_src_node) * nr;
5317                 free(pstore, M_TEMP);
5318                 break;
5319         }
5320
5321         case DIOCCLRSRCNODES: {
5322                 pf_clear_srcnodes(NULL);
5323                 pf_purge_expired_src_nodes();
5324                 break;
5325         }
5326
5327         case DIOCKILLSRCNODES:
5328                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5329                 break;
5330
5331         case DIOCKEEPCOUNTERS:
5332                 error = pf_keepcounters((struct pfioc_nv *)addr);
5333                 break;
5334
5335         case DIOCSETHOSTID: {
5336                 u_int32_t       *hostid = (u_int32_t *)addr;
5337
5338                 PF_RULES_WLOCK();
5339                 if (*hostid == 0)
5340                         V_pf_status.hostid = arc4random();
5341                 else
5342                         V_pf_status.hostid = *hostid;
5343                 PF_RULES_WUNLOCK();
5344                 break;
5345         }
5346
5347         case DIOCOSFPFLUSH:
5348                 PF_RULES_WLOCK();
5349                 pf_osfp_flush();
5350                 PF_RULES_WUNLOCK();
5351                 break;
5352
5353         case DIOCIGETIFACES: {
5354                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5355                 struct pfi_kif *ifstore;
5356                 size_t bufsiz;
5357
5358                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5359                         error = ENODEV;
5360                         break;
5361                 }
5362
5363                 if (io->pfiio_size < 0 ||
5364                     io->pfiio_size > pf_ioctl_maxcount ||
5365                     WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5366                         error = EINVAL;
5367                         break;
5368                 }
5369
5370                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5371                 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5372                     M_TEMP, M_NOWAIT);
5373                 if (ifstore == NULL) {
5374                         error = ENOMEM;
5375                         break;
5376                 }
5377
5378                 PF_RULES_RLOCK();
5379                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5380                 PF_RULES_RUNLOCK();
5381                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5382                 free(ifstore, M_TEMP);
5383                 break;
5384         }
5385
5386         case DIOCSETIFFLAG: {
5387                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5388
5389                 PF_RULES_WLOCK();
5390                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5391                 PF_RULES_WUNLOCK();
5392                 break;
5393         }
5394
5395         case DIOCCLRIFFLAG: {
5396                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5397
5398                 PF_RULES_WLOCK();
5399                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5400                 PF_RULES_WUNLOCK();
5401                 break;
5402         }
5403
5404         default:
5405                 error = ENODEV;
5406                 break;
5407         }
5408 fail:
5409         if (sx_xlocked(&pf_ioctl_lock))
5410                 sx_xunlock(&pf_ioctl_lock);
5411         CURVNET_RESTORE();
5412
5413 #undef ERROUT_IOCTL
5414
5415         return (error);
5416 }
5417
5418 void
5419 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
5420 {
5421         bzero(sp, sizeof(struct pfsync_state));
5422
5423         /* copy from state key */
5424         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5425         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5426         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5427         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5428         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5429         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5430         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5431         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5432         sp->proto = st->key[PF_SK_WIRE]->proto;
5433         sp->af = st->key[PF_SK_WIRE]->af;
5434
5435         /* copy from state */
5436         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5437         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5438         sp->creation = htonl(time_uptime - st->creation);
5439         sp->expire = pf_state_expires(st);
5440         if (sp->expire <= time_uptime)
5441                 sp->expire = htonl(0);
5442         else
5443                 sp->expire = htonl(sp->expire - time_uptime);
5444
5445         sp->direction = st->direction;
5446         sp->log = st->log;
5447         sp->timeout = st->timeout;
5448         sp->state_flags = st->state_flags;
5449         if (st->src_node)
5450                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5451         if (st->nat_src_node)
5452                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5453
5454         sp->id = st->id;
5455         sp->creatorid = st->creatorid;
5456         pf_state_peer_hton(&st->src, &sp->src);
5457         pf_state_peer_hton(&st->dst, &sp->dst);
5458
5459         if (st->rule.ptr == NULL)
5460                 sp->rule = htonl(-1);
5461         else
5462                 sp->rule = htonl(st->rule.ptr->nr);
5463         if (st->anchor.ptr == NULL)
5464                 sp->anchor = htonl(-1);
5465         else
5466                 sp->anchor = htonl(st->anchor.ptr->nr);
5467         if (st->nat_rule.ptr == NULL)
5468                 sp->nat_rule = htonl(-1);
5469         else
5470                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5471
5472         pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
5473             sp->packets[0]);
5474         pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
5475             sp->packets[1]);
5476         pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
5477         pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
5478
5479 }
5480
5481 static void
5482 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5483 {
5484         struct pfr_ktable *kt;
5485
5486         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5487
5488         kt = aw->p.tbl;
5489         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5490                 kt = kt->pfrkt_root;
5491         aw->p.tbl = NULL;
5492         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5493                 kt->pfrkt_cnt : -1;
5494 }
5495
5496 /*
5497  * XXX - Check for version missmatch!!!
5498  */
5499 static void
5500 pf_clear_all_states(void)
5501 {
5502         struct pf_state *s;
5503         u_int i;
5504
5505         for (i = 0; i <= pf_hashmask; i++) {
5506                 struct pf_idhash *ih = &V_pf_idhash[i];
5507 relock:
5508                 PF_HASHROW_LOCK(ih);
5509                 LIST_FOREACH(s, &ih->states, entry) {
5510                         s->timeout = PFTM_PURGE;
5511                         /* Don't send out individual delete messages. */
5512                         s->state_flags |= PFSTATE_NOSYNC;
5513                         pf_unlink_state(s, PF_ENTER_LOCKED);
5514                         goto relock;
5515                 }
5516                 PF_HASHROW_UNLOCK(ih);
5517         }
5518 }
5519
5520 static int
5521 pf_clear_tables(void)
5522 {
5523         struct pfioc_table io;
5524         int error;
5525
5526         bzero(&io, sizeof(io));
5527
5528         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5529             io.pfrio_flags);
5530
5531         return (error);
5532 }
5533
5534 static void
5535 pf_clear_srcnodes(struct pf_ksrc_node *n)
5536 {
5537         struct pf_state *s;
5538         int i;
5539
5540         for (i = 0; i <= pf_hashmask; i++) {
5541                 struct pf_idhash *ih = &V_pf_idhash[i];
5542
5543                 PF_HASHROW_LOCK(ih);
5544                 LIST_FOREACH(s, &ih->states, entry) {
5545                         if (n == NULL || n == s->src_node)
5546                                 s->src_node = NULL;
5547                         if (n == NULL || n == s->nat_src_node)
5548                                 s->nat_src_node = NULL;
5549                 }
5550                 PF_HASHROW_UNLOCK(ih);
5551         }
5552
5553         if (n == NULL) {
5554                 struct pf_srchash *sh;
5555
5556                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5557                     i++, sh++) {
5558                         PF_HASHROW_LOCK(sh);
5559                         LIST_FOREACH(n, &sh->nodes, entry) {
5560                                 n->expire = 1;
5561                                 n->states = 0;
5562                         }
5563                         PF_HASHROW_UNLOCK(sh);
5564                 }
5565         } else {
5566                 /* XXX: hash slot should already be locked here. */
5567                 n->expire = 1;
5568                 n->states = 0;
5569         }
5570 }
5571
5572 static void
5573 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5574 {
5575         struct pf_ksrc_node_list         kill;
5576
5577         LIST_INIT(&kill);
5578         for (int i = 0; i <= pf_srchashmask; i++) {
5579                 struct pf_srchash *sh = &V_pf_srchash[i];
5580                 struct pf_ksrc_node *sn, *tmp;
5581
5582                 PF_HASHROW_LOCK(sh);
5583                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5584                         if (PF_MATCHA(psnk->psnk_src.neg,
5585                               &psnk->psnk_src.addr.v.a.addr,
5586                               &psnk->psnk_src.addr.v.a.mask,
5587                               &sn->addr, sn->af) &&
5588                             PF_MATCHA(psnk->psnk_dst.neg,
5589                               &psnk->psnk_dst.addr.v.a.addr,
5590                               &psnk->psnk_dst.addr.v.a.mask,
5591                               &sn->raddr, sn->af)) {
5592                                 pf_unlink_src_node(sn);
5593                                 LIST_INSERT_HEAD(&kill, sn, entry);
5594                                 sn->expire = 1;
5595                         }
5596                 PF_HASHROW_UNLOCK(sh);
5597         }
5598
5599         for (int i = 0; i <= pf_hashmask; i++) {
5600                 struct pf_idhash *ih = &V_pf_idhash[i];
5601                 struct pf_state *s;
5602
5603                 PF_HASHROW_LOCK(ih);
5604                 LIST_FOREACH(s, &ih->states, entry) {
5605                         if (s->src_node && s->src_node->expire == 1)
5606                                 s->src_node = NULL;
5607                         if (s->nat_src_node && s->nat_src_node->expire == 1)
5608                                 s->nat_src_node = NULL;
5609                 }
5610                 PF_HASHROW_UNLOCK(ih);
5611         }
5612
5613         psnk->psnk_killed = pf_free_src_nodes(&kill);
5614 }
5615
5616 static int
5617 pf_keepcounters(struct pfioc_nv *nv)
5618 {
5619         nvlist_t        *nvl = NULL;
5620         void            *nvlpacked = NULL;
5621         int              error = 0;
5622
5623 #define ERROUT(x)       ERROUT_FUNCTION(on_error, x)
5624
5625         if (nv->len > pf_ioctl_maxcount)
5626                 ERROUT(ENOMEM);
5627
5628         nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5629         if (nvlpacked == NULL)
5630                 ERROUT(ENOMEM);
5631
5632         error = copyin(nv->data, nvlpacked, nv->len);
5633         if (error)
5634                 ERROUT(error);
5635
5636         nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5637         if (nvl == NULL)
5638                 ERROUT(EBADMSG);
5639
5640         if (! nvlist_exists_bool(nvl, "keep_counters"))
5641                 ERROUT(EBADMSG);
5642
5643         V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5644
5645 on_error:
5646         nvlist_destroy(nvl);
5647         free(nvlpacked, M_TEMP);
5648         return (error);
5649 }
5650
5651 static unsigned int
5652 pf_clear_states(const struct pf_kstate_kill *kill)
5653 {
5654         struct pf_state_key_cmp  match_key;
5655         struct pf_state *s;
5656         int              idx;
5657         unsigned int     killed = 0, dir;
5658
5659         for (unsigned int i = 0; i <= pf_hashmask; i++) {
5660                 struct pf_idhash *ih = &V_pf_idhash[i];
5661
5662 relock_DIOCCLRSTATES:
5663                 PF_HASHROW_LOCK(ih);
5664                 LIST_FOREACH(s, &ih->states, entry) {
5665                         if (kill->psk_ifname[0] &&
5666                             strcmp(kill->psk_ifname,
5667                             s->kif->pfik_name))
5668                                 continue;
5669
5670                         if (kill->psk_kill_match) {
5671                                 bzero(&match_key, sizeof(match_key));
5672
5673                                 if (s->direction == PF_OUT) {
5674                                         dir = PF_IN;
5675                                         idx = PF_SK_STACK;
5676                                 } else {
5677                                         dir = PF_OUT;
5678                                         idx = PF_SK_WIRE;
5679                                 }
5680
5681                                 match_key.af = s->key[idx]->af;
5682                                 match_key.proto = s->key[idx]->proto;
5683                                 PF_ACPY(&match_key.addr[0],
5684                                     &s->key[idx]->addr[1], match_key.af);
5685                                 match_key.port[0] = s->key[idx]->port[1];
5686                                 PF_ACPY(&match_key.addr[1],
5687                                     &s->key[idx]->addr[0], match_key.af);
5688                                 match_key.port[1] = s->key[idx]->port[0];
5689                         }
5690
5691                         /*
5692                          * Don't send out individual
5693                          * delete messages.
5694                          */
5695                         s->state_flags |= PFSTATE_NOSYNC;
5696                         pf_unlink_state(s, PF_ENTER_LOCKED);
5697                         killed++;
5698
5699                         if (kill->psk_kill_match)
5700                                 killed += pf_kill_matching_state(&match_key,
5701                                     dir);
5702
5703                         goto relock_DIOCCLRSTATES;
5704                 }
5705                 PF_HASHROW_UNLOCK(ih);
5706         }
5707
5708         if (V_pfsync_clear_states_ptr != NULL)
5709                 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
5710
5711         return (killed);
5712 }
5713
5714 static int
5715 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
5716 {
5717         struct pf_state         *s;
5718
5719         if (kill->psk_pfcmp.id) {
5720                 if (kill->psk_pfcmp.creatorid == 0)
5721                         kill->psk_pfcmp.creatorid = V_pf_status.hostid;
5722                 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
5723                     kill->psk_pfcmp.creatorid))) {
5724                         pf_unlink_state(s, PF_ENTER_LOCKED);
5725                         *killed = 1;
5726                 }
5727                 return (0);
5728         }
5729
5730         for (unsigned int i = 0; i <= pf_hashmask; i++)
5731                 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
5732
5733         return (0);
5734 }
5735
5736 static int
5737 pf_killstates_nv(struct pfioc_nv *nv)
5738 {
5739         struct pf_kstate_kill    kill;
5740         nvlist_t                *nvl = NULL;
5741         void                    *nvlpacked = NULL;
5742         int                      error = 0;
5743         unsigned int             killed = 0;
5744
5745 #define ERROUT(x)       ERROUT_FUNCTION(on_error, x)
5746
5747         if (nv->len > pf_ioctl_maxcount)
5748                 ERROUT(ENOMEM);
5749
5750         nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5751         if (nvlpacked == NULL)
5752                 ERROUT(ENOMEM);
5753
5754         error = copyin(nv->data, nvlpacked, nv->len);
5755         if (error)
5756                 ERROUT(error);
5757
5758         nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5759         if (nvl == NULL)
5760                 ERROUT(EBADMSG);
5761
5762         error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5763         if (error)
5764                 ERROUT(error);
5765
5766         error = pf_killstates(&kill, &killed);
5767
5768         free(nvlpacked, M_TEMP);
5769         nvlpacked = NULL;
5770         nvlist_destroy(nvl);
5771         nvl = nvlist_create(0);
5772         if (nvl == NULL)
5773                 ERROUT(ENOMEM);
5774
5775         nvlist_add_number(nvl, "killed", killed);
5776
5777         nvlpacked = nvlist_pack(nvl, &nv->len);
5778         if (nvlpacked == NULL)
5779                 ERROUT(ENOMEM);
5780
5781         if (nv->size == 0)
5782                 ERROUT(0);
5783         else if (nv->size < nv->len)
5784                 ERROUT(ENOSPC);
5785
5786         error = copyout(nvlpacked, nv->data, nv->len);
5787
5788 on_error:
5789         nvlist_destroy(nvl);
5790         free(nvlpacked, M_TEMP);
5791         return (error);
5792 }
5793
5794 static int
5795 pf_clearstates_nv(struct pfioc_nv *nv)
5796 {
5797         struct pf_kstate_kill    kill;
5798         nvlist_t                *nvl = NULL;
5799         void                    *nvlpacked = NULL;
5800         int                      error = 0;
5801         unsigned int             killed;
5802
5803 #define ERROUT(x)       ERROUT_FUNCTION(on_error, x)
5804
5805         if (nv->len > pf_ioctl_maxcount)
5806                 ERROUT(ENOMEM);
5807
5808         nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5809         if (nvlpacked == NULL)
5810                 ERROUT(ENOMEM);
5811
5812         error = copyin(nv->data, nvlpacked, nv->len);
5813         if (error)
5814                 ERROUT(error);
5815
5816         nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5817         if (nvl == NULL)
5818                 ERROUT(EBADMSG);
5819
5820         error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5821         if (error)
5822                 ERROUT(error);
5823
5824         killed = pf_clear_states(&kill);
5825
5826         free(nvlpacked, M_TEMP);
5827         nvlpacked = NULL;
5828         nvlist_destroy(nvl);
5829         nvl = nvlist_create(0);
5830         if (nvl == NULL)
5831                 ERROUT(ENOMEM);
5832
5833         nvlist_add_number(nvl, "killed", killed);
5834
5835         nvlpacked = nvlist_pack(nvl, &nv->len);
5836         if (nvlpacked == NULL)
5837                 ERROUT(ENOMEM);
5838
5839         if (nv->size == 0)
5840                 ERROUT(0);
5841         else if (nv->size < nv->len)
5842                 ERROUT(ENOSPC);
5843
5844         error = copyout(nvlpacked, nv->data, nv->len);
5845
5846 #undef ERROUT
5847 on_error:
5848         nvlist_destroy(nvl);
5849         free(nvlpacked, M_TEMP);
5850         return (error);
5851 }
5852
5853 static int
5854 pf_getstate(struct pfioc_nv *nv)
5855 {
5856         nvlist_t        *nvl = NULL, *nvls;
5857         void            *nvlpacked = NULL;
5858         struct pf_state *s = NULL;
5859         int              error = 0;
5860         uint64_t         id, creatorid;
5861
5862 #define ERROUT(x)       ERROUT_FUNCTION(errout, x)
5863
5864         if (nv->len > pf_ioctl_maxcount)
5865                 ERROUT(ENOMEM);
5866
5867         nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5868         if (nvlpacked == NULL)
5869                 ERROUT(ENOMEM);
5870
5871         error = copyin(nv->data, nvlpacked, nv->len);
5872         if (error)
5873                 ERROUT(error);
5874
5875         nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5876         if (nvl == NULL)
5877                 ERROUT(EBADMSG);
5878
5879         PFNV_CHK(pf_nvuint64(nvl, "id", &id));
5880         PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
5881
5882         s = pf_find_state_byid(id, creatorid);
5883         if (s == NULL)
5884                 ERROUT(ENOENT);
5885
5886         free(nvlpacked, M_TEMP);
5887         nvlpacked = NULL;
5888         nvlist_destroy(nvl);
5889         nvl = nvlist_create(0);
5890         if (nvl == NULL)
5891                 ERROUT(ENOMEM);
5892
5893         nvls = pf_state_to_nvstate(s);
5894         if (nvls == NULL)
5895                 ERROUT(ENOMEM);
5896
5897         nvlist_add_nvlist(nvl, "state", nvls);
5898
5899         nvlpacked = nvlist_pack(nvl, &nv->len);
5900         if (nvlpacked == NULL)
5901                 ERROUT(ENOMEM);
5902
5903         if (nv->size == 0)
5904                 ERROUT(0);
5905         else if (nv->size < nv->len)
5906                 ERROUT(ENOSPC);
5907
5908         error = copyout(nvlpacked, nv->data, nv->len);
5909
5910 #undef ERROUT
5911 errout:
5912         if (s != NULL)
5913                 PF_STATE_UNLOCK(s);
5914         free(nvlpacked, M_TEMP);
5915         nvlist_destroy(nvl);
5916         return (error);
5917 }
5918
5919 /*
5920  * XXX - Check for version missmatch!!!
5921  */
5922
5923 /*
5924  * Duplicate pfctl -Fa operation to get rid of as much as we can.
5925  */
5926 static int
5927 shutdown_pf(void)
5928 {
5929         int error = 0;
5930         u_int32_t t[5];
5931         char nn = '\0';
5932
5933         do {
5934                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
5935                     != 0) {
5936                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
5937                         break;
5938                 }
5939                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
5940                     != 0) {
5941                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
5942                         break;          /* XXX: rollback? */
5943                 }
5944                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
5945                     != 0) {
5946                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
5947                         break;          /* XXX: rollback? */
5948                 }
5949                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
5950                     != 0) {
5951                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
5952                         break;          /* XXX: rollback? */
5953                 }
5954                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
5955                     != 0) {
5956                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
5957                         break;          /* XXX: rollback? */
5958                 }
5959
5960                 /* XXX: these should always succeed here */
5961                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
5962                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
5963                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
5964                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
5965                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
5966
5967                 if ((error = pf_clear_tables()) != 0)
5968                         break;
5969
5970 #ifdef ALTQ
5971                 if ((error = pf_begin_altq(&t[0])) != 0) {
5972                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
5973                         break;
5974                 }
5975                 pf_commit_altq(t[0]);
5976 #endif
5977
5978                 pf_clear_all_states();
5979
5980                 pf_clear_srcnodes(NULL);
5981
5982                 /* status does not use malloced mem so no need to cleanup */
5983                 /* fingerprints and interfaces have their own cleanup code */
5984         } while(0);
5985
5986         return (error);
5987 }
5988
5989 static pfil_return_t
5990 pf_check_return(int chk, struct mbuf **m)
5991 {
5992
5993         switch (chk) {
5994         case PF_PASS:
5995                 if (*m == NULL)
5996                         return (PFIL_CONSUMED);
5997                 else
5998                         return (PFIL_PASS);
5999                 break;
6000         default:
6001                 if (*m != NULL) {
6002                         m_freem(*m);
6003                         *m = NULL;
6004                 }
6005                 return (PFIL_DROPPED);
6006         }
6007 }
6008
6009 #ifdef INET
6010 static pfil_return_t
6011 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6012     void *ruleset __unused, struct inpcb *inp)
6013 {
6014         int chk;
6015
6016         chk = pf_test(PF_IN, flags, ifp, m, inp);
6017
6018         return (pf_check_return(chk, m));
6019 }
6020
6021 static pfil_return_t
6022 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6023     void *ruleset __unused,  struct inpcb *inp)
6024 {
6025         int chk;
6026
6027         chk = pf_test(PF_OUT, flags, ifp, m, inp);
6028
6029         return (pf_check_return(chk, m));
6030 }
6031 #endif
6032
6033 #ifdef INET6
6034 static pfil_return_t
6035 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6036     void *ruleset __unused,  struct inpcb *inp)
6037 {
6038         int chk;
6039
6040         /*
6041          * In case of loopback traffic IPv6 uses the real interface in
6042          * order to support scoped addresses. In order to support stateful
6043          * filtering we have change this to lo0 as it is the case in IPv4.
6044          */
6045         CURVNET_SET(ifp->if_vnet);
6046         chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
6047         CURVNET_RESTORE();
6048
6049         return (pf_check_return(chk, m));
6050 }
6051
6052 static pfil_return_t
6053 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6054     void *ruleset __unused,  struct inpcb *inp)
6055 {
6056         int chk;
6057
6058         CURVNET_SET(ifp->if_vnet);
6059         chk = pf_test6(PF_OUT, flags, ifp, m, inp);
6060         CURVNET_RESTORE();
6061
6062         return (pf_check_return(chk, m));
6063 }
6064 #endif /* INET6 */
6065
6066 #ifdef INET
6067 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6068 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6069 #define V_pf_ip4_in_hook        VNET(pf_ip4_in_hook)
6070 #define V_pf_ip4_out_hook       VNET(pf_ip4_out_hook)
6071 #endif
6072 #ifdef INET6
6073 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6074 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6075 #define V_pf_ip6_in_hook        VNET(pf_ip6_in_hook)
6076 #define V_pf_ip6_out_hook       VNET(pf_ip6_out_hook)
6077 #endif
6078
6079 static void
6080 hook_pf(void)
6081 {
6082         struct pfil_hook_args pha;
6083         struct pfil_link_args pla;
6084         int ret;
6085
6086         if (V_pf_pfil_hooked)
6087                 return;
6088
6089         pha.pa_version = PFIL_VERSION;
6090         pha.pa_modname = "pf";
6091         pha.pa_ruleset = NULL;
6092
6093         pla.pa_version = PFIL_VERSION;
6094
6095 #ifdef INET
6096         pha.pa_type = PFIL_TYPE_IP4;
6097         pha.pa_func = pf_check_in;
6098         pha.pa_flags = PFIL_IN;
6099         pha.pa_rulname = "default-in";
6100         V_pf_ip4_in_hook = pfil_add_hook(&pha);
6101         pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6102         pla.pa_head = V_inet_pfil_head;
6103         pla.pa_hook = V_pf_ip4_in_hook;
6104         ret = pfil_link(&pla);
6105         MPASS(ret == 0);
6106         pha.pa_func = pf_check_out;
6107         pha.pa_flags = PFIL_OUT;
6108         pha.pa_rulname = "default-out";
6109         V_pf_ip4_out_hook = pfil_add_hook(&pha);
6110         pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6111         pla.pa_head = V_inet_pfil_head;
6112         pla.pa_hook = V_pf_ip4_out_hook;
6113         ret = pfil_link(&pla);
6114         MPASS(ret == 0);
6115 #endif
6116 #ifdef INET6
6117         pha.pa_type = PFIL_TYPE_IP6;
6118         pha.pa_func = pf_check6_in;
6119         pha.pa_flags = PFIL_IN;
6120         pha.pa_rulname = "default-in6";
6121         V_pf_ip6_in_hook = pfil_add_hook(&pha);
6122         pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6123         pla.pa_head = V_inet6_pfil_head;
6124         pla.pa_hook = V_pf_ip6_in_hook;
6125         ret = pfil_link(&pla);
6126         MPASS(ret == 0);
6127         pha.pa_func = pf_check6_out;
6128         pha.pa_rulname = "default-out6";
6129         pha.pa_flags = PFIL_OUT;
6130         V_pf_ip6_out_hook = pfil_add_hook(&pha);
6131         pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6132         pla.pa_head = V_inet6_pfil_head;
6133         pla.pa_hook = V_pf_ip6_out_hook;
6134         ret = pfil_link(&pla);
6135         MPASS(ret == 0);
6136 #endif
6137
6138         V_pf_pfil_hooked = 1;
6139 }
6140
6141 static void
6142 dehook_pf(void)
6143 {
6144
6145         if (V_pf_pfil_hooked == 0)
6146                 return;
6147
6148 #ifdef INET
6149         pfil_remove_hook(V_pf_ip4_in_hook);
6150         pfil_remove_hook(V_pf_ip4_out_hook);
6151 #endif
6152 #ifdef INET6
6153         pfil_remove_hook(V_pf_ip6_in_hook);
6154         pfil_remove_hook(V_pf_ip6_out_hook);
6155 #endif
6156
6157         V_pf_pfil_hooked = 0;
6158 }
6159
6160 static void
6161 pf_load_vnet(void)
6162 {
6163         V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6164             NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6165
6166         pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6167             PF_RULE_TAG_HASH_SIZE_DEFAULT);
6168 #ifdef ALTQ
6169         pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6170             PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6171 #endif
6172
6173         pfattach_vnet();
6174         V_pf_vnet_active = 1;
6175 }
6176
6177 static int
6178 pf_load(void)
6179 {
6180         int error;
6181
6182         rm_init(&pf_rules_lock, "pf rulesets");
6183         sx_init(&pf_ioctl_lock, "pf ioctl");
6184         sx_init(&pf_end_lock, "pf end thread");
6185
6186         pf_mtag_initialize();
6187
6188         pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6189         if (pf_dev == NULL)
6190                 return (ENOMEM);
6191
6192         pf_end_threads = 0;
6193         error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6194         if (error != 0)
6195                 return (error);
6196
6197         pfi_initialize();
6198
6199         return (0);
6200 }
6201
6202 static void
6203 pf_unload_vnet(void)
6204 {
6205         int ret;
6206
6207         V_pf_vnet_active = 0;
6208         V_pf_status.running = 0;
6209         dehook_pf();
6210
6211         PF_RULES_WLOCK();
6212         shutdown_pf();
6213         PF_RULES_WUNLOCK();
6214
6215         ret = swi_remove(V_pf_swi_cookie);
6216         MPASS(ret == 0);
6217         ret = intr_event_destroy(V_pf_swi_ie);
6218         MPASS(ret == 0);
6219
6220         pf_unload_vnet_purge();
6221
6222         pf_normalize_cleanup();
6223         PF_RULES_WLOCK();
6224         pfi_cleanup_vnet();
6225         PF_RULES_WUNLOCK();
6226         pfr_cleanup();
6227         pf_osfp_flush();
6228         pf_cleanup();
6229         if (IS_DEFAULT_VNET(curvnet))
6230                 pf_mtag_cleanup();
6231
6232         pf_cleanup_tagset(&V_pf_tags);
6233 #ifdef ALTQ
6234         pf_cleanup_tagset(&V_pf_qids);
6235 #endif
6236         uma_zdestroy(V_pf_tag_z);
6237
6238         /* Free counters last as we updated them during shutdown. */
6239         counter_u64_free(V_pf_default_rule.evaluations);
6240         for (int i = 0; i < 2; i++) {
6241                 counter_u64_free(V_pf_default_rule.packets[i]);
6242                 counter_u64_free(V_pf_default_rule.bytes[i]);
6243         }
6244         counter_u64_free(V_pf_default_rule.states_cur);
6245         counter_u64_free(V_pf_default_rule.states_tot);
6246         counter_u64_free(V_pf_default_rule.src_nodes);
6247
6248         for (int i = 0; i < PFRES_MAX; i++)
6249                 counter_u64_free(V_pf_status.counters[i]);
6250         for (int i = 0; i < LCNT_MAX; i++)
6251                 counter_u64_free(V_pf_status.lcounters[i]);
6252         for (int i = 0; i < FCNT_MAX; i++)
6253                 counter_u64_free(V_pf_status.fcounters[i]);
6254         for (int i = 0; i < SCNT_MAX; i++)
6255                 counter_u64_free(V_pf_status.scounters[i]);
6256 }
6257
6258 static void
6259 pf_unload(void)
6260 {
6261
6262         sx_xlock(&pf_end_lock);
6263         pf_end_threads = 1;
6264         while (pf_end_threads < 2) {
6265                 wakeup_one(pf_purge_thread);
6266                 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6267         }
6268         sx_xunlock(&pf_end_lock);
6269
6270         if (pf_dev != NULL)
6271                 destroy_dev(pf_dev);
6272
6273         pfi_cleanup();
6274
6275         rm_destroy(&pf_rules_lock);
6276         sx_destroy(&pf_ioctl_lock);
6277         sx_destroy(&pf_end_lock);
6278 }
6279
6280 static void
6281 vnet_pf_init(void *unused __unused)
6282 {
6283
6284         pf_load_vnet();
6285 }
6286 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6287     vnet_pf_init, NULL);
6288
6289 static void
6290 vnet_pf_uninit(const void *unused __unused)
6291 {
6292
6293         pf_unload_vnet();
6294 }
6295 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6296 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6297     vnet_pf_uninit, NULL);
6298
6299 static int
6300 pf_modevent(module_t mod, int type, void *data)
6301 {
6302         int error = 0;
6303
6304         switch(type) {
6305         case MOD_LOAD:
6306                 error = pf_load();
6307                 break;
6308         case MOD_UNLOAD:
6309                 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6310                  * the vnet_pf_uninit()s */
6311                 break;
6312         default:
6313                 error = EINVAL;
6314                 break;
6315         }
6316
6317         return (error);
6318 }
6319
6320 static moduledata_t pf_mod = {
6321         "pf",
6322         pf_modevent,
6323         0
6324 };
6325
6326 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6327 MODULE_VERSION(pf, PF_MODVER);