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 #include "opt_inet.h" 42 #include "opt_inet6.h" 43 #include "opt_bpf.h" 44 #include "opt_pf.h" 45 46 #include <sys/param.h> 47 #include <sys/_bitset.h> 48 #include <sys/bitset.h> 49 #include <sys/bus.h> 50 #include <sys/conf.h> 51 #include <sys/endian.h> 52 #include <sys/fcntl.h> 53 #include <sys/filio.h> 54 #include <sys/hash.h> 55 #include <sys/interrupt.h> 56 #include <sys/jail.h> 57 #include <sys/kernel.h> 58 #include <sys/kthread.h> 59 #include <sys/lock.h> 60 #include <sys/mbuf.h> 61 #include <sys/module.h> 62 #include <sys/nv.h> 63 #include <sys/proc.h> 64 #include <sys/sdt.h> 65 #include <sys/smp.h> 66 #include <sys/socket.h> 67 #include <sys/sysctl.h> 68 #include <sys/md5.h> 69 #include <sys/ucred.h> 70 71 #include <net/if.h> 72 #include <net/if_var.h> 73 #include <net/if_private.h> 74 #include <net/vnet.h> 75 #include <net/route.h> 76 #include <net/pfil.h> 77 #include <net/pfvar.h> 78 #include <net/if_pfsync.h> 79 #include <net/if_pflog.h> 80 81 #include <netinet/in.h> 82 #include <netinet/ip.h> 83 #include <netinet/ip_var.h> 84 #include <netinet6/ip6_var.h> 85 #include <netinet/ip_icmp.h> 86 #include <netpfil/pf/pf_nl.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_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int"); 98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int"); 99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int"); 100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int"); 101 102 static struct pf_kpool *pf_get_kpool(const char *, u_int32_t, u_int8_t, 103 u_int32_t, u_int8_t, u_int8_t, u_int8_t, int); 104 105 static void pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *); 106 static void pf_empty_kpool(struct pf_kpalist *); 107 static int pfioctl(struct cdev *, u_long, caddr_t, int, 108 struct thread *); 109 static int pf_begin_eth(uint32_t *, const char *); 110 static int pf_rollback_eth(uint32_t, const char *); 111 static int pf_commit_eth(uint32_t, const char *); 112 static void pf_free_eth_rule(struct pf_keth_rule *); 113 #ifdef ALTQ 114 static int pf_begin_altq(u_int32_t *); 115 static int pf_rollback_altq(u_int32_t); 116 static int pf_commit_altq(u_int32_t); 117 static int pf_enable_altq(struct pf_altq *); 118 static int pf_disable_altq(struct pf_altq *); 119 static uint16_t pf_qname2qid(const char *); 120 static void pf_qid_unref(uint16_t); 121 #endif /* ALTQ */ 122 static int pf_begin_rules(u_int32_t *, int, const char *); 123 static int pf_rollback_rules(u_int32_t, int, char *); 124 static int pf_setup_pfsync_matching(struct pf_kruleset *); 125 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *); 126 static void pf_hash_rule(struct pf_krule *); 127 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *); 128 static int pf_commit_rules(u_int32_t, int, char *); 129 static int pf_addr_setup(struct pf_kruleset *, 130 struct pf_addr_wrap *, sa_family_t); 131 static void pf_src_node_copy(const struct pf_ksrc_node *, 132 struct pf_src_node *); 133 #ifdef ALTQ 134 static int pf_export_kaltq(struct pf_altq *, 135 struct pfioc_altq_v1 *, size_t); 136 static int pf_import_kaltq(struct pfioc_altq_v1 *, 137 struct pf_altq *, size_t); 138 #endif /* ALTQ */ 139 140 VNET_DEFINE(struct pf_krule, pf_default_rule); 141 142 static __inline int pf_krule_compare(struct pf_krule *, 143 struct pf_krule *); 144 145 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare); 146 147 #ifdef ALTQ 148 VNET_DEFINE_STATIC(int, pf_altq_running); 149 #define V_pf_altq_running VNET(pf_altq_running) 150 #endif 151 152 #define TAGID_MAX 50000 153 struct pf_tagname { 154 TAILQ_ENTRY(pf_tagname) namehash_entries; 155 TAILQ_ENTRY(pf_tagname) taghash_entries; 156 char name[PF_TAG_NAME_SIZE]; 157 uint16_t tag; 158 int ref; 159 }; 160 161 struct pf_tagset { 162 TAILQ_HEAD(, pf_tagname) *namehash; 163 TAILQ_HEAD(, pf_tagname) *taghash; 164 unsigned int mask; 165 uint32_t seed; 166 BITSET_DEFINE(, TAGID_MAX) avail; 167 }; 168 169 VNET_DEFINE(struct pf_tagset, pf_tags); 170 #define V_pf_tags VNET(pf_tags) 171 static unsigned int pf_rule_tag_hashsize; 172 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128 173 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN, 174 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT, 175 "Size of pf(4) rule tag hashtable"); 176 177 #ifdef ALTQ 178 VNET_DEFINE(struct pf_tagset, pf_qids); 179 #define V_pf_qids VNET(pf_qids) 180 static unsigned int pf_queue_tag_hashsize; 181 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128 182 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN, 183 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT, 184 "Size of pf(4) queue tag hashtable"); 185 #endif 186 VNET_DEFINE(uma_zone_t, pf_tag_z); 187 #define V_pf_tag_z VNET(pf_tag_z) 188 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db"); 189 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules"); 190 191 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE) 192 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE 193 #endif 194 195 VNET_DEFINE_STATIC(bool, pf_filter_local) = false; 196 #define V_pf_filter_local VNET(pf_filter_local) 197 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW, 198 &VNET_NAME(pf_filter_local), false, 199 "Enable filtering for packets delivered to local network stack"); 200 201 #ifdef PF_DEFAULT_TO_DROP 202 VNET_DEFINE_STATIC(bool, default_to_drop) = true; 203 #else 204 VNET_DEFINE_STATIC(bool, default_to_drop); 205 #endif 206 #define V_default_to_drop VNET(default_to_drop) 207 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET, 208 &VNET_NAME(default_to_drop), false, 209 "Make the default rule drop all packets."); 210 211 static void pf_init_tagset(struct pf_tagset *, unsigned int *, 212 unsigned int); 213 static void pf_cleanup_tagset(struct pf_tagset *); 214 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *); 215 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t); 216 static u_int16_t tagname2tag(struct pf_tagset *, const char *); 217 static u_int16_t pf_tagname2tag(const char *); 218 static void tag_unref(struct pf_tagset *, u_int16_t); 219 220 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x 221 222 struct cdev *pf_dev; 223 224 /* 225 * XXX - These are new and need to be checked when moveing to a new version 226 */ 227 static void pf_clear_all_states(void); 228 static int pf_killstates_row(struct pf_kstate_kill *, 229 struct pf_idhash *); 230 static int pf_killstates_nv(struct pfioc_nv *); 231 static int pf_clearstates_nv(struct pfioc_nv *); 232 static int pf_getstate(struct pfioc_nv *); 233 static int pf_getstatus(struct pfioc_nv *); 234 static int pf_clear_tables(void); 235 static void pf_kill_srcnodes(struct pfioc_src_node_kill *); 236 static int pf_keepcounters(struct pfioc_nv *); 237 static void pf_tbladdr_copyout(struct pf_addr_wrap *); 238 239 /* 240 * Wrapper functions for pfil(9) hooks 241 */ 242 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, 243 int flags, void *ruleset __unused, struct inpcb *inp); 244 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, 245 int flags, void *ruleset __unused, struct inpcb *inp); 246 #ifdef INET 247 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp, 248 int flags, void *ruleset __unused, struct inpcb *inp); 249 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp, 250 int flags, void *ruleset __unused, struct inpcb *inp); 251 #endif 252 #ifdef INET6 253 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp, 254 int flags, void *ruleset __unused, struct inpcb *inp); 255 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp, 256 int flags, void *ruleset __unused, struct inpcb *inp); 257 #endif 258 259 static void hook_pf_eth(void); 260 static void hook_pf(void); 261 static void dehook_pf_eth(void); 262 static void dehook_pf(void); 263 static int shutdown_pf(void); 264 static int pf_load(void); 265 static void pf_unload(void); 266 267 static struct cdevsw pf_cdevsw = { 268 .d_ioctl = pfioctl, 269 .d_name = PF_NAME, 270 .d_version = D_VERSION, 271 }; 272 273 VNET_DEFINE_STATIC(bool, pf_pfil_hooked); 274 #define V_pf_pfil_hooked VNET(pf_pfil_hooked) 275 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked); 276 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked) 277 278 /* 279 * We need a flag that is neither hooked nor running to know when 280 * the VNET is "valid". We primarily need this to control (global) 281 * external event, e.g., eventhandlers. 282 */ 283 VNET_DEFINE(int, pf_vnet_active); 284 #define V_pf_vnet_active VNET(pf_vnet_active) 285 286 int pf_end_threads; 287 struct proc *pf_purge_proc; 288 289 VNET_DEFINE(struct rmlock, pf_rules_lock); 290 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock); 291 #define V_pf_ioctl_lock VNET(pf_ioctl_lock) 292 struct sx pf_end_lock; 293 294 /* pfsync */ 295 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr); 296 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr); 297 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr); 298 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr); 299 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr); 300 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr); 301 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr); 302 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr; 303 304 /* pflog */ 305 pflog_packet_t *pflog_packet_ptr = NULL; 306 307 /* 308 * Copy a user-provided string, returning an error if truncation would occur. 309 * Avoid scanning past "sz" bytes in the source string since there's no 310 * guarantee that it's nul-terminated. 311 */ 312 static int 313 pf_user_strcpy(char *dst, const char *src, size_t sz) 314 { 315 if (strnlen(src, sz) == sz) 316 return (EINVAL); 317 (void)strlcpy(dst, src, sz); 318 return (0); 319 } 320 321 static void 322 pfattach_vnet(void) 323 { 324 u_int32_t *my_timeout = V_pf_default_rule.timeout; 325 326 bzero(&V_pf_status, sizeof(V_pf_status)); 327 328 pf_initialize(); 329 pfr_initialize(); 330 pfi_initialize_vnet(); 331 pf_normalize_init(); 332 pf_syncookies_init(); 333 334 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT; 335 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT; 336 337 RB_INIT(&V_pf_anchors); 338 pf_init_kruleset(&pf_main_ruleset); 339 340 pf_init_keth(V_pf_keth); 341 342 /* default rule should never be garbage collected */ 343 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; 344 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS; 345 V_pf_default_rule.nr = (uint32_t)-1; 346 V_pf_default_rule.rtableid = -1; 347 348 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK); 349 for (int i = 0; i < 2; i++) { 350 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK); 351 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK); 352 } 353 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); 354 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); 355 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++) 356 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 357 358 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 359 M_WAITOK | M_ZERO); 360 361 #ifdef PF_WANT_32_TO_64_COUNTER 362 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO); 363 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO); 364 PF_RULES_WLOCK(); 365 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); 366 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist); 367 V_pf_allrulecount++; 368 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); 369 PF_RULES_WUNLOCK(); 370 #endif 371 372 /* initialize default timeouts */ 373 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 374 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; 375 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 376 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; 377 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; 378 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; 379 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 380 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL; 381 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 382 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL; 383 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL; 384 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; 385 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; 386 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; 387 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; 388 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; 389 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; 390 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; 391 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; 392 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; 393 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; 394 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; 395 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; 396 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; 397 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; 398 399 V_pf_status.debug = PF_DEBUG_URGENT; 400 /* 401 * XXX This is different than in OpenBSD where reassembly is enabled by 402 * defult. In FreeBSD we expect people to still use scrub rules and 403 * switch to the new syntax later. Only when they switch they must 404 * explicitly enable reassemle. We could change the default once the 405 * scrub rule functionality is hopefully removed some day in future. 406 */ 407 V_pf_status.reass = 0; 408 409 V_pf_pfil_hooked = false; 410 V_pf_pfil_eth_hooked = false; 411 412 /* XXX do our best to avoid a conflict */ 413 V_pf_status.hostid = arc4random(); 414 415 for (int i = 0; i < PFRES_MAX; i++) 416 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); 417 for (int i = 0; i < KLCNT_MAX; i++) 418 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); 419 for (int i = 0; i < FCNT_MAX; i++) 420 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK); 421 for (int i = 0; i < SCNT_MAX; i++) 422 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); 423 424 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET, 425 INTR_MPSAFE, &V_pf_swi_cookie) != 0) 426 /* XXXGL: leaked all above. */ 427 return; 428 } 429 430 static struct pf_kpool * 431 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action, 432 u_int32_t rule_number, u_int8_t r_last, u_int8_t active, 433 u_int8_t check_ticket, int which) 434 { 435 struct pf_kruleset *ruleset; 436 struct pf_krule *rule; 437 int rs_num; 438 439 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT); 440 441 ruleset = pf_find_kruleset(anchor); 442 if (ruleset == NULL) 443 return (NULL); 444 rs_num = pf_get_ruleset_number(rule_action); 445 if (rs_num >= PF_RULESET_MAX) 446 return (NULL); 447 if (active) { 448 if (check_ticket && ticket != 449 ruleset->rules[rs_num].active.ticket) 450 return (NULL); 451 if (r_last) 452 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 453 pf_krulequeue); 454 else 455 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 456 } else { 457 if (check_ticket && ticket != 458 ruleset->rules[rs_num].inactive.ticket) 459 return (NULL); 460 if (r_last) 461 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 462 pf_krulequeue); 463 else 464 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); 465 } 466 if (!r_last) { 467 while ((rule != NULL) && (rule->nr != rule_number)) 468 rule = TAILQ_NEXT(rule, entries); 469 } 470 if (rule == NULL) 471 return (NULL); 472 473 switch (which) { 474 case PF_RDR: 475 return (&rule->rdr); 476 case PF_NAT: 477 return (&rule->nat); 478 case PF_RT: 479 return (&rule->route); 480 default: 481 panic("Unknow pool type %d", which); 482 } 483 } 484 485 static void 486 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb) 487 { 488 struct pf_kpooladdr *mv_pool_pa; 489 490 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { 491 TAILQ_REMOVE(poola, mv_pool_pa, entries); 492 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); 493 } 494 } 495 496 static void 497 pf_empty_kpool(struct pf_kpalist *poola) 498 { 499 struct pf_kpooladdr *pa; 500 501 while ((pa = TAILQ_FIRST(poola)) != NULL) { 502 switch (pa->addr.type) { 503 case PF_ADDR_DYNIFTL: 504 pfi_dynaddr_remove(pa->addr.p.dyn); 505 break; 506 case PF_ADDR_TABLE: 507 /* XXX: this could be unfinished pooladdr on pabuf */ 508 if (pa->addr.p.tbl != NULL) 509 pfr_detach_table(pa->addr.p.tbl); 510 break; 511 } 512 if (pa->kif) 513 pfi_kkif_unref(pa->kif); 514 TAILQ_REMOVE(poola, pa, entries); 515 free(pa, M_PFRULE); 516 } 517 } 518 519 static void 520 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 521 { 522 523 PF_RULES_WASSERT(); 524 PF_UNLNKDRULES_ASSERT(); 525 526 TAILQ_REMOVE(rulequeue, rule, entries); 527 528 rule->rule_ref |= PFRULE_REFS; 529 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); 530 } 531 532 static void 533 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 534 { 535 536 PF_RULES_WASSERT(); 537 538 PF_UNLNKDRULES_LOCK(); 539 pf_unlink_rule_locked(rulequeue, rule); 540 PF_UNLNKDRULES_UNLOCK(); 541 } 542 543 static void 544 pf_free_eth_rule(struct pf_keth_rule *rule) 545 { 546 PF_RULES_WASSERT(); 547 548 if (rule == NULL) 549 return; 550 551 if (rule->tag) 552 tag_unref(&V_pf_tags, rule->tag); 553 if (rule->match_tag) 554 tag_unref(&V_pf_tags, rule->match_tag); 555 #ifdef ALTQ 556 pf_qid_unref(rule->qid); 557 #endif 558 559 if (rule->bridge_to) 560 pfi_kkif_unref(rule->bridge_to); 561 if (rule->kif) 562 pfi_kkif_unref(rule->kif); 563 564 if (rule->ipsrc.addr.type == PF_ADDR_TABLE) 565 pfr_detach_table(rule->ipsrc.addr.p.tbl); 566 if (rule->ipdst.addr.type == PF_ADDR_TABLE) 567 pfr_detach_table(rule->ipdst.addr.p.tbl); 568 569 counter_u64_free(rule->evaluations); 570 for (int i = 0; i < 2; i++) { 571 counter_u64_free(rule->packets[i]); 572 counter_u64_free(rule->bytes[i]); 573 } 574 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 575 pf_keth_anchor_remove(rule); 576 577 free(rule, M_PFRULE); 578 } 579 580 void 581 pf_free_rule(struct pf_krule *rule) 582 { 583 584 PF_RULES_WASSERT(); 585 PF_CONFIG_ASSERT(); 586 587 if (rule->tag) 588 tag_unref(&V_pf_tags, rule->tag); 589 if (rule->match_tag) 590 tag_unref(&V_pf_tags, rule->match_tag); 591 #ifdef ALTQ 592 if (rule->pqid != rule->qid) 593 pf_qid_unref(rule->pqid); 594 pf_qid_unref(rule->qid); 595 #endif 596 switch (rule->src.addr.type) { 597 case PF_ADDR_DYNIFTL: 598 pfi_dynaddr_remove(rule->src.addr.p.dyn); 599 break; 600 case PF_ADDR_TABLE: 601 pfr_detach_table(rule->src.addr.p.tbl); 602 break; 603 } 604 switch (rule->dst.addr.type) { 605 case PF_ADDR_DYNIFTL: 606 pfi_dynaddr_remove(rule->dst.addr.p.dyn); 607 break; 608 case PF_ADDR_TABLE: 609 pfr_detach_table(rule->dst.addr.p.tbl); 610 break; 611 } 612 if (rule->overload_tbl) 613 pfr_detach_table(rule->overload_tbl); 614 if (rule->kif) 615 pfi_kkif_unref(rule->kif); 616 if (rule->rcv_kif) 617 pfi_kkif_unref(rule->rcv_kif); 618 pf_kanchor_remove(rule); 619 pf_empty_kpool(&rule->rdr.list); 620 pf_empty_kpool(&rule->nat.list); 621 pf_empty_kpool(&rule->route.list); 622 623 pf_krule_free(rule); 624 } 625 626 static void 627 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size, 628 unsigned int default_size) 629 { 630 unsigned int i; 631 unsigned int hashsize; 632 633 if (*tunable_size == 0 || !powerof2(*tunable_size)) 634 *tunable_size = default_size; 635 636 hashsize = *tunable_size; 637 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH, 638 M_WAITOK); 639 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH, 640 M_WAITOK); 641 ts->mask = hashsize - 1; 642 ts->seed = arc4random(); 643 for (i = 0; i < hashsize; i++) { 644 TAILQ_INIT(&ts->namehash[i]); 645 TAILQ_INIT(&ts->taghash[i]); 646 } 647 BIT_FILL(TAGID_MAX, &ts->avail); 648 } 649 650 static void 651 pf_cleanup_tagset(struct pf_tagset *ts) 652 { 653 unsigned int i; 654 unsigned int hashsize; 655 struct pf_tagname *t, *tmp; 656 657 /* 658 * Only need to clean up one of the hashes as each tag is hashed 659 * into each table. 660 */ 661 hashsize = ts->mask + 1; 662 for (i = 0; i < hashsize; i++) 663 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp) 664 uma_zfree(V_pf_tag_z, t); 665 666 free(ts->namehash, M_PFHASH); 667 free(ts->taghash, M_PFHASH); 668 } 669 670 static uint16_t 671 tagname2hashindex(const struct pf_tagset *ts, const char *tagname) 672 { 673 size_t len; 674 675 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1); 676 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask); 677 } 678 679 static uint16_t 680 tag2hashindex(const struct pf_tagset *ts, uint16_t tag) 681 { 682 683 return (tag & ts->mask); 684 } 685 686 static u_int16_t 687 tagname2tag(struct pf_tagset *ts, const char *tagname) 688 { 689 struct pf_tagname *tag; 690 u_int32_t index; 691 u_int16_t new_tagid; 692 693 PF_RULES_WASSERT(); 694 695 index = tagname2hashindex(ts, tagname); 696 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 697 if (strcmp(tagname, tag->name) == 0) { 698 tag->ref++; 699 return (tag->tag); 700 } 701 702 /* 703 * new entry 704 * 705 * to avoid fragmentation, we do a linear search from the beginning 706 * and take the first free slot we find. 707 */ 708 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail); 709 /* 710 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX]. 711 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits 712 * set. It may also return a bit number greater than TAGID_MAX due 713 * to rounding of the number of bits in the vector up to a multiple 714 * of the vector word size at declaration/allocation time. 715 */ 716 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) 717 return (0); 718 719 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */ 720 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail); 721 722 /* allocate and fill new struct pf_tagname */ 723 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT); 724 if (tag == NULL) 725 return (0); 726 strlcpy(tag->name, tagname, sizeof(tag->name)); 727 tag->tag = new_tagid; 728 tag->ref = 1; 729 730 /* Insert into namehash */ 731 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries); 732 733 /* Insert into taghash */ 734 index = tag2hashindex(ts, new_tagid); 735 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries); 736 737 return (tag->tag); 738 } 739 740 static void 741 tag_unref(struct pf_tagset *ts, u_int16_t tag) 742 { 743 struct pf_tagname *t; 744 uint16_t index; 745 746 PF_RULES_WASSERT(); 747 748 index = tag2hashindex(ts, tag); 749 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 750 if (tag == t->tag) { 751 if (--t->ref == 0) { 752 TAILQ_REMOVE(&ts->taghash[index], t, 753 taghash_entries); 754 index = tagname2hashindex(ts, t->name); 755 TAILQ_REMOVE(&ts->namehash[index], t, 756 namehash_entries); 757 /* Bits are 0-based for BIT_SET() */ 758 BIT_SET(TAGID_MAX, tag - 1, &ts->avail); 759 uma_zfree(V_pf_tag_z, t); 760 } 761 break; 762 } 763 } 764 765 static uint16_t 766 pf_tagname2tag(const char *tagname) 767 { 768 return (tagname2tag(&V_pf_tags, tagname)); 769 } 770 771 static int 772 pf_begin_eth(uint32_t *ticket, const char *anchor) 773 { 774 struct pf_keth_rule *rule, *tmp; 775 struct pf_keth_ruleset *rs; 776 777 PF_RULES_WASSERT(); 778 779 rs = pf_find_or_create_keth_ruleset(anchor); 780 if (rs == NULL) 781 return (EINVAL); 782 783 /* Purge old inactive rules. */ 784 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 785 tmp) { 786 TAILQ_REMOVE(rs->inactive.rules, rule, 787 entries); 788 pf_free_eth_rule(rule); 789 } 790 791 *ticket = ++rs->inactive.ticket; 792 rs->inactive.open = 1; 793 794 return (0); 795 } 796 797 static int 798 pf_rollback_eth(uint32_t ticket, const char *anchor) 799 { 800 struct pf_keth_rule *rule, *tmp; 801 struct pf_keth_ruleset *rs; 802 803 PF_RULES_WASSERT(); 804 805 rs = pf_find_keth_ruleset(anchor); 806 if (rs == NULL) 807 return (EINVAL); 808 809 if (!rs->inactive.open || 810 ticket != rs->inactive.ticket) 811 return (0); 812 813 /* Purge old inactive rules. */ 814 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 815 tmp) { 816 TAILQ_REMOVE(rs->inactive.rules, rule, entries); 817 pf_free_eth_rule(rule); 818 } 819 820 rs->inactive.open = 0; 821 822 pf_remove_if_empty_keth_ruleset(rs); 823 824 return (0); 825 } 826 827 #define PF_SET_SKIP_STEPS(i) \ 828 do { \ 829 while (head[i] != cur) { \ 830 head[i]->skip[i].ptr = cur; \ 831 head[i] = TAILQ_NEXT(head[i], entries); \ 832 } \ 833 } while (0) 834 835 static void 836 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules) 837 { 838 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT]; 839 int i; 840 841 cur = TAILQ_FIRST(rules); 842 prev = cur; 843 for (i = 0; i < PFE_SKIP_COUNT; ++i) 844 head[i] = cur; 845 while (cur != NULL) { 846 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 847 PF_SET_SKIP_STEPS(PFE_SKIP_IFP); 848 if (cur->direction != prev->direction) 849 PF_SET_SKIP_STEPS(PFE_SKIP_DIR); 850 if (cur->proto != prev->proto) 851 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO); 852 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0) 853 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR); 854 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0) 855 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR); 856 if (cur->ipsrc.neg != prev->ipsrc.neg || 857 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr)) 858 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR); 859 if (cur->ipdst.neg != prev->ipdst.neg || 860 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr)) 861 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR); 862 863 prev = cur; 864 cur = TAILQ_NEXT(cur, entries); 865 } 866 for (i = 0; i < PFE_SKIP_COUNT; ++i) 867 PF_SET_SKIP_STEPS(i); 868 } 869 870 static int 871 pf_commit_eth(uint32_t ticket, const char *anchor) 872 { 873 struct pf_keth_ruleq *rules; 874 struct pf_keth_ruleset *rs; 875 876 rs = pf_find_keth_ruleset(anchor); 877 if (rs == NULL) { 878 return (EINVAL); 879 } 880 881 if (!rs->inactive.open || 882 ticket != rs->inactive.ticket) 883 return (EBUSY); 884 885 PF_RULES_WASSERT(); 886 887 pf_eth_calc_skip_steps(rs->inactive.rules); 888 889 rules = rs->active.rules; 890 atomic_store_ptr(&rs->active.rules, rs->inactive.rules); 891 rs->inactive.rules = rules; 892 rs->inactive.ticket = rs->active.ticket; 893 894 return (pf_rollback_eth(rs->inactive.ticket, 895 rs->anchor ? rs->anchor->path : "")); 896 } 897 898 #ifdef ALTQ 899 static uint16_t 900 pf_qname2qid(const char *qname) 901 { 902 return (tagname2tag(&V_pf_qids, qname)); 903 } 904 905 static void 906 pf_qid_unref(uint16_t qid) 907 { 908 tag_unref(&V_pf_qids, qid); 909 } 910 911 static int 912 pf_begin_altq(u_int32_t *ticket) 913 { 914 struct pf_altq *altq, *tmp; 915 int error = 0; 916 917 PF_RULES_WASSERT(); 918 919 /* Purge the old altq lists */ 920 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 921 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 922 /* detach and destroy the discipline */ 923 error = altq_remove(altq); 924 } 925 free(altq, M_PFALTQ); 926 } 927 TAILQ_INIT(V_pf_altq_ifs_inactive); 928 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 929 pf_qid_unref(altq->qid); 930 free(altq, M_PFALTQ); 931 } 932 TAILQ_INIT(V_pf_altqs_inactive); 933 if (error) 934 return (error); 935 *ticket = ++V_ticket_altqs_inactive; 936 V_altqs_inactive_open = 1; 937 return (0); 938 } 939 940 static int 941 pf_rollback_altq(u_int32_t ticket) 942 { 943 struct pf_altq *altq, *tmp; 944 int error = 0; 945 946 PF_RULES_WASSERT(); 947 948 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 949 return (0); 950 /* Purge the old altq lists */ 951 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 952 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 953 /* detach and destroy the discipline */ 954 error = altq_remove(altq); 955 } 956 free(altq, M_PFALTQ); 957 } 958 TAILQ_INIT(V_pf_altq_ifs_inactive); 959 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 960 pf_qid_unref(altq->qid); 961 free(altq, M_PFALTQ); 962 } 963 TAILQ_INIT(V_pf_altqs_inactive); 964 V_altqs_inactive_open = 0; 965 return (error); 966 } 967 968 static int 969 pf_commit_altq(u_int32_t ticket) 970 { 971 struct pf_altqqueue *old_altqs, *old_altq_ifs; 972 struct pf_altq *altq, *tmp; 973 int err, error = 0; 974 975 PF_RULES_WASSERT(); 976 977 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 978 return (EBUSY); 979 980 /* swap altqs, keep the old. */ 981 old_altqs = V_pf_altqs_active; 982 old_altq_ifs = V_pf_altq_ifs_active; 983 V_pf_altqs_active = V_pf_altqs_inactive; 984 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive; 985 V_pf_altqs_inactive = old_altqs; 986 V_pf_altq_ifs_inactive = old_altq_ifs; 987 V_ticket_altqs_active = V_ticket_altqs_inactive; 988 989 /* Attach new disciplines */ 990 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 991 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 992 /* attach the discipline */ 993 error = altq_pfattach(altq); 994 if (error == 0 && V_pf_altq_running) 995 error = pf_enable_altq(altq); 996 if (error != 0) 997 return (error); 998 } 999 } 1000 1001 /* Purge the old altq lists */ 1002 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 1003 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 1004 /* detach and destroy the discipline */ 1005 if (V_pf_altq_running) 1006 error = pf_disable_altq(altq); 1007 err = altq_pfdetach(altq); 1008 if (err != 0 && error == 0) 1009 error = err; 1010 err = altq_remove(altq); 1011 if (err != 0 && error == 0) 1012 error = err; 1013 } 1014 free(altq, M_PFALTQ); 1015 } 1016 TAILQ_INIT(V_pf_altq_ifs_inactive); 1017 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1018 pf_qid_unref(altq->qid); 1019 free(altq, M_PFALTQ); 1020 } 1021 TAILQ_INIT(V_pf_altqs_inactive); 1022 1023 V_altqs_inactive_open = 0; 1024 return (error); 1025 } 1026 1027 static int 1028 pf_enable_altq(struct pf_altq *altq) 1029 { 1030 struct ifnet *ifp; 1031 struct tb_profile tb; 1032 int error = 0; 1033 1034 if ((ifp = ifunit(altq->ifname)) == NULL) 1035 return (EINVAL); 1036 1037 if (ifp->if_snd.altq_type != ALTQT_NONE) 1038 error = altq_enable(&ifp->if_snd); 1039 1040 /* set tokenbucket regulator */ 1041 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { 1042 tb.rate = altq->ifbandwidth; 1043 tb.depth = altq->tbrsize; 1044 error = tbr_set(&ifp->if_snd, &tb); 1045 } 1046 1047 return (error); 1048 } 1049 1050 static int 1051 pf_disable_altq(struct pf_altq *altq) 1052 { 1053 struct ifnet *ifp; 1054 struct tb_profile tb; 1055 int error; 1056 1057 if ((ifp = ifunit(altq->ifname)) == NULL) 1058 return (EINVAL); 1059 1060 /* 1061 * when the discipline is no longer referenced, it was overridden 1062 * by a new one. if so, just return. 1063 */ 1064 if (altq->altq_disc != ifp->if_snd.altq_disc) 1065 return (0); 1066 1067 error = altq_disable(&ifp->if_snd); 1068 1069 if (error == 0) { 1070 /* clear tokenbucket regulator */ 1071 tb.rate = 0; 1072 error = tbr_set(&ifp->if_snd, &tb); 1073 } 1074 1075 return (error); 1076 } 1077 1078 static int 1079 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket, 1080 struct pf_altq *altq) 1081 { 1082 struct ifnet *ifp1; 1083 int error = 0; 1084 1085 /* Deactivate the interface in question */ 1086 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; 1087 if ((ifp1 = ifunit(altq->ifname)) == NULL || 1088 (remove && ifp1 == ifp)) { 1089 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 1090 } else { 1091 error = altq_add(ifp1, altq); 1092 1093 if (ticket != V_ticket_altqs_inactive) 1094 error = EBUSY; 1095 1096 if (error) 1097 free(altq, M_PFALTQ); 1098 } 1099 1100 return (error); 1101 } 1102 1103 void 1104 pf_altq_ifnet_event(struct ifnet *ifp, int remove) 1105 { 1106 struct pf_altq *a1, *a2, *a3; 1107 u_int32_t ticket; 1108 int error = 0; 1109 1110 /* 1111 * No need to re-evaluate the configuration for events on interfaces 1112 * that do not support ALTQ, as it's not possible for such 1113 * interfaces to be part of the configuration. 1114 */ 1115 if (!ALTQ_IS_READY(&ifp->if_snd)) 1116 return; 1117 1118 /* Interrupt userland queue modifications */ 1119 if (V_altqs_inactive_open) 1120 pf_rollback_altq(V_ticket_altqs_inactive); 1121 1122 /* Start new altq ruleset */ 1123 if (pf_begin_altq(&ticket)) 1124 return; 1125 1126 /* Copy the current active set */ 1127 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) { 1128 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1129 if (a2 == NULL) { 1130 error = ENOMEM; 1131 break; 1132 } 1133 bcopy(a1, a2, sizeof(struct pf_altq)); 1134 1135 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1136 if (error) 1137 break; 1138 1139 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries); 1140 } 1141 if (error) 1142 goto out; 1143 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { 1144 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1145 if (a2 == NULL) { 1146 error = ENOMEM; 1147 break; 1148 } 1149 bcopy(a1, a2, sizeof(struct pf_altq)); 1150 1151 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) { 1152 error = EBUSY; 1153 free(a2, M_PFALTQ); 1154 break; 1155 } 1156 a2->altq_disc = NULL; 1157 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) { 1158 if (strncmp(a3->ifname, a2->ifname, 1159 IFNAMSIZ) == 0) { 1160 a2->altq_disc = a3->altq_disc; 1161 break; 1162 } 1163 } 1164 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1165 if (error) 1166 break; 1167 1168 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); 1169 } 1170 1171 out: 1172 if (error != 0) 1173 pf_rollback_altq(ticket); 1174 else 1175 pf_commit_altq(ticket); 1176 } 1177 #endif /* ALTQ */ 1178 1179 static struct pf_krule_global * 1180 pf_rule_tree_alloc(int flags) 1181 { 1182 struct pf_krule_global *tree; 1183 1184 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags); 1185 if (tree == NULL) 1186 return (NULL); 1187 RB_INIT(tree); 1188 return (tree); 1189 } 1190 1191 static void 1192 pf_rule_tree_free(struct pf_krule_global *tree) 1193 { 1194 1195 free(tree, M_TEMP); 1196 } 1197 1198 static int 1199 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) 1200 { 1201 struct pf_krule_global *tree; 1202 struct pf_kruleset *rs; 1203 struct pf_krule *rule; 1204 1205 PF_RULES_WASSERT(); 1206 1207 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1208 return (EINVAL); 1209 tree = pf_rule_tree_alloc(M_NOWAIT); 1210 if (tree == NULL) 1211 return (ENOMEM); 1212 rs = pf_find_or_create_kruleset(anchor); 1213 if (rs == NULL) { 1214 free(tree, M_TEMP); 1215 return (EINVAL); 1216 } 1217 pf_rule_tree_free(rs->rules[rs_num].inactive.tree); 1218 rs->rules[rs_num].inactive.tree = tree; 1219 1220 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1221 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1222 rs->rules[rs_num].inactive.rcount--; 1223 } 1224 *ticket = ++rs->rules[rs_num].inactive.ticket; 1225 rs->rules[rs_num].inactive.open = 1; 1226 return (0); 1227 } 1228 1229 static int 1230 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) 1231 { 1232 struct pf_kruleset *rs; 1233 struct pf_krule *rule; 1234 1235 PF_RULES_WASSERT(); 1236 1237 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1238 return (EINVAL); 1239 rs = pf_find_kruleset(anchor); 1240 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1241 rs->rules[rs_num].inactive.ticket != ticket) 1242 return (0); 1243 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1244 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1245 rs->rules[rs_num].inactive.rcount--; 1246 } 1247 rs->rules[rs_num].inactive.open = 0; 1248 return (0); 1249 } 1250 1251 #define PF_MD5_UPD(st, elm) \ 1252 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) 1253 1254 #define PF_MD5_UPD_STR(st, elm) \ 1255 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) 1256 1257 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ 1258 (stor) = htonl((st)->elm); \ 1259 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ 1260 } while (0) 1261 1262 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ 1263 (stor) = htons((st)->elm); \ 1264 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ 1265 } while (0) 1266 1267 static void 1268 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) 1269 { 1270 PF_MD5_UPD(pfr, addr.type); 1271 switch (pfr->addr.type) { 1272 case PF_ADDR_DYNIFTL: 1273 PF_MD5_UPD(pfr, addr.v.ifname); 1274 PF_MD5_UPD(pfr, addr.iflags); 1275 break; 1276 case PF_ADDR_TABLE: 1277 PF_MD5_UPD(pfr, addr.v.tblname); 1278 break; 1279 case PF_ADDR_ADDRMASK: 1280 /* XXX ignore af? */ 1281 PF_MD5_UPD(pfr, addr.v.a.addr.addr32); 1282 PF_MD5_UPD(pfr, addr.v.a.mask.addr32); 1283 break; 1284 } 1285 1286 PF_MD5_UPD(pfr, port[0]); 1287 PF_MD5_UPD(pfr, port[1]); 1288 PF_MD5_UPD(pfr, neg); 1289 PF_MD5_UPD(pfr, port_op); 1290 } 1291 1292 static void 1293 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule) 1294 { 1295 u_int16_t x; 1296 u_int32_t y; 1297 1298 pf_hash_rule_addr(ctx, &rule->src); 1299 pf_hash_rule_addr(ctx, &rule->dst); 1300 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) 1301 PF_MD5_UPD_STR(rule, label[i]); 1302 PF_MD5_UPD_STR(rule, ifname); 1303 PF_MD5_UPD_STR(rule, rcv_ifname); 1304 PF_MD5_UPD_STR(rule, match_tagname); 1305 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ 1306 PF_MD5_UPD_HTONL(rule, os_fingerprint, y); 1307 PF_MD5_UPD_HTONL(rule, prob, y); 1308 PF_MD5_UPD_HTONL(rule, uid.uid[0], y); 1309 PF_MD5_UPD_HTONL(rule, uid.uid[1], y); 1310 PF_MD5_UPD(rule, uid.op); 1311 PF_MD5_UPD_HTONL(rule, gid.gid[0], y); 1312 PF_MD5_UPD_HTONL(rule, gid.gid[1], y); 1313 PF_MD5_UPD(rule, gid.op); 1314 PF_MD5_UPD_HTONL(rule, rule_flag, y); 1315 PF_MD5_UPD(rule, action); 1316 PF_MD5_UPD(rule, direction); 1317 PF_MD5_UPD(rule, af); 1318 PF_MD5_UPD(rule, quick); 1319 PF_MD5_UPD(rule, ifnot); 1320 PF_MD5_UPD(rule, rcvifnot); 1321 PF_MD5_UPD(rule, match_tag_not); 1322 PF_MD5_UPD(rule, natpass); 1323 PF_MD5_UPD(rule, keep_state); 1324 PF_MD5_UPD(rule, proto); 1325 PF_MD5_UPD(rule, type); 1326 PF_MD5_UPD(rule, code); 1327 PF_MD5_UPD(rule, flags); 1328 PF_MD5_UPD(rule, flagset); 1329 PF_MD5_UPD(rule, allow_opts); 1330 PF_MD5_UPD(rule, rt); 1331 PF_MD5_UPD(rule, tos); 1332 PF_MD5_UPD(rule, scrub_flags); 1333 PF_MD5_UPD(rule, min_ttl); 1334 PF_MD5_UPD(rule, set_tos); 1335 if (rule->anchor != NULL) 1336 PF_MD5_UPD_STR(rule, anchor->path); 1337 } 1338 1339 static void 1340 pf_hash_rule(struct pf_krule *rule) 1341 { 1342 MD5_CTX ctx; 1343 1344 MD5Init(&ctx); 1345 pf_hash_rule_rolling(&ctx, rule); 1346 MD5Final(rule->md5sum, &ctx); 1347 } 1348 1349 static int 1350 pf_krule_compare(struct pf_krule *a, struct pf_krule *b) 1351 { 1352 1353 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH)); 1354 } 1355 1356 static int 1357 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) 1358 { 1359 struct pf_kruleset *rs; 1360 struct pf_krule *rule, **old_array, *old_rule; 1361 struct pf_krulequeue *old_rules; 1362 struct pf_krule_global *old_tree; 1363 int error; 1364 u_int32_t old_rcount; 1365 1366 PF_RULES_WASSERT(); 1367 1368 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1369 return (EINVAL); 1370 rs = pf_find_kruleset(anchor); 1371 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1372 ticket != rs->rules[rs_num].inactive.ticket) 1373 return (EBUSY); 1374 1375 /* Calculate checksum for the main ruleset */ 1376 if (rs == &pf_main_ruleset) { 1377 error = pf_setup_pfsync_matching(rs); 1378 if (error != 0) 1379 return (error); 1380 } 1381 1382 /* Swap rules, keep the old. */ 1383 old_rules = rs->rules[rs_num].active.ptr; 1384 old_rcount = rs->rules[rs_num].active.rcount; 1385 old_array = rs->rules[rs_num].active.ptr_array; 1386 old_tree = rs->rules[rs_num].active.tree; 1387 1388 rs->rules[rs_num].active.ptr = 1389 rs->rules[rs_num].inactive.ptr; 1390 rs->rules[rs_num].active.ptr_array = 1391 rs->rules[rs_num].inactive.ptr_array; 1392 rs->rules[rs_num].active.tree = 1393 rs->rules[rs_num].inactive.tree; 1394 rs->rules[rs_num].active.rcount = 1395 rs->rules[rs_num].inactive.rcount; 1396 1397 /* Attempt to preserve counter information. */ 1398 if (V_pf_status.keep_counters && old_tree != NULL) { 1399 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr, 1400 entries) { 1401 old_rule = RB_FIND(pf_krule_global, old_tree, rule); 1402 if (old_rule == NULL) { 1403 continue; 1404 } 1405 pf_counter_u64_critical_enter(); 1406 pf_counter_u64_rollup_protected(&rule->evaluations, 1407 pf_counter_u64_fetch(&old_rule->evaluations)); 1408 pf_counter_u64_rollup_protected(&rule->packets[0], 1409 pf_counter_u64_fetch(&old_rule->packets[0])); 1410 pf_counter_u64_rollup_protected(&rule->packets[1], 1411 pf_counter_u64_fetch(&old_rule->packets[1])); 1412 pf_counter_u64_rollup_protected(&rule->bytes[0], 1413 pf_counter_u64_fetch(&old_rule->bytes[0])); 1414 pf_counter_u64_rollup_protected(&rule->bytes[1], 1415 pf_counter_u64_fetch(&old_rule->bytes[1])); 1416 pf_counter_u64_critical_exit(); 1417 } 1418 } 1419 1420 rs->rules[rs_num].inactive.ptr = old_rules; 1421 rs->rules[rs_num].inactive.ptr_array = old_array; 1422 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ 1423 rs->rules[rs_num].inactive.rcount = old_rcount; 1424 1425 rs->rules[rs_num].active.ticket = 1426 rs->rules[rs_num].inactive.ticket; 1427 pf_calc_skip_steps(rs->rules[rs_num].active.ptr); 1428 1429 /* Purge the old rule list. */ 1430 PF_UNLNKDRULES_LOCK(); 1431 while ((rule = TAILQ_FIRST(old_rules)) != NULL) 1432 pf_unlink_rule_locked(old_rules, rule); 1433 PF_UNLNKDRULES_UNLOCK(); 1434 if (rs->rules[rs_num].inactive.ptr_array) 1435 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP); 1436 rs->rules[rs_num].inactive.ptr_array = NULL; 1437 rs->rules[rs_num].inactive.rcount = 0; 1438 rs->rules[rs_num].inactive.open = 0; 1439 pf_remove_if_empty_kruleset(rs); 1440 free(old_tree, M_TEMP); 1441 1442 return (0); 1443 } 1444 1445 static int 1446 pf_setup_pfsync_matching(struct pf_kruleset *rs) 1447 { 1448 MD5_CTX ctx; 1449 struct pf_krule *rule; 1450 int rs_cnt; 1451 u_int8_t digest[PF_MD5_DIGEST_LENGTH]; 1452 1453 MD5Init(&ctx); 1454 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { 1455 /* XXX PF_RULESET_SCRUB as well? */ 1456 if (rs_cnt == PF_RULESET_SCRUB) 1457 continue; 1458 1459 if (rs->rules[rs_cnt].inactive.ptr_array) 1460 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP); 1461 rs->rules[rs_cnt].inactive.ptr_array = NULL; 1462 1463 if (rs->rules[rs_cnt].inactive.rcount) { 1464 rs->rules[rs_cnt].inactive.ptr_array = 1465 mallocarray(rs->rules[rs_cnt].inactive.rcount, 1466 sizeof(struct pf_rule **), 1467 M_TEMP, M_NOWAIT); 1468 1469 if (!rs->rules[rs_cnt].inactive.ptr_array) 1470 return (ENOMEM); 1471 } 1472 1473 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, 1474 entries) { 1475 pf_hash_rule_rolling(&ctx, rule); 1476 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule; 1477 } 1478 } 1479 1480 MD5Final(digest, &ctx); 1481 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); 1482 return (0); 1483 } 1484 1485 static int 1486 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr) 1487 { 1488 int error = 0; 1489 1490 switch (addr->type) { 1491 case PF_ADDR_TABLE: 1492 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname); 1493 if (addr->p.tbl == NULL) 1494 error = ENOMEM; 1495 break; 1496 default: 1497 error = EINVAL; 1498 } 1499 1500 return (error); 1501 } 1502 1503 static int 1504 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr, 1505 sa_family_t af) 1506 { 1507 int error = 0; 1508 1509 switch (addr->type) { 1510 case PF_ADDR_TABLE: 1511 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); 1512 if (addr->p.tbl == NULL) 1513 error = ENOMEM; 1514 break; 1515 case PF_ADDR_DYNIFTL: 1516 error = pfi_dynaddr_setup(addr, af); 1517 break; 1518 } 1519 1520 return (error); 1521 } 1522 1523 void 1524 pf_addr_copyout(struct pf_addr_wrap *addr) 1525 { 1526 1527 switch (addr->type) { 1528 case PF_ADDR_DYNIFTL: 1529 pfi_dynaddr_copyout(addr); 1530 break; 1531 case PF_ADDR_TABLE: 1532 pf_tbladdr_copyout(addr); 1533 break; 1534 } 1535 } 1536 1537 static void 1538 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out) 1539 { 1540 int secs = time_uptime, diff; 1541 1542 bzero(out, sizeof(struct pf_src_node)); 1543 1544 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr)); 1545 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr)); 1546 1547 if (in->rule != NULL) 1548 out->rule.nr = in->rule->nr; 1549 1550 for (int i = 0; i < 2; i++) { 1551 out->bytes[i] = counter_u64_fetch(in->bytes[i]); 1552 out->packets[i] = counter_u64_fetch(in->packets[i]); 1553 } 1554 1555 out->states = in->states; 1556 out->conn = in->conn; 1557 out->af = in->af; 1558 out->ruletype = in->ruletype; 1559 1560 out->creation = secs - in->creation; 1561 if (out->expire > secs) 1562 out->expire -= secs; 1563 else 1564 out->expire = 0; 1565 1566 /* Adjust the connection rate estimate. */ 1567 out->conn_rate = in->conn_rate; 1568 diff = secs - in->conn_rate.last; 1569 if (diff >= in->conn_rate.seconds) 1570 out->conn_rate.count = 0; 1571 else 1572 out->conn_rate.count -= 1573 in->conn_rate.count * diff / 1574 in->conn_rate.seconds; 1575 } 1576 1577 #ifdef ALTQ 1578 /* 1579 * Handle export of struct pf_kaltq to user binaries that may be using any 1580 * version of struct pf_altq. 1581 */ 1582 static int 1583 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size) 1584 { 1585 u_int32_t version; 1586 1587 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1588 version = 0; 1589 else 1590 version = pa->version; 1591 1592 if (version > PFIOC_ALTQ_VERSION) 1593 return (EINVAL); 1594 1595 #define ASSIGN(x) exported_q->x = q->x 1596 #define COPY(x) \ 1597 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x))) 1598 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX) 1599 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX) 1600 1601 switch (version) { 1602 case 0: { 1603 struct pf_altq_v0 *exported_q = 1604 &((struct pfioc_altq_v0 *)pa)->altq; 1605 1606 COPY(ifname); 1607 1608 ASSIGN(scheduler); 1609 ASSIGN(tbrsize); 1610 exported_q->tbrsize = SATU16(q->tbrsize); 1611 exported_q->ifbandwidth = SATU32(q->ifbandwidth); 1612 1613 COPY(qname); 1614 COPY(parent); 1615 ASSIGN(parent_qid); 1616 exported_q->bandwidth = SATU32(q->bandwidth); 1617 ASSIGN(priority); 1618 ASSIGN(local_flags); 1619 1620 ASSIGN(qlimit); 1621 ASSIGN(flags); 1622 1623 if (q->scheduler == ALTQT_HFSC) { 1624 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x 1625 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \ 1626 SATU32(q->pq_u.hfsc_opts.x) 1627 1628 ASSIGN_OPT_SATU32(rtsc_m1); 1629 ASSIGN_OPT(rtsc_d); 1630 ASSIGN_OPT_SATU32(rtsc_m2); 1631 1632 ASSIGN_OPT_SATU32(lssc_m1); 1633 ASSIGN_OPT(lssc_d); 1634 ASSIGN_OPT_SATU32(lssc_m2); 1635 1636 ASSIGN_OPT_SATU32(ulsc_m1); 1637 ASSIGN_OPT(ulsc_d); 1638 ASSIGN_OPT_SATU32(ulsc_m2); 1639 1640 ASSIGN_OPT(flags); 1641 1642 #undef ASSIGN_OPT 1643 #undef ASSIGN_OPT_SATU32 1644 } else 1645 COPY(pq_u); 1646 1647 ASSIGN(qid); 1648 break; 1649 } 1650 case 1: { 1651 struct pf_altq_v1 *exported_q = 1652 &((struct pfioc_altq_v1 *)pa)->altq; 1653 1654 COPY(ifname); 1655 1656 ASSIGN(scheduler); 1657 ASSIGN(tbrsize); 1658 ASSIGN(ifbandwidth); 1659 1660 COPY(qname); 1661 COPY(parent); 1662 ASSIGN(parent_qid); 1663 ASSIGN(bandwidth); 1664 ASSIGN(priority); 1665 ASSIGN(local_flags); 1666 1667 ASSIGN(qlimit); 1668 ASSIGN(flags); 1669 COPY(pq_u); 1670 1671 ASSIGN(qid); 1672 break; 1673 } 1674 default: 1675 panic("%s: unhandled struct pfioc_altq version", __func__); 1676 break; 1677 } 1678 1679 #undef ASSIGN 1680 #undef COPY 1681 #undef SATU16 1682 #undef SATU32 1683 1684 return (0); 1685 } 1686 1687 /* 1688 * Handle import to struct pf_kaltq of struct pf_altq from user binaries 1689 * that may be using any version of it. 1690 */ 1691 static int 1692 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size) 1693 { 1694 u_int32_t version; 1695 1696 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1697 version = 0; 1698 else 1699 version = pa->version; 1700 1701 if (version > PFIOC_ALTQ_VERSION) 1702 return (EINVAL); 1703 1704 #define ASSIGN(x) q->x = imported_q->x 1705 #define COPY(x) \ 1706 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x))) 1707 1708 switch (version) { 1709 case 0: { 1710 struct pf_altq_v0 *imported_q = 1711 &((struct pfioc_altq_v0 *)pa)->altq; 1712 1713 COPY(ifname); 1714 1715 ASSIGN(scheduler); 1716 ASSIGN(tbrsize); /* 16-bit -> 32-bit */ 1717 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */ 1718 1719 COPY(qname); 1720 COPY(parent); 1721 ASSIGN(parent_qid); 1722 ASSIGN(bandwidth); /* 32-bit -> 64-bit */ 1723 ASSIGN(priority); 1724 ASSIGN(local_flags); 1725 1726 ASSIGN(qlimit); 1727 ASSIGN(flags); 1728 1729 if (imported_q->scheduler == ALTQT_HFSC) { 1730 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x 1731 1732 /* 1733 * The m1 and m2 parameters are being copied from 1734 * 32-bit to 64-bit. 1735 */ 1736 ASSIGN_OPT(rtsc_m1); 1737 ASSIGN_OPT(rtsc_d); 1738 ASSIGN_OPT(rtsc_m2); 1739 1740 ASSIGN_OPT(lssc_m1); 1741 ASSIGN_OPT(lssc_d); 1742 ASSIGN_OPT(lssc_m2); 1743 1744 ASSIGN_OPT(ulsc_m1); 1745 ASSIGN_OPT(ulsc_d); 1746 ASSIGN_OPT(ulsc_m2); 1747 1748 ASSIGN_OPT(flags); 1749 1750 #undef ASSIGN_OPT 1751 } else 1752 COPY(pq_u); 1753 1754 ASSIGN(qid); 1755 break; 1756 } 1757 case 1: { 1758 struct pf_altq_v1 *imported_q = 1759 &((struct pfioc_altq_v1 *)pa)->altq; 1760 1761 COPY(ifname); 1762 1763 ASSIGN(scheduler); 1764 ASSIGN(tbrsize); 1765 ASSIGN(ifbandwidth); 1766 1767 COPY(qname); 1768 COPY(parent); 1769 ASSIGN(parent_qid); 1770 ASSIGN(bandwidth); 1771 ASSIGN(priority); 1772 ASSIGN(local_flags); 1773 1774 ASSIGN(qlimit); 1775 ASSIGN(flags); 1776 COPY(pq_u); 1777 1778 ASSIGN(qid); 1779 break; 1780 } 1781 default: 1782 panic("%s: unhandled struct pfioc_altq version", __func__); 1783 break; 1784 } 1785 1786 #undef ASSIGN 1787 #undef COPY 1788 1789 return (0); 1790 } 1791 1792 static struct pf_altq * 1793 pf_altq_get_nth_active(u_int32_t n) 1794 { 1795 struct pf_altq *altq; 1796 u_int32_t nr; 1797 1798 nr = 0; 1799 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1800 if (nr == n) 1801 return (altq); 1802 nr++; 1803 } 1804 1805 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { 1806 if (nr == n) 1807 return (altq); 1808 nr++; 1809 } 1810 1811 return (NULL); 1812 } 1813 #endif /* ALTQ */ 1814 1815 struct pf_krule * 1816 pf_krule_alloc(void) 1817 { 1818 struct pf_krule *rule; 1819 1820 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO); 1821 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF); 1822 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF); 1823 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF); 1824 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 1825 M_WAITOK | M_ZERO); 1826 return (rule); 1827 } 1828 1829 void 1830 pf_krule_free(struct pf_krule *rule) 1831 { 1832 #ifdef PF_WANT_32_TO_64_COUNTER 1833 bool wowned; 1834 #endif 1835 1836 if (rule == NULL) 1837 return; 1838 1839 #ifdef PF_WANT_32_TO_64_COUNTER 1840 if (rule->allrulelinked) { 1841 wowned = PF_RULES_WOWNED(); 1842 if (!wowned) 1843 PF_RULES_WLOCK(); 1844 LIST_REMOVE(rule, allrulelist); 1845 V_pf_allrulecount--; 1846 if (!wowned) 1847 PF_RULES_WUNLOCK(); 1848 } 1849 #endif 1850 1851 pf_counter_u64_deinit(&rule->evaluations); 1852 for (int i = 0; i < 2; i++) { 1853 pf_counter_u64_deinit(&rule->packets[i]); 1854 pf_counter_u64_deinit(&rule->bytes[i]); 1855 } 1856 counter_u64_free(rule->states_cur); 1857 counter_u64_free(rule->states_tot); 1858 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 1859 counter_u64_free(rule->src_nodes[sn_type]); 1860 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 1861 1862 mtx_destroy(&rule->nat.mtx); 1863 mtx_destroy(&rule->rdr.mtx); 1864 mtx_destroy(&rule->route.mtx); 1865 free(rule, M_PFRULE); 1866 } 1867 1868 void 1869 pf_krule_clear_counters(struct pf_krule *rule) 1870 { 1871 pf_counter_u64_zero(&rule->evaluations); 1872 for (int i = 0; i < 2; i++) { 1873 pf_counter_u64_zero(&rule->packets[i]); 1874 pf_counter_u64_zero(&rule->bytes[i]); 1875 } 1876 counter_u64_zero(rule->states_tot); 1877 } 1878 1879 static void 1880 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool, 1881 struct pf_pooladdr *pool) 1882 { 1883 1884 bzero(pool, sizeof(*pool)); 1885 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr)); 1886 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname)); 1887 } 1888 1889 static int 1890 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool, 1891 struct pf_kpooladdr *kpool) 1892 { 1893 int ret; 1894 1895 bzero(kpool, sizeof(*kpool)); 1896 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr)); 1897 ret = pf_user_strcpy(kpool->ifname, pool->ifname, 1898 sizeof(kpool->ifname)); 1899 return (ret); 1900 } 1901 1902 static void 1903 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool) 1904 { 1905 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), ""); 1906 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), ""); 1907 1908 bcopy(&pool->key, &kpool->key, sizeof(kpool->key)); 1909 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter)); 1910 1911 kpool->tblidx = pool->tblidx; 1912 kpool->proxy_port[0] = pool->proxy_port[0]; 1913 kpool->proxy_port[1] = pool->proxy_port[1]; 1914 kpool->opts = pool->opts; 1915 } 1916 1917 static int 1918 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule) 1919 { 1920 int ret; 1921 1922 #ifndef INET 1923 if (rule->af == AF_INET) { 1924 return (EAFNOSUPPORT); 1925 } 1926 #endif /* INET */ 1927 #ifndef INET6 1928 if (rule->af == AF_INET6) { 1929 return (EAFNOSUPPORT); 1930 } 1931 #endif /* INET6 */ 1932 1933 ret = pf_check_rule_addr(&rule->src); 1934 if (ret != 0) 1935 return (ret); 1936 ret = pf_check_rule_addr(&rule->dst); 1937 if (ret != 0) 1938 return (ret); 1939 1940 bcopy(&rule->src, &krule->src, sizeof(rule->src)); 1941 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst)); 1942 1943 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label)); 1944 if (ret != 0) 1945 return (ret); 1946 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname)); 1947 if (ret != 0) 1948 return (ret); 1949 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname)); 1950 if (ret != 0) 1951 return (ret); 1952 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname)); 1953 if (ret != 0) 1954 return (ret); 1955 ret = pf_user_strcpy(krule->tagname, rule->tagname, 1956 sizeof(rule->tagname)); 1957 if (ret != 0) 1958 return (ret); 1959 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname, 1960 sizeof(rule->match_tagname)); 1961 if (ret != 0) 1962 return (ret); 1963 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname, 1964 sizeof(rule->overload_tblname)); 1965 if (ret != 0) 1966 return (ret); 1967 1968 pf_pool_to_kpool(&rule->rpool, &krule->rdr); 1969 1970 /* Don't allow userspace to set evaluations, packets or bytes. */ 1971 /* kif, anchor, overload_tbl are not copied over. */ 1972 1973 krule->os_fingerprint = rule->os_fingerprint; 1974 1975 krule->rtableid = rule->rtableid; 1976 /* pf_rule->timeout is smaller than pf_krule->timeout */ 1977 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout)); 1978 krule->max_states = rule->max_states; 1979 krule->max_src_nodes = rule->max_src_nodes; 1980 krule->max_src_states = rule->max_src_states; 1981 krule->max_src_conn = rule->max_src_conn; 1982 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit; 1983 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds; 1984 krule->qid = rule->qid; 1985 krule->pqid = rule->pqid; 1986 krule->nr = rule->nr; 1987 krule->prob = rule->prob; 1988 krule->cuid = rule->cuid; 1989 krule->cpid = rule->cpid; 1990 1991 krule->return_icmp = rule->return_icmp; 1992 krule->return_icmp6 = rule->return_icmp6; 1993 krule->max_mss = rule->max_mss; 1994 krule->tag = rule->tag; 1995 krule->match_tag = rule->match_tag; 1996 krule->scrub_flags = rule->scrub_flags; 1997 1998 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid)); 1999 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid)); 2000 2001 krule->rule_flag = rule->rule_flag; 2002 krule->action = rule->action; 2003 krule->direction = rule->direction; 2004 krule->log = rule->log; 2005 krule->logif = rule->logif; 2006 krule->quick = rule->quick; 2007 krule->ifnot = rule->ifnot; 2008 krule->match_tag_not = rule->match_tag_not; 2009 krule->natpass = rule->natpass; 2010 2011 krule->keep_state = rule->keep_state; 2012 krule->af = rule->af; 2013 krule->proto = rule->proto; 2014 krule->type = rule->type; 2015 krule->code = rule->code; 2016 krule->flags = rule->flags; 2017 krule->flagset = rule->flagset; 2018 krule->min_ttl = rule->min_ttl; 2019 krule->allow_opts = rule->allow_opts; 2020 krule->rt = rule->rt; 2021 krule->return_ttl = rule->return_ttl; 2022 krule->tos = rule->tos; 2023 krule->set_tos = rule->set_tos; 2024 2025 krule->flush = rule->flush; 2026 krule->prio = rule->prio; 2027 krule->set_prio[0] = rule->set_prio[0]; 2028 krule->set_prio[1] = rule->set_prio[1]; 2029 2030 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert)); 2031 2032 return (0); 2033 } 2034 2035 int 2036 pf_ioctl_getrules(struct pfioc_rule *pr) 2037 { 2038 struct pf_kruleset *ruleset; 2039 struct pf_krule *tail; 2040 int rs_num; 2041 2042 PF_RULES_WLOCK(); 2043 ruleset = pf_find_kruleset(pr->anchor); 2044 if (ruleset == NULL) { 2045 PF_RULES_WUNLOCK(); 2046 return (EINVAL); 2047 } 2048 rs_num = pf_get_ruleset_number(pr->rule.action); 2049 if (rs_num >= PF_RULESET_MAX) { 2050 PF_RULES_WUNLOCK(); 2051 return (EINVAL); 2052 } 2053 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 2054 pf_krulequeue); 2055 if (tail) 2056 pr->nr = tail->nr + 1; 2057 else 2058 pr->nr = 0; 2059 pr->ticket = ruleset->rules[rs_num].active.ticket; 2060 PF_RULES_WUNLOCK(); 2061 2062 return (0); 2063 } 2064 2065 int 2066 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, 2067 uint32_t pool_ticket, const char *anchor, const char *anchor_call, 2068 uid_t uid, pid_t pid) 2069 { 2070 struct pf_kruleset *ruleset; 2071 struct pf_krule *tail; 2072 struct pf_kpooladdr *pa; 2073 struct pfi_kkif *kif = NULL, *rcv_kif = NULL; 2074 int rs_num; 2075 int error = 0; 2076 2077 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) { 2078 error = EINVAL; 2079 goto errout_unlocked; 2080 } 2081 2082 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 2083 2084 if (rule->ifname[0]) 2085 kif = pf_kkif_create(M_WAITOK); 2086 if (rule->rcv_ifname[0]) 2087 rcv_kif = pf_kkif_create(M_WAITOK); 2088 pf_counter_u64_init(&rule->evaluations, M_WAITOK); 2089 for (int i = 0; i < 2; i++) { 2090 pf_counter_u64_init(&rule->packets[i], M_WAITOK); 2091 pf_counter_u64_init(&rule->bytes[i], M_WAITOK); 2092 } 2093 rule->states_cur = counter_u64_alloc(M_WAITOK); 2094 rule->states_tot = counter_u64_alloc(M_WAITOK); 2095 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 2096 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 2097 rule->cuid = uid; 2098 rule->cpid = pid; 2099 TAILQ_INIT(&rule->rdr.list); 2100 TAILQ_INIT(&rule->nat.list); 2101 TAILQ_INIT(&rule->route.list); 2102 2103 PF_CONFIG_LOCK(); 2104 PF_RULES_WLOCK(); 2105 #ifdef PF_WANT_32_TO_64_COUNTER 2106 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist); 2107 MPASS(!rule->allrulelinked); 2108 rule->allrulelinked = true; 2109 V_pf_allrulecount++; 2110 #endif 2111 ruleset = pf_find_kruleset(anchor); 2112 if (ruleset == NULL) 2113 ERROUT(EINVAL); 2114 rs_num = pf_get_ruleset_number(rule->action); 2115 if (rs_num >= PF_RULESET_MAX) 2116 ERROUT(EINVAL); 2117 if (ticket != ruleset->rules[rs_num].inactive.ticket) { 2118 DPFPRINTF(PF_DEBUG_MISC, 2119 ("ticket: %d != [%d]%d\n", ticket, rs_num, 2120 ruleset->rules[rs_num].inactive.ticket)); 2121 ERROUT(EBUSY); 2122 } 2123 if (pool_ticket != V_ticket_pabuf) { 2124 DPFPRINTF(PF_DEBUG_MISC, 2125 ("pool_ticket: %d != %d\n", pool_ticket, 2126 V_ticket_pabuf)); 2127 ERROUT(EBUSY); 2128 } 2129 /* 2130 * XXXMJG hack: there is no mechanism to ensure they started the 2131 * transaction. Ticket checked above may happen to match by accident, 2132 * even if nobody called DIOCXBEGIN, let alone this process. 2133 * Partially work around it by checking if the RB tree got allocated, 2134 * see pf_begin_rules. 2135 */ 2136 if (ruleset->rules[rs_num].inactive.tree == NULL) { 2137 ERROUT(EINVAL); 2138 } 2139 2140 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 2141 pf_krulequeue); 2142 if (tail) 2143 rule->nr = tail->nr + 1; 2144 else 2145 rule->nr = 0; 2146 if (rule->ifname[0]) { 2147 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2148 kif = NULL; 2149 pfi_kkif_ref(rule->kif); 2150 } else 2151 rule->kif = NULL; 2152 2153 if (rule->rcv_ifname[0]) { 2154 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname); 2155 rcv_kif = NULL; 2156 pfi_kkif_ref(rule->rcv_kif); 2157 } else 2158 rule->rcv_kif = NULL; 2159 2160 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) 2161 error = EBUSY; 2162 2163 #ifdef ALTQ 2164 /* set queue IDs */ 2165 if (rule->qname[0] != 0) { 2166 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2167 error = EBUSY; 2168 else if (rule->pqname[0] != 0) { 2169 if ((rule->pqid = 2170 pf_qname2qid(rule->pqname)) == 0) 2171 error = EBUSY; 2172 } else 2173 rule->pqid = rule->qid; 2174 } 2175 #endif 2176 if (rule->tagname[0]) 2177 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2178 error = EBUSY; 2179 if (rule->match_tagname[0]) 2180 if ((rule->match_tag = 2181 pf_tagname2tag(rule->match_tagname)) == 0) 2182 error = EBUSY; 2183 if (rule->rt && !rule->direction) 2184 error = EINVAL; 2185 if (!rule->log) 2186 rule->logif = 0; 2187 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) 2188 error = ENOMEM; 2189 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) 2190 error = ENOMEM; 2191 if (pf_kanchor_setup(rule, ruleset, anchor_call)) 2192 error = EINVAL; 2193 if (rule->scrub_flags & PFSTATE_SETPRIO && 2194 (rule->set_prio[0] > PF_PRIO_MAX || 2195 rule->set_prio[1] > PF_PRIO_MAX)) 2196 error = EINVAL; 2197 for (int i = 0; i < 3; i++) { 2198 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 2199 if (pa->addr.type == PF_ADDR_TABLE) { 2200 pa->addr.p.tbl = pfr_attach_table(ruleset, 2201 pa->addr.v.tblname); 2202 if (pa->addr.p.tbl == NULL) 2203 error = ENOMEM; 2204 } 2205 } 2206 2207 rule->overload_tbl = NULL; 2208 if (rule->overload_tblname[0]) { 2209 if ((rule->overload_tbl = pfr_attach_table(ruleset, 2210 rule->overload_tblname)) == NULL) 2211 error = EINVAL; 2212 else 2213 rule->overload_tbl->pfrkt_flags |= 2214 PFR_TFLAG_ACTIVE; 2215 } 2216 2217 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list); 2218 2219 /* 2220 * Old version of pfctl provide route redirection pools in single 2221 * common redirection pool rdr. New versions use rdr only for 2222 * rdr-to rules. 2223 */ 2224 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) { 2225 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list); 2226 } else { 2227 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list); 2228 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list); 2229 } 2230 2231 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) || 2232 (rule->action == PF_BINAT)) && rule->anchor == NULL && 2233 TAILQ_FIRST(&rule->rdr.list) == NULL) { 2234 error = EINVAL; 2235 } 2236 2237 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) { 2238 error = EINVAL; 2239 } 2240 2241 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR || 2242 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) { 2243 error = EINVAL; 2244 } 2245 2246 if (error) { 2247 pf_free_rule(rule); 2248 rule = NULL; 2249 ERROUT(error); 2250 } 2251 2252 rule->nat.cur = TAILQ_FIRST(&rule->nat.list); 2253 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list); 2254 rule->route.cur = TAILQ_FIRST(&rule->route.list); 2255 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, 2256 rule, entries); 2257 ruleset->rules[rs_num].inactive.rcount++; 2258 2259 PF_RULES_WUNLOCK(); 2260 pf_hash_rule(rule); 2261 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) { 2262 PF_RULES_WLOCK(); 2263 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries); 2264 ruleset->rules[rs_num].inactive.rcount--; 2265 pf_free_rule(rule); 2266 rule = NULL; 2267 ERROUT(EEXIST); 2268 } 2269 PF_CONFIG_UNLOCK(); 2270 2271 return (0); 2272 2273 #undef ERROUT 2274 errout: 2275 PF_RULES_WUNLOCK(); 2276 PF_CONFIG_UNLOCK(); 2277 errout_unlocked: 2278 pf_kkif_free(rcv_kif); 2279 pf_kkif_free(kif); 2280 pf_krule_free(rule); 2281 return (error); 2282 } 2283 2284 static bool 2285 pf_label_match(const struct pf_krule *rule, const char *label) 2286 { 2287 int i = 0; 2288 2289 while (*rule->label[i]) { 2290 if (strcmp(rule->label[i], label) == 0) 2291 return (true); 2292 i++; 2293 } 2294 2295 return (false); 2296 } 2297 2298 static unsigned int 2299 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir) 2300 { 2301 struct pf_kstate *s; 2302 int more = 0; 2303 2304 s = pf_find_state_all(key, dir, &more); 2305 if (s == NULL) 2306 return (0); 2307 2308 if (more) { 2309 PF_STATE_UNLOCK(s); 2310 return (0); 2311 } 2312 2313 pf_remove_state(s); 2314 return (1); 2315 } 2316 2317 static int 2318 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih) 2319 { 2320 struct pf_kstate *s; 2321 struct pf_state_key *sk; 2322 struct pf_addr *srcaddr, *dstaddr; 2323 struct pf_state_key_cmp match_key; 2324 int idx, killed = 0; 2325 unsigned int dir; 2326 u_int16_t srcport, dstport; 2327 struct pfi_kkif *kif; 2328 2329 relock_DIOCKILLSTATES: 2330 PF_HASHROW_LOCK(ih); 2331 LIST_FOREACH(s, &ih->states, entry) { 2332 /* For floating states look at the original kif. */ 2333 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 2334 2335 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE]; 2336 if (s->direction == PF_OUT) { 2337 srcaddr = &sk->addr[1]; 2338 dstaddr = &sk->addr[0]; 2339 srcport = sk->port[1]; 2340 dstport = sk->port[0]; 2341 } else { 2342 srcaddr = &sk->addr[0]; 2343 dstaddr = &sk->addr[1]; 2344 srcport = sk->port[0]; 2345 dstport = sk->port[1]; 2346 } 2347 2348 if (psk->psk_af && sk->af != psk->psk_af) 2349 continue; 2350 2351 if (psk->psk_proto && psk->psk_proto != sk->proto) 2352 continue; 2353 2354 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr, 2355 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af)) 2356 continue; 2357 2358 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr, 2359 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af)) 2360 continue; 2361 2362 if (! PF_MATCHA(psk->psk_rt_addr.neg, 2363 &psk->psk_rt_addr.addr.v.a.addr, 2364 &psk->psk_rt_addr.addr.v.a.mask, 2365 &s->act.rt_addr, sk->af)) 2366 continue; 2367 2368 if (psk->psk_src.port_op != 0 && 2369 ! pf_match_port(psk->psk_src.port_op, 2370 psk->psk_src.port[0], psk->psk_src.port[1], srcport)) 2371 continue; 2372 2373 if (psk->psk_dst.port_op != 0 && 2374 ! pf_match_port(psk->psk_dst.port_op, 2375 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) 2376 continue; 2377 2378 if (psk->psk_label[0] && 2379 ! pf_label_match(s->rule, psk->psk_label)) 2380 continue; 2381 2382 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname, 2383 kif->pfik_name)) 2384 continue; 2385 2386 if (psk->psk_kill_match) { 2387 /* Create the key to find matching states, with lock 2388 * held. */ 2389 2390 bzero(&match_key, sizeof(match_key)); 2391 2392 if (s->direction == PF_OUT) { 2393 dir = PF_IN; 2394 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK; 2395 } else { 2396 dir = PF_OUT; 2397 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE; 2398 } 2399 2400 match_key.af = s->key[idx]->af; 2401 match_key.proto = s->key[idx]->proto; 2402 PF_ACPY(&match_key.addr[0], 2403 &s->key[idx]->addr[1], match_key.af); 2404 match_key.port[0] = s->key[idx]->port[1]; 2405 PF_ACPY(&match_key.addr[1], 2406 &s->key[idx]->addr[0], match_key.af); 2407 match_key.port[1] = s->key[idx]->port[0]; 2408 } 2409 2410 pf_remove_state(s); 2411 killed++; 2412 2413 if (psk->psk_kill_match) 2414 killed += pf_kill_matching_state(&match_key, dir); 2415 2416 goto relock_DIOCKILLSTATES; 2417 } 2418 PF_HASHROW_UNLOCK(ih); 2419 2420 return (killed); 2421 } 2422 2423 void 2424 unhandled_af(int af) 2425 { 2426 panic("unhandled af %d", af); 2427 } 2428 2429 int 2430 pf_start(void) 2431 { 2432 int error = 0; 2433 2434 sx_xlock(&V_pf_ioctl_lock); 2435 if (V_pf_status.running) 2436 error = EEXIST; 2437 else { 2438 hook_pf(); 2439 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 2440 hook_pf_eth(); 2441 V_pf_status.running = 1; 2442 V_pf_status.since = time_second; 2443 new_unrhdr64(&V_pf_stateid, time_second); 2444 2445 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n")); 2446 } 2447 sx_xunlock(&V_pf_ioctl_lock); 2448 2449 return (error); 2450 } 2451 2452 int 2453 pf_stop(void) 2454 { 2455 int error = 0; 2456 2457 sx_xlock(&V_pf_ioctl_lock); 2458 if (!V_pf_status.running) 2459 error = ENOENT; 2460 else { 2461 V_pf_status.running = 0; 2462 dehook_pf(); 2463 dehook_pf_eth(); 2464 V_pf_status.since = time_second; 2465 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n")); 2466 } 2467 sx_xunlock(&V_pf_ioctl_lock); 2468 2469 return (error); 2470 } 2471 2472 void 2473 pf_ioctl_clear_status(void) 2474 { 2475 PF_RULES_WLOCK(); 2476 for (int i = 0; i < PFRES_MAX; i++) 2477 counter_u64_zero(V_pf_status.counters[i]); 2478 for (int i = 0; i < FCNT_MAX; i++) 2479 pf_counter_u64_zero(&V_pf_status.fcounters[i]); 2480 for (int i = 0; i < SCNT_MAX; i++) 2481 counter_u64_zero(V_pf_status.scounters[i]); 2482 for (int i = 0; i < KLCNT_MAX; i++) 2483 counter_u64_zero(V_pf_status.lcounters[i]); 2484 V_pf_status.since = time_second; 2485 if (*V_pf_status.ifname) 2486 pfi_update_status(V_pf_status.ifname, NULL); 2487 PF_RULES_WUNLOCK(); 2488 } 2489 2490 int 2491 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds) 2492 { 2493 uint32_t old; 2494 2495 if (timeout < 0 || timeout >= PFTM_MAX || 2496 seconds < 0) 2497 return (EINVAL); 2498 2499 PF_RULES_WLOCK(); 2500 old = V_pf_default_rule.timeout[timeout]; 2501 if (timeout == PFTM_INTERVAL && seconds == 0) 2502 seconds = 1; 2503 V_pf_default_rule.timeout[timeout] = seconds; 2504 if (timeout == PFTM_INTERVAL && seconds < old) 2505 wakeup(pf_purge_thread); 2506 2507 if (prev_seconds != NULL) 2508 *prev_seconds = old; 2509 2510 PF_RULES_WUNLOCK(); 2511 2512 return (0); 2513 } 2514 2515 int 2516 pf_ioctl_get_timeout(int timeout, int *seconds) 2517 { 2518 PF_RULES_RLOCK_TRACKER; 2519 2520 if (timeout < 0 || timeout >= PFTM_MAX) 2521 return (EINVAL); 2522 2523 PF_RULES_RLOCK(); 2524 *seconds = V_pf_default_rule.timeout[timeout]; 2525 PF_RULES_RUNLOCK(); 2526 2527 return (0); 2528 } 2529 2530 int 2531 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit) 2532 { 2533 2534 PF_RULES_WLOCK(); 2535 if (index < 0 || index >= PF_LIMIT_MAX || 2536 V_pf_limits[index].zone == NULL) { 2537 PF_RULES_WUNLOCK(); 2538 return (EINVAL); 2539 } 2540 uma_zone_set_max(V_pf_limits[index].zone, 2541 limit == 0 ? INT_MAX : limit); 2542 if (old_limit != NULL) 2543 *old_limit = V_pf_limits[index].limit; 2544 V_pf_limits[index].limit = limit; 2545 PF_RULES_WUNLOCK(); 2546 2547 return (0); 2548 } 2549 2550 int 2551 pf_ioctl_get_limit(int index, unsigned int *limit) 2552 { 2553 PF_RULES_RLOCK_TRACKER; 2554 2555 if (index < 0 || index >= PF_LIMIT_MAX) 2556 return (EINVAL); 2557 2558 PF_RULES_RLOCK(); 2559 *limit = V_pf_limits[index].limit; 2560 PF_RULES_RUNLOCK(); 2561 2562 return (0); 2563 } 2564 2565 int 2566 pf_ioctl_begin_addrs(uint32_t *ticket) 2567 { 2568 PF_RULES_WLOCK(); 2569 pf_empty_kpool(&V_pf_pabuf[0]); 2570 pf_empty_kpool(&V_pf_pabuf[1]); 2571 pf_empty_kpool(&V_pf_pabuf[2]); 2572 *ticket = ++V_ticket_pabuf; 2573 PF_RULES_WUNLOCK(); 2574 2575 return (0); 2576 } 2577 2578 int 2579 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp) 2580 { 2581 struct pf_kpooladdr *pa = NULL; 2582 struct pfi_kkif *kif = NULL; 2583 int error; 2584 2585 if (pp->which != PF_RDR && pp->which != PF_NAT && 2586 pp->which != PF_RT) 2587 return (EINVAL); 2588 2589 switch (pp->af) { 2590 #ifdef INET 2591 case AF_INET: 2592 /* FALLTHROUGH */ 2593 #endif /* INET */ 2594 #ifdef INET6 2595 case AF_INET6: 2596 /* FALLTHROUGH */ 2597 #endif /* INET6 */ 2598 case AF_UNSPEC: 2599 break; 2600 default: 2601 return (EAFNOSUPPORT); 2602 } 2603 2604 if (pp->addr.addr.type != PF_ADDR_ADDRMASK && 2605 pp->addr.addr.type != PF_ADDR_DYNIFTL && 2606 pp->addr.addr.type != PF_ADDR_TABLE) 2607 return (EINVAL); 2608 2609 if (pp->addr.addr.p.dyn != NULL) 2610 return (EINVAL); 2611 2612 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); 2613 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa); 2614 if (error != 0) 2615 goto out; 2616 if (pa->ifname[0]) 2617 kif = pf_kkif_create(M_WAITOK); 2618 PF_RULES_WLOCK(); 2619 if (pp->ticket != V_ticket_pabuf) { 2620 PF_RULES_WUNLOCK(); 2621 if (pa->ifname[0]) 2622 pf_kkif_free(kif); 2623 error = EBUSY; 2624 goto out; 2625 } 2626 if (pa->ifname[0]) { 2627 pa->kif = pfi_kkif_attach(kif, pa->ifname); 2628 kif = NULL; 2629 pfi_kkif_ref(pa->kif); 2630 } else 2631 pa->kif = NULL; 2632 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = 2633 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { 2634 if (pa->ifname[0]) 2635 pfi_kkif_unref(pa->kif); 2636 PF_RULES_WUNLOCK(); 2637 goto out; 2638 } 2639 switch (pp->which) { 2640 case PF_NAT: 2641 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries); 2642 break; 2643 case PF_RDR: 2644 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries); 2645 break; 2646 case PF_RT: 2647 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries); 2648 break; 2649 } 2650 PF_RULES_WUNLOCK(); 2651 2652 return (0); 2653 2654 out: 2655 free(pa, M_PFRULE); 2656 return (error); 2657 } 2658 2659 int 2660 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp) 2661 { 2662 struct pf_kpool *pool; 2663 struct pf_kpooladdr *pa; 2664 2665 PF_RULES_RLOCK_TRACKER; 2666 2667 if (pp->which != PF_RDR && pp->which != PF_NAT && 2668 pp->which != PF_RT) 2669 return (EINVAL); 2670 2671 pp->anchor[sizeof(pp->anchor) - 1] = 0; 2672 pp->nr = 0; 2673 2674 PF_RULES_RLOCK(); 2675 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2676 pp->r_num, 0, 1, 0, pp->which); 2677 if (pool == NULL) { 2678 PF_RULES_RUNLOCK(); 2679 return (EBUSY); 2680 } 2681 TAILQ_FOREACH(pa, &pool->list, entries) 2682 pp->nr++; 2683 PF_RULES_RUNLOCK(); 2684 2685 return (0); 2686 } 2687 2688 int 2689 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp) 2690 { 2691 struct pf_kpool *pool; 2692 struct pf_kpooladdr *pa; 2693 u_int32_t nr = 0; 2694 2695 if (pp->which != PF_RDR && pp->which != PF_NAT && 2696 pp->which != PF_RT) 2697 return (EINVAL); 2698 2699 PF_RULES_RLOCK_TRACKER; 2700 2701 pp->anchor[sizeof(pp->anchor) - 1] = 0; 2702 2703 PF_RULES_RLOCK(); 2704 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2705 pp->r_num, 0, 1, 1, pp->which); 2706 if (pool == NULL) { 2707 PF_RULES_RUNLOCK(); 2708 return (EBUSY); 2709 } 2710 pa = TAILQ_FIRST(&pool->list); 2711 while ((pa != NULL) && (nr < pp->nr)) { 2712 pa = TAILQ_NEXT(pa, entries); 2713 nr++; 2714 } 2715 if (pa == NULL) { 2716 PF_RULES_RUNLOCK(); 2717 return (EBUSY); 2718 } 2719 pf_kpooladdr_to_pooladdr(pa, &pp->addr); 2720 pf_addr_copyout(&pp->addr.addr); 2721 PF_RULES_RUNLOCK(); 2722 2723 return (0); 2724 } 2725 2726 int 2727 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr) 2728 { 2729 struct pf_kruleset *ruleset; 2730 struct pf_kanchor *anchor; 2731 2732 PF_RULES_RLOCK_TRACKER; 2733 2734 pr->path[sizeof(pr->path) - 1] = 0; 2735 2736 PF_RULES_RLOCK(); 2737 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2738 PF_RULES_RUNLOCK(); 2739 return (ENOENT); 2740 } 2741 pr->nr = 0; 2742 if (ruleset->anchor == NULL) { 2743 /* XXX kludge for pf_main_ruleset */ 2744 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2745 if (anchor->parent == NULL) 2746 pr->nr++; 2747 } else { 2748 RB_FOREACH(anchor, pf_kanchor_node, 2749 &ruleset->anchor->children) 2750 pr->nr++; 2751 } 2752 PF_RULES_RUNLOCK(); 2753 2754 return (0); 2755 } 2756 2757 int 2758 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr) 2759 { 2760 struct pf_kruleset *ruleset; 2761 struct pf_kanchor *anchor; 2762 u_int32_t nr = 0; 2763 int error = 0; 2764 2765 PF_RULES_RLOCK_TRACKER; 2766 2767 PF_RULES_RLOCK(); 2768 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2769 PF_RULES_RUNLOCK(); 2770 return (ENOENT); 2771 } 2772 2773 pr->name[0] = 0; 2774 if (ruleset->anchor == NULL) { 2775 /* XXX kludge for pf_main_ruleset */ 2776 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2777 if (anchor->parent == NULL && nr++ == pr->nr) { 2778 strlcpy(pr->name, anchor->name, 2779 sizeof(pr->name)); 2780 break; 2781 } 2782 } else { 2783 RB_FOREACH(anchor, pf_kanchor_node, 2784 &ruleset->anchor->children) 2785 if (nr++ == pr->nr) { 2786 strlcpy(pr->name, anchor->name, 2787 sizeof(pr->name)); 2788 break; 2789 } 2790 } 2791 if (!pr->name[0]) 2792 error = EBUSY; 2793 PF_RULES_RUNLOCK(); 2794 2795 return (error); 2796 } 2797 2798 static int 2799 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) 2800 { 2801 int error = 0; 2802 PF_RULES_RLOCK_TRACKER; 2803 2804 #define ERROUT_IOCTL(target, x) \ 2805 do { \ 2806 error = (x); \ 2807 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \ 2808 goto target; \ 2809 } while (0) 2810 2811 2812 /* XXX keep in sync with switch() below */ 2813 if (securelevel_gt(td->td_ucred, 2)) 2814 switch (cmd) { 2815 case DIOCGETRULES: 2816 case DIOCGETRULENV: 2817 case DIOCGETADDRS: 2818 case DIOCGETADDR: 2819 case DIOCGETSTATE: 2820 case DIOCGETSTATENV: 2821 case DIOCSETSTATUSIF: 2822 case DIOCGETSTATUSNV: 2823 case DIOCCLRSTATUS: 2824 case DIOCNATLOOK: 2825 case DIOCSETDEBUG: 2826 #ifdef COMPAT_FREEBSD14 2827 case DIOCGETSTATES: 2828 case DIOCGETSTATESV2: 2829 #endif 2830 case DIOCGETTIMEOUT: 2831 case DIOCCLRRULECTRS: 2832 case DIOCGETLIMIT: 2833 case DIOCGETALTQSV0: 2834 case DIOCGETALTQSV1: 2835 case DIOCGETALTQV0: 2836 case DIOCGETALTQV1: 2837 case DIOCGETQSTATSV0: 2838 case DIOCGETQSTATSV1: 2839 case DIOCGETRULESETS: 2840 case DIOCGETRULESET: 2841 case DIOCRGETTABLES: 2842 case DIOCRGETTSTATS: 2843 case DIOCRCLRTSTATS: 2844 case DIOCRCLRADDRS: 2845 case DIOCRADDADDRS: 2846 case DIOCRDELADDRS: 2847 case DIOCRSETADDRS: 2848 case DIOCRGETADDRS: 2849 case DIOCRGETASTATS: 2850 case DIOCRCLRASTATS: 2851 case DIOCRTSTADDRS: 2852 case DIOCOSFPGET: 2853 case DIOCGETSRCNODES: 2854 case DIOCCLRSRCNODES: 2855 case DIOCGETSYNCOOKIES: 2856 case DIOCIGETIFACES: 2857 case DIOCGIFSPEEDV0: 2858 case DIOCGIFSPEEDV1: 2859 case DIOCSETIFFLAG: 2860 case DIOCCLRIFFLAG: 2861 case DIOCGETETHRULES: 2862 case DIOCGETETHRULE: 2863 case DIOCGETETHRULESETS: 2864 case DIOCGETETHRULESET: 2865 break; 2866 case DIOCRCLRTABLES: 2867 case DIOCRADDTABLES: 2868 case DIOCRDELTABLES: 2869 case DIOCRSETTFLAGS: 2870 if (((struct pfioc_table *)addr)->pfrio_flags & 2871 PFR_FLAG_DUMMY) 2872 break; /* dummy operation ok */ 2873 return (EPERM); 2874 default: 2875 return (EPERM); 2876 } 2877 2878 if (!(flags & FWRITE)) 2879 switch (cmd) { 2880 case DIOCGETRULES: 2881 case DIOCGETADDRS: 2882 case DIOCGETADDR: 2883 case DIOCGETSTATE: 2884 case DIOCGETSTATENV: 2885 case DIOCGETSTATUSNV: 2886 #ifdef COMPAT_FREEBSD14 2887 case DIOCGETSTATES: 2888 case DIOCGETSTATESV2: 2889 #endif 2890 case DIOCGETTIMEOUT: 2891 case DIOCGETLIMIT: 2892 case DIOCGETALTQSV0: 2893 case DIOCGETALTQSV1: 2894 case DIOCGETALTQV0: 2895 case DIOCGETALTQV1: 2896 case DIOCGETQSTATSV0: 2897 case DIOCGETQSTATSV1: 2898 case DIOCGETRULESETS: 2899 case DIOCGETRULESET: 2900 case DIOCNATLOOK: 2901 case DIOCRGETTABLES: 2902 case DIOCRGETTSTATS: 2903 case DIOCRGETADDRS: 2904 case DIOCRGETASTATS: 2905 case DIOCRTSTADDRS: 2906 case DIOCOSFPGET: 2907 case DIOCGETSRCNODES: 2908 case DIOCGETSYNCOOKIES: 2909 case DIOCIGETIFACES: 2910 case DIOCGIFSPEEDV1: 2911 case DIOCGIFSPEEDV0: 2912 case DIOCGETRULENV: 2913 case DIOCGETETHRULES: 2914 case DIOCGETETHRULE: 2915 case DIOCGETETHRULESETS: 2916 case DIOCGETETHRULESET: 2917 break; 2918 case DIOCRCLRTABLES: 2919 case DIOCRADDTABLES: 2920 case DIOCRDELTABLES: 2921 case DIOCRCLRTSTATS: 2922 case DIOCRCLRADDRS: 2923 case DIOCRADDADDRS: 2924 case DIOCRDELADDRS: 2925 case DIOCRSETADDRS: 2926 case DIOCRSETTFLAGS: 2927 if (((struct pfioc_table *)addr)->pfrio_flags & 2928 PFR_FLAG_DUMMY) { 2929 flags |= FWRITE; /* need write lock for dummy */ 2930 break; /* dummy operation ok */ 2931 } 2932 return (EACCES); 2933 default: 2934 return (EACCES); 2935 } 2936 2937 CURVNET_SET(TD_TO_VNET(td)); 2938 2939 switch (cmd) { 2940 #ifdef COMPAT_FREEBSD14 2941 case DIOCSTART: 2942 error = pf_start(); 2943 break; 2944 2945 case DIOCSTOP: 2946 error = pf_stop(); 2947 break; 2948 #endif 2949 2950 case DIOCGETETHRULES: { 2951 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2952 nvlist_t *nvl; 2953 void *packed; 2954 struct pf_keth_rule *tail; 2955 struct pf_keth_ruleset *rs; 2956 u_int32_t ticket, nr; 2957 const char *anchor = ""; 2958 2959 nvl = NULL; 2960 packed = NULL; 2961 2962 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x) 2963 2964 if (nv->len > pf_ioctl_maxcount) 2965 ERROUT(ENOMEM); 2966 2967 /* Copy the request in */ 2968 packed = malloc(nv->len, M_NVLIST, M_WAITOK); 2969 error = copyin(nv->data, packed, nv->len); 2970 if (error) 2971 ERROUT(error); 2972 2973 nvl = nvlist_unpack(packed, nv->len, 0); 2974 if (nvl == NULL) 2975 ERROUT(EBADMSG); 2976 2977 if (! nvlist_exists_string(nvl, "anchor")) 2978 ERROUT(EBADMSG); 2979 2980 anchor = nvlist_get_string(nvl, "anchor"); 2981 2982 rs = pf_find_keth_ruleset(anchor); 2983 2984 nvlist_destroy(nvl); 2985 nvl = NULL; 2986 free(packed, M_NVLIST); 2987 packed = NULL; 2988 2989 if (rs == NULL) 2990 ERROUT(ENOENT); 2991 2992 /* Reply */ 2993 nvl = nvlist_create(0); 2994 if (nvl == NULL) 2995 ERROUT(ENOMEM); 2996 2997 PF_RULES_RLOCK(); 2998 2999 ticket = rs->active.ticket; 3000 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq); 3001 if (tail) 3002 nr = tail->nr + 1; 3003 else 3004 nr = 0; 3005 3006 PF_RULES_RUNLOCK(); 3007 3008 nvlist_add_number(nvl, "ticket", ticket); 3009 nvlist_add_number(nvl, "nr", nr); 3010 3011 packed = nvlist_pack(nvl, &nv->len); 3012 if (packed == NULL) 3013 ERROUT(ENOMEM); 3014 3015 if (nv->size == 0) 3016 ERROUT(0); 3017 else if (nv->size < nv->len) 3018 ERROUT(ENOSPC); 3019 3020 error = copyout(packed, nv->data, nv->len); 3021 3022 #undef ERROUT 3023 DIOCGETETHRULES_error: 3024 free(packed, M_NVLIST); 3025 nvlist_destroy(nvl); 3026 break; 3027 } 3028 3029 case DIOCGETETHRULE: { 3030 struct epoch_tracker et; 3031 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3032 nvlist_t *nvl = NULL; 3033 void *nvlpacked = NULL; 3034 struct pf_keth_rule *rule = NULL; 3035 struct pf_keth_ruleset *rs; 3036 u_int32_t ticket, nr; 3037 bool clear = false; 3038 const char *anchor; 3039 3040 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x) 3041 3042 if (nv->len > pf_ioctl_maxcount) 3043 ERROUT(ENOMEM); 3044 3045 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3046 error = copyin(nv->data, nvlpacked, nv->len); 3047 if (error) 3048 ERROUT(error); 3049 3050 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3051 if (nvl == NULL) 3052 ERROUT(EBADMSG); 3053 if (! nvlist_exists_number(nvl, "ticket")) 3054 ERROUT(EBADMSG); 3055 ticket = nvlist_get_number(nvl, "ticket"); 3056 if (! nvlist_exists_string(nvl, "anchor")) 3057 ERROUT(EBADMSG); 3058 anchor = nvlist_get_string(nvl, "anchor"); 3059 3060 if (nvlist_exists_bool(nvl, "clear")) 3061 clear = nvlist_get_bool(nvl, "clear"); 3062 3063 if (clear && !(flags & FWRITE)) 3064 ERROUT(EACCES); 3065 3066 if (! nvlist_exists_number(nvl, "nr")) 3067 ERROUT(EBADMSG); 3068 nr = nvlist_get_number(nvl, "nr"); 3069 3070 PF_RULES_RLOCK(); 3071 rs = pf_find_keth_ruleset(anchor); 3072 if (rs == NULL) { 3073 PF_RULES_RUNLOCK(); 3074 ERROUT(ENOENT); 3075 } 3076 if (ticket != rs->active.ticket) { 3077 PF_RULES_RUNLOCK(); 3078 ERROUT(EBUSY); 3079 } 3080 3081 nvlist_destroy(nvl); 3082 nvl = NULL; 3083 free(nvlpacked, M_NVLIST); 3084 nvlpacked = NULL; 3085 3086 rule = TAILQ_FIRST(rs->active.rules); 3087 while ((rule != NULL) && (rule->nr != nr)) 3088 rule = TAILQ_NEXT(rule, entries); 3089 if (rule == NULL) { 3090 PF_RULES_RUNLOCK(); 3091 ERROUT(ENOENT); 3092 } 3093 /* Make sure rule can't go away. */ 3094 NET_EPOCH_ENTER(et); 3095 PF_RULES_RUNLOCK(); 3096 nvl = pf_keth_rule_to_nveth_rule(rule); 3097 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) { 3098 NET_EPOCH_EXIT(et); 3099 ERROUT(EBUSY); 3100 } 3101 NET_EPOCH_EXIT(et); 3102 if (nvl == NULL) 3103 ERROUT(ENOMEM); 3104 3105 nvlpacked = nvlist_pack(nvl, &nv->len); 3106 if (nvlpacked == NULL) 3107 ERROUT(ENOMEM); 3108 3109 if (nv->size == 0) 3110 ERROUT(0); 3111 else if (nv->size < nv->len) 3112 ERROUT(ENOSPC); 3113 3114 error = copyout(nvlpacked, nv->data, nv->len); 3115 if (error == 0 && clear) { 3116 counter_u64_zero(rule->evaluations); 3117 for (int i = 0; i < 2; i++) { 3118 counter_u64_zero(rule->packets[i]); 3119 counter_u64_zero(rule->bytes[i]); 3120 } 3121 } 3122 3123 #undef ERROUT 3124 DIOCGETETHRULE_error: 3125 free(nvlpacked, M_NVLIST); 3126 nvlist_destroy(nvl); 3127 break; 3128 } 3129 3130 case DIOCADDETHRULE: { 3131 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3132 nvlist_t *nvl = NULL; 3133 void *nvlpacked = NULL; 3134 struct pf_keth_rule *rule = NULL, *tail = NULL; 3135 struct pf_keth_ruleset *ruleset = NULL; 3136 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL; 3137 const char *anchor = "", *anchor_call = ""; 3138 3139 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x) 3140 3141 if (nv->len > pf_ioctl_maxcount) 3142 ERROUT(ENOMEM); 3143 3144 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3145 error = copyin(nv->data, nvlpacked, nv->len); 3146 if (error) 3147 ERROUT(error); 3148 3149 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3150 if (nvl == NULL) 3151 ERROUT(EBADMSG); 3152 3153 if (! nvlist_exists_number(nvl, "ticket")) 3154 ERROUT(EBADMSG); 3155 3156 if (nvlist_exists_string(nvl, "anchor")) 3157 anchor = nvlist_get_string(nvl, "anchor"); 3158 if (nvlist_exists_string(nvl, "anchor_call")) 3159 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3160 3161 ruleset = pf_find_keth_ruleset(anchor); 3162 if (ruleset == NULL) 3163 ERROUT(EINVAL); 3164 3165 if (nvlist_get_number(nvl, "ticket") != 3166 ruleset->inactive.ticket) { 3167 DPFPRINTF(PF_DEBUG_MISC, 3168 ("ticket: %d != %d\n", 3169 (u_int32_t)nvlist_get_number(nvl, "ticket"), 3170 ruleset->inactive.ticket)); 3171 ERROUT(EBUSY); 3172 } 3173 3174 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); 3175 rule->timestamp = NULL; 3176 3177 error = pf_nveth_rule_to_keth_rule(nvl, rule); 3178 if (error != 0) 3179 ERROUT(error); 3180 3181 if (rule->ifname[0]) 3182 kif = pf_kkif_create(M_WAITOK); 3183 if (rule->bridge_to_name[0]) 3184 bridge_to_kif = pf_kkif_create(M_WAITOK); 3185 rule->evaluations = counter_u64_alloc(M_WAITOK); 3186 for (int i = 0; i < 2; i++) { 3187 rule->packets[i] = counter_u64_alloc(M_WAITOK); 3188 rule->bytes[i] = counter_u64_alloc(M_WAITOK); 3189 } 3190 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 3191 M_WAITOK | M_ZERO); 3192 3193 PF_RULES_WLOCK(); 3194 3195 if (rule->ifname[0]) { 3196 rule->kif = pfi_kkif_attach(kif, rule->ifname); 3197 pfi_kkif_ref(rule->kif); 3198 } else 3199 rule->kif = NULL; 3200 if (rule->bridge_to_name[0]) { 3201 rule->bridge_to = pfi_kkif_attach(bridge_to_kif, 3202 rule->bridge_to_name); 3203 pfi_kkif_ref(rule->bridge_to); 3204 } else 3205 rule->bridge_to = NULL; 3206 3207 #ifdef ALTQ 3208 /* set queue IDs */ 3209 if (rule->qname[0] != 0) { 3210 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 3211 error = EBUSY; 3212 else 3213 rule->qid = rule->qid; 3214 } 3215 #endif 3216 if (rule->tagname[0]) 3217 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 3218 error = EBUSY; 3219 if (rule->match_tagname[0]) 3220 if ((rule->match_tag = pf_tagname2tag( 3221 rule->match_tagname)) == 0) 3222 error = EBUSY; 3223 3224 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE) 3225 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr); 3226 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE) 3227 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr); 3228 3229 if (error) { 3230 pf_free_eth_rule(rule); 3231 PF_RULES_WUNLOCK(); 3232 ERROUT(error); 3233 } 3234 3235 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) { 3236 pf_free_eth_rule(rule); 3237 PF_RULES_WUNLOCK(); 3238 ERROUT(EINVAL); 3239 } 3240 3241 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq); 3242 if (tail) 3243 rule->nr = tail->nr + 1; 3244 else 3245 rule->nr = 0; 3246 3247 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries); 3248 3249 PF_RULES_WUNLOCK(); 3250 3251 #undef ERROUT 3252 DIOCADDETHRULE_error: 3253 nvlist_destroy(nvl); 3254 free(nvlpacked, M_NVLIST); 3255 break; 3256 } 3257 3258 case DIOCGETETHRULESETS: { 3259 struct epoch_tracker et; 3260 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3261 nvlist_t *nvl = NULL; 3262 void *nvlpacked = NULL; 3263 struct pf_keth_ruleset *ruleset; 3264 struct pf_keth_anchor *anchor; 3265 int nr = 0; 3266 3267 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x) 3268 3269 if (nv->len > pf_ioctl_maxcount) 3270 ERROUT(ENOMEM); 3271 3272 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3273 error = copyin(nv->data, nvlpacked, nv->len); 3274 if (error) 3275 ERROUT(error); 3276 3277 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3278 if (nvl == NULL) 3279 ERROUT(EBADMSG); 3280 if (! nvlist_exists_string(nvl, "path")) 3281 ERROUT(EBADMSG); 3282 3283 NET_EPOCH_ENTER(et); 3284 3285 if ((ruleset = pf_find_keth_ruleset( 3286 nvlist_get_string(nvl, "path"))) == NULL) { 3287 NET_EPOCH_EXIT(et); 3288 ERROUT(ENOENT); 3289 } 3290 3291 if (ruleset->anchor == NULL) { 3292 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors) 3293 if (anchor->parent == NULL) 3294 nr++; 3295 } else { 3296 RB_FOREACH(anchor, pf_keth_anchor_node, 3297 &ruleset->anchor->children) 3298 nr++; 3299 } 3300 3301 NET_EPOCH_EXIT(et); 3302 3303 nvlist_destroy(nvl); 3304 nvl = NULL; 3305 free(nvlpacked, M_NVLIST); 3306 nvlpacked = NULL; 3307 3308 nvl = nvlist_create(0); 3309 if (nvl == NULL) 3310 ERROUT(ENOMEM); 3311 3312 nvlist_add_number(nvl, "nr", nr); 3313 3314 nvlpacked = nvlist_pack(nvl, &nv->len); 3315 if (nvlpacked == NULL) 3316 ERROUT(ENOMEM); 3317 3318 if (nv->size == 0) 3319 ERROUT(0); 3320 else if (nv->size < nv->len) 3321 ERROUT(ENOSPC); 3322 3323 error = copyout(nvlpacked, nv->data, nv->len); 3324 3325 #undef ERROUT 3326 DIOCGETETHRULESETS_error: 3327 free(nvlpacked, M_NVLIST); 3328 nvlist_destroy(nvl); 3329 break; 3330 } 3331 3332 case DIOCGETETHRULESET: { 3333 struct epoch_tracker et; 3334 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3335 nvlist_t *nvl = NULL; 3336 void *nvlpacked = NULL; 3337 struct pf_keth_ruleset *ruleset; 3338 struct pf_keth_anchor *anchor; 3339 int nr = 0, req_nr = 0; 3340 bool found = false; 3341 3342 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x) 3343 3344 if (nv->len > pf_ioctl_maxcount) 3345 ERROUT(ENOMEM); 3346 3347 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3348 error = copyin(nv->data, nvlpacked, nv->len); 3349 if (error) 3350 ERROUT(error); 3351 3352 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3353 if (nvl == NULL) 3354 ERROUT(EBADMSG); 3355 if (! nvlist_exists_string(nvl, "path")) 3356 ERROUT(EBADMSG); 3357 if (! nvlist_exists_number(nvl, "nr")) 3358 ERROUT(EBADMSG); 3359 3360 req_nr = nvlist_get_number(nvl, "nr"); 3361 3362 NET_EPOCH_ENTER(et); 3363 3364 if ((ruleset = pf_find_keth_ruleset( 3365 nvlist_get_string(nvl, "path"))) == NULL) { 3366 NET_EPOCH_EXIT(et); 3367 ERROUT(ENOENT); 3368 } 3369 3370 nvlist_destroy(nvl); 3371 nvl = NULL; 3372 free(nvlpacked, M_NVLIST); 3373 nvlpacked = NULL; 3374 3375 nvl = nvlist_create(0); 3376 if (nvl == NULL) { 3377 NET_EPOCH_EXIT(et); 3378 ERROUT(ENOMEM); 3379 } 3380 3381 if (ruleset->anchor == NULL) { 3382 RB_FOREACH(anchor, pf_keth_anchor_global, 3383 &V_pf_keth_anchors) { 3384 if (anchor->parent == NULL && nr++ == req_nr) { 3385 found = true; 3386 break; 3387 } 3388 } 3389 } else { 3390 RB_FOREACH(anchor, pf_keth_anchor_node, 3391 &ruleset->anchor->children) { 3392 if (nr++ == req_nr) { 3393 found = true; 3394 break; 3395 } 3396 } 3397 } 3398 3399 NET_EPOCH_EXIT(et); 3400 if (found) { 3401 nvlist_add_number(nvl, "nr", nr); 3402 nvlist_add_string(nvl, "name", anchor->name); 3403 if (ruleset->anchor) 3404 nvlist_add_string(nvl, "path", 3405 ruleset->anchor->path); 3406 else 3407 nvlist_add_string(nvl, "path", ""); 3408 } else { 3409 ERROUT(EBUSY); 3410 } 3411 3412 nvlpacked = nvlist_pack(nvl, &nv->len); 3413 if (nvlpacked == NULL) 3414 ERROUT(ENOMEM); 3415 3416 if (nv->size == 0) 3417 ERROUT(0); 3418 else if (nv->size < nv->len) 3419 ERROUT(ENOSPC); 3420 3421 error = copyout(nvlpacked, nv->data, nv->len); 3422 3423 #undef ERROUT 3424 DIOCGETETHRULESET_error: 3425 free(nvlpacked, M_NVLIST); 3426 nvlist_destroy(nvl); 3427 break; 3428 } 3429 3430 case DIOCADDRULENV: { 3431 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3432 nvlist_t *nvl = NULL; 3433 void *nvlpacked = NULL; 3434 struct pf_krule *rule = NULL; 3435 const char *anchor = "", *anchor_call = ""; 3436 uint32_t ticket = 0, pool_ticket = 0; 3437 3438 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x) 3439 3440 if (nv->len > pf_ioctl_maxcount) 3441 ERROUT(ENOMEM); 3442 3443 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3444 error = copyin(nv->data, nvlpacked, nv->len); 3445 if (error) 3446 ERROUT(error); 3447 3448 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3449 if (nvl == NULL) 3450 ERROUT(EBADMSG); 3451 3452 if (! nvlist_exists_number(nvl, "ticket")) 3453 ERROUT(EINVAL); 3454 ticket = nvlist_get_number(nvl, "ticket"); 3455 3456 if (! nvlist_exists_number(nvl, "pool_ticket")) 3457 ERROUT(EINVAL); 3458 pool_ticket = nvlist_get_number(nvl, "pool_ticket"); 3459 3460 if (! nvlist_exists_nvlist(nvl, "rule")) 3461 ERROUT(EINVAL); 3462 3463 rule = pf_krule_alloc(); 3464 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"), 3465 rule); 3466 if (error) 3467 ERROUT(error); 3468 3469 if (nvlist_exists_string(nvl, "anchor")) 3470 anchor = nvlist_get_string(nvl, "anchor"); 3471 if (nvlist_exists_string(nvl, "anchor_call")) 3472 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3473 3474 if ((error = nvlist_error(nvl))) 3475 ERROUT(error); 3476 3477 /* Frees rule on error */ 3478 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor, 3479 anchor_call, td->td_ucred->cr_ruid, 3480 td->td_proc ? td->td_proc->p_pid : 0); 3481 3482 nvlist_destroy(nvl); 3483 free(nvlpacked, M_NVLIST); 3484 break; 3485 #undef ERROUT 3486 DIOCADDRULENV_error: 3487 pf_krule_free(rule); 3488 nvlist_destroy(nvl); 3489 free(nvlpacked, M_NVLIST); 3490 3491 break; 3492 } 3493 case DIOCADDRULE: { 3494 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3495 struct pf_krule *rule; 3496 3497 rule = pf_krule_alloc(); 3498 error = pf_rule_to_krule(&pr->rule, rule); 3499 if (error != 0) { 3500 pf_krule_free(rule); 3501 break; 3502 } 3503 3504 pr->anchor[sizeof(pr->anchor) - 1] = 0; 3505 3506 /* Frees rule on error */ 3507 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, 3508 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid, 3509 td->td_proc ? td->td_proc->p_pid : 0); 3510 break; 3511 } 3512 3513 case DIOCGETRULES: { 3514 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3515 3516 pr->anchor[sizeof(pr->anchor) - 1] = 0; 3517 3518 error = pf_ioctl_getrules(pr); 3519 3520 break; 3521 } 3522 3523 case DIOCGETRULENV: { 3524 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3525 nvlist_t *nvrule = NULL; 3526 nvlist_t *nvl = NULL; 3527 struct pf_kruleset *ruleset; 3528 struct pf_krule *rule; 3529 void *nvlpacked = NULL; 3530 int rs_num, nr; 3531 bool clear_counter = false; 3532 3533 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x) 3534 3535 if (nv->len > pf_ioctl_maxcount) 3536 ERROUT(ENOMEM); 3537 3538 /* Copy the request in */ 3539 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3540 error = copyin(nv->data, nvlpacked, nv->len); 3541 if (error) 3542 ERROUT(error); 3543 3544 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3545 if (nvl == NULL) 3546 ERROUT(EBADMSG); 3547 3548 if (! nvlist_exists_string(nvl, "anchor")) 3549 ERROUT(EBADMSG); 3550 if (! nvlist_exists_number(nvl, "ruleset")) 3551 ERROUT(EBADMSG); 3552 if (! nvlist_exists_number(nvl, "ticket")) 3553 ERROUT(EBADMSG); 3554 if (! nvlist_exists_number(nvl, "nr")) 3555 ERROUT(EBADMSG); 3556 3557 if (nvlist_exists_bool(nvl, "clear_counter")) 3558 clear_counter = nvlist_get_bool(nvl, "clear_counter"); 3559 3560 if (clear_counter && !(flags & FWRITE)) 3561 ERROUT(EACCES); 3562 3563 nr = nvlist_get_number(nvl, "nr"); 3564 3565 PF_RULES_WLOCK(); 3566 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor")); 3567 if (ruleset == NULL) { 3568 PF_RULES_WUNLOCK(); 3569 ERROUT(ENOENT); 3570 } 3571 3572 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset")); 3573 if (rs_num >= PF_RULESET_MAX) { 3574 PF_RULES_WUNLOCK(); 3575 ERROUT(EINVAL); 3576 } 3577 3578 if (nvlist_get_number(nvl, "ticket") != 3579 ruleset->rules[rs_num].active.ticket) { 3580 PF_RULES_WUNLOCK(); 3581 ERROUT(EBUSY); 3582 } 3583 3584 if ((error = nvlist_error(nvl))) { 3585 PF_RULES_WUNLOCK(); 3586 ERROUT(error); 3587 } 3588 3589 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3590 while ((rule != NULL) && (rule->nr != nr)) 3591 rule = TAILQ_NEXT(rule, entries); 3592 if (rule == NULL) { 3593 PF_RULES_WUNLOCK(); 3594 ERROUT(EBUSY); 3595 } 3596 3597 nvrule = pf_krule_to_nvrule(rule); 3598 3599 nvlist_destroy(nvl); 3600 nvl = nvlist_create(0); 3601 if (nvl == NULL) { 3602 PF_RULES_WUNLOCK(); 3603 ERROUT(ENOMEM); 3604 } 3605 nvlist_add_number(nvl, "nr", nr); 3606 nvlist_add_nvlist(nvl, "rule", nvrule); 3607 nvlist_destroy(nvrule); 3608 nvrule = NULL; 3609 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) { 3610 PF_RULES_WUNLOCK(); 3611 ERROUT(EBUSY); 3612 } 3613 3614 free(nvlpacked, M_NVLIST); 3615 nvlpacked = nvlist_pack(nvl, &nv->len); 3616 if (nvlpacked == NULL) { 3617 PF_RULES_WUNLOCK(); 3618 ERROUT(ENOMEM); 3619 } 3620 3621 if (nv->size == 0) { 3622 PF_RULES_WUNLOCK(); 3623 ERROUT(0); 3624 } 3625 else if (nv->size < nv->len) { 3626 PF_RULES_WUNLOCK(); 3627 ERROUT(ENOSPC); 3628 } 3629 3630 if (clear_counter) 3631 pf_krule_clear_counters(rule); 3632 3633 PF_RULES_WUNLOCK(); 3634 3635 error = copyout(nvlpacked, nv->data, nv->len); 3636 3637 #undef ERROUT 3638 DIOCGETRULENV_error: 3639 free(nvlpacked, M_NVLIST); 3640 nvlist_destroy(nvrule); 3641 nvlist_destroy(nvl); 3642 3643 break; 3644 } 3645 3646 case DIOCCHANGERULE: { 3647 struct pfioc_rule *pcr = (struct pfioc_rule *)addr; 3648 struct pf_kruleset *ruleset; 3649 struct pf_krule *oldrule = NULL, *newrule = NULL; 3650 struct pfi_kkif *kif = NULL; 3651 struct pf_kpooladdr *pa; 3652 u_int32_t nr = 0; 3653 int rs_num; 3654 3655 pcr->anchor[sizeof(pcr->anchor) - 1] = 0; 3656 3657 if (pcr->action < PF_CHANGE_ADD_HEAD || 3658 pcr->action > PF_CHANGE_GET_TICKET) { 3659 error = EINVAL; 3660 break; 3661 } 3662 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { 3663 error = EINVAL; 3664 break; 3665 } 3666 3667 if (pcr->action != PF_CHANGE_REMOVE) { 3668 newrule = pf_krule_alloc(); 3669 error = pf_rule_to_krule(&pcr->rule, newrule); 3670 if (error != 0) { 3671 pf_krule_free(newrule); 3672 break; 3673 } 3674 3675 if (newrule->ifname[0]) 3676 kif = pf_kkif_create(M_WAITOK); 3677 pf_counter_u64_init(&newrule->evaluations, M_WAITOK); 3678 for (int i = 0; i < 2; i++) { 3679 pf_counter_u64_init(&newrule->packets[i], M_WAITOK); 3680 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK); 3681 } 3682 newrule->states_cur = counter_u64_alloc(M_WAITOK); 3683 newrule->states_tot = counter_u64_alloc(M_WAITOK); 3684 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 3685 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 3686 newrule->cuid = td->td_ucred->cr_ruid; 3687 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 3688 TAILQ_INIT(&newrule->nat.list); 3689 TAILQ_INIT(&newrule->rdr.list); 3690 TAILQ_INIT(&newrule->route.list); 3691 } 3692 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x) 3693 3694 PF_CONFIG_LOCK(); 3695 PF_RULES_WLOCK(); 3696 #ifdef PF_WANT_32_TO_64_COUNTER 3697 if (newrule != NULL) { 3698 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist); 3699 newrule->allrulelinked = true; 3700 V_pf_allrulecount++; 3701 } 3702 #endif 3703 3704 if (!(pcr->action == PF_CHANGE_REMOVE || 3705 pcr->action == PF_CHANGE_GET_TICKET) && 3706 pcr->pool_ticket != V_ticket_pabuf) 3707 ERROUT(EBUSY); 3708 3709 ruleset = pf_find_kruleset(pcr->anchor); 3710 if (ruleset == NULL) 3711 ERROUT(EINVAL); 3712 3713 rs_num = pf_get_ruleset_number(pcr->rule.action); 3714 if (rs_num >= PF_RULESET_MAX) 3715 ERROUT(EINVAL); 3716 3717 /* 3718 * XXXMJG: there is no guarantee that the ruleset was 3719 * created by the usual route of calling DIOCXBEGIN. 3720 * As a result it is possible the rule tree will not 3721 * be allocated yet. Hack around it by doing it here. 3722 * Note it is fine to let the tree persist in case of 3723 * error as it will be freed down the road on future 3724 * updates (if need be). 3725 */ 3726 if (ruleset->rules[rs_num].active.tree == NULL) { 3727 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT); 3728 if (ruleset->rules[rs_num].active.tree == NULL) { 3729 ERROUT(ENOMEM); 3730 } 3731 } 3732 3733 if (pcr->action == PF_CHANGE_GET_TICKET) { 3734 pcr->ticket = ++ruleset->rules[rs_num].active.ticket; 3735 ERROUT(0); 3736 } else if (pcr->ticket != 3737 ruleset->rules[rs_num].active.ticket) 3738 ERROUT(EINVAL); 3739 3740 if (pcr->action != PF_CHANGE_REMOVE) { 3741 if (newrule->ifname[0]) { 3742 newrule->kif = pfi_kkif_attach(kif, 3743 newrule->ifname); 3744 kif = NULL; 3745 pfi_kkif_ref(newrule->kif); 3746 } else 3747 newrule->kif = NULL; 3748 3749 if (newrule->rtableid > 0 && 3750 newrule->rtableid >= rt_numfibs) 3751 error = EBUSY; 3752 3753 #ifdef ALTQ 3754 /* set queue IDs */ 3755 if (newrule->qname[0] != 0) { 3756 if ((newrule->qid = 3757 pf_qname2qid(newrule->qname)) == 0) 3758 error = EBUSY; 3759 else if (newrule->pqname[0] != 0) { 3760 if ((newrule->pqid = 3761 pf_qname2qid(newrule->pqname)) == 0) 3762 error = EBUSY; 3763 } else 3764 newrule->pqid = newrule->qid; 3765 } 3766 #endif /* ALTQ */ 3767 if (newrule->tagname[0]) 3768 if ((newrule->tag = 3769 pf_tagname2tag(newrule->tagname)) == 0) 3770 error = EBUSY; 3771 if (newrule->match_tagname[0]) 3772 if ((newrule->match_tag = pf_tagname2tag( 3773 newrule->match_tagname)) == 0) 3774 error = EBUSY; 3775 if (newrule->rt && !newrule->direction) 3776 error = EINVAL; 3777 if (!newrule->log) 3778 newrule->logif = 0; 3779 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) 3780 error = ENOMEM; 3781 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) 3782 error = ENOMEM; 3783 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call)) 3784 error = EINVAL; 3785 for (int i = 0; i < 3; i++) { 3786 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 3787 if (pa->addr.type == PF_ADDR_TABLE) { 3788 pa->addr.p.tbl = 3789 pfr_attach_table(ruleset, 3790 pa->addr.v.tblname); 3791 if (pa->addr.p.tbl == NULL) 3792 error = ENOMEM; 3793 } 3794 } 3795 3796 newrule->overload_tbl = NULL; 3797 if (newrule->overload_tblname[0]) { 3798 if ((newrule->overload_tbl = pfr_attach_table( 3799 ruleset, newrule->overload_tblname)) == 3800 NULL) 3801 error = EINVAL; 3802 else 3803 newrule->overload_tbl->pfrkt_flags |= 3804 PFR_TFLAG_ACTIVE; 3805 } 3806 3807 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list); 3808 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list); 3809 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list); 3810 if (((((newrule->action == PF_NAT) || 3811 (newrule->action == PF_RDR) || 3812 (newrule->action == PF_BINAT) || 3813 (newrule->rt > PF_NOPFROUTE)) && 3814 !newrule->anchor)) && 3815 (TAILQ_FIRST(&newrule->rdr.list) == NULL)) 3816 error = EINVAL; 3817 3818 if (error) { 3819 pf_free_rule(newrule); 3820 PF_RULES_WUNLOCK(); 3821 PF_CONFIG_UNLOCK(); 3822 break; 3823 } 3824 3825 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list); 3826 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list); 3827 } 3828 pf_empty_kpool(&V_pf_pabuf[0]); 3829 pf_empty_kpool(&V_pf_pabuf[1]); 3830 pf_empty_kpool(&V_pf_pabuf[2]); 3831 3832 if (pcr->action == PF_CHANGE_ADD_HEAD) 3833 oldrule = TAILQ_FIRST( 3834 ruleset->rules[rs_num].active.ptr); 3835 else if (pcr->action == PF_CHANGE_ADD_TAIL) 3836 oldrule = TAILQ_LAST( 3837 ruleset->rules[rs_num].active.ptr, pf_krulequeue); 3838 else { 3839 oldrule = TAILQ_FIRST( 3840 ruleset->rules[rs_num].active.ptr); 3841 while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) 3842 oldrule = TAILQ_NEXT(oldrule, entries); 3843 if (oldrule == NULL) { 3844 if (newrule != NULL) 3845 pf_free_rule(newrule); 3846 PF_RULES_WUNLOCK(); 3847 PF_CONFIG_UNLOCK(); 3848 error = EINVAL; 3849 break; 3850 } 3851 } 3852 3853 if (pcr->action == PF_CHANGE_REMOVE) { 3854 pf_unlink_rule(ruleset->rules[rs_num].active.ptr, 3855 oldrule); 3856 RB_REMOVE(pf_krule_global, 3857 ruleset->rules[rs_num].active.tree, oldrule); 3858 ruleset->rules[rs_num].active.rcount--; 3859 } else { 3860 pf_hash_rule(newrule); 3861 if (RB_INSERT(pf_krule_global, 3862 ruleset->rules[rs_num].active.tree, newrule) != NULL) { 3863 pf_free_rule(newrule); 3864 PF_RULES_WUNLOCK(); 3865 PF_CONFIG_UNLOCK(); 3866 error = EEXIST; 3867 break; 3868 } 3869 3870 if (oldrule == NULL) 3871 TAILQ_INSERT_TAIL( 3872 ruleset->rules[rs_num].active.ptr, 3873 newrule, entries); 3874 else if (pcr->action == PF_CHANGE_ADD_HEAD || 3875 pcr->action == PF_CHANGE_ADD_BEFORE) 3876 TAILQ_INSERT_BEFORE(oldrule, newrule, entries); 3877 else 3878 TAILQ_INSERT_AFTER( 3879 ruleset->rules[rs_num].active.ptr, 3880 oldrule, newrule, entries); 3881 ruleset->rules[rs_num].active.rcount++; 3882 } 3883 3884 nr = 0; 3885 TAILQ_FOREACH(oldrule, 3886 ruleset->rules[rs_num].active.ptr, entries) 3887 oldrule->nr = nr++; 3888 3889 ruleset->rules[rs_num].active.ticket++; 3890 3891 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); 3892 pf_remove_if_empty_kruleset(ruleset); 3893 3894 PF_RULES_WUNLOCK(); 3895 PF_CONFIG_UNLOCK(); 3896 break; 3897 3898 #undef ERROUT 3899 DIOCCHANGERULE_error: 3900 PF_RULES_WUNLOCK(); 3901 PF_CONFIG_UNLOCK(); 3902 pf_krule_free(newrule); 3903 pf_kkif_free(kif); 3904 break; 3905 } 3906 3907 case DIOCCLRSTATESNV: { 3908 error = pf_clearstates_nv((struct pfioc_nv *)addr); 3909 break; 3910 } 3911 3912 case DIOCKILLSTATESNV: { 3913 error = pf_killstates_nv((struct pfioc_nv *)addr); 3914 break; 3915 } 3916 3917 case DIOCADDSTATE: { 3918 struct pfioc_state *ps = (struct pfioc_state *)addr; 3919 struct pfsync_state_1301 *sp = &ps->state; 3920 3921 if (sp->timeout >= PFTM_MAX) { 3922 error = EINVAL; 3923 break; 3924 } 3925 if (V_pfsync_state_import_ptr != NULL) { 3926 PF_RULES_RLOCK(); 3927 error = V_pfsync_state_import_ptr( 3928 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL, 3929 PFSYNC_MSG_VERSION_1301); 3930 PF_RULES_RUNLOCK(); 3931 } else 3932 error = EOPNOTSUPP; 3933 break; 3934 } 3935 3936 case DIOCGETSTATE: { 3937 struct pfioc_state *ps = (struct pfioc_state *)addr; 3938 struct pf_kstate *s; 3939 3940 s = pf_find_state_byid(ps->state.id, ps->state.creatorid); 3941 if (s == NULL) { 3942 error = ENOENT; 3943 break; 3944 } 3945 3946 pfsync_state_export((union pfsync_state_union*)&ps->state, 3947 s, PFSYNC_MSG_VERSION_1301); 3948 PF_STATE_UNLOCK(s); 3949 break; 3950 } 3951 3952 case DIOCGETSTATENV: { 3953 error = pf_getstate((struct pfioc_nv *)addr); 3954 break; 3955 } 3956 3957 #ifdef COMPAT_FREEBSD14 3958 case DIOCGETSTATES: { 3959 struct pfioc_states *ps = (struct pfioc_states *)addr; 3960 struct pf_kstate *s; 3961 struct pfsync_state_1301 *pstore, *p; 3962 int i, nr; 3963 size_t slice_count = 16, count; 3964 void *out; 3965 3966 if (ps->ps_len <= 0) { 3967 nr = uma_zone_get_cur(V_pf_state_z); 3968 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 3969 break; 3970 } 3971 3972 out = ps->ps_states; 3973 pstore = mallocarray(slice_count, 3974 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO); 3975 nr = 0; 3976 3977 for (i = 0; i <= V_pf_hashmask; i++) { 3978 struct pf_idhash *ih = &V_pf_idhash[i]; 3979 3980 DIOCGETSTATES_retry: 3981 p = pstore; 3982 3983 if (LIST_EMPTY(&ih->states)) 3984 continue; 3985 3986 PF_HASHROW_LOCK(ih); 3987 count = 0; 3988 LIST_FOREACH(s, &ih->states, entry) { 3989 if (s->timeout == PFTM_UNLINKED) 3990 continue; 3991 count++; 3992 } 3993 3994 if (count > slice_count) { 3995 PF_HASHROW_UNLOCK(ih); 3996 free(pstore, M_TEMP); 3997 slice_count = count * 2; 3998 pstore = mallocarray(slice_count, 3999 sizeof(struct pfsync_state_1301), M_TEMP, 4000 M_WAITOK | M_ZERO); 4001 goto DIOCGETSTATES_retry; 4002 } 4003 4004 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4005 PF_HASHROW_UNLOCK(ih); 4006 goto DIOCGETSTATES_full; 4007 } 4008 4009 LIST_FOREACH(s, &ih->states, entry) { 4010 if (s->timeout == PFTM_UNLINKED) 4011 continue; 4012 4013 pfsync_state_export((union pfsync_state_union*)p, 4014 s, PFSYNC_MSG_VERSION_1301); 4015 p++; 4016 nr++; 4017 } 4018 PF_HASHROW_UNLOCK(ih); 4019 error = copyout(pstore, out, 4020 sizeof(struct pfsync_state_1301) * count); 4021 if (error) 4022 break; 4023 out = ps->ps_states + nr; 4024 } 4025 DIOCGETSTATES_full: 4026 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4027 free(pstore, M_TEMP); 4028 4029 break; 4030 } 4031 4032 case DIOCGETSTATESV2: { 4033 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr; 4034 struct pf_kstate *s; 4035 struct pf_state_export *pstore, *p; 4036 int i, nr; 4037 size_t slice_count = 16, count; 4038 void *out; 4039 4040 if (ps->ps_req_version > PF_STATE_VERSION) { 4041 error = ENOTSUP; 4042 break; 4043 } 4044 4045 if (ps->ps_len <= 0) { 4046 nr = uma_zone_get_cur(V_pf_state_z); 4047 ps->ps_len = sizeof(struct pf_state_export) * nr; 4048 break; 4049 } 4050 4051 out = ps->ps_states; 4052 pstore = mallocarray(slice_count, 4053 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO); 4054 nr = 0; 4055 4056 for (i = 0; i <= V_pf_hashmask; i++) { 4057 struct pf_idhash *ih = &V_pf_idhash[i]; 4058 4059 DIOCGETSTATESV2_retry: 4060 p = pstore; 4061 4062 if (LIST_EMPTY(&ih->states)) 4063 continue; 4064 4065 PF_HASHROW_LOCK(ih); 4066 count = 0; 4067 LIST_FOREACH(s, &ih->states, entry) { 4068 if (s->timeout == PFTM_UNLINKED) 4069 continue; 4070 count++; 4071 } 4072 4073 if (count > slice_count) { 4074 PF_HASHROW_UNLOCK(ih); 4075 free(pstore, M_TEMP); 4076 slice_count = count * 2; 4077 pstore = mallocarray(slice_count, 4078 sizeof(struct pf_state_export), M_TEMP, 4079 M_WAITOK | M_ZERO); 4080 goto DIOCGETSTATESV2_retry; 4081 } 4082 4083 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4084 PF_HASHROW_UNLOCK(ih); 4085 goto DIOCGETSTATESV2_full; 4086 } 4087 4088 LIST_FOREACH(s, &ih->states, entry) { 4089 if (s->timeout == PFTM_UNLINKED) 4090 continue; 4091 4092 pf_state_export(p, s); 4093 p++; 4094 nr++; 4095 } 4096 PF_HASHROW_UNLOCK(ih); 4097 error = copyout(pstore, out, 4098 sizeof(struct pf_state_export) * count); 4099 if (error) 4100 break; 4101 out = ps->ps_states + nr; 4102 } 4103 DIOCGETSTATESV2_full: 4104 ps->ps_len = nr * sizeof(struct pf_state_export); 4105 free(pstore, M_TEMP); 4106 4107 break; 4108 } 4109 #endif 4110 case DIOCGETSTATUSNV: { 4111 error = pf_getstatus((struct pfioc_nv *)addr); 4112 break; 4113 } 4114 4115 case DIOCSETSTATUSIF: { 4116 struct pfioc_if *pi = (struct pfioc_if *)addr; 4117 4118 if (pi->ifname[0] == 0) { 4119 bzero(V_pf_status.ifname, IFNAMSIZ); 4120 break; 4121 } 4122 PF_RULES_WLOCK(); 4123 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); 4124 PF_RULES_WUNLOCK(); 4125 break; 4126 } 4127 4128 case DIOCCLRSTATUS: { 4129 pf_ioctl_clear_status(); 4130 break; 4131 } 4132 4133 case DIOCNATLOOK: { 4134 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; 4135 struct pf_state_key *sk; 4136 struct pf_kstate *state; 4137 struct pf_state_key_cmp key; 4138 int m = 0, direction = pnl->direction; 4139 int sidx, didx; 4140 4141 /* NATLOOK src and dst are reversed, so reverse sidx/didx */ 4142 sidx = (direction == PF_IN) ? 1 : 0; 4143 didx = (direction == PF_IN) ? 0 : 1; 4144 4145 if (!pnl->proto || 4146 PF_AZERO(&pnl->saddr, pnl->af) || 4147 PF_AZERO(&pnl->daddr, pnl->af) || 4148 ((pnl->proto == IPPROTO_TCP || 4149 pnl->proto == IPPROTO_UDP) && 4150 (!pnl->dport || !pnl->sport))) 4151 error = EINVAL; 4152 else { 4153 bzero(&key, sizeof(key)); 4154 key.af = pnl->af; 4155 key.proto = pnl->proto; 4156 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af); 4157 key.port[sidx] = pnl->sport; 4158 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af); 4159 key.port[didx] = pnl->dport; 4160 4161 state = pf_find_state_all(&key, direction, &m); 4162 if (state == NULL) { 4163 error = ENOENT; 4164 } else { 4165 if (m > 1) { 4166 PF_STATE_UNLOCK(state); 4167 error = E2BIG; /* more than one state */ 4168 } else { 4169 sk = state->key[sidx]; 4170 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af); 4171 pnl->rsport = sk->port[sidx]; 4172 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af); 4173 pnl->rdport = sk->port[didx]; 4174 PF_STATE_UNLOCK(state); 4175 } 4176 } 4177 } 4178 break; 4179 } 4180 4181 case DIOCSETTIMEOUT: { 4182 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4183 4184 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds, 4185 &pt->seconds); 4186 break; 4187 } 4188 4189 case DIOCGETTIMEOUT: { 4190 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4191 4192 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds); 4193 break; 4194 } 4195 4196 case DIOCGETLIMIT: { 4197 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4198 4199 error = pf_ioctl_get_limit(pl->index, &pl->limit); 4200 break; 4201 } 4202 4203 case DIOCSETLIMIT: { 4204 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4205 unsigned int old_limit; 4206 4207 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit); 4208 pl->limit = old_limit; 4209 break; 4210 } 4211 4212 case DIOCSETDEBUG: { 4213 u_int32_t *level = (u_int32_t *)addr; 4214 4215 PF_RULES_WLOCK(); 4216 V_pf_status.debug = *level; 4217 PF_RULES_WUNLOCK(); 4218 break; 4219 } 4220 4221 case DIOCCLRRULECTRS: { 4222 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ 4223 struct pf_kruleset *ruleset = &pf_main_ruleset; 4224 struct pf_krule *rule; 4225 4226 PF_RULES_WLOCK(); 4227 TAILQ_FOREACH(rule, 4228 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { 4229 pf_counter_u64_zero(&rule->evaluations); 4230 for (int i = 0; i < 2; i++) { 4231 pf_counter_u64_zero(&rule->packets[i]); 4232 pf_counter_u64_zero(&rule->bytes[i]); 4233 } 4234 } 4235 PF_RULES_WUNLOCK(); 4236 break; 4237 } 4238 4239 case DIOCGIFSPEEDV0: 4240 case DIOCGIFSPEEDV1: { 4241 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr; 4242 struct pf_ifspeed_v1 ps; 4243 struct ifnet *ifp; 4244 4245 if (psp->ifname[0] == '\0') { 4246 error = EINVAL; 4247 break; 4248 } 4249 4250 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); 4251 if (error != 0) 4252 break; 4253 ifp = ifunit(ps.ifname); 4254 if (ifp != NULL) { 4255 psp->baudrate32 = 4256 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX); 4257 if (cmd == DIOCGIFSPEEDV1) 4258 psp->baudrate = ifp->if_baudrate; 4259 } else { 4260 error = EINVAL; 4261 } 4262 break; 4263 } 4264 4265 #ifdef ALTQ 4266 case DIOCSTARTALTQ: { 4267 struct pf_altq *altq; 4268 4269 PF_RULES_WLOCK(); 4270 /* enable all altq interfaces on active list */ 4271 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4272 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4273 error = pf_enable_altq(altq); 4274 if (error != 0) 4275 break; 4276 } 4277 } 4278 if (error == 0) 4279 V_pf_altq_running = 1; 4280 PF_RULES_WUNLOCK(); 4281 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n")); 4282 break; 4283 } 4284 4285 case DIOCSTOPALTQ: { 4286 struct pf_altq *altq; 4287 4288 PF_RULES_WLOCK(); 4289 /* disable all altq interfaces on active list */ 4290 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4291 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4292 error = pf_disable_altq(altq); 4293 if (error != 0) 4294 break; 4295 } 4296 } 4297 if (error == 0) 4298 V_pf_altq_running = 0; 4299 PF_RULES_WUNLOCK(); 4300 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n")); 4301 break; 4302 } 4303 4304 case DIOCADDALTQV0: 4305 case DIOCADDALTQV1: { 4306 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4307 struct pf_altq *altq, *a; 4308 struct ifnet *ifp; 4309 4310 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); 4311 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); 4312 if (error) 4313 break; 4314 altq->local_flags = 0; 4315 4316 PF_RULES_WLOCK(); 4317 if (pa->ticket != V_ticket_altqs_inactive) { 4318 PF_RULES_WUNLOCK(); 4319 free(altq, M_PFALTQ); 4320 error = EBUSY; 4321 break; 4322 } 4323 4324 /* 4325 * if this is for a queue, find the discipline and 4326 * copy the necessary fields 4327 */ 4328 if (altq->qname[0] != 0) { 4329 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) { 4330 PF_RULES_WUNLOCK(); 4331 error = EBUSY; 4332 free(altq, M_PFALTQ); 4333 break; 4334 } 4335 altq->altq_disc = NULL; 4336 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { 4337 if (strncmp(a->ifname, altq->ifname, 4338 IFNAMSIZ) == 0) { 4339 altq->altq_disc = a->altq_disc; 4340 break; 4341 } 4342 } 4343 } 4344 4345 if ((ifp = ifunit(altq->ifname)) == NULL) 4346 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 4347 else 4348 error = altq_add(ifp, altq); 4349 4350 if (error) { 4351 PF_RULES_WUNLOCK(); 4352 free(altq, M_PFALTQ); 4353 break; 4354 } 4355 4356 if (altq->qname[0] != 0) 4357 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); 4358 else 4359 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries); 4360 /* version error check done on import above */ 4361 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4362 PF_RULES_WUNLOCK(); 4363 break; 4364 } 4365 4366 case DIOCGETALTQSV0: 4367 case DIOCGETALTQSV1: { 4368 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4369 struct pf_altq *altq; 4370 4371 PF_RULES_RLOCK(); 4372 pa->nr = 0; 4373 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) 4374 pa->nr++; 4375 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) 4376 pa->nr++; 4377 pa->ticket = V_ticket_altqs_active; 4378 PF_RULES_RUNLOCK(); 4379 break; 4380 } 4381 4382 case DIOCGETALTQV0: 4383 case DIOCGETALTQV1: { 4384 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4385 struct pf_altq *altq; 4386 4387 PF_RULES_RLOCK(); 4388 if (pa->ticket != V_ticket_altqs_active) { 4389 PF_RULES_RUNLOCK(); 4390 error = EBUSY; 4391 break; 4392 } 4393 altq = pf_altq_get_nth_active(pa->nr); 4394 if (altq == NULL) { 4395 PF_RULES_RUNLOCK(); 4396 error = EBUSY; 4397 break; 4398 } 4399 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4400 PF_RULES_RUNLOCK(); 4401 break; 4402 } 4403 4404 case DIOCCHANGEALTQV0: 4405 case DIOCCHANGEALTQV1: 4406 /* CHANGEALTQ not supported yet! */ 4407 error = ENODEV; 4408 break; 4409 4410 case DIOCGETQSTATSV0: 4411 case DIOCGETQSTATSV1: { 4412 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr; 4413 struct pf_altq *altq; 4414 int nbytes; 4415 u_int32_t version; 4416 4417 PF_RULES_RLOCK(); 4418 if (pq->ticket != V_ticket_altqs_active) { 4419 PF_RULES_RUNLOCK(); 4420 error = EBUSY; 4421 break; 4422 } 4423 nbytes = pq->nbytes; 4424 altq = pf_altq_get_nth_active(pq->nr); 4425 if (altq == NULL) { 4426 PF_RULES_RUNLOCK(); 4427 error = EBUSY; 4428 break; 4429 } 4430 4431 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { 4432 PF_RULES_RUNLOCK(); 4433 error = ENXIO; 4434 break; 4435 } 4436 PF_RULES_RUNLOCK(); 4437 if (cmd == DIOCGETQSTATSV0) 4438 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */ 4439 else 4440 version = pq->version; 4441 error = altq_getqstats(altq, pq->buf, &nbytes, version); 4442 if (error == 0) { 4443 pq->scheduler = altq->scheduler; 4444 pq->nbytes = nbytes; 4445 } 4446 break; 4447 } 4448 #endif /* ALTQ */ 4449 4450 case DIOCBEGINADDRS: { 4451 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4452 4453 error = pf_ioctl_begin_addrs(&pp->ticket); 4454 break; 4455 } 4456 4457 case DIOCADDADDR: { 4458 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4459 struct pf_nl_pooladdr npp = {}; 4460 4461 npp.which = PF_RDR; 4462 memcpy(&npp, pp, sizeof(*pp)); 4463 error = pf_ioctl_add_addr(&npp); 4464 break; 4465 } 4466 4467 case DIOCGETADDRS: { 4468 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4469 struct pf_nl_pooladdr npp = {}; 4470 4471 npp.which = PF_RDR; 4472 memcpy(&npp, pp, sizeof(*pp)); 4473 error = pf_ioctl_get_addrs(&npp); 4474 memcpy(pp, &npp, sizeof(*pp)); 4475 4476 break; 4477 } 4478 4479 case DIOCGETADDR: { 4480 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4481 struct pf_nl_pooladdr npp = {}; 4482 4483 npp.which = PF_RDR; 4484 memcpy(&npp, pp, sizeof(*pp)); 4485 error = pf_ioctl_get_addr(&npp); 4486 memcpy(pp, &npp, sizeof(*pp)); 4487 4488 break; 4489 } 4490 4491 case DIOCCHANGEADDR: { 4492 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; 4493 struct pf_kpool *pool; 4494 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL; 4495 struct pf_kruleset *ruleset; 4496 struct pfi_kkif *kif = NULL; 4497 4498 pca->anchor[sizeof(pca->anchor) - 1] = 0; 4499 4500 if (pca->action < PF_CHANGE_ADD_HEAD || 4501 pca->action > PF_CHANGE_REMOVE) { 4502 error = EINVAL; 4503 break; 4504 } 4505 if (pca->addr.addr.type != PF_ADDR_ADDRMASK && 4506 pca->addr.addr.type != PF_ADDR_DYNIFTL && 4507 pca->addr.addr.type != PF_ADDR_TABLE) { 4508 error = EINVAL; 4509 break; 4510 } 4511 if (pca->addr.addr.p.dyn != NULL) { 4512 error = EINVAL; 4513 break; 4514 } 4515 4516 if (pca->action != PF_CHANGE_REMOVE) { 4517 #ifndef INET 4518 if (pca->af == AF_INET) { 4519 error = EAFNOSUPPORT; 4520 break; 4521 } 4522 #endif /* INET */ 4523 #ifndef INET6 4524 if (pca->af == AF_INET6) { 4525 error = EAFNOSUPPORT; 4526 break; 4527 } 4528 #endif /* INET6 */ 4529 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); 4530 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); 4531 if (newpa->ifname[0]) 4532 kif = pf_kkif_create(M_WAITOK); 4533 newpa->kif = NULL; 4534 } 4535 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x) 4536 PF_RULES_WLOCK(); 4537 ruleset = pf_find_kruleset(pca->anchor); 4538 if (ruleset == NULL) 4539 ERROUT(EBUSY); 4540 4541 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action, 4542 pca->r_num, pca->r_last, 1, 1, PF_RDR); 4543 if (pool == NULL) 4544 ERROUT(EBUSY); 4545 4546 if (pca->action != PF_CHANGE_REMOVE) { 4547 if (newpa->ifname[0]) { 4548 newpa->kif = pfi_kkif_attach(kif, newpa->ifname); 4549 pfi_kkif_ref(newpa->kif); 4550 kif = NULL; 4551 } 4552 4553 switch (newpa->addr.type) { 4554 case PF_ADDR_DYNIFTL: 4555 error = pfi_dynaddr_setup(&newpa->addr, 4556 pca->af); 4557 break; 4558 case PF_ADDR_TABLE: 4559 newpa->addr.p.tbl = pfr_attach_table(ruleset, 4560 newpa->addr.v.tblname); 4561 if (newpa->addr.p.tbl == NULL) 4562 error = ENOMEM; 4563 break; 4564 } 4565 if (error) 4566 goto DIOCCHANGEADDR_error; 4567 } 4568 4569 switch (pca->action) { 4570 case PF_CHANGE_ADD_HEAD: 4571 oldpa = TAILQ_FIRST(&pool->list); 4572 break; 4573 case PF_CHANGE_ADD_TAIL: 4574 oldpa = TAILQ_LAST(&pool->list, pf_kpalist); 4575 break; 4576 default: 4577 oldpa = TAILQ_FIRST(&pool->list); 4578 for (int i = 0; oldpa && i < pca->nr; i++) 4579 oldpa = TAILQ_NEXT(oldpa, entries); 4580 4581 if (oldpa == NULL) 4582 ERROUT(EINVAL); 4583 } 4584 4585 if (pca->action == PF_CHANGE_REMOVE) { 4586 TAILQ_REMOVE(&pool->list, oldpa, entries); 4587 switch (oldpa->addr.type) { 4588 case PF_ADDR_DYNIFTL: 4589 pfi_dynaddr_remove(oldpa->addr.p.dyn); 4590 break; 4591 case PF_ADDR_TABLE: 4592 pfr_detach_table(oldpa->addr.p.tbl); 4593 break; 4594 } 4595 if (oldpa->kif) 4596 pfi_kkif_unref(oldpa->kif); 4597 free(oldpa, M_PFRULE); 4598 } else { 4599 if (oldpa == NULL) 4600 TAILQ_INSERT_TAIL(&pool->list, newpa, entries); 4601 else if (pca->action == PF_CHANGE_ADD_HEAD || 4602 pca->action == PF_CHANGE_ADD_BEFORE) 4603 TAILQ_INSERT_BEFORE(oldpa, newpa, entries); 4604 else 4605 TAILQ_INSERT_AFTER(&pool->list, oldpa, 4606 newpa, entries); 4607 } 4608 4609 pool->cur = TAILQ_FIRST(&pool->list); 4610 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); 4611 PF_RULES_WUNLOCK(); 4612 break; 4613 4614 #undef ERROUT 4615 DIOCCHANGEADDR_error: 4616 if (newpa != NULL) { 4617 if (newpa->kif) 4618 pfi_kkif_unref(newpa->kif); 4619 free(newpa, M_PFRULE); 4620 } 4621 PF_RULES_WUNLOCK(); 4622 pf_kkif_free(kif); 4623 break; 4624 } 4625 4626 case DIOCGETRULESETS: { 4627 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4628 4629 pr->path[sizeof(pr->path) - 1] = 0; 4630 4631 error = pf_ioctl_get_rulesets(pr); 4632 break; 4633 } 4634 4635 case DIOCGETRULESET: { 4636 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4637 4638 pr->path[sizeof(pr->path) - 1] = 0; 4639 4640 error = pf_ioctl_get_ruleset(pr); 4641 break; 4642 } 4643 4644 case DIOCRCLRTABLES: { 4645 struct pfioc_table *io = (struct pfioc_table *)addr; 4646 4647 if (io->pfrio_esize != 0) { 4648 error = ENODEV; 4649 break; 4650 } 4651 PF_RULES_WLOCK(); 4652 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, 4653 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4654 PF_RULES_WUNLOCK(); 4655 break; 4656 } 4657 4658 case DIOCRADDTABLES: { 4659 struct pfioc_table *io = (struct pfioc_table *)addr; 4660 struct pfr_table *pfrts; 4661 size_t totlen; 4662 4663 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4664 error = ENODEV; 4665 break; 4666 } 4667 4668 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4669 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4670 error = ENOMEM; 4671 break; 4672 } 4673 4674 totlen = io->pfrio_size * sizeof(struct pfr_table); 4675 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4676 M_TEMP, M_WAITOK); 4677 error = copyin(io->pfrio_buffer, pfrts, totlen); 4678 if (error) { 4679 free(pfrts, M_TEMP); 4680 break; 4681 } 4682 PF_RULES_WLOCK(); 4683 error = pfr_add_tables(pfrts, io->pfrio_size, 4684 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4685 PF_RULES_WUNLOCK(); 4686 free(pfrts, M_TEMP); 4687 break; 4688 } 4689 4690 case DIOCRDELTABLES: { 4691 struct pfioc_table *io = (struct pfioc_table *)addr; 4692 struct pfr_table *pfrts; 4693 size_t totlen; 4694 4695 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4696 error = ENODEV; 4697 break; 4698 } 4699 4700 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4701 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4702 error = ENOMEM; 4703 break; 4704 } 4705 4706 totlen = io->pfrio_size * sizeof(struct pfr_table); 4707 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4708 M_TEMP, M_WAITOK); 4709 error = copyin(io->pfrio_buffer, pfrts, totlen); 4710 if (error) { 4711 free(pfrts, M_TEMP); 4712 break; 4713 } 4714 PF_RULES_WLOCK(); 4715 error = pfr_del_tables(pfrts, io->pfrio_size, 4716 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4717 PF_RULES_WUNLOCK(); 4718 free(pfrts, M_TEMP); 4719 break; 4720 } 4721 4722 case DIOCRGETTABLES: { 4723 struct pfioc_table *io = (struct pfioc_table *)addr; 4724 struct pfr_table *pfrts; 4725 size_t totlen; 4726 int n; 4727 4728 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4729 error = ENODEV; 4730 break; 4731 } 4732 PF_RULES_RLOCK(); 4733 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4734 if (n < 0) { 4735 PF_RULES_RUNLOCK(); 4736 error = EINVAL; 4737 break; 4738 } 4739 io->pfrio_size = min(io->pfrio_size, n); 4740 4741 totlen = io->pfrio_size * sizeof(struct pfr_table); 4742 4743 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4744 M_TEMP, M_NOWAIT | M_ZERO); 4745 if (pfrts == NULL) { 4746 error = ENOMEM; 4747 PF_RULES_RUNLOCK(); 4748 break; 4749 } 4750 error = pfr_get_tables(&io->pfrio_table, pfrts, 4751 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4752 PF_RULES_RUNLOCK(); 4753 if (error == 0) 4754 error = copyout(pfrts, io->pfrio_buffer, totlen); 4755 free(pfrts, M_TEMP); 4756 break; 4757 } 4758 4759 case DIOCRGETTSTATS: { 4760 struct pfioc_table *io = (struct pfioc_table *)addr; 4761 struct pfr_tstats *pfrtstats; 4762 size_t totlen; 4763 int n; 4764 4765 if (io->pfrio_esize != sizeof(struct pfr_tstats)) { 4766 error = ENODEV; 4767 break; 4768 } 4769 PF_TABLE_STATS_LOCK(); 4770 PF_RULES_RLOCK(); 4771 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4772 if (n < 0) { 4773 PF_RULES_RUNLOCK(); 4774 PF_TABLE_STATS_UNLOCK(); 4775 error = EINVAL; 4776 break; 4777 } 4778 io->pfrio_size = min(io->pfrio_size, n); 4779 4780 totlen = io->pfrio_size * sizeof(struct pfr_tstats); 4781 pfrtstats = mallocarray(io->pfrio_size, 4782 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO); 4783 if (pfrtstats == NULL) { 4784 error = ENOMEM; 4785 PF_RULES_RUNLOCK(); 4786 PF_TABLE_STATS_UNLOCK(); 4787 break; 4788 } 4789 error = pfr_get_tstats(&io->pfrio_table, pfrtstats, 4790 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4791 PF_RULES_RUNLOCK(); 4792 PF_TABLE_STATS_UNLOCK(); 4793 if (error == 0) 4794 error = copyout(pfrtstats, io->pfrio_buffer, totlen); 4795 free(pfrtstats, M_TEMP); 4796 break; 4797 } 4798 4799 case DIOCRCLRTSTATS: { 4800 struct pfioc_table *io = (struct pfioc_table *)addr; 4801 struct pfr_table *pfrts; 4802 size_t totlen; 4803 4804 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4805 error = ENODEV; 4806 break; 4807 } 4808 4809 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4810 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4811 /* We used to count tables and use the minimum required 4812 * size, so we didn't fail on overly large requests. 4813 * Keep doing so. */ 4814 io->pfrio_size = pf_ioctl_maxcount; 4815 break; 4816 } 4817 4818 totlen = io->pfrio_size * sizeof(struct pfr_table); 4819 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4820 M_TEMP, M_WAITOK); 4821 error = copyin(io->pfrio_buffer, pfrts, totlen); 4822 if (error) { 4823 free(pfrts, M_TEMP); 4824 break; 4825 } 4826 4827 PF_TABLE_STATS_LOCK(); 4828 PF_RULES_RLOCK(); 4829 error = pfr_clr_tstats(pfrts, io->pfrio_size, 4830 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4831 PF_RULES_RUNLOCK(); 4832 PF_TABLE_STATS_UNLOCK(); 4833 free(pfrts, M_TEMP); 4834 break; 4835 } 4836 4837 case DIOCRSETTFLAGS: { 4838 struct pfioc_table *io = (struct pfioc_table *)addr; 4839 struct pfr_table *pfrts; 4840 size_t totlen; 4841 int n; 4842 4843 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4844 error = ENODEV; 4845 break; 4846 } 4847 4848 PF_RULES_RLOCK(); 4849 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4850 if (n < 0) { 4851 PF_RULES_RUNLOCK(); 4852 error = EINVAL; 4853 break; 4854 } 4855 4856 io->pfrio_size = min(io->pfrio_size, n); 4857 PF_RULES_RUNLOCK(); 4858 4859 totlen = io->pfrio_size * sizeof(struct pfr_table); 4860 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4861 M_TEMP, M_WAITOK); 4862 error = copyin(io->pfrio_buffer, pfrts, totlen); 4863 if (error) { 4864 free(pfrts, M_TEMP); 4865 break; 4866 } 4867 PF_RULES_WLOCK(); 4868 error = pfr_set_tflags(pfrts, io->pfrio_size, 4869 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, 4870 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4871 PF_RULES_WUNLOCK(); 4872 free(pfrts, M_TEMP); 4873 break; 4874 } 4875 4876 case DIOCRCLRADDRS: { 4877 struct pfioc_table *io = (struct pfioc_table *)addr; 4878 4879 if (io->pfrio_esize != 0) { 4880 error = ENODEV; 4881 break; 4882 } 4883 PF_RULES_WLOCK(); 4884 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, 4885 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4886 PF_RULES_WUNLOCK(); 4887 break; 4888 } 4889 4890 case DIOCRADDADDRS: { 4891 struct pfioc_table *io = (struct pfioc_table *)addr; 4892 struct pfr_addr *pfras; 4893 size_t totlen; 4894 4895 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4896 error = ENODEV; 4897 break; 4898 } 4899 if (io->pfrio_size < 0 || 4900 io->pfrio_size > pf_ioctl_maxcount || 4901 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4902 error = EINVAL; 4903 break; 4904 } 4905 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4906 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4907 M_TEMP, M_WAITOK); 4908 error = copyin(io->pfrio_buffer, pfras, totlen); 4909 if (error) { 4910 free(pfras, M_TEMP); 4911 break; 4912 } 4913 PF_RULES_WLOCK(); 4914 error = pfr_add_addrs(&io->pfrio_table, pfras, 4915 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | 4916 PFR_FLAG_USERIOCTL); 4917 PF_RULES_WUNLOCK(); 4918 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4919 error = copyout(pfras, io->pfrio_buffer, totlen); 4920 free(pfras, M_TEMP); 4921 break; 4922 } 4923 4924 case DIOCRDELADDRS: { 4925 struct pfioc_table *io = (struct pfioc_table *)addr; 4926 struct pfr_addr *pfras; 4927 size_t totlen; 4928 4929 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4930 error = ENODEV; 4931 break; 4932 } 4933 if (io->pfrio_size < 0 || 4934 io->pfrio_size > pf_ioctl_maxcount || 4935 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4936 error = EINVAL; 4937 break; 4938 } 4939 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4940 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4941 M_TEMP, M_WAITOK); 4942 error = copyin(io->pfrio_buffer, pfras, totlen); 4943 if (error) { 4944 free(pfras, M_TEMP); 4945 break; 4946 } 4947 PF_RULES_WLOCK(); 4948 error = pfr_del_addrs(&io->pfrio_table, pfras, 4949 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | 4950 PFR_FLAG_USERIOCTL); 4951 PF_RULES_WUNLOCK(); 4952 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4953 error = copyout(pfras, io->pfrio_buffer, totlen); 4954 free(pfras, M_TEMP); 4955 break; 4956 } 4957 4958 case DIOCRSETADDRS: { 4959 struct pfioc_table *io = (struct pfioc_table *)addr; 4960 struct pfr_addr *pfras; 4961 size_t totlen, count; 4962 4963 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4964 error = ENODEV; 4965 break; 4966 } 4967 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { 4968 error = EINVAL; 4969 break; 4970 } 4971 count = max(io->pfrio_size, io->pfrio_size2); 4972 if (count > pf_ioctl_maxcount || 4973 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { 4974 error = EINVAL; 4975 break; 4976 } 4977 totlen = count * sizeof(struct pfr_addr); 4978 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP, 4979 M_WAITOK); 4980 error = copyin(io->pfrio_buffer, pfras, totlen); 4981 if (error) { 4982 free(pfras, M_TEMP); 4983 break; 4984 } 4985 PF_RULES_WLOCK(); 4986 error = pfr_set_addrs(&io->pfrio_table, pfras, 4987 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, 4988 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | 4989 PFR_FLAG_USERIOCTL, 0); 4990 PF_RULES_WUNLOCK(); 4991 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4992 error = copyout(pfras, io->pfrio_buffer, totlen); 4993 free(pfras, M_TEMP); 4994 break; 4995 } 4996 4997 case DIOCRGETADDRS: { 4998 struct pfioc_table *io = (struct pfioc_table *)addr; 4999 struct pfr_addr *pfras; 5000 size_t totlen; 5001 5002 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5003 error = ENODEV; 5004 break; 5005 } 5006 if (io->pfrio_size < 0 || 5007 io->pfrio_size > pf_ioctl_maxcount || 5008 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5009 error = EINVAL; 5010 break; 5011 } 5012 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5013 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5014 M_TEMP, M_WAITOK | M_ZERO); 5015 PF_RULES_RLOCK(); 5016 error = pfr_get_addrs(&io->pfrio_table, pfras, 5017 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5018 PF_RULES_RUNLOCK(); 5019 if (error == 0) 5020 error = copyout(pfras, io->pfrio_buffer, totlen); 5021 free(pfras, M_TEMP); 5022 break; 5023 } 5024 5025 case DIOCRGETASTATS: { 5026 struct pfioc_table *io = (struct pfioc_table *)addr; 5027 struct pfr_astats *pfrastats; 5028 size_t totlen; 5029 5030 if (io->pfrio_esize != sizeof(struct pfr_astats)) { 5031 error = ENODEV; 5032 break; 5033 } 5034 if (io->pfrio_size < 0 || 5035 io->pfrio_size > pf_ioctl_maxcount || 5036 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { 5037 error = EINVAL; 5038 break; 5039 } 5040 totlen = io->pfrio_size * sizeof(struct pfr_astats); 5041 pfrastats = mallocarray(io->pfrio_size, 5042 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO); 5043 PF_RULES_RLOCK(); 5044 error = pfr_get_astats(&io->pfrio_table, pfrastats, 5045 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5046 PF_RULES_RUNLOCK(); 5047 if (error == 0) 5048 error = copyout(pfrastats, io->pfrio_buffer, totlen); 5049 free(pfrastats, M_TEMP); 5050 break; 5051 } 5052 5053 case DIOCRCLRASTATS: { 5054 struct pfioc_table *io = (struct pfioc_table *)addr; 5055 struct pfr_addr *pfras; 5056 size_t totlen; 5057 5058 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5059 error = ENODEV; 5060 break; 5061 } 5062 if (io->pfrio_size < 0 || 5063 io->pfrio_size > pf_ioctl_maxcount || 5064 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5065 error = EINVAL; 5066 break; 5067 } 5068 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5069 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5070 M_TEMP, M_WAITOK); 5071 error = copyin(io->pfrio_buffer, pfras, totlen); 5072 if (error) { 5073 free(pfras, M_TEMP); 5074 break; 5075 } 5076 PF_RULES_WLOCK(); 5077 error = pfr_clr_astats(&io->pfrio_table, pfras, 5078 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | 5079 PFR_FLAG_USERIOCTL); 5080 PF_RULES_WUNLOCK(); 5081 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5082 error = copyout(pfras, io->pfrio_buffer, totlen); 5083 free(pfras, M_TEMP); 5084 break; 5085 } 5086 5087 case DIOCRTSTADDRS: { 5088 struct pfioc_table *io = (struct pfioc_table *)addr; 5089 struct pfr_addr *pfras; 5090 size_t totlen; 5091 5092 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5093 error = ENODEV; 5094 break; 5095 } 5096 if (io->pfrio_size < 0 || 5097 io->pfrio_size > pf_ioctl_maxcount || 5098 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5099 error = EINVAL; 5100 break; 5101 } 5102 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5103 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5104 M_TEMP, M_WAITOK); 5105 error = copyin(io->pfrio_buffer, pfras, totlen); 5106 if (error) { 5107 free(pfras, M_TEMP); 5108 break; 5109 } 5110 PF_RULES_RLOCK(); 5111 error = pfr_tst_addrs(&io->pfrio_table, pfras, 5112 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | 5113 PFR_FLAG_USERIOCTL); 5114 PF_RULES_RUNLOCK(); 5115 if (error == 0) 5116 error = copyout(pfras, io->pfrio_buffer, totlen); 5117 free(pfras, M_TEMP); 5118 break; 5119 } 5120 5121 case DIOCRINADEFINE: { 5122 struct pfioc_table *io = (struct pfioc_table *)addr; 5123 struct pfr_addr *pfras; 5124 size_t totlen; 5125 5126 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5127 error = ENODEV; 5128 break; 5129 } 5130 if (io->pfrio_size < 0 || 5131 io->pfrio_size > pf_ioctl_maxcount || 5132 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5133 error = EINVAL; 5134 break; 5135 } 5136 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5137 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5138 M_TEMP, M_WAITOK); 5139 error = copyin(io->pfrio_buffer, pfras, totlen); 5140 if (error) { 5141 free(pfras, M_TEMP); 5142 break; 5143 } 5144 PF_RULES_WLOCK(); 5145 error = pfr_ina_define(&io->pfrio_table, pfras, 5146 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, 5147 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5148 PF_RULES_WUNLOCK(); 5149 free(pfras, M_TEMP); 5150 break; 5151 } 5152 5153 case DIOCOSFPADD: { 5154 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5155 PF_RULES_WLOCK(); 5156 error = pf_osfp_add(io); 5157 PF_RULES_WUNLOCK(); 5158 break; 5159 } 5160 5161 case DIOCOSFPGET: { 5162 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5163 PF_RULES_RLOCK(); 5164 error = pf_osfp_get(io); 5165 PF_RULES_RUNLOCK(); 5166 break; 5167 } 5168 5169 case DIOCXBEGIN: { 5170 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5171 struct pfioc_trans_e *ioes, *ioe; 5172 size_t totlen; 5173 int i; 5174 5175 if (io->esize != sizeof(*ioe)) { 5176 error = ENODEV; 5177 break; 5178 } 5179 if (io->size < 0 || 5180 io->size > pf_ioctl_maxcount || 5181 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5182 error = EINVAL; 5183 break; 5184 } 5185 totlen = sizeof(struct pfioc_trans_e) * io->size; 5186 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5187 M_TEMP, M_WAITOK); 5188 error = copyin(io->array, ioes, totlen); 5189 if (error) { 5190 free(ioes, M_TEMP); 5191 break; 5192 } 5193 PF_RULES_WLOCK(); 5194 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5195 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5196 switch (ioe->rs_num) { 5197 case PF_RULESET_ETH: 5198 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) { 5199 PF_RULES_WUNLOCK(); 5200 free(ioes, M_TEMP); 5201 goto fail; 5202 } 5203 break; 5204 #ifdef ALTQ 5205 case PF_RULESET_ALTQ: 5206 if (ioe->anchor[0]) { 5207 PF_RULES_WUNLOCK(); 5208 free(ioes, M_TEMP); 5209 error = EINVAL; 5210 goto fail; 5211 } 5212 if ((error = pf_begin_altq(&ioe->ticket))) { 5213 PF_RULES_WUNLOCK(); 5214 free(ioes, M_TEMP); 5215 goto fail; 5216 } 5217 break; 5218 #endif /* ALTQ */ 5219 case PF_RULESET_TABLE: 5220 { 5221 struct pfr_table table; 5222 5223 bzero(&table, sizeof(table)); 5224 strlcpy(table.pfrt_anchor, ioe->anchor, 5225 sizeof(table.pfrt_anchor)); 5226 if ((error = pfr_ina_begin(&table, 5227 &ioe->ticket, NULL, 0))) { 5228 PF_RULES_WUNLOCK(); 5229 free(ioes, M_TEMP); 5230 goto fail; 5231 } 5232 break; 5233 } 5234 default: 5235 if ((error = pf_begin_rules(&ioe->ticket, 5236 ioe->rs_num, ioe->anchor))) { 5237 PF_RULES_WUNLOCK(); 5238 free(ioes, M_TEMP); 5239 goto fail; 5240 } 5241 break; 5242 } 5243 } 5244 PF_RULES_WUNLOCK(); 5245 error = copyout(ioes, io->array, totlen); 5246 free(ioes, M_TEMP); 5247 break; 5248 } 5249 5250 case DIOCXROLLBACK: { 5251 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5252 struct pfioc_trans_e *ioe, *ioes; 5253 size_t totlen; 5254 int i; 5255 5256 if (io->esize != sizeof(*ioe)) { 5257 error = ENODEV; 5258 break; 5259 } 5260 if (io->size < 0 || 5261 io->size > pf_ioctl_maxcount || 5262 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5263 error = EINVAL; 5264 break; 5265 } 5266 totlen = sizeof(struct pfioc_trans_e) * io->size; 5267 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5268 M_TEMP, M_WAITOK); 5269 error = copyin(io->array, ioes, totlen); 5270 if (error) { 5271 free(ioes, M_TEMP); 5272 break; 5273 } 5274 PF_RULES_WLOCK(); 5275 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5276 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5277 switch (ioe->rs_num) { 5278 case PF_RULESET_ETH: 5279 if ((error = pf_rollback_eth(ioe->ticket, 5280 ioe->anchor))) { 5281 PF_RULES_WUNLOCK(); 5282 free(ioes, M_TEMP); 5283 goto fail; /* really bad */ 5284 } 5285 break; 5286 #ifdef ALTQ 5287 case PF_RULESET_ALTQ: 5288 if (ioe->anchor[0]) { 5289 PF_RULES_WUNLOCK(); 5290 free(ioes, M_TEMP); 5291 error = EINVAL; 5292 goto fail; 5293 } 5294 if ((error = pf_rollback_altq(ioe->ticket))) { 5295 PF_RULES_WUNLOCK(); 5296 free(ioes, M_TEMP); 5297 goto fail; /* really bad */ 5298 } 5299 break; 5300 #endif /* ALTQ */ 5301 case PF_RULESET_TABLE: 5302 { 5303 struct pfr_table table; 5304 5305 bzero(&table, sizeof(table)); 5306 strlcpy(table.pfrt_anchor, ioe->anchor, 5307 sizeof(table.pfrt_anchor)); 5308 if ((error = pfr_ina_rollback(&table, 5309 ioe->ticket, NULL, 0))) { 5310 PF_RULES_WUNLOCK(); 5311 free(ioes, M_TEMP); 5312 goto fail; /* really bad */ 5313 } 5314 break; 5315 } 5316 default: 5317 if ((error = pf_rollback_rules(ioe->ticket, 5318 ioe->rs_num, ioe->anchor))) { 5319 PF_RULES_WUNLOCK(); 5320 free(ioes, M_TEMP); 5321 goto fail; /* really bad */ 5322 } 5323 break; 5324 } 5325 } 5326 PF_RULES_WUNLOCK(); 5327 free(ioes, M_TEMP); 5328 break; 5329 } 5330 5331 case DIOCXCOMMIT: { 5332 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5333 struct pfioc_trans_e *ioe, *ioes; 5334 struct pf_kruleset *rs; 5335 struct pf_keth_ruleset *ers; 5336 size_t totlen; 5337 int i; 5338 5339 if (io->esize != sizeof(*ioe)) { 5340 error = ENODEV; 5341 break; 5342 } 5343 5344 if (io->size < 0 || 5345 io->size > pf_ioctl_maxcount || 5346 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5347 error = EINVAL; 5348 break; 5349 } 5350 5351 totlen = sizeof(struct pfioc_trans_e) * io->size; 5352 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5353 M_TEMP, M_WAITOK); 5354 error = copyin(io->array, ioes, totlen); 5355 if (error) { 5356 free(ioes, M_TEMP); 5357 break; 5358 } 5359 PF_RULES_WLOCK(); 5360 /* First makes sure everything will succeed. */ 5361 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5362 ioe->anchor[sizeof(ioe->anchor) - 1] = 0; 5363 switch (ioe->rs_num) { 5364 case PF_RULESET_ETH: 5365 ers = pf_find_keth_ruleset(ioe->anchor); 5366 if (ers == NULL || ioe->ticket == 0 || 5367 ioe->ticket != ers->inactive.ticket) { 5368 PF_RULES_WUNLOCK(); 5369 free(ioes, M_TEMP); 5370 error = EINVAL; 5371 goto fail; 5372 } 5373 break; 5374 #ifdef ALTQ 5375 case PF_RULESET_ALTQ: 5376 if (ioe->anchor[0]) { 5377 PF_RULES_WUNLOCK(); 5378 free(ioes, M_TEMP); 5379 error = EINVAL; 5380 goto fail; 5381 } 5382 if (!V_altqs_inactive_open || ioe->ticket != 5383 V_ticket_altqs_inactive) { 5384 PF_RULES_WUNLOCK(); 5385 free(ioes, M_TEMP); 5386 error = EBUSY; 5387 goto fail; 5388 } 5389 break; 5390 #endif /* ALTQ */ 5391 case PF_RULESET_TABLE: 5392 rs = pf_find_kruleset(ioe->anchor); 5393 if (rs == NULL || !rs->topen || ioe->ticket != 5394 rs->tticket) { 5395 PF_RULES_WUNLOCK(); 5396 free(ioes, M_TEMP); 5397 error = EBUSY; 5398 goto fail; 5399 } 5400 break; 5401 default: 5402 if (ioe->rs_num < 0 || ioe->rs_num >= 5403 PF_RULESET_MAX) { 5404 PF_RULES_WUNLOCK(); 5405 free(ioes, M_TEMP); 5406 error = EINVAL; 5407 goto fail; 5408 } 5409 rs = pf_find_kruleset(ioe->anchor); 5410 if (rs == NULL || 5411 !rs->rules[ioe->rs_num].inactive.open || 5412 rs->rules[ioe->rs_num].inactive.ticket != 5413 ioe->ticket) { 5414 PF_RULES_WUNLOCK(); 5415 free(ioes, M_TEMP); 5416 error = EBUSY; 5417 goto fail; 5418 } 5419 break; 5420 } 5421 } 5422 /* Now do the commit - no errors should happen here. */ 5423 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5424 switch (ioe->rs_num) { 5425 case PF_RULESET_ETH: 5426 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) { 5427 PF_RULES_WUNLOCK(); 5428 free(ioes, M_TEMP); 5429 goto fail; /* really bad */ 5430 } 5431 break; 5432 #ifdef ALTQ 5433 case PF_RULESET_ALTQ: 5434 if ((error = pf_commit_altq(ioe->ticket))) { 5435 PF_RULES_WUNLOCK(); 5436 free(ioes, M_TEMP); 5437 goto fail; /* really bad */ 5438 } 5439 break; 5440 #endif /* ALTQ */ 5441 case PF_RULESET_TABLE: 5442 { 5443 struct pfr_table table; 5444 5445 bzero(&table, sizeof(table)); 5446 (void)strlcpy(table.pfrt_anchor, ioe->anchor, 5447 sizeof(table.pfrt_anchor)); 5448 if ((error = pfr_ina_commit(&table, 5449 ioe->ticket, NULL, NULL, 0))) { 5450 PF_RULES_WUNLOCK(); 5451 free(ioes, M_TEMP); 5452 goto fail; /* really bad */ 5453 } 5454 break; 5455 } 5456 default: 5457 if ((error = pf_commit_rules(ioe->ticket, 5458 ioe->rs_num, ioe->anchor))) { 5459 PF_RULES_WUNLOCK(); 5460 free(ioes, M_TEMP); 5461 goto fail; /* really bad */ 5462 } 5463 break; 5464 } 5465 } 5466 PF_RULES_WUNLOCK(); 5467 5468 /* Only hook into EtherNet taffic if we've got rules for it. */ 5469 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 5470 hook_pf_eth(); 5471 else 5472 dehook_pf_eth(); 5473 5474 free(ioes, M_TEMP); 5475 break; 5476 } 5477 5478 case DIOCGETSRCNODES: { 5479 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; 5480 struct pf_srchash *sh; 5481 struct pf_ksrc_node *n; 5482 struct pf_src_node *p, *pstore; 5483 uint32_t i, nr = 0; 5484 5485 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5486 i++, sh++) { 5487 PF_HASHROW_LOCK(sh); 5488 LIST_FOREACH(n, &sh->nodes, entry) 5489 nr++; 5490 PF_HASHROW_UNLOCK(sh); 5491 } 5492 5493 psn->psn_len = min(psn->psn_len, 5494 sizeof(struct pf_src_node) * nr); 5495 5496 if (psn->psn_len == 0) { 5497 psn->psn_len = sizeof(struct pf_src_node) * nr; 5498 break; 5499 } 5500 5501 nr = 0; 5502 5503 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO); 5504 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5505 i++, sh++) { 5506 PF_HASHROW_LOCK(sh); 5507 LIST_FOREACH(n, &sh->nodes, entry) { 5508 5509 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) 5510 break; 5511 5512 pf_src_node_copy(n, p); 5513 5514 p++; 5515 nr++; 5516 } 5517 PF_HASHROW_UNLOCK(sh); 5518 } 5519 error = copyout(pstore, psn->psn_src_nodes, 5520 sizeof(struct pf_src_node) * nr); 5521 if (error) { 5522 free(pstore, M_TEMP); 5523 break; 5524 } 5525 psn->psn_len = sizeof(struct pf_src_node) * nr; 5526 free(pstore, M_TEMP); 5527 break; 5528 } 5529 5530 case DIOCCLRSRCNODES: { 5531 pf_kill_srcnodes(NULL); 5532 break; 5533 } 5534 5535 case DIOCKILLSRCNODES: 5536 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); 5537 break; 5538 5539 #ifdef COMPAT_FREEBSD13 5540 case DIOCKEEPCOUNTERS_FREEBSD13: 5541 #endif 5542 case DIOCKEEPCOUNTERS: 5543 error = pf_keepcounters((struct pfioc_nv *)addr); 5544 break; 5545 5546 case DIOCGETSYNCOOKIES: 5547 error = pf_get_syncookies((struct pfioc_nv *)addr); 5548 break; 5549 5550 case DIOCSETSYNCOOKIES: 5551 error = pf_set_syncookies((struct pfioc_nv *)addr); 5552 break; 5553 5554 case DIOCSETHOSTID: { 5555 u_int32_t *hostid = (u_int32_t *)addr; 5556 5557 PF_RULES_WLOCK(); 5558 if (*hostid == 0) 5559 V_pf_status.hostid = arc4random(); 5560 else 5561 V_pf_status.hostid = *hostid; 5562 PF_RULES_WUNLOCK(); 5563 break; 5564 } 5565 5566 case DIOCOSFPFLUSH: 5567 PF_RULES_WLOCK(); 5568 pf_osfp_flush(); 5569 PF_RULES_WUNLOCK(); 5570 break; 5571 5572 case DIOCIGETIFACES: { 5573 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5574 struct pfi_kif *ifstore; 5575 size_t bufsiz; 5576 5577 if (io->pfiio_esize != sizeof(struct pfi_kif)) { 5578 error = ENODEV; 5579 break; 5580 } 5581 5582 if (io->pfiio_size < 0 || 5583 io->pfiio_size > pf_ioctl_maxcount || 5584 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { 5585 error = EINVAL; 5586 break; 5587 } 5588 5589 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5590 5591 bufsiz = io->pfiio_size * sizeof(struct pfi_kif); 5592 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif), 5593 M_TEMP, M_WAITOK | M_ZERO); 5594 5595 PF_RULES_RLOCK(); 5596 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); 5597 PF_RULES_RUNLOCK(); 5598 error = copyout(ifstore, io->pfiio_buffer, bufsiz); 5599 free(ifstore, M_TEMP); 5600 break; 5601 } 5602 5603 case DIOCSETIFFLAG: { 5604 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5605 5606 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5607 5608 PF_RULES_WLOCK(); 5609 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags); 5610 PF_RULES_WUNLOCK(); 5611 break; 5612 } 5613 5614 case DIOCCLRIFFLAG: { 5615 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5616 5617 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5618 5619 PF_RULES_WLOCK(); 5620 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); 5621 PF_RULES_WUNLOCK(); 5622 break; 5623 } 5624 5625 case DIOCSETREASS: { 5626 u_int32_t *reass = (u_int32_t *)addr; 5627 5628 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF); 5629 /* Removal of DF flag without reassembly enabled is not a 5630 * valid combination. Disable reassembly in such case. */ 5631 if (!(V_pf_status.reass & PF_REASS_ENABLED)) 5632 V_pf_status.reass = 0; 5633 break; 5634 } 5635 5636 default: 5637 error = ENODEV; 5638 break; 5639 } 5640 fail: 5641 CURVNET_RESTORE(); 5642 5643 #undef ERROUT_IOCTL 5644 5645 return (error); 5646 } 5647 5648 void 5649 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version) 5650 { 5651 bzero(sp, sizeof(union pfsync_state_union)); 5652 5653 /* copy from state key */ 5654 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5655 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5656 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5657 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5658 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5659 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5660 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5661 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5662 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto; 5663 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af; 5664 5665 /* copy from state */ 5666 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname)); 5667 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr)); 5668 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000)); 5669 sp->pfs_1301.expire = pf_state_expires(st); 5670 if (sp->pfs_1301.expire <= time_uptime) 5671 sp->pfs_1301.expire = htonl(0); 5672 else 5673 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime); 5674 5675 sp->pfs_1301.direction = st->direction; 5676 sp->pfs_1301.log = st->act.log; 5677 sp->pfs_1301.timeout = st->timeout; 5678 5679 switch (msg_version) { 5680 case PFSYNC_MSG_VERSION_1301: 5681 sp->pfs_1301.state_flags = st->state_flags; 5682 break; 5683 case PFSYNC_MSG_VERSION_1400: 5684 sp->pfs_1400.state_flags = htons(st->state_flags); 5685 sp->pfs_1400.qid = htons(st->act.qid); 5686 sp->pfs_1400.pqid = htons(st->act.pqid); 5687 sp->pfs_1400.dnpipe = htons(st->act.dnpipe); 5688 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe); 5689 sp->pfs_1400.rtableid = htonl(st->act.rtableid); 5690 sp->pfs_1400.min_ttl = st->act.min_ttl; 5691 sp->pfs_1400.set_tos = st->act.set_tos; 5692 sp->pfs_1400.max_mss = htons(st->act.max_mss); 5693 sp->pfs_1400.set_prio[0] = st->act.set_prio[0]; 5694 sp->pfs_1400.set_prio[1] = st->act.set_prio[1]; 5695 sp->pfs_1400.rt = st->act.rt; 5696 if (st->act.rt_kif) 5697 strlcpy(sp->pfs_1400.rt_ifname, 5698 st->act.rt_kif->pfik_name, 5699 sizeof(sp->pfs_1400.rt_ifname)); 5700 break; 5701 default: 5702 panic("%s: Unsupported pfsync_msg_version %d", 5703 __func__, msg_version); 5704 } 5705 5706 /* 5707 * XXX Why do we bother pfsyncing source node information if source 5708 * nodes are not synced? Showing users that there is source tracking 5709 * when there is none seems useless. 5710 */ 5711 if (st->sns[PF_SN_LIMIT] != NULL) 5712 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE; 5713 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE]) 5714 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5715 5716 sp->pfs_1301.id = st->id; 5717 sp->pfs_1301.creatorid = st->creatorid; 5718 pf_state_peer_hton(&st->src, &sp->pfs_1301.src); 5719 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst); 5720 5721 if (st->rule == NULL) 5722 sp->pfs_1301.rule = htonl(-1); 5723 else 5724 sp->pfs_1301.rule = htonl(st->rule->nr); 5725 if (st->anchor == NULL) 5726 sp->pfs_1301.anchor = htonl(-1); 5727 else 5728 sp->pfs_1301.anchor = htonl(st->anchor->nr); 5729 if (st->nat_rule == NULL) 5730 sp->pfs_1301.nat_rule = htonl(-1); 5731 else 5732 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr); 5733 5734 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]); 5735 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]); 5736 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]); 5737 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]); 5738 } 5739 5740 void 5741 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st) 5742 { 5743 bzero(sp, sizeof(*sp)); 5744 5745 sp->version = PF_STATE_VERSION; 5746 5747 /* copy from state key */ 5748 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5749 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5750 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5751 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5752 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5753 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5754 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5755 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5756 sp->proto = st->key[PF_SK_WIRE]->proto; 5757 sp->af = st->key[PF_SK_WIRE]->af; 5758 5759 /* copy from state */ 5760 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5761 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name, 5762 sizeof(sp->orig_ifname)); 5763 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr)); 5764 sp->creation = htonl(time_uptime - (st->creation / 1000)); 5765 sp->expire = pf_state_expires(st); 5766 if (sp->expire <= time_uptime) 5767 sp->expire = htonl(0); 5768 else 5769 sp->expire = htonl(sp->expire - time_uptime); 5770 5771 sp->direction = st->direction; 5772 sp->log = st->act.log; 5773 sp->timeout = st->timeout; 5774 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */ 5775 sp->state_flags_compat = st->state_flags; 5776 sp->state_flags = htons(st->state_flags); 5777 if (st->sns[PF_SN_LIMIT] != NULL) 5778 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5779 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL) 5780 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5781 sp->id = st->id; 5782 sp->creatorid = st->creatorid; 5783 pf_state_peer_hton(&st->src, &sp->src); 5784 pf_state_peer_hton(&st->dst, &sp->dst); 5785 5786 if (st->rule == NULL) 5787 sp->rule = htonl(-1); 5788 else 5789 sp->rule = htonl(st->rule->nr); 5790 if (st->anchor == NULL) 5791 sp->anchor = htonl(-1); 5792 else 5793 sp->anchor = htonl(st->anchor->nr); 5794 if (st->nat_rule == NULL) 5795 sp->nat_rule = htonl(-1); 5796 else 5797 sp->nat_rule = htonl(st->nat_rule->nr); 5798 5799 sp->packets[0] = st->packets[0]; 5800 sp->packets[1] = st->packets[1]; 5801 sp->bytes[0] = st->bytes[0]; 5802 sp->bytes[1] = st->bytes[1]; 5803 5804 sp->qid = htons(st->act.qid); 5805 sp->pqid = htons(st->act.pqid); 5806 sp->dnpipe = htons(st->act.dnpipe); 5807 sp->dnrpipe = htons(st->act.dnrpipe); 5808 sp->rtableid = htonl(st->act.rtableid); 5809 sp->min_ttl = st->act.min_ttl; 5810 sp->set_tos = st->act.set_tos; 5811 sp->max_mss = htons(st->act.max_mss); 5812 sp->rt = st->act.rt; 5813 if (st->act.rt_kif) 5814 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name, 5815 sizeof(sp->rt_ifname)); 5816 sp->set_prio[0] = st->act.set_prio[0]; 5817 sp->set_prio[1] = st->act.set_prio[1]; 5818 5819 } 5820 5821 static void 5822 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 5823 { 5824 struct pfr_ktable *kt; 5825 5826 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); 5827 5828 kt = aw->p.tbl; 5829 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 5830 kt = kt->pfrkt_root; 5831 aw->p.tbl = NULL; 5832 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 5833 kt->pfrkt_cnt : -1; 5834 } 5835 5836 static int 5837 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters, 5838 size_t number, char **names) 5839 { 5840 nvlist_t *nvc; 5841 5842 nvc = nvlist_create(0); 5843 if (nvc == NULL) 5844 return (ENOMEM); 5845 5846 for (int i = 0; i < number; i++) { 5847 nvlist_append_number_array(nvc, "counters", 5848 counter_u64_fetch(counters[i])); 5849 nvlist_append_string_array(nvc, "names", 5850 names[i]); 5851 nvlist_append_number_array(nvc, "ids", 5852 i); 5853 } 5854 nvlist_add_nvlist(nvl, name, nvc); 5855 nvlist_destroy(nvc); 5856 5857 return (0); 5858 } 5859 5860 static int 5861 pf_getstatus(struct pfioc_nv *nv) 5862 { 5863 nvlist_t *nvl = NULL, *nvc = NULL; 5864 void *nvlpacked = NULL; 5865 int error; 5866 struct pf_status s; 5867 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES; 5868 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES; 5869 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES; 5870 PF_RULES_RLOCK_TRACKER; 5871 5872 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 5873 5874 PF_RULES_RLOCK(); 5875 5876 nvl = nvlist_create(0); 5877 if (nvl == NULL) 5878 ERROUT(ENOMEM); 5879 5880 nvlist_add_bool(nvl, "running", V_pf_status.running); 5881 nvlist_add_number(nvl, "since", V_pf_status.since); 5882 nvlist_add_number(nvl, "debug", V_pf_status.debug); 5883 nvlist_add_number(nvl, "hostid", V_pf_status.hostid); 5884 nvlist_add_number(nvl, "states", V_pf_status.states); 5885 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes); 5886 nvlist_add_number(nvl, "reass", V_pf_status.reass); 5887 nvlist_add_bool(nvl, "syncookies_active", 5888 V_pf_status.syncookies_active); 5889 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen); 5890 5891 /* counters */ 5892 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters, 5893 PFRES_MAX, pf_reasons); 5894 if (error != 0) 5895 ERROUT(error); 5896 5897 /* lcounters */ 5898 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters, 5899 KLCNT_MAX, pf_lcounter); 5900 if (error != 0) 5901 ERROUT(error); 5902 5903 /* fcounters */ 5904 nvc = nvlist_create(0); 5905 if (nvc == NULL) 5906 ERROUT(ENOMEM); 5907 5908 for (int i = 0; i < FCNT_MAX; i++) { 5909 nvlist_append_number_array(nvc, "counters", 5910 pf_counter_u64_fetch(&V_pf_status.fcounters[i])); 5911 nvlist_append_string_array(nvc, "names", 5912 pf_fcounter[i]); 5913 nvlist_append_number_array(nvc, "ids", 5914 i); 5915 } 5916 nvlist_add_nvlist(nvl, "fcounters", nvc); 5917 nvlist_destroy(nvc); 5918 nvc = NULL; 5919 5920 /* scounters */ 5921 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters, 5922 SCNT_MAX, pf_fcounter); 5923 if (error != 0) 5924 ERROUT(error); 5925 5926 nvlist_add_string(nvl, "ifname", V_pf_status.ifname); 5927 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum, 5928 PF_MD5_DIGEST_LENGTH); 5929 5930 pfi_update_status(V_pf_status.ifname, &s); 5931 5932 /* pcounters / bcounters */ 5933 for (int i = 0; i < 2; i++) { 5934 for (int j = 0; j < 2; j++) { 5935 for (int k = 0; k < 2; k++) { 5936 nvlist_append_number_array(nvl, "pcounters", 5937 s.pcounters[i][j][k]); 5938 } 5939 nvlist_append_number_array(nvl, "bcounters", 5940 s.bcounters[i][j]); 5941 } 5942 } 5943 5944 nvlpacked = nvlist_pack(nvl, &nv->len); 5945 if (nvlpacked == NULL) 5946 ERROUT(ENOMEM); 5947 5948 if (nv->size == 0) 5949 ERROUT(0); 5950 else if (nv->size < nv->len) 5951 ERROUT(ENOSPC); 5952 5953 PF_RULES_RUNLOCK(); 5954 error = copyout(nvlpacked, nv->data, nv->len); 5955 goto done; 5956 5957 #undef ERROUT 5958 errout: 5959 PF_RULES_RUNLOCK(); 5960 done: 5961 free(nvlpacked, M_NVLIST); 5962 nvlist_destroy(nvc); 5963 nvlist_destroy(nvl); 5964 5965 return (error); 5966 } 5967 5968 /* 5969 * XXX - Check for version mismatch!!! 5970 */ 5971 static void 5972 pf_clear_all_states(void) 5973 { 5974 struct epoch_tracker et; 5975 struct pf_kstate *s; 5976 u_int i; 5977 5978 NET_EPOCH_ENTER(et); 5979 for (i = 0; i <= V_pf_hashmask; i++) { 5980 struct pf_idhash *ih = &V_pf_idhash[i]; 5981 relock: 5982 PF_HASHROW_LOCK(ih); 5983 LIST_FOREACH(s, &ih->states, entry) { 5984 s->timeout = PFTM_PURGE; 5985 /* Don't send out individual delete messages. */ 5986 s->state_flags |= PFSTATE_NOSYNC; 5987 pf_remove_state(s); 5988 goto relock; 5989 } 5990 PF_HASHROW_UNLOCK(ih); 5991 } 5992 NET_EPOCH_EXIT(et); 5993 } 5994 5995 static int 5996 pf_clear_tables(void) 5997 { 5998 struct pfioc_table io; 5999 int error; 6000 6001 bzero(&io, sizeof(io)); 6002 io.pfrio_flags |= PFR_FLAG_ALLRSETS; 6003 6004 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, 6005 io.pfrio_flags); 6006 6007 return (error); 6008 } 6009 6010 static void 6011 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) 6012 { 6013 struct pf_ksrc_node_list kill; 6014 u_int killed; 6015 6016 LIST_INIT(&kill); 6017 for (int i = 0; i <= V_pf_srchashmask; i++) { 6018 struct pf_srchash *sh = &V_pf_srchash[i]; 6019 struct pf_ksrc_node *sn, *tmp; 6020 6021 PF_HASHROW_LOCK(sh); 6022 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) 6023 if (psnk == NULL || 6024 (PF_MATCHA(psnk->psnk_src.neg, 6025 &psnk->psnk_src.addr.v.a.addr, 6026 &psnk->psnk_src.addr.v.a.mask, 6027 &sn->addr, sn->af) && 6028 PF_MATCHA(psnk->psnk_dst.neg, 6029 &psnk->psnk_dst.addr.v.a.addr, 6030 &psnk->psnk_dst.addr.v.a.mask, 6031 &sn->raddr, sn->af))) { 6032 pf_unlink_src_node(sn); 6033 LIST_INSERT_HEAD(&kill, sn, entry); 6034 sn->expire = 1; 6035 } 6036 PF_HASHROW_UNLOCK(sh); 6037 } 6038 6039 for (int i = 0; i <= V_pf_hashmask; i++) { 6040 struct pf_idhash *ih = &V_pf_idhash[i]; 6041 struct pf_kstate *s; 6042 6043 PF_HASHROW_LOCK(ih); 6044 LIST_FOREACH(s, &ih->states, entry) { 6045 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; 6046 sn_type++) { 6047 if (s->sns[sn_type] && 6048 s->sns[sn_type]->expire == 1) { 6049 s->sns[sn_type] = NULL; 6050 } 6051 } 6052 } 6053 PF_HASHROW_UNLOCK(ih); 6054 } 6055 6056 killed = pf_free_src_nodes(&kill); 6057 6058 if (psnk != NULL) 6059 psnk->psnk_killed = killed; 6060 } 6061 6062 static int 6063 pf_keepcounters(struct pfioc_nv *nv) 6064 { 6065 nvlist_t *nvl = NULL; 6066 void *nvlpacked = NULL; 6067 int error = 0; 6068 6069 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6070 6071 if (nv->len > pf_ioctl_maxcount) 6072 ERROUT(ENOMEM); 6073 6074 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6075 error = copyin(nv->data, nvlpacked, nv->len); 6076 if (error) 6077 ERROUT(error); 6078 6079 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6080 if (nvl == NULL) 6081 ERROUT(EBADMSG); 6082 6083 if (! nvlist_exists_bool(nvl, "keep_counters")) 6084 ERROUT(EBADMSG); 6085 6086 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters"); 6087 6088 on_error: 6089 nvlist_destroy(nvl); 6090 free(nvlpacked, M_NVLIST); 6091 return (error); 6092 } 6093 6094 unsigned int 6095 pf_clear_states(const struct pf_kstate_kill *kill) 6096 { 6097 struct pf_state_key_cmp match_key; 6098 struct pf_kstate *s; 6099 struct pfi_kkif *kif; 6100 int idx; 6101 unsigned int killed = 0, dir; 6102 6103 NET_EPOCH_ASSERT(); 6104 6105 for (unsigned int i = 0; i <= V_pf_hashmask; i++) { 6106 struct pf_idhash *ih = &V_pf_idhash[i]; 6107 6108 relock_DIOCCLRSTATES: 6109 PF_HASHROW_LOCK(ih); 6110 LIST_FOREACH(s, &ih->states, entry) { 6111 /* For floating states look at the original kif. */ 6112 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 6113 6114 if (kill->psk_ifname[0] && 6115 strcmp(kill->psk_ifname, 6116 kif->pfik_name)) 6117 continue; 6118 6119 if (kill->psk_kill_match) { 6120 bzero(&match_key, sizeof(match_key)); 6121 6122 if (s->direction == PF_OUT) { 6123 dir = PF_IN; 6124 idx = PF_SK_STACK; 6125 } else { 6126 dir = PF_OUT; 6127 idx = PF_SK_WIRE; 6128 } 6129 6130 match_key.af = s->key[idx]->af; 6131 match_key.proto = s->key[idx]->proto; 6132 PF_ACPY(&match_key.addr[0], 6133 &s->key[idx]->addr[1], match_key.af); 6134 match_key.port[0] = s->key[idx]->port[1]; 6135 PF_ACPY(&match_key.addr[1], 6136 &s->key[idx]->addr[0], match_key.af); 6137 match_key.port[1] = s->key[idx]->port[0]; 6138 } 6139 6140 /* 6141 * Don't send out individual 6142 * delete messages. 6143 */ 6144 s->state_flags |= PFSTATE_NOSYNC; 6145 pf_remove_state(s); 6146 killed++; 6147 6148 if (kill->psk_kill_match) 6149 killed += pf_kill_matching_state(&match_key, 6150 dir); 6151 6152 goto relock_DIOCCLRSTATES; 6153 } 6154 PF_HASHROW_UNLOCK(ih); 6155 } 6156 6157 if (V_pfsync_clear_states_ptr != NULL) 6158 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname); 6159 6160 return (killed); 6161 } 6162 6163 void 6164 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed) 6165 { 6166 struct pf_kstate *s; 6167 6168 NET_EPOCH_ASSERT(); 6169 if (kill->psk_pfcmp.id) { 6170 if (kill->psk_pfcmp.creatorid == 0) 6171 kill->psk_pfcmp.creatorid = V_pf_status.hostid; 6172 if ((s = pf_find_state_byid(kill->psk_pfcmp.id, 6173 kill->psk_pfcmp.creatorid))) { 6174 pf_remove_state(s); 6175 *killed = 1; 6176 } 6177 return; 6178 } 6179 6180 for (unsigned int i = 0; i <= V_pf_hashmask; i++) 6181 *killed += pf_killstates_row(kill, &V_pf_idhash[i]); 6182 } 6183 6184 static int 6185 pf_killstates_nv(struct pfioc_nv *nv) 6186 { 6187 struct pf_kstate_kill kill; 6188 struct epoch_tracker et; 6189 nvlist_t *nvl = NULL; 6190 void *nvlpacked = NULL; 6191 int error = 0; 6192 unsigned int killed = 0; 6193 6194 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6195 6196 if (nv->len > pf_ioctl_maxcount) 6197 ERROUT(ENOMEM); 6198 6199 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6200 error = copyin(nv->data, nvlpacked, nv->len); 6201 if (error) 6202 ERROUT(error); 6203 6204 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6205 if (nvl == NULL) 6206 ERROUT(EBADMSG); 6207 6208 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6209 if (error) 6210 ERROUT(error); 6211 6212 NET_EPOCH_ENTER(et); 6213 pf_killstates(&kill, &killed); 6214 NET_EPOCH_EXIT(et); 6215 6216 free(nvlpacked, M_NVLIST); 6217 nvlpacked = NULL; 6218 nvlist_destroy(nvl); 6219 nvl = nvlist_create(0); 6220 if (nvl == NULL) 6221 ERROUT(ENOMEM); 6222 6223 nvlist_add_number(nvl, "killed", killed); 6224 6225 nvlpacked = nvlist_pack(nvl, &nv->len); 6226 if (nvlpacked == NULL) 6227 ERROUT(ENOMEM); 6228 6229 if (nv->size == 0) 6230 ERROUT(0); 6231 else if (nv->size < nv->len) 6232 ERROUT(ENOSPC); 6233 6234 error = copyout(nvlpacked, nv->data, nv->len); 6235 6236 on_error: 6237 nvlist_destroy(nvl); 6238 free(nvlpacked, M_NVLIST); 6239 return (error); 6240 } 6241 6242 static int 6243 pf_clearstates_nv(struct pfioc_nv *nv) 6244 { 6245 struct pf_kstate_kill kill; 6246 struct epoch_tracker et; 6247 nvlist_t *nvl = NULL; 6248 void *nvlpacked = NULL; 6249 int error = 0; 6250 unsigned int killed; 6251 6252 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6253 6254 if (nv->len > pf_ioctl_maxcount) 6255 ERROUT(ENOMEM); 6256 6257 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6258 error = copyin(nv->data, nvlpacked, nv->len); 6259 if (error) 6260 ERROUT(error); 6261 6262 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6263 if (nvl == NULL) 6264 ERROUT(EBADMSG); 6265 6266 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6267 if (error) 6268 ERROUT(error); 6269 6270 NET_EPOCH_ENTER(et); 6271 killed = pf_clear_states(&kill); 6272 NET_EPOCH_EXIT(et); 6273 6274 free(nvlpacked, M_NVLIST); 6275 nvlpacked = NULL; 6276 nvlist_destroy(nvl); 6277 nvl = nvlist_create(0); 6278 if (nvl == NULL) 6279 ERROUT(ENOMEM); 6280 6281 nvlist_add_number(nvl, "killed", killed); 6282 6283 nvlpacked = nvlist_pack(nvl, &nv->len); 6284 if (nvlpacked == NULL) 6285 ERROUT(ENOMEM); 6286 6287 if (nv->size == 0) 6288 ERROUT(0); 6289 else if (nv->size < nv->len) 6290 ERROUT(ENOSPC); 6291 6292 error = copyout(nvlpacked, nv->data, nv->len); 6293 6294 #undef ERROUT 6295 on_error: 6296 nvlist_destroy(nvl); 6297 free(nvlpacked, M_NVLIST); 6298 return (error); 6299 } 6300 6301 static int 6302 pf_getstate(struct pfioc_nv *nv) 6303 { 6304 nvlist_t *nvl = NULL, *nvls; 6305 void *nvlpacked = NULL; 6306 struct pf_kstate *s = NULL; 6307 int error = 0; 6308 uint64_t id, creatorid; 6309 6310 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6311 6312 if (nv->len > pf_ioctl_maxcount) 6313 ERROUT(ENOMEM); 6314 6315 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6316 error = copyin(nv->data, nvlpacked, nv->len); 6317 if (error) 6318 ERROUT(error); 6319 6320 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6321 if (nvl == NULL) 6322 ERROUT(EBADMSG); 6323 6324 PFNV_CHK(pf_nvuint64(nvl, "id", &id)); 6325 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid)); 6326 6327 s = pf_find_state_byid(id, creatorid); 6328 if (s == NULL) 6329 ERROUT(ENOENT); 6330 6331 free(nvlpacked, M_NVLIST); 6332 nvlpacked = NULL; 6333 nvlist_destroy(nvl); 6334 nvl = nvlist_create(0); 6335 if (nvl == NULL) 6336 ERROUT(ENOMEM); 6337 6338 nvls = pf_state_to_nvstate(s); 6339 if (nvls == NULL) 6340 ERROUT(ENOMEM); 6341 6342 nvlist_add_nvlist(nvl, "state", nvls); 6343 nvlist_destroy(nvls); 6344 6345 nvlpacked = nvlist_pack(nvl, &nv->len); 6346 if (nvlpacked == NULL) 6347 ERROUT(ENOMEM); 6348 6349 if (nv->size == 0) 6350 ERROUT(0); 6351 else if (nv->size < nv->len) 6352 ERROUT(ENOSPC); 6353 6354 error = copyout(nvlpacked, nv->data, nv->len); 6355 6356 #undef ERROUT 6357 errout: 6358 if (s != NULL) 6359 PF_STATE_UNLOCK(s); 6360 free(nvlpacked, M_NVLIST); 6361 nvlist_destroy(nvl); 6362 return (error); 6363 } 6364 6365 /* 6366 * XXX - Check for version mismatch!!! 6367 */ 6368 6369 /* 6370 * Duplicate pfctl -Fa operation to get rid of as much as we can. 6371 */ 6372 static int 6373 shutdown_pf(void) 6374 { 6375 int error = 0; 6376 u_int32_t t[5]; 6377 char nn = '\0'; 6378 struct pf_kanchor *anchor; 6379 struct pf_keth_anchor *eth_anchor; 6380 int rs_num; 6381 6382 do { 6383 /* Unlink rules of all user defined anchors */ 6384 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) { 6385 /* Wildcard based anchors may not have a respective 6386 * explicit anchor rule or they may be left empty 6387 * without rules. It leads to anchor.refcnt=0, and the 6388 * rest of the logic does not expect it. */ 6389 if (anchor->refcnt == 0) 6390 anchor->refcnt = 1; 6391 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6392 if ((error = pf_begin_rules(&t[rs_num], rs_num, 6393 anchor->path)) != 0) { 6394 DPFPRINTF(PF_DEBUG_MISC, ("%s: " 6395 "anchor.path=%s rs_num=%d\n", 6396 __func__, anchor->path, rs_num)); 6397 goto error; /* XXX: rollback? */ 6398 } 6399 } 6400 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6401 error = pf_commit_rules(t[rs_num], rs_num, 6402 anchor->path); 6403 MPASS(error == 0); 6404 } 6405 } 6406 6407 /* Unlink rules of all user defined ether anchors */ 6408 RB_FOREACH(eth_anchor, pf_keth_anchor_global, 6409 &V_pf_keth_anchors) { 6410 /* Wildcard based anchors may not have a respective 6411 * explicit anchor rule or they may be left empty 6412 * without rules. It leads to anchor.refcnt=0, and the 6413 * rest of the logic does not expect it. */ 6414 if (eth_anchor->refcnt == 0) 6415 eth_anchor->refcnt = 1; 6416 if ((error = pf_begin_eth(&t[0], eth_anchor->path)) 6417 != 0) { 6418 DPFPRINTF(PF_DEBUG_MISC, ("%s: eth " 6419 "anchor.path=%s\n", __func__, 6420 eth_anchor->path)); 6421 goto error; 6422 } 6423 error = pf_commit_eth(t[0], eth_anchor->path); 6424 MPASS(error == 0); 6425 } 6426 6427 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) 6428 != 0) { 6429 DPFPRINTF(PF_DEBUG_MISC, ("%s: SCRUB\n", __func__)); 6430 break; 6431 } 6432 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) 6433 != 0) { 6434 DPFPRINTF(PF_DEBUG_MISC, ("%s: FILTER\n", __func__)); 6435 break; /* XXX: rollback? */ 6436 } 6437 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) 6438 != 0) { 6439 DPFPRINTF(PF_DEBUG_MISC, ("%s: NAT\n", __func__)); 6440 break; /* XXX: rollback? */ 6441 } 6442 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) 6443 != 0) { 6444 DPFPRINTF(PF_DEBUG_MISC, ("%s: BINAT\n", __func__)); 6445 break; /* XXX: rollback? */ 6446 } 6447 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) 6448 != 0) { 6449 DPFPRINTF(PF_DEBUG_MISC, ("%s: RDR\n", __func__)); 6450 break; /* XXX: rollback? */ 6451 } 6452 6453 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn); 6454 MPASS(error == 0); 6455 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn); 6456 MPASS(error == 0); 6457 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn); 6458 MPASS(error == 0); 6459 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn); 6460 MPASS(error == 0); 6461 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn); 6462 MPASS(error == 0); 6463 6464 if ((error = pf_clear_tables()) != 0) 6465 break; 6466 6467 if ((error = pf_begin_eth(&t[0], &nn)) != 0) { 6468 DPFPRINTF(PF_DEBUG_MISC, ("%s: eth\n", __func__)); 6469 break; 6470 } 6471 error = pf_commit_eth(t[0], &nn); 6472 MPASS(error == 0); 6473 6474 #ifdef ALTQ 6475 if ((error = pf_begin_altq(&t[0])) != 0) { 6476 DPFPRINTF(PF_DEBUG_MISC, ("%s: ALTQ\n", __func__)); 6477 break; 6478 } 6479 pf_commit_altq(t[0]); 6480 #endif 6481 6482 pf_clear_all_states(); 6483 6484 pf_kill_srcnodes(NULL); 6485 6486 /* status does not use malloced mem so no need to cleanup */ 6487 /* fingerprints and interfaces have their own cleanup code */ 6488 } while(0); 6489 6490 error: 6491 return (error); 6492 } 6493 6494 static pfil_return_t 6495 pf_check_return(int chk, struct mbuf **m) 6496 { 6497 6498 switch (chk) { 6499 case PF_PASS: 6500 if (*m == NULL) 6501 return (PFIL_CONSUMED); 6502 else 6503 return (PFIL_PASS); 6504 break; 6505 default: 6506 if (*m != NULL) { 6507 m_freem(*m); 6508 *m = NULL; 6509 } 6510 return (PFIL_DROPPED); 6511 } 6512 } 6513 6514 static pfil_return_t 6515 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6516 void *ruleset __unused, struct inpcb *inp) 6517 { 6518 int chk; 6519 6520 CURVNET_ASSERT_SET(); 6521 6522 chk = pf_test_eth(PF_IN, flags, ifp, m, inp); 6523 6524 return (pf_check_return(chk, m)); 6525 } 6526 6527 static pfil_return_t 6528 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6529 void *ruleset __unused, struct inpcb *inp) 6530 { 6531 int chk; 6532 6533 CURVNET_ASSERT_SET(); 6534 6535 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp); 6536 6537 return (pf_check_return(chk, m)); 6538 } 6539 6540 #ifdef INET 6541 static pfil_return_t 6542 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6543 void *ruleset __unused, struct inpcb *inp) 6544 { 6545 int chk; 6546 6547 CURVNET_ASSERT_SET(); 6548 6549 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL); 6550 6551 return (pf_check_return(chk, m)); 6552 } 6553 6554 static pfil_return_t 6555 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6556 void *ruleset __unused, struct inpcb *inp) 6557 { 6558 int chk; 6559 6560 CURVNET_ASSERT_SET(); 6561 6562 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL); 6563 6564 return (pf_check_return(chk, m)); 6565 } 6566 #endif 6567 6568 #ifdef INET6 6569 static pfil_return_t 6570 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags, 6571 void *ruleset __unused, struct inpcb *inp) 6572 { 6573 int chk; 6574 6575 CURVNET_ASSERT_SET(); 6576 6577 /* 6578 * In case of loopback traffic IPv6 uses the real interface in 6579 * order to support scoped addresses. In order to support stateful 6580 * filtering we have change this to lo0 as it is the case in IPv4. 6581 */ 6582 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, 6583 m, inp, NULL); 6584 6585 return (pf_check_return(chk, m)); 6586 } 6587 6588 static pfil_return_t 6589 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags, 6590 void *ruleset __unused, struct inpcb *inp) 6591 { 6592 int chk; 6593 6594 CURVNET_ASSERT_SET(); 6595 6596 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL); 6597 6598 return (pf_check_return(chk, m)); 6599 } 6600 #endif /* INET6 */ 6601 6602 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook); 6603 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook); 6604 #define V_pf_eth_in_hook VNET(pf_eth_in_hook) 6605 #define V_pf_eth_out_hook VNET(pf_eth_out_hook) 6606 6607 #ifdef INET 6608 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook); 6609 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook); 6610 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook) 6611 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook) 6612 #endif 6613 #ifdef INET6 6614 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook); 6615 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook); 6616 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook) 6617 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook) 6618 #endif 6619 6620 static void 6621 hook_pf_eth(void) 6622 { 6623 struct pfil_hook_args pha = { 6624 .pa_version = PFIL_VERSION, 6625 .pa_modname = "pf", 6626 .pa_type = PFIL_TYPE_ETHERNET, 6627 }; 6628 struct pfil_link_args pla = { 6629 .pa_version = PFIL_VERSION, 6630 }; 6631 int ret __diagused; 6632 6633 if (atomic_load_bool(&V_pf_pfil_eth_hooked)) 6634 return; 6635 6636 pha.pa_mbuf_chk = pf_eth_check_in; 6637 pha.pa_flags = PFIL_IN; 6638 pha.pa_rulname = "eth-in"; 6639 V_pf_eth_in_hook = pfil_add_hook(&pha); 6640 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6641 pla.pa_head = V_link_pfil_head; 6642 pla.pa_hook = V_pf_eth_in_hook; 6643 ret = pfil_link(&pla); 6644 MPASS(ret == 0); 6645 pha.pa_mbuf_chk = pf_eth_check_out; 6646 pha.pa_flags = PFIL_OUT; 6647 pha.pa_rulname = "eth-out"; 6648 V_pf_eth_out_hook = pfil_add_hook(&pha); 6649 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6650 pla.pa_head = V_link_pfil_head; 6651 pla.pa_hook = V_pf_eth_out_hook; 6652 ret = pfil_link(&pla); 6653 MPASS(ret == 0); 6654 6655 atomic_store_bool(&V_pf_pfil_eth_hooked, true); 6656 } 6657 6658 static void 6659 hook_pf(void) 6660 { 6661 struct pfil_hook_args pha = { 6662 .pa_version = PFIL_VERSION, 6663 .pa_modname = "pf", 6664 }; 6665 struct pfil_link_args pla = { 6666 .pa_version = PFIL_VERSION, 6667 }; 6668 int ret __diagused; 6669 6670 if (atomic_load_bool(&V_pf_pfil_hooked)) 6671 return; 6672 6673 #ifdef INET 6674 pha.pa_type = PFIL_TYPE_IP4; 6675 pha.pa_mbuf_chk = pf_check_in; 6676 pha.pa_flags = PFIL_IN; 6677 pha.pa_rulname = "default-in"; 6678 V_pf_ip4_in_hook = pfil_add_hook(&pha); 6679 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6680 pla.pa_head = V_inet_pfil_head; 6681 pla.pa_hook = V_pf_ip4_in_hook; 6682 ret = pfil_link(&pla); 6683 MPASS(ret == 0); 6684 pha.pa_mbuf_chk = pf_check_out; 6685 pha.pa_flags = PFIL_OUT; 6686 pha.pa_rulname = "default-out"; 6687 V_pf_ip4_out_hook = pfil_add_hook(&pha); 6688 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6689 pla.pa_head = V_inet_pfil_head; 6690 pla.pa_hook = V_pf_ip4_out_hook; 6691 ret = pfil_link(&pla); 6692 MPASS(ret == 0); 6693 if (V_pf_filter_local) { 6694 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6695 pla.pa_head = V_inet_local_pfil_head; 6696 pla.pa_hook = V_pf_ip4_out_hook; 6697 ret = pfil_link(&pla); 6698 MPASS(ret == 0); 6699 } 6700 #endif 6701 #ifdef INET6 6702 pha.pa_type = PFIL_TYPE_IP6; 6703 pha.pa_mbuf_chk = pf_check6_in; 6704 pha.pa_flags = PFIL_IN; 6705 pha.pa_rulname = "default-in6"; 6706 V_pf_ip6_in_hook = pfil_add_hook(&pha); 6707 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6708 pla.pa_head = V_inet6_pfil_head; 6709 pla.pa_hook = V_pf_ip6_in_hook; 6710 ret = pfil_link(&pla); 6711 MPASS(ret == 0); 6712 pha.pa_mbuf_chk = pf_check6_out; 6713 pha.pa_rulname = "default-out6"; 6714 pha.pa_flags = PFIL_OUT; 6715 V_pf_ip6_out_hook = pfil_add_hook(&pha); 6716 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6717 pla.pa_head = V_inet6_pfil_head; 6718 pla.pa_hook = V_pf_ip6_out_hook; 6719 ret = pfil_link(&pla); 6720 MPASS(ret == 0); 6721 if (V_pf_filter_local) { 6722 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6723 pla.pa_head = V_inet6_local_pfil_head; 6724 pla.pa_hook = V_pf_ip6_out_hook; 6725 ret = pfil_link(&pla); 6726 MPASS(ret == 0); 6727 } 6728 #endif 6729 6730 atomic_store_bool(&V_pf_pfil_hooked, true); 6731 } 6732 6733 static void 6734 dehook_pf_eth(void) 6735 { 6736 6737 if (!atomic_load_bool(&V_pf_pfil_eth_hooked)) 6738 return; 6739 6740 pfil_remove_hook(V_pf_eth_in_hook); 6741 pfil_remove_hook(V_pf_eth_out_hook); 6742 6743 atomic_store_bool(&V_pf_pfil_eth_hooked, false); 6744 } 6745 6746 static void 6747 dehook_pf(void) 6748 { 6749 6750 if (!atomic_load_bool(&V_pf_pfil_hooked)) 6751 return; 6752 6753 #ifdef INET 6754 pfil_remove_hook(V_pf_ip4_in_hook); 6755 pfil_remove_hook(V_pf_ip4_out_hook); 6756 #endif 6757 #ifdef INET6 6758 pfil_remove_hook(V_pf_ip6_in_hook); 6759 pfil_remove_hook(V_pf_ip6_out_hook); 6760 #endif 6761 6762 atomic_store_bool(&V_pf_pfil_hooked, false); 6763 } 6764 6765 static void 6766 pf_load_vnet(void) 6767 { 6768 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname), 6769 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 6770 6771 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE); 6772 sx_init(&V_pf_ioctl_lock, "pf ioctl"); 6773 6774 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize, 6775 PF_RULE_TAG_HASH_SIZE_DEFAULT); 6776 #ifdef ALTQ 6777 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize, 6778 PF_QUEUE_TAG_HASH_SIZE_DEFAULT); 6779 #endif 6780 6781 V_pf_keth = &V_pf_main_keth_anchor.ruleset; 6782 6783 pfattach_vnet(); 6784 V_pf_vnet_active = 1; 6785 } 6786 6787 static int 6788 pf_load(void) 6789 { 6790 int error; 6791 6792 sx_init(&pf_end_lock, "pf end thread"); 6793 6794 pf_mtag_initialize(); 6795 6796 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME); 6797 if (pf_dev == NULL) 6798 return (ENOMEM); 6799 6800 pf_end_threads = 0; 6801 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge"); 6802 if (error != 0) 6803 return (error); 6804 6805 pfi_initialize(); 6806 6807 return (0); 6808 } 6809 6810 static void 6811 pf_unload_vnet(void) 6812 { 6813 int ret __diagused; 6814 6815 V_pf_vnet_active = 0; 6816 V_pf_status.running = 0; 6817 dehook_pf(); 6818 dehook_pf_eth(); 6819 6820 PF_RULES_WLOCK(); 6821 pf_syncookies_cleanup(); 6822 shutdown_pf(); 6823 PF_RULES_WUNLOCK(); 6824 6825 ret = swi_remove(V_pf_swi_cookie); 6826 MPASS(ret == 0); 6827 ret = intr_event_destroy(V_pf_swi_ie); 6828 MPASS(ret == 0); 6829 6830 pf_unload_vnet_purge(); 6831 6832 pf_normalize_cleanup(); 6833 PF_RULES_WLOCK(); 6834 pfi_cleanup_vnet(); 6835 PF_RULES_WUNLOCK(); 6836 pfr_cleanup(); 6837 pf_osfp_flush(); 6838 pf_cleanup(); 6839 if (IS_DEFAULT_VNET(curvnet)) 6840 pf_mtag_cleanup(); 6841 6842 pf_cleanup_tagset(&V_pf_tags); 6843 #ifdef ALTQ 6844 pf_cleanup_tagset(&V_pf_qids); 6845 #endif 6846 uma_zdestroy(V_pf_tag_z); 6847 6848 #ifdef PF_WANT_32_TO_64_COUNTER 6849 PF_RULES_WLOCK(); 6850 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); 6851 6852 MPASS(LIST_EMPTY(&V_pf_allkiflist)); 6853 MPASS(V_pf_allkifcount == 0); 6854 6855 LIST_REMOVE(&V_pf_default_rule, allrulelist); 6856 V_pf_allrulecount--; 6857 LIST_REMOVE(V_pf_rulemarker, allrulelist); 6858 6859 MPASS(LIST_EMPTY(&V_pf_allrulelist)); 6860 MPASS(V_pf_allrulecount == 0); 6861 6862 PF_RULES_WUNLOCK(); 6863 6864 free(V_pf_kifmarker, PFI_MTYPE); 6865 free(V_pf_rulemarker, M_PFRULE); 6866 #endif 6867 6868 /* Free counters last as we updated them during shutdown. */ 6869 pf_counter_u64_deinit(&V_pf_default_rule.evaluations); 6870 for (int i = 0; i < 2; i++) { 6871 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]); 6872 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]); 6873 } 6874 counter_u64_free(V_pf_default_rule.states_cur); 6875 counter_u64_free(V_pf_default_rule.states_tot); 6876 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 6877 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]); 6878 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp); 6879 6880 for (int i = 0; i < PFRES_MAX; i++) 6881 counter_u64_free(V_pf_status.counters[i]); 6882 for (int i = 0; i < KLCNT_MAX; i++) 6883 counter_u64_free(V_pf_status.lcounters[i]); 6884 for (int i = 0; i < FCNT_MAX; i++) 6885 pf_counter_u64_deinit(&V_pf_status.fcounters[i]); 6886 for (int i = 0; i < SCNT_MAX; i++) 6887 counter_u64_free(V_pf_status.scounters[i]); 6888 6889 rm_destroy(&V_pf_rules_lock); 6890 sx_destroy(&V_pf_ioctl_lock); 6891 } 6892 6893 static void 6894 pf_unload(void) 6895 { 6896 6897 sx_xlock(&pf_end_lock); 6898 pf_end_threads = 1; 6899 while (pf_end_threads < 2) { 6900 wakeup_one(pf_purge_thread); 6901 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0); 6902 } 6903 sx_xunlock(&pf_end_lock); 6904 6905 pf_nl_unregister(); 6906 6907 if (pf_dev != NULL) 6908 destroy_dev(pf_dev); 6909 6910 pfi_cleanup(); 6911 6912 sx_destroy(&pf_end_lock); 6913 } 6914 6915 static void 6916 vnet_pf_init(void *unused __unused) 6917 { 6918 6919 pf_load_vnet(); 6920 } 6921 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6922 vnet_pf_init, NULL); 6923 6924 static void 6925 vnet_pf_uninit(const void *unused __unused) 6926 { 6927 6928 pf_unload_vnet(); 6929 } 6930 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL); 6931 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6932 vnet_pf_uninit, NULL); 6933 6934 static int 6935 pf_modevent(module_t mod, int type, void *data) 6936 { 6937 int error = 0; 6938 6939 switch(type) { 6940 case MOD_LOAD: 6941 error = pf_load(); 6942 pf_nl_register(); 6943 break; 6944 case MOD_UNLOAD: 6945 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after 6946 * the vnet_pf_uninit()s */ 6947 break; 6948 default: 6949 error = EINVAL; 6950 break; 6951 } 6952 6953 return (error); 6954 } 6955 6956 static moduledata_t pf_mod = { 6957 "pf", 6958 pf_modevent, 6959 0 6960 }; 6961 6962 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND); 6963 MODULE_DEPEND(pf, netlink, 1, 1, 1); 6964 MODULE_DEPEND(pf, crypto, 1, 1, 1); 6965 MODULE_VERSION(pf, PF_MODVER); 6966