1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 2001 Daniel Hartmeier 5 * Copyright (c) 2002,2003 Henning Brauer 6 * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org> 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 13 * - Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * - Redistributions in binary form must reproduce the above 16 * copyright notice, this list of conditions and the following 17 * disclaimer in the documentation and/or other materials provided 18 * with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 * 33 * Effort sponsored in part by the Defense Advanced Research Projects 34 * Agency (DARPA) and Air Force Research Laboratory, Air Force 35 * Materiel Command, USAF, under agreement number F30602-01-2-0537. 36 * 37 * $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $ 38 */ 39 40 #include <sys/cdefs.h> 41 __FBSDID("$FreeBSD$"); 42 43 #include "opt_inet.h" 44 #include "opt_inet6.h" 45 #include "opt_bpf.h" 46 #include "opt_pf.h" 47 48 #include <sys/param.h> 49 #include <sys/_bitset.h> 50 #include <sys/bitset.h> 51 #include <sys/bus.h> 52 #include <sys/conf.h> 53 #include <sys/endian.h> 54 #include <sys/fcntl.h> 55 #include <sys/filio.h> 56 #include <sys/hash.h> 57 #include <sys/interrupt.h> 58 #include <sys/jail.h> 59 #include <sys/kernel.h> 60 #include <sys/kthread.h> 61 #include <sys/lock.h> 62 #include <sys/mbuf.h> 63 #include <sys/module.h> 64 #include <sys/nv.h> 65 #include <sys/proc.h> 66 #include <sys/sdt.h> 67 #include <sys/smp.h> 68 #include <sys/socket.h> 69 #include <sys/sysctl.h> 70 #include <sys/md5.h> 71 #include <sys/ucred.h> 72 73 #include <net/if.h> 74 #include <net/if_var.h> 75 #include <net/if_private.h> 76 #include <net/vnet.h> 77 #include <net/route.h> 78 #include <net/pfil.h> 79 #include <net/pfvar.h> 80 #include <net/if_pfsync.h> 81 #include <net/if_pflog.h> 82 83 #include <netinet/in.h> 84 #include <netinet/ip.h> 85 #include <netinet/ip_var.h> 86 #include <netinet6/ip6_var.h> 87 #include <netinet/ip_icmp.h> 88 #include <netpfil/pf/pf_nv.h> 89 90 #ifdef INET6 91 #include <netinet/ip6.h> 92 #endif /* INET6 */ 93 94 #ifdef ALTQ 95 #include <net/altq/altq.h> 96 #endif 97 98 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int"); 99 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int"); 100 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int"); 101 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int"); 102 103 static struct pf_kpool *pf_get_kpool(const char *, u_int32_t, u_int8_t, 104 u_int32_t, u_int8_t, u_int8_t, u_int8_t); 105 106 static void pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *); 107 static void pf_empty_kpool(struct pf_kpalist *); 108 static int pfioctl(struct cdev *, u_long, caddr_t, int, 109 struct thread *); 110 static int pf_begin_eth(uint32_t *, const char *); 111 static void pf_rollback_eth_cb(struct epoch_context *); 112 static int pf_rollback_eth(uint32_t, const char *); 113 static int pf_commit_eth(uint32_t, const char *); 114 static void pf_free_eth_rule(struct pf_keth_rule *); 115 #ifdef ALTQ 116 static int pf_begin_altq(u_int32_t *); 117 static int pf_rollback_altq(u_int32_t); 118 static int pf_commit_altq(u_int32_t); 119 static int pf_enable_altq(struct pf_altq *); 120 static int pf_disable_altq(struct pf_altq *); 121 static uint16_t pf_qname2qid(const char *); 122 static void pf_qid_unref(uint16_t); 123 #endif /* ALTQ */ 124 static int pf_begin_rules(u_int32_t *, int, const char *); 125 static int pf_rollback_rules(u_int32_t, int, char *); 126 static int pf_setup_pfsync_matching(struct pf_kruleset *); 127 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *); 128 static void pf_hash_rule(struct pf_krule *); 129 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *); 130 static int pf_commit_rules(u_int32_t, int, char *); 131 static int pf_addr_setup(struct pf_kruleset *, 132 struct pf_addr_wrap *, sa_family_t); 133 static void pf_addr_copyout(struct pf_addr_wrap *); 134 static void pf_src_node_copy(const struct pf_ksrc_node *, 135 struct pf_src_node *); 136 #ifdef ALTQ 137 static int pf_export_kaltq(struct pf_altq *, 138 struct pfioc_altq_v1 *, size_t); 139 static int pf_import_kaltq(struct pfioc_altq_v1 *, 140 struct pf_altq *, size_t); 141 #endif /* ALTQ */ 142 143 VNET_DEFINE(struct pf_krule, pf_default_rule); 144 145 static __inline int pf_krule_compare(struct pf_krule *, 146 struct pf_krule *); 147 148 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare); 149 150 #ifdef ALTQ 151 VNET_DEFINE_STATIC(int, pf_altq_running); 152 #define V_pf_altq_running VNET(pf_altq_running) 153 #endif 154 155 #define TAGID_MAX 50000 156 struct pf_tagname { 157 TAILQ_ENTRY(pf_tagname) namehash_entries; 158 TAILQ_ENTRY(pf_tagname) taghash_entries; 159 char name[PF_TAG_NAME_SIZE]; 160 uint16_t tag; 161 int ref; 162 }; 163 164 struct pf_tagset { 165 TAILQ_HEAD(, pf_tagname) *namehash; 166 TAILQ_HEAD(, pf_tagname) *taghash; 167 unsigned int mask; 168 uint32_t seed; 169 BITSET_DEFINE(, TAGID_MAX) avail; 170 }; 171 172 VNET_DEFINE(struct pf_tagset, pf_tags); 173 #define V_pf_tags VNET(pf_tags) 174 static unsigned int pf_rule_tag_hashsize; 175 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128 176 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN, 177 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT, 178 "Size of pf(4) rule tag hashtable"); 179 180 #ifdef ALTQ 181 VNET_DEFINE(struct pf_tagset, pf_qids); 182 #define V_pf_qids VNET(pf_qids) 183 static unsigned int pf_queue_tag_hashsize; 184 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128 185 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN, 186 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT, 187 "Size of pf(4) queue tag hashtable"); 188 #endif 189 VNET_DEFINE(uma_zone_t, pf_tag_z); 190 #define V_pf_tag_z VNET(pf_tag_z) 191 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db"); 192 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules"); 193 194 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE) 195 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE 196 #endif 197 198 static void pf_init_tagset(struct pf_tagset *, unsigned int *, 199 unsigned int); 200 static void pf_cleanup_tagset(struct pf_tagset *); 201 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *); 202 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t); 203 static u_int16_t tagname2tag(struct pf_tagset *, const char *); 204 static u_int16_t pf_tagname2tag(const char *); 205 static void tag_unref(struct pf_tagset *, u_int16_t); 206 207 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x 208 209 struct cdev *pf_dev; 210 211 /* 212 * XXX - These are new and need to be checked when moveing to a new version 213 */ 214 static void pf_clear_all_states(void); 215 static unsigned int pf_clear_states(const struct pf_kstate_kill *); 216 static void pf_killstates(struct pf_kstate_kill *, 217 unsigned int *); 218 static int pf_killstates_row(struct pf_kstate_kill *, 219 struct pf_idhash *); 220 static int pf_killstates_nv(struct pfioc_nv *); 221 static int pf_clearstates_nv(struct pfioc_nv *); 222 static int pf_getstate(struct pfioc_nv *); 223 static int pf_getstatus(struct pfioc_nv *); 224 static int pf_clear_tables(void); 225 static void pf_clear_srcnodes(struct pf_ksrc_node *); 226 static void pf_kill_srcnodes(struct pfioc_src_node_kill *); 227 static int pf_keepcounters(struct pfioc_nv *); 228 static void pf_tbladdr_copyout(struct pf_addr_wrap *); 229 230 /* 231 * Wrapper functions for pfil(9) hooks 232 */ 233 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, 234 int flags, void *ruleset __unused, struct inpcb *inp); 235 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, 236 int flags, void *ruleset __unused, struct inpcb *inp); 237 #ifdef INET 238 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp, 239 int flags, void *ruleset __unused, struct inpcb *inp); 240 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp, 241 int flags, void *ruleset __unused, struct inpcb *inp); 242 #endif 243 #ifdef INET6 244 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp, 245 int flags, void *ruleset __unused, struct inpcb *inp); 246 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp, 247 int flags, void *ruleset __unused, struct inpcb *inp); 248 #endif 249 250 static void hook_pf_eth(void); 251 static void hook_pf(void); 252 static void dehook_pf_eth(void); 253 static void dehook_pf(void); 254 static int shutdown_pf(void); 255 static int pf_load(void); 256 static void pf_unload(void); 257 258 static struct cdevsw pf_cdevsw = { 259 .d_ioctl = pfioctl, 260 .d_name = PF_NAME, 261 .d_version = D_VERSION, 262 }; 263 264 VNET_DEFINE_STATIC(bool, pf_pfil_hooked); 265 #define V_pf_pfil_hooked VNET(pf_pfil_hooked) 266 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked); 267 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked) 268 269 /* 270 * We need a flag that is neither hooked nor running to know when 271 * the VNET is "valid". We primarily need this to control (global) 272 * external event, e.g., eventhandlers. 273 */ 274 VNET_DEFINE(int, pf_vnet_active); 275 #define V_pf_vnet_active VNET(pf_vnet_active) 276 277 int pf_end_threads; 278 struct proc *pf_purge_proc; 279 280 VNET_DEFINE(struct rmlock, pf_rules_lock); 281 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock); 282 #define V_pf_ioctl_lock VNET(pf_ioctl_lock) 283 struct sx pf_end_lock; 284 285 /* pfsync */ 286 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr); 287 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr); 288 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr); 289 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr); 290 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr); 291 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr); 292 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr; 293 294 /* pflog */ 295 pflog_packet_t *pflog_packet_ptr = NULL; 296 297 /* 298 * Copy a user-provided string, returning an error if truncation would occur. 299 * Avoid scanning past "sz" bytes in the source string since there's no 300 * guarantee that it's nul-terminated. 301 */ 302 static int 303 pf_user_strcpy(char *dst, const char *src, size_t sz) 304 { 305 if (strnlen(src, sz) == sz) 306 return (EINVAL); 307 (void)strlcpy(dst, src, sz); 308 return (0); 309 } 310 311 static void 312 pfattach_vnet(void) 313 { 314 u_int32_t *my_timeout = V_pf_default_rule.timeout; 315 316 bzero(&V_pf_status, sizeof(V_pf_status)); 317 318 pf_initialize(); 319 pfr_initialize(); 320 pfi_initialize_vnet(); 321 pf_normalize_init(); 322 pf_syncookies_init(); 323 324 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT; 325 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT; 326 327 RB_INIT(&V_pf_anchors); 328 pf_init_kruleset(&pf_main_ruleset); 329 330 pf_init_keth(V_pf_keth); 331 332 /* default rule should never be garbage collected */ 333 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; 334 #ifdef PF_DEFAULT_TO_DROP 335 V_pf_default_rule.action = PF_DROP; 336 #else 337 V_pf_default_rule.action = PF_PASS; 338 #endif 339 V_pf_default_rule.nr = -1; 340 V_pf_default_rule.rtableid = -1; 341 342 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK); 343 for (int i = 0; i < 2; i++) { 344 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK); 345 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK); 346 } 347 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); 348 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); 349 V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK); 350 351 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 352 M_WAITOK | M_ZERO); 353 354 #ifdef PF_WANT_32_TO_64_COUNTER 355 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO); 356 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO); 357 PF_RULES_WLOCK(); 358 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); 359 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist); 360 V_pf_allrulecount++; 361 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); 362 PF_RULES_WUNLOCK(); 363 #endif 364 365 /* initialize default timeouts */ 366 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 367 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; 368 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 369 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; 370 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; 371 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; 372 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; 373 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; 374 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; 375 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; 376 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; 377 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; 378 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; 379 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; 380 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; 381 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; 382 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; 383 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; 384 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; 385 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; 386 387 V_pf_status.debug = PF_DEBUG_URGENT; 388 /* 389 * XXX This is different than in OpenBSD where reassembly is enabled by 390 * defult. In FreeBSD we expect people to still use scrub rules and 391 * switch to the new syntax later. Only when they switch they must 392 * explicitly enable reassemle. We could change the default once the 393 * scrub rule functionality is hopefully removed some day in future. 394 */ 395 V_pf_status.reass = 0; 396 397 V_pf_pfil_hooked = false; 398 V_pf_pfil_eth_hooked = false; 399 400 /* XXX do our best to avoid a conflict */ 401 V_pf_status.hostid = arc4random(); 402 403 for (int i = 0; i < PFRES_MAX; i++) 404 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); 405 for (int i = 0; i < KLCNT_MAX; i++) 406 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); 407 for (int i = 0; i < FCNT_MAX; i++) 408 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK); 409 for (int i = 0; i < SCNT_MAX; i++) 410 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); 411 412 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET, 413 INTR_MPSAFE, &V_pf_swi_cookie) != 0) 414 /* XXXGL: leaked all above. */ 415 return; 416 } 417 418 static struct pf_kpool * 419 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action, 420 u_int32_t rule_number, u_int8_t r_last, u_int8_t active, 421 u_int8_t check_ticket) 422 { 423 struct pf_kruleset *ruleset; 424 struct pf_krule *rule; 425 int rs_num; 426 427 ruleset = pf_find_kruleset(anchor); 428 if (ruleset == NULL) 429 return (NULL); 430 rs_num = pf_get_ruleset_number(rule_action); 431 if (rs_num >= PF_RULESET_MAX) 432 return (NULL); 433 if (active) { 434 if (check_ticket && ticket != 435 ruleset->rules[rs_num].active.ticket) 436 return (NULL); 437 if (r_last) 438 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 439 pf_krulequeue); 440 else 441 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 442 } else { 443 if (check_ticket && ticket != 444 ruleset->rules[rs_num].inactive.ticket) 445 return (NULL); 446 if (r_last) 447 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 448 pf_krulequeue); 449 else 450 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); 451 } 452 if (!r_last) { 453 while ((rule != NULL) && (rule->nr != rule_number)) 454 rule = TAILQ_NEXT(rule, entries); 455 } 456 if (rule == NULL) 457 return (NULL); 458 459 return (&rule->rpool); 460 } 461 462 static void 463 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb) 464 { 465 struct pf_kpooladdr *mv_pool_pa; 466 467 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { 468 TAILQ_REMOVE(poola, mv_pool_pa, entries); 469 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); 470 } 471 } 472 473 static void 474 pf_empty_kpool(struct pf_kpalist *poola) 475 { 476 struct pf_kpooladdr *pa; 477 478 while ((pa = TAILQ_FIRST(poola)) != NULL) { 479 switch (pa->addr.type) { 480 case PF_ADDR_DYNIFTL: 481 pfi_dynaddr_remove(pa->addr.p.dyn); 482 break; 483 case PF_ADDR_TABLE: 484 /* XXX: this could be unfinished pooladdr on pabuf */ 485 if (pa->addr.p.tbl != NULL) 486 pfr_detach_table(pa->addr.p.tbl); 487 break; 488 } 489 if (pa->kif) 490 pfi_kkif_unref(pa->kif); 491 TAILQ_REMOVE(poola, pa, entries); 492 free(pa, M_PFRULE); 493 } 494 } 495 496 static void 497 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 498 { 499 500 PF_RULES_WASSERT(); 501 PF_UNLNKDRULES_ASSERT(); 502 503 TAILQ_REMOVE(rulequeue, rule, entries); 504 505 rule->rule_ref |= PFRULE_REFS; 506 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); 507 } 508 509 static void 510 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 511 { 512 513 PF_RULES_WASSERT(); 514 515 PF_UNLNKDRULES_LOCK(); 516 pf_unlink_rule_locked(rulequeue, rule); 517 PF_UNLNKDRULES_UNLOCK(); 518 } 519 520 static void 521 pf_free_eth_rule(struct pf_keth_rule *rule) 522 { 523 PF_RULES_WASSERT(); 524 525 if (rule == NULL) 526 return; 527 528 if (rule->tag) 529 tag_unref(&V_pf_tags, rule->tag); 530 if (rule->match_tag) 531 tag_unref(&V_pf_tags, rule->match_tag); 532 #ifdef ALTQ 533 pf_qid_unref(rule->qid); 534 #endif 535 536 if (rule->bridge_to) 537 pfi_kkif_unref(rule->bridge_to); 538 if (rule->kif) 539 pfi_kkif_unref(rule->kif); 540 541 if (rule->ipsrc.addr.type == PF_ADDR_TABLE) 542 pfr_detach_table(rule->ipsrc.addr.p.tbl); 543 if (rule->ipdst.addr.type == PF_ADDR_TABLE) 544 pfr_detach_table(rule->ipdst.addr.p.tbl); 545 546 counter_u64_free(rule->evaluations); 547 for (int i = 0; i < 2; i++) { 548 counter_u64_free(rule->packets[i]); 549 counter_u64_free(rule->bytes[i]); 550 } 551 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 552 pf_keth_anchor_remove(rule); 553 554 free(rule, M_PFRULE); 555 } 556 557 void 558 pf_free_rule(struct pf_krule *rule) 559 { 560 561 PF_RULES_WASSERT(); 562 PF_CONFIG_ASSERT(); 563 564 if (rule->tag) 565 tag_unref(&V_pf_tags, rule->tag); 566 if (rule->match_tag) 567 tag_unref(&V_pf_tags, rule->match_tag); 568 #ifdef ALTQ 569 if (rule->pqid != rule->qid) 570 pf_qid_unref(rule->pqid); 571 pf_qid_unref(rule->qid); 572 #endif 573 switch (rule->src.addr.type) { 574 case PF_ADDR_DYNIFTL: 575 pfi_dynaddr_remove(rule->src.addr.p.dyn); 576 break; 577 case PF_ADDR_TABLE: 578 pfr_detach_table(rule->src.addr.p.tbl); 579 break; 580 } 581 switch (rule->dst.addr.type) { 582 case PF_ADDR_DYNIFTL: 583 pfi_dynaddr_remove(rule->dst.addr.p.dyn); 584 break; 585 case PF_ADDR_TABLE: 586 pfr_detach_table(rule->dst.addr.p.tbl); 587 break; 588 } 589 if (rule->overload_tbl) 590 pfr_detach_table(rule->overload_tbl); 591 if (rule->kif) 592 pfi_kkif_unref(rule->kif); 593 pf_kanchor_remove(rule); 594 pf_empty_kpool(&rule->rpool.list); 595 596 pf_krule_free(rule); 597 } 598 599 static void 600 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size, 601 unsigned int default_size) 602 { 603 unsigned int i; 604 unsigned int hashsize; 605 606 if (*tunable_size == 0 || !powerof2(*tunable_size)) 607 *tunable_size = default_size; 608 609 hashsize = *tunable_size; 610 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH, 611 M_WAITOK); 612 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH, 613 M_WAITOK); 614 ts->mask = hashsize - 1; 615 ts->seed = arc4random(); 616 for (i = 0; i < hashsize; i++) { 617 TAILQ_INIT(&ts->namehash[i]); 618 TAILQ_INIT(&ts->taghash[i]); 619 } 620 BIT_FILL(TAGID_MAX, &ts->avail); 621 } 622 623 static void 624 pf_cleanup_tagset(struct pf_tagset *ts) 625 { 626 unsigned int i; 627 unsigned int hashsize; 628 struct pf_tagname *t, *tmp; 629 630 /* 631 * Only need to clean up one of the hashes as each tag is hashed 632 * into each table. 633 */ 634 hashsize = ts->mask + 1; 635 for (i = 0; i < hashsize; i++) 636 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp) 637 uma_zfree(V_pf_tag_z, t); 638 639 free(ts->namehash, M_PFHASH); 640 free(ts->taghash, M_PFHASH); 641 } 642 643 static uint16_t 644 tagname2hashindex(const struct pf_tagset *ts, const char *tagname) 645 { 646 size_t len; 647 648 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1); 649 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask); 650 } 651 652 static uint16_t 653 tag2hashindex(const struct pf_tagset *ts, uint16_t tag) 654 { 655 656 return (tag & ts->mask); 657 } 658 659 static u_int16_t 660 tagname2tag(struct pf_tagset *ts, const char *tagname) 661 { 662 struct pf_tagname *tag; 663 u_int32_t index; 664 u_int16_t new_tagid; 665 666 PF_RULES_WASSERT(); 667 668 index = tagname2hashindex(ts, tagname); 669 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 670 if (strcmp(tagname, tag->name) == 0) { 671 tag->ref++; 672 return (tag->tag); 673 } 674 675 /* 676 * new entry 677 * 678 * to avoid fragmentation, we do a linear search from the beginning 679 * and take the first free slot we find. 680 */ 681 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail); 682 /* 683 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX]. 684 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits 685 * set. It may also return a bit number greater than TAGID_MAX due 686 * to rounding of the number of bits in the vector up to a multiple 687 * of the vector word size at declaration/allocation time. 688 */ 689 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) 690 return (0); 691 692 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */ 693 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail); 694 695 /* allocate and fill new struct pf_tagname */ 696 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT); 697 if (tag == NULL) 698 return (0); 699 strlcpy(tag->name, tagname, sizeof(tag->name)); 700 tag->tag = new_tagid; 701 tag->ref = 1; 702 703 /* Insert into namehash */ 704 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries); 705 706 /* Insert into taghash */ 707 index = tag2hashindex(ts, new_tagid); 708 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries); 709 710 return (tag->tag); 711 } 712 713 static void 714 tag_unref(struct pf_tagset *ts, u_int16_t tag) 715 { 716 struct pf_tagname *t; 717 uint16_t index; 718 719 PF_RULES_WASSERT(); 720 721 index = tag2hashindex(ts, tag); 722 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 723 if (tag == t->tag) { 724 if (--t->ref == 0) { 725 TAILQ_REMOVE(&ts->taghash[index], t, 726 taghash_entries); 727 index = tagname2hashindex(ts, t->name); 728 TAILQ_REMOVE(&ts->namehash[index], t, 729 namehash_entries); 730 /* Bits are 0-based for BIT_SET() */ 731 BIT_SET(TAGID_MAX, tag - 1, &ts->avail); 732 uma_zfree(V_pf_tag_z, t); 733 } 734 break; 735 } 736 } 737 738 static uint16_t 739 pf_tagname2tag(const char *tagname) 740 { 741 return (tagname2tag(&V_pf_tags, tagname)); 742 } 743 744 static int 745 pf_begin_eth(uint32_t *ticket, const char *anchor) 746 { 747 struct pf_keth_rule *rule, *tmp; 748 struct pf_keth_ruleset *rs; 749 750 PF_RULES_WASSERT(); 751 752 rs = pf_find_or_create_keth_ruleset(anchor); 753 if (rs == NULL) 754 return (EINVAL); 755 756 /* Purge old inactive rules. */ 757 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 758 tmp) { 759 TAILQ_REMOVE(rs->inactive.rules, rule, 760 entries); 761 pf_free_eth_rule(rule); 762 } 763 764 *ticket = ++rs->inactive.ticket; 765 rs->inactive.open = 1; 766 767 return (0); 768 } 769 770 static void 771 pf_rollback_eth_cb(struct epoch_context *ctx) 772 { 773 struct pf_keth_ruleset *rs; 774 775 rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx); 776 777 CURVNET_SET(rs->vnet); 778 779 PF_RULES_WLOCK(); 780 pf_rollback_eth(rs->inactive.ticket, 781 rs->anchor ? rs->anchor->path : ""); 782 PF_RULES_WUNLOCK(); 783 784 CURVNET_RESTORE(); 785 } 786 787 static int 788 pf_rollback_eth(uint32_t ticket, const char *anchor) 789 { 790 struct pf_keth_rule *rule, *tmp; 791 struct pf_keth_ruleset *rs; 792 793 PF_RULES_WASSERT(); 794 795 rs = pf_find_keth_ruleset(anchor); 796 if (rs == NULL) 797 return (EINVAL); 798 799 if (!rs->inactive.open || 800 ticket != rs->inactive.ticket) 801 return (0); 802 803 /* Purge old inactive rules. */ 804 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 805 tmp) { 806 TAILQ_REMOVE(rs->inactive.rules, rule, entries); 807 pf_free_eth_rule(rule); 808 } 809 810 rs->inactive.open = 0; 811 812 pf_remove_if_empty_keth_ruleset(rs); 813 814 return (0); 815 } 816 817 #define PF_SET_SKIP_STEPS(i) \ 818 do { \ 819 while (head[i] != cur) { \ 820 head[i]->skip[i].ptr = cur; \ 821 head[i] = TAILQ_NEXT(head[i], entries); \ 822 } \ 823 } while (0) 824 825 static void 826 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules) 827 { 828 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT]; 829 int i; 830 831 cur = TAILQ_FIRST(rules); 832 prev = cur; 833 for (i = 0; i < PFE_SKIP_COUNT; ++i) 834 head[i] = cur; 835 while (cur != NULL) { 836 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 837 PF_SET_SKIP_STEPS(PFE_SKIP_IFP); 838 if (cur->direction != prev->direction) 839 PF_SET_SKIP_STEPS(PFE_SKIP_DIR); 840 if (cur->proto != prev->proto) 841 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO); 842 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0) 843 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR); 844 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0) 845 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR); 846 if (cur->ipsrc.neg != prev->ipsrc.neg || 847 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr)) 848 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR); 849 if (cur->ipdst.neg != prev->ipdst.neg || 850 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr)) 851 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR); 852 853 prev = cur; 854 cur = TAILQ_NEXT(cur, entries); 855 } 856 for (i = 0; i < PFE_SKIP_COUNT; ++i) 857 PF_SET_SKIP_STEPS(i); 858 } 859 860 static int 861 pf_commit_eth(uint32_t ticket, const char *anchor) 862 { 863 struct pf_keth_ruleq *rules; 864 struct pf_keth_ruleset *rs; 865 866 rs = pf_find_keth_ruleset(anchor); 867 if (rs == NULL) { 868 return (EINVAL); 869 } 870 871 if (!rs->inactive.open || 872 ticket != rs->inactive.ticket) 873 return (EBUSY); 874 875 PF_RULES_WASSERT(); 876 877 pf_eth_calc_skip_steps(rs->inactive.rules); 878 879 rules = rs->active.rules; 880 ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules); 881 rs->inactive.rules = rules; 882 rs->inactive.ticket = rs->active.ticket; 883 884 /* Clean up inactive rules (i.e. previously active rules), only when 885 * we're sure they're no longer used. */ 886 NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx); 887 888 return (0); 889 } 890 891 #ifdef ALTQ 892 static uint16_t 893 pf_qname2qid(const char *qname) 894 { 895 return (tagname2tag(&V_pf_qids, qname)); 896 } 897 898 static void 899 pf_qid_unref(uint16_t qid) 900 { 901 tag_unref(&V_pf_qids, qid); 902 } 903 904 static int 905 pf_begin_altq(u_int32_t *ticket) 906 { 907 struct pf_altq *altq, *tmp; 908 int error = 0; 909 910 PF_RULES_WASSERT(); 911 912 /* Purge the old altq lists */ 913 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 914 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 915 /* detach and destroy the discipline */ 916 error = altq_remove(altq); 917 } 918 free(altq, M_PFALTQ); 919 } 920 TAILQ_INIT(V_pf_altq_ifs_inactive); 921 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 922 pf_qid_unref(altq->qid); 923 free(altq, M_PFALTQ); 924 } 925 TAILQ_INIT(V_pf_altqs_inactive); 926 if (error) 927 return (error); 928 *ticket = ++V_ticket_altqs_inactive; 929 V_altqs_inactive_open = 1; 930 return (0); 931 } 932 933 static int 934 pf_rollback_altq(u_int32_t ticket) 935 { 936 struct pf_altq *altq, *tmp; 937 int error = 0; 938 939 PF_RULES_WASSERT(); 940 941 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 942 return (0); 943 /* Purge the old altq lists */ 944 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 945 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 946 /* detach and destroy the discipline */ 947 error = altq_remove(altq); 948 } 949 free(altq, M_PFALTQ); 950 } 951 TAILQ_INIT(V_pf_altq_ifs_inactive); 952 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 953 pf_qid_unref(altq->qid); 954 free(altq, M_PFALTQ); 955 } 956 TAILQ_INIT(V_pf_altqs_inactive); 957 V_altqs_inactive_open = 0; 958 return (error); 959 } 960 961 static int 962 pf_commit_altq(u_int32_t ticket) 963 { 964 struct pf_altqqueue *old_altqs, *old_altq_ifs; 965 struct pf_altq *altq, *tmp; 966 int err, error = 0; 967 968 PF_RULES_WASSERT(); 969 970 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 971 return (EBUSY); 972 973 /* swap altqs, keep the old. */ 974 old_altqs = V_pf_altqs_active; 975 old_altq_ifs = V_pf_altq_ifs_active; 976 V_pf_altqs_active = V_pf_altqs_inactive; 977 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive; 978 V_pf_altqs_inactive = old_altqs; 979 V_pf_altq_ifs_inactive = old_altq_ifs; 980 V_ticket_altqs_active = V_ticket_altqs_inactive; 981 982 /* Attach new disciplines */ 983 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 984 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 985 /* attach the discipline */ 986 error = altq_pfattach(altq); 987 if (error == 0 && V_pf_altq_running) 988 error = pf_enable_altq(altq); 989 if (error != 0) 990 return (error); 991 } 992 } 993 994 /* Purge the old altq lists */ 995 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 996 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 997 /* detach and destroy the discipline */ 998 if (V_pf_altq_running) 999 error = pf_disable_altq(altq); 1000 err = altq_pfdetach(altq); 1001 if (err != 0 && error == 0) 1002 error = err; 1003 err = altq_remove(altq); 1004 if (err != 0 && error == 0) 1005 error = err; 1006 } 1007 free(altq, M_PFALTQ); 1008 } 1009 TAILQ_INIT(V_pf_altq_ifs_inactive); 1010 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1011 pf_qid_unref(altq->qid); 1012 free(altq, M_PFALTQ); 1013 } 1014 TAILQ_INIT(V_pf_altqs_inactive); 1015 1016 V_altqs_inactive_open = 0; 1017 return (error); 1018 } 1019 1020 static int 1021 pf_enable_altq(struct pf_altq *altq) 1022 { 1023 struct ifnet *ifp; 1024 struct tb_profile tb; 1025 int error = 0; 1026 1027 if ((ifp = ifunit(altq->ifname)) == NULL) 1028 return (EINVAL); 1029 1030 if (ifp->if_snd.altq_type != ALTQT_NONE) 1031 error = altq_enable(&ifp->if_snd); 1032 1033 /* set tokenbucket regulator */ 1034 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { 1035 tb.rate = altq->ifbandwidth; 1036 tb.depth = altq->tbrsize; 1037 error = tbr_set(&ifp->if_snd, &tb); 1038 } 1039 1040 return (error); 1041 } 1042 1043 static int 1044 pf_disable_altq(struct pf_altq *altq) 1045 { 1046 struct ifnet *ifp; 1047 struct tb_profile tb; 1048 int error; 1049 1050 if ((ifp = ifunit(altq->ifname)) == NULL) 1051 return (EINVAL); 1052 1053 /* 1054 * when the discipline is no longer referenced, it was overridden 1055 * by a new one. if so, just return. 1056 */ 1057 if (altq->altq_disc != ifp->if_snd.altq_disc) 1058 return (0); 1059 1060 error = altq_disable(&ifp->if_snd); 1061 1062 if (error == 0) { 1063 /* clear tokenbucket regulator */ 1064 tb.rate = 0; 1065 error = tbr_set(&ifp->if_snd, &tb); 1066 } 1067 1068 return (error); 1069 } 1070 1071 static int 1072 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket, 1073 struct pf_altq *altq) 1074 { 1075 struct ifnet *ifp1; 1076 int error = 0; 1077 1078 /* Deactivate the interface in question */ 1079 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; 1080 if ((ifp1 = ifunit(altq->ifname)) == NULL || 1081 (remove && ifp1 == ifp)) { 1082 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 1083 } else { 1084 error = altq_add(ifp1, altq); 1085 1086 if (ticket != V_ticket_altqs_inactive) 1087 error = EBUSY; 1088 1089 if (error) 1090 free(altq, M_PFALTQ); 1091 } 1092 1093 return (error); 1094 } 1095 1096 void 1097 pf_altq_ifnet_event(struct ifnet *ifp, int remove) 1098 { 1099 struct pf_altq *a1, *a2, *a3; 1100 u_int32_t ticket; 1101 int error = 0; 1102 1103 /* 1104 * No need to re-evaluate the configuration for events on interfaces 1105 * that do not support ALTQ, as it's not possible for such 1106 * interfaces to be part of the configuration. 1107 */ 1108 if (!ALTQ_IS_READY(&ifp->if_snd)) 1109 return; 1110 1111 /* Interrupt userland queue modifications */ 1112 if (V_altqs_inactive_open) 1113 pf_rollback_altq(V_ticket_altqs_inactive); 1114 1115 /* Start new altq ruleset */ 1116 if (pf_begin_altq(&ticket)) 1117 return; 1118 1119 /* Copy the current active set */ 1120 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) { 1121 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1122 if (a2 == NULL) { 1123 error = ENOMEM; 1124 break; 1125 } 1126 bcopy(a1, a2, sizeof(struct pf_altq)); 1127 1128 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1129 if (error) 1130 break; 1131 1132 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries); 1133 } 1134 if (error) 1135 goto out; 1136 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { 1137 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1138 if (a2 == NULL) { 1139 error = ENOMEM; 1140 break; 1141 } 1142 bcopy(a1, a2, sizeof(struct pf_altq)); 1143 1144 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) { 1145 error = EBUSY; 1146 free(a2, M_PFALTQ); 1147 break; 1148 } 1149 a2->altq_disc = NULL; 1150 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) { 1151 if (strncmp(a3->ifname, a2->ifname, 1152 IFNAMSIZ) == 0) { 1153 a2->altq_disc = a3->altq_disc; 1154 break; 1155 } 1156 } 1157 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1158 if (error) 1159 break; 1160 1161 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); 1162 } 1163 1164 out: 1165 if (error != 0) 1166 pf_rollback_altq(ticket); 1167 else 1168 pf_commit_altq(ticket); 1169 } 1170 #endif /* ALTQ */ 1171 1172 static struct pf_krule_global * 1173 pf_rule_tree_alloc(int flags) 1174 { 1175 struct pf_krule_global *tree; 1176 1177 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags); 1178 if (tree == NULL) 1179 return (NULL); 1180 RB_INIT(tree); 1181 return (tree); 1182 } 1183 1184 static void 1185 pf_rule_tree_free(struct pf_krule_global *tree) 1186 { 1187 1188 free(tree, M_TEMP); 1189 } 1190 1191 static int 1192 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) 1193 { 1194 struct pf_krule_global *tree; 1195 struct pf_kruleset *rs; 1196 struct pf_krule *rule; 1197 1198 PF_RULES_WASSERT(); 1199 1200 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1201 return (EINVAL); 1202 tree = pf_rule_tree_alloc(M_NOWAIT); 1203 if (tree == NULL) 1204 return (ENOMEM); 1205 rs = pf_find_or_create_kruleset(anchor); 1206 if (rs == NULL) { 1207 free(tree, M_TEMP); 1208 return (EINVAL); 1209 } 1210 pf_rule_tree_free(rs->rules[rs_num].inactive.tree); 1211 rs->rules[rs_num].inactive.tree = tree; 1212 1213 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1214 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1215 rs->rules[rs_num].inactive.rcount--; 1216 } 1217 *ticket = ++rs->rules[rs_num].inactive.ticket; 1218 rs->rules[rs_num].inactive.open = 1; 1219 return (0); 1220 } 1221 1222 static int 1223 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) 1224 { 1225 struct pf_kruleset *rs; 1226 struct pf_krule *rule; 1227 1228 PF_RULES_WASSERT(); 1229 1230 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1231 return (EINVAL); 1232 rs = pf_find_kruleset(anchor); 1233 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1234 rs->rules[rs_num].inactive.ticket != ticket) 1235 return (0); 1236 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1237 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1238 rs->rules[rs_num].inactive.rcount--; 1239 } 1240 rs->rules[rs_num].inactive.open = 0; 1241 return (0); 1242 } 1243 1244 #define PF_MD5_UPD(st, elm) \ 1245 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) 1246 1247 #define PF_MD5_UPD_STR(st, elm) \ 1248 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) 1249 1250 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ 1251 (stor) = htonl((st)->elm); \ 1252 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ 1253 } while (0) 1254 1255 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ 1256 (stor) = htons((st)->elm); \ 1257 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ 1258 } while (0) 1259 1260 static void 1261 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) 1262 { 1263 PF_MD5_UPD(pfr, addr.type); 1264 switch (pfr->addr.type) { 1265 case PF_ADDR_DYNIFTL: 1266 PF_MD5_UPD(pfr, addr.v.ifname); 1267 PF_MD5_UPD(pfr, addr.iflags); 1268 break; 1269 case PF_ADDR_TABLE: 1270 PF_MD5_UPD(pfr, addr.v.tblname); 1271 break; 1272 case PF_ADDR_ADDRMASK: 1273 /* XXX ignore af? */ 1274 PF_MD5_UPD(pfr, addr.v.a.addr.addr32); 1275 PF_MD5_UPD(pfr, addr.v.a.mask.addr32); 1276 break; 1277 } 1278 1279 PF_MD5_UPD(pfr, port[0]); 1280 PF_MD5_UPD(pfr, port[1]); 1281 PF_MD5_UPD(pfr, neg); 1282 PF_MD5_UPD(pfr, port_op); 1283 } 1284 1285 static void 1286 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule) 1287 { 1288 u_int16_t x; 1289 u_int32_t y; 1290 1291 pf_hash_rule_addr(ctx, &rule->src); 1292 pf_hash_rule_addr(ctx, &rule->dst); 1293 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) 1294 PF_MD5_UPD_STR(rule, label[i]); 1295 PF_MD5_UPD_STR(rule, ifname); 1296 PF_MD5_UPD_STR(rule, match_tagname); 1297 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ 1298 PF_MD5_UPD_HTONL(rule, os_fingerprint, y); 1299 PF_MD5_UPD_HTONL(rule, prob, y); 1300 PF_MD5_UPD_HTONL(rule, uid.uid[0], y); 1301 PF_MD5_UPD_HTONL(rule, uid.uid[1], y); 1302 PF_MD5_UPD(rule, uid.op); 1303 PF_MD5_UPD_HTONL(rule, gid.gid[0], y); 1304 PF_MD5_UPD_HTONL(rule, gid.gid[1], y); 1305 PF_MD5_UPD(rule, gid.op); 1306 PF_MD5_UPD_HTONL(rule, rule_flag, y); 1307 PF_MD5_UPD(rule, action); 1308 PF_MD5_UPD(rule, direction); 1309 PF_MD5_UPD(rule, af); 1310 PF_MD5_UPD(rule, quick); 1311 PF_MD5_UPD(rule, ifnot); 1312 PF_MD5_UPD(rule, match_tag_not); 1313 PF_MD5_UPD(rule, natpass); 1314 PF_MD5_UPD(rule, keep_state); 1315 PF_MD5_UPD(rule, proto); 1316 PF_MD5_UPD(rule, type); 1317 PF_MD5_UPD(rule, code); 1318 PF_MD5_UPD(rule, flags); 1319 PF_MD5_UPD(rule, flagset); 1320 PF_MD5_UPD(rule, allow_opts); 1321 PF_MD5_UPD(rule, rt); 1322 PF_MD5_UPD(rule, tos); 1323 PF_MD5_UPD(rule, scrub_flags); 1324 PF_MD5_UPD(rule, min_ttl); 1325 PF_MD5_UPD(rule, set_tos); 1326 if (rule->anchor != NULL) 1327 PF_MD5_UPD_STR(rule, anchor->path); 1328 } 1329 1330 static void 1331 pf_hash_rule(struct pf_krule *rule) 1332 { 1333 MD5_CTX ctx; 1334 1335 MD5Init(&ctx); 1336 pf_hash_rule_rolling(&ctx, rule); 1337 MD5Final(rule->md5sum, &ctx); 1338 } 1339 1340 static int 1341 pf_krule_compare(struct pf_krule *a, struct pf_krule *b) 1342 { 1343 1344 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH)); 1345 } 1346 1347 static int 1348 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) 1349 { 1350 struct pf_kruleset *rs; 1351 struct pf_krule *rule, **old_array, *old_rule; 1352 struct pf_krulequeue *old_rules; 1353 struct pf_krule_global *old_tree; 1354 int error; 1355 u_int32_t old_rcount; 1356 1357 PF_RULES_WASSERT(); 1358 1359 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1360 return (EINVAL); 1361 rs = pf_find_kruleset(anchor); 1362 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1363 ticket != rs->rules[rs_num].inactive.ticket) 1364 return (EBUSY); 1365 1366 /* Calculate checksum for the main ruleset */ 1367 if (rs == &pf_main_ruleset) { 1368 error = pf_setup_pfsync_matching(rs); 1369 if (error != 0) 1370 return (error); 1371 } 1372 1373 /* Swap rules, keep the old. */ 1374 old_rules = rs->rules[rs_num].active.ptr; 1375 old_rcount = rs->rules[rs_num].active.rcount; 1376 old_array = rs->rules[rs_num].active.ptr_array; 1377 old_tree = rs->rules[rs_num].active.tree; 1378 1379 rs->rules[rs_num].active.ptr = 1380 rs->rules[rs_num].inactive.ptr; 1381 rs->rules[rs_num].active.ptr_array = 1382 rs->rules[rs_num].inactive.ptr_array; 1383 rs->rules[rs_num].active.tree = 1384 rs->rules[rs_num].inactive.tree; 1385 rs->rules[rs_num].active.rcount = 1386 rs->rules[rs_num].inactive.rcount; 1387 1388 /* Attempt to preserve counter information. */ 1389 if (V_pf_status.keep_counters && old_tree != NULL) { 1390 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr, 1391 entries) { 1392 old_rule = RB_FIND(pf_krule_global, old_tree, rule); 1393 if (old_rule == NULL) { 1394 continue; 1395 } 1396 pf_counter_u64_critical_enter(); 1397 pf_counter_u64_add_protected(&rule->evaluations, 1398 pf_counter_u64_fetch(&old_rule->evaluations)); 1399 pf_counter_u64_add_protected(&rule->packets[0], 1400 pf_counter_u64_fetch(&old_rule->packets[0])); 1401 pf_counter_u64_add_protected(&rule->packets[1], 1402 pf_counter_u64_fetch(&old_rule->packets[1])); 1403 pf_counter_u64_add_protected(&rule->bytes[0], 1404 pf_counter_u64_fetch(&old_rule->bytes[0])); 1405 pf_counter_u64_add_protected(&rule->bytes[1], 1406 pf_counter_u64_fetch(&old_rule->bytes[1])); 1407 pf_counter_u64_critical_exit(); 1408 } 1409 } 1410 1411 rs->rules[rs_num].inactive.ptr = old_rules; 1412 rs->rules[rs_num].inactive.ptr_array = old_array; 1413 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ 1414 rs->rules[rs_num].inactive.rcount = old_rcount; 1415 1416 rs->rules[rs_num].active.ticket = 1417 rs->rules[rs_num].inactive.ticket; 1418 pf_calc_skip_steps(rs->rules[rs_num].active.ptr); 1419 1420 /* Purge the old rule list. */ 1421 PF_UNLNKDRULES_LOCK(); 1422 while ((rule = TAILQ_FIRST(old_rules)) != NULL) 1423 pf_unlink_rule_locked(old_rules, rule); 1424 PF_UNLNKDRULES_UNLOCK(); 1425 if (rs->rules[rs_num].inactive.ptr_array) 1426 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP); 1427 rs->rules[rs_num].inactive.ptr_array = NULL; 1428 rs->rules[rs_num].inactive.rcount = 0; 1429 rs->rules[rs_num].inactive.open = 0; 1430 pf_remove_if_empty_kruleset(rs); 1431 free(old_tree, M_TEMP); 1432 1433 return (0); 1434 } 1435 1436 static int 1437 pf_setup_pfsync_matching(struct pf_kruleset *rs) 1438 { 1439 MD5_CTX ctx; 1440 struct pf_krule *rule; 1441 int rs_cnt; 1442 u_int8_t digest[PF_MD5_DIGEST_LENGTH]; 1443 1444 MD5Init(&ctx); 1445 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { 1446 /* XXX PF_RULESET_SCRUB as well? */ 1447 if (rs_cnt == PF_RULESET_SCRUB) 1448 continue; 1449 1450 if (rs->rules[rs_cnt].inactive.ptr_array) 1451 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP); 1452 rs->rules[rs_cnt].inactive.ptr_array = NULL; 1453 1454 if (rs->rules[rs_cnt].inactive.rcount) { 1455 rs->rules[rs_cnt].inactive.ptr_array = 1456 mallocarray(rs->rules[rs_cnt].inactive.rcount, 1457 sizeof(struct pf_rule **), 1458 M_TEMP, M_NOWAIT); 1459 1460 if (!rs->rules[rs_cnt].inactive.ptr_array) 1461 return (ENOMEM); 1462 } 1463 1464 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, 1465 entries) { 1466 pf_hash_rule_rolling(&ctx, rule); 1467 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule; 1468 } 1469 } 1470 1471 MD5Final(digest, &ctx); 1472 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); 1473 return (0); 1474 } 1475 1476 static int 1477 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr) 1478 { 1479 int error = 0; 1480 1481 switch (addr->type) { 1482 case PF_ADDR_TABLE: 1483 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname); 1484 if (addr->p.tbl == NULL) 1485 error = ENOMEM; 1486 break; 1487 default: 1488 error = EINVAL; 1489 } 1490 1491 return (error); 1492 } 1493 1494 static int 1495 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr, 1496 sa_family_t af) 1497 { 1498 int error = 0; 1499 1500 switch (addr->type) { 1501 case PF_ADDR_TABLE: 1502 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); 1503 if (addr->p.tbl == NULL) 1504 error = ENOMEM; 1505 break; 1506 case PF_ADDR_DYNIFTL: 1507 error = pfi_dynaddr_setup(addr, af); 1508 break; 1509 } 1510 1511 return (error); 1512 } 1513 1514 static void 1515 pf_addr_copyout(struct pf_addr_wrap *addr) 1516 { 1517 1518 switch (addr->type) { 1519 case PF_ADDR_DYNIFTL: 1520 pfi_dynaddr_copyout(addr); 1521 break; 1522 case PF_ADDR_TABLE: 1523 pf_tbladdr_copyout(addr); 1524 break; 1525 } 1526 } 1527 1528 static void 1529 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out) 1530 { 1531 int secs = time_uptime, diff; 1532 1533 bzero(out, sizeof(struct pf_src_node)); 1534 1535 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr)); 1536 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr)); 1537 1538 if (in->rule.ptr != NULL) 1539 out->rule.nr = in->rule.ptr->nr; 1540 1541 for (int i = 0; i < 2; i++) { 1542 out->bytes[i] = counter_u64_fetch(in->bytes[i]); 1543 out->packets[i] = counter_u64_fetch(in->packets[i]); 1544 } 1545 1546 out->states = in->states; 1547 out->conn = in->conn; 1548 out->af = in->af; 1549 out->ruletype = in->ruletype; 1550 1551 out->creation = secs - in->creation; 1552 if (out->expire > secs) 1553 out->expire -= secs; 1554 else 1555 out->expire = 0; 1556 1557 /* Adjust the connection rate estimate. */ 1558 diff = secs - in->conn_rate.last; 1559 if (diff >= in->conn_rate.seconds) 1560 out->conn_rate.count = 0; 1561 else 1562 out->conn_rate.count -= 1563 in->conn_rate.count * diff / 1564 in->conn_rate.seconds; 1565 } 1566 1567 #ifdef ALTQ 1568 /* 1569 * Handle export of struct pf_kaltq to user binaries that may be using any 1570 * version of struct pf_altq. 1571 */ 1572 static int 1573 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size) 1574 { 1575 u_int32_t version; 1576 1577 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1578 version = 0; 1579 else 1580 version = pa->version; 1581 1582 if (version > PFIOC_ALTQ_VERSION) 1583 return (EINVAL); 1584 1585 #define ASSIGN(x) exported_q->x = q->x 1586 #define COPY(x) \ 1587 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x))) 1588 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX) 1589 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX) 1590 1591 switch (version) { 1592 case 0: { 1593 struct pf_altq_v0 *exported_q = 1594 &((struct pfioc_altq_v0 *)pa)->altq; 1595 1596 COPY(ifname); 1597 1598 ASSIGN(scheduler); 1599 ASSIGN(tbrsize); 1600 exported_q->tbrsize = SATU16(q->tbrsize); 1601 exported_q->ifbandwidth = SATU32(q->ifbandwidth); 1602 1603 COPY(qname); 1604 COPY(parent); 1605 ASSIGN(parent_qid); 1606 exported_q->bandwidth = SATU32(q->bandwidth); 1607 ASSIGN(priority); 1608 ASSIGN(local_flags); 1609 1610 ASSIGN(qlimit); 1611 ASSIGN(flags); 1612 1613 if (q->scheduler == ALTQT_HFSC) { 1614 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x 1615 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \ 1616 SATU32(q->pq_u.hfsc_opts.x) 1617 1618 ASSIGN_OPT_SATU32(rtsc_m1); 1619 ASSIGN_OPT(rtsc_d); 1620 ASSIGN_OPT_SATU32(rtsc_m2); 1621 1622 ASSIGN_OPT_SATU32(lssc_m1); 1623 ASSIGN_OPT(lssc_d); 1624 ASSIGN_OPT_SATU32(lssc_m2); 1625 1626 ASSIGN_OPT_SATU32(ulsc_m1); 1627 ASSIGN_OPT(ulsc_d); 1628 ASSIGN_OPT_SATU32(ulsc_m2); 1629 1630 ASSIGN_OPT(flags); 1631 1632 #undef ASSIGN_OPT 1633 #undef ASSIGN_OPT_SATU32 1634 } else 1635 COPY(pq_u); 1636 1637 ASSIGN(qid); 1638 break; 1639 } 1640 case 1: { 1641 struct pf_altq_v1 *exported_q = 1642 &((struct pfioc_altq_v1 *)pa)->altq; 1643 1644 COPY(ifname); 1645 1646 ASSIGN(scheduler); 1647 ASSIGN(tbrsize); 1648 ASSIGN(ifbandwidth); 1649 1650 COPY(qname); 1651 COPY(parent); 1652 ASSIGN(parent_qid); 1653 ASSIGN(bandwidth); 1654 ASSIGN(priority); 1655 ASSIGN(local_flags); 1656 1657 ASSIGN(qlimit); 1658 ASSIGN(flags); 1659 COPY(pq_u); 1660 1661 ASSIGN(qid); 1662 break; 1663 } 1664 default: 1665 panic("%s: unhandled struct pfioc_altq version", __func__); 1666 break; 1667 } 1668 1669 #undef ASSIGN 1670 #undef COPY 1671 #undef SATU16 1672 #undef SATU32 1673 1674 return (0); 1675 } 1676 1677 /* 1678 * Handle import to struct pf_kaltq of struct pf_altq from user binaries 1679 * that may be using any version of it. 1680 */ 1681 static int 1682 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size) 1683 { 1684 u_int32_t version; 1685 1686 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1687 version = 0; 1688 else 1689 version = pa->version; 1690 1691 if (version > PFIOC_ALTQ_VERSION) 1692 return (EINVAL); 1693 1694 #define ASSIGN(x) q->x = imported_q->x 1695 #define COPY(x) \ 1696 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x))) 1697 1698 switch (version) { 1699 case 0: { 1700 struct pf_altq_v0 *imported_q = 1701 &((struct pfioc_altq_v0 *)pa)->altq; 1702 1703 COPY(ifname); 1704 1705 ASSIGN(scheduler); 1706 ASSIGN(tbrsize); /* 16-bit -> 32-bit */ 1707 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */ 1708 1709 COPY(qname); 1710 COPY(parent); 1711 ASSIGN(parent_qid); 1712 ASSIGN(bandwidth); /* 32-bit -> 64-bit */ 1713 ASSIGN(priority); 1714 ASSIGN(local_flags); 1715 1716 ASSIGN(qlimit); 1717 ASSIGN(flags); 1718 1719 if (imported_q->scheduler == ALTQT_HFSC) { 1720 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x 1721 1722 /* 1723 * The m1 and m2 parameters are being copied from 1724 * 32-bit to 64-bit. 1725 */ 1726 ASSIGN_OPT(rtsc_m1); 1727 ASSIGN_OPT(rtsc_d); 1728 ASSIGN_OPT(rtsc_m2); 1729 1730 ASSIGN_OPT(lssc_m1); 1731 ASSIGN_OPT(lssc_d); 1732 ASSIGN_OPT(lssc_m2); 1733 1734 ASSIGN_OPT(ulsc_m1); 1735 ASSIGN_OPT(ulsc_d); 1736 ASSIGN_OPT(ulsc_m2); 1737 1738 ASSIGN_OPT(flags); 1739 1740 #undef ASSIGN_OPT 1741 } else 1742 COPY(pq_u); 1743 1744 ASSIGN(qid); 1745 break; 1746 } 1747 case 1: { 1748 struct pf_altq_v1 *imported_q = 1749 &((struct pfioc_altq_v1 *)pa)->altq; 1750 1751 COPY(ifname); 1752 1753 ASSIGN(scheduler); 1754 ASSIGN(tbrsize); 1755 ASSIGN(ifbandwidth); 1756 1757 COPY(qname); 1758 COPY(parent); 1759 ASSIGN(parent_qid); 1760 ASSIGN(bandwidth); 1761 ASSIGN(priority); 1762 ASSIGN(local_flags); 1763 1764 ASSIGN(qlimit); 1765 ASSIGN(flags); 1766 COPY(pq_u); 1767 1768 ASSIGN(qid); 1769 break; 1770 } 1771 default: 1772 panic("%s: unhandled struct pfioc_altq version", __func__); 1773 break; 1774 } 1775 1776 #undef ASSIGN 1777 #undef COPY 1778 1779 return (0); 1780 } 1781 1782 static struct pf_altq * 1783 pf_altq_get_nth_active(u_int32_t n) 1784 { 1785 struct pf_altq *altq; 1786 u_int32_t nr; 1787 1788 nr = 0; 1789 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1790 if (nr == n) 1791 return (altq); 1792 nr++; 1793 } 1794 1795 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { 1796 if (nr == n) 1797 return (altq); 1798 nr++; 1799 } 1800 1801 return (NULL); 1802 } 1803 #endif /* ALTQ */ 1804 1805 struct pf_krule * 1806 pf_krule_alloc(void) 1807 { 1808 struct pf_krule *rule; 1809 1810 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO); 1811 mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF); 1812 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 1813 M_WAITOK | M_ZERO); 1814 return (rule); 1815 } 1816 1817 void 1818 pf_krule_free(struct pf_krule *rule) 1819 { 1820 #ifdef PF_WANT_32_TO_64_COUNTER 1821 bool wowned; 1822 #endif 1823 1824 if (rule == NULL) 1825 return; 1826 1827 #ifdef PF_WANT_32_TO_64_COUNTER 1828 if (rule->allrulelinked) { 1829 wowned = PF_RULES_WOWNED(); 1830 if (!wowned) 1831 PF_RULES_WLOCK(); 1832 LIST_REMOVE(rule, allrulelist); 1833 V_pf_allrulecount--; 1834 if (!wowned) 1835 PF_RULES_WUNLOCK(); 1836 } 1837 #endif 1838 1839 pf_counter_u64_deinit(&rule->evaluations); 1840 for (int i = 0; i < 2; i++) { 1841 pf_counter_u64_deinit(&rule->packets[i]); 1842 pf_counter_u64_deinit(&rule->bytes[i]); 1843 } 1844 counter_u64_free(rule->states_cur); 1845 counter_u64_free(rule->states_tot); 1846 counter_u64_free(rule->src_nodes); 1847 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 1848 1849 mtx_destroy(&rule->rpool.mtx); 1850 free(rule, M_PFRULE); 1851 } 1852 1853 static void 1854 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool, 1855 struct pf_pooladdr *pool) 1856 { 1857 1858 bzero(pool, sizeof(*pool)); 1859 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr)); 1860 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname)); 1861 } 1862 1863 static int 1864 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool, 1865 struct pf_kpooladdr *kpool) 1866 { 1867 int ret; 1868 1869 bzero(kpool, sizeof(*kpool)); 1870 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr)); 1871 ret = pf_user_strcpy(kpool->ifname, pool->ifname, 1872 sizeof(kpool->ifname)); 1873 return (ret); 1874 } 1875 1876 static void 1877 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool) 1878 { 1879 bzero(pool, sizeof(*pool)); 1880 1881 bcopy(&kpool->key, &pool->key, sizeof(pool->key)); 1882 bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter)); 1883 1884 pool->tblidx = kpool->tblidx; 1885 pool->proxy_port[0] = kpool->proxy_port[0]; 1886 pool->proxy_port[1] = kpool->proxy_port[1]; 1887 pool->opts = kpool->opts; 1888 } 1889 1890 static void 1891 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool) 1892 { 1893 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), ""); 1894 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), ""); 1895 1896 bcopy(&pool->key, &kpool->key, sizeof(kpool->key)); 1897 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter)); 1898 1899 kpool->tblidx = pool->tblidx; 1900 kpool->proxy_port[0] = pool->proxy_port[0]; 1901 kpool->proxy_port[1] = pool->proxy_port[1]; 1902 kpool->opts = pool->opts; 1903 } 1904 1905 static void 1906 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule) 1907 { 1908 1909 bzero(rule, sizeof(*rule)); 1910 1911 bcopy(&krule->src, &rule->src, sizeof(rule->src)); 1912 bcopy(&krule->dst, &rule->dst, sizeof(rule->dst)); 1913 1914 for (int i = 0; i < PF_SKIP_COUNT; ++i) { 1915 if (rule->skip[i].ptr == NULL) 1916 rule->skip[i].nr = -1; 1917 else 1918 rule->skip[i].nr = krule->skip[i].ptr->nr; 1919 } 1920 1921 strlcpy(rule->label, krule->label[0], sizeof(rule->label)); 1922 strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname)); 1923 strlcpy(rule->qname, krule->qname, sizeof(rule->qname)); 1924 strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname)); 1925 strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname)); 1926 strlcpy(rule->match_tagname, krule->match_tagname, 1927 sizeof(rule->match_tagname)); 1928 strlcpy(rule->overload_tblname, krule->overload_tblname, 1929 sizeof(rule->overload_tblname)); 1930 1931 pf_kpool_to_pool(&krule->rpool, &rule->rpool); 1932 1933 rule->evaluations = pf_counter_u64_fetch(&krule->evaluations); 1934 for (int i = 0; i < 2; i++) { 1935 rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]); 1936 rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]); 1937 } 1938 1939 /* kif, anchor, overload_tbl are not copied over. */ 1940 1941 rule->os_fingerprint = krule->os_fingerprint; 1942 1943 rule->rtableid = krule->rtableid; 1944 bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout)); 1945 rule->max_states = krule->max_states; 1946 rule->max_src_nodes = krule->max_src_nodes; 1947 rule->max_src_states = krule->max_src_states; 1948 rule->max_src_conn = krule->max_src_conn; 1949 rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit; 1950 rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds; 1951 rule->qid = krule->qid; 1952 rule->pqid = krule->pqid; 1953 rule->nr = krule->nr; 1954 rule->prob = krule->prob; 1955 rule->cuid = krule->cuid; 1956 rule->cpid = krule->cpid; 1957 1958 rule->return_icmp = krule->return_icmp; 1959 rule->return_icmp6 = krule->return_icmp6; 1960 rule->max_mss = krule->max_mss; 1961 rule->tag = krule->tag; 1962 rule->match_tag = krule->match_tag; 1963 rule->scrub_flags = krule->scrub_flags; 1964 1965 bcopy(&krule->uid, &rule->uid, sizeof(krule->uid)); 1966 bcopy(&krule->gid, &rule->gid, sizeof(krule->gid)); 1967 1968 rule->rule_flag = krule->rule_flag; 1969 rule->action = krule->action; 1970 rule->direction = krule->direction; 1971 rule->log = krule->log; 1972 rule->logif = krule->logif; 1973 rule->quick = krule->quick; 1974 rule->ifnot = krule->ifnot; 1975 rule->match_tag_not = krule->match_tag_not; 1976 rule->natpass = krule->natpass; 1977 1978 rule->keep_state = krule->keep_state; 1979 rule->af = krule->af; 1980 rule->proto = krule->proto; 1981 rule->type = krule->type; 1982 rule->code = krule->code; 1983 rule->flags = krule->flags; 1984 rule->flagset = krule->flagset; 1985 rule->min_ttl = krule->min_ttl; 1986 rule->allow_opts = krule->allow_opts; 1987 rule->rt = krule->rt; 1988 rule->return_ttl = krule->return_ttl; 1989 rule->tos = krule->tos; 1990 rule->set_tos = krule->set_tos; 1991 rule->anchor_relative = krule->anchor_relative; 1992 rule->anchor_wildcard = krule->anchor_wildcard; 1993 1994 rule->flush = krule->flush; 1995 rule->prio = krule->prio; 1996 rule->set_prio[0] = krule->set_prio[0]; 1997 rule->set_prio[1] = krule->set_prio[1]; 1998 1999 bcopy(&krule->divert, &rule->divert, sizeof(krule->divert)); 2000 2001 rule->u_states_cur = counter_u64_fetch(krule->states_cur); 2002 rule->u_states_tot = counter_u64_fetch(krule->states_tot); 2003 rule->u_src_nodes = counter_u64_fetch(krule->src_nodes); 2004 } 2005 2006 static int 2007 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule) 2008 { 2009 int ret; 2010 2011 #ifndef INET 2012 if (rule->af == AF_INET) { 2013 return (EAFNOSUPPORT); 2014 } 2015 #endif /* INET */ 2016 #ifndef INET6 2017 if (rule->af == AF_INET6) { 2018 return (EAFNOSUPPORT); 2019 } 2020 #endif /* INET6 */ 2021 2022 ret = pf_check_rule_addr(&rule->src); 2023 if (ret != 0) 2024 return (ret); 2025 ret = pf_check_rule_addr(&rule->dst); 2026 if (ret != 0) 2027 return (ret); 2028 2029 bcopy(&rule->src, &krule->src, sizeof(rule->src)); 2030 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst)); 2031 2032 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label)); 2033 if (ret != 0) 2034 return (ret); 2035 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname)); 2036 if (ret != 0) 2037 return (ret); 2038 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname)); 2039 if (ret != 0) 2040 return (ret); 2041 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname)); 2042 if (ret != 0) 2043 return (ret); 2044 ret = pf_user_strcpy(krule->tagname, rule->tagname, 2045 sizeof(rule->tagname)); 2046 if (ret != 0) 2047 return (ret); 2048 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname, 2049 sizeof(rule->match_tagname)); 2050 if (ret != 0) 2051 return (ret); 2052 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname, 2053 sizeof(rule->overload_tblname)); 2054 if (ret != 0) 2055 return (ret); 2056 2057 pf_pool_to_kpool(&rule->rpool, &krule->rpool); 2058 2059 /* Don't allow userspace to set evaluations, packets or bytes. */ 2060 /* kif, anchor, overload_tbl are not copied over. */ 2061 2062 krule->os_fingerprint = rule->os_fingerprint; 2063 2064 krule->rtableid = rule->rtableid; 2065 bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout)); 2066 krule->max_states = rule->max_states; 2067 krule->max_src_nodes = rule->max_src_nodes; 2068 krule->max_src_states = rule->max_src_states; 2069 krule->max_src_conn = rule->max_src_conn; 2070 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit; 2071 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds; 2072 krule->qid = rule->qid; 2073 krule->pqid = rule->pqid; 2074 krule->nr = rule->nr; 2075 krule->prob = rule->prob; 2076 krule->cuid = rule->cuid; 2077 krule->cpid = rule->cpid; 2078 2079 krule->return_icmp = rule->return_icmp; 2080 krule->return_icmp6 = rule->return_icmp6; 2081 krule->max_mss = rule->max_mss; 2082 krule->tag = rule->tag; 2083 krule->match_tag = rule->match_tag; 2084 krule->scrub_flags = rule->scrub_flags; 2085 2086 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid)); 2087 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid)); 2088 2089 krule->rule_flag = rule->rule_flag; 2090 krule->action = rule->action; 2091 krule->direction = rule->direction; 2092 krule->log = rule->log; 2093 krule->logif = rule->logif; 2094 krule->quick = rule->quick; 2095 krule->ifnot = rule->ifnot; 2096 krule->match_tag_not = rule->match_tag_not; 2097 krule->natpass = rule->natpass; 2098 2099 krule->keep_state = rule->keep_state; 2100 krule->af = rule->af; 2101 krule->proto = rule->proto; 2102 krule->type = rule->type; 2103 krule->code = rule->code; 2104 krule->flags = rule->flags; 2105 krule->flagset = rule->flagset; 2106 krule->min_ttl = rule->min_ttl; 2107 krule->allow_opts = rule->allow_opts; 2108 krule->rt = rule->rt; 2109 krule->return_ttl = rule->return_ttl; 2110 krule->tos = rule->tos; 2111 krule->set_tos = rule->set_tos; 2112 2113 krule->flush = rule->flush; 2114 krule->prio = rule->prio; 2115 krule->set_prio[0] = rule->set_prio[0]; 2116 krule->set_prio[1] = rule->set_prio[1]; 2117 2118 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert)); 2119 2120 return (0); 2121 } 2122 2123 static int 2124 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk, 2125 struct pf_kstate_kill *kill) 2126 { 2127 int ret; 2128 2129 bzero(kill, sizeof(*kill)); 2130 2131 bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp)); 2132 kill->psk_af = psk->psk_af; 2133 kill->psk_proto = psk->psk_proto; 2134 bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src)); 2135 bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst)); 2136 ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname, 2137 sizeof(kill->psk_ifname)); 2138 if (ret != 0) 2139 return (ret); 2140 ret = pf_user_strcpy(kill->psk_label, psk->psk_label, 2141 sizeof(kill->psk_label)); 2142 if (ret != 0) 2143 return (ret); 2144 2145 return (0); 2146 } 2147 2148 static int 2149 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, 2150 uint32_t pool_ticket, const char *anchor, const char *anchor_call, 2151 struct thread *td) 2152 { 2153 struct pf_kruleset *ruleset; 2154 struct pf_krule *tail; 2155 struct pf_kpooladdr *pa; 2156 struct pfi_kkif *kif = NULL; 2157 int rs_num; 2158 int error = 0; 2159 2160 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) { 2161 error = EINVAL; 2162 goto errout_unlocked; 2163 } 2164 2165 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 2166 2167 if (rule->ifname[0]) 2168 kif = pf_kkif_create(M_WAITOK); 2169 pf_counter_u64_init(&rule->evaluations, M_WAITOK); 2170 for (int i = 0; i < 2; i++) { 2171 pf_counter_u64_init(&rule->packets[i], M_WAITOK); 2172 pf_counter_u64_init(&rule->bytes[i], M_WAITOK); 2173 } 2174 rule->states_cur = counter_u64_alloc(M_WAITOK); 2175 rule->states_tot = counter_u64_alloc(M_WAITOK); 2176 rule->src_nodes = counter_u64_alloc(M_WAITOK); 2177 rule->cuid = td->td_ucred->cr_ruid; 2178 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 2179 TAILQ_INIT(&rule->rpool.list); 2180 2181 PF_CONFIG_LOCK(); 2182 PF_RULES_WLOCK(); 2183 #ifdef PF_WANT_32_TO_64_COUNTER 2184 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist); 2185 MPASS(!rule->allrulelinked); 2186 rule->allrulelinked = true; 2187 V_pf_allrulecount++; 2188 #endif 2189 ruleset = pf_find_kruleset(anchor); 2190 if (ruleset == NULL) 2191 ERROUT(EINVAL); 2192 rs_num = pf_get_ruleset_number(rule->action); 2193 if (rs_num >= PF_RULESET_MAX) 2194 ERROUT(EINVAL); 2195 if (ticket != ruleset->rules[rs_num].inactive.ticket) { 2196 DPFPRINTF(PF_DEBUG_MISC, 2197 ("ticket: %d != [%d]%d\n", ticket, rs_num, 2198 ruleset->rules[rs_num].inactive.ticket)); 2199 ERROUT(EBUSY); 2200 } 2201 if (pool_ticket != V_ticket_pabuf) { 2202 DPFPRINTF(PF_DEBUG_MISC, 2203 ("pool_ticket: %d != %d\n", pool_ticket, 2204 V_ticket_pabuf)); 2205 ERROUT(EBUSY); 2206 } 2207 /* 2208 * XXXMJG hack: there is no mechanism to ensure they started the 2209 * transaction. Ticket checked above may happen to match by accident, 2210 * even if nobody called DIOCXBEGIN, let alone this process. 2211 * Partially work around it by checking if the RB tree got allocated, 2212 * see pf_begin_rules. 2213 */ 2214 if (ruleset->rules[rs_num].inactive.tree == NULL) { 2215 ERROUT(EINVAL); 2216 } 2217 2218 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 2219 pf_krulequeue); 2220 if (tail) 2221 rule->nr = tail->nr + 1; 2222 else 2223 rule->nr = 0; 2224 if (rule->ifname[0]) { 2225 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2226 kif = NULL; 2227 pfi_kkif_ref(rule->kif); 2228 } else 2229 rule->kif = NULL; 2230 2231 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) 2232 error = EBUSY; 2233 2234 #ifdef ALTQ 2235 /* set queue IDs */ 2236 if (rule->qname[0] != 0) { 2237 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2238 error = EBUSY; 2239 else if (rule->pqname[0] != 0) { 2240 if ((rule->pqid = 2241 pf_qname2qid(rule->pqname)) == 0) 2242 error = EBUSY; 2243 } else 2244 rule->pqid = rule->qid; 2245 } 2246 #endif 2247 if (rule->tagname[0]) 2248 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2249 error = EBUSY; 2250 if (rule->match_tagname[0]) 2251 if ((rule->match_tag = 2252 pf_tagname2tag(rule->match_tagname)) == 0) 2253 error = EBUSY; 2254 if (rule->rt && !rule->direction) 2255 error = EINVAL; 2256 if (!rule->log) 2257 rule->logif = 0; 2258 if (rule->logif >= PFLOGIFS_MAX) 2259 error = EINVAL; 2260 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) 2261 error = ENOMEM; 2262 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) 2263 error = ENOMEM; 2264 if (pf_kanchor_setup(rule, ruleset, anchor_call)) 2265 error = EINVAL; 2266 if (rule->scrub_flags & PFSTATE_SETPRIO && 2267 (rule->set_prio[0] > PF_PRIO_MAX || 2268 rule->set_prio[1] > PF_PRIO_MAX)) 2269 error = EINVAL; 2270 TAILQ_FOREACH(pa, &V_pf_pabuf, entries) 2271 if (pa->addr.type == PF_ADDR_TABLE) { 2272 pa->addr.p.tbl = pfr_attach_table(ruleset, 2273 pa->addr.v.tblname); 2274 if (pa->addr.p.tbl == NULL) 2275 error = ENOMEM; 2276 } 2277 2278 rule->overload_tbl = NULL; 2279 if (rule->overload_tblname[0]) { 2280 if ((rule->overload_tbl = pfr_attach_table(ruleset, 2281 rule->overload_tblname)) == NULL) 2282 error = EINVAL; 2283 else 2284 rule->overload_tbl->pfrkt_flags |= 2285 PFR_TFLAG_ACTIVE; 2286 } 2287 2288 pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list); 2289 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) || 2290 (rule->action == PF_BINAT)) && rule->anchor == NULL) || 2291 (rule->rt > PF_NOPFROUTE)) && 2292 (TAILQ_FIRST(&rule->rpool.list) == NULL)) 2293 error = EINVAL; 2294 2295 if (error) { 2296 pf_free_rule(rule); 2297 rule = NULL; 2298 ERROUT(error); 2299 } 2300 2301 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list); 2302 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, 2303 rule, entries); 2304 ruleset->rules[rs_num].inactive.rcount++; 2305 2306 PF_RULES_WUNLOCK(); 2307 pf_hash_rule(rule); 2308 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) { 2309 PF_RULES_WLOCK(); 2310 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries); 2311 ruleset->rules[rs_num].inactive.rcount--; 2312 pf_free_rule(rule); 2313 rule = NULL; 2314 ERROUT(EEXIST); 2315 } 2316 PF_CONFIG_UNLOCK(); 2317 2318 return (0); 2319 2320 #undef ERROUT 2321 errout: 2322 PF_RULES_WUNLOCK(); 2323 PF_CONFIG_UNLOCK(); 2324 errout_unlocked: 2325 pf_kkif_free(kif); 2326 pf_krule_free(rule); 2327 return (error); 2328 } 2329 2330 static bool 2331 pf_label_match(const struct pf_krule *rule, const char *label) 2332 { 2333 int i = 0; 2334 2335 while (*rule->label[i]) { 2336 if (strcmp(rule->label[i], label) == 0) 2337 return (true); 2338 i++; 2339 } 2340 2341 return (false); 2342 } 2343 2344 static unsigned int 2345 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir) 2346 { 2347 struct pf_kstate *s; 2348 int more = 0; 2349 2350 s = pf_find_state_all(key, dir, &more); 2351 if (s == NULL) 2352 return (0); 2353 2354 if (more) { 2355 PF_STATE_UNLOCK(s); 2356 return (0); 2357 } 2358 2359 pf_unlink_state(s); 2360 return (1); 2361 } 2362 2363 static int 2364 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih) 2365 { 2366 struct pf_kstate *s; 2367 struct pf_state_key *sk; 2368 struct pf_addr *srcaddr, *dstaddr; 2369 struct pf_state_key_cmp match_key; 2370 int idx, killed = 0; 2371 unsigned int dir; 2372 u_int16_t srcport, dstport; 2373 struct pfi_kkif *kif; 2374 2375 relock_DIOCKILLSTATES: 2376 PF_HASHROW_LOCK(ih); 2377 LIST_FOREACH(s, &ih->states, entry) { 2378 /* For floating states look at the original kif. */ 2379 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 2380 2381 sk = s->key[PF_SK_WIRE]; 2382 if (s->direction == PF_OUT) { 2383 srcaddr = &sk->addr[1]; 2384 dstaddr = &sk->addr[0]; 2385 srcport = sk->port[1]; 2386 dstport = sk->port[0]; 2387 } else { 2388 srcaddr = &sk->addr[0]; 2389 dstaddr = &sk->addr[1]; 2390 srcport = sk->port[0]; 2391 dstport = sk->port[1]; 2392 } 2393 2394 if (psk->psk_af && sk->af != psk->psk_af) 2395 continue; 2396 2397 if (psk->psk_proto && psk->psk_proto != sk->proto) 2398 continue; 2399 2400 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr, 2401 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af)) 2402 continue; 2403 2404 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr, 2405 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af)) 2406 continue; 2407 2408 if (! PF_MATCHA(psk->psk_rt_addr.neg, 2409 &psk->psk_rt_addr.addr.v.a.addr, 2410 &psk->psk_rt_addr.addr.v.a.mask, 2411 &s->rt_addr, sk->af)) 2412 continue; 2413 2414 if (psk->psk_src.port_op != 0 && 2415 ! pf_match_port(psk->psk_src.port_op, 2416 psk->psk_src.port[0], psk->psk_src.port[1], srcport)) 2417 continue; 2418 2419 if (psk->psk_dst.port_op != 0 && 2420 ! pf_match_port(psk->psk_dst.port_op, 2421 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) 2422 continue; 2423 2424 if (psk->psk_label[0] && 2425 ! pf_label_match(s->rule.ptr, psk->psk_label)) 2426 continue; 2427 2428 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname, 2429 kif->pfik_name)) 2430 continue; 2431 2432 if (psk->psk_kill_match) { 2433 /* Create the key to find matching states, with lock 2434 * held. */ 2435 2436 bzero(&match_key, sizeof(match_key)); 2437 2438 if (s->direction == PF_OUT) { 2439 dir = PF_IN; 2440 idx = PF_SK_STACK; 2441 } else { 2442 dir = PF_OUT; 2443 idx = PF_SK_WIRE; 2444 } 2445 2446 match_key.af = s->key[idx]->af; 2447 match_key.proto = s->key[idx]->proto; 2448 PF_ACPY(&match_key.addr[0], 2449 &s->key[idx]->addr[1], match_key.af); 2450 match_key.port[0] = s->key[idx]->port[1]; 2451 PF_ACPY(&match_key.addr[1], 2452 &s->key[idx]->addr[0], match_key.af); 2453 match_key.port[1] = s->key[idx]->port[0]; 2454 } 2455 2456 pf_unlink_state(s); 2457 killed++; 2458 2459 if (psk->psk_kill_match) 2460 killed += pf_kill_matching_state(&match_key, dir); 2461 2462 goto relock_DIOCKILLSTATES; 2463 } 2464 PF_HASHROW_UNLOCK(ih); 2465 2466 return (killed); 2467 } 2468 2469 static int 2470 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) 2471 { 2472 int error = 0; 2473 PF_RULES_RLOCK_TRACKER; 2474 2475 #define ERROUT_IOCTL(target, x) \ 2476 do { \ 2477 error = (x); \ 2478 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \ 2479 goto target; \ 2480 } while (0) 2481 2482 2483 /* XXX keep in sync with switch() below */ 2484 if (securelevel_gt(td->td_ucred, 2)) 2485 switch (cmd) { 2486 case DIOCGETRULES: 2487 case DIOCGETRULE: 2488 case DIOCGETRULENV: 2489 case DIOCGETADDRS: 2490 case DIOCGETADDR: 2491 case DIOCGETSTATE: 2492 case DIOCGETSTATENV: 2493 case DIOCSETSTATUSIF: 2494 case DIOCGETSTATUS: 2495 case DIOCGETSTATUSNV: 2496 case DIOCCLRSTATUS: 2497 case DIOCNATLOOK: 2498 case DIOCSETDEBUG: 2499 case DIOCGETSTATES: 2500 case DIOCGETSTATESV2: 2501 case DIOCGETTIMEOUT: 2502 case DIOCCLRRULECTRS: 2503 case DIOCGETLIMIT: 2504 case DIOCGETALTQSV0: 2505 case DIOCGETALTQSV1: 2506 case DIOCGETALTQV0: 2507 case DIOCGETALTQV1: 2508 case DIOCGETQSTATSV0: 2509 case DIOCGETQSTATSV1: 2510 case DIOCGETRULESETS: 2511 case DIOCGETRULESET: 2512 case DIOCRGETTABLES: 2513 case DIOCRGETTSTATS: 2514 case DIOCRCLRTSTATS: 2515 case DIOCRCLRADDRS: 2516 case DIOCRADDADDRS: 2517 case DIOCRDELADDRS: 2518 case DIOCRSETADDRS: 2519 case DIOCRGETADDRS: 2520 case DIOCRGETASTATS: 2521 case DIOCRCLRASTATS: 2522 case DIOCRTSTADDRS: 2523 case DIOCOSFPGET: 2524 case DIOCGETSRCNODES: 2525 case DIOCCLRSRCNODES: 2526 case DIOCGETSYNCOOKIES: 2527 case DIOCIGETIFACES: 2528 case DIOCGIFSPEEDV0: 2529 case DIOCGIFSPEEDV1: 2530 case DIOCSETIFFLAG: 2531 case DIOCCLRIFFLAG: 2532 case DIOCGETETHRULES: 2533 case DIOCGETETHRULE: 2534 case DIOCGETETHRULESETS: 2535 case DIOCGETETHRULESET: 2536 break; 2537 case DIOCRCLRTABLES: 2538 case DIOCRADDTABLES: 2539 case DIOCRDELTABLES: 2540 case DIOCRSETTFLAGS: 2541 if (((struct pfioc_table *)addr)->pfrio_flags & 2542 PFR_FLAG_DUMMY) 2543 break; /* dummy operation ok */ 2544 return (EPERM); 2545 default: 2546 return (EPERM); 2547 } 2548 2549 if (!(flags & FWRITE)) 2550 switch (cmd) { 2551 case DIOCGETRULES: 2552 case DIOCGETADDRS: 2553 case DIOCGETADDR: 2554 case DIOCGETSTATE: 2555 case DIOCGETSTATENV: 2556 case DIOCGETSTATUS: 2557 case DIOCGETSTATUSNV: 2558 case DIOCGETSTATES: 2559 case DIOCGETSTATESV2: 2560 case DIOCGETTIMEOUT: 2561 case DIOCGETLIMIT: 2562 case DIOCGETALTQSV0: 2563 case DIOCGETALTQSV1: 2564 case DIOCGETALTQV0: 2565 case DIOCGETALTQV1: 2566 case DIOCGETQSTATSV0: 2567 case DIOCGETQSTATSV1: 2568 case DIOCGETRULESETS: 2569 case DIOCGETRULESET: 2570 case DIOCNATLOOK: 2571 case DIOCRGETTABLES: 2572 case DIOCRGETTSTATS: 2573 case DIOCRGETADDRS: 2574 case DIOCRGETASTATS: 2575 case DIOCRTSTADDRS: 2576 case DIOCOSFPGET: 2577 case DIOCGETSRCNODES: 2578 case DIOCGETSYNCOOKIES: 2579 case DIOCIGETIFACES: 2580 case DIOCGIFSPEEDV1: 2581 case DIOCGIFSPEEDV0: 2582 case DIOCGETRULENV: 2583 case DIOCGETETHRULES: 2584 case DIOCGETETHRULE: 2585 case DIOCGETETHRULESETS: 2586 case DIOCGETETHRULESET: 2587 break; 2588 case DIOCRCLRTABLES: 2589 case DIOCRADDTABLES: 2590 case DIOCRDELTABLES: 2591 case DIOCRCLRTSTATS: 2592 case DIOCRCLRADDRS: 2593 case DIOCRADDADDRS: 2594 case DIOCRDELADDRS: 2595 case DIOCRSETADDRS: 2596 case DIOCRSETTFLAGS: 2597 if (((struct pfioc_table *)addr)->pfrio_flags & 2598 PFR_FLAG_DUMMY) { 2599 flags |= FWRITE; /* need write lock for dummy */ 2600 break; /* dummy operation ok */ 2601 } 2602 return (EACCES); 2603 case DIOCGETRULE: 2604 if (((struct pfioc_rule *)addr)->action == 2605 PF_GET_CLR_CNTR) 2606 return (EACCES); 2607 break; 2608 default: 2609 return (EACCES); 2610 } 2611 2612 CURVNET_SET(TD_TO_VNET(td)); 2613 2614 switch (cmd) { 2615 case DIOCSTART: 2616 sx_xlock(&V_pf_ioctl_lock); 2617 if (V_pf_status.running) 2618 error = EEXIST; 2619 else { 2620 hook_pf(); 2621 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 2622 hook_pf_eth(); 2623 V_pf_status.running = 1; 2624 V_pf_status.since = time_second; 2625 new_unrhdr64(&V_pf_stateid, time_second); 2626 2627 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n")); 2628 } 2629 break; 2630 2631 case DIOCSTOP: 2632 sx_xlock(&V_pf_ioctl_lock); 2633 if (!V_pf_status.running) 2634 error = ENOENT; 2635 else { 2636 V_pf_status.running = 0; 2637 dehook_pf(); 2638 dehook_pf_eth(); 2639 V_pf_status.since = time_second; 2640 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n")); 2641 } 2642 break; 2643 2644 case DIOCGETETHRULES: { 2645 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2646 nvlist_t *nvl; 2647 void *packed; 2648 struct pf_keth_rule *tail; 2649 struct pf_keth_ruleset *rs; 2650 u_int32_t ticket, nr; 2651 const char *anchor = ""; 2652 2653 nvl = NULL; 2654 packed = NULL; 2655 2656 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x) 2657 2658 if (nv->len > pf_ioctl_maxcount) 2659 ERROUT(ENOMEM); 2660 2661 /* Copy the request in */ 2662 packed = malloc(nv->len, M_NVLIST, M_WAITOK); 2663 if (packed == NULL) 2664 ERROUT(ENOMEM); 2665 2666 error = copyin(nv->data, packed, nv->len); 2667 if (error) 2668 ERROUT(error); 2669 2670 nvl = nvlist_unpack(packed, nv->len, 0); 2671 if (nvl == NULL) 2672 ERROUT(EBADMSG); 2673 2674 if (! nvlist_exists_string(nvl, "anchor")) 2675 ERROUT(EBADMSG); 2676 2677 anchor = nvlist_get_string(nvl, "anchor"); 2678 2679 rs = pf_find_keth_ruleset(anchor); 2680 2681 nvlist_destroy(nvl); 2682 nvl = NULL; 2683 free(packed, M_NVLIST); 2684 packed = NULL; 2685 2686 if (rs == NULL) 2687 ERROUT(ENOENT); 2688 2689 /* Reply */ 2690 nvl = nvlist_create(0); 2691 if (nvl == NULL) 2692 ERROUT(ENOMEM); 2693 2694 PF_RULES_RLOCK(); 2695 2696 ticket = rs->active.ticket; 2697 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq); 2698 if (tail) 2699 nr = tail->nr + 1; 2700 else 2701 nr = 0; 2702 2703 PF_RULES_RUNLOCK(); 2704 2705 nvlist_add_number(nvl, "ticket", ticket); 2706 nvlist_add_number(nvl, "nr", nr); 2707 2708 packed = nvlist_pack(nvl, &nv->len); 2709 if (packed == NULL) 2710 ERROUT(ENOMEM); 2711 2712 if (nv->size == 0) 2713 ERROUT(0); 2714 else if (nv->size < nv->len) 2715 ERROUT(ENOSPC); 2716 2717 error = copyout(packed, nv->data, nv->len); 2718 2719 #undef ERROUT 2720 DIOCGETETHRULES_error: 2721 free(packed, M_NVLIST); 2722 nvlist_destroy(nvl); 2723 break; 2724 } 2725 2726 case DIOCGETETHRULE: { 2727 struct epoch_tracker et; 2728 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2729 nvlist_t *nvl = NULL; 2730 void *nvlpacked = NULL; 2731 struct pf_keth_rule *rule = NULL; 2732 struct pf_keth_ruleset *rs; 2733 u_int32_t ticket, nr; 2734 bool clear = false; 2735 const char *anchor; 2736 2737 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x) 2738 2739 if (nv->len > pf_ioctl_maxcount) 2740 ERROUT(ENOMEM); 2741 2742 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 2743 if (nvlpacked == NULL) 2744 ERROUT(ENOMEM); 2745 2746 error = copyin(nv->data, nvlpacked, nv->len); 2747 if (error) 2748 ERROUT(error); 2749 2750 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 2751 if (nvl == NULL) 2752 ERROUT(EBADMSG); 2753 if (! nvlist_exists_number(nvl, "ticket")) 2754 ERROUT(EBADMSG); 2755 ticket = nvlist_get_number(nvl, "ticket"); 2756 if (! nvlist_exists_string(nvl, "anchor")) 2757 ERROUT(EBADMSG); 2758 anchor = nvlist_get_string(nvl, "anchor"); 2759 2760 if (nvlist_exists_bool(nvl, "clear")) 2761 clear = nvlist_get_bool(nvl, "clear"); 2762 2763 if (clear && !(flags & FWRITE)) 2764 ERROUT(EACCES); 2765 2766 if (! nvlist_exists_number(nvl, "nr")) 2767 ERROUT(EBADMSG); 2768 nr = nvlist_get_number(nvl, "nr"); 2769 2770 PF_RULES_RLOCK(); 2771 rs = pf_find_keth_ruleset(anchor); 2772 if (rs == NULL) { 2773 PF_RULES_RUNLOCK(); 2774 ERROUT(ENOENT); 2775 } 2776 if (ticket != rs->active.ticket) { 2777 PF_RULES_RUNLOCK(); 2778 ERROUT(EBUSY); 2779 } 2780 2781 nvlist_destroy(nvl); 2782 nvl = NULL; 2783 free(nvlpacked, M_NVLIST); 2784 nvlpacked = NULL; 2785 2786 rule = TAILQ_FIRST(rs->active.rules); 2787 while ((rule != NULL) && (rule->nr != nr)) 2788 rule = TAILQ_NEXT(rule, entries); 2789 if (rule == NULL) { 2790 PF_RULES_RUNLOCK(); 2791 ERROUT(ENOENT); 2792 } 2793 /* Make sure rule can't go away. */ 2794 NET_EPOCH_ENTER(et); 2795 PF_RULES_RUNLOCK(); 2796 nvl = pf_keth_rule_to_nveth_rule(rule); 2797 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) 2798 ERROUT(EBUSY); 2799 NET_EPOCH_EXIT(et); 2800 if (nvl == NULL) 2801 ERROUT(ENOMEM); 2802 2803 nvlpacked = nvlist_pack(nvl, &nv->len); 2804 if (nvlpacked == NULL) 2805 ERROUT(ENOMEM); 2806 2807 if (nv->size == 0) 2808 ERROUT(0); 2809 else if (nv->size < nv->len) 2810 ERROUT(ENOSPC); 2811 2812 error = copyout(nvlpacked, nv->data, nv->len); 2813 if (error == 0 && clear) { 2814 counter_u64_zero(rule->evaluations); 2815 for (int i = 0; i < 2; i++) { 2816 counter_u64_zero(rule->packets[i]); 2817 counter_u64_zero(rule->bytes[i]); 2818 } 2819 } 2820 2821 #undef ERROUT 2822 DIOCGETETHRULE_error: 2823 free(nvlpacked, M_NVLIST); 2824 nvlist_destroy(nvl); 2825 break; 2826 } 2827 2828 case DIOCADDETHRULE: { 2829 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2830 nvlist_t *nvl = NULL; 2831 void *nvlpacked = NULL; 2832 struct pf_keth_rule *rule = NULL, *tail = NULL; 2833 struct pf_keth_ruleset *ruleset = NULL; 2834 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL; 2835 const char *anchor = "", *anchor_call = ""; 2836 2837 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x) 2838 2839 if (nv->len > pf_ioctl_maxcount) 2840 ERROUT(ENOMEM); 2841 2842 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 2843 if (nvlpacked == NULL) 2844 ERROUT(ENOMEM); 2845 2846 error = copyin(nv->data, nvlpacked, nv->len); 2847 if (error) 2848 ERROUT(error); 2849 2850 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 2851 if (nvl == NULL) 2852 ERROUT(EBADMSG); 2853 2854 if (! nvlist_exists_number(nvl, "ticket")) 2855 ERROUT(EBADMSG); 2856 2857 if (nvlist_exists_string(nvl, "anchor")) 2858 anchor = nvlist_get_string(nvl, "anchor"); 2859 if (nvlist_exists_string(nvl, "anchor_call")) 2860 anchor_call = nvlist_get_string(nvl, "anchor_call"); 2861 2862 ruleset = pf_find_keth_ruleset(anchor); 2863 if (ruleset == NULL) 2864 ERROUT(EINVAL); 2865 2866 if (nvlist_get_number(nvl, "ticket") != 2867 ruleset->inactive.ticket) { 2868 DPFPRINTF(PF_DEBUG_MISC, 2869 ("ticket: %d != %d\n", 2870 (u_int32_t)nvlist_get_number(nvl, "ticket"), 2871 ruleset->inactive.ticket)); 2872 ERROUT(EBUSY); 2873 } 2874 2875 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); 2876 if (rule == NULL) 2877 ERROUT(ENOMEM); 2878 rule->timestamp = NULL; 2879 2880 error = pf_nveth_rule_to_keth_rule(nvl, rule); 2881 if (error != 0) 2882 ERROUT(error); 2883 2884 if (rule->ifname[0]) 2885 kif = pf_kkif_create(M_WAITOK); 2886 if (rule->bridge_to_name[0]) 2887 bridge_to_kif = pf_kkif_create(M_WAITOK); 2888 rule->evaluations = counter_u64_alloc(M_WAITOK); 2889 for (int i = 0; i < 2; i++) { 2890 rule->packets[i] = counter_u64_alloc(M_WAITOK); 2891 rule->bytes[i] = counter_u64_alloc(M_WAITOK); 2892 } 2893 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 2894 M_WAITOK | M_ZERO); 2895 2896 PF_RULES_WLOCK(); 2897 2898 if (rule->ifname[0]) { 2899 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2900 pfi_kkif_ref(rule->kif); 2901 } else 2902 rule->kif = NULL; 2903 if (rule->bridge_to_name[0]) { 2904 rule->bridge_to = pfi_kkif_attach(bridge_to_kif, 2905 rule->bridge_to_name); 2906 pfi_kkif_ref(rule->bridge_to); 2907 } else 2908 rule->bridge_to = NULL; 2909 2910 #ifdef ALTQ 2911 /* set queue IDs */ 2912 if (rule->qname[0] != 0) { 2913 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2914 error = EBUSY; 2915 else 2916 rule->qid = rule->qid; 2917 } 2918 #endif 2919 if (rule->tagname[0]) 2920 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2921 error = EBUSY; 2922 if (rule->match_tagname[0]) 2923 if ((rule->match_tag = pf_tagname2tag( 2924 rule->match_tagname)) == 0) 2925 error = EBUSY; 2926 2927 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE) 2928 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr); 2929 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE) 2930 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr); 2931 2932 if (error) { 2933 pf_free_eth_rule(rule); 2934 PF_RULES_WUNLOCK(); 2935 ERROUT(error); 2936 } 2937 2938 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) { 2939 pf_free_eth_rule(rule); 2940 PF_RULES_WUNLOCK(); 2941 ERROUT(EINVAL); 2942 } 2943 2944 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq); 2945 if (tail) 2946 rule->nr = tail->nr + 1; 2947 else 2948 rule->nr = 0; 2949 2950 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries); 2951 2952 PF_RULES_WUNLOCK(); 2953 2954 #undef ERROUT 2955 DIOCADDETHRULE_error: 2956 nvlist_destroy(nvl); 2957 free(nvlpacked, M_NVLIST); 2958 break; 2959 } 2960 2961 case DIOCGETETHRULESETS: { 2962 struct epoch_tracker et; 2963 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2964 nvlist_t *nvl = NULL; 2965 void *nvlpacked = NULL; 2966 struct pf_keth_ruleset *ruleset; 2967 struct pf_keth_anchor *anchor; 2968 int nr = 0; 2969 2970 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x) 2971 2972 if (nv->len > pf_ioctl_maxcount) 2973 ERROUT(ENOMEM); 2974 2975 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 2976 if (nvlpacked == NULL) 2977 ERROUT(ENOMEM); 2978 2979 error = copyin(nv->data, nvlpacked, nv->len); 2980 if (error) 2981 ERROUT(error); 2982 2983 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 2984 if (nvl == NULL) 2985 ERROUT(EBADMSG); 2986 if (! nvlist_exists_string(nvl, "path")) 2987 ERROUT(EBADMSG); 2988 2989 NET_EPOCH_ENTER(et); 2990 2991 if ((ruleset = pf_find_keth_ruleset( 2992 nvlist_get_string(nvl, "path"))) == NULL) { 2993 NET_EPOCH_EXIT(et); 2994 ERROUT(ENOENT); 2995 } 2996 2997 if (ruleset->anchor == NULL) { 2998 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors) 2999 if (anchor->parent == NULL) 3000 nr++; 3001 } else { 3002 RB_FOREACH(anchor, pf_keth_anchor_node, 3003 &ruleset->anchor->children) 3004 nr++; 3005 } 3006 3007 NET_EPOCH_EXIT(et); 3008 3009 nvlist_destroy(nvl); 3010 nvl = NULL; 3011 free(nvlpacked, M_NVLIST); 3012 nvlpacked = NULL; 3013 3014 nvl = nvlist_create(0); 3015 if (nvl == NULL) 3016 ERROUT(ENOMEM); 3017 3018 nvlist_add_number(nvl, "nr", nr); 3019 3020 nvlpacked = nvlist_pack(nvl, &nv->len); 3021 if (nvlpacked == NULL) 3022 ERROUT(ENOMEM); 3023 3024 if (nv->size == 0) 3025 ERROUT(0); 3026 else if (nv->size < nv->len) 3027 ERROUT(ENOSPC); 3028 3029 error = copyout(nvlpacked, nv->data, nv->len); 3030 3031 #undef ERROUT 3032 DIOCGETETHRULESETS_error: 3033 free(nvlpacked, M_NVLIST); 3034 nvlist_destroy(nvl); 3035 break; 3036 } 3037 3038 case DIOCGETETHRULESET: { 3039 struct epoch_tracker et; 3040 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3041 nvlist_t *nvl = NULL; 3042 void *nvlpacked = NULL; 3043 struct pf_keth_ruleset *ruleset; 3044 struct pf_keth_anchor *anchor; 3045 int nr = 0, req_nr = 0; 3046 bool found = false; 3047 3048 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x) 3049 3050 if (nv->len > pf_ioctl_maxcount) 3051 ERROUT(ENOMEM); 3052 3053 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3054 if (nvlpacked == NULL) 3055 ERROUT(ENOMEM); 3056 3057 error = copyin(nv->data, nvlpacked, nv->len); 3058 if (error) 3059 ERROUT(error); 3060 3061 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3062 if (nvl == NULL) 3063 ERROUT(EBADMSG); 3064 if (! nvlist_exists_string(nvl, "path")) 3065 ERROUT(EBADMSG); 3066 if (! nvlist_exists_number(nvl, "nr")) 3067 ERROUT(EBADMSG); 3068 3069 req_nr = nvlist_get_number(nvl, "nr"); 3070 3071 NET_EPOCH_ENTER(et); 3072 3073 if ((ruleset = pf_find_keth_ruleset( 3074 nvlist_get_string(nvl, "path"))) == NULL) { 3075 NET_EPOCH_EXIT(et); 3076 ERROUT(ENOENT); 3077 } 3078 3079 nvlist_destroy(nvl); 3080 nvl = NULL; 3081 free(nvlpacked, M_NVLIST); 3082 nvlpacked = NULL; 3083 3084 nvl = nvlist_create(0); 3085 if (nvl == NULL) { 3086 NET_EPOCH_EXIT(et); 3087 ERROUT(ENOMEM); 3088 } 3089 3090 if (ruleset->anchor == NULL) { 3091 RB_FOREACH(anchor, pf_keth_anchor_global, 3092 &V_pf_keth_anchors) { 3093 if (anchor->parent == NULL && nr++ == req_nr) { 3094 found = true; 3095 break; 3096 } 3097 } 3098 } else { 3099 RB_FOREACH(anchor, pf_keth_anchor_node, 3100 &ruleset->anchor->children) { 3101 if (nr++ == req_nr) { 3102 found = true; 3103 break; 3104 } 3105 } 3106 } 3107 3108 NET_EPOCH_EXIT(et); 3109 if (found) { 3110 nvlist_add_number(nvl, "nr", nr); 3111 nvlist_add_string(nvl, "name", anchor->name); 3112 if (ruleset->anchor) 3113 nvlist_add_string(nvl, "path", 3114 ruleset->anchor->path); 3115 else 3116 nvlist_add_string(nvl, "path", ""); 3117 } else { 3118 ERROUT(EBUSY); 3119 } 3120 3121 nvlpacked = nvlist_pack(nvl, &nv->len); 3122 if (nvlpacked == NULL) 3123 ERROUT(ENOMEM); 3124 3125 if (nv->size == 0) 3126 ERROUT(0); 3127 else if (nv->size < nv->len) 3128 ERROUT(ENOSPC); 3129 3130 error = copyout(nvlpacked, nv->data, nv->len); 3131 3132 #undef ERROUT 3133 DIOCGETETHRULESET_error: 3134 free(nvlpacked, M_NVLIST); 3135 nvlist_destroy(nvl); 3136 break; 3137 } 3138 3139 case DIOCADDRULENV: { 3140 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3141 nvlist_t *nvl = NULL; 3142 void *nvlpacked = NULL; 3143 struct pf_krule *rule = NULL; 3144 const char *anchor = "", *anchor_call = ""; 3145 uint32_t ticket = 0, pool_ticket = 0; 3146 3147 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x) 3148 3149 if (nv->len > pf_ioctl_maxcount) 3150 ERROUT(ENOMEM); 3151 3152 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3153 error = copyin(nv->data, nvlpacked, nv->len); 3154 if (error) 3155 ERROUT(error); 3156 3157 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3158 if (nvl == NULL) 3159 ERROUT(EBADMSG); 3160 3161 if (! nvlist_exists_number(nvl, "ticket")) 3162 ERROUT(EINVAL); 3163 ticket = nvlist_get_number(nvl, "ticket"); 3164 3165 if (! nvlist_exists_number(nvl, "pool_ticket")) 3166 ERROUT(EINVAL); 3167 pool_ticket = nvlist_get_number(nvl, "pool_ticket"); 3168 3169 if (! nvlist_exists_nvlist(nvl, "rule")) 3170 ERROUT(EINVAL); 3171 3172 rule = pf_krule_alloc(); 3173 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"), 3174 rule); 3175 if (error) 3176 ERROUT(error); 3177 3178 if (nvlist_exists_string(nvl, "anchor")) 3179 anchor = nvlist_get_string(nvl, "anchor"); 3180 if (nvlist_exists_string(nvl, "anchor_call")) 3181 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3182 3183 if ((error = nvlist_error(nvl))) 3184 ERROUT(error); 3185 3186 /* Frees rule on error */ 3187 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor, 3188 anchor_call, td); 3189 3190 nvlist_destroy(nvl); 3191 free(nvlpacked, M_NVLIST); 3192 break; 3193 #undef ERROUT 3194 DIOCADDRULENV_error: 3195 pf_krule_free(rule); 3196 nvlist_destroy(nvl); 3197 free(nvlpacked, M_NVLIST); 3198 3199 break; 3200 } 3201 case DIOCADDRULE: { 3202 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3203 struct pf_krule *rule; 3204 3205 rule = pf_krule_alloc(); 3206 error = pf_rule_to_krule(&pr->rule, rule); 3207 if (error != 0) { 3208 pf_krule_free(rule); 3209 break; 3210 } 3211 3212 pr->anchor[sizeof(pr->anchor) - 1] = 0; 3213 3214 /* Frees rule on error */ 3215 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, 3216 pr->anchor, pr->anchor_call, td); 3217 break; 3218 } 3219 3220 case DIOCGETRULES: { 3221 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3222 struct pf_kruleset *ruleset; 3223 struct pf_krule *tail; 3224 int rs_num; 3225 3226 pr->anchor[sizeof(pr->anchor) - 1] = 0; 3227 3228 PF_RULES_WLOCK(); 3229 ruleset = pf_find_kruleset(pr->anchor); 3230 if (ruleset == NULL) { 3231 PF_RULES_WUNLOCK(); 3232 error = EINVAL; 3233 break; 3234 } 3235 rs_num = pf_get_ruleset_number(pr->rule.action); 3236 if (rs_num >= PF_RULESET_MAX) { 3237 PF_RULES_WUNLOCK(); 3238 error = EINVAL; 3239 break; 3240 } 3241 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 3242 pf_krulequeue); 3243 if (tail) 3244 pr->nr = tail->nr + 1; 3245 else 3246 pr->nr = 0; 3247 pr->ticket = ruleset->rules[rs_num].active.ticket; 3248 PF_RULES_WUNLOCK(); 3249 break; 3250 } 3251 3252 case DIOCGETRULE: { 3253 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3254 struct pf_kruleset *ruleset; 3255 struct pf_krule *rule; 3256 int rs_num; 3257 3258 pr->anchor[sizeof(pr->anchor) - 1] = 0; 3259 3260 PF_RULES_WLOCK(); 3261 ruleset = pf_find_kruleset(pr->anchor); 3262 if (ruleset == NULL) { 3263 PF_RULES_WUNLOCK(); 3264 error = EINVAL; 3265 break; 3266 } 3267 rs_num = pf_get_ruleset_number(pr->rule.action); 3268 if (rs_num >= PF_RULESET_MAX) { 3269 PF_RULES_WUNLOCK(); 3270 error = EINVAL; 3271 break; 3272 } 3273 if (pr->ticket != ruleset->rules[rs_num].active.ticket) { 3274 PF_RULES_WUNLOCK(); 3275 error = EBUSY; 3276 break; 3277 } 3278 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3279 while ((rule != NULL) && (rule->nr != pr->nr)) 3280 rule = TAILQ_NEXT(rule, entries); 3281 if (rule == NULL) { 3282 PF_RULES_WUNLOCK(); 3283 error = EBUSY; 3284 break; 3285 } 3286 3287 pf_krule_to_rule(rule, &pr->rule); 3288 3289 if (pf_kanchor_copyout(ruleset, rule, pr)) { 3290 PF_RULES_WUNLOCK(); 3291 error = EBUSY; 3292 break; 3293 } 3294 pf_addr_copyout(&pr->rule.src.addr); 3295 pf_addr_copyout(&pr->rule.dst.addr); 3296 3297 if (pr->action == PF_GET_CLR_CNTR) { 3298 pf_counter_u64_zero(&rule->evaluations); 3299 for (int i = 0; i < 2; i++) { 3300 pf_counter_u64_zero(&rule->packets[i]); 3301 pf_counter_u64_zero(&rule->bytes[i]); 3302 } 3303 counter_u64_zero(rule->states_tot); 3304 } 3305 PF_RULES_WUNLOCK(); 3306 break; 3307 } 3308 3309 case DIOCGETRULENV: { 3310 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3311 nvlist_t *nvrule = NULL; 3312 nvlist_t *nvl = NULL; 3313 struct pf_kruleset *ruleset; 3314 struct pf_krule *rule; 3315 void *nvlpacked = NULL; 3316 int rs_num, nr; 3317 bool clear_counter = false; 3318 3319 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x) 3320 3321 if (nv->len > pf_ioctl_maxcount) 3322 ERROUT(ENOMEM); 3323 3324 /* Copy the request in */ 3325 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3326 if (nvlpacked == NULL) 3327 ERROUT(ENOMEM); 3328 3329 error = copyin(nv->data, nvlpacked, nv->len); 3330 if (error) 3331 ERROUT(error); 3332 3333 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3334 if (nvl == NULL) 3335 ERROUT(EBADMSG); 3336 3337 if (! nvlist_exists_string(nvl, "anchor")) 3338 ERROUT(EBADMSG); 3339 if (! nvlist_exists_number(nvl, "ruleset")) 3340 ERROUT(EBADMSG); 3341 if (! nvlist_exists_number(nvl, "ticket")) 3342 ERROUT(EBADMSG); 3343 if (! nvlist_exists_number(nvl, "nr")) 3344 ERROUT(EBADMSG); 3345 3346 if (nvlist_exists_bool(nvl, "clear_counter")) 3347 clear_counter = nvlist_get_bool(nvl, "clear_counter"); 3348 3349 if (clear_counter && !(flags & FWRITE)) 3350 ERROUT(EACCES); 3351 3352 nr = nvlist_get_number(nvl, "nr"); 3353 3354 PF_RULES_WLOCK(); 3355 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor")); 3356 if (ruleset == NULL) { 3357 PF_RULES_WUNLOCK(); 3358 ERROUT(ENOENT); 3359 } 3360 3361 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset")); 3362 if (rs_num >= PF_RULESET_MAX) { 3363 PF_RULES_WUNLOCK(); 3364 ERROUT(EINVAL); 3365 } 3366 3367 if (nvlist_get_number(nvl, "ticket") != 3368 ruleset->rules[rs_num].active.ticket) { 3369 PF_RULES_WUNLOCK(); 3370 ERROUT(EBUSY); 3371 } 3372 3373 if ((error = nvlist_error(nvl))) { 3374 PF_RULES_WUNLOCK(); 3375 ERROUT(error); 3376 } 3377 3378 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3379 while ((rule != NULL) && (rule->nr != nr)) 3380 rule = TAILQ_NEXT(rule, entries); 3381 if (rule == NULL) { 3382 PF_RULES_WUNLOCK(); 3383 ERROUT(EBUSY); 3384 } 3385 3386 nvrule = pf_krule_to_nvrule(rule); 3387 3388 nvlist_destroy(nvl); 3389 nvl = nvlist_create(0); 3390 if (nvl == NULL) { 3391 PF_RULES_WUNLOCK(); 3392 ERROUT(ENOMEM); 3393 } 3394 nvlist_add_number(nvl, "nr", nr); 3395 nvlist_add_nvlist(nvl, "rule", nvrule); 3396 nvlist_destroy(nvrule); 3397 nvrule = NULL; 3398 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) { 3399 PF_RULES_WUNLOCK(); 3400 ERROUT(EBUSY); 3401 } 3402 3403 free(nvlpacked, M_NVLIST); 3404 nvlpacked = nvlist_pack(nvl, &nv->len); 3405 if (nvlpacked == NULL) { 3406 PF_RULES_WUNLOCK(); 3407 ERROUT(ENOMEM); 3408 } 3409 3410 if (nv->size == 0) { 3411 PF_RULES_WUNLOCK(); 3412 ERROUT(0); 3413 } 3414 else if (nv->size < nv->len) { 3415 PF_RULES_WUNLOCK(); 3416 ERROUT(ENOSPC); 3417 } 3418 3419 if (clear_counter) { 3420 pf_counter_u64_zero(&rule->evaluations); 3421 for (int i = 0; i < 2; i++) { 3422 pf_counter_u64_zero(&rule->packets[i]); 3423 pf_counter_u64_zero(&rule->bytes[i]); 3424 } 3425 counter_u64_zero(rule->states_tot); 3426 } 3427 PF_RULES_WUNLOCK(); 3428 3429 error = copyout(nvlpacked, nv->data, nv->len); 3430 3431 #undef ERROUT 3432 DIOCGETRULENV_error: 3433 free(nvlpacked, M_NVLIST); 3434 nvlist_destroy(nvrule); 3435 nvlist_destroy(nvl); 3436 3437 break; 3438 } 3439 3440 case DIOCCHANGERULE: { 3441 struct pfioc_rule *pcr = (struct pfioc_rule *)addr; 3442 struct pf_kruleset *ruleset; 3443 struct pf_krule *oldrule = NULL, *newrule = NULL; 3444 struct pfi_kkif *kif = NULL; 3445 struct pf_kpooladdr *pa; 3446 u_int32_t nr = 0; 3447 int rs_num; 3448 3449 pcr->anchor[sizeof(pcr->anchor) - 1] = 0; 3450 3451 if (pcr->action < PF_CHANGE_ADD_HEAD || 3452 pcr->action > PF_CHANGE_GET_TICKET) { 3453 error = EINVAL; 3454 break; 3455 } 3456 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { 3457 error = EINVAL; 3458 break; 3459 } 3460 3461 if (pcr->action != PF_CHANGE_REMOVE) { 3462 newrule = pf_krule_alloc(); 3463 error = pf_rule_to_krule(&pcr->rule, newrule); 3464 if (error != 0) { 3465 pf_krule_free(newrule); 3466 break; 3467 } 3468 3469 if (newrule->ifname[0]) 3470 kif = pf_kkif_create(M_WAITOK); 3471 pf_counter_u64_init(&newrule->evaluations, M_WAITOK); 3472 for (int i = 0; i < 2; i++) { 3473 pf_counter_u64_init(&newrule->packets[i], M_WAITOK); 3474 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK); 3475 } 3476 newrule->states_cur = counter_u64_alloc(M_WAITOK); 3477 newrule->states_tot = counter_u64_alloc(M_WAITOK); 3478 newrule->src_nodes = counter_u64_alloc(M_WAITOK); 3479 newrule->cuid = td->td_ucred->cr_ruid; 3480 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 3481 TAILQ_INIT(&newrule->rpool.list); 3482 } 3483 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x) 3484 3485 PF_CONFIG_LOCK(); 3486 PF_RULES_WLOCK(); 3487 #ifdef PF_WANT_32_TO_64_COUNTER 3488 if (newrule != NULL) { 3489 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist); 3490 newrule->allrulelinked = true; 3491 V_pf_allrulecount++; 3492 } 3493 #endif 3494 3495 if (!(pcr->action == PF_CHANGE_REMOVE || 3496 pcr->action == PF_CHANGE_GET_TICKET) && 3497 pcr->pool_ticket != V_ticket_pabuf) 3498 ERROUT(EBUSY); 3499 3500 ruleset = pf_find_kruleset(pcr->anchor); 3501 if (ruleset == NULL) 3502 ERROUT(EINVAL); 3503 3504 rs_num = pf_get_ruleset_number(pcr->rule.action); 3505 if (rs_num >= PF_RULESET_MAX) 3506 ERROUT(EINVAL); 3507 3508 /* 3509 * XXXMJG: there is no guarantee that the ruleset was 3510 * created by the usual route of calling DIOCXBEGIN. 3511 * As a result it is possible the rule tree will not 3512 * be allocated yet. Hack around it by doing it here. 3513 * Note it is fine to let the tree persist in case of 3514 * error as it will be freed down the road on future 3515 * updates (if need be). 3516 */ 3517 if (ruleset->rules[rs_num].active.tree == NULL) { 3518 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT); 3519 if (ruleset->rules[rs_num].active.tree == NULL) { 3520 ERROUT(ENOMEM); 3521 } 3522 } 3523 3524 if (pcr->action == PF_CHANGE_GET_TICKET) { 3525 pcr->ticket = ++ruleset->rules[rs_num].active.ticket; 3526 ERROUT(0); 3527 } else if (pcr->ticket != 3528 ruleset->rules[rs_num].active.ticket) 3529 ERROUT(EINVAL); 3530 3531 if (pcr->action != PF_CHANGE_REMOVE) { 3532 if (newrule->ifname[0]) { 3533 newrule->kif = pfi_kkif_attach(kif, 3534 newrule->ifname); 3535 kif = NULL; 3536 pfi_kkif_ref(newrule->kif); 3537 } else 3538 newrule->kif = NULL; 3539 3540 if (newrule->rtableid > 0 && 3541 newrule->rtableid >= rt_numfibs) 3542 error = EBUSY; 3543 3544 #ifdef ALTQ 3545 /* set queue IDs */ 3546 if (newrule->qname[0] != 0) { 3547 if ((newrule->qid = 3548 pf_qname2qid(newrule->qname)) == 0) 3549 error = EBUSY; 3550 else if (newrule->pqname[0] != 0) { 3551 if ((newrule->pqid = 3552 pf_qname2qid(newrule->pqname)) == 0) 3553 error = EBUSY; 3554 } else 3555 newrule->pqid = newrule->qid; 3556 } 3557 #endif /* ALTQ */ 3558 if (newrule->tagname[0]) 3559 if ((newrule->tag = 3560 pf_tagname2tag(newrule->tagname)) == 0) 3561 error = EBUSY; 3562 if (newrule->match_tagname[0]) 3563 if ((newrule->match_tag = pf_tagname2tag( 3564 newrule->match_tagname)) == 0) 3565 error = EBUSY; 3566 if (newrule->rt && !newrule->direction) 3567 error = EINVAL; 3568 if (!newrule->log) 3569 newrule->logif = 0; 3570 if (newrule->logif >= PFLOGIFS_MAX) 3571 error = EINVAL; 3572 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) 3573 error = ENOMEM; 3574 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) 3575 error = ENOMEM; 3576 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call)) 3577 error = EINVAL; 3578 TAILQ_FOREACH(pa, &V_pf_pabuf, entries) 3579 if (pa->addr.type == PF_ADDR_TABLE) { 3580 pa->addr.p.tbl = 3581 pfr_attach_table(ruleset, 3582 pa->addr.v.tblname); 3583 if (pa->addr.p.tbl == NULL) 3584 error = ENOMEM; 3585 } 3586 3587 newrule->overload_tbl = NULL; 3588 if (newrule->overload_tblname[0]) { 3589 if ((newrule->overload_tbl = pfr_attach_table( 3590 ruleset, newrule->overload_tblname)) == 3591 NULL) 3592 error = EINVAL; 3593 else 3594 newrule->overload_tbl->pfrkt_flags |= 3595 PFR_TFLAG_ACTIVE; 3596 } 3597 3598 pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list); 3599 if (((((newrule->action == PF_NAT) || 3600 (newrule->action == PF_RDR) || 3601 (newrule->action == PF_BINAT) || 3602 (newrule->rt > PF_NOPFROUTE)) && 3603 !newrule->anchor)) && 3604 (TAILQ_FIRST(&newrule->rpool.list) == NULL)) 3605 error = EINVAL; 3606 3607 if (error) { 3608 pf_free_rule(newrule); 3609 PF_RULES_WUNLOCK(); 3610 PF_CONFIG_UNLOCK(); 3611 break; 3612 } 3613 3614 newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list); 3615 } 3616 pf_empty_kpool(&V_pf_pabuf); 3617 3618 if (pcr->action == PF_CHANGE_ADD_HEAD) 3619 oldrule = TAILQ_FIRST( 3620 ruleset->rules[rs_num].active.ptr); 3621 else if (pcr->action == PF_CHANGE_ADD_TAIL) 3622 oldrule = TAILQ_LAST( 3623 ruleset->rules[rs_num].active.ptr, pf_krulequeue); 3624 else { 3625 oldrule = TAILQ_FIRST( 3626 ruleset->rules[rs_num].active.ptr); 3627 while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) 3628 oldrule = TAILQ_NEXT(oldrule, entries); 3629 if (oldrule == NULL) { 3630 if (newrule != NULL) 3631 pf_free_rule(newrule); 3632 PF_RULES_WUNLOCK(); 3633 PF_CONFIG_UNLOCK(); 3634 error = EINVAL; 3635 break; 3636 } 3637 } 3638 3639 if (pcr->action == PF_CHANGE_REMOVE) { 3640 pf_unlink_rule(ruleset->rules[rs_num].active.ptr, 3641 oldrule); 3642 RB_REMOVE(pf_krule_global, 3643 ruleset->rules[rs_num].active.tree, oldrule); 3644 ruleset->rules[rs_num].active.rcount--; 3645 } else { 3646 pf_hash_rule(newrule); 3647 if (RB_INSERT(pf_krule_global, 3648 ruleset->rules[rs_num].active.tree, newrule) != NULL) { 3649 pf_free_rule(newrule); 3650 PF_RULES_WUNLOCK(); 3651 PF_CONFIG_UNLOCK(); 3652 error = EEXIST; 3653 break; 3654 } 3655 3656 if (oldrule == NULL) 3657 TAILQ_INSERT_TAIL( 3658 ruleset->rules[rs_num].active.ptr, 3659 newrule, entries); 3660 else if (pcr->action == PF_CHANGE_ADD_HEAD || 3661 pcr->action == PF_CHANGE_ADD_BEFORE) 3662 TAILQ_INSERT_BEFORE(oldrule, newrule, entries); 3663 else 3664 TAILQ_INSERT_AFTER( 3665 ruleset->rules[rs_num].active.ptr, 3666 oldrule, newrule, entries); 3667 ruleset->rules[rs_num].active.rcount++; 3668 } 3669 3670 nr = 0; 3671 TAILQ_FOREACH(oldrule, 3672 ruleset->rules[rs_num].active.ptr, entries) 3673 oldrule->nr = nr++; 3674 3675 ruleset->rules[rs_num].active.ticket++; 3676 3677 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); 3678 pf_remove_if_empty_kruleset(ruleset); 3679 3680 PF_RULES_WUNLOCK(); 3681 PF_CONFIG_UNLOCK(); 3682 break; 3683 3684 #undef ERROUT 3685 DIOCCHANGERULE_error: 3686 PF_RULES_WUNLOCK(); 3687 PF_CONFIG_UNLOCK(); 3688 pf_krule_free(newrule); 3689 pf_kkif_free(kif); 3690 break; 3691 } 3692 3693 case DIOCCLRSTATES: { 3694 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr; 3695 struct pf_kstate_kill kill; 3696 3697 error = pf_state_kill_to_kstate_kill(psk, &kill); 3698 if (error) 3699 break; 3700 3701 psk->psk_killed = pf_clear_states(&kill); 3702 break; 3703 } 3704 3705 case DIOCCLRSTATESNV: { 3706 error = pf_clearstates_nv((struct pfioc_nv *)addr); 3707 break; 3708 } 3709 3710 case DIOCKILLSTATES: { 3711 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr; 3712 struct pf_kstate_kill kill; 3713 3714 error = pf_state_kill_to_kstate_kill(psk, &kill); 3715 if (error) 3716 break; 3717 3718 psk->psk_killed = 0; 3719 pf_killstates(&kill, &psk->psk_killed); 3720 break; 3721 } 3722 3723 case DIOCKILLSTATESNV: { 3724 error = pf_killstates_nv((struct pfioc_nv *)addr); 3725 break; 3726 } 3727 3728 case DIOCADDSTATE: { 3729 struct pfioc_state *ps = (struct pfioc_state *)addr; 3730 struct pfsync_state_1301 *sp = &ps->state; 3731 3732 if (sp->timeout >= PFTM_MAX) { 3733 error = EINVAL; 3734 break; 3735 } 3736 if (V_pfsync_state_import_ptr != NULL) { 3737 PF_RULES_RLOCK(); 3738 error = V_pfsync_state_import_ptr( 3739 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL, 3740 PFSYNC_MSG_VERSION_1301); 3741 PF_RULES_RUNLOCK(); 3742 } else 3743 error = EOPNOTSUPP; 3744 break; 3745 } 3746 3747 case DIOCGETSTATE: { 3748 struct pfioc_state *ps = (struct pfioc_state *)addr; 3749 struct pf_kstate *s; 3750 3751 s = pf_find_state_byid(ps->state.id, ps->state.creatorid); 3752 if (s == NULL) { 3753 error = ENOENT; 3754 break; 3755 } 3756 3757 pfsync_state_export((union pfsync_state_union*)&ps->state, 3758 s, PFSYNC_MSG_VERSION_1301); 3759 PF_STATE_UNLOCK(s); 3760 break; 3761 } 3762 3763 case DIOCGETSTATENV: { 3764 error = pf_getstate((struct pfioc_nv *)addr); 3765 break; 3766 } 3767 3768 case DIOCGETSTATES: { 3769 struct pfioc_states *ps = (struct pfioc_states *)addr; 3770 struct pf_kstate *s; 3771 struct pfsync_state_1301 *pstore, *p; 3772 int i, nr; 3773 size_t slice_count = 16, count; 3774 void *out; 3775 3776 if (ps->ps_len <= 0) { 3777 nr = uma_zone_get_cur(V_pf_state_z); 3778 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 3779 break; 3780 } 3781 3782 out = ps->ps_states; 3783 pstore = mallocarray(slice_count, 3784 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO); 3785 nr = 0; 3786 3787 for (i = 0; i <= pf_hashmask; i++) { 3788 struct pf_idhash *ih = &V_pf_idhash[i]; 3789 3790 DIOCGETSTATES_retry: 3791 p = pstore; 3792 3793 if (LIST_EMPTY(&ih->states)) 3794 continue; 3795 3796 PF_HASHROW_LOCK(ih); 3797 count = 0; 3798 LIST_FOREACH(s, &ih->states, entry) { 3799 if (s->timeout == PFTM_UNLINKED) 3800 continue; 3801 count++; 3802 } 3803 3804 if (count > slice_count) { 3805 PF_HASHROW_UNLOCK(ih); 3806 free(pstore, M_TEMP); 3807 slice_count = count * 2; 3808 pstore = mallocarray(slice_count, 3809 sizeof(struct pfsync_state_1301), M_TEMP, 3810 M_WAITOK | M_ZERO); 3811 goto DIOCGETSTATES_retry; 3812 } 3813 3814 if ((nr+count) * sizeof(*p) > ps->ps_len) { 3815 PF_HASHROW_UNLOCK(ih); 3816 goto DIOCGETSTATES_full; 3817 } 3818 3819 LIST_FOREACH(s, &ih->states, entry) { 3820 if (s->timeout == PFTM_UNLINKED) 3821 continue; 3822 3823 pfsync_state_export((union pfsync_state_union*)p, 3824 s, PFSYNC_MSG_VERSION_1301); 3825 p++; 3826 nr++; 3827 } 3828 PF_HASHROW_UNLOCK(ih); 3829 error = copyout(pstore, out, 3830 sizeof(struct pfsync_state_1301) * count); 3831 if (error) 3832 break; 3833 out = ps->ps_states + nr; 3834 } 3835 DIOCGETSTATES_full: 3836 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 3837 free(pstore, M_TEMP); 3838 3839 break; 3840 } 3841 3842 case DIOCGETSTATESV2: { 3843 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr; 3844 struct pf_kstate *s; 3845 struct pf_state_export *pstore, *p; 3846 int i, nr; 3847 size_t slice_count = 16, count; 3848 void *out; 3849 3850 if (ps->ps_req_version > PF_STATE_VERSION) { 3851 error = ENOTSUP; 3852 break; 3853 } 3854 3855 if (ps->ps_len <= 0) { 3856 nr = uma_zone_get_cur(V_pf_state_z); 3857 ps->ps_len = sizeof(struct pf_state_export) * nr; 3858 break; 3859 } 3860 3861 out = ps->ps_states; 3862 pstore = mallocarray(slice_count, 3863 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO); 3864 nr = 0; 3865 3866 for (i = 0; i <= pf_hashmask; i++) { 3867 struct pf_idhash *ih = &V_pf_idhash[i]; 3868 3869 DIOCGETSTATESV2_retry: 3870 p = pstore; 3871 3872 if (LIST_EMPTY(&ih->states)) 3873 continue; 3874 3875 PF_HASHROW_LOCK(ih); 3876 count = 0; 3877 LIST_FOREACH(s, &ih->states, entry) { 3878 if (s->timeout == PFTM_UNLINKED) 3879 continue; 3880 count++; 3881 } 3882 3883 if (count > slice_count) { 3884 PF_HASHROW_UNLOCK(ih); 3885 free(pstore, M_TEMP); 3886 slice_count = count * 2; 3887 pstore = mallocarray(slice_count, 3888 sizeof(struct pf_state_export), M_TEMP, 3889 M_WAITOK | M_ZERO); 3890 goto DIOCGETSTATESV2_retry; 3891 } 3892 3893 if ((nr+count) * sizeof(*p) > ps->ps_len) { 3894 PF_HASHROW_UNLOCK(ih); 3895 goto DIOCGETSTATESV2_full; 3896 } 3897 3898 LIST_FOREACH(s, &ih->states, entry) { 3899 if (s->timeout == PFTM_UNLINKED) 3900 continue; 3901 3902 pf_state_export(p, s); 3903 p++; 3904 nr++; 3905 } 3906 PF_HASHROW_UNLOCK(ih); 3907 error = copyout(pstore, out, 3908 sizeof(struct pf_state_export) * count); 3909 if (error) 3910 break; 3911 out = ps->ps_states + nr; 3912 } 3913 DIOCGETSTATESV2_full: 3914 ps->ps_len = nr * sizeof(struct pf_state_export); 3915 free(pstore, M_TEMP); 3916 3917 break; 3918 } 3919 3920 case DIOCGETSTATUS: { 3921 struct pf_status *s = (struct pf_status *)addr; 3922 3923 PF_RULES_RLOCK(); 3924 s->running = V_pf_status.running; 3925 s->since = V_pf_status.since; 3926 s->debug = V_pf_status.debug; 3927 s->hostid = V_pf_status.hostid; 3928 s->states = V_pf_status.states; 3929 s->src_nodes = V_pf_status.src_nodes; 3930 3931 for (int i = 0; i < PFRES_MAX; i++) 3932 s->counters[i] = 3933 counter_u64_fetch(V_pf_status.counters[i]); 3934 for (int i = 0; i < LCNT_MAX; i++) 3935 s->lcounters[i] = 3936 counter_u64_fetch(V_pf_status.lcounters[i]); 3937 for (int i = 0; i < FCNT_MAX; i++) 3938 s->fcounters[i] = 3939 pf_counter_u64_fetch(&V_pf_status.fcounters[i]); 3940 for (int i = 0; i < SCNT_MAX; i++) 3941 s->scounters[i] = 3942 counter_u64_fetch(V_pf_status.scounters[i]); 3943 3944 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ); 3945 bcopy(V_pf_status.pf_chksum, s->pf_chksum, 3946 PF_MD5_DIGEST_LENGTH); 3947 3948 pfi_update_status(s->ifname, s); 3949 PF_RULES_RUNLOCK(); 3950 break; 3951 } 3952 3953 case DIOCGETSTATUSNV: { 3954 error = pf_getstatus((struct pfioc_nv *)addr); 3955 break; 3956 } 3957 3958 case DIOCSETSTATUSIF: { 3959 struct pfioc_if *pi = (struct pfioc_if *)addr; 3960 3961 if (pi->ifname[0] == 0) { 3962 bzero(V_pf_status.ifname, IFNAMSIZ); 3963 break; 3964 } 3965 PF_RULES_WLOCK(); 3966 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); 3967 PF_RULES_WUNLOCK(); 3968 break; 3969 } 3970 3971 case DIOCCLRSTATUS: { 3972 PF_RULES_WLOCK(); 3973 for (int i = 0; i < PFRES_MAX; i++) 3974 counter_u64_zero(V_pf_status.counters[i]); 3975 for (int i = 0; i < FCNT_MAX; i++) 3976 pf_counter_u64_zero(&V_pf_status.fcounters[i]); 3977 for (int i = 0; i < SCNT_MAX; i++) 3978 counter_u64_zero(V_pf_status.scounters[i]); 3979 for (int i = 0; i < KLCNT_MAX; i++) 3980 counter_u64_zero(V_pf_status.lcounters[i]); 3981 V_pf_status.since = time_second; 3982 if (*V_pf_status.ifname) 3983 pfi_update_status(V_pf_status.ifname, NULL); 3984 PF_RULES_WUNLOCK(); 3985 break; 3986 } 3987 3988 case DIOCNATLOOK: { 3989 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; 3990 struct pf_state_key *sk; 3991 struct pf_kstate *state; 3992 struct pf_state_key_cmp key; 3993 int m = 0, direction = pnl->direction; 3994 int sidx, didx; 3995 3996 /* NATLOOK src and dst are reversed, so reverse sidx/didx */ 3997 sidx = (direction == PF_IN) ? 1 : 0; 3998 didx = (direction == PF_IN) ? 0 : 1; 3999 4000 if (!pnl->proto || 4001 PF_AZERO(&pnl->saddr, pnl->af) || 4002 PF_AZERO(&pnl->daddr, pnl->af) || 4003 ((pnl->proto == IPPROTO_TCP || 4004 pnl->proto == IPPROTO_UDP) && 4005 (!pnl->dport || !pnl->sport))) 4006 error = EINVAL; 4007 else { 4008 bzero(&key, sizeof(key)); 4009 key.af = pnl->af; 4010 key.proto = pnl->proto; 4011 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af); 4012 key.port[sidx] = pnl->sport; 4013 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af); 4014 key.port[didx] = pnl->dport; 4015 4016 state = pf_find_state_all(&key, direction, &m); 4017 if (state == NULL) { 4018 error = ENOENT; 4019 } else { 4020 if (m > 1) { 4021 PF_STATE_UNLOCK(state); 4022 error = E2BIG; /* more than one state */ 4023 } else { 4024 sk = state->key[sidx]; 4025 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af); 4026 pnl->rsport = sk->port[sidx]; 4027 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af); 4028 pnl->rdport = sk->port[didx]; 4029 PF_STATE_UNLOCK(state); 4030 } 4031 } 4032 } 4033 break; 4034 } 4035 4036 case DIOCSETTIMEOUT: { 4037 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4038 int old; 4039 4040 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX || 4041 pt->seconds < 0) { 4042 error = EINVAL; 4043 break; 4044 } 4045 PF_RULES_WLOCK(); 4046 old = V_pf_default_rule.timeout[pt->timeout]; 4047 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0) 4048 pt->seconds = 1; 4049 V_pf_default_rule.timeout[pt->timeout] = pt->seconds; 4050 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old) 4051 wakeup(pf_purge_thread); 4052 pt->seconds = old; 4053 PF_RULES_WUNLOCK(); 4054 break; 4055 } 4056 4057 case DIOCGETTIMEOUT: { 4058 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4059 4060 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) { 4061 error = EINVAL; 4062 break; 4063 } 4064 PF_RULES_RLOCK(); 4065 pt->seconds = V_pf_default_rule.timeout[pt->timeout]; 4066 PF_RULES_RUNLOCK(); 4067 break; 4068 } 4069 4070 case DIOCGETLIMIT: { 4071 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4072 4073 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) { 4074 error = EINVAL; 4075 break; 4076 } 4077 PF_RULES_RLOCK(); 4078 pl->limit = V_pf_limits[pl->index].limit; 4079 PF_RULES_RUNLOCK(); 4080 break; 4081 } 4082 4083 case DIOCSETLIMIT: { 4084 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4085 int old_limit; 4086 4087 PF_RULES_WLOCK(); 4088 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX || 4089 V_pf_limits[pl->index].zone == NULL) { 4090 PF_RULES_WUNLOCK(); 4091 error = EINVAL; 4092 break; 4093 } 4094 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit); 4095 old_limit = V_pf_limits[pl->index].limit; 4096 V_pf_limits[pl->index].limit = pl->limit; 4097 pl->limit = old_limit; 4098 PF_RULES_WUNLOCK(); 4099 break; 4100 } 4101 4102 case DIOCSETDEBUG: { 4103 u_int32_t *level = (u_int32_t *)addr; 4104 4105 PF_RULES_WLOCK(); 4106 V_pf_status.debug = *level; 4107 PF_RULES_WUNLOCK(); 4108 break; 4109 } 4110 4111 case DIOCCLRRULECTRS: { 4112 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ 4113 struct pf_kruleset *ruleset = &pf_main_ruleset; 4114 struct pf_krule *rule; 4115 4116 PF_RULES_WLOCK(); 4117 TAILQ_FOREACH(rule, 4118 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { 4119 pf_counter_u64_zero(&rule->evaluations); 4120 for (int i = 0; i < 2; i++) { 4121 pf_counter_u64_zero(&rule->packets[i]); 4122 pf_counter_u64_zero(&rule->bytes[i]); 4123 } 4124 } 4125 PF_RULES_WUNLOCK(); 4126 break; 4127 } 4128 4129 case DIOCGIFSPEEDV0: 4130 case DIOCGIFSPEEDV1: { 4131 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr; 4132 struct pf_ifspeed_v1 ps; 4133 struct ifnet *ifp; 4134 4135 if (psp->ifname[0] == '\0') { 4136 error = EINVAL; 4137 break; 4138 } 4139 4140 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); 4141 if (error != 0) 4142 break; 4143 ifp = ifunit(ps.ifname); 4144 if (ifp != NULL) { 4145 psp->baudrate32 = 4146 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX); 4147 if (cmd == DIOCGIFSPEEDV1) 4148 psp->baudrate = ifp->if_baudrate; 4149 } else { 4150 error = EINVAL; 4151 } 4152 break; 4153 } 4154 4155 #ifdef ALTQ 4156 case DIOCSTARTALTQ: { 4157 struct pf_altq *altq; 4158 4159 PF_RULES_WLOCK(); 4160 /* enable all altq interfaces on active list */ 4161 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4162 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4163 error = pf_enable_altq(altq); 4164 if (error != 0) 4165 break; 4166 } 4167 } 4168 if (error == 0) 4169 V_pf_altq_running = 1; 4170 PF_RULES_WUNLOCK(); 4171 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n")); 4172 break; 4173 } 4174 4175 case DIOCSTOPALTQ: { 4176 struct pf_altq *altq; 4177 4178 PF_RULES_WLOCK(); 4179 /* disable all altq interfaces on active list */ 4180 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4181 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4182 error = pf_disable_altq(altq); 4183 if (error != 0) 4184 break; 4185 } 4186 } 4187 if (error == 0) 4188 V_pf_altq_running = 0; 4189 PF_RULES_WUNLOCK(); 4190 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n")); 4191 break; 4192 } 4193 4194 case DIOCADDALTQV0: 4195 case DIOCADDALTQV1: { 4196 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4197 struct pf_altq *altq, *a; 4198 struct ifnet *ifp; 4199 4200 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); 4201 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); 4202 if (error) 4203 break; 4204 altq->local_flags = 0; 4205 4206 PF_RULES_WLOCK(); 4207 if (pa->ticket != V_ticket_altqs_inactive) { 4208 PF_RULES_WUNLOCK(); 4209 free(altq, M_PFALTQ); 4210 error = EBUSY; 4211 break; 4212 } 4213 4214 /* 4215 * if this is for a queue, find the discipline and 4216 * copy the necessary fields 4217 */ 4218 if (altq->qname[0] != 0) { 4219 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) { 4220 PF_RULES_WUNLOCK(); 4221 error = EBUSY; 4222 free(altq, M_PFALTQ); 4223 break; 4224 } 4225 altq->altq_disc = NULL; 4226 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { 4227 if (strncmp(a->ifname, altq->ifname, 4228 IFNAMSIZ) == 0) { 4229 altq->altq_disc = a->altq_disc; 4230 break; 4231 } 4232 } 4233 } 4234 4235 if ((ifp = ifunit(altq->ifname)) == NULL) 4236 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 4237 else 4238 error = altq_add(ifp, altq); 4239 4240 if (error) { 4241 PF_RULES_WUNLOCK(); 4242 free(altq, M_PFALTQ); 4243 break; 4244 } 4245 4246 if (altq->qname[0] != 0) 4247 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); 4248 else 4249 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries); 4250 /* version error check done on import above */ 4251 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4252 PF_RULES_WUNLOCK(); 4253 break; 4254 } 4255 4256 case DIOCGETALTQSV0: 4257 case DIOCGETALTQSV1: { 4258 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4259 struct pf_altq *altq; 4260 4261 PF_RULES_RLOCK(); 4262 pa->nr = 0; 4263 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) 4264 pa->nr++; 4265 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) 4266 pa->nr++; 4267 pa->ticket = V_ticket_altqs_active; 4268 PF_RULES_RUNLOCK(); 4269 break; 4270 } 4271 4272 case DIOCGETALTQV0: 4273 case DIOCGETALTQV1: { 4274 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4275 struct pf_altq *altq; 4276 4277 PF_RULES_RLOCK(); 4278 if (pa->ticket != V_ticket_altqs_active) { 4279 PF_RULES_RUNLOCK(); 4280 error = EBUSY; 4281 break; 4282 } 4283 altq = pf_altq_get_nth_active(pa->nr); 4284 if (altq == NULL) { 4285 PF_RULES_RUNLOCK(); 4286 error = EBUSY; 4287 break; 4288 } 4289 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4290 PF_RULES_RUNLOCK(); 4291 break; 4292 } 4293 4294 case DIOCCHANGEALTQV0: 4295 case DIOCCHANGEALTQV1: 4296 /* CHANGEALTQ not supported yet! */ 4297 error = ENODEV; 4298 break; 4299 4300 case DIOCGETQSTATSV0: 4301 case DIOCGETQSTATSV1: { 4302 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr; 4303 struct pf_altq *altq; 4304 int nbytes; 4305 u_int32_t version; 4306 4307 PF_RULES_RLOCK(); 4308 if (pq->ticket != V_ticket_altqs_active) { 4309 PF_RULES_RUNLOCK(); 4310 error = EBUSY; 4311 break; 4312 } 4313 nbytes = pq->nbytes; 4314 altq = pf_altq_get_nth_active(pq->nr); 4315 if (altq == NULL) { 4316 PF_RULES_RUNLOCK(); 4317 error = EBUSY; 4318 break; 4319 } 4320 4321 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { 4322 PF_RULES_RUNLOCK(); 4323 error = ENXIO; 4324 break; 4325 } 4326 PF_RULES_RUNLOCK(); 4327 if (cmd == DIOCGETQSTATSV0) 4328 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */ 4329 else 4330 version = pq->version; 4331 error = altq_getqstats(altq, pq->buf, &nbytes, version); 4332 if (error == 0) { 4333 pq->scheduler = altq->scheduler; 4334 pq->nbytes = nbytes; 4335 } 4336 break; 4337 } 4338 #endif /* ALTQ */ 4339 4340 case DIOCBEGINADDRS: { 4341 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4342 4343 PF_RULES_WLOCK(); 4344 pf_empty_kpool(&V_pf_pabuf); 4345 pp->ticket = ++V_ticket_pabuf; 4346 PF_RULES_WUNLOCK(); 4347 break; 4348 } 4349 4350 case DIOCADDADDR: { 4351 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4352 struct pf_kpooladdr *pa; 4353 struct pfi_kkif *kif = NULL; 4354 4355 #ifndef INET 4356 if (pp->af == AF_INET) { 4357 error = EAFNOSUPPORT; 4358 break; 4359 } 4360 #endif /* INET */ 4361 #ifndef INET6 4362 if (pp->af == AF_INET6) { 4363 error = EAFNOSUPPORT; 4364 break; 4365 } 4366 #endif /* INET6 */ 4367 if (pp->addr.addr.type != PF_ADDR_ADDRMASK && 4368 pp->addr.addr.type != PF_ADDR_DYNIFTL && 4369 pp->addr.addr.type != PF_ADDR_TABLE) { 4370 error = EINVAL; 4371 break; 4372 } 4373 if (pp->addr.addr.p.dyn != NULL) { 4374 error = EINVAL; 4375 break; 4376 } 4377 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); 4378 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa); 4379 if (error != 0) 4380 break; 4381 if (pa->ifname[0]) 4382 kif = pf_kkif_create(M_WAITOK); 4383 PF_RULES_WLOCK(); 4384 if (pp->ticket != V_ticket_pabuf) { 4385 PF_RULES_WUNLOCK(); 4386 if (pa->ifname[0]) 4387 pf_kkif_free(kif); 4388 free(pa, M_PFRULE); 4389 error = EBUSY; 4390 break; 4391 } 4392 if (pa->ifname[0]) { 4393 pa->kif = pfi_kkif_attach(kif, pa->ifname); 4394 kif = NULL; 4395 pfi_kkif_ref(pa->kif); 4396 } else 4397 pa->kif = NULL; 4398 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = 4399 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { 4400 if (pa->ifname[0]) 4401 pfi_kkif_unref(pa->kif); 4402 PF_RULES_WUNLOCK(); 4403 free(pa, M_PFRULE); 4404 break; 4405 } 4406 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries); 4407 PF_RULES_WUNLOCK(); 4408 break; 4409 } 4410 4411 case DIOCGETADDRS: { 4412 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4413 struct pf_kpool *pool; 4414 struct pf_kpooladdr *pa; 4415 4416 pp->anchor[sizeof(pp->anchor) - 1] = 0; 4417 pp->nr = 0; 4418 4419 PF_RULES_RLOCK(); 4420 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 4421 pp->r_num, 0, 1, 0); 4422 if (pool == NULL) { 4423 PF_RULES_RUNLOCK(); 4424 error = EBUSY; 4425 break; 4426 } 4427 TAILQ_FOREACH(pa, &pool->list, entries) 4428 pp->nr++; 4429 PF_RULES_RUNLOCK(); 4430 break; 4431 } 4432 4433 case DIOCGETADDR: { 4434 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4435 struct pf_kpool *pool; 4436 struct pf_kpooladdr *pa; 4437 u_int32_t nr = 0; 4438 4439 pp->anchor[sizeof(pp->anchor) - 1] = 0; 4440 4441 PF_RULES_RLOCK(); 4442 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 4443 pp->r_num, 0, 1, 1); 4444 if (pool == NULL) { 4445 PF_RULES_RUNLOCK(); 4446 error = EBUSY; 4447 break; 4448 } 4449 pa = TAILQ_FIRST(&pool->list); 4450 while ((pa != NULL) && (nr < pp->nr)) { 4451 pa = TAILQ_NEXT(pa, entries); 4452 nr++; 4453 } 4454 if (pa == NULL) { 4455 PF_RULES_RUNLOCK(); 4456 error = EBUSY; 4457 break; 4458 } 4459 pf_kpooladdr_to_pooladdr(pa, &pp->addr); 4460 pf_addr_copyout(&pp->addr.addr); 4461 PF_RULES_RUNLOCK(); 4462 break; 4463 } 4464 4465 case DIOCCHANGEADDR: { 4466 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; 4467 struct pf_kpool *pool; 4468 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL; 4469 struct pf_kruleset *ruleset; 4470 struct pfi_kkif *kif = NULL; 4471 4472 pca->anchor[sizeof(pca->anchor) - 1] = 0; 4473 4474 if (pca->action < PF_CHANGE_ADD_HEAD || 4475 pca->action > PF_CHANGE_REMOVE) { 4476 error = EINVAL; 4477 break; 4478 } 4479 if (pca->addr.addr.type != PF_ADDR_ADDRMASK && 4480 pca->addr.addr.type != PF_ADDR_DYNIFTL && 4481 pca->addr.addr.type != PF_ADDR_TABLE) { 4482 error = EINVAL; 4483 break; 4484 } 4485 if (pca->addr.addr.p.dyn != NULL) { 4486 error = EINVAL; 4487 break; 4488 } 4489 4490 if (pca->action != PF_CHANGE_REMOVE) { 4491 #ifndef INET 4492 if (pca->af == AF_INET) { 4493 error = EAFNOSUPPORT; 4494 break; 4495 } 4496 #endif /* INET */ 4497 #ifndef INET6 4498 if (pca->af == AF_INET6) { 4499 error = EAFNOSUPPORT; 4500 break; 4501 } 4502 #endif /* INET6 */ 4503 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); 4504 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); 4505 if (newpa->ifname[0]) 4506 kif = pf_kkif_create(M_WAITOK); 4507 newpa->kif = NULL; 4508 } 4509 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x) 4510 PF_RULES_WLOCK(); 4511 ruleset = pf_find_kruleset(pca->anchor); 4512 if (ruleset == NULL) 4513 ERROUT(EBUSY); 4514 4515 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action, 4516 pca->r_num, pca->r_last, 1, 1); 4517 if (pool == NULL) 4518 ERROUT(EBUSY); 4519 4520 if (pca->action != PF_CHANGE_REMOVE) { 4521 if (newpa->ifname[0]) { 4522 newpa->kif = pfi_kkif_attach(kif, newpa->ifname); 4523 pfi_kkif_ref(newpa->kif); 4524 kif = NULL; 4525 } 4526 4527 switch (newpa->addr.type) { 4528 case PF_ADDR_DYNIFTL: 4529 error = pfi_dynaddr_setup(&newpa->addr, 4530 pca->af); 4531 break; 4532 case PF_ADDR_TABLE: 4533 newpa->addr.p.tbl = pfr_attach_table(ruleset, 4534 newpa->addr.v.tblname); 4535 if (newpa->addr.p.tbl == NULL) 4536 error = ENOMEM; 4537 break; 4538 } 4539 if (error) 4540 goto DIOCCHANGEADDR_error; 4541 } 4542 4543 switch (pca->action) { 4544 case PF_CHANGE_ADD_HEAD: 4545 oldpa = TAILQ_FIRST(&pool->list); 4546 break; 4547 case PF_CHANGE_ADD_TAIL: 4548 oldpa = TAILQ_LAST(&pool->list, pf_kpalist); 4549 break; 4550 default: 4551 oldpa = TAILQ_FIRST(&pool->list); 4552 for (int i = 0; oldpa && i < pca->nr; i++) 4553 oldpa = TAILQ_NEXT(oldpa, entries); 4554 4555 if (oldpa == NULL) 4556 ERROUT(EINVAL); 4557 } 4558 4559 if (pca->action == PF_CHANGE_REMOVE) { 4560 TAILQ_REMOVE(&pool->list, oldpa, entries); 4561 switch (oldpa->addr.type) { 4562 case PF_ADDR_DYNIFTL: 4563 pfi_dynaddr_remove(oldpa->addr.p.dyn); 4564 break; 4565 case PF_ADDR_TABLE: 4566 pfr_detach_table(oldpa->addr.p.tbl); 4567 break; 4568 } 4569 if (oldpa->kif) 4570 pfi_kkif_unref(oldpa->kif); 4571 free(oldpa, M_PFRULE); 4572 } else { 4573 if (oldpa == NULL) 4574 TAILQ_INSERT_TAIL(&pool->list, newpa, entries); 4575 else if (pca->action == PF_CHANGE_ADD_HEAD || 4576 pca->action == PF_CHANGE_ADD_BEFORE) 4577 TAILQ_INSERT_BEFORE(oldpa, newpa, entries); 4578 else 4579 TAILQ_INSERT_AFTER(&pool->list, oldpa, 4580 newpa, entries); 4581 } 4582 4583 pool->cur = TAILQ_FIRST(&pool->list); 4584 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); 4585 PF_RULES_WUNLOCK(); 4586 break; 4587 4588 #undef ERROUT 4589 DIOCCHANGEADDR_error: 4590 if (newpa != NULL) { 4591 if (newpa->kif) 4592 pfi_kkif_unref(newpa->kif); 4593 free(newpa, M_PFRULE); 4594 } 4595 PF_RULES_WUNLOCK(); 4596 pf_kkif_free(kif); 4597 break; 4598 } 4599 4600 case DIOCGETRULESETS: { 4601 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4602 struct pf_kruleset *ruleset; 4603 struct pf_kanchor *anchor; 4604 4605 pr->path[sizeof(pr->path) - 1] = 0; 4606 4607 PF_RULES_RLOCK(); 4608 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 4609 PF_RULES_RUNLOCK(); 4610 error = ENOENT; 4611 break; 4612 } 4613 pr->nr = 0; 4614 if (ruleset->anchor == NULL) { 4615 /* XXX kludge for pf_main_ruleset */ 4616 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 4617 if (anchor->parent == NULL) 4618 pr->nr++; 4619 } else { 4620 RB_FOREACH(anchor, pf_kanchor_node, 4621 &ruleset->anchor->children) 4622 pr->nr++; 4623 } 4624 PF_RULES_RUNLOCK(); 4625 break; 4626 } 4627 4628 case DIOCGETRULESET: { 4629 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4630 struct pf_kruleset *ruleset; 4631 struct pf_kanchor *anchor; 4632 u_int32_t nr = 0; 4633 4634 pr->path[sizeof(pr->path) - 1] = 0; 4635 4636 PF_RULES_RLOCK(); 4637 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 4638 PF_RULES_RUNLOCK(); 4639 error = ENOENT; 4640 break; 4641 } 4642 pr->name[0] = 0; 4643 if (ruleset->anchor == NULL) { 4644 /* XXX kludge for pf_main_ruleset */ 4645 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 4646 if (anchor->parent == NULL && nr++ == pr->nr) { 4647 strlcpy(pr->name, anchor->name, 4648 sizeof(pr->name)); 4649 break; 4650 } 4651 } else { 4652 RB_FOREACH(anchor, pf_kanchor_node, 4653 &ruleset->anchor->children) 4654 if (nr++ == pr->nr) { 4655 strlcpy(pr->name, anchor->name, 4656 sizeof(pr->name)); 4657 break; 4658 } 4659 } 4660 if (!pr->name[0]) 4661 error = EBUSY; 4662 PF_RULES_RUNLOCK(); 4663 break; 4664 } 4665 4666 case DIOCRCLRTABLES: { 4667 struct pfioc_table *io = (struct pfioc_table *)addr; 4668 4669 if (io->pfrio_esize != 0) { 4670 error = ENODEV; 4671 break; 4672 } 4673 PF_RULES_WLOCK(); 4674 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, 4675 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4676 PF_RULES_WUNLOCK(); 4677 break; 4678 } 4679 4680 case DIOCRADDTABLES: { 4681 struct pfioc_table *io = (struct pfioc_table *)addr; 4682 struct pfr_table *pfrts; 4683 size_t totlen; 4684 4685 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4686 error = ENODEV; 4687 break; 4688 } 4689 4690 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4691 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4692 error = ENOMEM; 4693 break; 4694 } 4695 4696 totlen = io->pfrio_size * sizeof(struct pfr_table); 4697 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4698 M_TEMP, M_WAITOK); 4699 error = copyin(io->pfrio_buffer, pfrts, totlen); 4700 if (error) { 4701 free(pfrts, M_TEMP); 4702 break; 4703 } 4704 PF_RULES_WLOCK(); 4705 error = pfr_add_tables(pfrts, io->pfrio_size, 4706 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4707 PF_RULES_WUNLOCK(); 4708 free(pfrts, M_TEMP); 4709 break; 4710 } 4711 4712 case DIOCRDELTABLES: { 4713 struct pfioc_table *io = (struct pfioc_table *)addr; 4714 struct pfr_table *pfrts; 4715 size_t totlen; 4716 4717 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4718 error = ENODEV; 4719 break; 4720 } 4721 4722 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4723 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4724 error = ENOMEM; 4725 break; 4726 } 4727 4728 totlen = io->pfrio_size * sizeof(struct pfr_table); 4729 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4730 M_TEMP, M_WAITOK); 4731 error = copyin(io->pfrio_buffer, pfrts, totlen); 4732 if (error) { 4733 free(pfrts, M_TEMP); 4734 break; 4735 } 4736 PF_RULES_WLOCK(); 4737 error = pfr_del_tables(pfrts, io->pfrio_size, 4738 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4739 PF_RULES_WUNLOCK(); 4740 free(pfrts, M_TEMP); 4741 break; 4742 } 4743 4744 case DIOCRGETTABLES: { 4745 struct pfioc_table *io = (struct pfioc_table *)addr; 4746 struct pfr_table *pfrts; 4747 size_t totlen; 4748 int n; 4749 4750 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4751 error = ENODEV; 4752 break; 4753 } 4754 PF_RULES_RLOCK(); 4755 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4756 if (n < 0) { 4757 PF_RULES_RUNLOCK(); 4758 error = EINVAL; 4759 break; 4760 } 4761 io->pfrio_size = min(io->pfrio_size, n); 4762 4763 totlen = io->pfrio_size * sizeof(struct pfr_table); 4764 4765 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4766 M_TEMP, M_NOWAIT | M_ZERO); 4767 if (pfrts == NULL) { 4768 error = ENOMEM; 4769 PF_RULES_RUNLOCK(); 4770 break; 4771 } 4772 error = pfr_get_tables(&io->pfrio_table, pfrts, 4773 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4774 PF_RULES_RUNLOCK(); 4775 if (error == 0) 4776 error = copyout(pfrts, io->pfrio_buffer, totlen); 4777 free(pfrts, M_TEMP); 4778 break; 4779 } 4780 4781 case DIOCRGETTSTATS: { 4782 struct pfioc_table *io = (struct pfioc_table *)addr; 4783 struct pfr_tstats *pfrtstats; 4784 size_t totlen; 4785 int n; 4786 4787 if (io->pfrio_esize != sizeof(struct pfr_tstats)) { 4788 error = ENODEV; 4789 break; 4790 } 4791 PF_TABLE_STATS_LOCK(); 4792 PF_RULES_RLOCK(); 4793 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4794 if (n < 0) { 4795 PF_RULES_RUNLOCK(); 4796 PF_TABLE_STATS_UNLOCK(); 4797 error = EINVAL; 4798 break; 4799 } 4800 io->pfrio_size = min(io->pfrio_size, n); 4801 4802 totlen = io->pfrio_size * sizeof(struct pfr_tstats); 4803 pfrtstats = mallocarray(io->pfrio_size, 4804 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO); 4805 if (pfrtstats == NULL) { 4806 error = ENOMEM; 4807 PF_RULES_RUNLOCK(); 4808 PF_TABLE_STATS_UNLOCK(); 4809 break; 4810 } 4811 error = pfr_get_tstats(&io->pfrio_table, pfrtstats, 4812 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4813 PF_RULES_RUNLOCK(); 4814 PF_TABLE_STATS_UNLOCK(); 4815 if (error == 0) 4816 error = copyout(pfrtstats, io->pfrio_buffer, totlen); 4817 free(pfrtstats, M_TEMP); 4818 break; 4819 } 4820 4821 case DIOCRCLRTSTATS: { 4822 struct pfioc_table *io = (struct pfioc_table *)addr; 4823 struct pfr_table *pfrts; 4824 size_t totlen; 4825 4826 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4827 error = ENODEV; 4828 break; 4829 } 4830 4831 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4832 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4833 /* We used to count tables and use the minimum required 4834 * size, so we didn't fail on overly large requests. 4835 * Keep doing so. */ 4836 io->pfrio_size = pf_ioctl_maxcount; 4837 break; 4838 } 4839 4840 totlen = io->pfrio_size * sizeof(struct pfr_table); 4841 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4842 M_TEMP, M_WAITOK); 4843 error = copyin(io->pfrio_buffer, pfrts, totlen); 4844 if (error) { 4845 free(pfrts, M_TEMP); 4846 break; 4847 } 4848 4849 PF_TABLE_STATS_LOCK(); 4850 PF_RULES_RLOCK(); 4851 error = pfr_clr_tstats(pfrts, io->pfrio_size, 4852 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4853 PF_RULES_RUNLOCK(); 4854 PF_TABLE_STATS_UNLOCK(); 4855 free(pfrts, M_TEMP); 4856 break; 4857 } 4858 4859 case DIOCRSETTFLAGS: { 4860 struct pfioc_table *io = (struct pfioc_table *)addr; 4861 struct pfr_table *pfrts; 4862 size_t totlen; 4863 int n; 4864 4865 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4866 error = ENODEV; 4867 break; 4868 } 4869 4870 PF_RULES_RLOCK(); 4871 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4872 if (n < 0) { 4873 PF_RULES_RUNLOCK(); 4874 error = EINVAL; 4875 break; 4876 } 4877 4878 io->pfrio_size = min(io->pfrio_size, n); 4879 PF_RULES_RUNLOCK(); 4880 4881 totlen = io->pfrio_size * sizeof(struct pfr_table); 4882 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4883 M_TEMP, M_WAITOK); 4884 error = copyin(io->pfrio_buffer, pfrts, totlen); 4885 if (error) { 4886 free(pfrts, M_TEMP); 4887 break; 4888 } 4889 PF_RULES_WLOCK(); 4890 error = pfr_set_tflags(pfrts, io->pfrio_size, 4891 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, 4892 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4893 PF_RULES_WUNLOCK(); 4894 free(pfrts, M_TEMP); 4895 break; 4896 } 4897 4898 case DIOCRCLRADDRS: { 4899 struct pfioc_table *io = (struct pfioc_table *)addr; 4900 4901 if (io->pfrio_esize != 0) { 4902 error = ENODEV; 4903 break; 4904 } 4905 PF_RULES_WLOCK(); 4906 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, 4907 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4908 PF_RULES_WUNLOCK(); 4909 break; 4910 } 4911 4912 case DIOCRADDADDRS: { 4913 struct pfioc_table *io = (struct pfioc_table *)addr; 4914 struct pfr_addr *pfras; 4915 size_t totlen; 4916 4917 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4918 error = ENODEV; 4919 break; 4920 } 4921 if (io->pfrio_size < 0 || 4922 io->pfrio_size > pf_ioctl_maxcount || 4923 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4924 error = EINVAL; 4925 break; 4926 } 4927 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4928 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4929 M_TEMP, M_WAITOK); 4930 error = copyin(io->pfrio_buffer, pfras, totlen); 4931 if (error) { 4932 free(pfras, M_TEMP); 4933 break; 4934 } 4935 PF_RULES_WLOCK(); 4936 error = pfr_add_addrs(&io->pfrio_table, pfras, 4937 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | 4938 PFR_FLAG_USERIOCTL); 4939 PF_RULES_WUNLOCK(); 4940 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4941 error = copyout(pfras, io->pfrio_buffer, totlen); 4942 free(pfras, M_TEMP); 4943 break; 4944 } 4945 4946 case DIOCRDELADDRS: { 4947 struct pfioc_table *io = (struct pfioc_table *)addr; 4948 struct pfr_addr *pfras; 4949 size_t totlen; 4950 4951 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4952 error = ENODEV; 4953 break; 4954 } 4955 if (io->pfrio_size < 0 || 4956 io->pfrio_size > pf_ioctl_maxcount || 4957 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4958 error = EINVAL; 4959 break; 4960 } 4961 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4962 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4963 M_TEMP, M_WAITOK); 4964 error = copyin(io->pfrio_buffer, pfras, totlen); 4965 if (error) { 4966 free(pfras, M_TEMP); 4967 break; 4968 } 4969 PF_RULES_WLOCK(); 4970 error = pfr_del_addrs(&io->pfrio_table, pfras, 4971 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | 4972 PFR_FLAG_USERIOCTL); 4973 PF_RULES_WUNLOCK(); 4974 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4975 error = copyout(pfras, io->pfrio_buffer, totlen); 4976 free(pfras, M_TEMP); 4977 break; 4978 } 4979 4980 case DIOCRSETADDRS: { 4981 struct pfioc_table *io = (struct pfioc_table *)addr; 4982 struct pfr_addr *pfras; 4983 size_t totlen, count; 4984 4985 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4986 error = ENODEV; 4987 break; 4988 } 4989 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { 4990 error = EINVAL; 4991 break; 4992 } 4993 count = max(io->pfrio_size, io->pfrio_size2); 4994 if (count > pf_ioctl_maxcount || 4995 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { 4996 error = EINVAL; 4997 break; 4998 } 4999 totlen = count * sizeof(struct pfr_addr); 5000 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP, 5001 M_WAITOK); 5002 error = copyin(io->pfrio_buffer, pfras, totlen); 5003 if (error) { 5004 free(pfras, M_TEMP); 5005 break; 5006 } 5007 PF_RULES_WLOCK(); 5008 error = pfr_set_addrs(&io->pfrio_table, pfras, 5009 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, 5010 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | 5011 PFR_FLAG_USERIOCTL, 0); 5012 PF_RULES_WUNLOCK(); 5013 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5014 error = copyout(pfras, io->pfrio_buffer, totlen); 5015 free(pfras, M_TEMP); 5016 break; 5017 } 5018 5019 case DIOCRGETADDRS: { 5020 struct pfioc_table *io = (struct pfioc_table *)addr; 5021 struct pfr_addr *pfras; 5022 size_t totlen; 5023 5024 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5025 error = ENODEV; 5026 break; 5027 } 5028 if (io->pfrio_size < 0 || 5029 io->pfrio_size > pf_ioctl_maxcount || 5030 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5031 error = EINVAL; 5032 break; 5033 } 5034 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5035 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5036 M_TEMP, M_WAITOK | M_ZERO); 5037 PF_RULES_RLOCK(); 5038 error = pfr_get_addrs(&io->pfrio_table, pfras, 5039 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5040 PF_RULES_RUNLOCK(); 5041 if (error == 0) 5042 error = copyout(pfras, io->pfrio_buffer, totlen); 5043 free(pfras, M_TEMP); 5044 break; 5045 } 5046 5047 case DIOCRGETASTATS: { 5048 struct pfioc_table *io = (struct pfioc_table *)addr; 5049 struct pfr_astats *pfrastats; 5050 size_t totlen; 5051 5052 if (io->pfrio_esize != sizeof(struct pfr_astats)) { 5053 error = ENODEV; 5054 break; 5055 } 5056 if (io->pfrio_size < 0 || 5057 io->pfrio_size > pf_ioctl_maxcount || 5058 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { 5059 error = EINVAL; 5060 break; 5061 } 5062 totlen = io->pfrio_size * sizeof(struct pfr_astats); 5063 pfrastats = mallocarray(io->pfrio_size, 5064 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO); 5065 PF_RULES_RLOCK(); 5066 error = pfr_get_astats(&io->pfrio_table, pfrastats, 5067 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5068 PF_RULES_RUNLOCK(); 5069 if (error == 0) 5070 error = copyout(pfrastats, io->pfrio_buffer, totlen); 5071 free(pfrastats, M_TEMP); 5072 break; 5073 } 5074 5075 case DIOCRCLRASTATS: { 5076 struct pfioc_table *io = (struct pfioc_table *)addr; 5077 struct pfr_addr *pfras; 5078 size_t totlen; 5079 5080 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5081 error = ENODEV; 5082 break; 5083 } 5084 if (io->pfrio_size < 0 || 5085 io->pfrio_size > pf_ioctl_maxcount || 5086 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5087 error = EINVAL; 5088 break; 5089 } 5090 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5091 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5092 M_TEMP, M_WAITOK); 5093 error = copyin(io->pfrio_buffer, pfras, totlen); 5094 if (error) { 5095 free(pfras, M_TEMP); 5096 break; 5097 } 5098 PF_RULES_WLOCK(); 5099 error = pfr_clr_astats(&io->pfrio_table, pfras, 5100 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | 5101 PFR_FLAG_USERIOCTL); 5102 PF_RULES_WUNLOCK(); 5103 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5104 error = copyout(pfras, io->pfrio_buffer, totlen); 5105 free(pfras, M_TEMP); 5106 break; 5107 } 5108 5109 case DIOCRTSTADDRS: { 5110 struct pfioc_table *io = (struct pfioc_table *)addr; 5111 struct pfr_addr *pfras; 5112 size_t totlen; 5113 5114 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5115 error = ENODEV; 5116 break; 5117 } 5118 if (io->pfrio_size < 0 || 5119 io->pfrio_size > pf_ioctl_maxcount || 5120 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5121 error = EINVAL; 5122 break; 5123 } 5124 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5125 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5126 M_TEMP, M_WAITOK); 5127 error = copyin(io->pfrio_buffer, pfras, totlen); 5128 if (error) { 5129 free(pfras, M_TEMP); 5130 break; 5131 } 5132 PF_RULES_RLOCK(); 5133 error = pfr_tst_addrs(&io->pfrio_table, pfras, 5134 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | 5135 PFR_FLAG_USERIOCTL); 5136 PF_RULES_RUNLOCK(); 5137 if (error == 0) 5138 error = copyout(pfras, io->pfrio_buffer, totlen); 5139 free(pfras, M_TEMP); 5140 break; 5141 } 5142 5143 case DIOCRINADEFINE: { 5144 struct pfioc_table *io = (struct pfioc_table *)addr; 5145 struct pfr_addr *pfras; 5146 size_t totlen; 5147 5148 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5149 error = ENODEV; 5150 break; 5151 } 5152 if (io->pfrio_size < 0 || 5153 io->pfrio_size > pf_ioctl_maxcount || 5154 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5155 error = EINVAL; 5156 break; 5157 } 5158 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5159 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5160 M_TEMP, M_WAITOK); 5161 error = copyin(io->pfrio_buffer, pfras, totlen); 5162 if (error) { 5163 free(pfras, M_TEMP); 5164 break; 5165 } 5166 PF_RULES_WLOCK(); 5167 error = pfr_ina_define(&io->pfrio_table, pfras, 5168 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, 5169 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5170 PF_RULES_WUNLOCK(); 5171 free(pfras, M_TEMP); 5172 break; 5173 } 5174 5175 case DIOCOSFPADD: { 5176 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5177 PF_RULES_WLOCK(); 5178 error = pf_osfp_add(io); 5179 PF_RULES_WUNLOCK(); 5180 break; 5181 } 5182 5183 case DIOCOSFPGET: { 5184 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5185 PF_RULES_RLOCK(); 5186 error = pf_osfp_get(io); 5187 PF_RULES_RUNLOCK(); 5188 break; 5189 } 5190 5191 case DIOCXBEGIN: { 5192 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5193 struct pfioc_trans_e *ioes, *ioe; 5194 size_t totlen; 5195 int i; 5196 5197 if (io->esize != sizeof(*ioe)) { 5198 error = ENODEV; 5199 break; 5200 } 5201 if (io->size < 0 || 5202 io->size > pf_ioctl_maxcount || 5203 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5204 error = EINVAL; 5205 break; 5206 } 5207 totlen = sizeof(struct pfioc_trans_e) * io->size; 5208 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5209 M_TEMP, M_WAITOK); 5210 error = copyin(io->array, ioes, totlen); 5211 if (error) { 5212 free(ioes, M_TEMP); 5213 break; 5214 } 5215 /* Ensure there's no more ethernet rules to clean up. */ 5216 NET_EPOCH_DRAIN_CALLBACKS(); 5217 PF_RULES_WLOCK(); 5218 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5219 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5220 switch (ioe->rs_num) { 5221 case PF_RULESET_ETH: 5222 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) { 5223 PF_RULES_WUNLOCK(); 5224 free(ioes, M_TEMP); 5225 goto fail; 5226 } 5227 break; 5228 #ifdef ALTQ 5229 case PF_RULESET_ALTQ: 5230 if (ioe->anchor[0]) { 5231 PF_RULES_WUNLOCK(); 5232 free(ioes, M_TEMP); 5233 error = EINVAL; 5234 goto fail; 5235 } 5236 if ((error = pf_begin_altq(&ioe->ticket))) { 5237 PF_RULES_WUNLOCK(); 5238 free(ioes, M_TEMP); 5239 goto fail; 5240 } 5241 break; 5242 #endif /* ALTQ */ 5243 case PF_RULESET_TABLE: 5244 { 5245 struct pfr_table table; 5246 5247 bzero(&table, sizeof(table)); 5248 strlcpy(table.pfrt_anchor, ioe->anchor, 5249 sizeof(table.pfrt_anchor)); 5250 if ((error = pfr_ina_begin(&table, 5251 &ioe->ticket, NULL, 0))) { 5252 PF_RULES_WUNLOCK(); 5253 free(ioes, M_TEMP); 5254 goto fail; 5255 } 5256 break; 5257 } 5258 default: 5259 if ((error = pf_begin_rules(&ioe->ticket, 5260 ioe->rs_num, ioe->anchor))) { 5261 PF_RULES_WUNLOCK(); 5262 free(ioes, M_TEMP); 5263 goto fail; 5264 } 5265 break; 5266 } 5267 } 5268 PF_RULES_WUNLOCK(); 5269 error = copyout(ioes, io->array, totlen); 5270 free(ioes, M_TEMP); 5271 break; 5272 } 5273 5274 case DIOCXROLLBACK: { 5275 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5276 struct pfioc_trans_e *ioe, *ioes; 5277 size_t totlen; 5278 int i; 5279 5280 if (io->esize != sizeof(*ioe)) { 5281 error = ENODEV; 5282 break; 5283 } 5284 if (io->size < 0 || 5285 io->size > pf_ioctl_maxcount || 5286 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5287 error = EINVAL; 5288 break; 5289 } 5290 totlen = sizeof(struct pfioc_trans_e) * io->size; 5291 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5292 M_TEMP, M_WAITOK); 5293 error = copyin(io->array, ioes, totlen); 5294 if (error) { 5295 free(ioes, M_TEMP); 5296 break; 5297 } 5298 PF_RULES_WLOCK(); 5299 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5300 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5301 switch (ioe->rs_num) { 5302 case PF_RULESET_ETH: 5303 if ((error = pf_rollback_eth(ioe->ticket, 5304 ioe->anchor))) { 5305 PF_RULES_WUNLOCK(); 5306 free(ioes, M_TEMP); 5307 goto fail; /* really bad */ 5308 } 5309 break; 5310 #ifdef ALTQ 5311 case PF_RULESET_ALTQ: 5312 if (ioe->anchor[0]) { 5313 PF_RULES_WUNLOCK(); 5314 free(ioes, M_TEMP); 5315 error = EINVAL; 5316 goto fail; 5317 } 5318 if ((error = pf_rollback_altq(ioe->ticket))) { 5319 PF_RULES_WUNLOCK(); 5320 free(ioes, M_TEMP); 5321 goto fail; /* really bad */ 5322 } 5323 break; 5324 #endif /* ALTQ */ 5325 case PF_RULESET_TABLE: 5326 { 5327 struct pfr_table table; 5328 5329 bzero(&table, sizeof(table)); 5330 strlcpy(table.pfrt_anchor, ioe->anchor, 5331 sizeof(table.pfrt_anchor)); 5332 if ((error = pfr_ina_rollback(&table, 5333 ioe->ticket, NULL, 0))) { 5334 PF_RULES_WUNLOCK(); 5335 free(ioes, M_TEMP); 5336 goto fail; /* really bad */ 5337 } 5338 break; 5339 } 5340 default: 5341 if ((error = pf_rollback_rules(ioe->ticket, 5342 ioe->rs_num, ioe->anchor))) { 5343 PF_RULES_WUNLOCK(); 5344 free(ioes, M_TEMP); 5345 goto fail; /* really bad */ 5346 } 5347 break; 5348 } 5349 } 5350 PF_RULES_WUNLOCK(); 5351 free(ioes, M_TEMP); 5352 break; 5353 } 5354 5355 case DIOCXCOMMIT: { 5356 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5357 struct pfioc_trans_e *ioe, *ioes; 5358 struct pf_kruleset *rs; 5359 struct pf_keth_ruleset *ers; 5360 size_t totlen; 5361 int i; 5362 5363 if (io->esize != sizeof(*ioe)) { 5364 error = ENODEV; 5365 break; 5366 } 5367 5368 if (io->size < 0 || 5369 io->size > pf_ioctl_maxcount || 5370 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5371 error = EINVAL; 5372 break; 5373 } 5374 5375 totlen = sizeof(struct pfioc_trans_e) * io->size; 5376 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5377 M_TEMP, M_WAITOK); 5378 error = copyin(io->array, ioes, totlen); 5379 if (error) { 5380 free(ioes, M_TEMP); 5381 break; 5382 } 5383 PF_RULES_WLOCK(); 5384 /* First makes sure everything will succeed. */ 5385 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5386 ioe->anchor[sizeof(ioe->anchor) - 1] = 0; 5387 switch (ioe->rs_num) { 5388 case PF_RULESET_ETH: 5389 ers = pf_find_keth_ruleset(ioe->anchor); 5390 if (ers == NULL || ioe->ticket == 0 || 5391 ioe->ticket != ers->inactive.ticket) { 5392 PF_RULES_WUNLOCK(); 5393 free(ioes, M_TEMP); 5394 error = EINVAL; 5395 goto fail; 5396 } 5397 break; 5398 #ifdef ALTQ 5399 case PF_RULESET_ALTQ: 5400 if (ioe->anchor[0]) { 5401 PF_RULES_WUNLOCK(); 5402 free(ioes, M_TEMP); 5403 error = EINVAL; 5404 goto fail; 5405 } 5406 if (!V_altqs_inactive_open || ioe->ticket != 5407 V_ticket_altqs_inactive) { 5408 PF_RULES_WUNLOCK(); 5409 free(ioes, M_TEMP); 5410 error = EBUSY; 5411 goto fail; 5412 } 5413 break; 5414 #endif /* ALTQ */ 5415 case PF_RULESET_TABLE: 5416 rs = pf_find_kruleset(ioe->anchor); 5417 if (rs == NULL || !rs->topen || ioe->ticket != 5418 rs->tticket) { 5419 PF_RULES_WUNLOCK(); 5420 free(ioes, M_TEMP); 5421 error = EBUSY; 5422 goto fail; 5423 } 5424 break; 5425 default: 5426 if (ioe->rs_num < 0 || ioe->rs_num >= 5427 PF_RULESET_MAX) { 5428 PF_RULES_WUNLOCK(); 5429 free(ioes, M_TEMP); 5430 error = EINVAL; 5431 goto fail; 5432 } 5433 rs = pf_find_kruleset(ioe->anchor); 5434 if (rs == NULL || 5435 !rs->rules[ioe->rs_num].inactive.open || 5436 rs->rules[ioe->rs_num].inactive.ticket != 5437 ioe->ticket) { 5438 PF_RULES_WUNLOCK(); 5439 free(ioes, M_TEMP); 5440 error = EBUSY; 5441 goto fail; 5442 } 5443 break; 5444 } 5445 } 5446 /* Now do the commit - no errors should happen here. */ 5447 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5448 switch (ioe->rs_num) { 5449 case PF_RULESET_ETH: 5450 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) { 5451 PF_RULES_WUNLOCK(); 5452 free(ioes, M_TEMP); 5453 goto fail; /* really bad */ 5454 } 5455 break; 5456 #ifdef ALTQ 5457 case PF_RULESET_ALTQ: 5458 if ((error = pf_commit_altq(ioe->ticket))) { 5459 PF_RULES_WUNLOCK(); 5460 free(ioes, M_TEMP); 5461 goto fail; /* really bad */ 5462 } 5463 break; 5464 #endif /* ALTQ */ 5465 case PF_RULESET_TABLE: 5466 { 5467 struct pfr_table table; 5468 5469 bzero(&table, sizeof(table)); 5470 (void)strlcpy(table.pfrt_anchor, ioe->anchor, 5471 sizeof(table.pfrt_anchor)); 5472 if ((error = pfr_ina_commit(&table, 5473 ioe->ticket, NULL, NULL, 0))) { 5474 PF_RULES_WUNLOCK(); 5475 free(ioes, M_TEMP); 5476 goto fail; /* really bad */ 5477 } 5478 break; 5479 } 5480 default: 5481 if ((error = pf_commit_rules(ioe->ticket, 5482 ioe->rs_num, ioe->anchor))) { 5483 PF_RULES_WUNLOCK(); 5484 free(ioes, M_TEMP); 5485 goto fail; /* really bad */ 5486 } 5487 break; 5488 } 5489 } 5490 PF_RULES_WUNLOCK(); 5491 5492 /* Only hook into EtherNet taffic if we've got rules for it. */ 5493 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 5494 hook_pf_eth(); 5495 else 5496 dehook_pf_eth(); 5497 5498 free(ioes, M_TEMP); 5499 break; 5500 } 5501 5502 case DIOCGETSRCNODES: { 5503 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; 5504 struct pf_srchash *sh; 5505 struct pf_ksrc_node *n; 5506 struct pf_src_node *p, *pstore; 5507 uint32_t i, nr = 0; 5508 5509 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; 5510 i++, sh++) { 5511 PF_HASHROW_LOCK(sh); 5512 LIST_FOREACH(n, &sh->nodes, entry) 5513 nr++; 5514 PF_HASHROW_UNLOCK(sh); 5515 } 5516 5517 psn->psn_len = min(psn->psn_len, 5518 sizeof(struct pf_src_node) * nr); 5519 5520 if (psn->psn_len == 0) { 5521 psn->psn_len = sizeof(struct pf_src_node) * nr; 5522 break; 5523 } 5524 5525 nr = 0; 5526 5527 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO); 5528 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; 5529 i++, sh++) { 5530 PF_HASHROW_LOCK(sh); 5531 LIST_FOREACH(n, &sh->nodes, entry) { 5532 5533 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) 5534 break; 5535 5536 pf_src_node_copy(n, p); 5537 5538 p++; 5539 nr++; 5540 } 5541 PF_HASHROW_UNLOCK(sh); 5542 } 5543 error = copyout(pstore, psn->psn_src_nodes, 5544 sizeof(struct pf_src_node) * nr); 5545 if (error) { 5546 free(pstore, M_TEMP); 5547 break; 5548 } 5549 psn->psn_len = sizeof(struct pf_src_node) * nr; 5550 free(pstore, M_TEMP); 5551 break; 5552 } 5553 5554 case DIOCCLRSRCNODES: { 5555 pf_clear_srcnodes(NULL); 5556 pf_purge_expired_src_nodes(); 5557 break; 5558 } 5559 5560 case DIOCKILLSRCNODES: 5561 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); 5562 break; 5563 5564 #ifdef COMPAT_FREEBSD13 5565 case DIOCKEEPCOUNTERS_FREEBSD13: 5566 #endif 5567 case DIOCKEEPCOUNTERS: 5568 error = pf_keepcounters((struct pfioc_nv *)addr); 5569 break; 5570 5571 case DIOCGETSYNCOOKIES: 5572 error = pf_get_syncookies((struct pfioc_nv *)addr); 5573 break; 5574 5575 case DIOCSETSYNCOOKIES: 5576 error = pf_set_syncookies((struct pfioc_nv *)addr); 5577 break; 5578 5579 case DIOCSETHOSTID: { 5580 u_int32_t *hostid = (u_int32_t *)addr; 5581 5582 PF_RULES_WLOCK(); 5583 if (*hostid == 0) 5584 V_pf_status.hostid = arc4random(); 5585 else 5586 V_pf_status.hostid = *hostid; 5587 PF_RULES_WUNLOCK(); 5588 break; 5589 } 5590 5591 case DIOCOSFPFLUSH: 5592 PF_RULES_WLOCK(); 5593 pf_osfp_flush(); 5594 PF_RULES_WUNLOCK(); 5595 break; 5596 5597 case DIOCIGETIFACES: { 5598 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5599 struct pfi_kif *ifstore; 5600 size_t bufsiz; 5601 5602 if (io->pfiio_esize != sizeof(struct pfi_kif)) { 5603 error = ENODEV; 5604 break; 5605 } 5606 5607 if (io->pfiio_size < 0 || 5608 io->pfiio_size > pf_ioctl_maxcount || 5609 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { 5610 error = EINVAL; 5611 break; 5612 } 5613 5614 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5615 5616 bufsiz = io->pfiio_size * sizeof(struct pfi_kif); 5617 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif), 5618 M_TEMP, M_WAITOK | M_ZERO); 5619 5620 PF_RULES_RLOCK(); 5621 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); 5622 PF_RULES_RUNLOCK(); 5623 error = copyout(ifstore, io->pfiio_buffer, bufsiz); 5624 free(ifstore, M_TEMP); 5625 break; 5626 } 5627 5628 case DIOCSETIFFLAG: { 5629 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5630 5631 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5632 5633 PF_RULES_WLOCK(); 5634 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags); 5635 PF_RULES_WUNLOCK(); 5636 break; 5637 } 5638 5639 case DIOCCLRIFFLAG: { 5640 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5641 5642 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5643 5644 PF_RULES_WLOCK(); 5645 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); 5646 PF_RULES_WUNLOCK(); 5647 break; 5648 } 5649 5650 case DIOCSETREASS: { 5651 u_int32_t *reass = (u_int32_t *)addr; 5652 5653 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF); 5654 /* Removal of DF flag without reassembly enabled is not a 5655 * valid combination. Disable reassembly in such case. */ 5656 if (!(V_pf_status.reass & PF_REASS_ENABLED)) 5657 V_pf_status.reass = 0; 5658 break; 5659 } 5660 5661 default: 5662 error = ENODEV; 5663 break; 5664 } 5665 fail: 5666 if (sx_xlocked(&V_pf_ioctl_lock)) 5667 sx_xunlock(&V_pf_ioctl_lock); 5668 CURVNET_RESTORE(); 5669 5670 #undef ERROUT_IOCTL 5671 5672 return (error); 5673 } 5674 5675 void 5676 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version) 5677 { 5678 bzero(sp, sizeof(union pfsync_state_union)); 5679 5680 /* copy from state key */ 5681 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5682 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5683 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5684 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5685 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5686 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5687 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5688 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5689 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto; 5690 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af; 5691 5692 /* copy from state */ 5693 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname)); 5694 bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr)); 5695 sp->pfs_1301.creation = htonl(time_uptime - st->creation); 5696 sp->pfs_1301.expire = pf_state_expires(st); 5697 if (sp->pfs_1301.expire <= time_uptime) 5698 sp->pfs_1301.expire = htonl(0); 5699 else 5700 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime); 5701 5702 sp->pfs_1301.direction = st->direction; 5703 sp->pfs_1301.log = st->log; 5704 sp->pfs_1301.timeout = st->timeout; 5705 5706 switch (msg_version) { 5707 case PFSYNC_MSG_VERSION_1301: 5708 sp->pfs_1301.state_flags = st->state_flags; 5709 break; 5710 case PFSYNC_MSG_VERSION_1400: 5711 sp->pfs_1400.state_flags = htons(st->state_flags); 5712 sp->pfs_1400.qid = htons(st->qid); 5713 sp->pfs_1400.pqid = htons(st->pqid); 5714 sp->pfs_1400.dnpipe = htons(st->dnpipe); 5715 sp->pfs_1400.dnrpipe = htons(st->dnrpipe); 5716 sp->pfs_1400.rtableid = htonl(st->rtableid); 5717 sp->pfs_1400.min_ttl = st->min_ttl; 5718 sp->pfs_1400.set_tos = st->set_tos; 5719 sp->pfs_1400.max_mss = htons(st->max_mss); 5720 sp->pfs_1400.set_prio[0] = st->set_prio[0]; 5721 sp->pfs_1400.set_prio[1] = st->set_prio[1]; 5722 sp->pfs_1400.rt = st->rt; 5723 if (st->rt_kif) 5724 strlcpy(sp->pfs_1400.rt_ifname, 5725 st->rt_kif->pfik_name, 5726 sizeof(sp->pfs_1400.rt_ifname)); 5727 break; 5728 default: 5729 panic("%s: Unsupported pfsync_msg_version %d", 5730 __func__, msg_version); 5731 } 5732 5733 if (st->src_node) 5734 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE; 5735 if (st->nat_src_node) 5736 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5737 5738 sp->pfs_1301.id = st->id; 5739 sp->pfs_1301.creatorid = st->creatorid; 5740 pf_state_peer_hton(&st->src, &sp->pfs_1301.src); 5741 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst); 5742 5743 if (st->rule.ptr == NULL) 5744 sp->pfs_1301.rule = htonl(-1); 5745 else 5746 sp->pfs_1301.rule = htonl(st->rule.ptr->nr); 5747 if (st->anchor.ptr == NULL) 5748 sp->pfs_1301.anchor = htonl(-1); 5749 else 5750 sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr); 5751 if (st->nat_rule.ptr == NULL) 5752 sp->pfs_1301.nat_rule = htonl(-1); 5753 else 5754 sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr); 5755 5756 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]); 5757 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]); 5758 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]); 5759 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]); 5760 } 5761 5762 void 5763 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st) 5764 { 5765 bzero(sp, sizeof(*sp)); 5766 5767 sp->version = PF_STATE_VERSION; 5768 5769 /* copy from state key */ 5770 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5771 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5772 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5773 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5774 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5775 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5776 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5777 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5778 sp->proto = st->key[PF_SK_WIRE]->proto; 5779 sp->af = st->key[PF_SK_WIRE]->af; 5780 5781 /* copy from state */ 5782 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5783 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name, 5784 sizeof(sp->orig_ifname)); 5785 bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr)); 5786 sp->creation = htonl(time_uptime - st->creation); 5787 sp->expire = pf_state_expires(st); 5788 if (sp->expire <= time_uptime) 5789 sp->expire = htonl(0); 5790 else 5791 sp->expire = htonl(sp->expire - time_uptime); 5792 5793 sp->direction = st->direction; 5794 sp->log = st->log; 5795 sp->timeout = st->timeout; 5796 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */ 5797 sp->state_flags_compat = st->state_flags; 5798 sp->state_flags = htons(st->state_flags); 5799 if (st->src_node) 5800 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5801 if (st->nat_src_node) 5802 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5803 5804 sp->id = st->id; 5805 sp->creatorid = st->creatorid; 5806 pf_state_peer_hton(&st->src, &sp->src); 5807 pf_state_peer_hton(&st->dst, &sp->dst); 5808 5809 if (st->rule.ptr == NULL) 5810 sp->rule = htonl(-1); 5811 else 5812 sp->rule = htonl(st->rule.ptr->nr); 5813 if (st->anchor.ptr == NULL) 5814 sp->anchor = htonl(-1); 5815 else 5816 sp->anchor = htonl(st->anchor.ptr->nr); 5817 if (st->nat_rule.ptr == NULL) 5818 sp->nat_rule = htonl(-1); 5819 else 5820 sp->nat_rule = htonl(st->nat_rule.ptr->nr); 5821 5822 sp->packets[0] = st->packets[0]; 5823 sp->packets[1] = st->packets[1]; 5824 sp->bytes[0] = st->bytes[0]; 5825 sp->bytes[1] = st->bytes[1]; 5826 5827 sp->qid = htons(st->qid); 5828 sp->pqid = htons(st->pqid); 5829 sp->dnpipe = htons(st->dnpipe); 5830 sp->dnrpipe = htons(st->dnrpipe); 5831 sp->rtableid = htonl(st->rtableid); 5832 sp->min_ttl = st->min_ttl; 5833 sp->set_tos = st->set_tos; 5834 sp->max_mss = htons(st->max_mss); 5835 sp->rt = st->rt; 5836 if (st->rt_kif) 5837 strlcpy(sp->rt_ifname, st->rt_kif->pfik_name, 5838 sizeof(sp->rt_ifname)); 5839 sp->set_prio[0] = st->set_prio[0]; 5840 sp->set_prio[1] = st->set_prio[1]; 5841 5842 } 5843 5844 static void 5845 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 5846 { 5847 struct pfr_ktable *kt; 5848 5849 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); 5850 5851 kt = aw->p.tbl; 5852 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 5853 kt = kt->pfrkt_root; 5854 aw->p.tbl = NULL; 5855 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 5856 kt->pfrkt_cnt : -1; 5857 } 5858 5859 static int 5860 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters, 5861 size_t number, char **names) 5862 { 5863 nvlist_t *nvc; 5864 5865 nvc = nvlist_create(0); 5866 if (nvc == NULL) 5867 return (ENOMEM); 5868 5869 for (int i = 0; i < number; i++) { 5870 nvlist_append_number_array(nvc, "counters", 5871 counter_u64_fetch(counters[i])); 5872 nvlist_append_string_array(nvc, "names", 5873 names[i]); 5874 nvlist_append_number_array(nvc, "ids", 5875 i); 5876 } 5877 nvlist_add_nvlist(nvl, name, nvc); 5878 nvlist_destroy(nvc); 5879 5880 return (0); 5881 } 5882 5883 static int 5884 pf_getstatus(struct pfioc_nv *nv) 5885 { 5886 nvlist_t *nvl = NULL, *nvc = NULL; 5887 void *nvlpacked = NULL; 5888 int error; 5889 struct pf_status s; 5890 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES; 5891 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES; 5892 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES; 5893 PF_RULES_RLOCK_TRACKER; 5894 5895 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 5896 5897 PF_RULES_RLOCK(); 5898 5899 nvl = nvlist_create(0); 5900 if (nvl == NULL) 5901 ERROUT(ENOMEM); 5902 5903 nvlist_add_bool(nvl, "running", V_pf_status.running); 5904 nvlist_add_number(nvl, "since", V_pf_status.since); 5905 nvlist_add_number(nvl, "debug", V_pf_status.debug); 5906 nvlist_add_number(nvl, "hostid", V_pf_status.hostid); 5907 nvlist_add_number(nvl, "states", V_pf_status.states); 5908 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes); 5909 nvlist_add_number(nvl, "reass", V_pf_status.reass); 5910 nvlist_add_bool(nvl, "syncookies_active", 5911 V_pf_status.syncookies_active); 5912 5913 /* counters */ 5914 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters, 5915 PFRES_MAX, pf_reasons); 5916 if (error != 0) 5917 ERROUT(error); 5918 5919 /* lcounters */ 5920 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters, 5921 KLCNT_MAX, pf_lcounter); 5922 if (error != 0) 5923 ERROUT(error); 5924 5925 /* fcounters */ 5926 nvc = nvlist_create(0); 5927 if (nvc == NULL) 5928 ERROUT(ENOMEM); 5929 5930 for (int i = 0; i < FCNT_MAX; i++) { 5931 nvlist_append_number_array(nvc, "counters", 5932 pf_counter_u64_fetch(&V_pf_status.fcounters[i])); 5933 nvlist_append_string_array(nvc, "names", 5934 pf_fcounter[i]); 5935 nvlist_append_number_array(nvc, "ids", 5936 i); 5937 } 5938 nvlist_add_nvlist(nvl, "fcounters", nvc); 5939 nvlist_destroy(nvc); 5940 nvc = NULL; 5941 5942 /* scounters */ 5943 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters, 5944 SCNT_MAX, pf_fcounter); 5945 if (error != 0) 5946 ERROUT(error); 5947 5948 nvlist_add_string(nvl, "ifname", V_pf_status.ifname); 5949 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum, 5950 PF_MD5_DIGEST_LENGTH); 5951 5952 pfi_update_status(V_pf_status.ifname, &s); 5953 5954 /* pcounters / bcounters */ 5955 for (int i = 0; i < 2; i++) { 5956 for (int j = 0; j < 2; j++) { 5957 for (int k = 0; k < 2; k++) { 5958 nvlist_append_number_array(nvl, "pcounters", 5959 s.pcounters[i][j][k]); 5960 } 5961 nvlist_append_number_array(nvl, "bcounters", 5962 s.bcounters[i][j]); 5963 } 5964 } 5965 5966 nvlpacked = nvlist_pack(nvl, &nv->len); 5967 if (nvlpacked == NULL) 5968 ERROUT(ENOMEM); 5969 5970 if (nv->size == 0) 5971 ERROUT(0); 5972 else if (nv->size < nv->len) 5973 ERROUT(ENOSPC); 5974 5975 PF_RULES_RUNLOCK(); 5976 error = copyout(nvlpacked, nv->data, nv->len); 5977 goto done; 5978 5979 #undef ERROUT 5980 errout: 5981 PF_RULES_RUNLOCK(); 5982 done: 5983 free(nvlpacked, M_NVLIST); 5984 nvlist_destroy(nvc); 5985 nvlist_destroy(nvl); 5986 5987 return (error); 5988 } 5989 5990 /* 5991 * XXX - Check for version mismatch!!! 5992 */ 5993 static void 5994 pf_clear_all_states(void) 5995 { 5996 struct pf_kstate *s; 5997 u_int i; 5998 5999 for (i = 0; i <= pf_hashmask; i++) { 6000 struct pf_idhash *ih = &V_pf_idhash[i]; 6001 relock: 6002 PF_HASHROW_LOCK(ih); 6003 LIST_FOREACH(s, &ih->states, entry) { 6004 s->timeout = PFTM_PURGE; 6005 /* Don't send out individual delete messages. */ 6006 s->state_flags |= PFSTATE_NOSYNC; 6007 pf_unlink_state(s); 6008 goto relock; 6009 } 6010 PF_HASHROW_UNLOCK(ih); 6011 } 6012 } 6013 6014 static int 6015 pf_clear_tables(void) 6016 { 6017 struct pfioc_table io; 6018 int error; 6019 6020 bzero(&io, sizeof(io)); 6021 6022 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, 6023 io.pfrio_flags); 6024 6025 return (error); 6026 } 6027 6028 static void 6029 pf_clear_srcnodes(struct pf_ksrc_node *n) 6030 { 6031 struct pf_kstate *s; 6032 int i; 6033 6034 for (i = 0; i <= pf_hashmask; i++) { 6035 struct pf_idhash *ih = &V_pf_idhash[i]; 6036 6037 PF_HASHROW_LOCK(ih); 6038 LIST_FOREACH(s, &ih->states, entry) { 6039 if (n == NULL || n == s->src_node) 6040 s->src_node = NULL; 6041 if (n == NULL || n == s->nat_src_node) 6042 s->nat_src_node = NULL; 6043 } 6044 PF_HASHROW_UNLOCK(ih); 6045 } 6046 6047 if (n == NULL) { 6048 struct pf_srchash *sh; 6049 6050 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; 6051 i++, sh++) { 6052 PF_HASHROW_LOCK(sh); 6053 LIST_FOREACH(n, &sh->nodes, entry) { 6054 n->expire = 1; 6055 n->states = 0; 6056 } 6057 PF_HASHROW_UNLOCK(sh); 6058 } 6059 } else { 6060 /* XXX: hash slot should already be locked here. */ 6061 n->expire = 1; 6062 n->states = 0; 6063 } 6064 } 6065 6066 static void 6067 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) 6068 { 6069 struct pf_ksrc_node_list kill; 6070 6071 LIST_INIT(&kill); 6072 for (int i = 0; i <= pf_srchashmask; i++) { 6073 struct pf_srchash *sh = &V_pf_srchash[i]; 6074 struct pf_ksrc_node *sn, *tmp; 6075 6076 PF_HASHROW_LOCK(sh); 6077 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) 6078 if (PF_MATCHA(psnk->psnk_src.neg, 6079 &psnk->psnk_src.addr.v.a.addr, 6080 &psnk->psnk_src.addr.v.a.mask, 6081 &sn->addr, sn->af) && 6082 PF_MATCHA(psnk->psnk_dst.neg, 6083 &psnk->psnk_dst.addr.v.a.addr, 6084 &psnk->psnk_dst.addr.v.a.mask, 6085 &sn->raddr, sn->af)) { 6086 pf_unlink_src_node(sn); 6087 LIST_INSERT_HEAD(&kill, sn, entry); 6088 sn->expire = 1; 6089 } 6090 PF_HASHROW_UNLOCK(sh); 6091 } 6092 6093 for (int i = 0; i <= pf_hashmask; i++) { 6094 struct pf_idhash *ih = &V_pf_idhash[i]; 6095 struct pf_kstate *s; 6096 6097 PF_HASHROW_LOCK(ih); 6098 LIST_FOREACH(s, &ih->states, entry) { 6099 if (s->src_node && s->src_node->expire == 1) 6100 s->src_node = NULL; 6101 if (s->nat_src_node && s->nat_src_node->expire == 1) 6102 s->nat_src_node = NULL; 6103 } 6104 PF_HASHROW_UNLOCK(ih); 6105 } 6106 6107 psnk->psnk_killed = pf_free_src_nodes(&kill); 6108 } 6109 6110 static int 6111 pf_keepcounters(struct pfioc_nv *nv) 6112 { 6113 nvlist_t *nvl = NULL; 6114 void *nvlpacked = NULL; 6115 int error = 0; 6116 6117 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6118 6119 if (nv->len > pf_ioctl_maxcount) 6120 ERROUT(ENOMEM); 6121 6122 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6123 if (nvlpacked == NULL) 6124 ERROUT(ENOMEM); 6125 6126 error = copyin(nv->data, nvlpacked, nv->len); 6127 if (error) 6128 ERROUT(error); 6129 6130 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6131 if (nvl == NULL) 6132 ERROUT(EBADMSG); 6133 6134 if (! nvlist_exists_bool(nvl, "keep_counters")) 6135 ERROUT(EBADMSG); 6136 6137 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters"); 6138 6139 on_error: 6140 nvlist_destroy(nvl); 6141 free(nvlpacked, M_NVLIST); 6142 return (error); 6143 } 6144 6145 static unsigned int 6146 pf_clear_states(const struct pf_kstate_kill *kill) 6147 { 6148 struct pf_state_key_cmp match_key; 6149 struct pf_kstate *s; 6150 struct pfi_kkif *kif; 6151 int idx; 6152 unsigned int killed = 0, dir; 6153 6154 for (unsigned int i = 0; i <= pf_hashmask; i++) { 6155 struct pf_idhash *ih = &V_pf_idhash[i]; 6156 6157 relock_DIOCCLRSTATES: 6158 PF_HASHROW_LOCK(ih); 6159 LIST_FOREACH(s, &ih->states, entry) { 6160 /* For floating states look at the original kif. */ 6161 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 6162 6163 if (kill->psk_ifname[0] && 6164 strcmp(kill->psk_ifname, 6165 kif->pfik_name)) 6166 continue; 6167 6168 if (kill->psk_kill_match) { 6169 bzero(&match_key, sizeof(match_key)); 6170 6171 if (s->direction == PF_OUT) { 6172 dir = PF_IN; 6173 idx = PF_SK_STACK; 6174 } else { 6175 dir = PF_OUT; 6176 idx = PF_SK_WIRE; 6177 } 6178 6179 match_key.af = s->key[idx]->af; 6180 match_key.proto = s->key[idx]->proto; 6181 PF_ACPY(&match_key.addr[0], 6182 &s->key[idx]->addr[1], match_key.af); 6183 match_key.port[0] = s->key[idx]->port[1]; 6184 PF_ACPY(&match_key.addr[1], 6185 &s->key[idx]->addr[0], match_key.af); 6186 match_key.port[1] = s->key[idx]->port[0]; 6187 } 6188 6189 /* 6190 * Don't send out individual 6191 * delete messages. 6192 */ 6193 s->state_flags |= PFSTATE_NOSYNC; 6194 pf_unlink_state(s); 6195 killed++; 6196 6197 if (kill->psk_kill_match) 6198 killed += pf_kill_matching_state(&match_key, 6199 dir); 6200 6201 goto relock_DIOCCLRSTATES; 6202 } 6203 PF_HASHROW_UNLOCK(ih); 6204 } 6205 6206 if (V_pfsync_clear_states_ptr != NULL) 6207 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname); 6208 6209 return (killed); 6210 } 6211 6212 static void 6213 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed) 6214 { 6215 struct pf_kstate *s; 6216 6217 if (kill->psk_pfcmp.id) { 6218 if (kill->psk_pfcmp.creatorid == 0) 6219 kill->psk_pfcmp.creatorid = V_pf_status.hostid; 6220 if ((s = pf_find_state_byid(kill->psk_pfcmp.id, 6221 kill->psk_pfcmp.creatorid))) { 6222 pf_unlink_state(s); 6223 *killed = 1; 6224 } 6225 return; 6226 } 6227 6228 for (unsigned int i = 0; i <= pf_hashmask; i++) 6229 *killed += pf_killstates_row(kill, &V_pf_idhash[i]); 6230 6231 return; 6232 } 6233 6234 static int 6235 pf_killstates_nv(struct pfioc_nv *nv) 6236 { 6237 struct pf_kstate_kill kill; 6238 nvlist_t *nvl = NULL; 6239 void *nvlpacked = NULL; 6240 int error = 0; 6241 unsigned int killed = 0; 6242 6243 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6244 6245 if (nv->len > pf_ioctl_maxcount) 6246 ERROUT(ENOMEM); 6247 6248 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6249 if (nvlpacked == NULL) 6250 ERROUT(ENOMEM); 6251 6252 error = copyin(nv->data, nvlpacked, nv->len); 6253 if (error) 6254 ERROUT(error); 6255 6256 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6257 if (nvl == NULL) 6258 ERROUT(EBADMSG); 6259 6260 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6261 if (error) 6262 ERROUT(error); 6263 6264 pf_killstates(&kill, &killed); 6265 6266 free(nvlpacked, M_NVLIST); 6267 nvlpacked = NULL; 6268 nvlist_destroy(nvl); 6269 nvl = nvlist_create(0); 6270 if (nvl == NULL) 6271 ERROUT(ENOMEM); 6272 6273 nvlist_add_number(nvl, "killed", killed); 6274 6275 nvlpacked = nvlist_pack(nvl, &nv->len); 6276 if (nvlpacked == NULL) 6277 ERROUT(ENOMEM); 6278 6279 if (nv->size == 0) 6280 ERROUT(0); 6281 else if (nv->size < nv->len) 6282 ERROUT(ENOSPC); 6283 6284 error = copyout(nvlpacked, nv->data, nv->len); 6285 6286 on_error: 6287 nvlist_destroy(nvl); 6288 free(nvlpacked, M_NVLIST); 6289 return (error); 6290 } 6291 6292 static int 6293 pf_clearstates_nv(struct pfioc_nv *nv) 6294 { 6295 struct pf_kstate_kill kill; 6296 nvlist_t *nvl = NULL; 6297 void *nvlpacked = NULL; 6298 int error = 0; 6299 unsigned int killed; 6300 6301 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6302 6303 if (nv->len > pf_ioctl_maxcount) 6304 ERROUT(ENOMEM); 6305 6306 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6307 if (nvlpacked == NULL) 6308 ERROUT(ENOMEM); 6309 6310 error = copyin(nv->data, nvlpacked, nv->len); 6311 if (error) 6312 ERROUT(error); 6313 6314 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6315 if (nvl == NULL) 6316 ERROUT(EBADMSG); 6317 6318 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6319 if (error) 6320 ERROUT(error); 6321 6322 killed = pf_clear_states(&kill); 6323 6324 free(nvlpacked, M_NVLIST); 6325 nvlpacked = NULL; 6326 nvlist_destroy(nvl); 6327 nvl = nvlist_create(0); 6328 if (nvl == NULL) 6329 ERROUT(ENOMEM); 6330 6331 nvlist_add_number(nvl, "killed", killed); 6332 6333 nvlpacked = nvlist_pack(nvl, &nv->len); 6334 if (nvlpacked == NULL) 6335 ERROUT(ENOMEM); 6336 6337 if (nv->size == 0) 6338 ERROUT(0); 6339 else if (nv->size < nv->len) 6340 ERROUT(ENOSPC); 6341 6342 error = copyout(nvlpacked, nv->data, nv->len); 6343 6344 #undef ERROUT 6345 on_error: 6346 nvlist_destroy(nvl); 6347 free(nvlpacked, M_NVLIST); 6348 return (error); 6349 } 6350 6351 static int 6352 pf_getstate(struct pfioc_nv *nv) 6353 { 6354 nvlist_t *nvl = NULL, *nvls; 6355 void *nvlpacked = NULL; 6356 struct pf_kstate *s = NULL; 6357 int error = 0; 6358 uint64_t id, creatorid; 6359 6360 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6361 6362 if (nv->len > pf_ioctl_maxcount) 6363 ERROUT(ENOMEM); 6364 6365 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6366 if (nvlpacked == NULL) 6367 ERROUT(ENOMEM); 6368 6369 error = copyin(nv->data, nvlpacked, nv->len); 6370 if (error) 6371 ERROUT(error); 6372 6373 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6374 if (nvl == NULL) 6375 ERROUT(EBADMSG); 6376 6377 PFNV_CHK(pf_nvuint64(nvl, "id", &id)); 6378 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid)); 6379 6380 s = pf_find_state_byid(id, creatorid); 6381 if (s == NULL) 6382 ERROUT(ENOENT); 6383 6384 free(nvlpacked, M_NVLIST); 6385 nvlpacked = NULL; 6386 nvlist_destroy(nvl); 6387 nvl = nvlist_create(0); 6388 if (nvl == NULL) 6389 ERROUT(ENOMEM); 6390 6391 nvls = pf_state_to_nvstate(s); 6392 if (nvls == NULL) 6393 ERROUT(ENOMEM); 6394 6395 nvlist_add_nvlist(nvl, "state", nvls); 6396 nvlist_destroy(nvls); 6397 6398 nvlpacked = nvlist_pack(nvl, &nv->len); 6399 if (nvlpacked == NULL) 6400 ERROUT(ENOMEM); 6401 6402 if (nv->size == 0) 6403 ERROUT(0); 6404 else if (nv->size < nv->len) 6405 ERROUT(ENOSPC); 6406 6407 error = copyout(nvlpacked, nv->data, nv->len); 6408 6409 #undef ERROUT 6410 errout: 6411 if (s != NULL) 6412 PF_STATE_UNLOCK(s); 6413 free(nvlpacked, M_NVLIST); 6414 nvlist_destroy(nvl); 6415 return (error); 6416 } 6417 6418 /* 6419 * XXX - Check for version mismatch!!! 6420 */ 6421 6422 /* 6423 * Duplicate pfctl -Fa operation to get rid of as much as we can. 6424 */ 6425 static int 6426 shutdown_pf(void) 6427 { 6428 int error = 0; 6429 u_int32_t t[5]; 6430 char nn = '\0'; 6431 6432 do { 6433 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) 6434 != 0) { 6435 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n")); 6436 break; 6437 } 6438 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) 6439 != 0) { 6440 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n")); 6441 break; /* XXX: rollback? */ 6442 } 6443 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) 6444 != 0) { 6445 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n")); 6446 break; /* XXX: rollback? */ 6447 } 6448 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) 6449 != 0) { 6450 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n")); 6451 break; /* XXX: rollback? */ 6452 } 6453 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) 6454 != 0) { 6455 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n")); 6456 break; /* XXX: rollback? */ 6457 } 6458 6459 /* XXX: these should always succeed here */ 6460 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn); 6461 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn); 6462 pf_commit_rules(t[2], PF_RULESET_NAT, &nn); 6463 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn); 6464 pf_commit_rules(t[4], PF_RULESET_RDR, &nn); 6465 6466 if ((error = pf_clear_tables()) != 0) 6467 break; 6468 6469 if ((error = pf_begin_eth(&t[0], &nn)) != 0) { 6470 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n")); 6471 break; 6472 } 6473 pf_commit_eth(t[0], &nn); 6474 6475 #ifdef ALTQ 6476 if ((error = pf_begin_altq(&t[0])) != 0) { 6477 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n")); 6478 break; 6479 } 6480 pf_commit_altq(t[0]); 6481 #endif 6482 6483 pf_clear_all_states(); 6484 6485 pf_clear_srcnodes(NULL); 6486 6487 /* status does not use malloced mem so no need to cleanup */ 6488 /* fingerprints and interfaces have their own cleanup code */ 6489 } while(0); 6490 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 chk = pf_test_eth(PF_IN, flags, ifp, m, inp); 6521 6522 return (pf_check_return(chk, m)); 6523 } 6524 6525 static pfil_return_t 6526 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6527 void *ruleset __unused, struct inpcb *inp) 6528 { 6529 int chk; 6530 6531 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp); 6532 6533 return (pf_check_return(chk, m)); 6534 } 6535 6536 #ifdef INET 6537 static pfil_return_t 6538 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6539 void *ruleset __unused, struct inpcb *inp) 6540 { 6541 int chk; 6542 6543 chk = pf_test(PF_IN, flags, ifp, m, inp, NULL); 6544 6545 return (pf_check_return(chk, m)); 6546 } 6547 6548 static pfil_return_t 6549 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6550 void *ruleset __unused, struct inpcb *inp) 6551 { 6552 int chk; 6553 6554 chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL); 6555 6556 return (pf_check_return(chk, m)); 6557 } 6558 #endif 6559 6560 #ifdef INET6 6561 static pfil_return_t 6562 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags, 6563 void *ruleset __unused, struct inpcb *inp) 6564 { 6565 int chk; 6566 6567 /* 6568 * In case of loopback traffic IPv6 uses the real interface in 6569 * order to support scoped addresses. In order to support stateful 6570 * filtering we have change this to lo0 as it is the case in IPv4. 6571 */ 6572 CURVNET_SET(ifp->if_vnet); 6573 chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, 6574 m, inp, NULL); 6575 CURVNET_RESTORE(); 6576 6577 return (pf_check_return(chk, m)); 6578 } 6579 6580 static pfil_return_t 6581 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags, 6582 void *ruleset __unused, struct inpcb *inp) 6583 { 6584 int chk; 6585 6586 CURVNET_SET(ifp->if_vnet); 6587 chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL); 6588 CURVNET_RESTORE(); 6589 6590 return (pf_check_return(chk, m)); 6591 } 6592 #endif /* INET6 */ 6593 6594 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook); 6595 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook); 6596 #define V_pf_eth_in_hook VNET(pf_eth_in_hook) 6597 #define V_pf_eth_out_hook VNET(pf_eth_out_hook) 6598 6599 #ifdef INET 6600 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook); 6601 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook); 6602 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook) 6603 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook) 6604 #endif 6605 #ifdef INET6 6606 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook); 6607 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook); 6608 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook) 6609 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook) 6610 #endif 6611 6612 static void 6613 hook_pf_eth(void) 6614 { 6615 struct pfil_hook_args pha = { 6616 .pa_version = PFIL_VERSION, 6617 .pa_modname = "pf", 6618 .pa_type = PFIL_TYPE_ETHERNET, 6619 }; 6620 struct pfil_link_args pla = { 6621 .pa_version = PFIL_VERSION, 6622 }; 6623 int ret __diagused; 6624 6625 if (atomic_load_bool(&V_pf_pfil_eth_hooked)) 6626 return; 6627 6628 pha.pa_mbuf_chk = pf_eth_check_in; 6629 pha.pa_flags = PFIL_IN; 6630 pha.pa_rulname = "eth-in"; 6631 V_pf_eth_in_hook = pfil_add_hook(&pha); 6632 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6633 pla.pa_head = V_link_pfil_head; 6634 pla.pa_hook = V_pf_eth_in_hook; 6635 ret = pfil_link(&pla); 6636 MPASS(ret == 0); 6637 pha.pa_mbuf_chk = pf_eth_check_out; 6638 pha.pa_flags = PFIL_OUT; 6639 pha.pa_rulname = "eth-out"; 6640 V_pf_eth_out_hook = pfil_add_hook(&pha); 6641 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6642 pla.pa_head = V_link_pfil_head; 6643 pla.pa_hook = V_pf_eth_out_hook; 6644 ret = pfil_link(&pla); 6645 MPASS(ret == 0); 6646 6647 atomic_store_bool(&V_pf_pfil_eth_hooked, true); 6648 } 6649 6650 static void 6651 hook_pf(void) 6652 { 6653 struct pfil_hook_args pha = { 6654 .pa_version = PFIL_VERSION, 6655 .pa_modname = "pf", 6656 }; 6657 struct pfil_link_args pla = { 6658 .pa_version = PFIL_VERSION, 6659 }; 6660 int ret __diagused; 6661 6662 if (atomic_load_bool(&V_pf_pfil_hooked)) 6663 return; 6664 6665 #ifdef INET 6666 pha.pa_type = PFIL_TYPE_IP4; 6667 pha.pa_mbuf_chk = pf_check_in; 6668 pha.pa_flags = PFIL_IN; 6669 pha.pa_rulname = "default-in"; 6670 V_pf_ip4_in_hook = pfil_add_hook(&pha); 6671 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6672 pla.pa_head = V_inet_pfil_head; 6673 pla.pa_hook = V_pf_ip4_in_hook; 6674 ret = pfil_link(&pla); 6675 MPASS(ret == 0); 6676 pha.pa_mbuf_chk = pf_check_out; 6677 pha.pa_flags = PFIL_OUT; 6678 pha.pa_rulname = "default-out"; 6679 V_pf_ip4_out_hook = pfil_add_hook(&pha); 6680 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6681 pla.pa_head = V_inet_pfil_head; 6682 pla.pa_hook = V_pf_ip4_out_hook; 6683 ret = pfil_link(&pla); 6684 MPASS(ret == 0); 6685 #endif 6686 #ifdef INET6 6687 pha.pa_type = PFIL_TYPE_IP6; 6688 pha.pa_mbuf_chk = pf_check6_in; 6689 pha.pa_flags = PFIL_IN; 6690 pha.pa_rulname = "default-in6"; 6691 V_pf_ip6_in_hook = pfil_add_hook(&pha); 6692 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6693 pla.pa_head = V_inet6_pfil_head; 6694 pla.pa_hook = V_pf_ip6_in_hook; 6695 ret = pfil_link(&pla); 6696 MPASS(ret == 0); 6697 pha.pa_mbuf_chk = pf_check6_out; 6698 pha.pa_rulname = "default-out6"; 6699 pha.pa_flags = PFIL_OUT; 6700 V_pf_ip6_out_hook = pfil_add_hook(&pha); 6701 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6702 pla.pa_head = V_inet6_pfil_head; 6703 pla.pa_hook = V_pf_ip6_out_hook; 6704 ret = pfil_link(&pla); 6705 MPASS(ret == 0); 6706 #endif 6707 6708 atomic_store_bool(&V_pf_pfil_hooked, true); 6709 } 6710 6711 static void 6712 dehook_pf_eth(void) 6713 { 6714 6715 if (!atomic_load_bool(&V_pf_pfil_eth_hooked)) 6716 return; 6717 6718 pfil_remove_hook(V_pf_eth_in_hook); 6719 pfil_remove_hook(V_pf_eth_out_hook); 6720 6721 atomic_store_bool(&V_pf_pfil_eth_hooked, false); 6722 } 6723 6724 static void 6725 dehook_pf(void) 6726 { 6727 6728 if (!atomic_load_bool(&V_pf_pfil_hooked)) 6729 return; 6730 6731 #ifdef INET 6732 pfil_remove_hook(V_pf_ip4_in_hook); 6733 pfil_remove_hook(V_pf_ip4_out_hook); 6734 #endif 6735 #ifdef INET6 6736 pfil_remove_hook(V_pf_ip6_in_hook); 6737 pfil_remove_hook(V_pf_ip6_out_hook); 6738 #endif 6739 6740 atomic_store_bool(&V_pf_pfil_hooked, false); 6741 } 6742 6743 static void 6744 pf_load_vnet(void) 6745 { 6746 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname), 6747 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 6748 6749 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE); 6750 sx_init(&V_pf_ioctl_lock, "pf ioctl"); 6751 6752 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize, 6753 PF_RULE_TAG_HASH_SIZE_DEFAULT); 6754 #ifdef ALTQ 6755 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize, 6756 PF_QUEUE_TAG_HASH_SIZE_DEFAULT); 6757 #endif 6758 6759 V_pf_keth = &V_pf_main_keth_anchor.ruleset; 6760 6761 pfattach_vnet(); 6762 V_pf_vnet_active = 1; 6763 } 6764 6765 static int 6766 pf_load(void) 6767 { 6768 int error; 6769 6770 sx_init(&pf_end_lock, "pf end thread"); 6771 6772 pf_mtag_initialize(); 6773 6774 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME); 6775 if (pf_dev == NULL) 6776 return (ENOMEM); 6777 6778 pf_end_threads = 0; 6779 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge"); 6780 if (error != 0) 6781 return (error); 6782 6783 pfi_initialize(); 6784 6785 return (0); 6786 } 6787 6788 static void 6789 pf_unload_vnet(void) 6790 { 6791 int ret __diagused; 6792 6793 V_pf_vnet_active = 0; 6794 V_pf_status.running = 0; 6795 dehook_pf(); 6796 dehook_pf_eth(); 6797 6798 PF_RULES_WLOCK(); 6799 pf_syncookies_cleanup(); 6800 shutdown_pf(); 6801 PF_RULES_WUNLOCK(); 6802 6803 /* Make sure we've cleaned up ethernet rules before we continue. */ 6804 NET_EPOCH_DRAIN_CALLBACKS(); 6805 6806 ret = swi_remove(V_pf_swi_cookie); 6807 MPASS(ret == 0); 6808 ret = intr_event_destroy(V_pf_swi_ie); 6809 MPASS(ret == 0); 6810 6811 pf_unload_vnet_purge(); 6812 6813 pf_normalize_cleanup(); 6814 PF_RULES_WLOCK(); 6815 pfi_cleanup_vnet(); 6816 PF_RULES_WUNLOCK(); 6817 pfr_cleanup(); 6818 pf_osfp_flush(); 6819 pf_cleanup(); 6820 if (IS_DEFAULT_VNET(curvnet)) 6821 pf_mtag_cleanup(); 6822 6823 pf_cleanup_tagset(&V_pf_tags); 6824 #ifdef ALTQ 6825 pf_cleanup_tagset(&V_pf_qids); 6826 #endif 6827 uma_zdestroy(V_pf_tag_z); 6828 6829 #ifdef PF_WANT_32_TO_64_COUNTER 6830 PF_RULES_WLOCK(); 6831 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); 6832 6833 MPASS(LIST_EMPTY(&V_pf_allkiflist)); 6834 MPASS(V_pf_allkifcount == 0); 6835 6836 LIST_REMOVE(&V_pf_default_rule, allrulelist); 6837 V_pf_allrulecount--; 6838 LIST_REMOVE(V_pf_rulemarker, allrulelist); 6839 6840 /* 6841 * There are known pf rule leaks when running the test suite. 6842 */ 6843 #ifdef notyet 6844 MPASS(LIST_EMPTY(&V_pf_allrulelist)); 6845 MPASS(V_pf_allrulecount == 0); 6846 #endif 6847 6848 PF_RULES_WUNLOCK(); 6849 6850 free(V_pf_kifmarker, PFI_MTYPE); 6851 free(V_pf_rulemarker, M_PFRULE); 6852 #endif 6853 6854 /* Free counters last as we updated them during shutdown. */ 6855 pf_counter_u64_deinit(&V_pf_default_rule.evaluations); 6856 for (int i = 0; i < 2; i++) { 6857 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]); 6858 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]); 6859 } 6860 counter_u64_free(V_pf_default_rule.states_cur); 6861 counter_u64_free(V_pf_default_rule.states_tot); 6862 counter_u64_free(V_pf_default_rule.src_nodes); 6863 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp); 6864 6865 for (int i = 0; i < PFRES_MAX; i++) 6866 counter_u64_free(V_pf_status.counters[i]); 6867 for (int i = 0; i < KLCNT_MAX; i++) 6868 counter_u64_free(V_pf_status.lcounters[i]); 6869 for (int i = 0; i < FCNT_MAX; i++) 6870 pf_counter_u64_deinit(&V_pf_status.fcounters[i]); 6871 for (int i = 0; i < SCNT_MAX; i++) 6872 counter_u64_free(V_pf_status.scounters[i]); 6873 6874 rm_destroy(&V_pf_rules_lock); 6875 sx_destroy(&V_pf_ioctl_lock); 6876 } 6877 6878 static void 6879 pf_unload(void) 6880 { 6881 6882 sx_xlock(&pf_end_lock); 6883 pf_end_threads = 1; 6884 while (pf_end_threads < 2) { 6885 wakeup_one(pf_purge_thread); 6886 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0); 6887 } 6888 sx_xunlock(&pf_end_lock); 6889 6890 if (pf_dev != NULL) 6891 destroy_dev(pf_dev); 6892 6893 pfi_cleanup(); 6894 6895 sx_destroy(&pf_end_lock); 6896 } 6897 6898 static void 6899 vnet_pf_init(void *unused __unused) 6900 { 6901 6902 pf_load_vnet(); 6903 } 6904 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6905 vnet_pf_init, NULL); 6906 6907 static void 6908 vnet_pf_uninit(const void *unused __unused) 6909 { 6910 6911 pf_unload_vnet(); 6912 } 6913 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL); 6914 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6915 vnet_pf_uninit, NULL); 6916 6917 static int 6918 pf_modevent(module_t mod, int type, void *data) 6919 { 6920 int error = 0; 6921 6922 switch(type) { 6923 case MOD_LOAD: 6924 error = pf_load(); 6925 break; 6926 case MOD_UNLOAD: 6927 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after 6928 * the vnet_pf_uninit()s */ 6929 break; 6930 default: 6931 error = EINVAL; 6932 break; 6933 } 6934 6935 return (error); 6936 } 6937 6938 static moduledata_t pf_mod = { 6939 "pf", 6940 pf_modevent, 6941 0 6942 }; 6943 6944 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND); 6945 MODULE_VERSION(pf, PF_MODVER); 6946