1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 2001 Daniel Hartmeier 5 * Copyright (c) 2002,2003 Henning Brauer 6 * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org> 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 13 * - Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * - Redistributions in binary form must reproduce the above 16 * copyright notice, this list of conditions and the following 17 * disclaimer in the documentation and/or other materials provided 18 * with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 * 33 * Effort sponsored in part by the Defense Advanced Research Projects 34 * Agency (DARPA) and Air Force Research Laboratory, Air Force 35 * Materiel Command, USAF, under agreement number F30602-01-2-0537. 36 * 37 * $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $ 38 */ 39 40 #include <sys/cdefs.h> 41 __FBSDID("$FreeBSD$"); 42 43 #include "opt_inet.h" 44 #include "opt_inet6.h" 45 #include "opt_bpf.h" 46 #include "opt_pf.h" 47 48 #include <sys/param.h> 49 #include <sys/_bitset.h> 50 #include <sys/bitset.h> 51 #include <sys/bus.h> 52 #include <sys/conf.h> 53 #include <sys/endian.h> 54 #include <sys/fcntl.h> 55 #include <sys/filio.h> 56 #include <sys/hash.h> 57 #include <sys/interrupt.h> 58 #include <sys/jail.h> 59 #include <sys/kernel.h> 60 #include <sys/kthread.h> 61 #include <sys/lock.h> 62 #include <sys/mbuf.h> 63 #include <sys/module.h> 64 #include <sys/nv.h> 65 #include <sys/proc.h> 66 #include <sys/sdt.h> 67 #include <sys/smp.h> 68 #include <sys/socket.h> 69 #include <sys/sysctl.h> 70 #include <sys/md5.h> 71 #include <sys/ucred.h> 72 73 #include <net/if.h> 74 #include <net/if_var.h> 75 #include <net/vnet.h> 76 #include <net/route.h> 77 #include <net/pfil.h> 78 #include <net/pfvar.h> 79 #include <net/if_pfsync.h> 80 #include <net/if_pflog.h> 81 82 #include <netinet/in.h> 83 #include <netinet/ip.h> 84 #include <netinet/ip_var.h> 85 #include <netinet6/ip6_var.h> 86 #include <netinet/ip_icmp.h> 87 #include <netpfil/pf/pf_nv.h> 88 89 #ifdef INET6 90 #include <netinet/ip6.h> 91 #endif /* INET6 */ 92 93 #ifdef ALTQ 94 #include <net/altq/altq.h> 95 #endif 96 97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int"); 98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int"); 99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int"); 100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int"); 101 102 static struct pf_kpool *pf_get_kpool(const char *, u_int32_t, u_int8_t, 103 u_int32_t, u_int8_t, u_int8_t, u_int8_t); 104 105 static void pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *); 106 static void pf_empty_kpool(struct pf_kpalist *); 107 static int pfioctl(struct cdev *, u_long, caddr_t, int, 108 struct thread *); 109 static int pf_begin_eth(uint32_t *, const char *); 110 static void pf_rollback_eth_cb(struct epoch_context *); 111 static int pf_rollback_eth(uint32_t, const char *); 112 static int pf_commit_eth(uint32_t, const char *); 113 static void pf_free_eth_rule(struct pf_keth_rule *); 114 #ifdef ALTQ 115 static int pf_begin_altq(u_int32_t *); 116 static int pf_rollback_altq(u_int32_t); 117 static int pf_commit_altq(u_int32_t); 118 static int pf_enable_altq(struct pf_altq *); 119 static int pf_disable_altq(struct pf_altq *); 120 static uint16_t pf_qname2qid(const char *); 121 static void pf_qid_unref(uint16_t); 122 #endif /* ALTQ */ 123 static int pf_begin_rules(u_int32_t *, int, const char *); 124 static int pf_rollback_rules(u_int32_t, int, char *); 125 static int pf_setup_pfsync_matching(struct pf_kruleset *); 126 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *); 127 static void pf_hash_rule(struct pf_krule *); 128 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *); 129 static int pf_commit_rules(u_int32_t, int, char *); 130 static int pf_addr_setup(struct pf_kruleset *, 131 struct pf_addr_wrap *, sa_family_t); 132 static void pf_addr_copyout(struct pf_addr_wrap *); 133 static void pf_src_node_copy(const struct pf_ksrc_node *, 134 struct pf_src_node *); 135 #ifdef ALTQ 136 static int pf_export_kaltq(struct pf_altq *, 137 struct pfioc_altq_v1 *, size_t); 138 static int pf_import_kaltq(struct pfioc_altq_v1 *, 139 struct pf_altq *, size_t); 140 #endif /* ALTQ */ 141 142 VNET_DEFINE(struct pf_krule, pf_default_rule); 143 144 static __inline int pf_krule_compare(struct pf_krule *, 145 struct pf_krule *); 146 147 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare); 148 149 #ifdef ALTQ 150 VNET_DEFINE_STATIC(int, pf_altq_running); 151 #define V_pf_altq_running VNET(pf_altq_running) 152 #endif 153 154 #define TAGID_MAX 50000 155 struct pf_tagname { 156 TAILQ_ENTRY(pf_tagname) namehash_entries; 157 TAILQ_ENTRY(pf_tagname) taghash_entries; 158 char name[PF_TAG_NAME_SIZE]; 159 uint16_t tag; 160 int ref; 161 }; 162 163 struct pf_tagset { 164 TAILQ_HEAD(, pf_tagname) *namehash; 165 TAILQ_HEAD(, pf_tagname) *taghash; 166 unsigned int mask; 167 uint32_t seed; 168 BITSET_DEFINE(, TAGID_MAX) avail; 169 }; 170 171 VNET_DEFINE(struct pf_tagset, pf_tags); 172 #define V_pf_tags VNET(pf_tags) 173 static unsigned int pf_rule_tag_hashsize; 174 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128 175 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN, 176 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT, 177 "Size of pf(4) rule tag hashtable"); 178 179 #ifdef ALTQ 180 VNET_DEFINE(struct pf_tagset, pf_qids); 181 #define V_pf_qids VNET(pf_qids) 182 static unsigned int pf_queue_tag_hashsize; 183 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128 184 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN, 185 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT, 186 "Size of pf(4) queue tag hashtable"); 187 #endif 188 VNET_DEFINE(uma_zone_t, pf_tag_z); 189 #define V_pf_tag_z VNET(pf_tag_z) 190 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db"); 191 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules"); 192 193 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE) 194 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE 195 #endif 196 197 static void pf_init_tagset(struct pf_tagset *, unsigned int *, 198 unsigned int); 199 static void pf_cleanup_tagset(struct pf_tagset *); 200 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *); 201 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t); 202 static u_int16_t tagname2tag(struct pf_tagset *, const char *); 203 static u_int16_t pf_tagname2tag(const char *); 204 static void tag_unref(struct pf_tagset *, u_int16_t); 205 206 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x 207 208 struct cdev *pf_dev; 209 210 /* 211 * XXX - These are new and need to be checked when moveing to a new version 212 */ 213 static void pf_clear_all_states(void); 214 static unsigned int pf_clear_states(const struct pf_kstate_kill *); 215 static void pf_killstates(struct pf_kstate_kill *, 216 unsigned int *); 217 static int pf_killstates_row(struct pf_kstate_kill *, 218 struct pf_idhash *); 219 static int pf_killstates_nv(struct pfioc_nv *); 220 static int pf_clearstates_nv(struct pfioc_nv *); 221 static int pf_getstate(struct pfioc_nv *); 222 static int pf_getstatus(struct pfioc_nv *); 223 static int pf_clear_tables(void); 224 static void pf_clear_srcnodes(struct pf_ksrc_node *); 225 static void pf_kill_srcnodes(struct pfioc_src_node_kill *); 226 static int pf_keepcounters(struct pfioc_nv *); 227 static void pf_tbladdr_copyout(struct pf_addr_wrap *); 228 229 /* 230 * Wrapper functions for pfil(9) hooks 231 */ 232 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, 233 int flags, void *ruleset __unused, struct inpcb *inp); 234 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, 235 int flags, void *ruleset __unused, struct inpcb *inp); 236 #ifdef INET 237 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp, 238 int flags, void *ruleset __unused, struct inpcb *inp); 239 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp, 240 int flags, void *ruleset __unused, struct inpcb *inp); 241 #endif 242 #ifdef INET6 243 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp, 244 int flags, void *ruleset __unused, struct inpcb *inp); 245 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp, 246 int flags, void *ruleset __unused, struct inpcb *inp); 247 #endif 248 249 static void hook_pf_eth(void); 250 static void hook_pf(void); 251 static void dehook_pf_eth(void); 252 static void dehook_pf(void); 253 static int shutdown_pf(void); 254 static int pf_load(void); 255 static void pf_unload(void); 256 257 static struct cdevsw pf_cdevsw = { 258 .d_ioctl = pfioctl, 259 .d_name = PF_NAME, 260 .d_version = D_VERSION, 261 }; 262 263 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked); 264 #define V_pf_pfil_hooked VNET(pf_pfil_hooked) 265 volatile VNET_DEFINE_STATIC(int, pf_pfil_eth_hooked); 266 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked) 267 268 /* 269 * We need a flag that is neither hooked nor running to know when 270 * the VNET is "valid". We primarily need this to control (global) 271 * external event, e.g., eventhandlers. 272 */ 273 VNET_DEFINE(int, pf_vnet_active); 274 #define V_pf_vnet_active VNET(pf_vnet_active) 275 276 int pf_end_threads; 277 struct proc *pf_purge_proc; 278 279 struct rmlock pf_rules_lock; 280 struct sx pf_ioctl_lock; 281 struct sx pf_end_lock; 282 283 /* pfsync */ 284 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr); 285 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr); 286 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr); 287 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr); 288 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr); 289 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr); 290 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr; 291 292 /* pflog */ 293 pflog_packet_t *pflog_packet_ptr = NULL; 294 295 /* 296 * Copy a user-provided string, returning an error if truncation would occur. 297 * Avoid scanning past "sz" bytes in the source string since there's no 298 * guarantee that it's nul-terminated. 299 */ 300 static int 301 pf_user_strcpy(char *dst, const char *src, size_t sz) 302 { 303 if (strnlen(src, sz) == sz) 304 return (EINVAL); 305 (void)strlcpy(dst, src, sz); 306 return (0); 307 } 308 309 static void 310 pfattach_vnet(void) 311 { 312 u_int32_t *my_timeout = V_pf_default_rule.timeout; 313 314 pf_initialize(); 315 pfr_initialize(); 316 pfi_initialize_vnet(); 317 pf_normalize_init(); 318 pf_syncookies_init(); 319 320 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT; 321 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT; 322 323 RB_INIT(&V_pf_anchors); 324 pf_init_kruleset(&pf_main_ruleset); 325 326 pf_init_keth(V_pf_keth); 327 328 /* default rule should never be garbage collected */ 329 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; 330 #ifdef PF_DEFAULT_TO_DROP 331 V_pf_default_rule.action = PF_DROP; 332 #else 333 V_pf_default_rule.action = PF_PASS; 334 #endif 335 V_pf_default_rule.nr = -1; 336 V_pf_default_rule.rtableid = -1; 337 338 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK); 339 for (int i = 0; i < 2; i++) { 340 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK); 341 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK); 342 } 343 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); 344 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); 345 V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK); 346 347 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pcpu_zone_4, M_WAITOK | M_ZERO); 348 349 #ifdef PF_WANT_32_TO_64_COUNTER 350 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO); 351 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO); 352 PF_RULES_WLOCK(); 353 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); 354 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist); 355 V_pf_allrulecount++; 356 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); 357 PF_RULES_WUNLOCK(); 358 #endif 359 360 /* initialize default timeouts */ 361 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 362 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; 363 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 364 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; 365 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; 366 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; 367 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; 368 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; 369 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; 370 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; 371 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; 372 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; 373 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; 374 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; 375 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; 376 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; 377 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; 378 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; 379 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; 380 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; 381 382 bzero(&V_pf_status, sizeof(V_pf_status)); 383 V_pf_status.debug = PF_DEBUG_URGENT; 384 385 V_pf_pfil_hooked = 0; 386 V_pf_pfil_eth_hooked = 0; 387 388 /* XXX do our best to avoid a conflict */ 389 V_pf_status.hostid = arc4random(); 390 391 for (int i = 0; i < PFRES_MAX; i++) 392 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); 393 for (int i = 0; i < KLCNT_MAX; i++) 394 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); 395 for (int i = 0; i < FCNT_MAX; i++) 396 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK); 397 for (int i = 0; i < SCNT_MAX; i++) 398 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); 399 400 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET, 401 INTR_MPSAFE, &V_pf_swi_cookie) != 0) 402 /* XXXGL: leaked all above. */ 403 return; 404 } 405 406 static struct pf_kpool * 407 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action, 408 u_int32_t rule_number, u_int8_t r_last, u_int8_t active, 409 u_int8_t check_ticket) 410 { 411 struct pf_kruleset *ruleset; 412 struct pf_krule *rule; 413 int rs_num; 414 415 ruleset = pf_find_kruleset(anchor); 416 if (ruleset == NULL) 417 return (NULL); 418 rs_num = pf_get_ruleset_number(rule_action); 419 if (rs_num >= PF_RULESET_MAX) 420 return (NULL); 421 if (active) { 422 if (check_ticket && ticket != 423 ruleset->rules[rs_num].active.ticket) 424 return (NULL); 425 if (r_last) 426 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 427 pf_krulequeue); 428 else 429 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 430 } else { 431 if (check_ticket && ticket != 432 ruleset->rules[rs_num].inactive.ticket) 433 return (NULL); 434 if (r_last) 435 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 436 pf_krulequeue); 437 else 438 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); 439 } 440 if (!r_last) { 441 while ((rule != NULL) && (rule->nr != rule_number)) 442 rule = TAILQ_NEXT(rule, entries); 443 } 444 if (rule == NULL) 445 return (NULL); 446 447 return (&rule->rpool); 448 } 449 450 static void 451 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb) 452 { 453 struct pf_kpooladdr *mv_pool_pa; 454 455 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { 456 TAILQ_REMOVE(poola, mv_pool_pa, entries); 457 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); 458 } 459 } 460 461 static void 462 pf_empty_kpool(struct pf_kpalist *poola) 463 { 464 struct pf_kpooladdr *pa; 465 466 while ((pa = TAILQ_FIRST(poola)) != NULL) { 467 switch (pa->addr.type) { 468 case PF_ADDR_DYNIFTL: 469 pfi_dynaddr_remove(pa->addr.p.dyn); 470 break; 471 case PF_ADDR_TABLE: 472 /* XXX: this could be unfinished pooladdr on pabuf */ 473 if (pa->addr.p.tbl != NULL) 474 pfr_detach_table(pa->addr.p.tbl); 475 break; 476 } 477 if (pa->kif) 478 pfi_kkif_unref(pa->kif); 479 TAILQ_REMOVE(poola, pa, entries); 480 free(pa, M_PFRULE); 481 } 482 } 483 484 static void 485 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 486 { 487 488 PF_RULES_WASSERT(); 489 PF_UNLNKDRULES_ASSERT(); 490 491 TAILQ_REMOVE(rulequeue, rule, entries); 492 493 rule->rule_ref |= PFRULE_REFS; 494 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); 495 } 496 497 static void 498 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 499 { 500 501 PF_RULES_WASSERT(); 502 503 PF_UNLNKDRULES_LOCK(); 504 pf_unlink_rule_locked(rulequeue, rule); 505 PF_UNLNKDRULES_UNLOCK(); 506 } 507 508 static void 509 pf_free_eth_rule(struct pf_keth_rule *rule) 510 { 511 PF_RULES_WASSERT(); 512 513 if (rule == NULL) 514 return; 515 516 if (rule->tag) 517 tag_unref(&V_pf_tags, rule->tag); 518 #ifdef ALTQ 519 pf_qid_unref(rule->qid); 520 #endif 521 522 if (rule->kif) 523 pfi_kkif_unref(rule->kif); 524 525 if (rule->ipsrc.addr.type == PF_ADDR_TABLE) 526 pfr_detach_table(rule->ipsrc.addr.p.tbl); 527 if (rule->ipdst.addr.type == PF_ADDR_TABLE) 528 pfr_detach_table(rule->ipdst.addr.p.tbl); 529 530 counter_u64_free(rule->evaluations); 531 for (int i = 0; i < 2; i++) { 532 counter_u64_free(rule->packets[i]); 533 counter_u64_free(rule->bytes[i]); 534 } 535 uma_zfree_pcpu(pcpu_zone_4, rule->timestamp); 536 pf_keth_anchor_remove(rule); 537 538 free(rule, M_PFRULE); 539 } 540 541 void 542 pf_free_rule(struct pf_krule *rule) 543 { 544 545 PF_RULES_WASSERT(); 546 PF_CONFIG_ASSERT(); 547 548 if (rule->tag) 549 tag_unref(&V_pf_tags, rule->tag); 550 if (rule->match_tag) 551 tag_unref(&V_pf_tags, rule->match_tag); 552 #ifdef ALTQ 553 if (rule->pqid != rule->qid) 554 pf_qid_unref(rule->pqid); 555 pf_qid_unref(rule->qid); 556 #endif 557 switch (rule->src.addr.type) { 558 case PF_ADDR_DYNIFTL: 559 pfi_dynaddr_remove(rule->src.addr.p.dyn); 560 break; 561 case PF_ADDR_TABLE: 562 pfr_detach_table(rule->src.addr.p.tbl); 563 break; 564 } 565 switch (rule->dst.addr.type) { 566 case PF_ADDR_DYNIFTL: 567 pfi_dynaddr_remove(rule->dst.addr.p.dyn); 568 break; 569 case PF_ADDR_TABLE: 570 pfr_detach_table(rule->dst.addr.p.tbl); 571 break; 572 } 573 if (rule->overload_tbl) 574 pfr_detach_table(rule->overload_tbl); 575 if (rule->kif) 576 pfi_kkif_unref(rule->kif); 577 pf_kanchor_remove(rule); 578 pf_empty_kpool(&rule->rpool.list); 579 580 pf_krule_free(rule); 581 } 582 583 static void 584 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size, 585 unsigned int default_size) 586 { 587 unsigned int i; 588 unsigned int hashsize; 589 590 if (*tunable_size == 0 || !powerof2(*tunable_size)) 591 *tunable_size = default_size; 592 593 hashsize = *tunable_size; 594 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH, 595 M_WAITOK); 596 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH, 597 M_WAITOK); 598 ts->mask = hashsize - 1; 599 ts->seed = arc4random(); 600 for (i = 0; i < hashsize; i++) { 601 TAILQ_INIT(&ts->namehash[i]); 602 TAILQ_INIT(&ts->taghash[i]); 603 } 604 BIT_FILL(TAGID_MAX, &ts->avail); 605 } 606 607 static void 608 pf_cleanup_tagset(struct pf_tagset *ts) 609 { 610 unsigned int i; 611 unsigned int hashsize; 612 struct pf_tagname *t, *tmp; 613 614 /* 615 * Only need to clean up one of the hashes as each tag is hashed 616 * into each table. 617 */ 618 hashsize = ts->mask + 1; 619 for (i = 0; i < hashsize; i++) 620 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp) 621 uma_zfree(V_pf_tag_z, t); 622 623 free(ts->namehash, M_PFHASH); 624 free(ts->taghash, M_PFHASH); 625 } 626 627 static uint16_t 628 tagname2hashindex(const struct pf_tagset *ts, const char *tagname) 629 { 630 size_t len; 631 632 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1); 633 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask); 634 } 635 636 static uint16_t 637 tag2hashindex(const struct pf_tagset *ts, uint16_t tag) 638 { 639 640 return (tag & ts->mask); 641 } 642 643 static u_int16_t 644 tagname2tag(struct pf_tagset *ts, const char *tagname) 645 { 646 struct pf_tagname *tag; 647 u_int32_t index; 648 u_int16_t new_tagid; 649 650 PF_RULES_WASSERT(); 651 652 index = tagname2hashindex(ts, tagname); 653 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 654 if (strcmp(tagname, tag->name) == 0) { 655 tag->ref++; 656 return (tag->tag); 657 } 658 659 /* 660 * new entry 661 * 662 * to avoid fragmentation, we do a linear search from the beginning 663 * and take the first free slot we find. 664 */ 665 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail); 666 /* 667 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX]. 668 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits 669 * set. It may also return a bit number greater than TAGID_MAX due 670 * to rounding of the number of bits in the vector up to a multiple 671 * of the vector word size at declaration/allocation time. 672 */ 673 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) 674 return (0); 675 676 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */ 677 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail); 678 679 /* allocate and fill new struct pf_tagname */ 680 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT); 681 if (tag == NULL) 682 return (0); 683 strlcpy(tag->name, tagname, sizeof(tag->name)); 684 tag->tag = new_tagid; 685 tag->ref = 1; 686 687 /* Insert into namehash */ 688 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries); 689 690 /* Insert into taghash */ 691 index = tag2hashindex(ts, new_tagid); 692 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries); 693 694 return (tag->tag); 695 } 696 697 static void 698 tag_unref(struct pf_tagset *ts, u_int16_t tag) 699 { 700 struct pf_tagname *t; 701 uint16_t index; 702 703 PF_RULES_WASSERT(); 704 705 index = tag2hashindex(ts, tag); 706 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 707 if (tag == t->tag) { 708 if (--t->ref == 0) { 709 TAILQ_REMOVE(&ts->taghash[index], t, 710 taghash_entries); 711 index = tagname2hashindex(ts, t->name); 712 TAILQ_REMOVE(&ts->namehash[index], t, 713 namehash_entries); 714 /* Bits are 0-based for BIT_SET() */ 715 BIT_SET(TAGID_MAX, tag - 1, &ts->avail); 716 uma_zfree(V_pf_tag_z, t); 717 } 718 break; 719 } 720 } 721 722 static uint16_t 723 pf_tagname2tag(const char *tagname) 724 { 725 return (tagname2tag(&V_pf_tags, tagname)); 726 } 727 728 static int 729 pf_begin_eth(uint32_t *ticket, const char *anchor) 730 { 731 struct pf_keth_rule *rule, *tmp; 732 struct pf_keth_ruleset *rs; 733 734 PF_RULES_WASSERT(); 735 736 rs = pf_find_or_create_keth_ruleset(anchor); 737 if (rs == NULL) 738 return (EINVAL); 739 740 if (rs->inactive.open) 741 /* We may be waiting for NET_EPOCH_CALL(pf_rollback_eth_cb) to 742 * finish. */ 743 return (EBUSY); 744 745 /* Purge old inactive rules. */ 746 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 747 tmp) { 748 TAILQ_REMOVE(rs->inactive.rules, rule, 749 entries); 750 pf_free_eth_rule(rule); 751 } 752 753 *ticket = ++rs->inactive.ticket; 754 rs->inactive.open = 1; 755 756 return (0); 757 } 758 759 static void 760 pf_rollback_eth_cb(struct epoch_context *ctx) 761 { 762 struct pf_keth_ruleset *rs; 763 764 rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx); 765 766 CURVNET_SET(rs->vnet); 767 768 PF_RULES_WLOCK(); 769 pf_rollback_eth(rs->inactive.ticket, 770 rs->anchor ? rs->anchor->path : ""); 771 PF_RULES_WUNLOCK(); 772 773 CURVNET_RESTORE(); 774 } 775 776 static int 777 pf_rollback_eth(uint32_t ticket, const char *anchor) 778 { 779 struct pf_keth_rule *rule, *tmp; 780 struct pf_keth_ruleset *rs; 781 782 PF_RULES_WASSERT(); 783 784 rs = pf_find_keth_ruleset(anchor); 785 if (rs == NULL) 786 return (EINVAL); 787 788 if (!rs->inactive.open || 789 ticket != rs->inactive.ticket) 790 return (0); 791 792 /* Purge old inactive rules. */ 793 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 794 tmp) { 795 TAILQ_REMOVE(rs->inactive.rules, rule, entries); 796 pf_free_eth_rule(rule); 797 } 798 799 rs->inactive.open = 0; 800 801 pf_remove_if_empty_keth_ruleset(rs); 802 803 return (0); 804 } 805 806 #define PF_SET_SKIP_STEPS(i) \ 807 do { \ 808 while (head[i] != cur) { \ 809 head[i]->skip[i].ptr = cur; \ 810 head[i] = TAILQ_NEXT(head[i], entries); \ 811 } \ 812 } while (0) 813 814 static void 815 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules) 816 { 817 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT]; 818 int i; 819 820 cur = TAILQ_FIRST(rules); 821 prev = cur; 822 for (i = 0; i < PFE_SKIP_COUNT; ++i) 823 head[i] = cur; 824 while (cur != NULL) { 825 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 826 PF_SET_SKIP_STEPS(PFE_SKIP_IFP); 827 if (cur->direction != prev->direction) 828 PF_SET_SKIP_STEPS(PFE_SKIP_DIR); 829 if (cur->proto != prev->proto) 830 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO); 831 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0) 832 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR); 833 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0) 834 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR); 835 836 prev = cur; 837 cur = TAILQ_NEXT(cur, entries); 838 } 839 for (i = 0; i < PFE_SKIP_COUNT; ++i) 840 PF_SET_SKIP_STEPS(i); 841 } 842 843 static int 844 pf_commit_eth(uint32_t ticket, const char *anchor) 845 { 846 struct pf_keth_ruleq *rules; 847 struct pf_keth_ruleset *rs; 848 849 rs = pf_find_keth_ruleset(anchor); 850 if (rs == NULL) { 851 return (EINVAL); 852 } 853 854 if (!rs->inactive.open || 855 ticket != rs->inactive.ticket) 856 return (EBUSY); 857 858 PF_RULES_WASSERT(); 859 860 pf_eth_calc_skip_steps(rs->inactive.rules); 861 862 rules = rs->active.rules; 863 ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules); 864 rs->inactive.rules = rules; 865 rs->inactive.ticket = rs->active.ticket; 866 867 /* Clean up inactive rules (i.e. previously active rules), only when 868 * we're sure they're no longer used. */ 869 NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx); 870 871 return (0); 872 } 873 874 #ifdef ALTQ 875 static uint16_t 876 pf_qname2qid(const char *qname) 877 { 878 return (tagname2tag(&V_pf_qids, qname)); 879 } 880 881 static void 882 pf_qid_unref(uint16_t qid) 883 { 884 tag_unref(&V_pf_qids, qid); 885 } 886 887 static int 888 pf_begin_altq(u_int32_t *ticket) 889 { 890 struct pf_altq *altq, *tmp; 891 int error = 0; 892 893 PF_RULES_WASSERT(); 894 895 /* Purge the old altq lists */ 896 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 897 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 898 /* detach and destroy the discipline */ 899 error = altq_remove(altq); 900 } 901 free(altq, M_PFALTQ); 902 } 903 TAILQ_INIT(V_pf_altq_ifs_inactive); 904 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 905 pf_qid_unref(altq->qid); 906 free(altq, M_PFALTQ); 907 } 908 TAILQ_INIT(V_pf_altqs_inactive); 909 if (error) 910 return (error); 911 *ticket = ++V_ticket_altqs_inactive; 912 V_altqs_inactive_open = 1; 913 return (0); 914 } 915 916 static int 917 pf_rollback_altq(u_int32_t ticket) 918 { 919 struct pf_altq *altq, *tmp; 920 int error = 0; 921 922 PF_RULES_WASSERT(); 923 924 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 925 return (0); 926 /* Purge the old altq lists */ 927 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 928 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 929 /* detach and destroy the discipline */ 930 error = altq_remove(altq); 931 } 932 free(altq, M_PFALTQ); 933 } 934 TAILQ_INIT(V_pf_altq_ifs_inactive); 935 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 936 pf_qid_unref(altq->qid); 937 free(altq, M_PFALTQ); 938 } 939 TAILQ_INIT(V_pf_altqs_inactive); 940 V_altqs_inactive_open = 0; 941 return (error); 942 } 943 944 static int 945 pf_commit_altq(u_int32_t ticket) 946 { 947 struct pf_altqqueue *old_altqs, *old_altq_ifs; 948 struct pf_altq *altq, *tmp; 949 int err, error = 0; 950 951 PF_RULES_WASSERT(); 952 953 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 954 return (EBUSY); 955 956 /* swap altqs, keep the old. */ 957 old_altqs = V_pf_altqs_active; 958 old_altq_ifs = V_pf_altq_ifs_active; 959 V_pf_altqs_active = V_pf_altqs_inactive; 960 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive; 961 V_pf_altqs_inactive = old_altqs; 962 V_pf_altq_ifs_inactive = old_altq_ifs; 963 V_ticket_altqs_active = V_ticket_altqs_inactive; 964 965 /* Attach new disciplines */ 966 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 967 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 968 /* attach the discipline */ 969 error = altq_pfattach(altq); 970 if (error == 0 && V_pf_altq_running) 971 error = pf_enable_altq(altq); 972 if (error != 0) 973 return (error); 974 } 975 } 976 977 /* Purge the old altq lists */ 978 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 979 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 980 /* detach and destroy the discipline */ 981 if (V_pf_altq_running) 982 error = pf_disable_altq(altq); 983 err = altq_pfdetach(altq); 984 if (err != 0 && error == 0) 985 error = err; 986 err = altq_remove(altq); 987 if (err != 0 && error == 0) 988 error = err; 989 } 990 free(altq, M_PFALTQ); 991 } 992 TAILQ_INIT(V_pf_altq_ifs_inactive); 993 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 994 pf_qid_unref(altq->qid); 995 free(altq, M_PFALTQ); 996 } 997 TAILQ_INIT(V_pf_altqs_inactive); 998 999 V_altqs_inactive_open = 0; 1000 return (error); 1001 } 1002 1003 static int 1004 pf_enable_altq(struct pf_altq *altq) 1005 { 1006 struct ifnet *ifp; 1007 struct tb_profile tb; 1008 int error = 0; 1009 1010 if ((ifp = ifunit(altq->ifname)) == NULL) 1011 return (EINVAL); 1012 1013 if (ifp->if_snd.altq_type != ALTQT_NONE) 1014 error = altq_enable(&ifp->if_snd); 1015 1016 /* set tokenbucket regulator */ 1017 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { 1018 tb.rate = altq->ifbandwidth; 1019 tb.depth = altq->tbrsize; 1020 error = tbr_set(&ifp->if_snd, &tb); 1021 } 1022 1023 return (error); 1024 } 1025 1026 static int 1027 pf_disable_altq(struct pf_altq *altq) 1028 { 1029 struct ifnet *ifp; 1030 struct tb_profile tb; 1031 int error; 1032 1033 if ((ifp = ifunit(altq->ifname)) == NULL) 1034 return (EINVAL); 1035 1036 /* 1037 * when the discipline is no longer referenced, it was overridden 1038 * by a new one. if so, just return. 1039 */ 1040 if (altq->altq_disc != ifp->if_snd.altq_disc) 1041 return (0); 1042 1043 error = altq_disable(&ifp->if_snd); 1044 1045 if (error == 0) { 1046 /* clear tokenbucket regulator */ 1047 tb.rate = 0; 1048 error = tbr_set(&ifp->if_snd, &tb); 1049 } 1050 1051 return (error); 1052 } 1053 1054 static int 1055 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket, 1056 struct pf_altq *altq) 1057 { 1058 struct ifnet *ifp1; 1059 int error = 0; 1060 1061 /* Deactivate the interface in question */ 1062 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; 1063 if ((ifp1 = ifunit(altq->ifname)) == NULL || 1064 (remove && ifp1 == ifp)) { 1065 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 1066 } else { 1067 error = altq_add(ifp1, altq); 1068 1069 if (ticket != V_ticket_altqs_inactive) 1070 error = EBUSY; 1071 1072 if (error) 1073 free(altq, M_PFALTQ); 1074 } 1075 1076 return (error); 1077 } 1078 1079 void 1080 pf_altq_ifnet_event(struct ifnet *ifp, int remove) 1081 { 1082 struct pf_altq *a1, *a2, *a3; 1083 u_int32_t ticket; 1084 int error = 0; 1085 1086 /* 1087 * No need to re-evaluate the configuration for events on interfaces 1088 * that do not support ALTQ, as it's not possible for such 1089 * interfaces to be part of the configuration. 1090 */ 1091 if (!ALTQ_IS_READY(&ifp->if_snd)) 1092 return; 1093 1094 /* Interrupt userland queue modifications */ 1095 if (V_altqs_inactive_open) 1096 pf_rollback_altq(V_ticket_altqs_inactive); 1097 1098 /* Start new altq ruleset */ 1099 if (pf_begin_altq(&ticket)) 1100 return; 1101 1102 /* Copy the current active set */ 1103 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) { 1104 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1105 if (a2 == NULL) { 1106 error = ENOMEM; 1107 break; 1108 } 1109 bcopy(a1, a2, sizeof(struct pf_altq)); 1110 1111 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1112 if (error) 1113 break; 1114 1115 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries); 1116 } 1117 if (error) 1118 goto out; 1119 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { 1120 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1121 if (a2 == NULL) { 1122 error = ENOMEM; 1123 break; 1124 } 1125 bcopy(a1, a2, sizeof(struct pf_altq)); 1126 1127 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) { 1128 error = EBUSY; 1129 free(a2, M_PFALTQ); 1130 break; 1131 } 1132 a2->altq_disc = NULL; 1133 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) { 1134 if (strncmp(a3->ifname, a2->ifname, 1135 IFNAMSIZ) == 0) { 1136 a2->altq_disc = a3->altq_disc; 1137 break; 1138 } 1139 } 1140 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1141 if (error) 1142 break; 1143 1144 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); 1145 } 1146 1147 out: 1148 if (error != 0) 1149 pf_rollback_altq(ticket); 1150 else 1151 pf_commit_altq(ticket); 1152 } 1153 #endif /* ALTQ */ 1154 1155 static struct pf_krule_global * 1156 pf_rule_tree_alloc(int flags) 1157 { 1158 struct pf_krule_global *tree; 1159 1160 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags); 1161 if (tree == NULL) 1162 return (NULL); 1163 RB_INIT(tree); 1164 return (tree); 1165 } 1166 1167 static void 1168 pf_rule_tree_free(struct pf_krule_global *tree) 1169 { 1170 1171 free(tree, M_TEMP); 1172 } 1173 1174 static int 1175 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) 1176 { 1177 struct pf_krule_global *tree; 1178 struct pf_kruleset *rs; 1179 struct pf_krule *rule; 1180 1181 PF_RULES_WASSERT(); 1182 1183 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1184 return (EINVAL); 1185 tree = pf_rule_tree_alloc(M_NOWAIT); 1186 if (tree == NULL) 1187 return (ENOMEM); 1188 rs = pf_find_or_create_kruleset(anchor); 1189 if (rs == NULL) { 1190 free(tree, M_TEMP); 1191 return (EINVAL); 1192 } 1193 pf_rule_tree_free(rs->rules[rs_num].inactive.tree); 1194 rs->rules[rs_num].inactive.tree = tree; 1195 1196 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1197 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1198 rs->rules[rs_num].inactive.rcount--; 1199 } 1200 *ticket = ++rs->rules[rs_num].inactive.ticket; 1201 rs->rules[rs_num].inactive.open = 1; 1202 return (0); 1203 } 1204 1205 static int 1206 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) 1207 { 1208 struct pf_kruleset *rs; 1209 struct pf_krule *rule; 1210 1211 PF_RULES_WASSERT(); 1212 1213 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1214 return (EINVAL); 1215 rs = pf_find_kruleset(anchor); 1216 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1217 rs->rules[rs_num].inactive.ticket != ticket) 1218 return (0); 1219 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1220 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1221 rs->rules[rs_num].inactive.rcount--; 1222 } 1223 rs->rules[rs_num].inactive.open = 0; 1224 return (0); 1225 } 1226 1227 #define PF_MD5_UPD(st, elm) \ 1228 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) 1229 1230 #define PF_MD5_UPD_STR(st, elm) \ 1231 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) 1232 1233 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ 1234 (stor) = htonl((st)->elm); \ 1235 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ 1236 } while (0) 1237 1238 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ 1239 (stor) = htons((st)->elm); \ 1240 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ 1241 } while (0) 1242 1243 static void 1244 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) 1245 { 1246 PF_MD5_UPD(pfr, addr.type); 1247 switch (pfr->addr.type) { 1248 case PF_ADDR_DYNIFTL: 1249 PF_MD5_UPD(pfr, addr.v.ifname); 1250 PF_MD5_UPD(pfr, addr.iflags); 1251 break; 1252 case PF_ADDR_TABLE: 1253 PF_MD5_UPD(pfr, addr.v.tblname); 1254 break; 1255 case PF_ADDR_ADDRMASK: 1256 /* XXX ignore af? */ 1257 PF_MD5_UPD(pfr, addr.v.a.addr.addr32); 1258 PF_MD5_UPD(pfr, addr.v.a.mask.addr32); 1259 break; 1260 } 1261 1262 PF_MD5_UPD(pfr, port[0]); 1263 PF_MD5_UPD(pfr, port[1]); 1264 PF_MD5_UPD(pfr, neg); 1265 PF_MD5_UPD(pfr, port_op); 1266 } 1267 1268 static void 1269 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule) 1270 { 1271 u_int16_t x; 1272 u_int32_t y; 1273 1274 pf_hash_rule_addr(ctx, &rule->src); 1275 pf_hash_rule_addr(ctx, &rule->dst); 1276 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) 1277 PF_MD5_UPD_STR(rule, label[i]); 1278 PF_MD5_UPD_STR(rule, ifname); 1279 PF_MD5_UPD_STR(rule, match_tagname); 1280 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ 1281 PF_MD5_UPD_HTONL(rule, os_fingerprint, y); 1282 PF_MD5_UPD_HTONL(rule, prob, y); 1283 PF_MD5_UPD_HTONL(rule, uid.uid[0], y); 1284 PF_MD5_UPD_HTONL(rule, uid.uid[1], y); 1285 PF_MD5_UPD(rule, uid.op); 1286 PF_MD5_UPD_HTONL(rule, gid.gid[0], y); 1287 PF_MD5_UPD_HTONL(rule, gid.gid[1], y); 1288 PF_MD5_UPD(rule, gid.op); 1289 PF_MD5_UPD_HTONL(rule, rule_flag, y); 1290 PF_MD5_UPD(rule, action); 1291 PF_MD5_UPD(rule, direction); 1292 PF_MD5_UPD(rule, af); 1293 PF_MD5_UPD(rule, quick); 1294 PF_MD5_UPD(rule, ifnot); 1295 PF_MD5_UPD(rule, match_tag_not); 1296 PF_MD5_UPD(rule, natpass); 1297 PF_MD5_UPD(rule, keep_state); 1298 PF_MD5_UPD(rule, proto); 1299 PF_MD5_UPD(rule, type); 1300 PF_MD5_UPD(rule, code); 1301 PF_MD5_UPD(rule, flags); 1302 PF_MD5_UPD(rule, flagset); 1303 PF_MD5_UPD(rule, allow_opts); 1304 PF_MD5_UPD(rule, rt); 1305 PF_MD5_UPD(rule, tos); 1306 if (rule->anchor != NULL) 1307 PF_MD5_UPD_STR(rule, anchor->path); 1308 } 1309 1310 static void 1311 pf_hash_rule(struct pf_krule *rule) 1312 { 1313 MD5_CTX ctx; 1314 1315 MD5Init(&ctx); 1316 pf_hash_rule_rolling(&ctx, rule); 1317 MD5Final(rule->md5sum, &ctx); 1318 } 1319 1320 static int 1321 pf_krule_compare(struct pf_krule *a, struct pf_krule *b) 1322 { 1323 1324 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH)); 1325 } 1326 1327 static int 1328 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) 1329 { 1330 struct pf_kruleset *rs; 1331 struct pf_krule *rule, **old_array, *old_rule; 1332 struct pf_krulequeue *old_rules; 1333 struct pf_krule_global *old_tree; 1334 int error; 1335 u_int32_t old_rcount; 1336 1337 PF_RULES_WASSERT(); 1338 1339 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1340 return (EINVAL); 1341 rs = pf_find_kruleset(anchor); 1342 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1343 ticket != rs->rules[rs_num].inactive.ticket) 1344 return (EBUSY); 1345 1346 /* Calculate checksum for the main ruleset */ 1347 if (rs == &pf_main_ruleset) { 1348 error = pf_setup_pfsync_matching(rs); 1349 if (error != 0) 1350 return (error); 1351 } 1352 1353 /* Swap rules, keep the old. */ 1354 old_rules = rs->rules[rs_num].active.ptr; 1355 old_rcount = rs->rules[rs_num].active.rcount; 1356 old_array = rs->rules[rs_num].active.ptr_array; 1357 old_tree = rs->rules[rs_num].active.tree; 1358 1359 rs->rules[rs_num].active.ptr = 1360 rs->rules[rs_num].inactive.ptr; 1361 rs->rules[rs_num].active.ptr_array = 1362 rs->rules[rs_num].inactive.ptr_array; 1363 rs->rules[rs_num].active.tree = 1364 rs->rules[rs_num].inactive.tree; 1365 rs->rules[rs_num].active.rcount = 1366 rs->rules[rs_num].inactive.rcount; 1367 1368 /* Attempt to preserve counter information. */ 1369 if (V_pf_status.keep_counters && old_tree != NULL) { 1370 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr, 1371 entries) { 1372 old_rule = RB_FIND(pf_krule_global, old_tree, rule); 1373 if (old_rule == NULL) { 1374 continue; 1375 } 1376 pf_counter_u64_critical_enter(); 1377 pf_counter_u64_add_protected(&rule->evaluations, 1378 pf_counter_u64_fetch(&old_rule->evaluations)); 1379 pf_counter_u64_add_protected(&rule->packets[0], 1380 pf_counter_u64_fetch(&old_rule->packets[0])); 1381 pf_counter_u64_add_protected(&rule->packets[1], 1382 pf_counter_u64_fetch(&old_rule->packets[1])); 1383 pf_counter_u64_add_protected(&rule->bytes[0], 1384 pf_counter_u64_fetch(&old_rule->bytes[0])); 1385 pf_counter_u64_add_protected(&rule->bytes[1], 1386 pf_counter_u64_fetch(&old_rule->bytes[1])); 1387 pf_counter_u64_critical_exit(); 1388 } 1389 } 1390 1391 rs->rules[rs_num].inactive.ptr = old_rules; 1392 rs->rules[rs_num].inactive.ptr_array = old_array; 1393 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ 1394 rs->rules[rs_num].inactive.rcount = old_rcount; 1395 1396 rs->rules[rs_num].active.ticket = 1397 rs->rules[rs_num].inactive.ticket; 1398 pf_calc_skip_steps(rs->rules[rs_num].active.ptr); 1399 1400 /* Purge the old rule list. */ 1401 PF_UNLNKDRULES_LOCK(); 1402 while ((rule = TAILQ_FIRST(old_rules)) != NULL) 1403 pf_unlink_rule_locked(old_rules, rule); 1404 PF_UNLNKDRULES_UNLOCK(); 1405 if (rs->rules[rs_num].inactive.ptr_array) 1406 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP); 1407 rs->rules[rs_num].inactive.ptr_array = NULL; 1408 rs->rules[rs_num].inactive.rcount = 0; 1409 rs->rules[rs_num].inactive.open = 0; 1410 pf_remove_if_empty_kruleset(rs); 1411 free(old_tree, M_TEMP); 1412 1413 return (0); 1414 } 1415 1416 static int 1417 pf_setup_pfsync_matching(struct pf_kruleset *rs) 1418 { 1419 MD5_CTX ctx; 1420 struct pf_krule *rule; 1421 int rs_cnt; 1422 u_int8_t digest[PF_MD5_DIGEST_LENGTH]; 1423 1424 MD5Init(&ctx); 1425 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { 1426 /* XXX PF_RULESET_SCRUB as well? */ 1427 if (rs_cnt == PF_RULESET_SCRUB) 1428 continue; 1429 1430 if (rs->rules[rs_cnt].inactive.ptr_array) 1431 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP); 1432 rs->rules[rs_cnt].inactive.ptr_array = NULL; 1433 1434 if (rs->rules[rs_cnt].inactive.rcount) { 1435 rs->rules[rs_cnt].inactive.ptr_array = 1436 malloc(sizeof(caddr_t) * 1437 rs->rules[rs_cnt].inactive.rcount, 1438 M_TEMP, M_NOWAIT); 1439 1440 if (!rs->rules[rs_cnt].inactive.ptr_array) 1441 return (ENOMEM); 1442 } 1443 1444 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, 1445 entries) { 1446 pf_hash_rule_rolling(&ctx, rule); 1447 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule; 1448 } 1449 } 1450 1451 MD5Final(digest, &ctx); 1452 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); 1453 return (0); 1454 } 1455 1456 static int 1457 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr) 1458 { 1459 int error = 0; 1460 1461 switch (addr->type) { 1462 case PF_ADDR_TABLE: 1463 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname); 1464 if (addr->p.tbl == NULL) 1465 error = ENOMEM; 1466 break; 1467 default: 1468 error = EINVAL; 1469 } 1470 1471 return (error); 1472 } 1473 1474 static int 1475 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr, 1476 sa_family_t af) 1477 { 1478 int error = 0; 1479 1480 switch (addr->type) { 1481 case PF_ADDR_TABLE: 1482 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); 1483 if (addr->p.tbl == NULL) 1484 error = ENOMEM; 1485 break; 1486 case PF_ADDR_DYNIFTL: 1487 error = pfi_dynaddr_setup(addr, af); 1488 break; 1489 } 1490 1491 return (error); 1492 } 1493 1494 static void 1495 pf_addr_copyout(struct pf_addr_wrap *addr) 1496 { 1497 1498 switch (addr->type) { 1499 case PF_ADDR_DYNIFTL: 1500 pfi_dynaddr_copyout(addr); 1501 break; 1502 case PF_ADDR_TABLE: 1503 pf_tbladdr_copyout(addr); 1504 break; 1505 } 1506 } 1507 1508 static void 1509 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out) 1510 { 1511 int secs = time_uptime, diff; 1512 1513 bzero(out, sizeof(struct pf_src_node)); 1514 1515 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr)); 1516 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr)); 1517 1518 if (in->rule.ptr != NULL) 1519 out->rule.nr = in->rule.ptr->nr; 1520 1521 for (int i = 0; i < 2; i++) { 1522 out->bytes[i] = counter_u64_fetch(in->bytes[i]); 1523 out->packets[i] = counter_u64_fetch(in->packets[i]); 1524 } 1525 1526 out->states = in->states; 1527 out->conn = in->conn; 1528 out->af = in->af; 1529 out->ruletype = in->ruletype; 1530 1531 out->creation = secs - in->creation; 1532 if (out->expire > secs) 1533 out->expire -= secs; 1534 else 1535 out->expire = 0; 1536 1537 /* Adjust the connection rate estimate. */ 1538 diff = secs - in->conn_rate.last; 1539 if (diff >= in->conn_rate.seconds) 1540 out->conn_rate.count = 0; 1541 else 1542 out->conn_rate.count -= 1543 in->conn_rate.count * diff / 1544 in->conn_rate.seconds; 1545 } 1546 1547 #ifdef ALTQ 1548 /* 1549 * Handle export of struct pf_kaltq to user binaries that may be using any 1550 * version of struct pf_altq. 1551 */ 1552 static int 1553 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size) 1554 { 1555 u_int32_t version; 1556 1557 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1558 version = 0; 1559 else 1560 version = pa->version; 1561 1562 if (version > PFIOC_ALTQ_VERSION) 1563 return (EINVAL); 1564 1565 #define ASSIGN(x) exported_q->x = q->x 1566 #define COPY(x) \ 1567 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x))) 1568 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX) 1569 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX) 1570 1571 switch (version) { 1572 case 0: { 1573 struct pf_altq_v0 *exported_q = 1574 &((struct pfioc_altq_v0 *)pa)->altq; 1575 1576 COPY(ifname); 1577 1578 ASSIGN(scheduler); 1579 ASSIGN(tbrsize); 1580 exported_q->tbrsize = SATU16(q->tbrsize); 1581 exported_q->ifbandwidth = SATU32(q->ifbandwidth); 1582 1583 COPY(qname); 1584 COPY(parent); 1585 ASSIGN(parent_qid); 1586 exported_q->bandwidth = SATU32(q->bandwidth); 1587 ASSIGN(priority); 1588 ASSIGN(local_flags); 1589 1590 ASSIGN(qlimit); 1591 ASSIGN(flags); 1592 1593 if (q->scheduler == ALTQT_HFSC) { 1594 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x 1595 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \ 1596 SATU32(q->pq_u.hfsc_opts.x) 1597 1598 ASSIGN_OPT_SATU32(rtsc_m1); 1599 ASSIGN_OPT(rtsc_d); 1600 ASSIGN_OPT_SATU32(rtsc_m2); 1601 1602 ASSIGN_OPT_SATU32(lssc_m1); 1603 ASSIGN_OPT(lssc_d); 1604 ASSIGN_OPT_SATU32(lssc_m2); 1605 1606 ASSIGN_OPT_SATU32(ulsc_m1); 1607 ASSIGN_OPT(ulsc_d); 1608 ASSIGN_OPT_SATU32(ulsc_m2); 1609 1610 ASSIGN_OPT(flags); 1611 1612 #undef ASSIGN_OPT 1613 #undef ASSIGN_OPT_SATU32 1614 } else 1615 COPY(pq_u); 1616 1617 ASSIGN(qid); 1618 break; 1619 } 1620 case 1: { 1621 struct pf_altq_v1 *exported_q = 1622 &((struct pfioc_altq_v1 *)pa)->altq; 1623 1624 COPY(ifname); 1625 1626 ASSIGN(scheduler); 1627 ASSIGN(tbrsize); 1628 ASSIGN(ifbandwidth); 1629 1630 COPY(qname); 1631 COPY(parent); 1632 ASSIGN(parent_qid); 1633 ASSIGN(bandwidth); 1634 ASSIGN(priority); 1635 ASSIGN(local_flags); 1636 1637 ASSIGN(qlimit); 1638 ASSIGN(flags); 1639 COPY(pq_u); 1640 1641 ASSIGN(qid); 1642 break; 1643 } 1644 default: 1645 panic("%s: unhandled struct pfioc_altq version", __func__); 1646 break; 1647 } 1648 1649 #undef ASSIGN 1650 #undef COPY 1651 #undef SATU16 1652 #undef SATU32 1653 1654 return (0); 1655 } 1656 1657 /* 1658 * Handle import to struct pf_kaltq of struct pf_altq from user binaries 1659 * that may be using any version of it. 1660 */ 1661 static int 1662 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size) 1663 { 1664 u_int32_t version; 1665 1666 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1667 version = 0; 1668 else 1669 version = pa->version; 1670 1671 if (version > PFIOC_ALTQ_VERSION) 1672 return (EINVAL); 1673 1674 #define ASSIGN(x) q->x = imported_q->x 1675 #define COPY(x) \ 1676 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x))) 1677 1678 switch (version) { 1679 case 0: { 1680 struct pf_altq_v0 *imported_q = 1681 &((struct pfioc_altq_v0 *)pa)->altq; 1682 1683 COPY(ifname); 1684 1685 ASSIGN(scheduler); 1686 ASSIGN(tbrsize); /* 16-bit -> 32-bit */ 1687 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */ 1688 1689 COPY(qname); 1690 COPY(parent); 1691 ASSIGN(parent_qid); 1692 ASSIGN(bandwidth); /* 32-bit -> 64-bit */ 1693 ASSIGN(priority); 1694 ASSIGN(local_flags); 1695 1696 ASSIGN(qlimit); 1697 ASSIGN(flags); 1698 1699 if (imported_q->scheduler == ALTQT_HFSC) { 1700 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x 1701 1702 /* 1703 * The m1 and m2 parameters are being copied from 1704 * 32-bit to 64-bit. 1705 */ 1706 ASSIGN_OPT(rtsc_m1); 1707 ASSIGN_OPT(rtsc_d); 1708 ASSIGN_OPT(rtsc_m2); 1709 1710 ASSIGN_OPT(lssc_m1); 1711 ASSIGN_OPT(lssc_d); 1712 ASSIGN_OPT(lssc_m2); 1713 1714 ASSIGN_OPT(ulsc_m1); 1715 ASSIGN_OPT(ulsc_d); 1716 ASSIGN_OPT(ulsc_m2); 1717 1718 ASSIGN_OPT(flags); 1719 1720 #undef ASSIGN_OPT 1721 } else 1722 COPY(pq_u); 1723 1724 ASSIGN(qid); 1725 break; 1726 } 1727 case 1: { 1728 struct pf_altq_v1 *imported_q = 1729 &((struct pfioc_altq_v1 *)pa)->altq; 1730 1731 COPY(ifname); 1732 1733 ASSIGN(scheduler); 1734 ASSIGN(tbrsize); 1735 ASSIGN(ifbandwidth); 1736 1737 COPY(qname); 1738 COPY(parent); 1739 ASSIGN(parent_qid); 1740 ASSIGN(bandwidth); 1741 ASSIGN(priority); 1742 ASSIGN(local_flags); 1743 1744 ASSIGN(qlimit); 1745 ASSIGN(flags); 1746 COPY(pq_u); 1747 1748 ASSIGN(qid); 1749 break; 1750 } 1751 default: 1752 panic("%s: unhandled struct pfioc_altq version", __func__); 1753 break; 1754 } 1755 1756 #undef ASSIGN 1757 #undef COPY 1758 1759 return (0); 1760 } 1761 1762 static struct pf_altq * 1763 pf_altq_get_nth_active(u_int32_t n) 1764 { 1765 struct pf_altq *altq; 1766 u_int32_t nr; 1767 1768 nr = 0; 1769 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1770 if (nr == n) 1771 return (altq); 1772 nr++; 1773 } 1774 1775 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { 1776 if (nr == n) 1777 return (altq); 1778 nr++; 1779 } 1780 1781 return (NULL); 1782 } 1783 #endif /* ALTQ */ 1784 1785 struct pf_krule * 1786 pf_krule_alloc(void) 1787 { 1788 struct pf_krule *rule; 1789 1790 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO); 1791 mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF); 1792 rule->timestamp = uma_zalloc_pcpu(pcpu_zone_4, M_WAITOK | M_ZERO); 1793 return (rule); 1794 } 1795 1796 void 1797 pf_krule_free(struct pf_krule *rule) 1798 { 1799 #ifdef PF_WANT_32_TO_64_COUNTER 1800 bool wowned; 1801 #endif 1802 1803 if (rule == NULL) 1804 return; 1805 1806 #ifdef PF_WANT_32_TO_64_COUNTER 1807 if (rule->allrulelinked) { 1808 wowned = PF_RULES_WOWNED(); 1809 if (!wowned) 1810 PF_RULES_WLOCK(); 1811 LIST_REMOVE(rule, allrulelist); 1812 V_pf_allrulecount--; 1813 if (!wowned) 1814 PF_RULES_WUNLOCK(); 1815 } 1816 #endif 1817 1818 pf_counter_u64_deinit(&rule->evaluations); 1819 for (int i = 0; i < 2; i++) { 1820 pf_counter_u64_deinit(&rule->packets[i]); 1821 pf_counter_u64_deinit(&rule->bytes[i]); 1822 } 1823 counter_u64_free(rule->states_cur); 1824 counter_u64_free(rule->states_tot); 1825 counter_u64_free(rule->src_nodes); 1826 uma_zfree_pcpu(pcpu_zone_4, rule->timestamp); 1827 1828 mtx_destroy(&rule->rpool.mtx); 1829 free(rule, M_PFRULE); 1830 } 1831 1832 static void 1833 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool, 1834 struct pf_pooladdr *pool) 1835 { 1836 1837 bzero(pool, sizeof(*pool)); 1838 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr)); 1839 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname)); 1840 } 1841 1842 static int 1843 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool, 1844 struct pf_kpooladdr *kpool) 1845 { 1846 int ret; 1847 1848 bzero(kpool, sizeof(*kpool)); 1849 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr)); 1850 ret = pf_user_strcpy(kpool->ifname, pool->ifname, 1851 sizeof(kpool->ifname)); 1852 return (ret); 1853 } 1854 1855 static void 1856 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool) 1857 { 1858 bzero(pool, sizeof(*pool)); 1859 1860 bcopy(&kpool->key, &pool->key, sizeof(pool->key)); 1861 bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter)); 1862 1863 pool->tblidx = kpool->tblidx; 1864 pool->proxy_port[0] = kpool->proxy_port[0]; 1865 pool->proxy_port[1] = kpool->proxy_port[1]; 1866 pool->opts = kpool->opts; 1867 } 1868 1869 static void 1870 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool) 1871 { 1872 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), ""); 1873 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), ""); 1874 1875 bcopy(&pool->key, &kpool->key, sizeof(kpool->key)); 1876 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter)); 1877 1878 kpool->tblidx = pool->tblidx; 1879 kpool->proxy_port[0] = pool->proxy_port[0]; 1880 kpool->proxy_port[1] = pool->proxy_port[1]; 1881 kpool->opts = pool->opts; 1882 } 1883 1884 static void 1885 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule) 1886 { 1887 1888 bzero(rule, sizeof(*rule)); 1889 1890 bcopy(&krule->src, &rule->src, sizeof(rule->src)); 1891 bcopy(&krule->dst, &rule->dst, sizeof(rule->dst)); 1892 1893 for (int i = 0; i < PF_SKIP_COUNT; ++i) { 1894 if (rule->skip[i].ptr == NULL) 1895 rule->skip[i].nr = -1; 1896 else 1897 rule->skip[i].nr = krule->skip[i].ptr->nr; 1898 } 1899 1900 strlcpy(rule->label, krule->label[0], sizeof(rule->label)); 1901 strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname)); 1902 strlcpy(rule->qname, krule->qname, sizeof(rule->qname)); 1903 strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname)); 1904 strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname)); 1905 strlcpy(rule->match_tagname, krule->match_tagname, 1906 sizeof(rule->match_tagname)); 1907 strlcpy(rule->overload_tblname, krule->overload_tblname, 1908 sizeof(rule->overload_tblname)); 1909 1910 pf_kpool_to_pool(&krule->rpool, &rule->rpool); 1911 1912 rule->evaluations = pf_counter_u64_fetch(&krule->evaluations); 1913 for (int i = 0; i < 2; i++) { 1914 rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]); 1915 rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]); 1916 } 1917 1918 /* kif, anchor, overload_tbl are not copied over. */ 1919 1920 rule->os_fingerprint = krule->os_fingerprint; 1921 1922 rule->rtableid = krule->rtableid; 1923 bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout)); 1924 rule->max_states = krule->max_states; 1925 rule->max_src_nodes = krule->max_src_nodes; 1926 rule->max_src_states = krule->max_src_states; 1927 rule->max_src_conn = krule->max_src_conn; 1928 rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit; 1929 rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds; 1930 rule->qid = krule->qid; 1931 rule->pqid = krule->pqid; 1932 rule->nr = krule->nr; 1933 rule->prob = krule->prob; 1934 rule->cuid = krule->cuid; 1935 rule->cpid = krule->cpid; 1936 1937 rule->return_icmp = krule->return_icmp; 1938 rule->return_icmp6 = krule->return_icmp6; 1939 rule->max_mss = krule->max_mss; 1940 rule->tag = krule->tag; 1941 rule->match_tag = krule->match_tag; 1942 rule->scrub_flags = krule->scrub_flags; 1943 1944 bcopy(&krule->uid, &rule->uid, sizeof(krule->uid)); 1945 bcopy(&krule->gid, &rule->gid, sizeof(krule->gid)); 1946 1947 rule->rule_flag = krule->rule_flag; 1948 rule->action = krule->action; 1949 rule->direction = krule->direction; 1950 rule->log = krule->log; 1951 rule->logif = krule->logif; 1952 rule->quick = krule->quick; 1953 rule->ifnot = krule->ifnot; 1954 rule->match_tag_not = krule->match_tag_not; 1955 rule->natpass = krule->natpass; 1956 1957 rule->keep_state = krule->keep_state; 1958 rule->af = krule->af; 1959 rule->proto = krule->proto; 1960 rule->type = krule->type; 1961 rule->code = krule->code; 1962 rule->flags = krule->flags; 1963 rule->flagset = krule->flagset; 1964 rule->min_ttl = krule->min_ttl; 1965 rule->allow_opts = krule->allow_opts; 1966 rule->rt = krule->rt; 1967 rule->return_ttl = krule->return_ttl; 1968 rule->tos = krule->tos; 1969 rule->set_tos = krule->set_tos; 1970 rule->anchor_relative = krule->anchor_relative; 1971 rule->anchor_wildcard = krule->anchor_wildcard; 1972 1973 rule->flush = krule->flush; 1974 rule->prio = krule->prio; 1975 rule->set_prio[0] = krule->set_prio[0]; 1976 rule->set_prio[1] = krule->set_prio[1]; 1977 1978 bcopy(&krule->divert, &rule->divert, sizeof(krule->divert)); 1979 1980 rule->u_states_cur = counter_u64_fetch(krule->states_cur); 1981 rule->u_states_tot = counter_u64_fetch(krule->states_tot); 1982 rule->u_src_nodes = counter_u64_fetch(krule->src_nodes); 1983 } 1984 1985 static int 1986 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule) 1987 { 1988 int ret; 1989 1990 #ifndef INET 1991 if (rule->af == AF_INET) { 1992 return (EAFNOSUPPORT); 1993 } 1994 #endif /* INET */ 1995 #ifndef INET6 1996 if (rule->af == AF_INET6) { 1997 return (EAFNOSUPPORT); 1998 } 1999 #endif /* INET6 */ 2000 2001 ret = pf_check_rule_addr(&rule->src); 2002 if (ret != 0) 2003 return (ret); 2004 ret = pf_check_rule_addr(&rule->dst); 2005 if (ret != 0) 2006 return (ret); 2007 2008 bcopy(&rule->src, &krule->src, sizeof(rule->src)); 2009 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst)); 2010 2011 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label)); 2012 if (ret != 0) 2013 return (ret); 2014 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname)); 2015 if (ret != 0) 2016 return (ret); 2017 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname)); 2018 if (ret != 0) 2019 return (ret); 2020 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname)); 2021 if (ret != 0) 2022 return (ret); 2023 ret = pf_user_strcpy(krule->tagname, rule->tagname, 2024 sizeof(rule->tagname)); 2025 if (ret != 0) 2026 return (ret); 2027 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname, 2028 sizeof(rule->match_tagname)); 2029 if (ret != 0) 2030 return (ret); 2031 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname, 2032 sizeof(rule->overload_tblname)); 2033 if (ret != 0) 2034 return (ret); 2035 2036 pf_pool_to_kpool(&rule->rpool, &krule->rpool); 2037 2038 /* Don't allow userspace to set evaulations, packets or bytes. */ 2039 /* kif, anchor, overload_tbl are not copied over. */ 2040 2041 krule->os_fingerprint = rule->os_fingerprint; 2042 2043 krule->rtableid = rule->rtableid; 2044 bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout)); 2045 krule->max_states = rule->max_states; 2046 krule->max_src_nodes = rule->max_src_nodes; 2047 krule->max_src_states = rule->max_src_states; 2048 krule->max_src_conn = rule->max_src_conn; 2049 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit; 2050 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds; 2051 krule->qid = rule->qid; 2052 krule->pqid = rule->pqid; 2053 krule->nr = rule->nr; 2054 krule->prob = rule->prob; 2055 krule->cuid = rule->cuid; 2056 krule->cpid = rule->cpid; 2057 2058 krule->return_icmp = rule->return_icmp; 2059 krule->return_icmp6 = rule->return_icmp6; 2060 krule->max_mss = rule->max_mss; 2061 krule->tag = rule->tag; 2062 krule->match_tag = rule->match_tag; 2063 krule->scrub_flags = rule->scrub_flags; 2064 2065 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid)); 2066 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid)); 2067 2068 krule->rule_flag = rule->rule_flag; 2069 krule->action = rule->action; 2070 krule->direction = rule->direction; 2071 krule->log = rule->log; 2072 krule->logif = rule->logif; 2073 krule->quick = rule->quick; 2074 krule->ifnot = rule->ifnot; 2075 krule->match_tag_not = rule->match_tag_not; 2076 krule->natpass = rule->natpass; 2077 2078 krule->keep_state = rule->keep_state; 2079 krule->af = rule->af; 2080 krule->proto = rule->proto; 2081 krule->type = rule->type; 2082 krule->code = rule->code; 2083 krule->flags = rule->flags; 2084 krule->flagset = rule->flagset; 2085 krule->min_ttl = rule->min_ttl; 2086 krule->allow_opts = rule->allow_opts; 2087 krule->rt = rule->rt; 2088 krule->return_ttl = rule->return_ttl; 2089 krule->tos = rule->tos; 2090 krule->set_tos = rule->set_tos; 2091 2092 krule->flush = rule->flush; 2093 krule->prio = rule->prio; 2094 krule->set_prio[0] = rule->set_prio[0]; 2095 krule->set_prio[1] = rule->set_prio[1]; 2096 2097 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert)); 2098 2099 return (0); 2100 } 2101 2102 static int 2103 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk, 2104 struct pf_kstate_kill *kill) 2105 { 2106 int ret; 2107 2108 bzero(kill, sizeof(*kill)); 2109 2110 bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp)); 2111 kill->psk_af = psk->psk_af; 2112 kill->psk_proto = psk->psk_proto; 2113 bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src)); 2114 bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst)); 2115 ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname, 2116 sizeof(kill->psk_ifname)); 2117 if (ret != 0) 2118 return (ret); 2119 ret = pf_user_strcpy(kill->psk_label, psk->psk_label, 2120 sizeof(kill->psk_label)); 2121 if (ret != 0) 2122 return (ret); 2123 2124 return (0); 2125 } 2126 2127 static int 2128 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, 2129 uint32_t pool_ticket, const char *anchor, const char *anchor_call, 2130 struct thread *td) 2131 { 2132 struct pf_kruleset *ruleset; 2133 struct pf_krule *tail; 2134 struct pf_kpooladdr *pa; 2135 struct pfi_kkif *kif = NULL; 2136 int rs_num; 2137 int error = 0; 2138 2139 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) { 2140 error = EINVAL; 2141 goto errout_unlocked; 2142 } 2143 2144 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 2145 2146 if (rule->ifname[0]) 2147 kif = pf_kkif_create(M_WAITOK); 2148 pf_counter_u64_init(&rule->evaluations, M_WAITOK); 2149 for (int i = 0; i < 2; i++) { 2150 pf_counter_u64_init(&rule->packets[i], M_WAITOK); 2151 pf_counter_u64_init(&rule->bytes[i], M_WAITOK); 2152 } 2153 rule->states_cur = counter_u64_alloc(M_WAITOK); 2154 rule->states_tot = counter_u64_alloc(M_WAITOK); 2155 rule->src_nodes = counter_u64_alloc(M_WAITOK); 2156 rule->cuid = td->td_ucred->cr_ruid; 2157 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 2158 TAILQ_INIT(&rule->rpool.list); 2159 2160 PF_CONFIG_LOCK(); 2161 PF_RULES_WLOCK(); 2162 #ifdef PF_WANT_32_TO_64_COUNTER 2163 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist); 2164 MPASS(!rule->allrulelinked); 2165 rule->allrulelinked = true; 2166 V_pf_allrulecount++; 2167 #endif 2168 ruleset = pf_find_kruleset(anchor); 2169 if (ruleset == NULL) 2170 ERROUT(EINVAL); 2171 rs_num = pf_get_ruleset_number(rule->action); 2172 if (rs_num >= PF_RULESET_MAX) 2173 ERROUT(EINVAL); 2174 if (ticket != ruleset->rules[rs_num].inactive.ticket) { 2175 DPFPRINTF(PF_DEBUG_MISC, 2176 ("ticket: %d != [%d]%d\n", ticket, rs_num, 2177 ruleset->rules[rs_num].inactive.ticket)); 2178 ERROUT(EBUSY); 2179 } 2180 if (pool_ticket != V_ticket_pabuf) { 2181 DPFPRINTF(PF_DEBUG_MISC, 2182 ("pool_ticket: %d != %d\n", pool_ticket, 2183 V_ticket_pabuf)); 2184 ERROUT(EBUSY); 2185 } 2186 /* 2187 * XXXMJG hack: there is no mechanism to ensure they started the 2188 * transaction. Ticket checked above may happen to match by accident, 2189 * even if nobody called DIOCXBEGIN, let alone this process. 2190 * Partially work around it by checking if the RB tree got allocated, 2191 * see pf_begin_rules. 2192 */ 2193 if (ruleset->rules[rs_num].inactive.tree == NULL) { 2194 ERROUT(EINVAL); 2195 } 2196 2197 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 2198 pf_krulequeue); 2199 if (tail) 2200 rule->nr = tail->nr + 1; 2201 else 2202 rule->nr = 0; 2203 if (rule->ifname[0]) { 2204 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2205 kif = NULL; 2206 pfi_kkif_ref(rule->kif); 2207 } else 2208 rule->kif = NULL; 2209 2210 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) 2211 error = EBUSY; 2212 2213 #ifdef ALTQ 2214 /* set queue IDs */ 2215 if (rule->qname[0] != 0) { 2216 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2217 error = EBUSY; 2218 else if (rule->pqname[0] != 0) { 2219 if ((rule->pqid = 2220 pf_qname2qid(rule->pqname)) == 0) 2221 error = EBUSY; 2222 } else 2223 rule->pqid = rule->qid; 2224 } 2225 #endif 2226 if (rule->tagname[0]) 2227 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2228 error = EBUSY; 2229 if (rule->match_tagname[0]) 2230 if ((rule->match_tag = 2231 pf_tagname2tag(rule->match_tagname)) == 0) 2232 error = EBUSY; 2233 if (rule->rt && !rule->direction) 2234 error = EINVAL; 2235 if (!rule->log) 2236 rule->logif = 0; 2237 if (rule->logif >= PFLOGIFS_MAX) 2238 error = EINVAL; 2239 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) 2240 error = ENOMEM; 2241 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) 2242 error = ENOMEM; 2243 if (pf_kanchor_setup(rule, ruleset, anchor_call)) 2244 error = EINVAL; 2245 if (rule->scrub_flags & PFSTATE_SETPRIO && 2246 (rule->set_prio[0] > PF_PRIO_MAX || 2247 rule->set_prio[1] > PF_PRIO_MAX)) 2248 error = EINVAL; 2249 TAILQ_FOREACH(pa, &V_pf_pabuf, entries) 2250 if (pa->addr.type == PF_ADDR_TABLE) { 2251 pa->addr.p.tbl = pfr_attach_table(ruleset, 2252 pa->addr.v.tblname); 2253 if (pa->addr.p.tbl == NULL) 2254 error = ENOMEM; 2255 } 2256 2257 rule->overload_tbl = NULL; 2258 if (rule->overload_tblname[0]) { 2259 if ((rule->overload_tbl = pfr_attach_table(ruleset, 2260 rule->overload_tblname)) == NULL) 2261 error = EINVAL; 2262 else 2263 rule->overload_tbl->pfrkt_flags |= 2264 PFR_TFLAG_ACTIVE; 2265 } 2266 2267 pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list); 2268 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) || 2269 (rule->action == PF_BINAT)) && rule->anchor == NULL) || 2270 (rule->rt > PF_NOPFROUTE)) && 2271 (TAILQ_FIRST(&rule->rpool.list) == NULL)) 2272 error = EINVAL; 2273 2274 if (error) { 2275 pf_free_rule(rule); 2276 rule = NULL; 2277 ERROUT(error); 2278 } 2279 2280 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list); 2281 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, 2282 rule, entries); 2283 ruleset->rules[rs_num].inactive.rcount++; 2284 2285 PF_RULES_WUNLOCK(); 2286 pf_hash_rule(rule); 2287 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) { 2288 PF_RULES_WLOCK(); 2289 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries); 2290 ruleset->rules[rs_num].inactive.rcount--; 2291 pf_free_rule(rule); 2292 rule = NULL; 2293 ERROUT(EEXIST); 2294 } 2295 PF_CONFIG_UNLOCK(); 2296 2297 return (0); 2298 2299 #undef ERROUT 2300 errout: 2301 PF_RULES_WUNLOCK(); 2302 PF_CONFIG_UNLOCK(); 2303 errout_unlocked: 2304 pf_kkif_free(kif); 2305 pf_krule_free(rule); 2306 return (error); 2307 } 2308 2309 static bool 2310 pf_label_match(const struct pf_krule *rule, const char *label) 2311 { 2312 int i = 0; 2313 2314 while (*rule->label[i]) { 2315 if (strcmp(rule->label[i], label) == 0) 2316 return (true); 2317 i++; 2318 } 2319 2320 return (false); 2321 } 2322 2323 static unsigned int 2324 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir) 2325 { 2326 struct pf_kstate *s; 2327 int more = 0; 2328 2329 s = pf_find_state_all(key, dir, &more); 2330 if (s == NULL) 2331 return (0); 2332 2333 if (more) { 2334 PF_STATE_UNLOCK(s); 2335 return (0); 2336 } 2337 2338 pf_unlink_state(s); 2339 return (1); 2340 } 2341 2342 static int 2343 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih) 2344 { 2345 struct pf_kstate *s; 2346 struct pf_state_key *sk; 2347 struct pf_addr *srcaddr, *dstaddr; 2348 struct pf_state_key_cmp match_key; 2349 int idx, killed = 0; 2350 unsigned int dir; 2351 u_int16_t srcport, dstport; 2352 struct pfi_kkif *kif; 2353 2354 relock_DIOCKILLSTATES: 2355 PF_HASHROW_LOCK(ih); 2356 LIST_FOREACH(s, &ih->states, entry) { 2357 /* For floating states look at the original kif. */ 2358 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 2359 2360 sk = s->key[PF_SK_WIRE]; 2361 if (s->direction == PF_OUT) { 2362 srcaddr = &sk->addr[1]; 2363 dstaddr = &sk->addr[0]; 2364 srcport = sk->port[1]; 2365 dstport = sk->port[0]; 2366 } else { 2367 srcaddr = &sk->addr[0]; 2368 dstaddr = &sk->addr[1]; 2369 srcport = sk->port[0]; 2370 dstport = sk->port[1]; 2371 } 2372 2373 if (psk->psk_af && sk->af != psk->psk_af) 2374 continue; 2375 2376 if (psk->psk_proto && psk->psk_proto != sk->proto) 2377 continue; 2378 2379 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr, 2380 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af)) 2381 continue; 2382 2383 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr, 2384 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af)) 2385 continue; 2386 2387 if (! PF_MATCHA(psk->psk_rt_addr.neg, 2388 &psk->psk_rt_addr.addr.v.a.addr, 2389 &psk->psk_rt_addr.addr.v.a.mask, 2390 &s->rt_addr, sk->af)) 2391 continue; 2392 2393 if (psk->psk_src.port_op != 0 && 2394 ! pf_match_port(psk->psk_src.port_op, 2395 psk->psk_src.port[0], psk->psk_src.port[1], srcport)) 2396 continue; 2397 2398 if (psk->psk_dst.port_op != 0 && 2399 ! pf_match_port(psk->psk_dst.port_op, 2400 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) 2401 continue; 2402 2403 if (psk->psk_label[0] && 2404 ! pf_label_match(s->rule.ptr, psk->psk_label)) 2405 continue; 2406 2407 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname, 2408 kif->pfik_name)) 2409 continue; 2410 2411 if (psk->psk_kill_match) { 2412 /* Create the key to find matching states, with lock 2413 * held. */ 2414 2415 bzero(&match_key, sizeof(match_key)); 2416 2417 if (s->direction == PF_OUT) { 2418 dir = PF_IN; 2419 idx = PF_SK_STACK; 2420 } else { 2421 dir = PF_OUT; 2422 idx = PF_SK_WIRE; 2423 } 2424 2425 match_key.af = s->key[idx]->af; 2426 match_key.proto = s->key[idx]->proto; 2427 PF_ACPY(&match_key.addr[0], 2428 &s->key[idx]->addr[1], match_key.af); 2429 match_key.port[0] = s->key[idx]->port[1]; 2430 PF_ACPY(&match_key.addr[1], 2431 &s->key[idx]->addr[0], match_key.af); 2432 match_key.port[1] = s->key[idx]->port[0]; 2433 } 2434 2435 pf_unlink_state(s); 2436 killed++; 2437 2438 if (psk->psk_kill_match) 2439 killed += pf_kill_matching_state(&match_key, dir); 2440 2441 goto relock_DIOCKILLSTATES; 2442 } 2443 PF_HASHROW_UNLOCK(ih); 2444 2445 return (killed); 2446 } 2447 2448 static int 2449 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) 2450 { 2451 int error = 0; 2452 PF_RULES_RLOCK_TRACKER; 2453 2454 #define ERROUT_IOCTL(target, x) \ 2455 do { \ 2456 error = (x); \ 2457 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \ 2458 goto target; \ 2459 } while (0) 2460 2461 2462 /* XXX keep in sync with switch() below */ 2463 if (securelevel_gt(td->td_ucred, 2)) 2464 switch (cmd) { 2465 case DIOCGETRULES: 2466 case DIOCGETRULE: 2467 case DIOCGETRULENV: 2468 case DIOCGETADDRS: 2469 case DIOCGETADDR: 2470 case DIOCGETSTATE: 2471 case DIOCGETSTATENV: 2472 case DIOCSETSTATUSIF: 2473 case DIOCGETSTATUS: 2474 case DIOCGETSTATUSNV: 2475 case DIOCCLRSTATUS: 2476 case DIOCNATLOOK: 2477 case DIOCSETDEBUG: 2478 case DIOCGETSTATES: 2479 case DIOCGETSTATESV2: 2480 case DIOCGETTIMEOUT: 2481 case DIOCCLRRULECTRS: 2482 case DIOCGETLIMIT: 2483 case DIOCGETALTQSV0: 2484 case DIOCGETALTQSV1: 2485 case DIOCGETALTQV0: 2486 case DIOCGETALTQV1: 2487 case DIOCGETQSTATSV0: 2488 case DIOCGETQSTATSV1: 2489 case DIOCGETRULESETS: 2490 case DIOCGETRULESET: 2491 case DIOCRGETTABLES: 2492 case DIOCRGETTSTATS: 2493 case DIOCRCLRTSTATS: 2494 case DIOCRCLRADDRS: 2495 case DIOCRADDADDRS: 2496 case DIOCRDELADDRS: 2497 case DIOCRSETADDRS: 2498 case DIOCRGETADDRS: 2499 case DIOCRGETASTATS: 2500 case DIOCRCLRASTATS: 2501 case DIOCRTSTADDRS: 2502 case DIOCOSFPGET: 2503 case DIOCGETSRCNODES: 2504 case DIOCCLRSRCNODES: 2505 case DIOCGETSYNCOOKIES: 2506 case DIOCIGETIFACES: 2507 case DIOCGIFSPEEDV0: 2508 case DIOCGIFSPEEDV1: 2509 case DIOCSETIFFLAG: 2510 case DIOCCLRIFFLAG: 2511 case DIOCGETETHRULES: 2512 case DIOCGETETHRULE: 2513 case DIOCGETETHRULESETS: 2514 case DIOCGETETHRULESET: 2515 break; 2516 case DIOCRCLRTABLES: 2517 case DIOCRADDTABLES: 2518 case DIOCRDELTABLES: 2519 case DIOCRSETTFLAGS: 2520 if (((struct pfioc_table *)addr)->pfrio_flags & 2521 PFR_FLAG_DUMMY) 2522 break; /* dummy operation ok */ 2523 return (EPERM); 2524 default: 2525 return (EPERM); 2526 } 2527 2528 if (!(flags & FWRITE)) 2529 switch (cmd) { 2530 case DIOCGETRULES: 2531 case DIOCGETADDRS: 2532 case DIOCGETADDR: 2533 case DIOCGETSTATE: 2534 case DIOCGETSTATENV: 2535 case DIOCGETSTATUS: 2536 case DIOCGETSTATUSNV: 2537 case DIOCGETSTATES: 2538 case DIOCGETSTATESV2: 2539 case DIOCGETTIMEOUT: 2540 case DIOCGETLIMIT: 2541 case DIOCGETALTQSV0: 2542 case DIOCGETALTQSV1: 2543 case DIOCGETALTQV0: 2544 case DIOCGETALTQV1: 2545 case DIOCGETQSTATSV0: 2546 case DIOCGETQSTATSV1: 2547 case DIOCGETRULESETS: 2548 case DIOCGETRULESET: 2549 case DIOCNATLOOK: 2550 case DIOCRGETTABLES: 2551 case DIOCRGETTSTATS: 2552 case DIOCRGETADDRS: 2553 case DIOCRGETASTATS: 2554 case DIOCRTSTADDRS: 2555 case DIOCOSFPGET: 2556 case DIOCGETSRCNODES: 2557 case DIOCGETSYNCOOKIES: 2558 case DIOCIGETIFACES: 2559 case DIOCGIFSPEEDV1: 2560 case DIOCGIFSPEEDV0: 2561 case DIOCGETRULENV: 2562 case DIOCGETETHRULES: 2563 case DIOCGETETHRULE: 2564 case DIOCGETETHRULESETS: 2565 case DIOCGETETHRULESET: 2566 break; 2567 case DIOCRCLRTABLES: 2568 case DIOCRADDTABLES: 2569 case DIOCRDELTABLES: 2570 case DIOCRCLRTSTATS: 2571 case DIOCRCLRADDRS: 2572 case DIOCRADDADDRS: 2573 case DIOCRDELADDRS: 2574 case DIOCRSETADDRS: 2575 case DIOCRSETTFLAGS: 2576 if (((struct pfioc_table *)addr)->pfrio_flags & 2577 PFR_FLAG_DUMMY) { 2578 flags |= FWRITE; /* need write lock for dummy */ 2579 break; /* dummy operation ok */ 2580 } 2581 return (EACCES); 2582 case DIOCGETRULE: 2583 if (((struct pfioc_rule *)addr)->action == 2584 PF_GET_CLR_CNTR) 2585 return (EACCES); 2586 break; 2587 default: 2588 return (EACCES); 2589 } 2590 2591 CURVNET_SET(TD_TO_VNET(td)); 2592 2593 switch (cmd) { 2594 case DIOCSTART: 2595 sx_xlock(&pf_ioctl_lock); 2596 if (V_pf_status.running) 2597 error = EEXIST; 2598 else { 2599 int cpu; 2600 2601 hook_pf(); 2602 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 2603 hook_pf_eth(); 2604 V_pf_status.running = 1; 2605 V_pf_status.since = time_second; 2606 2607 CPU_FOREACH(cpu) 2608 V_pf_stateid[cpu] = time_second; 2609 2610 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n")); 2611 } 2612 break; 2613 2614 case DIOCSTOP: 2615 sx_xlock(&pf_ioctl_lock); 2616 if (!V_pf_status.running) 2617 error = ENOENT; 2618 else { 2619 V_pf_status.running = 0; 2620 dehook_pf(); 2621 dehook_pf_eth(); 2622 V_pf_status.since = time_second; 2623 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n")); 2624 } 2625 break; 2626 2627 case DIOCGETETHRULES: { 2628 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2629 nvlist_t *nvl; 2630 void *packed; 2631 struct pf_keth_rule *tail; 2632 struct pf_keth_ruleset *rs; 2633 u_int32_t ticket, nr; 2634 const char *anchor = ""; 2635 2636 nvl = NULL; 2637 packed = NULL; 2638 2639 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x) 2640 2641 if (nv->len > pf_ioctl_maxcount) 2642 ERROUT(ENOMEM); 2643 2644 /* Copy the request in */ 2645 packed = malloc(nv->len, M_NVLIST, M_WAITOK); 2646 if (packed == NULL) 2647 ERROUT(ENOMEM); 2648 2649 error = copyin(nv->data, packed, nv->len); 2650 if (error) 2651 ERROUT(error); 2652 2653 nvl = nvlist_unpack(packed, nv->len, 0); 2654 if (nvl == NULL) 2655 ERROUT(EBADMSG); 2656 2657 if (! nvlist_exists_string(nvl, "anchor")) 2658 ERROUT(EBADMSG); 2659 2660 anchor = nvlist_get_string(nvl, "anchor"); 2661 2662 rs = pf_find_keth_ruleset(anchor); 2663 2664 nvlist_destroy(nvl); 2665 nvl = NULL; 2666 free(packed, M_NVLIST); 2667 packed = NULL; 2668 2669 if (rs == NULL) 2670 ERROUT(ENOENT); 2671 2672 /* Reply */ 2673 nvl = nvlist_create(0); 2674 if (nvl == NULL) 2675 ERROUT(ENOMEM); 2676 2677 PF_RULES_RLOCK(); 2678 2679 ticket = rs->active.ticket; 2680 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq); 2681 if (tail) 2682 nr = tail->nr + 1; 2683 else 2684 nr = 0; 2685 2686 PF_RULES_RUNLOCK(); 2687 2688 nvlist_add_number(nvl, "ticket", ticket); 2689 nvlist_add_number(nvl, "nr", nr); 2690 2691 packed = nvlist_pack(nvl, &nv->len); 2692 if (packed == NULL) 2693 ERROUT(ENOMEM); 2694 2695 if (nv->size == 0) 2696 ERROUT(0); 2697 else if (nv->size < nv->len) 2698 ERROUT(ENOSPC); 2699 2700 error = copyout(packed, nv->data, nv->len); 2701 2702 #undef ERROUT 2703 DIOCGETETHRULES_error: 2704 free(packed, M_NVLIST); 2705 nvlist_destroy(nvl); 2706 break; 2707 } 2708 2709 case DIOCGETETHRULE: { 2710 struct epoch_tracker et; 2711 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2712 nvlist_t *nvl = NULL; 2713 void *nvlpacked = NULL; 2714 struct pf_keth_rule *rule = NULL; 2715 struct pf_keth_ruleset *rs; 2716 u_int32_t ticket, nr; 2717 bool clear = false; 2718 const char *anchor; 2719 2720 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x) 2721 2722 if (nv->len > pf_ioctl_maxcount) 2723 ERROUT(ENOMEM); 2724 2725 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 2726 if (nvlpacked == NULL) 2727 ERROUT(ENOMEM); 2728 2729 error = copyin(nv->data, nvlpacked, nv->len); 2730 if (error) 2731 ERROUT(error); 2732 2733 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 2734 if (nvl == NULL) 2735 ERROUT(EBADMSG); 2736 if (! nvlist_exists_number(nvl, "ticket")) 2737 ERROUT(EBADMSG); 2738 ticket = nvlist_get_number(nvl, "ticket"); 2739 if (! nvlist_exists_string(nvl, "anchor")) 2740 ERROUT(EBADMSG); 2741 anchor = nvlist_get_string(nvl, "anchor"); 2742 2743 if (nvlist_exists_bool(nvl, "clear")) 2744 clear = nvlist_get_bool(nvl, "clear"); 2745 2746 if (clear && !(flags & FWRITE)) 2747 ERROUT(EACCES); 2748 2749 if (! nvlist_exists_number(nvl, "nr")) 2750 ERROUT(EBADMSG); 2751 nr = nvlist_get_number(nvl, "nr"); 2752 2753 PF_RULES_RLOCK(); 2754 rs = pf_find_keth_ruleset(anchor); 2755 if (rs == NULL) { 2756 PF_RULES_RUNLOCK(); 2757 ERROUT(ENOENT); 2758 } 2759 if (ticket != rs->active.ticket) { 2760 PF_RULES_RUNLOCK(); 2761 ERROUT(EBUSY); 2762 } 2763 2764 nvlist_destroy(nvl); 2765 nvl = NULL; 2766 free(nvlpacked, M_NVLIST); 2767 nvlpacked = NULL; 2768 2769 rule = TAILQ_FIRST(rs->active.rules); 2770 while ((rule != NULL) && (rule->nr != nr)) 2771 rule = TAILQ_NEXT(rule, entries); 2772 if (rule == NULL) { 2773 PF_RULES_RUNLOCK(); 2774 ERROUT(ENOENT); 2775 } 2776 /* Make sure rule can't go away. */ 2777 NET_EPOCH_ENTER(et); 2778 PF_RULES_RUNLOCK(); 2779 nvl = pf_keth_rule_to_nveth_rule(rule); 2780 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) 2781 ERROUT(EBUSY); 2782 NET_EPOCH_EXIT(et); 2783 if (nvl == NULL) 2784 ERROUT(ENOMEM); 2785 2786 nvlpacked = nvlist_pack(nvl, &nv->len); 2787 if (nvlpacked == NULL) 2788 ERROUT(ENOMEM); 2789 2790 if (nv->size == 0) 2791 ERROUT(0); 2792 else if (nv->size < nv->len) 2793 ERROUT(ENOSPC); 2794 2795 error = copyout(nvlpacked, nv->data, nv->len); 2796 if (error == 0 && clear) { 2797 counter_u64_zero(rule->evaluations); 2798 for (int i = 0; i < 2; i++) { 2799 counter_u64_zero(rule->packets[i]); 2800 counter_u64_zero(rule->bytes[i]); 2801 } 2802 } 2803 2804 #undef ERROUT 2805 DIOCGETETHRULE_error: 2806 free(nvlpacked, M_NVLIST); 2807 nvlist_destroy(nvl); 2808 break; 2809 } 2810 2811 case DIOCADDETHRULE: { 2812 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2813 nvlist_t *nvl = NULL; 2814 void *nvlpacked = NULL; 2815 struct pf_keth_rule *rule = NULL, *tail = NULL; 2816 struct pf_keth_ruleset *ruleset = NULL; 2817 struct pfi_kkif *kif = NULL; 2818 const char *anchor = "", *anchor_call = ""; 2819 2820 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x) 2821 2822 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 2823 if (nvlpacked == NULL) 2824 ERROUT(ENOMEM); 2825 2826 error = copyin(nv->data, nvlpacked, nv->len); 2827 if (error) 2828 ERROUT(error); 2829 2830 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 2831 if (nvl == NULL) 2832 ERROUT(EBADMSG); 2833 2834 if (! nvlist_exists_number(nvl, "ticket")) 2835 ERROUT(EBADMSG); 2836 2837 if (nvlist_exists_string(nvl, "anchor")) 2838 anchor = nvlist_get_string(nvl, "anchor"); 2839 if (nvlist_exists_string(nvl, "anchor_call")) 2840 anchor_call = nvlist_get_string(nvl, "anchor_call"); 2841 2842 ruleset = pf_find_keth_ruleset(anchor); 2843 if (ruleset == NULL) 2844 ERROUT(EINVAL); 2845 2846 if (nvlist_get_number(nvl, "ticket") != 2847 ruleset->inactive.ticket) { 2848 DPFPRINTF(PF_DEBUG_MISC, 2849 ("ticket: %d != %d\n", 2850 (u_int32_t)nvlist_get_number(nvl, "ticket"), 2851 ruleset->inactive.ticket)); 2852 ERROUT(EBUSY); 2853 } 2854 2855 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); 2856 if (rule == NULL) 2857 ERROUT(ENOMEM); 2858 rule->timestamp = NULL; 2859 2860 error = pf_nveth_rule_to_keth_rule(nvl, rule); 2861 if (error != 0) 2862 ERROUT(error); 2863 2864 if (rule->ifname[0]) 2865 kif = pf_kkif_create(M_WAITOK); 2866 rule->evaluations = counter_u64_alloc(M_WAITOK); 2867 for (int i = 0; i < 2; i++) { 2868 rule->packets[i] = counter_u64_alloc(M_WAITOK); 2869 rule->bytes[i] = counter_u64_alloc(M_WAITOK); 2870 } 2871 rule->timestamp = uma_zalloc_pcpu(pcpu_zone_4, 2872 M_WAITOK | M_ZERO); 2873 2874 PF_RULES_WLOCK(); 2875 2876 if (rule->ifname[0]) { 2877 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2878 pfi_kkif_ref(rule->kif); 2879 } else 2880 rule->kif = NULL; 2881 2882 #ifdef ALTQ 2883 /* set queue IDs */ 2884 if (rule->qname[0] != 0) { 2885 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2886 error = EBUSY; 2887 else 2888 rule->qid = rule->qid; 2889 } 2890 #endif 2891 if (rule->tagname[0]) 2892 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2893 error = EBUSY; 2894 2895 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE) 2896 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr); 2897 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE) 2898 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr); 2899 2900 if (error) { 2901 pf_free_eth_rule(rule); 2902 PF_RULES_WUNLOCK(); 2903 ERROUT(error); 2904 } 2905 2906 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) { 2907 pf_free_eth_rule(rule); 2908 PF_RULES_WUNLOCK(); 2909 ERROUT(EINVAL); 2910 } 2911 2912 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq); 2913 if (tail) 2914 rule->nr = tail->nr + 1; 2915 else 2916 rule->nr = 0; 2917 2918 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries); 2919 2920 PF_RULES_WUNLOCK(); 2921 2922 #undef ERROUT 2923 DIOCADDETHRULE_error: 2924 nvlist_destroy(nvl); 2925 free(nvlpacked, M_NVLIST); 2926 break; 2927 } 2928 2929 case DIOCGETETHRULESETS: { 2930 struct epoch_tracker et; 2931 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 2932 nvlist_t *nvl = NULL; 2933 void *nvlpacked = NULL; 2934 struct pf_keth_ruleset *ruleset; 2935 struct pf_keth_anchor *anchor; 2936 int nr = 0; 2937 2938 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x) 2939 2940 if (nv->len > pf_ioctl_maxcount) 2941 ERROUT(ENOMEM); 2942 2943 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 2944 if (nvlpacked == NULL) 2945 ERROUT(ENOMEM); 2946 2947 error = copyin(nv->data, nvlpacked, nv->len); 2948 if (error) 2949 ERROUT(error); 2950 2951 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 2952 if (nvl == NULL) 2953 ERROUT(EBADMSG); 2954 if (! nvlist_exists_string(nvl, "path")) 2955 ERROUT(EBADMSG); 2956 2957 NET_EPOCH_ENTER(et); 2958 2959 if ((ruleset = pf_find_keth_ruleset( 2960 nvlist_get_string(nvl, "path"))) == NULL) { 2961 NET_EPOCH_EXIT(et); 2962 ERROUT(ENOENT); 2963 } 2964 2965 if (ruleset->anchor == NULL) { 2966 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors) 2967 if (anchor->parent == NULL) 2968 nr++; 2969 } else { 2970 RB_FOREACH(anchor, pf_keth_anchor_node, 2971 &ruleset->anchor->children) 2972 nr++; 2973 } 2974 2975 NET_EPOCH_EXIT(et); 2976 2977 nvlist_destroy(nvl); 2978 nvl = NULL; 2979 free(nvlpacked, M_NVLIST); 2980 nvlpacked = NULL; 2981 2982 nvl = nvlist_create(0); 2983 if (nvl == NULL) 2984 ERROUT(ENOMEM); 2985 2986 nvlist_add_number(nvl, "nr", nr); 2987 2988 nvlpacked = nvlist_pack(nvl, &nv->len); 2989 if (nvlpacked == NULL) 2990 ERROUT(ENOMEM); 2991 2992 if (nv->size == 0) 2993 ERROUT(0); 2994 else if (nv->size < nv->len) 2995 ERROUT(ENOSPC); 2996 2997 error = copyout(nvlpacked, nv->data, nv->len); 2998 2999 #undef ERROUT 3000 DIOCGETETHRULESETS_error: 3001 free(nvlpacked, M_NVLIST); 3002 nvlist_destroy(nvl); 3003 break; 3004 } 3005 3006 case DIOCGETETHRULESET: { 3007 struct epoch_tracker et; 3008 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3009 nvlist_t *nvl = NULL; 3010 void *nvlpacked = NULL; 3011 struct pf_keth_ruleset *ruleset; 3012 struct pf_keth_anchor *anchor; 3013 int nr = 0, req_nr = 0; 3014 bool found = false; 3015 3016 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x) 3017 3018 if (nv->len > pf_ioctl_maxcount) 3019 ERROUT(ENOMEM); 3020 3021 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3022 if (nvlpacked == NULL) 3023 ERROUT(ENOMEM); 3024 3025 error = copyin(nv->data, nvlpacked, nv->len); 3026 if (error) 3027 ERROUT(error); 3028 3029 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3030 if (nvl == NULL) 3031 ERROUT(EBADMSG); 3032 if (! nvlist_exists_string(nvl, "path")) 3033 ERROUT(EBADMSG); 3034 if (! nvlist_exists_number(nvl, "nr")) 3035 ERROUT(EBADMSG); 3036 3037 req_nr = nvlist_get_number(nvl, "nr"); 3038 3039 NET_EPOCH_ENTER(et); 3040 3041 if ((ruleset = pf_find_keth_ruleset( 3042 nvlist_get_string(nvl, "path"))) == NULL) { 3043 NET_EPOCH_EXIT(et); 3044 ERROUT(ENOENT); 3045 } 3046 3047 nvlist_destroy(nvl); 3048 nvl = NULL; 3049 free(nvlpacked, M_NVLIST); 3050 nvlpacked = NULL; 3051 3052 nvl = nvlist_create(0); 3053 if (nvl == NULL) { 3054 NET_EPOCH_EXIT(et); 3055 ERROUT(ENOMEM); 3056 } 3057 3058 if (ruleset->anchor == NULL) { 3059 RB_FOREACH(anchor, pf_keth_anchor_global, 3060 &V_pf_keth_anchors) { 3061 if (anchor->parent == NULL && nr++ == req_nr) { 3062 found = true; 3063 break; 3064 } 3065 } 3066 } else { 3067 RB_FOREACH(anchor, pf_keth_anchor_node, 3068 &ruleset->anchor->children) { 3069 if (nr++ == req_nr) { 3070 found = true; 3071 break; 3072 } 3073 } 3074 } 3075 3076 NET_EPOCH_EXIT(et); 3077 if (found) { 3078 nvlist_add_number(nvl, "nr", nr); 3079 nvlist_add_string(nvl, "name", anchor->name); 3080 if (ruleset->anchor) 3081 nvlist_add_string(nvl, "path", 3082 ruleset->anchor->path); 3083 else 3084 nvlist_add_string(nvl, "path", ""); 3085 } else { 3086 ERROUT(EBUSY); 3087 } 3088 3089 nvlpacked = nvlist_pack(nvl, &nv->len); 3090 if (nvlpacked == NULL) 3091 ERROUT(ENOMEM); 3092 3093 if (nv->size == 0) 3094 ERROUT(0); 3095 else if (nv->size < nv->len) 3096 ERROUT(ENOSPC); 3097 3098 error = copyout(nvlpacked, nv->data, nv->len); 3099 3100 #undef ERROUT 3101 DIOCGETETHRULESET_error: 3102 free(nvlpacked, M_NVLIST); 3103 nvlist_destroy(nvl); 3104 break; 3105 } 3106 3107 case DIOCADDRULENV: { 3108 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3109 nvlist_t *nvl = NULL; 3110 void *nvlpacked = NULL; 3111 struct pf_krule *rule = NULL; 3112 const char *anchor = "", *anchor_call = ""; 3113 uint32_t ticket = 0, pool_ticket = 0; 3114 3115 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x) 3116 3117 if (nv->len > pf_ioctl_maxcount) 3118 ERROUT(ENOMEM); 3119 3120 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3121 error = copyin(nv->data, nvlpacked, nv->len); 3122 if (error) 3123 ERROUT(error); 3124 3125 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3126 if (nvl == NULL) 3127 ERROUT(EBADMSG); 3128 3129 if (! nvlist_exists_number(nvl, "ticket")) 3130 ERROUT(EINVAL); 3131 ticket = nvlist_get_number(nvl, "ticket"); 3132 3133 if (! nvlist_exists_number(nvl, "pool_ticket")) 3134 ERROUT(EINVAL); 3135 pool_ticket = nvlist_get_number(nvl, "pool_ticket"); 3136 3137 if (! nvlist_exists_nvlist(nvl, "rule")) 3138 ERROUT(EINVAL); 3139 3140 rule = pf_krule_alloc(); 3141 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"), 3142 rule); 3143 if (error) 3144 ERROUT(error); 3145 3146 if (nvlist_exists_string(nvl, "anchor")) 3147 anchor = nvlist_get_string(nvl, "anchor"); 3148 if (nvlist_exists_string(nvl, "anchor_call")) 3149 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3150 3151 if ((error = nvlist_error(nvl))) 3152 ERROUT(error); 3153 3154 /* Frees rule on error */ 3155 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor, 3156 anchor_call, td); 3157 3158 nvlist_destroy(nvl); 3159 free(nvlpacked, M_NVLIST); 3160 break; 3161 #undef ERROUT 3162 DIOCADDRULENV_error: 3163 pf_krule_free(rule); 3164 nvlist_destroy(nvl); 3165 free(nvlpacked, M_NVLIST); 3166 3167 break; 3168 } 3169 case DIOCADDRULE: { 3170 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3171 struct pf_krule *rule; 3172 3173 rule = pf_krule_alloc(); 3174 error = pf_rule_to_krule(&pr->rule, rule); 3175 if (error != 0) { 3176 pf_krule_free(rule); 3177 break; 3178 } 3179 3180 pr->anchor[sizeof(pr->anchor) - 1] = 0; 3181 3182 /* Frees rule on error */ 3183 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, 3184 pr->anchor, pr->anchor_call, td); 3185 break; 3186 } 3187 3188 case DIOCGETRULES: { 3189 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3190 struct pf_kruleset *ruleset; 3191 struct pf_krule *tail; 3192 int rs_num; 3193 3194 pr->anchor[sizeof(pr->anchor) - 1] = 0; 3195 3196 PF_RULES_WLOCK(); 3197 ruleset = pf_find_kruleset(pr->anchor); 3198 if (ruleset == NULL) { 3199 PF_RULES_WUNLOCK(); 3200 error = EINVAL; 3201 break; 3202 } 3203 rs_num = pf_get_ruleset_number(pr->rule.action); 3204 if (rs_num >= PF_RULESET_MAX) { 3205 PF_RULES_WUNLOCK(); 3206 error = EINVAL; 3207 break; 3208 } 3209 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 3210 pf_krulequeue); 3211 if (tail) 3212 pr->nr = tail->nr + 1; 3213 else 3214 pr->nr = 0; 3215 pr->ticket = ruleset->rules[rs_num].active.ticket; 3216 PF_RULES_WUNLOCK(); 3217 break; 3218 } 3219 3220 case DIOCGETRULE: { 3221 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3222 struct pf_kruleset *ruleset; 3223 struct pf_krule *rule; 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 if (pr->ticket != ruleset->rules[rs_num].active.ticket) { 3242 PF_RULES_WUNLOCK(); 3243 error = EBUSY; 3244 break; 3245 } 3246 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3247 while ((rule != NULL) && (rule->nr != pr->nr)) 3248 rule = TAILQ_NEXT(rule, entries); 3249 if (rule == NULL) { 3250 PF_RULES_WUNLOCK(); 3251 error = EBUSY; 3252 break; 3253 } 3254 3255 pf_krule_to_rule(rule, &pr->rule); 3256 3257 if (pf_kanchor_copyout(ruleset, rule, pr)) { 3258 PF_RULES_WUNLOCK(); 3259 error = EBUSY; 3260 break; 3261 } 3262 pf_addr_copyout(&pr->rule.src.addr); 3263 pf_addr_copyout(&pr->rule.dst.addr); 3264 3265 if (pr->action == PF_GET_CLR_CNTR) { 3266 pf_counter_u64_zero(&rule->evaluations); 3267 for (int i = 0; i < 2; i++) { 3268 pf_counter_u64_zero(&rule->packets[i]); 3269 pf_counter_u64_zero(&rule->bytes[i]); 3270 } 3271 counter_u64_zero(rule->states_tot); 3272 } 3273 PF_RULES_WUNLOCK(); 3274 break; 3275 } 3276 3277 case DIOCGETRULENV: { 3278 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3279 nvlist_t *nvrule = NULL; 3280 nvlist_t *nvl = NULL; 3281 struct pf_kruleset *ruleset; 3282 struct pf_krule *rule; 3283 void *nvlpacked = NULL; 3284 int rs_num, nr; 3285 bool clear_counter = false; 3286 3287 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x) 3288 3289 if (nv->len > pf_ioctl_maxcount) 3290 ERROUT(ENOMEM); 3291 3292 /* Copy the request in */ 3293 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3294 if (nvlpacked == NULL) 3295 ERROUT(ENOMEM); 3296 3297 error = copyin(nv->data, nvlpacked, nv->len); 3298 if (error) 3299 ERROUT(error); 3300 3301 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3302 if (nvl == NULL) 3303 ERROUT(EBADMSG); 3304 3305 if (! nvlist_exists_string(nvl, "anchor")) 3306 ERROUT(EBADMSG); 3307 if (! nvlist_exists_number(nvl, "ruleset")) 3308 ERROUT(EBADMSG); 3309 if (! nvlist_exists_number(nvl, "ticket")) 3310 ERROUT(EBADMSG); 3311 if (! nvlist_exists_number(nvl, "nr")) 3312 ERROUT(EBADMSG); 3313 3314 if (nvlist_exists_bool(nvl, "clear_counter")) 3315 clear_counter = nvlist_get_bool(nvl, "clear_counter"); 3316 3317 if (clear_counter && !(flags & FWRITE)) 3318 ERROUT(EACCES); 3319 3320 nr = nvlist_get_number(nvl, "nr"); 3321 3322 PF_RULES_WLOCK(); 3323 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor")); 3324 if (ruleset == NULL) { 3325 PF_RULES_WUNLOCK(); 3326 ERROUT(ENOENT); 3327 } 3328 3329 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset")); 3330 if (rs_num >= PF_RULESET_MAX) { 3331 PF_RULES_WUNLOCK(); 3332 ERROUT(EINVAL); 3333 } 3334 3335 if (nvlist_get_number(nvl, "ticket") != 3336 ruleset->rules[rs_num].active.ticket) { 3337 PF_RULES_WUNLOCK(); 3338 ERROUT(EBUSY); 3339 } 3340 3341 if ((error = nvlist_error(nvl))) { 3342 PF_RULES_WUNLOCK(); 3343 ERROUT(error); 3344 } 3345 3346 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3347 while ((rule != NULL) && (rule->nr != nr)) 3348 rule = TAILQ_NEXT(rule, entries); 3349 if (rule == NULL) { 3350 PF_RULES_WUNLOCK(); 3351 ERROUT(EBUSY); 3352 } 3353 3354 nvrule = pf_krule_to_nvrule(rule); 3355 3356 nvlist_destroy(nvl); 3357 nvl = nvlist_create(0); 3358 if (nvl == NULL) { 3359 PF_RULES_WUNLOCK(); 3360 ERROUT(ENOMEM); 3361 } 3362 nvlist_add_number(nvl, "nr", nr); 3363 nvlist_add_nvlist(nvl, "rule", nvrule); 3364 nvlist_destroy(nvrule); 3365 nvrule = NULL; 3366 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) { 3367 PF_RULES_WUNLOCK(); 3368 ERROUT(EBUSY); 3369 } 3370 3371 free(nvlpacked, M_NVLIST); 3372 nvlpacked = nvlist_pack(nvl, &nv->len); 3373 if (nvlpacked == NULL) { 3374 PF_RULES_WUNLOCK(); 3375 ERROUT(ENOMEM); 3376 } 3377 3378 if (nv->size == 0) { 3379 PF_RULES_WUNLOCK(); 3380 ERROUT(0); 3381 } 3382 else if (nv->size < nv->len) { 3383 PF_RULES_WUNLOCK(); 3384 ERROUT(ENOSPC); 3385 } 3386 3387 if (clear_counter) { 3388 pf_counter_u64_zero(&rule->evaluations); 3389 for (int i = 0; i < 2; i++) { 3390 pf_counter_u64_zero(&rule->packets[i]); 3391 pf_counter_u64_zero(&rule->bytes[i]); 3392 } 3393 counter_u64_zero(rule->states_tot); 3394 } 3395 PF_RULES_WUNLOCK(); 3396 3397 error = copyout(nvlpacked, nv->data, nv->len); 3398 3399 #undef ERROUT 3400 DIOCGETRULENV_error: 3401 free(nvlpacked, M_NVLIST); 3402 nvlist_destroy(nvrule); 3403 nvlist_destroy(nvl); 3404 3405 break; 3406 } 3407 3408 case DIOCCHANGERULE: { 3409 struct pfioc_rule *pcr = (struct pfioc_rule *)addr; 3410 struct pf_kruleset *ruleset; 3411 struct pf_krule *oldrule = NULL, *newrule = NULL; 3412 struct pfi_kkif *kif = NULL; 3413 struct pf_kpooladdr *pa; 3414 u_int32_t nr = 0; 3415 int rs_num; 3416 3417 pcr->anchor[sizeof(pcr->anchor) - 1] = 0; 3418 3419 if (pcr->action < PF_CHANGE_ADD_HEAD || 3420 pcr->action > PF_CHANGE_GET_TICKET) { 3421 error = EINVAL; 3422 break; 3423 } 3424 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { 3425 error = EINVAL; 3426 break; 3427 } 3428 3429 if (pcr->action != PF_CHANGE_REMOVE) { 3430 newrule = pf_krule_alloc(); 3431 error = pf_rule_to_krule(&pcr->rule, newrule); 3432 if (error != 0) { 3433 pf_krule_free(newrule); 3434 break; 3435 } 3436 3437 if (newrule->ifname[0]) 3438 kif = pf_kkif_create(M_WAITOK); 3439 pf_counter_u64_init(&newrule->evaluations, M_WAITOK); 3440 for (int i = 0; i < 2; i++) { 3441 pf_counter_u64_init(&newrule->packets[i], M_WAITOK); 3442 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK); 3443 } 3444 newrule->states_cur = counter_u64_alloc(M_WAITOK); 3445 newrule->states_tot = counter_u64_alloc(M_WAITOK); 3446 newrule->src_nodes = counter_u64_alloc(M_WAITOK); 3447 newrule->cuid = td->td_ucred->cr_ruid; 3448 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 3449 TAILQ_INIT(&newrule->rpool.list); 3450 } 3451 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x) 3452 3453 PF_CONFIG_LOCK(); 3454 PF_RULES_WLOCK(); 3455 #ifdef PF_WANT_32_TO_64_COUNTER 3456 if (newrule != NULL) { 3457 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist); 3458 newrule->allrulelinked = true; 3459 V_pf_allrulecount++; 3460 } 3461 #endif 3462 3463 if (!(pcr->action == PF_CHANGE_REMOVE || 3464 pcr->action == PF_CHANGE_GET_TICKET) && 3465 pcr->pool_ticket != V_ticket_pabuf) 3466 ERROUT(EBUSY); 3467 3468 ruleset = pf_find_kruleset(pcr->anchor); 3469 if (ruleset == NULL) 3470 ERROUT(EINVAL); 3471 3472 rs_num = pf_get_ruleset_number(pcr->rule.action); 3473 if (rs_num >= PF_RULESET_MAX) 3474 ERROUT(EINVAL); 3475 3476 /* 3477 * XXXMJG: there is no guarantee that the ruleset was 3478 * created by the usual route of calling DIOCXBEGIN. 3479 * As a result it is possible the rule tree will not 3480 * be allocated yet. Hack around it by doing it here. 3481 * Note it is fine to let the tree persist in case of 3482 * error as it will be freed down the road on future 3483 * updates (if need be). 3484 */ 3485 if (ruleset->rules[rs_num].active.tree == NULL) { 3486 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT); 3487 if (ruleset->rules[rs_num].active.tree == NULL) { 3488 ERROUT(ENOMEM); 3489 } 3490 } 3491 3492 if (pcr->action == PF_CHANGE_GET_TICKET) { 3493 pcr->ticket = ++ruleset->rules[rs_num].active.ticket; 3494 ERROUT(0); 3495 } else if (pcr->ticket != 3496 ruleset->rules[rs_num].active.ticket) 3497 ERROUT(EINVAL); 3498 3499 if (pcr->action != PF_CHANGE_REMOVE) { 3500 if (newrule->ifname[0]) { 3501 newrule->kif = pfi_kkif_attach(kif, 3502 newrule->ifname); 3503 kif = NULL; 3504 pfi_kkif_ref(newrule->kif); 3505 } else 3506 newrule->kif = NULL; 3507 3508 if (newrule->rtableid > 0 && 3509 newrule->rtableid >= rt_numfibs) 3510 error = EBUSY; 3511 3512 #ifdef ALTQ 3513 /* set queue IDs */ 3514 if (newrule->qname[0] != 0) { 3515 if ((newrule->qid = 3516 pf_qname2qid(newrule->qname)) == 0) 3517 error = EBUSY; 3518 else if (newrule->pqname[0] != 0) { 3519 if ((newrule->pqid = 3520 pf_qname2qid(newrule->pqname)) == 0) 3521 error = EBUSY; 3522 } else 3523 newrule->pqid = newrule->qid; 3524 } 3525 #endif /* ALTQ */ 3526 if (newrule->tagname[0]) 3527 if ((newrule->tag = 3528 pf_tagname2tag(newrule->tagname)) == 0) 3529 error = EBUSY; 3530 if (newrule->match_tagname[0]) 3531 if ((newrule->match_tag = pf_tagname2tag( 3532 newrule->match_tagname)) == 0) 3533 error = EBUSY; 3534 if (newrule->rt && !newrule->direction) 3535 error = EINVAL; 3536 if (!newrule->log) 3537 newrule->logif = 0; 3538 if (newrule->logif >= PFLOGIFS_MAX) 3539 error = EINVAL; 3540 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) 3541 error = ENOMEM; 3542 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) 3543 error = ENOMEM; 3544 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call)) 3545 error = EINVAL; 3546 TAILQ_FOREACH(pa, &V_pf_pabuf, entries) 3547 if (pa->addr.type == PF_ADDR_TABLE) { 3548 pa->addr.p.tbl = 3549 pfr_attach_table(ruleset, 3550 pa->addr.v.tblname); 3551 if (pa->addr.p.tbl == NULL) 3552 error = ENOMEM; 3553 } 3554 3555 newrule->overload_tbl = NULL; 3556 if (newrule->overload_tblname[0]) { 3557 if ((newrule->overload_tbl = pfr_attach_table( 3558 ruleset, newrule->overload_tblname)) == 3559 NULL) 3560 error = EINVAL; 3561 else 3562 newrule->overload_tbl->pfrkt_flags |= 3563 PFR_TFLAG_ACTIVE; 3564 } 3565 3566 pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list); 3567 if (((((newrule->action == PF_NAT) || 3568 (newrule->action == PF_RDR) || 3569 (newrule->action == PF_BINAT) || 3570 (newrule->rt > PF_NOPFROUTE)) && 3571 !newrule->anchor)) && 3572 (TAILQ_FIRST(&newrule->rpool.list) == NULL)) 3573 error = EINVAL; 3574 3575 if (error) { 3576 pf_free_rule(newrule); 3577 PF_RULES_WUNLOCK(); 3578 PF_CONFIG_UNLOCK(); 3579 break; 3580 } 3581 3582 newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list); 3583 } 3584 pf_empty_kpool(&V_pf_pabuf); 3585 3586 if (pcr->action == PF_CHANGE_ADD_HEAD) 3587 oldrule = TAILQ_FIRST( 3588 ruleset->rules[rs_num].active.ptr); 3589 else if (pcr->action == PF_CHANGE_ADD_TAIL) 3590 oldrule = TAILQ_LAST( 3591 ruleset->rules[rs_num].active.ptr, pf_krulequeue); 3592 else { 3593 oldrule = TAILQ_FIRST( 3594 ruleset->rules[rs_num].active.ptr); 3595 while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) 3596 oldrule = TAILQ_NEXT(oldrule, entries); 3597 if (oldrule == NULL) { 3598 if (newrule != NULL) 3599 pf_free_rule(newrule); 3600 PF_RULES_WUNLOCK(); 3601 PF_CONFIG_UNLOCK(); 3602 error = EINVAL; 3603 break; 3604 } 3605 } 3606 3607 if (pcr->action == PF_CHANGE_REMOVE) { 3608 pf_unlink_rule(ruleset->rules[rs_num].active.ptr, 3609 oldrule); 3610 RB_REMOVE(pf_krule_global, 3611 ruleset->rules[rs_num].active.tree, oldrule); 3612 ruleset->rules[rs_num].active.rcount--; 3613 } else { 3614 pf_hash_rule(newrule); 3615 if (RB_INSERT(pf_krule_global, 3616 ruleset->rules[rs_num].active.tree, newrule) != NULL) { 3617 pf_free_rule(newrule); 3618 PF_RULES_WUNLOCK(); 3619 PF_CONFIG_UNLOCK(); 3620 error = EEXIST; 3621 break; 3622 } 3623 3624 if (oldrule == NULL) 3625 TAILQ_INSERT_TAIL( 3626 ruleset->rules[rs_num].active.ptr, 3627 newrule, entries); 3628 else if (pcr->action == PF_CHANGE_ADD_HEAD || 3629 pcr->action == PF_CHANGE_ADD_BEFORE) 3630 TAILQ_INSERT_BEFORE(oldrule, newrule, entries); 3631 else 3632 TAILQ_INSERT_AFTER( 3633 ruleset->rules[rs_num].active.ptr, 3634 oldrule, newrule, entries); 3635 ruleset->rules[rs_num].active.rcount++; 3636 } 3637 3638 nr = 0; 3639 TAILQ_FOREACH(oldrule, 3640 ruleset->rules[rs_num].active.ptr, entries) 3641 oldrule->nr = nr++; 3642 3643 ruleset->rules[rs_num].active.ticket++; 3644 3645 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); 3646 pf_remove_if_empty_kruleset(ruleset); 3647 3648 PF_RULES_WUNLOCK(); 3649 PF_CONFIG_UNLOCK(); 3650 break; 3651 3652 #undef ERROUT 3653 DIOCCHANGERULE_error: 3654 PF_RULES_WUNLOCK(); 3655 PF_CONFIG_UNLOCK(); 3656 pf_krule_free(newrule); 3657 pf_kkif_free(kif); 3658 break; 3659 } 3660 3661 case DIOCCLRSTATES: { 3662 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr; 3663 struct pf_kstate_kill kill; 3664 3665 error = pf_state_kill_to_kstate_kill(psk, &kill); 3666 if (error) 3667 break; 3668 3669 psk->psk_killed = pf_clear_states(&kill); 3670 break; 3671 } 3672 3673 case DIOCCLRSTATESNV: { 3674 error = pf_clearstates_nv((struct pfioc_nv *)addr); 3675 break; 3676 } 3677 3678 case DIOCKILLSTATES: { 3679 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr; 3680 struct pf_kstate_kill kill; 3681 3682 error = pf_state_kill_to_kstate_kill(psk, &kill); 3683 if (error) 3684 break; 3685 3686 psk->psk_killed = 0; 3687 pf_killstates(&kill, &psk->psk_killed); 3688 break; 3689 } 3690 3691 case DIOCKILLSTATESNV: { 3692 error = pf_killstates_nv((struct pfioc_nv *)addr); 3693 break; 3694 } 3695 3696 case DIOCADDSTATE: { 3697 struct pfioc_state *ps = (struct pfioc_state *)addr; 3698 struct pfsync_state *sp = &ps->state; 3699 3700 if (sp->timeout >= PFTM_MAX) { 3701 error = EINVAL; 3702 break; 3703 } 3704 if (V_pfsync_state_import_ptr != NULL) { 3705 PF_RULES_RLOCK(); 3706 error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL); 3707 PF_RULES_RUNLOCK(); 3708 } else 3709 error = EOPNOTSUPP; 3710 break; 3711 } 3712 3713 case DIOCGETSTATE: { 3714 struct pfioc_state *ps = (struct pfioc_state *)addr; 3715 struct pf_kstate *s; 3716 3717 s = pf_find_state_byid(ps->state.id, ps->state.creatorid); 3718 if (s == NULL) { 3719 error = ENOENT; 3720 break; 3721 } 3722 3723 pfsync_state_export(&ps->state, s); 3724 PF_STATE_UNLOCK(s); 3725 break; 3726 } 3727 3728 case DIOCGETSTATENV: { 3729 error = pf_getstate((struct pfioc_nv *)addr); 3730 break; 3731 } 3732 3733 case DIOCGETSTATES: { 3734 struct pfioc_states *ps = (struct pfioc_states *)addr; 3735 struct pf_kstate *s; 3736 struct pfsync_state *pstore, *p; 3737 int i, nr; 3738 size_t slice_count = 16, count; 3739 void *out; 3740 3741 if (ps->ps_len <= 0) { 3742 nr = uma_zone_get_cur(V_pf_state_z); 3743 ps->ps_len = sizeof(struct pfsync_state) * nr; 3744 break; 3745 } 3746 3747 out = ps->ps_states; 3748 pstore = mallocarray(slice_count, 3749 sizeof(struct pfsync_state), M_TEMP, M_WAITOK | M_ZERO); 3750 nr = 0; 3751 3752 for (i = 0; i <= pf_hashmask; i++) { 3753 struct pf_idhash *ih = &V_pf_idhash[i]; 3754 3755 DIOCGETSTATES_retry: 3756 p = pstore; 3757 3758 if (LIST_EMPTY(&ih->states)) 3759 continue; 3760 3761 PF_HASHROW_LOCK(ih); 3762 count = 0; 3763 LIST_FOREACH(s, &ih->states, entry) { 3764 if (s->timeout == PFTM_UNLINKED) 3765 continue; 3766 count++; 3767 } 3768 3769 if (count > slice_count) { 3770 PF_HASHROW_UNLOCK(ih); 3771 free(pstore, M_TEMP); 3772 slice_count = count * 2; 3773 pstore = mallocarray(slice_count, 3774 sizeof(struct pfsync_state), M_TEMP, 3775 M_WAITOK | M_ZERO); 3776 goto DIOCGETSTATES_retry; 3777 } 3778 3779 if ((nr+count) * sizeof(*p) > ps->ps_len) { 3780 PF_HASHROW_UNLOCK(ih); 3781 goto DIOCGETSTATES_full; 3782 } 3783 3784 LIST_FOREACH(s, &ih->states, entry) { 3785 if (s->timeout == PFTM_UNLINKED) 3786 continue; 3787 3788 pfsync_state_export(p, s); 3789 p++; 3790 nr++; 3791 } 3792 PF_HASHROW_UNLOCK(ih); 3793 error = copyout(pstore, out, 3794 sizeof(struct pfsync_state) * count); 3795 if (error) 3796 break; 3797 out = ps->ps_states + nr; 3798 } 3799 DIOCGETSTATES_full: 3800 ps->ps_len = sizeof(struct pfsync_state) * nr; 3801 free(pstore, M_TEMP); 3802 3803 break; 3804 } 3805 3806 case DIOCGETSTATESV2: { 3807 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr; 3808 struct pf_kstate *s; 3809 struct pf_state_export *pstore, *p; 3810 int i, nr; 3811 size_t slice_count = 16, count; 3812 void *out; 3813 3814 if (ps->ps_req_version > PF_STATE_VERSION) { 3815 error = ENOTSUP; 3816 break; 3817 } 3818 3819 if (ps->ps_len <= 0) { 3820 nr = uma_zone_get_cur(V_pf_state_z); 3821 ps->ps_len = sizeof(struct pf_state_export) * nr; 3822 break; 3823 } 3824 3825 out = ps->ps_states; 3826 pstore = mallocarray(slice_count, 3827 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO); 3828 nr = 0; 3829 3830 for (i = 0; i <= pf_hashmask; i++) { 3831 struct pf_idhash *ih = &V_pf_idhash[i]; 3832 3833 DIOCGETSTATESV2_retry: 3834 p = pstore; 3835 3836 if (LIST_EMPTY(&ih->states)) 3837 continue; 3838 3839 PF_HASHROW_LOCK(ih); 3840 count = 0; 3841 LIST_FOREACH(s, &ih->states, entry) { 3842 if (s->timeout == PFTM_UNLINKED) 3843 continue; 3844 count++; 3845 } 3846 3847 if (count > slice_count) { 3848 PF_HASHROW_UNLOCK(ih); 3849 free(pstore, M_TEMP); 3850 slice_count = count * 2; 3851 pstore = mallocarray(slice_count, 3852 sizeof(struct pf_state_export), M_TEMP, 3853 M_WAITOK | M_ZERO); 3854 goto DIOCGETSTATESV2_retry; 3855 } 3856 3857 if ((nr+count) * sizeof(*p) > ps->ps_len) { 3858 PF_HASHROW_UNLOCK(ih); 3859 goto DIOCGETSTATESV2_full; 3860 } 3861 3862 LIST_FOREACH(s, &ih->states, entry) { 3863 if (s->timeout == PFTM_UNLINKED) 3864 continue; 3865 3866 pf_state_export(p, s); 3867 p++; 3868 nr++; 3869 } 3870 PF_HASHROW_UNLOCK(ih); 3871 error = copyout(pstore, out, 3872 sizeof(struct pf_state_export) * count); 3873 if (error) 3874 break; 3875 out = ps->ps_states + nr; 3876 } 3877 DIOCGETSTATESV2_full: 3878 ps->ps_len = nr * sizeof(struct pf_state_export); 3879 free(pstore, M_TEMP); 3880 3881 break; 3882 } 3883 3884 case DIOCGETSTATUS: { 3885 struct pf_status *s = (struct pf_status *)addr; 3886 3887 PF_RULES_RLOCK(); 3888 s->running = V_pf_status.running; 3889 s->since = V_pf_status.since; 3890 s->debug = V_pf_status.debug; 3891 s->hostid = V_pf_status.hostid; 3892 s->states = V_pf_status.states; 3893 s->src_nodes = V_pf_status.src_nodes; 3894 3895 for (int i = 0; i < PFRES_MAX; i++) 3896 s->counters[i] = 3897 counter_u64_fetch(V_pf_status.counters[i]); 3898 for (int i = 0; i < LCNT_MAX; i++) 3899 s->lcounters[i] = 3900 counter_u64_fetch(V_pf_status.lcounters[i]); 3901 for (int i = 0; i < FCNT_MAX; i++) 3902 s->fcounters[i] = 3903 pf_counter_u64_fetch(&V_pf_status.fcounters[i]); 3904 for (int i = 0; i < SCNT_MAX; i++) 3905 s->scounters[i] = 3906 counter_u64_fetch(V_pf_status.scounters[i]); 3907 3908 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ); 3909 bcopy(V_pf_status.pf_chksum, s->pf_chksum, 3910 PF_MD5_DIGEST_LENGTH); 3911 3912 pfi_update_status(s->ifname, s); 3913 PF_RULES_RUNLOCK(); 3914 break; 3915 } 3916 3917 case DIOCGETSTATUSNV: { 3918 error = pf_getstatus((struct pfioc_nv *)addr); 3919 break; 3920 } 3921 3922 case DIOCSETSTATUSIF: { 3923 struct pfioc_if *pi = (struct pfioc_if *)addr; 3924 3925 if (pi->ifname[0] == 0) { 3926 bzero(V_pf_status.ifname, IFNAMSIZ); 3927 break; 3928 } 3929 PF_RULES_WLOCK(); 3930 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); 3931 PF_RULES_WUNLOCK(); 3932 break; 3933 } 3934 3935 case DIOCCLRSTATUS: { 3936 PF_RULES_WLOCK(); 3937 for (int i = 0; i < PFRES_MAX; i++) 3938 counter_u64_zero(V_pf_status.counters[i]); 3939 for (int i = 0; i < FCNT_MAX; i++) 3940 pf_counter_u64_zero(&V_pf_status.fcounters[i]); 3941 for (int i = 0; i < SCNT_MAX; i++) 3942 counter_u64_zero(V_pf_status.scounters[i]); 3943 for (int i = 0; i < KLCNT_MAX; i++) 3944 counter_u64_zero(V_pf_status.lcounters[i]); 3945 V_pf_status.since = time_second; 3946 if (*V_pf_status.ifname) 3947 pfi_update_status(V_pf_status.ifname, NULL); 3948 PF_RULES_WUNLOCK(); 3949 break; 3950 } 3951 3952 case DIOCNATLOOK: { 3953 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; 3954 struct pf_state_key *sk; 3955 struct pf_kstate *state; 3956 struct pf_state_key_cmp key; 3957 int m = 0, direction = pnl->direction; 3958 int sidx, didx; 3959 3960 /* NATLOOK src and dst are reversed, so reverse sidx/didx */ 3961 sidx = (direction == PF_IN) ? 1 : 0; 3962 didx = (direction == PF_IN) ? 0 : 1; 3963 3964 if (!pnl->proto || 3965 PF_AZERO(&pnl->saddr, pnl->af) || 3966 PF_AZERO(&pnl->daddr, pnl->af) || 3967 ((pnl->proto == IPPROTO_TCP || 3968 pnl->proto == IPPROTO_UDP) && 3969 (!pnl->dport || !pnl->sport))) 3970 error = EINVAL; 3971 else { 3972 bzero(&key, sizeof(key)); 3973 key.af = pnl->af; 3974 key.proto = pnl->proto; 3975 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af); 3976 key.port[sidx] = pnl->sport; 3977 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af); 3978 key.port[didx] = pnl->dport; 3979 3980 state = pf_find_state_all(&key, direction, &m); 3981 if (state == NULL) { 3982 error = ENOENT; 3983 } else { 3984 if (m > 1) { 3985 PF_STATE_UNLOCK(state); 3986 error = E2BIG; /* more than one state */ 3987 } else { 3988 sk = state->key[sidx]; 3989 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af); 3990 pnl->rsport = sk->port[sidx]; 3991 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af); 3992 pnl->rdport = sk->port[didx]; 3993 PF_STATE_UNLOCK(state); 3994 } 3995 } 3996 } 3997 break; 3998 } 3999 4000 case DIOCSETTIMEOUT: { 4001 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4002 int old; 4003 4004 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX || 4005 pt->seconds < 0) { 4006 error = EINVAL; 4007 break; 4008 } 4009 PF_RULES_WLOCK(); 4010 old = V_pf_default_rule.timeout[pt->timeout]; 4011 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0) 4012 pt->seconds = 1; 4013 V_pf_default_rule.timeout[pt->timeout] = pt->seconds; 4014 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old) 4015 wakeup(pf_purge_thread); 4016 pt->seconds = old; 4017 PF_RULES_WUNLOCK(); 4018 break; 4019 } 4020 4021 case DIOCGETTIMEOUT: { 4022 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4023 4024 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) { 4025 error = EINVAL; 4026 break; 4027 } 4028 PF_RULES_RLOCK(); 4029 pt->seconds = V_pf_default_rule.timeout[pt->timeout]; 4030 PF_RULES_RUNLOCK(); 4031 break; 4032 } 4033 4034 case DIOCGETLIMIT: { 4035 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4036 4037 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) { 4038 error = EINVAL; 4039 break; 4040 } 4041 PF_RULES_RLOCK(); 4042 pl->limit = V_pf_limits[pl->index].limit; 4043 PF_RULES_RUNLOCK(); 4044 break; 4045 } 4046 4047 case DIOCSETLIMIT: { 4048 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4049 int old_limit; 4050 4051 PF_RULES_WLOCK(); 4052 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX || 4053 V_pf_limits[pl->index].zone == NULL) { 4054 PF_RULES_WUNLOCK(); 4055 error = EINVAL; 4056 break; 4057 } 4058 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit); 4059 old_limit = V_pf_limits[pl->index].limit; 4060 V_pf_limits[pl->index].limit = pl->limit; 4061 pl->limit = old_limit; 4062 PF_RULES_WUNLOCK(); 4063 break; 4064 } 4065 4066 case DIOCSETDEBUG: { 4067 u_int32_t *level = (u_int32_t *)addr; 4068 4069 PF_RULES_WLOCK(); 4070 V_pf_status.debug = *level; 4071 PF_RULES_WUNLOCK(); 4072 break; 4073 } 4074 4075 case DIOCCLRRULECTRS: { 4076 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ 4077 struct pf_kruleset *ruleset = &pf_main_ruleset; 4078 struct pf_krule *rule; 4079 4080 PF_RULES_WLOCK(); 4081 TAILQ_FOREACH(rule, 4082 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { 4083 pf_counter_u64_zero(&rule->evaluations); 4084 for (int i = 0; i < 2; i++) { 4085 pf_counter_u64_zero(&rule->packets[i]); 4086 pf_counter_u64_zero(&rule->bytes[i]); 4087 } 4088 } 4089 PF_RULES_WUNLOCK(); 4090 break; 4091 } 4092 4093 case DIOCGIFSPEEDV0: 4094 case DIOCGIFSPEEDV1: { 4095 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr; 4096 struct pf_ifspeed_v1 ps; 4097 struct ifnet *ifp; 4098 4099 if (psp->ifname[0] == '\0') { 4100 error = EINVAL; 4101 break; 4102 } 4103 4104 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); 4105 if (error != 0) 4106 break; 4107 ifp = ifunit(ps.ifname); 4108 if (ifp != NULL) { 4109 psp->baudrate32 = 4110 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX); 4111 if (cmd == DIOCGIFSPEEDV1) 4112 psp->baudrate = ifp->if_baudrate; 4113 } else { 4114 error = EINVAL; 4115 } 4116 break; 4117 } 4118 4119 #ifdef ALTQ 4120 case DIOCSTARTALTQ: { 4121 struct pf_altq *altq; 4122 4123 PF_RULES_WLOCK(); 4124 /* enable all altq interfaces on active list */ 4125 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4126 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4127 error = pf_enable_altq(altq); 4128 if (error != 0) 4129 break; 4130 } 4131 } 4132 if (error == 0) 4133 V_pf_altq_running = 1; 4134 PF_RULES_WUNLOCK(); 4135 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n")); 4136 break; 4137 } 4138 4139 case DIOCSTOPALTQ: { 4140 struct pf_altq *altq; 4141 4142 PF_RULES_WLOCK(); 4143 /* disable all altq interfaces on active list */ 4144 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4145 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4146 error = pf_disable_altq(altq); 4147 if (error != 0) 4148 break; 4149 } 4150 } 4151 if (error == 0) 4152 V_pf_altq_running = 0; 4153 PF_RULES_WUNLOCK(); 4154 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n")); 4155 break; 4156 } 4157 4158 case DIOCADDALTQV0: 4159 case DIOCADDALTQV1: { 4160 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4161 struct pf_altq *altq, *a; 4162 struct ifnet *ifp; 4163 4164 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); 4165 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); 4166 if (error) 4167 break; 4168 altq->local_flags = 0; 4169 4170 PF_RULES_WLOCK(); 4171 if (pa->ticket != V_ticket_altqs_inactive) { 4172 PF_RULES_WUNLOCK(); 4173 free(altq, M_PFALTQ); 4174 error = EBUSY; 4175 break; 4176 } 4177 4178 /* 4179 * if this is for a queue, find the discipline and 4180 * copy the necessary fields 4181 */ 4182 if (altq->qname[0] != 0) { 4183 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) { 4184 PF_RULES_WUNLOCK(); 4185 error = EBUSY; 4186 free(altq, M_PFALTQ); 4187 break; 4188 } 4189 altq->altq_disc = NULL; 4190 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { 4191 if (strncmp(a->ifname, altq->ifname, 4192 IFNAMSIZ) == 0) { 4193 altq->altq_disc = a->altq_disc; 4194 break; 4195 } 4196 } 4197 } 4198 4199 if ((ifp = ifunit(altq->ifname)) == NULL) 4200 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 4201 else 4202 error = altq_add(ifp, altq); 4203 4204 if (error) { 4205 PF_RULES_WUNLOCK(); 4206 free(altq, M_PFALTQ); 4207 break; 4208 } 4209 4210 if (altq->qname[0] != 0) 4211 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); 4212 else 4213 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries); 4214 /* version error check done on import above */ 4215 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4216 PF_RULES_WUNLOCK(); 4217 break; 4218 } 4219 4220 case DIOCGETALTQSV0: 4221 case DIOCGETALTQSV1: { 4222 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4223 struct pf_altq *altq; 4224 4225 PF_RULES_RLOCK(); 4226 pa->nr = 0; 4227 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) 4228 pa->nr++; 4229 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) 4230 pa->nr++; 4231 pa->ticket = V_ticket_altqs_active; 4232 PF_RULES_RUNLOCK(); 4233 break; 4234 } 4235 4236 case DIOCGETALTQV0: 4237 case DIOCGETALTQV1: { 4238 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4239 struct pf_altq *altq; 4240 4241 PF_RULES_RLOCK(); 4242 if (pa->ticket != V_ticket_altqs_active) { 4243 PF_RULES_RUNLOCK(); 4244 error = EBUSY; 4245 break; 4246 } 4247 altq = pf_altq_get_nth_active(pa->nr); 4248 if (altq == NULL) { 4249 PF_RULES_RUNLOCK(); 4250 error = EBUSY; 4251 break; 4252 } 4253 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4254 PF_RULES_RUNLOCK(); 4255 break; 4256 } 4257 4258 case DIOCCHANGEALTQV0: 4259 case DIOCCHANGEALTQV1: 4260 /* CHANGEALTQ not supported yet! */ 4261 error = ENODEV; 4262 break; 4263 4264 case DIOCGETQSTATSV0: 4265 case DIOCGETQSTATSV1: { 4266 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr; 4267 struct pf_altq *altq; 4268 int nbytes; 4269 u_int32_t version; 4270 4271 PF_RULES_RLOCK(); 4272 if (pq->ticket != V_ticket_altqs_active) { 4273 PF_RULES_RUNLOCK(); 4274 error = EBUSY; 4275 break; 4276 } 4277 nbytes = pq->nbytes; 4278 altq = pf_altq_get_nth_active(pq->nr); 4279 if (altq == NULL) { 4280 PF_RULES_RUNLOCK(); 4281 error = EBUSY; 4282 break; 4283 } 4284 4285 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { 4286 PF_RULES_RUNLOCK(); 4287 error = ENXIO; 4288 break; 4289 } 4290 PF_RULES_RUNLOCK(); 4291 if (cmd == DIOCGETQSTATSV0) 4292 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */ 4293 else 4294 version = pq->version; 4295 error = altq_getqstats(altq, pq->buf, &nbytes, version); 4296 if (error == 0) { 4297 pq->scheduler = altq->scheduler; 4298 pq->nbytes = nbytes; 4299 } 4300 break; 4301 } 4302 #endif /* ALTQ */ 4303 4304 case DIOCBEGINADDRS: { 4305 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4306 4307 PF_RULES_WLOCK(); 4308 pf_empty_kpool(&V_pf_pabuf); 4309 pp->ticket = ++V_ticket_pabuf; 4310 PF_RULES_WUNLOCK(); 4311 break; 4312 } 4313 4314 case DIOCADDADDR: { 4315 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4316 struct pf_kpooladdr *pa; 4317 struct pfi_kkif *kif = NULL; 4318 4319 #ifndef INET 4320 if (pp->af == AF_INET) { 4321 error = EAFNOSUPPORT; 4322 break; 4323 } 4324 #endif /* INET */ 4325 #ifndef INET6 4326 if (pp->af == AF_INET6) { 4327 error = EAFNOSUPPORT; 4328 break; 4329 } 4330 #endif /* INET6 */ 4331 if (pp->addr.addr.type != PF_ADDR_ADDRMASK && 4332 pp->addr.addr.type != PF_ADDR_DYNIFTL && 4333 pp->addr.addr.type != PF_ADDR_TABLE) { 4334 error = EINVAL; 4335 break; 4336 } 4337 if (pp->addr.addr.p.dyn != NULL) { 4338 error = EINVAL; 4339 break; 4340 } 4341 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); 4342 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa); 4343 if (error != 0) 4344 break; 4345 if (pa->ifname[0]) 4346 kif = pf_kkif_create(M_WAITOK); 4347 PF_RULES_WLOCK(); 4348 if (pp->ticket != V_ticket_pabuf) { 4349 PF_RULES_WUNLOCK(); 4350 if (pa->ifname[0]) 4351 pf_kkif_free(kif); 4352 free(pa, M_PFRULE); 4353 error = EBUSY; 4354 break; 4355 } 4356 if (pa->ifname[0]) { 4357 pa->kif = pfi_kkif_attach(kif, pa->ifname); 4358 kif = NULL; 4359 pfi_kkif_ref(pa->kif); 4360 } else 4361 pa->kif = NULL; 4362 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = 4363 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { 4364 if (pa->ifname[0]) 4365 pfi_kkif_unref(pa->kif); 4366 PF_RULES_WUNLOCK(); 4367 free(pa, M_PFRULE); 4368 break; 4369 } 4370 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries); 4371 PF_RULES_WUNLOCK(); 4372 break; 4373 } 4374 4375 case DIOCGETADDRS: { 4376 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4377 struct pf_kpool *pool; 4378 struct pf_kpooladdr *pa; 4379 4380 pp->anchor[sizeof(pp->anchor) - 1] = 0; 4381 pp->nr = 0; 4382 4383 PF_RULES_RLOCK(); 4384 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 4385 pp->r_num, 0, 1, 0); 4386 if (pool == NULL) { 4387 PF_RULES_RUNLOCK(); 4388 error = EBUSY; 4389 break; 4390 } 4391 TAILQ_FOREACH(pa, &pool->list, entries) 4392 pp->nr++; 4393 PF_RULES_RUNLOCK(); 4394 break; 4395 } 4396 4397 case DIOCGETADDR: { 4398 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4399 struct pf_kpool *pool; 4400 struct pf_kpooladdr *pa; 4401 u_int32_t nr = 0; 4402 4403 pp->anchor[sizeof(pp->anchor) - 1] = 0; 4404 4405 PF_RULES_RLOCK(); 4406 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 4407 pp->r_num, 0, 1, 1); 4408 if (pool == NULL) { 4409 PF_RULES_RUNLOCK(); 4410 error = EBUSY; 4411 break; 4412 } 4413 pa = TAILQ_FIRST(&pool->list); 4414 while ((pa != NULL) && (nr < pp->nr)) { 4415 pa = TAILQ_NEXT(pa, entries); 4416 nr++; 4417 } 4418 if (pa == NULL) { 4419 PF_RULES_RUNLOCK(); 4420 error = EBUSY; 4421 break; 4422 } 4423 pf_kpooladdr_to_pooladdr(pa, &pp->addr); 4424 pf_addr_copyout(&pp->addr.addr); 4425 PF_RULES_RUNLOCK(); 4426 break; 4427 } 4428 4429 case DIOCCHANGEADDR: { 4430 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; 4431 struct pf_kpool *pool; 4432 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL; 4433 struct pf_kruleset *ruleset; 4434 struct pfi_kkif *kif = NULL; 4435 4436 pca->anchor[sizeof(pca->anchor) - 1] = 0; 4437 4438 if (pca->action < PF_CHANGE_ADD_HEAD || 4439 pca->action > PF_CHANGE_REMOVE) { 4440 error = EINVAL; 4441 break; 4442 } 4443 if (pca->addr.addr.type != PF_ADDR_ADDRMASK && 4444 pca->addr.addr.type != PF_ADDR_DYNIFTL && 4445 pca->addr.addr.type != PF_ADDR_TABLE) { 4446 error = EINVAL; 4447 break; 4448 } 4449 if (pca->addr.addr.p.dyn != NULL) { 4450 error = EINVAL; 4451 break; 4452 } 4453 4454 if (pca->action != PF_CHANGE_REMOVE) { 4455 #ifndef INET 4456 if (pca->af == AF_INET) { 4457 error = EAFNOSUPPORT; 4458 break; 4459 } 4460 #endif /* INET */ 4461 #ifndef INET6 4462 if (pca->af == AF_INET6) { 4463 error = EAFNOSUPPORT; 4464 break; 4465 } 4466 #endif /* INET6 */ 4467 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); 4468 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); 4469 if (newpa->ifname[0]) 4470 kif = pf_kkif_create(M_WAITOK); 4471 newpa->kif = NULL; 4472 } 4473 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x) 4474 PF_RULES_WLOCK(); 4475 ruleset = pf_find_kruleset(pca->anchor); 4476 if (ruleset == NULL) 4477 ERROUT(EBUSY); 4478 4479 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action, 4480 pca->r_num, pca->r_last, 1, 1); 4481 if (pool == NULL) 4482 ERROUT(EBUSY); 4483 4484 if (pca->action != PF_CHANGE_REMOVE) { 4485 if (newpa->ifname[0]) { 4486 newpa->kif = pfi_kkif_attach(kif, newpa->ifname); 4487 pfi_kkif_ref(newpa->kif); 4488 kif = NULL; 4489 } 4490 4491 switch (newpa->addr.type) { 4492 case PF_ADDR_DYNIFTL: 4493 error = pfi_dynaddr_setup(&newpa->addr, 4494 pca->af); 4495 break; 4496 case PF_ADDR_TABLE: 4497 newpa->addr.p.tbl = pfr_attach_table(ruleset, 4498 newpa->addr.v.tblname); 4499 if (newpa->addr.p.tbl == NULL) 4500 error = ENOMEM; 4501 break; 4502 } 4503 if (error) 4504 goto DIOCCHANGEADDR_error; 4505 } 4506 4507 switch (pca->action) { 4508 case PF_CHANGE_ADD_HEAD: 4509 oldpa = TAILQ_FIRST(&pool->list); 4510 break; 4511 case PF_CHANGE_ADD_TAIL: 4512 oldpa = TAILQ_LAST(&pool->list, pf_kpalist); 4513 break; 4514 default: 4515 oldpa = TAILQ_FIRST(&pool->list); 4516 for (int i = 0; oldpa && i < pca->nr; i++) 4517 oldpa = TAILQ_NEXT(oldpa, entries); 4518 4519 if (oldpa == NULL) 4520 ERROUT(EINVAL); 4521 } 4522 4523 if (pca->action == PF_CHANGE_REMOVE) { 4524 TAILQ_REMOVE(&pool->list, oldpa, entries); 4525 switch (oldpa->addr.type) { 4526 case PF_ADDR_DYNIFTL: 4527 pfi_dynaddr_remove(oldpa->addr.p.dyn); 4528 break; 4529 case PF_ADDR_TABLE: 4530 pfr_detach_table(oldpa->addr.p.tbl); 4531 break; 4532 } 4533 if (oldpa->kif) 4534 pfi_kkif_unref(oldpa->kif); 4535 free(oldpa, M_PFRULE); 4536 } else { 4537 if (oldpa == NULL) 4538 TAILQ_INSERT_TAIL(&pool->list, newpa, entries); 4539 else if (pca->action == PF_CHANGE_ADD_HEAD || 4540 pca->action == PF_CHANGE_ADD_BEFORE) 4541 TAILQ_INSERT_BEFORE(oldpa, newpa, entries); 4542 else 4543 TAILQ_INSERT_AFTER(&pool->list, oldpa, 4544 newpa, entries); 4545 } 4546 4547 pool->cur = TAILQ_FIRST(&pool->list); 4548 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); 4549 PF_RULES_WUNLOCK(); 4550 break; 4551 4552 #undef ERROUT 4553 DIOCCHANGEADDR_error: 4554 if (newpa != NULL) { 4555 if (newpa->kif) 4556 pfi_kkif_unref(newpa->kif); 4557 free(newpa, M_PFRULE); 4558 } 4559 PF_RULES_WUNLOCK(); 4560 pf_kkif_free(kif); 4561 break; 4562 } 4563 4564 case DIOCGETRULESETS: { 4565 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4566 struct pf_kruleset *ruleset; 4567 struct pf_kanchor *anchor; 4568 4569 pr->path[sizeof(pr->path) - 1] = 0; 4570 4571 PF_RULES_RLOCK(); 4572 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 4573 PF_RULES_RUNLOCK(); 4574 error = ENOENT; 4575 break; 4576 } 4577 pr->nr = 0; 4578 if (ruleset->anchor == NULL) { 4579 /* XXX kludge for pf_main_ruleset */ 4580 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 4581 if (anchor->parent == NULL) 4582 pr->nr++; 4583 } else { 4584 RB_FOREACH(anchor, pf_kanchor_node, 4585 &ruleset->anchor->children) 4586 pr->nr++; 4587 } 4588 PF_RULES_RUNLOCK(); 4589 break; 4590 } 4591 4592 case DIOCGETRULESET: { 4593 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4594 struct pf_kruleset *ruleset; 4595 struct pf_kanchor *anchor; 4596 u_int32_t nr = 0; 4597 4598 pr->path[sizeof(pr->path) - 1] = 0; 4599 4600 PF_RULES_RLOCK(); 4601 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 4602 PF_RULES_RUNLOCK(); 4603 error = ENOENT; 4604 break; 4605 } 4606 pr->name[0] = 0; 4607 if (ruleset->anchor == NULL) { 4608 /* XXX kludge for pf_main_ruleset */ 4609 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 4610 if (anchor->parent == NULL && nr++ == pr->nr) { 4611 strlcpy(pr->name, anchor->name, 4612 sizeof(pr->name)); 4613 break; 4614 } 4615 } else { 4616 RB_FOREACH(anchor, pf_kanchor_node, 4617 &ruleset->anchor->children) 4618 if (nr++ == pr->nr) { 4619 strlcpy(pr->name, anchor->name, 4620 sizeof(pr->name)); 4621 break; 4622 } 4623 } 4624 if (!pr->name[0]) 4625 error = EBUSY; 4626 PF_RULES_RUNLOCK(); 4627 break; 4628 } 4629 4630 case DIOCRCLRTABLES: { 4631 struct pfioc_table *io = (struct pfioc_table *)addr; 4632 4633 if (io->pfrio_esize != 0) { 4634 error = ENODEV; 4635 break; 4636 } 4637 PF_RULES_WLOCK(); 4638 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, 4639 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4640 PF_RULES_WUNLOCK(); 4641 break; 4642 } 4643 4644 case DIOCRADDTABLES: { 4645 struct pfioc_table *io = (struct pfioc_table *)addr; 4646 struct pfr_table *pfrts; 4647 size_t totlen; 4648 4649 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4650 error = ENODEV; 4651 break; 4652 } 4653 4654 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4655 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4656 error = ENOMEM; 4657 break; 4658 } 4659 4660 totlen = io->pfrio_size * sizeof(struct pfr_table); 4661 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4662 M_TEMP, M_WAITOK); 4663 error = copyin(io->pfrio_buffer, pfrts, totlen); 4664 if (error) { 4665 free(pfrts, M_TEMP); 4666 break; 4667 } 4668 PF_RULES_WLOCK(); 4669 error = pfr_add_tables(pfrts, io->pfrio_size, 4670 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4671 PF_RULES_WUNLOCK(); 4672 free(pfrts, M_TEMP); 4673 break; 4674 } 4675 4676 case DIOCRDELTABLES: { 4677 struct pfioc_table *io = (struct pfioc_table *)addr; 4678 struct pfr_table *pfrts; 4679 size_t totlen; 4680 4681 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4682 error = ENODEV; 4683 break; 4684 } 4685 4686 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4687 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4688 error = ENOMEM; 4689 break; 4690 } 4691 4692 totlen = io->pfrio_size * sizeof(struct pfr_table); 4693 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4694 M_TEMP, M_WAITOK); 4695 error = copyin(io->pfrio_buffer, pfrts, totlen); 4696 if (error) { 4697 free(pfrts, M_TEMP); 4698 break; 4699 } 4700 PF_RULES_WLOCK(); 4701 error = pfr_del_tables(pfrts, io->pfrio_size, 4702 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4703 PF_RULES_WUNLOCK(); 4704 free(pfrts, M_TEMP); 4705 break; 4706 } 4707 4708 case DIOCRGETTABLES: { 4709 struct pfioc_table *io = (struct pfioc_table *)addr; 4710 struct pfr_table *pfrts; 4711 size_t totlen; 4712 int n; 4713 4714 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4715 error = ENODEV; 4716 break; 4717 } 4718 PF_RULES_RLOCK(); 4719 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4720 if (n < 0) { 4721 PF_RULES_RUNLOCK(); 4722 error = EINVAL; 4723 break; 4724 } 4725 io->pfrio_size = min(io->pfrio_size, n); 4726 4727 totlen = io->pfrio_size * sizeof(struct pfr_table); 4728 4729 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4730 M_TEMP, M_NOWAIT | M_ZERO); 4731 if (pfrts == NULL) { 4732 error = ENOMEM; 4733 PF_RULES_RUNLOCK(); 4734 break; 4735 } 4736 error = pfr_get_tables(&io->pfrio_table, pfrts, 4737 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4738 PF_RULES_RUNLOCK(); 4739 if (error == 0) 4740 error = copyout(pfrts, io->pfrio_buffer, totlen); 4741 free(pfrts, M_TEMP); 4742 break; 4743 } 4744 4745 case DIOCRGETTSTATS: { 4746 struct pfioc_table *io = (struct pfioc_table *)addr; 4747 struct pfr_tstats *pfrtstats; 4748 size_t totlen; 4749 int n; 4750 4751 if (io->pfrio_esize != sizeof(struct pfr_tstats)) { 4752 error = ENODEV; 4753 break; 4754 } 4755 PF_TABLE_STATS_LOCK(); 4756 PF_RULES_RLOCK(); 4757 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4758 if (n < 0) { 4759 PF_RULES_RUNLOCK(); 4760 PF_TABLE_STATS_UNLOCK(); 4761 error = EINVAL; 4762 break; 4763 } 4764 io->pfrio_size = min(io->pfrio_size, n); 4765 4766 totlen = io->pfrio_size * sizeof(struct pfr_tstats); 4767 pfrtstats = mallocarray(io->pfrio_size, 4768 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO); 4769 if (pfrtstats == NULL) { 4770 error = ENOMEM; 4771 PF_RULES_RUNLOCK(); 4772 PF_TABLE_STATS_UNLOCK(); 4773 break; 4774 } 4775 error = pfr_get_tstats(&io->pfrio_table, pfrtstats, 4776 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4777 PF_RULES_RUNLOCK(); 4778 PF_TABLE_STATS_UNLOCK(); 4779 if (error == 0) 4780 error = copyout(pfrtstats, io->pfrio_buffer, totlen); 4781 free(pfrtstats, M_TEMP); 4782 break; 4783 } 4784 4785 case DIOCRCLRTSTATS: { 4786 struct pfioc_table *io = (struct pfioc_table *)addr; 4787 struct pfr_table *pfrts; 4788 size_t totlen; 4789 4790 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4791 error = ENODEV; 4792 break; 4793 } 4794 4795 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4796 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4797 /* We used to count tables and use the minimum required 4798 * size, so we didn't fail on overly large requests. 4799 * Keep doing so. */ 4800 io->pfrio_size = pf_ioctl_maxcount; 4801 break; 4802 } 4803 4804 totlen = io->pfrio_size * sizeof(struct pfr_table); 4805 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4806 M_TEMP, M_WAITOK); 4807 error = copyin(io->pfrio_buffer, pfrts, totlen); 4808 if (error) { 4809 free(pfrts, M_TEMP); 4810 break; 4811 } 4812 4813 PF_TABLE_STATS_LOCK(); 4814 PF_RULES_RLOCK(); 4815 error = pfr_clr_tstats(pfrts, io->pfrio_size, 4816 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4817 PF_RULES_RUNLOCK(); 4818 PF_TABLE_STATS_UNLOCK(); 4819 free(pfrts, M_TEMP); 4820 break; 4821 } 4822 4823 case DIOCRSETTFLAGS: { 4824 struct pfioc_table *io = (struct pfioc_table *)addr; 4825 struct pfr_table *pfrts; 4826 size_t totlen; 4827 int n; 4828 4829 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4830 error = ENODEV; 4831 break; 4832 } 4833 4834 PF_RULES_RLOCK(); 4835 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4836 if (n < 0) { 4837 PF_RULES_RUNLOCK(); 4838 error = EINVAL; 4839 break; 4840 } 4841 4842 io->pfrio_size = min(io->pfrio_size, n); 4843 PF_RULES_RUNLOCK(); 4844 4845 totlen = io->pfrio_size * sizeof(struct pfr_table); 4846 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4847 M_TEMP, M_WAITOK); 4848 error = copyin(io->pfrio_buffer, pfrts, totlen); 4849 if (error) { 4850 free(pfrts, M_TEMP); 4851 break; 4852 } 4853 PF_RULES_WLOCK(); 4854 error = pfr_set_tflags(pfrts, io->pfrio_size, 4855 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, 4856 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4857 PF_RULES_WUNLOCK(); 4858 free(pfrts, M_TEMP); 4859 break; 4860 } 4861 4862 case DIOCRCLRADDRS: { 4863 struct pfioc_table *io = (struct pfioc_table *)addr; 4864 4865 if (io->pfrio_esize != 0) { 4866 error = ENODEV; 4867 break; 4868 } 4869 PF_RULES_WLOCK(); 4870 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, 4871 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4872 PF_RULES_WUNLOCK(); 4873 break; 4874 } 4875 4876 case DIOCRADDADDRS: { 4877 struct pfioc_table *io = (struct pfioc_table *)addr; 4878 struct pfr_addr *pfras; 4879 size_t totlen; 4880 4881 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4882 error = ENODEV; 4883 break; 4884 } 4885 if (io->pfrio_size < 0 || 4886 io->pfrio_size > pf_ioctl_maxcount || 4887 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4888 error = EINVAL; 4889 break; 4890 } 4891 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4892 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4893 M_TEMP, M_WAITOK); 4894 error = copyin(io->pfrio_buffer, pfras, totlen); 4895 if (error) { 4896 free(pfras, M_TEMP); 4897 break; 4898 } 4899 PF_RULES_WLOCK(); 4900 error = pfr_add_addrs(&io->pfrio_table, pfras, 4901 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | 4902 PFR_FLAG_USERIOCTL); 4903 PF_RULES_WUNLOCK(); 4904 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4905 error = copyout(pfras, io->pfrio_buffer, totlen); 4906 free(pfras, M_TEMP); 4907 break; 4908 } 4909 4910 case DIOCRDELADDRS: { 4911 struct pfioc_table *io = (struct pfioc_table *)addr; 4912 struct pfr_addr *pfras; 4913 size_t totlen; 4914 4915 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4916 error = ENODEV; 4917 break; 4918 } 4919 if (io->pfrio_size < 0 || 4920 io->pfrio_size > pf_ioctl_maxcount || 4921 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4922 error = EINVAL; 4923 break; 4924 } 4925 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4926 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4927 M_TEMP, M_WAITOK); 4928 error = copyin(io->pfrio_buffer, pfras, totlen); 4929 if (error) { 4930 free(pfras, M_TEMP); 4931 break; 4932 } 4933 PF_RULES_WLOCK(); 4934 error = pfr_del_addrs(&io->pfrio_table, pfras, 4935 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | 4936 PFR_FLAG_USERIOCTL); 4937 PF_RULES_WUNLOCK(); 4938 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4939 error = copyout(pfras, io->pfrio_buffer, totlen); 4940 free(pfras, M_TEMP); 4941 break; 4942 } 4943 4944 case DIOCRSETADDRS: { 4945 struct pfioc_table *io = (struct pfioc_table *)addr; 4946 struct pfr_addr *pfras; 4947 size_t totlen, count; 4948 4949 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4950 error = ENODEV; 4951 break; 4952 } 4953 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { 4954 error = EINVAL; 4955 break; 4956 } 4957 count = max(io->pfrio_size, io->pfrio_size2); 4958 if (count > pf_ioctl_maxcount || 4959 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { 4960 error = EINVAL; 4961 break; 4962 } 4963 totlen = count * sizeof(struct pfr_addr); 4964 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP, 4965 M_WAITOK); 4966 error = copyin(io->pfrio_buffer, pfras, totlen); 4967 if (error) { 4968 free(pfras, M_TEMP); 4969 break; 4970 } 4971 PF_RULES_WLOCK(); 4972 error = pfr_set_addrs(&io->pfrio_table, pfras, 4973 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, 4974 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | 4975 PFR_FLAG_USERIOCTL, 0); 4976 PF_RULES_WUNLOCK(); 4977 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4978 error = copyout(pfras, io->pfrio_buffer, totlen); 4979 free(pfras, M_TEMP); 4980 break; 4981 } 4982 4983 case DIOCRGETADDRS: { 4984 struct pfioc_table *io = (struct pfioc_table *)addr; 4985 struct pfr_addr *pfras; 4986 size_t totlen; 4987 4988 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4989 error = ENODEV; 4990 break; 4991 } 4992 if (io->pfrio_size < 0 || 4993 io->pfrio_size > pf_ioctl_maxcount || 4994 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4995 error = EINVAL; 4996 break; 4997 } 4998 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4999 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5000 M_TEMP, M_WAITOK | M_ZERO); 5001 PF_RULES_RLOCK(); 5002 error = pfr_get_addrs(&io->pfrio_table, pfras, 5003 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5004 PF_RULES_RUNLOCK(); 5005 if (error == 0) 5006 error = copyout(pfras, io->pfrio_buffer, totlen); 5007 free(pfras, M_TEMP); 5008 break; 5009 } 5010 5011 case DIOCRGETASTATS: { 5012 struct pfioc_table *io = (struct pfioc_table *)addr; 5013 struct pfr_astats *pfrastats; 5014 size_t totlen; 5015 5016 if (io->pfrio_esize != sizeof(struct pfr_astats)) { 5017 error = ENODEV; 5018 break; 5019 } 5020 if (io->pfrio_size < 0 || 5021 io->pfrio_size > pf_ioctl_maxcount || 5022 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { 5023 error = EINVAL; 5024 break; 5025 } 5026 totlen = io->pfrio_size * sizeof(struct pfr_astats); 5027 pfrastats = mallocarray(io->pfrio_size, 5028 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO); 5029 PF_RULES_RLOCK(); 5030 error = pfr_get_astats(&io->pfrio_table, pfrastats, 5031 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5032 PF_RULES_RUNLOCK(); 5033 if (error == 0) 5034 error = copyout(pfrastats, io->pfrio_buffer, totlen); 5035 free(pfrastats, M_TEMP); 5036 break; 5037 } 5038 5039 case DIOCRCLRASTATS: { 5040 struct pfioc_table *io = (struct pfioc_table *)addr; 5041 struct pfr_addr *pfras; 5042 size_t totlen; 5043 5044 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5045 error = ENODEV; 5046 break; 5047 } 5048 if (io->pfrio_size < 0 || 5049 io->pfrio_size > pf_ioctl_maxcount || 5050 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5051 error = EINVAL; 5052 break; 5053 } 5054 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5055 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5056 M_TEMP, M_WAITOK); 5057 error = copyin(io->pfrio_buffer, pfras, totlen); 5058 if (error) { 5059 free(pfras, M_TEMP); 5060 break; 5061 } 5062 PF_RULES_WLOCK(); 5063 error = pfr_clr_astats(&io->pfrio_table, pfras, 5064 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | 5065 PFR_FLAG_USERIOCTL); 5066 PF_RULES_WUNLOCK(); 5067 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5068 error = copyout(pfras, io->pfrio_buffer, totlen); 5069 free(pfras, M_TEMP); 5070 break; 5071 } 5072 5073 case DIOCRTSTADDRS: { 5074 struct pfioc_table *io = (struct pfioc_table *)addr; 5075 struct pfr_addr *pfras; 5076 size_t totlen; 5077 5078 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5079 error = ENODEV; 5080 break; 5081 } 5082 if (io->pfrio_size < 0 || 5083 io->pfrio_size > pf_ioctl_maxcount || 5084 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5085 error = EINVAL; 5086 break; 5087 } 5088 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5089 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5090 M_TEMP, M_WAITOK); 5091 error = copyin(io->pfrio_buffer, pfras, totlen); 5092 if (error) { 5093 free(pfras, M_TEMP); 5094 break; 5095 } 5096 PF_RULES_RLOCK(); 5097 error = pfr_tst_addrs(&io->pfrio_table, pfras, 5098 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | 5099 PFR_FLAG_USERIOCTL); 5100 PF_RULES_RUNLOCK(); 5101 if (error == 0) 5102 error = copyout(pfras, io->pfrio_buffer, totlen); 5103 free(pfras, M_TEMP); 5104 break; 5105 } 5106 5107 case DIOCRINADEFINE: { 5108 struct pfioc_table *io = (struct pfioc_table *)addr; 5109 struct pfr_addr *pfras; 5110 size_t totlen; 5111 5112 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5113 error = ENODEV; 5114 break; 5115 } 5116 if (io->pfrio_size < 0 || 5117 io->pfrio_size > pf_ioctl_maxcount || 5118 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5119 error = EINVAL; 5120 break; 5121 } 5122 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5123 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5124 M_TEMP, M_WAITOK); 5125 error = copyin(io->pfrio_buffer, pfras, totlen); 5126 if (error) { 5127 free(pfras, M_TEMP); 5128 break; 5129 } 5130 PF_RULES_WLOCK(); 5131 error = pfr_ina_define(&io->pfrio_table, pfras, 5132 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, 5133 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5134 PF_RULES_WUNLOCK(); 5135 free(pfras, M_TEMP); 5136 break; 5137 } 5138 5139 case DIOCOSFPADD: { 5140 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5141 PF_RULES_WLOCK(); 5142 error = pf_osfp_add(io); 5143 PF_RULES_WUNLOCK(); 5144 break; 5145 } 5146 5147 case DIOCOSFPGET: { 5148 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5149 PF_RULES_RLOCK(); 5150 error = pf_osfp_get(io); 5151 PF_RULES_RUNLOCK(); 5152 break; 5153 } 5154 5155 case DIOCXBEGIN: { 5156 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5157 struct pfioc_trans_e *ioes, *ioe; 5158 size_t totlen; 5159 int i; 5160 5161 if (io->esize != sizeof(*ioe)) { 5162 error = ENODEV; 5163 break; 5164 } 5165 if (io->size < 0 || 5166 io->size > pf_ioctl_maxcount || 5167 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5168 error = EINVAL; 5169 break; 5170 } 5171 totlen = sizeof(struct pfioc_trans_e) * io->size; 5172 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5173 M_TEMP, M_WAITOK); 5174 error = copyin(io->array, ioes, totlen); 5175 if (error) { 5176 free(ioes, M_TEMP); 5177 break; 5178 } 5179 /* Ensure there's no more ethernet rules to clean up. */ 5180 epoch_drain_callbacks(net_epoch_preempt); 5181 PF_RULES_WLOCK(); 5182 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5183 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5184 switch (ioe->rs_num) { 5185 case PF_RULESET_ETH: 5186 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) { 5187 PF_RULES_WUNLOCK(); 5188 free(ioes, M_TEMP); 5189 goto fail; 5190 } 5191 break; 5192 #ifdef ALTQ 5193 case PF_RULESET_ALTQ: 5194 if (ioe->anchor[0]) { 5195 PF_RULES_WUNLOCK(); 5196 free(ioes, M_TEMP); 5197 error = EINVAL; 5198 goto fail; 5199 } 5200 if ((error = pf_begin_altq(&ioe->ticket))) { 5201 PF_RULES_WUNLOCK(); 5202 free(ioes, M_TEMP); 5203 goto fail; 5204 } 5205 break; 5206 #endif /* ALTQ */ 5207 case PF_RULESET_TABLE: 5208 { 5209 struct pfr_table table; 5210 5211 bzero(&table, sizeof(table)); 5212 strlcpy(table.pfrt_anchor, ioe->anchor, 5213 sizeof(table.pfrt_anchor)); 5214 if ((error = pfr_ina_begin(&table, 5215 &ioe->ticket, NULL, 0))) { 5216 PF_RULES_WUNLOCK(); 5217 free(ioes, M_TEMP); 5218 goto fail; 5219 } 5220 break; 5221 } 5222 default: 5223 if ((error = pf_begin_rules(&ioe->ticket, 5224 ioe->rs_num, ioe->anchor))) { 5225 PF_RULES_WUNLOCK(); 5226 free(ioes, M_TEMP); 5227 goto fail; 5228 } 5229 break; 5230 } 5231 } 5232 PF_RULES_WUNLOCK(); 5233 error = copyout(ioes, io->array, totlen); 5234 free(ioes, M_TEMP); 5235 break; 5236 } 5237 5238 case DIOCXROLLBACK: { 5239 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5240 struct pfioc_trans_e *ioe, *ioes; 5241 size_t totlen; 5242 int i; 5243 5244 if (io->esize != sizeof(*ioe)) { 5245 error = ENODEV; 5246 break; 5247 } 5248 if (io->size < 0 || 5249 io->size > pf_ioctl_maxcount || 5250 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5251 error = EINVAL; 5252 break; 5253 } 5254 totlen = sizeof(struct pfioc_trans_e) * io->size; 5255 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5256 M_TEMP, M_WAITOK); 5257 error = copyin(io->array, ioes, totlen); 5258 if (error) { 5259 free(ioes, M_TEMP); 5260 break; 5261 } 5262 PF_RULES_WLOCK(); 5263 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5264 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5265 switch (ioe->rs_num) { 5266 case PF_RULESET_ETH: 5267 if ((error = pf_rollback_eth(ioe->ticket, 5268 ioe->anchor))) { 5269 PF_RULES_WUNLOCK(); 5270 free(ioes, M_TEMP); 5271 goto fail; /* really bad */ 5272 } 5273 break; 5274 #ifdef ALTQ 5275 case PF_RULESET_ALTQ: 5276 if (ioe->anchor[0]) { 5277 PF_RULES_WUNLOCK(); 5278 free(ioes, M_TEMP); 5279 error = EINVAL; 5280 goto fail; 5281 } 5282 if ((error = pf_rollback_altq(ioe->ticket))) { 5283 PF_RULES_WUNLOCK(); 5284 free(ioes, M_TEMP); 5285 goto fail; /* really bad */ 5286 } 5287 break; 5288 #endif /* ALTQ */ 5289 case PF_RULESET_TABLE: 5290 { 5291 struct pfr_table table; 5292 5293 bzero(&table, sizeof(table)); 5294 strlcpy(table.pfrt_anchor, ioe->anchor, 5295 sizeof(table.pfrt_anchor)); 5296 if ((error = pfr_ina_rollback(&table, 5297 ioe->ticket, NULL, 0))) { 5298 PF_RULES_WUNLOCK(); 5299 free(ioes, M_TEMP); 5300 goto fail; /* really bad */ 5301 } 5302 break; 5303 } 5304 default: 5305 if ((error = pf_rollback_rules(ioe->ticket, 5306 ioe->rs_num, ioe->anchor))) { 5307 PF_RULES_WUNLOCK(); 5308 free(ioes, M_TEMP); 5309 goto fail; /* really bad */ 5310 } 5311 break; 5312 } 5313 } 5314 PF_RULES_WUNLOCK(); 5315 free(ioes, M_TEMP); 5316 break; 5317 } 5318 5319 case DIOCXCOMMIT: { 5320 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5321 struct pfioc_trans_e *ioe, *ioes; 5322 struct pf_kruleset *rs; 5323 struct pf_keth_ruleset *ers; 5324 size_t totlen; 5325 int i; 5326 5327 if (io->esize != sizeof(*ioe)) { 5328 error = ENODEV; 5329 break; 5330 } 5331 5332 if (io->size < 0 || 5333 io->size > pf_ioctl_maxcount || 5334 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5335 error = EINVAL; 5336 break; 5337 } 5338 5339 totlen = sizeof(struct pfioc_trans_e) * io->size; 5340 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5341 M_TEMP, M_WAITOK); 5342 error = copyin(io->array, ioes, totlen); 5343 if (error) { 5344 free(ioes, M_TEMP); 5345 break; 5346 } 5347 PF_RULES_WLOCK(); 5348 /* First makes sure everything will succeed. */ 5349 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5350 ioe->anchor[sizeof(ioe->anchor) - 1] = 0; 5351 switch (ioe->rs_num) { 5352 case PF_RULESET_ETH: 5353 ers = pf_find_keth_ruleset(ioe->anchor); 5354 if (ers == NULL || ioe->ticket == 0 || 5355 ioe->ticket != ers->inactive.ticket) { 5356 PF_RULES_WUNLOCK(); 5357 free(ioes, M_TEMP); 5358 error = EINVAL; 5359 goto fail; 5360 } 5361 break; 5362 #ifdef ALTQ 5363 case PF_RULESET_ALTQ: 5364 if (ioe->anchor[0]) { 5365 PF_RULES_WUNLOCK(); 5366 free(ioes, M_TEMP); 5367 error = EINVAL; 5368 goto fail; 5369 } 5370 if (!V_altqs_inactive_open || ioe->ticket != 5371 V_ticket_altqs_inactive) { 5372 PF_RULES_WUNLOCK(); 5373 free(ioes, M_TEMP); 5374 error = EBUSY; 5375 goto fail; 5376 } 5377 break; 5378 #endif /* ALTQ */ 5379 case PF_RULESET_TABLE: 5380 rs = pf_find_kruleset(ioe->anchor); 5381 if (rs == NULL || !rs->topen || ioe->ticket != 5382 rs->tticket) { 5383 PF_RULES_WUNLOCK(); 5384 free(ioes, M_TEMP); 5385 error = EBUSY; 5386 goto fail; 5387 } 5388 break; 5389 default: 5390 if (ioe->rs_num < 0 || ioe->rs_num >= 5391 PF_RULESET_MAX) { 5392 PF_RULES_WUNLOCK(); 5393 free(ioes, M_TEMP); 5394 error = EINVAL; 5395 goto fail; 5396 } 5397 rs = pf_find_kruleset(ioe->anchor); 5398 if (rs == NULL || 5399 !rs->rules[ioe->rs_num].inactive.open || 5400 rs->rules[ioe->rs_num].inactive.ticket != 5401 ioe->ticket) { 5402 PF_RULES_WUNLOCK(); 5403 free(ioes, M_TEMP); 5404 error = EBUSY; 5405 goto fail; 5406 } 5407 break; 5408 } 5409 } 5410 /* Now do the commit - no errors should happen here. */ 5411 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5412 switch (ioe->rs_num) { 5413 case PF_RULESET_ETH: 5414 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) { 5415 PF_RULES_WUNLOCK(); 5416 free(ioes, M_TEMP); 5417 goto fail; /* really bad */ 5418 } 5419 break; 5420 #ifdef ALTQ 5421 case PF_RULESET_ALTQ: 5422 if ((error = pf_commit_altq(ioe->ticket))) { 5423 PF_RULES_WUNLOCK(); 5424 free(ioes, M_TEMP); 5425 goto fail; /* really bad */ 5426 } 5427 break; 5428 #endif /* ALTQ */ 5429 case PF_RULESET_TABLE: 5430 { 5431 struct pfr_table table; 5432 5433 bzero(&table, sizeof(table)); 5434 (void)strlcpy(table.pfrt_anchor, ioe->anchor, 5435 sizeof(table.pfrt_anchor)); 5436 if ((error = pfr_ina_commit(&table, 5437 ioe->ticket, NULL, NULL, 0))) { 5438 PF_RULES_WUNLOCK(); 5439 free(ioes, M_TEMP); 5440 goto fail; /* really bad */ 5441 } 5442 break; 5443 } 5444 default: 5445 if ((error = pf_commit_rules(ioe->ticket, 5446 ioe->rs_num, ioe->anchor))) { 5447 PF_RULES_WUNLOCK(); 5448 free(ioes, M_TEMP); 5449 goto fail; /* really bad */ 5450 } 5451 break; 5452 } 5453 } 5454 PF_RULES_WUNLOCK(); 5455 5456 /* Only hook into EtherNet taffic if we've got rules for it. */ 5457 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 5458 hook_pf_eth(); 5459 else 5460 dehook_pf_eth(); 5461 5462 free(ioes, M_TEMP); 5463 break; 5464 } 5465 5466 case DIOCGETSRCNODES: { 5467 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; 5468 struct pf_srchash *sh; 5469 struct pf_ksrc_node *n; 5470 struct pf_src_node *p, *pstore; 5471 uint32_t i, nr = 0; 5472 5473 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; 5474 i++, sh++) { 5475 PF_HASHROW_LOCK(sh); 5476 LIST_FOREACH(n, &sh->nodes, entry) 5477 nr++; 5478 PF_HASHROW_UNLOCK(sh); 5479 } 5480 5481 psn->psn_len = min(psn->psn_len, 5482 sizeof(struct pf_src_node) * nr); 5483 5484 if (psn->psn_len == 0) { 5485 psn->psn_len = sizeof(struct pf_src_node) * nr; 5486 break; 5487 } 5488 5489 nr = 0; 5490 5491 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO); 5492 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; 5493 i++, sh++) { 5494 PF_HASHROW_LOCK(sh); 5495 LIST_FOREACH(n, &sh->nodes, entry) { 5496 5497 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) 5498 break; 5499 5500 pf_src_node_copy(n, p); 5501 5502 p++; 5503 nr++; 5504 } 5505 PF_HASHROW_UNLOCK(sh); 5506 } 5507 error = copyout(pstore, psn->psn_src_nodes, 5508 sizeof(struct pf_src_node) * nr); 5509 if (error) { 5510 free(pstore, M_TEMP); 5511 break; 5512 } 5513 psn->psn_len = sizeof(struct pf_src_node) * nr; 5514 free(pstore, M_TEMP); 5515 break; 5516 } 5517 5518 case DIOCCLRSRCNODES: { 5519 pf_clear_srcnodes(NULL); 5520 pf_purge_expired_src_nodes(); 5521 break; 5522 } 5523 5524 case DIOCKILLSRCNODES: 5525 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); 5526 break; 5527 5528 #ifdef COMPAT_FREEBSD13 5529 case DIOCKEEPCOUNTERS_FREEBSD13: 5530 #endif 5531 case DIOCKEEPCOUNTERS: 5532 error = pf_keepcounters((struct pfioc_nv *)addr); 5533 break; 5534 5535 case DIOCGETSYNCOOKIES: 5536 error = pf_get_syncookies((struct pfioc_nv *)addr); 5537 break; 5538 5539 case DIOCSETSYNCOOKIES: 5540 error = pf_set_syncookies((struct pfioc_nv *)addr); 5541 break; 5542 5543 case DIOCSETHOSTID: { 5544 u_int32_t *hostid = (u_int32_t *)addr; 5545 5546 PF_RULES_WLOCK(); 5547 if (*hostid == 0) 5548 V_pf_status.hostid = arc4random(); 5549 else 5550 V_pf_status.hostid = *hostid; 5551 PF_RULES_WUNLOCK(); 5552 break; 5553 } 5554 5555 case DIOCOSFPFLUSH: 5556 PF_RULES_WLOCK(); 5557 pf_osfp_flush(); 5558 PF_RULES_WUNLOCK(); 5559 break; 5560 5561 case DIOCIGETIFACES: { 5562 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5563 struct pfi_kif *ifstore; 5564 size_t bufsiz; 5565 5566 if (io->pfiio_esize != sizeof(struct pfi_kif)) { 5567 error = ENODEV; 5568 break; 5569 } 5570 5571 if (io->pfiio_size < 0 || 5572 io->pfiio_size > pf_ioctl_maxcount || 5573 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { 5574 error = EINVAL; 5575 break; 5576 } 5577 5578 bufsiz = io->pfiio_size * sizeof(struct pfi_kif); 5579 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif), 5580 M_TEMP, M_WAITOK | M_ZERO); 5581 5582 PF_RULES_RLOCK(); 5583 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); 5584 PF_RULES_RUNLOCK(); 5585 error = copyout(ifstore, io->pfiio_buffer, bufsiz); 5586 free(ifstore, M_TEMP); 5587 break; 5588 } 5589 5590 case DIOCSETIFFLAG: { 5591 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5592 5593 PF_RULES_WLOCK(); 5594 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags); 5595 PF_RULES_WUNLOCK(); 5596 break; 5597 } 5598 5599 case DIOCCLRIFFLAG: { 5600 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5601 5602 PF_RULES_WLOCK(); 5603 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); 5604 PF_RULES_WUNLOCK(); 5605 break; 5606 } 5607 5608 default: 5609 error = ENODEV; 5610 break; 5611 } 5612 fail: 5613 if (sx_xlocked(&pf_ioctl_lock)) 5614 sx_xunlock(&pf_ioctl_lock); 5615 CURVNET_RESTORE(); 5616 5617 #undef ERROUT_IOCTL 5618 5619 return (error); 5620 } 5621 5622 void 5623 pfsync_state_export(struct pfsync_state *sp, struct pf_kstate *st) 5624 { 5625 bzero(sp, sizeof(struct pfsync_state)); 5626 5627 /* copy from state key */ 5628 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5629 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5630 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5631 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5632 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5633 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5634 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5635 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5636 sp->proto = st->key[PF_SK_WIRE]->proto; 5637 sp->af = st->key[PF_SK_WIRE]->af; 5638 5639 /* copy from state */ 5640 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5641 bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr)); 5642 sp->creation = htonl(time_uptime - st->creation); 5643 sp->expire = pf_state_expires(st); 5644 if (sp->expire <= time_uptime) 5645 sp->expire = htonl(0); 5646 else 5647 sp->expire = htonl(sp->expire - time_uptime); 5648 5649 sp->direction = st->direction; 5650 sp->log = st->log; 5651 sp->timeout = st->timeout; 5652 sp->state_flags = st->state_flags; 5653 if (st->src_node) 5654 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5655 if (st->nat_src_node) 5656 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5657 5658 sp->id = st->id; 5659 sp->creatorid = st->creatorid; 5660 pf_state_peer_hton(&st->src, &sp->src); 5661 pf_state_peer_hton(&st->dst, &sp->dst); 5662 5663 if (st->rule.ptr == NULL) 5664 sp->rule = htonl(-1); 5665 else 5666 sp->rule = htonl(st->rule.ptr->nr); 5667 if (st->anchor.ptr == NULL) 5668 sp->anchor = htonl(-1); 5669 else 5670 sp->anchor = htonl(st->anchor.ptr->nr); 5671 if (st->nat_rule.ptr == NULL) 5672 sp->nat_rule = htonl(-1); 5673 else 5674 sp->nat_rule = htonl(st->nat_rule.ptr->nr); 5675 5676 pf_state_counter_hton(st->packets[0], sp->packets[0]); 5677 pf_state_counter_hton(st->packets[1], sp->packets[1]); 5678 pf_state_counter_hton(st->bytes[0], sp->bytes[0]); 5679 pf_state_counter_hton(st->bytes[1], sp->bytes[1]); 5680 } 5681 5682 void 5683 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st) 5684 { 5685 bzero(sp, sizeof(*sp)); 5686 5687 sp->version = PF_STATE_VERSION; 5688 5689 /* copy from state key */ 5690 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5691 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5692 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5693 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5694 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5695 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5696 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5697 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5698 sp->proto = st->key[PF_SK_WIRE]->proto; 5699 sp->af = st->key[PF_SK_WIRE]->af; 5700 5701 /* copy from state */ 5702 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5703 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name, 5704 sizeof(sp->orig_ifname)); 5705 bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr)); 5706 sp->creation = htonl(time_uptime - st->creation); 5707 sp->expire = pf_state_expires(st); 5708 if (sp->expire <= time_uptime) 5709 sp->expire = htonl(0); 5710 else 5711 sp->expire = htonl(sp->expire - time_uptime); 5712 5713 sp->direction = st->direction; 5714 sp->log = st->log; 5715 sp->timeout = st->timeout; 5716 sp->state_flags = st->state_flags; 5717 if (st->src_node) 5718 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5719 if (st->nat_src_node) 5720 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5721 5722 sp->id = st->id; 5723 sp->creatorid = st->creatorid; 5724 pf_state_peer_hton(&st->src, &sp->src); 5725 pf_state_peer_hton(&st->dst, &sp->dst); 5726 5727 if (st->rule.ptr == NULL) 5728 sp->rule = htonl(-1); 5729 else 5730 sp->rule = htonl(st->rule.ptr->nr); 5731 if (st->anchor.ptr == NULL) 5732 sp->anchor = htonl(-1); 5733 else 5734 sp->anchor = htonl(st->anchor.ptr->nr); 5735 if (st->nat_rule.ptr == NULL) 5736 sp->nat_rule = htonl(-1); 5737 else 5738 sp->nat_rule = htonl(st->nat_rule.ptr->nr); 5739 5740 sp->packets[0] = st->packets[0]; 5741 sp->packets[1] = st->packets[1]; 5742 sp->bytes[0] = st->bytes[0]; 5743 sp->bytes[1] = st->bytes[1]; 5744 } 5745 5746 static void 5747 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 5748 { 5749 struct pfr_ktable *kt; 5750 5751 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); 5752 5753 kt = aw->p.tbl; 5754 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 5755 kt = kt->pfrkt_root; 5756 aw->p.tbl = NULL; 5757 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 5758 kt->pfrkt_cnt : -1; 5759 } 5760 5761 static int 5762 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters, 5763 size_t number, char **names) 5764 { 5765 nvlist_t *nvc; 5766 5767 nvc = nvlist_create(0); 5768 if (nvc == NULL) 5769 return (ENOMEM); 5770 5771 for (int i = 0; i < number; i++) { 5772 nvlist_append_number_array(nvc, "counters", 5773 counter_u64_fetch(counters[i])); 5774 nvlist_append_string_array(nvc, "names", 5775 names[i]); 5776 nvlist_append_number_array(nvc, "ids", 5777 i); 5778 } 5779 nvlist_add_nvlist(nvl, name, nvc); 5780 nvlist_destroy(nvc); 5781 5782 return (0); 5783 } 5784 5785 static int 5786 pf_getstatus(struct pfioc_nv *nv) 5787 { 5788 nvlist_t *nvl = NULL, *nvc = NULL; 5789 void *nvlpacked = NULL; 5790 int error; 5791 struct pf_status s; 5792 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES; 5793 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES; 5794 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES; 5795 PF_RULES_RLOCK_TRACKER; 5796 5797 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 5798 5799 PF_RULES_RLOCK(); 5800 5801 nvl = nvlist_create(0); 5802 if (nvl == NULL) 5803 ERROUT(ENOMEM); 5804 5805 nvlist_add_bool(nvl, "running", V_pf_status.running); 5806 nvlist_add_number(nvl, "since", V_pf_status.since); 5807 nvlist_add_number(nvl, "debug", V_pf_status.debug); 5808 nvlist_add_number(nvl, "hostid", V_pf_status.hostid); 5809 nvlist_add_number(nvl, "states", V_pf_status.states); 5810 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes); 5811 5812 /* counters */ 5813 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters, 5814 PFRES_MAX, pf_reasons); 5815 if (error != 0) 5816 ERROUT(error); 5817 5818 /* lcounters */ 5819 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters, 5820 KLCNT_MAX, pf_lcounter); 5821 if (error != 0) 5822 ERROUT(error); 5823 5824 /* fcounters */ 5825 nvc = nvlist_create(0); 5826 if (nvc == NULL) 5827 ERROUT(ENOMEM); 5828 5829 for (int i = 0; i < FCNT_MAX; i++) { 5830 nvlist_append_number_array(nvc, "counters", 5831 pf_counter_u64_fetch(&V_pf_status.fcounters[i])); 5832 nvlist_append_string_array(nvc, "names", 5833 pf_fcounter[i]); 5834 nvlist_append_number_array(nvc, "ids", 5835 i); 5836 } 5837 nvlist_add_nvlist(nvl, "fcounters", nvc); 5838 nvlist_destroy(nvc); 5839 nvc = NULL; 5840 5841 /* scounters */ 5842 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters, 5843 SCNT_MAX, pf_fcounter); 5844 if (error != 0) 5845 ERROUT(error); 5846 5847 nvlist_add_string(nvl, "ifname", V_pf_status.ifname); 5848 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum, 5849 PF_MD5_DIGEST_LENGTH); 5850 5851 pfi_update_status(V_pf_status.ifname, &s); 5852 5853 /* pcounters / bcounters */ 5854 for (int i = 0; i < 2; i++) { 5855 for (int j = 0; j < 2; j++) { 5856 for (int k = 0; k < 2; k++) { 5857 nvlist_append_number_array(nvl, "pcounters", 5858 s.pcounters[i][j][k]); 5859 } 5860 nvlist_append_number_array(nvl, "bcounters", 5861 s.bcounters[i][j]); 5862 } 5863 } 5864 5865 nvlpacked = nvlist_pack(nvl, &nv->len); 5866 if (nvlpacked == NULL) 5867 ERROUT(ENOMEM); 5868 5869 if (nv->size == 0) 5870 ERROUT(0); 5871 else if (nv->size < nv->len) 5872 ERROUT(ENOSPC); 5873 5874 PF_RULES_RUNLOCK(); 5875 error = copyout(nvlpacked, nv->data, nv->len); 5876 goto done; 5877 5878 #undef ERROUT 5879 errout: 5880 PF_RULES_RUNLOCK(); 5881 done: 5882 free(nvlpacked, M_NVLIST); 5883 nvlist_destroy(nvc); 5884 nvlist_destroy(nvl); 5885 5886 return (error); 5887 } 5888 5889 /* 5890 * XXX - Check for version missmatch!!! 5891 */ 5892 static void 5893 pf_clear_all_states(void) 5894 { 5895 struct pf_kstate *s; 5896 u_int i; 5897 5898 for (i = 0; i <= pf_hashmask; i++) { 5899 struct pf_idhash *ih = &V_pf_idhash[i]; 5900 relock: 5901 PF_HASHROW_LOCK(ih); 5902 LIST_FOREACH(s, &ih->states, entry) { 5903 s->timeout = PFTM_PURGE; 5904 /* Don't send out individual delete messages. */ 5905 s->state_flags |= PFSTATE_NOSYNC; 5906 pf_unlink_state(s); 5907 goto relock; 5908 } 5909 PF_HASHROW_UNLOCK(ih); 5910 } 5911 } 5912 5913 static int 5914 pf_clear_tables(void) 5915 { 5916 struct pfioc_table io; 5917 int error; 5918 5919 bzero(&io, sizeof(io)); 5920 5921 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, 5922 io.pfrio_flags); 5923 5924 return (error); 5925 } 5926 5927 static void 5928 pf_clear_srcnodes(struct pf_ksrc_node *n) 5929 { 5930 struct pf_kstate *s; 5931 int i; 5932 5933 for (i = 0; i <= pf_hashmask; i++) { 5934 struct pf_idhash *ih = &V_pf_idhash[i]; 5935 5936 PF_HASHROW_LOCK(ih); 5937 LIST_FOREACH(s, &ih->states, entry) { 5938 if (n == NULL || n == s->src_node) 5939 s->src_node = NULL; 5940 if (n == NULL || n == s->nat_src_node) 5941 s->nat_src_node = NULL; 5942 } 5943 PF_HASHROW_UNLOCK(ih); 5944 } 5945 5946 if (n == NULL) { 5947 struct pf_srchash *sh; 5948 5949 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; 5950 i++, sh++) { 5951 PF_HASHROW_LOCK(sh); 5952 LIST_FOREACH(n, &sh->nodes, entry) { 5953 n->expire = 1; 5954 n->states = 0; 5955 } 5956 PF_HASHROW_UNLOCK(sh); 5957 } 5958 } else { 5959 /* XXX: hash slot should already be locked here. */ 5960 n->expire = 1; 5961 n->states = 0; 5962 } 5963 } 5964 5965 static void 5966 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) 5967 { 5968 struct pf_ksrc_node_list kill; 5969 5970 LIST_INIT(&kill); 5971 for (int i = 0; i <= pf_srchashmask; i++) { 5972 struct pf_srchash *sh = &V_pf_srchash[i]; 5973 struct pf_ksrc_node *sn, *tmp; 5974 5975 PF_HASHROW_LOCK(sh); 5976 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) 5977 if (PF_MATCHA(psnk->psnk_src.neg, 5978 &psnk->psnk_src.addr.v.a.addr, 5979 &psnk->psnk_src.addr.v.a.mask, 5980 &sn->addr, sn->af) && 5981 PF_MATCHA(psnk->psnk_dst.neg, 5982 &psnk->psnk_dst.addr.v.a.addr, 5983 &psnk->psnk_dst.addr.v.a.mask, 5984 &sn->raddr, sn->af)) { 5985 pf_unlink_src_node(sn); 5986 LIST_INSERT_HEAD(&kill, sn, entry); 5987 sn->expire = 1; 5988 } 5989 PF_HASHROW_UNLOCK(sh); 5990 } 5991 5992 for (int i = 0; i <= pf_hashmask; i++) { 5993 struct pf_idhash *ih = &V_pf_idhash[i]; 5994 struct pf_kstate *s; 5995 5996 PF_HASHROW_LOCK(ih); 5997 LIST_FOREACH(s, &ih->states, entry) { 5998 if (s->src_node && s->src_node->expire == 1) 5999 s->src_node = NULL; 6000 if (s->nat_src_node && s->nat_src_node->expire == 1) 6001 s->nat_src_node = NULL; 6002 } 6003 PF_HASHROW_UNLOCK(ih); 6004 } 6005 6006 psnk->psnk_killed = pf_free_src_nodes(&kill); 6007 } 6008 6009 static int 6010 pf_keepcounters(struct pfioc_nv *nv) 6011 { 6012 nvlist_t *nvl = NULL; 6013 void *nvlpacked = NULL; 6014 int error = 0; 6015 6016 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6017 6018 if (nv->len > pf_ioctl_maxcount) 6019 ERROUT(ENOMEM); 6020 6021 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6022 if (nvlpacked == NULL) 6023 ERROUT(ENOMEM); 6024 6025 error = copyin(nv->data, nvlpacked, nv->len); 6026 if (error) 6027 ERROUT(error); 6028 6029 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6030 if (nvl == NULL) 6031 ERROUT(EBADMSG); 6032 6033 if (! nvlist_exists_bool(nvl, "keep_counters")) 6034 ERROUT(EBADMSG); 6035 6036 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters"); 6037 6038 on_error: 6039 nvlist_destroy(nvl); 6040 free(nvlpacked, M_NVLIST); 6041 return (error); 6042 } 6043 6044 static unsigned int 6045 pf_clear_states(const struct pf_kstate_kill *kill) 6046 { 6047 struct pf_state_key_cmp match_key; 6048 struct pf_kstate *s; 6049 struct pfi_kkif *kif; 6050 int idx; 6051 unsigned int killed = 0, dir; 6052 6053 for (unsigned int i = 0; i <= pf_hashmask; i++) { 6054 struct pf_idhash *ih = &V_pf_idhash[i]; 6055 6056 relock_DIOCCLRSTATES: 6057 PF_HASHROW_LOCK(ih); 6058 LIST_FOREACH(s, &ih->states, entry) { 6059 /* For floating states look at the original kif. */ 6060 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 6061 6062 if (kill->psk_ifname[0] && 6063 strcmp(kill->psk_ifname, 6064 kif->pfik_name)) 6065 continue; 6066 6067 if (kill->psk_kill_match) { 6068 bzero(&match_key, sizeof(match_key)); 6069 6070 if (s->direction == PF_OUT) { 6071 dir = PF_IN; 6072 idx = PF_SK_STACK; 6073 } else { 6074 dir = PF_OUT; 6075 idx = PF_SK_WIRE; 6076 } 6077 6078 match_key.af = s->key[idx]->af; 6079 match_key.proto = s->key[idx]->proto; 6080 PF_ACPY(&match_key.addr[0], 6081 &s->key[idx]->addr[1], match_key.af); 6082 match_key.port[0] = s->key[idx]->port[1]; 6083 PF_ACPY(&match_key.addr[1], 6084 &s->key[idx]->addr[0], match_key.af); 6085 match_key.port[1] = s->key[idx]->port[0]; 6086 } 6087 6088 /* 6089 * Don't send out individual 6090 * delete messages. 6091 */ 6092 s->state_flags |= PFSTATE_NOSYNC; 6093 pf_unlink_state(s); 6094 killed++; 6095 6096 if (kill->psk_kill_match) 6097 killed += pf_kill_matching_state(&match_key, 6098 dir); 6099 6100 goto relock_DIOCCLRSTATES; 6101 } 6102 PF_HASHROW_UNLOCK(ih); 6103 } 6104 6105 if (V_pfsync_clear_states_ptr != NULL) 6106 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname); 6107 6108 return (killed); 6109 } 6110 6111 static void 6112 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed) 6113 { 6114 struct pf_kstate *s; 6115 6116 if (kill->psk_pfcmp.id) { 6117 if (kill->psk_pfcmp.creatorid == 0) 6118 kill->psk_pfcmp.creatorid = V_pf_status.hostid; 6119 if ((s = pf_find_state_byid(kill->psk_pfcmp.id, 6120 kill->psk_pfcmp.creatorid))) { 6121 pf_unlink_state(s); 6122 *killed = 1; 6123 } 6124 return; 6125 } 6126 6127 for (unsigned int i = 0; i <= pf_hashmask; i++) 6128 *killed += pf_killstates_row(kill, &V_pf_idhash[i]); 6129 6130 return; 6131 } 6132 6133 static int 6134 pf_killstates_nv(struct pfioc_nv *nv) 6135 { 6136 struct pf_kstate_kill kill; 6137 nvlist_t *nvl = NULL; 6138 void *nvlpacked = NULL; 6139 int error = 0; 6140 unsigned int killed = 0; 6141 6142 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6143 6144 if (nv->len > pf_ioctl_maxcount) 6145 ERROUT(ENOMEM); 6146 6147 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6148 if (nvlpacked == NULL) 6149 ERROUT(ENOMEM); 6150 6151 error = copyin(nv->data, nvlpacked, nv->len); 6152 if (error) 6153 ERROUT(error); 6154 6155 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6156 if (nvl == NULL) 6157 ERROUT(EBADMSG); 6158 6159 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6160 if (error) 6161 ERROUT(error); 6162 6163 pf_killstates(&kill, &killed); 6164 6165 free(nvlpacked, M_NVLIST); 6166 nvlpacked = NULL; 6167 nvlist_destroy(nvl); 6168 nvl = nvlist_create(0); 6169 if (nvl == NULL) 6170 ERROUT(ENOMEM); 6171 6172 nvlist_add_number(nvl, "killed", killed); 6173 6174 nvlpacked = nvlist_pack(nvl, &nv->len); 6175 if (nvlpacked == NULL) 6176 ERROUT(ENOMEM); 6177 6178 if (nv->size == 0) 6179 ERROUT(0); 6180 else if (nv->size < nv->len) 6181 ERROUT(ENOSPC); 6182 6183 error = copyout(nvlpacked, nv->data, nv->len); 6184 6185 on_error: 6186 nvlist_destroy(nvl); 6187 free(nvlpacked, M_NVLIST); 6188 return (error); 6189 } 6190 6191 static int 6192 pf_clearstates_nv(struct pfioc_nv *nv) 6193 { 6194 struct pf_kstate_kill kill; 6195 nvlist_t *nvl = NULL; 6196 void *nvlpacked = NULL; 6197 int error = 0; 6198 unsigned int killed; 6199 6200 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6201 6202 if (nv->len > pf_ioctl_maxcount) 6203 ERROUT(ENOMEM); 6204 6205 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6206 if (nvlpacked == NULL) 6207 ERROUT(ENOMEM); 6208 6209 error = copyin(nv->data, nvlpacked, nv->len); 6210 if (error) 6211 ERROUT(error); 6212 6213 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6214 if (nvl == NULL) 6215 ERROUT(EBADMSG); 6216 6217 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6218 if (error) 6219 ERROUT(error); 6220 6221 killed = pf_clear_states(&kill); 6222 6223 free(nvlpacked, M_NVLIST); 6224 nvlpacked = NULL; 6225 nvlist_destroy(nvl); 6226 nvl = nvlist_create(0); 6227 if (nvl == NULL) 6228 ERROUT(ENOMEM); 6229 6230 nvlist_add_number(nvl, "killed", killed); 6231 6232 nvlpacked = nvlist_pack(nvl, &nv->len); 6233 if (nvlpacked == NULL) 6234 ERROUT(ENOMEM); 6235 6236 if (nv->size == 0) 6237 ERROUT(0); 6238 else if (nv->size < nv->len) 6239 ERROUT(ENOSPC); 6240 6241 error = copyout(nvlpacked, nv->data, nv->len); 6242 6243 #undef ERROUT 6244 on_error: 6245 nvlist_destroy(nvl); 6246 free(nvlpacked, M_NVLIST); 6247 return (error); 6248 } 6249 6250 static int 6251 pf_getstate(struct pfioc_nv *nv) 6252 { 6253 nvlist_t *nvl = NULL, *nvls; 6254 void *nvlpacked = NULL; 6255 struct pf_kstate *s = NULL; 6256 int error = 0; 6257 uint64_t id, creatorid; 6258 6259 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6260 6261 if (nv->len > pf_ioctl_maxcount) 6262 ERROUT(ENOMEM); 6263 6264 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6265 if (nvlpacked == NULL) 6266 ERROUT(ENOMEM); 6267 6268 error = copyin(nv->data, nvlpacked, nv->len); 6269 if (error) 6270 ERROUT(error); 6271 6272 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6273 if (nvl == NULL) 6274 ERROUT(EBADMSG); 6275 6276 PFNV_CHK(pf_nvuint64(nvl, "id", &id)); 6277 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid)); 6278 6279 s = pf_find_state_byid(id, creatorid); 6280 if (s == NULL) 6281 ERROUT(ENOENT); 6282 6283 free(nvlpacked, M_NVLIST); 6284 nvlpacked = NULL; 6285 nvlist_destroy(nvl); 6286 nvl = nvlist_create(0); 6287 if (nvl == NULL) 6288 ERROUT(ENOMEM); 6289 6290 nvls = pf_state_to_nvstate(s); 6291 if (nvls == NULL) 6292 ERROUT(ENOMEM); 6293 6294 nvlist_add_nvlist(nvl, "state", nvls); 6295 nvlist_destroy(nvls); 6296 6297 nvlpacked = nvlist_pack(nvl, &nv->len); 6298 if (nvlpacked == NULL) 6299 ERROUT(ENOMEM); 6300 6301 if (nv->size == 0) 6302 ERROUT(0); 6303 else if (nv->size < nv->len) 6304 ERROUT(ENOSPC); 6305 6306 error = copyout(nvlpacked, nv->data, nv->len); 6307 6308 #undef ERROUT 6309 errout: 6310 if (s != NULL) 6311 PF_STATE_UNLOCK(s); 6312 free(nvlpacked, M_NVLIST); 6313 nvlist_destroy(nvl); 6314 return (error); 6315 } 6316 6317 /* 6318 * XXX - Check for version missmatch!!! 6319 */ 6320 6321 /* 6322 * Duplicate pfctl -Fa operation to get rid of as much as we can. 6323 */ 6324 static int 6325 shutdown_pf(void) 6326 { 6327 int error = 0; 6328 u_int32_t t[5]; 6329 char nn = '\0'; 6330 6331 do { 6332 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) 6333 != 0) { 6334 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n")); 6335 break; 6336 } 6337 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) 6338 != 0) { 6339 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n")); 6340 break; /* XXX: rollback? */ 6341 } 6342 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) 6343 != 0) { 6344 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n")); 6345 break; /* XXX: rollback? */ 6346 } 6347 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) 6348 != 0) { 6349 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n")); 6350 break; /* XXX: rollback? */ 6351 } 6352 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) 6353 != 0) { 6354 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n")); 6355 break; /* XXX: rollback? */ 6356 } 6357 6358 /* XXX: these should always succeed here */ 6359 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn); 6360 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn); 6361 pf_commit_rules(t[2], PF_RULESET_NAT, &nn); 6362 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn); 6363 pf_commit_rules(t[4], PF_RULESET_RDR, &nn); 6364 6365 if ((error = pf_clear_tables()) != 0) 6366 break; 6367 6368 if ((error = pf_begin_eth(&t[0], &nn)) != 0) { 6369 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n")); 6370 break; 6371 } 6372 pf_commit_eth(t[0], &nn); 6373 6374 #ifdef ALTQ 6375 if ((error = pf_begin_altq(&t[0])) != 0) { 6376 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n")); 6377 break; 6378 } 6379 pf_commit_altq(t[0]); 6380 #endif 6381 6382 pf_clear_all_states(); 6383 6384 pf_clear_srcnodes(NULL); 6385 6386 /* status does not use malloced mem so no need to cleanup */ 6387 /* fingerprints and interfaces have their own cleanup code */ 6388 } while(0); 6389 6390 return (error); 6391 } 6392 6393 static pfil_return_t 6394 pf_check_return(int chk, struct mbuf **m) 6395 { 6396 6397 switch (chk) { 6398 case PF_PASS: 6399 if (*m == NULL) 6400 return (PFIL_CONSUMED); 6401 else 6402 return (PFIL_PASS); 6403 break; 6404 default: 6405 if (*m != NULL) { 6406 m_freem(*m); 6407 *m = NULL; 6408 } 6409 return (PFIL_DROPPED); 6410 } 6411 } 6412 6413 static pfil_return_t 6414 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6415 void *ruleset __unused, struct inpcb *inp) 6416 { 6417 int chk; 6418 6419 chk = pf_test_eth(PF_IN, flags, ifp, m, inp); 6420 6421 return (pf_check_return(chk, m)); 6422 } 6423 6424 static pfil_return_t 6425 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6426 void *ruleset __unused, struct inpcb *inp) 6427 { 6428 int chk; 6429 6430 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp); 6431 6432 return (pf_check_return(chk, m)); 6433 } 6434 6435 #ifdef INET 6436 static pfil_return_t 6437 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6438 void *ruleset __unused, struct inpcb *inp) 6439 { 6440 int chk; 6441 6442 chk = pf_test(PF_IN, flags, ifp, m, inp); 6443 6444 return (pf_check_return(chk, m)); 6445 } 6446 6447 static pfil_return_t 6448 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6449 void *ruleset __unused, struct inpcb *inp) 6450 { 6451 int chk; 6452 6453 chk = pf_test(PF_OUT, flags, ifp, m, inp); 6454 6455 return (pf_check_return(chk, m)); 6456 } 6457 #endif 6458 6459 #ifdef INET6 6460 static pfil_return_t 6461 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags, 6462 void *ruleset __unused, struct inpcb *inp) 6463 { 6464 int chk; 6465 6466 /* 6467 * In case of loopback traffic IPv6 uses the real interface in 6468 * order to support scoped addresses. In order to support stateful 6469 * filtering we have change this to lo0 as it is the case in IPv4. 6470 */ 6471 CURVNET_SET(ifp->if_vnet); 6472 chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp); 6473 CURVNET_RESTORE(); 6474 6475 return (pf_check_return(chk, m)); 6476 } 6477 6478 static pfil_return_t 6479 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags, 6480 void *ruleset __unused, struct inpcb *inp) 6481 { 6482 int chk; 6483 6484 CURVNET_SET(ifp->if_vnet); 6485 chk = pf_test6(PF_OUT, flags, ifp, m, inp); 6486 CURVNET_RESTORE(); 6487 6488 return (pf_check_return(chk, m)); 6489 } 6490 #endif /* INET6 */ 6491 6492 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook); 6493 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook); 6494 #define V_pf_eth_in_hook VNET(pf_eth_in_hook) 6495 #define V_pf_eth_out_hook VNET(pf_eth_out_hook) 6496 6497 #ifdef INET 6498 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook); 6499 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook); 6500 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook) 6501 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook) 6502 #endif 6503 #ifdef INET6 6504 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook); 6505 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook); 6506 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook) 6507 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook) 6508 #endif 6509 6510 static void 6511 hook_pf_eth(void) 6512 { 6513 struct pfil_hook_args pha; 6514 struct pfil_link_args pla; 6515 int ret __diagused; 6516 6517 if (V_pf_pfil_eth_hooked) 6518 return; 6519 6520 pha.pa_version = PFIL_VERSION; 6521 pha.pa_modname = "pf"; 6522 pha.pa_ruleset = NULL; 6523 6524 pla.pa_version = PFIL_VERSION; 6525 6526 pha.pa_type = PFIL_TYPE_ETHERNET; 6527 pha.pa_func = pf_eth_check_in; 6528 pha.pa_flags = PFIL_IN; 6529 pha.pa_rulname = "eth-in"; 6530 V_pf_eth_in_hook = pfil_add_hook(&pha); 6531 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6532 pla.pa_head = V_link_pfil_head; 6533 pla.pa_hook = V_pf_eth_in_hook; 6534 ret = pfil_link(&pla); 6535 MPASS(ret == 0); 6536 pha.pa_func = pf_eth_check_out; 6537 pha.pa_flags = PFIL_OUT; 6538 pha.pa_rulname = "eth-out"; 6539 V_pf_eth_out_hook = pfil_add_hook(&pha); 6540 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6541 pla.pa_head = V_link_pfil_head; 6542 pla.pa_hook = V_pf_eth_out_hook; 6543 ret = pfil_link(&pla); 6544 MPASS(ret == 0); 6545 6546 V_pf_pfil_eth_hooked = 1; 6547 } 6548 6549 static void 6550 hook_pf(void) 6551 { 6552 struct pfil_hook_args pha; 6553 struct pfil_link_args pla; 6554 int ret __diagused; 6555 6556 if (V_pf_pfil_hooked) 6557 return; 6558 6559 pha.pa_version = PFIL_VERSION; 6560 pha.pa_modname = "pf"; 6561 pha.pa_ruleset = NULL; 6562 6563 pla.pa_version = PFIL_VERSION; 6564 6565 #ifdef INET 6566 pha.pa_type = PFIL_TYPE_IP4; 6567 pha.pa_func = pf_check_in; 6568 pha.pa_flags = PFIL_IN; 6569 pha.pa_rulname = "default-in"; 6570 V_pf_ip4_in_hook = pfil_add_hook(&pha); 6571 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6572 pla.pa_head = V_inet_pfil_head; 6573 pla.pa_hook = V_pf_ip4_in_hook; 6574 ret = pfil_link(&pla); 6575 MPASS(ret == 0); 6576 pha.pa_func = pf_check_out; 6577 pha.pa_flags = PFIL_OUT; 6578 pha.pa_rulname = "default-out"; 6579 V_pf_ip4_out_hook = pfil_add_hook(&pha); 6580 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6581 pla.pa_head = V_inet_pfil_head; 6582 pla.pa_hook = V_pf_ip4_out_hook; 6583 ret = pfil_link(&pla); 6584 MPASS(ret == 0); 6585 #endif 6586 #ifdef INET6 6587 pha.pa_type = PFIL_TYPE_IP6; 6588 pha.pa_func = pf_check6_in; 6589 pha.pa_flags = PFIL_IN; 6590 pha.pa_rulname = "default-in6"; 6591 V_pf_ip6_in_hook = pfil_add_hook(&pha); 6592 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6593 pla.pa_head = V_inet6_pfil_head; 6594 pla.pa_hook = V_pf_ip6_in_hook; 6595 ret = pfil_link(&pla); 6596 MPASS(ret == 0); 6597 pha.pa_func = pf_check6_out; 6598 pha.pa_rulname = "default-out6"; 6599 pha.pa_flags = PFIL_OUT; 6600 V_pf_ip6_out_hook = pfil_add_hook(&pha); 6601 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6602 pla.pa_head = V_inet6_pfil_head; 6603 pla.pa_hook = V_pf_ip6_out_hook; 6604 ret = pfil_link(&pla); 6605 MPASS(ret == 0); 6606 #endif 6607 6608 V_pf_pfil_hooked = 1; 6609 } 6610 6611 static void 6612 dehook_pf_eth(void) 6613 { 6614 6615 if (V_pf_pfil_eth_hooked == 0) 6616 return; 6617 6618 pfil_remove_hook(V_pf_eth_in_hook); 6619 pfil_remove_hook(V_pf_eth_out_hook); 6620 6621 V_pf_pfil_eth_hooked = 0; 6622 } 6623 6624 static void 6625 dehook_pf(void) 6626 { 6627 6628 if (V_pf_pfil_hooked == 0) 6629 return; 6630 6631 #ifdef INET 6632 pfil_remove_hook(V_pf_ip4_in_hook); 6633 pfil_remove_hook(V_pf_ip4_out_hook); 6634 #endif 6635 #ifdef INET6 6636 pfil_remove_hook(V_pf_ip6_in_hook); 6637 pfil_remove_hook(V_pf_ip6_out_hook); 6638 #endif 6639 6640 V_pf_pfil_hooked = 0; 6641 } 6642 6643 static void 6644 pf_load_vnet(void) 6645 { 6646 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname), 6647 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 6648 6649 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize, 6650 PF_RULE_TAG_HASH_SIZE_DEFAULT); 6651 #ifdef ALTQ 6652 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize, 6653 PF_QUEUE_TAG_HASH_SIZE_DEFAULT); 6654 #endif 6655 6656 V_pf_keth = &V_pf_main_keth_anchor.ruleset; 6657 6658 pfattach_vnet(); 6659 V_pf_vnet_active = 1; 6660 } 6661 6662 static int 6663 pf_load(void) 6664 { 6665 int error; 6666 6667 rm_init_flags(&pf_rules_lock, "pf rulesets", RM_RECURSE); 6668 sx_init(&pf_ioctl_lock, "pf ioctl"); 6669 sx_init(&pf_end_lock, "pf end thread"); 6670 6671 pf_mtag_initialize(); 6672 6673 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME); 6674 if (pf_dev == NULL) 6675 return (ENOMEM); 6676 6677 pf_end_threads = 0; 6678 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge"); 6679 if (error != 0) 6680 return (error); 6681 6682 pfi_initialize(); 6683 6684 return (0); 6685 } 6686 6687 static void 6688 pf_unload_vnet(void) 6689 { 6690 int ret __diagused; 6691 6692 V_pf_vnet_active = 0; 6693 V_pf_status.running = 0; 6694 dehook_pf(); 6695 dehook_pf_eth(); 6696 6697 PF_RULES_WLOCK(); 6698 pf_syncookies_cleanup(); 6699 shutdown_pf(); 6700 PF_RULES_WUNLOCK(); 6701 6702 ret = swi_remove(V_pf_swi_cookie); 6703 MPASS(ret == 0); 6704 ret = intr_event_destroy(V_pf_swi_ie); 6705 MPASS(ret == 0); 6706 6707 pf_unload_vnet_purge(); 6708 6709 pf_normalize_cleanup(); 6710 PF_RULES_WLOCK(); 6711 pfi_cleanup_vnet(); 6712 PF_RULES_WUNLOCK(); 6713 pfr_cleanup(); 6714 pf_osfp_flush(); 6715 pf_cleanup(); 6716 if (IS_DEFAULT_VNET(curvnet)) 6717 pf_mtag_cleanup(); 6718 6719 pf_cleanup_tagset(&V_pf_tags); 6720 #ifdef ALTQ 6721 pf_cleanup_tagset(&V_pf_qids); 6722 #endif 6723 uma_zdestroy(V_pf_tag_z); 6724 6725 #ifdef PF_WANT_32_TO_64_COUNTER 6726 PF_RULES_WLOCK(); 6727 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); 6728 6729 MPASS(LIST_EMPTY(&V_pf_allkiflist)); 6730 MPASS(V_pf_allkifcount == 0); 6731 6732 LIST_REMOVE(&V_pf_default_rule, allrulelist); 6733 V_pf_allrulecount--; 6734 LIST_REMOVE(V_pf_rulemarker, allrulelist); 6735 6736 /* 6737 * There are known pf rule leaks when running the test suite. 6738 */ 6739 #ifdef notyet 6740 MPASS(LIST_EMPTY(&V_pf_allrulelist)); 6741 MPASS(V_pf_allrulecount == 0); 6742 #endif 6743 6744 PF_RULES_WUNLOCK(); 6745 6746 free(V_pf_kifmarker, PFI_MTYPE); 6747 free(V_pf_rulemarker, M_PFRULE); 6748 #endif 6749 6750 /* Free counters last as we updated them during shutdown. */ 6751 pf_counter_u64_deinit(&V_pf_default_rule.evaluations); 6752 for (int i = 0; i < 2; i++) { 6753 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]); 6754 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]); 6755 } 6756 counter_u64_free(V_pf_default_rule.states_cur); 6757 counter_u64_free(V_pf_default_rule.states_tot); 6758 counter_u64_free(V_pf_default_rule.src_nodes); 6759 uma_zfree_pcpu(pcpu_zone_4, V_pf_default_rule.timestamp); 6760 6761 for (int i = 0; i < PFRES_MAX; i++) 6762 counter_u64_free(V_pf_status.counters[i]); 6763 for (int i = 0; i < KLCNT_MAX; i++) 6764 counter_u64_free(V_pf_status.lcounters[i]); 6765 for (int i = 0; i < FCNT_MAX; i++) 6766 pf_counter_u64_deinit(&V_pf_status.fcounters[i]); 6767 for (int i = 0; i < SCNT_MAX; i++) 6768 counter_u64_free(V_pf_status.scounters[i]); 6769 } 6770 6771 static void 6772 pf_unload(void) 6773 { 6774 6775 sx_xlock(&pf_end_lock); 6776 pf_end_threads = 1; 6777 while (pf_end_threads < 2) { 6778 wakeup_one(pf_purge_thread); 6779 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0); 6780 } 6781 sx_xunlock(&pf_end_lock); 6782 6783 if (pf_dev != NULL) 6784 destroy_dev(pf_dev); 6785 6786 pfi_cleanup(); 6787 6788 rm_destroy(&pf_rules_lock); 6789 sx_destroy(&pf_ioctl_lock); 6790 sx_destroy(&pf_end_lock); 6791 } 6792 6793 static void 6794 vnet_pf_init(void *unused __unused) 6795 { 6796 6797 pf_load_vnet(); 6798 } 6799 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6800 vnet_pf_init, NULL); 6801 6802 static void 6803 vnet_pf_uninit(const void *unused __unused) 6804 { 6805 6806 pf_unload_vnet(); 6807 } 6808 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL); 6809 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6810 vnet_pf_uninit, NULL); 6811 6812 static int 6813 pf_modevent(module_t mod, int type, void *data) 6814 { 6815 int error = 0; 6816 6817 switch(type) { 6818 case MOD_LOAD: 6819 error = pf_load(); 6820 break; 6821 case MOD_UNLOAD: 6822 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after 6823 * the vnet_pf_uninit()s */ 6824 break; 6825 default: 6826 error = EINVAL; 6827 break; 6828 } 6829 6830 return (error); 6831 } 6832 6833 static moduledata_t pf_mod = { 6834 "pf", 6835 pf_modevent, 6836 0 6837 }; 6838 6839 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND); 6840 MODULE_VERSION(pf, PF_MODVER); 6841