1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 2001 Daniel Hartmeier 5 * Copyright (c) 2002,2003 Henning Brauer 6 * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org> 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 13 * - Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * - Redistributions in binary form must reproduce the above 16 * copyright notice, this list of conditions and the following 17 * disclaimer in the documentation and/or other materials provided 18 * with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 * 33 * Effort sponsored in part by the Defense Advanced Research Projects 34 * Agency (DARPA) and Air Force Research Laboratory, Air Force 35 * Materiel Command, USAF, under agreement number F30602-01-2-0537. 36 * 37 * $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $ 38 */ 39 40 #include <sys/cdefs.h> 41 #include "opt_inet.h" 42 #include "opt_inet6.h" 43 #include "opt_bpf.h" 44 #include "opt_pf.h" 45 46 #include <sys/param.h> 47 #include <sys/_bitset.h> 48 #include <sys/bitset.h> 49 #include <sys/bus.h> 50 #include <sys/conf.h> 51 #include <sys/endian.h> 52 #include <sys/fcntl.h> 53 #include <sys/filio.h> 54 #include <sys/hash.h> 55 #include <sys/interrupt.h> 56 #include <sys/jail.h> 57 #include <sys/kernel.h> 58 #include <sys/kthread.h> 59 #include <sys/lock.h> 60 #include <sys/mbuf.h> 61 #include <sys/module.h> 62 #include <sys/nv.h> 63 #include <sys/proc.h> 64 #include <sys/sdt.h> 65 #include <sys/smp.h> 66 #include <sys/socket.h> 67 #include <sys/sysctl.h> 68 #include <sys/md5.h> 69 #include <sys/ucred.h> 70 71 #include <net/if.h> 72 #include <net/if_var.h> 73 #include <net/if_private.h> 74 #include <net/vnet.h> 75 #include <net/route.h> 76 #include <net/pfil.h> 77 #include <net/pfvar.h> 78 #include <net/if_pfsync.h> 79 #include <net/if_pflog.h> 80 81 #include <netinet/in.h> 82 #include <netinet/ip.h> 83 #include <netinet/ip_var.h> 84 #include <netinet6/ip6_var.h> 85 #include <netinet/ip_icmp.h> 86 #include <netpfil/pf/pf_nl.h> 87 #include <netpfil/pf/pf_nv.h> 88 89 #ifdef INET6 90 #include <netinet/ip6.h> 91 #endif /* INET6 */ 92 93 #ifdef ALTQ 94 #include <net/altq/altq.h> 95 #endif 96 97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int"); 98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int"); 99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int"); 100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int"); 101 102 static struct pf_kpool *pf_get_kpool(const char *, u_int32_t, u_int8_t, 103 u_int32_t, u_int8_t, u_int8_t, u_int8_t, int); 104 105 static void pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *); 106 static void pf_empty_kpool(struct pf_kpalist *); 107 static int pfioctl(struct cdev *, u_long, caddr_t, int, 108 struct thread *); 109 static int pf_begin_eth(uint32_t *, const char *); 110 static int pf_rollback_eth(uint32_t, const char *); 111 static int pf_commit_eth(uint32_t, const char *); 112 static void pf_free_eth_rule(struct pf_keth_rule *); 113 #ifdef ALTQ 114 static int pf_begin_altq(u_int32_t *); 115 static int pf_rollback_altq(u_int32_t); 116 static int pf_commit_altq(u_int32_t); 117 static int pf_enable_altq(struct pf_altq *); 118 static int pf_disable_altq(struct pf_altq *); 119 static uint16_t pf_qname2qid(const char *); 120 static void pf_qid_unref(uint16_t); 121 #endif /* ALTQ */ 122 static int pf_begin_rules(u_int32_t *, int, const char *); 123 static int pf_rollback_rules(u_int32_t, int, char *); 124 static int pf_setup_pfsync_matching(struct pf_kruleset *); 125 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *); 126 static void pf_hash_rule(struct pf_krule *); 127 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *); 128 static int pf_commit_rules(u_int32_t, int, char *); 129 static int pf_addr_setup(struct pf_kruleset *, 130 struct pf_addr_wrap *, sa_family_t); 131 static void pf_src_node_copy(const struct pf_ksrc_node *, 132 struct pf_src_node *); 133 #ifdef ALTQ 134 static int pf_export_kaltq(struct pf_altq *, 135 struct pfioc_altq_v1 *, size_t); 136 static int pf_import_kaltq(struct pfioc_altq_v1 *, 137 struct pf_altq *, size_t); 138 #endif /* ALTQ */ 139 140 VNET_DEFINE(struct pf_krule, pf_default_rule); 141 142 static __inline int pf_krule_compare(struct pf_krule *, 143 struct pf_krule *); 144 145 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare); 146 147 #ifdef ALTQ 148 VNET_DEFINE_STATIC(int, pf_altq_running); 149 #define V_pf_altq_running VNET(pf_altq_running) 150 #endif 151 152 #define TAGID_MAX 50000 153 struct pf_tagname { 154 TAILQ_ENTRY(pf_tagname) namehash_entries; 155 TAILQ_ENTRY(pf_tagname) taghash_entries; 156 char name[PF_TAG_NAME_SIZE]; 157 uint16_t tag; 158 int ref; 159 }; 160 161 struct pf_tagset { 162 TAILQ_HEAD(, pf_tagname) *namehash; 163 TAILQ_HEAD(, pf_tagname) *taghash; 164 unsigned int mask; 165 uint32_t seed; 166 BITSET_DEFINE(, TAGID_MAX) avail; 167 }; 168 169 VNET_DEFINE(struct pf_tagset, pf_tags); 170 #define V_pf_tags VNET(pf_tags) 171 static unsigned int pf_rule_tag_hashsize; 172 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128 173 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN, 174 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT, 175 "Size of pf(4) rule tag hashtable"); 176 177 #ifdef ALTQ 178 VNET_DEFINE(struct pf_tagset, pf_qids); 179 #define V_pf_qids VNET(pf_qids) 180 static unsigned int pf_queue_tag_hashsize; 181 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128 182 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN, 183 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT, 184 "Size of pf(4) queue tag hashtable"); 185 #endif 186 VNET_DEFINE(uma_zone_t, pf_tag_z); 187 #define V_pf_tag_z VNET(pf_tag_z) 188 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db"); 189 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules"); 190 191 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE) 192 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE 193 #endif 194 195 VNET_DEFINE_STATIC(bool, pf_filter_local) = false; 196 #define V_pf_filter_local VNET(pf_filter_local) 197 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW, 198 &VNET_NAME(pf_filter_local), false, 199 "Enable filtering for packets delivered to local network stack"); 200 201 #ifdef PF_DEFAULT_TO_DROP 202 VNET_DEFINE_STATIC(bool, default_to_drop) = true; 203 #else 204 VNET_DEFINE_STATIC(bool, default_to_drop); 205 #endif 206 #define V_default_to_drop VNET(default_to_drop) 207 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET, 208 &VNET_NAME(default_to_drop), false, 209 "Make the default rule drop all packets."); 210 211 static void pf_init_tagset(struct pf_tagset *, unsigned int *, 212 unsigned int); 213 static void pf_cleanup_tagset(struct pf_tagset *); 214 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *); 215 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t); 216 static u_int16_t tagname2tag(struct pf_tagset *, const char *); 217 static u_int16_t pf_tagname2tag(const char *); 218 static void tag_unref(struct pf_tagset *, u_int16_t); 219 220 struct cdev *pf_dev; 221 222 /* 223 * XXX - These are new and need to be checked when moveing to a new version 224 */ 225 static void pf_clear_all_states(void); 226 static int pf_killstates_row(struct pf_kstate_kill *, 227 struct pf_idhash *); 228 static int pf_killstates_nv(struct pfioc_nv *); 229 static int pf_clearstates_nv(struct pfioc_nv *); 230 static int pf_getstate(struct pfioc_nv *); 231 static int pf_getstatus(struct pfioc_nv *); 232 static int pf_clear_tables(void); 233 static void pf_kill_srcnodes(struct pfioc_src_node_kill *); 234 static int pf_keepcounters(struct pfioc_nv *); 235 static void pf_tbladdr_copyout(struct pf_addr_wrap *); 236 237 /* 238 * Wrapper functions for pfil(9) hooks 239 */ 240 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, 241 int flags, void *ruleset __unused, struct inpcb *inp); 242 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, 243 int flags, void *ruleset __unused, struct inpcb *inp); 244 #ifdef INET 245 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp, 246 int flags, void *ruleset __unused, struct inpcb *inp); 247 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp, 248 int flags, void *ruleset __unused, struct inpcb *inp); 249 #endif 250 #ifdef INET6 251 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp, 252 int flags, void *ruleset __unused, struct inpcb *inp); 253 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp, 254 int flags, void *ruleset __unused, struct inpcb *inp); 255 #endif 256 257 static void hook_pf_eth(void); 258 static void hook_pf(void); 259 static void dehook_pf_eth(void); 260 static void dehook_pf(void); 261 static int shutdown_pf(void); 262 static int pf_load(void); 263 static void pf_unload(void); 264 265 static struct cdevsw pf_cdevsw = { 266 .d_ioctl = pfioctl, 267 .d_name = PF_NAME, 268 .d_version = D_VERSION, 269 }; 270 271 VNET_DEFINE_STATIC(bool, pf_pfil_hooked); 272 #define V_pf_pfil_hooked VNET(pf_pfil_hooked) 273 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked); 274 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked) 275 276 /* 277 * We need a flag that is neither hooked nor running to know when 278 * the VNET is "valid". We primarily need this to control (global) 279 * external event, e.g., eventhandlers. 280 */ 281 VNET_DEFINE(int, pf_vnet_active); 282 #define V_pf_vnet_active VNET(pf_vnet_active) 283 284 int pf_end_threads; 285 struct proc *pf_purge_proc; 286 287 VNET_DEFINE(struct rmlock, pf_rules_lock); 288 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock); 289 #define V_pf_ioctl_lock VNET(pf_ioctl_lock) 290 struct sx pf_end_lock; 291 292 /* pfsync */ 293 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr); 294 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr); 295 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr); 296 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr); 297 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr); 298 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr); 299 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr); 300 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr; 301 302 /* pflog */ 303 pflog_packet_t *pflog_packet_ptr = NULL; 304 305 /* 306 * Copy a user-provided string, returning an error if truncation would occur. 307 * Avoid scanning past "sz" bytes in the source string since there's no 308 * guarantee that it's nul-terminated. 309 */ 310 static int 311 pf_user_strcpy(char *dst, const char *src, size_t sz) 312 { 313 if (strnlen(src, sz) == sz) 314 return (EINVAL); 315 (void)strlcpy(dst, src, sz); 316 return (0); 317 } 318 319 static void 320 pfattach_vnet(void) 321 { 322 u_int32_t *my_timeout = V_pf_default_rule.timeout; 323 324 bzero(&V_pf_status, sizeof(V_pf_status)); 325 326 pf_initialize(); 327 pfr_initialize(); 328 pfi_initialize_vnet(); 329 pf_normalize_init(); 330 pf_syncookies_init(); 331 332 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT; 333 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT; 334 V_pf_limits[PF_LIMIT_ANCHORS].limit = PF_ANCHOR_HIWAT; 335 V_pf_limits[PF_LIMIT_ETH_ANCHORS].limit = PF_ANCHOR_HIWAT; 336 337 RB_INIT(&V_pf_anchors); 338 pf_init_kruleset(&pf_main_ruleset); 339 340 pf_init_keth(V_pf_keth); 341 342 /* default rule should never be garbage collected */ 343 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; 344 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS; 345 V_pf_default_rule.nr = (uint32_t)-1; 346 V_pf_default_rule.rtableid = -1; 347 348 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK); 349 for (int i = 0; i < 2; i++) { 350 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK); 351 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK); 352 } 353 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); 354 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); 355 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++) 356 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 357 358 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 359 M_WAITOK | M_ZERO); 360 361 #ifdef PF_WANT_32_TO_64_COUNTER 362 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO); 363 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO); 364 PF_RULES_WLOCK(); 365 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); 366 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist); 367 V_pf_allrulecount++; 368 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); 369 PF_RULES_WUNLOCK(); 370 #endif 371 372 /* initialize default timeouts */ 373 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 374 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; 375 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 376 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; 377 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; 378 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; 379 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 380 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL; 381 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 382 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL; 383 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL; 384 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; 385 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; 386 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; 387 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; 388 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; 389 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; 390 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; 391 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; 392 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; 393 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; 394 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; 395 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; 396 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; 397 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; 398 399 V_pf_status.debug = PF_DEBUG_URGENT; 400 /* 401 * XXX This is different than in OpenBSD where reassembly is enabled by 402 * defult. In FreeBSD we expect people to still use scrub rules and 403 * switch to the new syntax later. Only when they switch they must 404 * explicitly enable reassemle. We could change the default once the 405 * scrub rule functionality is hopefully removed some day in future. 406 */ 407 V_pf_status.reass = 0; 408 409 V_pf_pfil_hooked = false; 410 V_pf_pfil_eth_hooked = false; 411 412 /* XXX do our best to avoid a conflict */ 413 V_pf_status.hostid = arc4random(); 414 415 for (int i = 0; i < PFRES_MAX; i++) 416 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); 417 for (int i = 0; i < KLCNT_MAX; i++) 418 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); 419 for (int i = 0; i < FCNT_MAX; i++) 420 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK); 421 for (int i = 0; i < SCNT_MAX; i++) 422 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); 423 424 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET, 425 INTR_MPSAFE, &V_pf_swi_cookie) != 0) 426 /* XXXGL: leaked all above. */ 427 return; 428 } 429 430 static struct pf_kpool * 431 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action, 432 u_int32_t rule_number, u_int8_t r_last, u_int8_t active, 433 u_int8_t check_ticket, int which) 434 { 435 struct pf_kruleset *ruleset; 436 struct pf_krule *rule; 437 int rs_num; 438 439 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT); 440 441 ruleset = pf_find_kruleset(anchor); 442 if (ruleset == NULL) 443 return (NULL); 444 rs_num = pf_get_ruleset_number(rule_action); 445 if (rs_num >= PF_RULESET_MAX) 446 return (NULL); 447 if (active) { 448 if (check_ticket && ticket != 449 ruleset->rules[rs_num].active.ticket) 450 return (NULL); 451 if (r_last) 452 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 453 pf_krulequeue); 454 else 455 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 456 } else { 457 if (check_ticket && ticket != 458 ruleset->rules[rs_num].inactive.ticket) 459 return (NULL); 460 if (r_last) 461 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 462 pf_krulequeue); 463 else 464 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); 465 } 466 if (!r_last) { 467 while ((rule != NULL) && (rule->nr != rule_number)) 468 rule = TAILQ_NEXT(rule, entries); 469 } 470 if (rule == NULL) 471 return (NULL); 472 473 switch (which) { 474 case PF_RDR: 475 return (&rule->rdr); 476 case PF_NAT: 477 return (&rule->nat); 478 case PF_RT: 479 return (&rule->route); 480 default: 481 panic("Unknow pool type %d", which); 482 } 483 } 484 485 static void 486 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb) 487 { 488 struct pf_kpooladdr *mv_pool_pa; 489 490 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { 491 TAILQ_REMOVE(poola, mv_pool_pa, entries); 492 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); 493 } 494 } 495 496 static void 497 pf_empty_kpool(struct pf_kpalist *poola) 498 { 499 struct pf_kpooladdr *pa; 500 501 while ((pa = TAILQ_FIRST(poola)) != NULL) { 502 switch (pa->addr.type) { 503 case PF_ADDR_DYNIFTL: 504 pfi_dynaddr_remove(pa->addr.p.dyn); 505 break; 506 case PF_ADDR_TABLE: 507 /* XXX: this could be unfinished pooladdr on pabuf */ 508 if (pa->addr.p.tbl != NULL) 509 pfr_detach_table(pa->addr.p.tbl); 510 break; 511 } 512 if (pa->kif) 513 pfi_kkif_unref(pa->kif); 514 TAILQ_REMOVE(poola, pa, entries); 515 free(pa, M_PFRULE); 516 } 517 } 518 519 static void 520 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 521 { 522 523 PF_RULES_WASSERT(); 524 PF_UNLNKDRULES_ASSERT(); 525 526 TAILQ_REMOVE(rulequeue, rule, entries); 527 528 rule->rule_ref |= PFRULE_REFS; 529 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); 530 } 531 532 static void 533 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 534 { 535 536 PF_RULES_WASSERT(); 537 538 PF_UNLNKDRULES_LOCK(); 539 pf_unlink_rule_locked(rulequeue, rule); 540 PF_UNLNKDRULES_UNLOCK(); 541 } 542 543 static void 544 pf_free_eth_rule(struct pf_keth_rule *rule) 545 { 546 PF_RULES_WASSERT(); 547 548 if (rule == NULL) 549 return; 550 551 if (rule->tag) 552 tag_unref(&V_pf_tags, rule->tag); 553 if (rule->match_tag) 554 tag_unref(&V_pf_tags, rule->match_tag); 555 #ifdef ALTQ 556 pf_qid_unref(rule->qid); 557 #endif 558 559 if (rule->bridge_to) 560 pfi_kkif_unref(rule->bridge_to); 561 if (rule->kif) 562 pfi_kkif_unref(rule->kif); 563 564 if (rule->ipsrc.addr.type == PF_ADDR_TABLE) 565 pfr_detach_table(rule->ipsrc.addr.p.tbl); 566 if (rule->ipdst.addr.type == PF_ADDR_TABLE) 567 pfr_detach_table(rule->ipdst.addr.p.tbl); 568 569 counter_u64_free(rule->evaluations); 570 for (int i = 0; i < 2; i++) { 571 counter_u64_free(rule->packets[i]); 572 counter_u64_free(rule->bytes[i]); 573 } 574 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 575 pf_keth_anchor_remove(rule); 576 577 free(rule, M_PFRULE); 578 } 579 580 void 581 pf_free_rule(struct pf_krule *rule) 582 { 583 584 PF_RULES_WASSERT(); 585 PF_CONFIG_ASSERT(); 586 587 if (rule->tag) 588 tag_unref(&V_pf_tags, rule->tag); 589 if (rule->match_tag) 590 tag_unref(&V_pf_tags, rule->match_tag); 591 #ifdef ALTQ 592 if (rule->pqid != rule->qid) 593 pf_qid_unref(rule->pqid); 594 pf_qid_unref(rule->qid); 595 #endif 596 switch (rule->src.addr.type) { 597 case PF_ADDR_DYNIFTL: 598 pfi_dynaddr_remove(rule->src.addr.p.dyn); 599 break; 600 case PF_ADDR_TABLE: 601 pfr_detach_table(rule->src.addr.p.tbl); 602 break; 603 } 604 switch (rule->dst.addr.type) { 605 case PF_ADDR_DYNIFTL: 606 pfi_dynaddr_remove(rule->dst.addr.p.dyn); 607 break; 608 case PF_ADDR_TABLE: 609 pfr_detach_table(rule->dst.addr.p.tbl); 610 break; 611 } 612 if (rule->overload_tbl) 613 pfr_detach_table(rule->overload_tbl); 614 if (rule->kif) 615 pfi_kkif_unref(rule->kif); 616 if (rule->rcv_kif) 617 pfi_kkif_unref(rule->rcv_kif); 618 pf_remove_kanchor(rule); 619 pf_empty_kpool(&rule->rdr.list); 620 pf_empty_kpool(&rule->nat.list); 621 pf_empty_kpool(&rule->route.list); 622 623 pf_krule_free(rule); 624 } 625 626 static void 627 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size, 628 unsigned int default_size) 629 { 630 unsigned int i; 631 unsigned int hashsize; 632 633 if (*tunable_size == 0 || !powerof2(*tunable_size)) 634 *tunable_size = default_size; 635 636 hashsize = *tunable_size; 637 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH, 638 M_WAITOK); 639 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH, 640 M_WAITOK); 641 ts->mask = hashsize - 1; 642 ts->seed = arc4random(); 643 for (i = 0; i < hashsize; i++) { 644 TAILQ_INIT(&ts->namehash[i]); 645 TAILQ_INIT(&ts->taghash[i]); 646 } 647 BIT_FILL(TAGID_MAX, &ts->avail); 648 } 649 650 static void 651 pf_cleanup_tagset(struct pf_tagset *ts) 652 { 653 unsigned int i; 654 unsigned int hashsize; 655 struct pf_tagname *t, *tmp; 656 657 /* 658 * Only need to clean up one of the hashes as each tag is hashed 659 * into each table. 660 */ 661 hashsize = ts->mask + 1; 662 for (i = 0; i < hashsize; i++) 663 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp) 664 uma_zfree(V_pf_tag_z, t); 665 666 free(ts->namehash, M_PFHASH); 667 free(ts->taghash, M_PFHASH); 668 } 669 670 static uint16_t 671 tagname2hashindex(const struct pf_tagset *ts, const char *tagname) 672 { 673 size_t len; 674 675 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1); 676 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask); 677 } 678 679 static uint16_t 680 tag2hashindex(const struct pf_tagset *ts, uint16_t tag) 681 { 682 683 return (tag & ts->mask); 684 } 685 686 static u_int16_t 687 tagname2tag(struct pf_tagset *ts, const char *tagname) 688 { 689 struct pf_tagname *tag; 690 u_int32_t index; 691 u_int16_t new_tagid; 692 693 PF_RULES_WASSERT(); 694 695 index = tagname2hashindex(ts, tagname); 696 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 697 if (strcmp(tagname, tag->name) == 0) { 698 tag->ref++; 699 return (tag->tag); 700 } 701 702 /* 703 * new entry 704 * 705 * to avoid fragmentation, we do a linear search from the beginning 706 * and take the first free slot we find. 707 */ 708 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail); 709 /* 710 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX]. 711 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits 712 * set. It may also return a bit number greater than TAGID_MAX due 713 * to rounding of the number of bits in the vector up to a multiple 714 * of the vector word size at declaration/allocation time. 715 */ 716 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) 717 return (0); 718 719 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */ 720 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail); 721 722 /* allocate and fill new struct pf_tagname */ 723 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT); 724 if (tag == NULL) 725 return (0); 726 strlcpy(tag->name, tagname, sizeof(tag->name)); 727 tag->tag = new_tagid; 728 tag->ref = 1; 729 730 /* Insert into namehash */ 731 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries); 732 733 /* Insert into taghash */ 734 index = tag2hashindex(ts, new_tagid); 735 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries); 736 737 return (tag->tag); 738 } 739 740 static void 741 tag_unref(struct pf_tagset *ts, u_int16_t tag) 742 { 743 struct pf_tagname *t; 744 uint16_t index; 745 746 PF_RULES_WASSERT(); 747 748 index = tag2hashindex(ts, tag); 749 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 750 if (tag == t->tag) { 751 if (--t->ref == 0) { 752 TAILQ_REMOVE(&ts->taghash[index], t, 753 taghash_entries); 754 index = tagname2hashindex(ts, t->name); 755 TAILQ_REMOVE(&ts->namehash[index], t, 756 namehash_entries); 757 /* Bits are 0-based for BIT_SET() */ 758 BIT_SET(TAGID_MAX, tag - 1, &ts->avail); 759 uma_zfree(V_pf_tag_z, t); 760 } 761 break; 762 } 763 } 764 765 static uint16_t 766 pf_tagname2tag(const char *tagname) 767 { 768 return (tagname2tag(&V_pf_tags, tagname)); 769 } 770 771 static int 772 pf_begin_eth(uint32_t *ticket, const char *anchor) 773 { 774 struct pf_keth_rule *rule, *tmp; 775 struct pf_keth_ruleset *rs; 776 777 PF_RULES_WASSERT(); 778 779 rs = pf_find_or_create_keth_ruleset(anchor); 780 if (rs == NULL) 781 return (EINVAL); 782 783 /* Purge old inactive rules. */ 784 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 785 tmp) { 786 TAILQ_REMOVE(rs->inactive.rules, rule, 787 entries); 788 pf_free_eth_rule(rule); 789 } 790 791 *ticket = ++rs->inactive.ticket; 792 rs->inactive.open = 1; 793 794 return (0); 795 } 796 797 static int 798 pf_rollback_eth(uint32_t ticket, const char *anchor) 799 { 800 struct pf_keth_rule *rule, *tmp; 801 struct pf_keth_ruleset *rs; 802 803 PF_RULES_WASSERT(); 804 805 rs = pf_find_keth_ruleset(anchor); 806 if (rs == NULL) 807 return (EINVAL); 808 809 if (!rs->inactive.open || 810 ticket != rs->inactive.ticket) 811 return (0); 812 813 /* Purge old inactive rules. */ 814 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 815 tmp) { 816 TAILQ_REMOVE(rs->inactive.rules, rule, entries); 817 pf_free_eth_rule(rule); 818 } 819 820 rs->inactive.open = 0; 821 822 pf_remove_if_empty_keth_ruleset(rs); 823 824 return (0); 825 } 826 827 #define PF_SET_SKIP_STEPS(i) \ 828 do { \ 829 while (head[i] != cur) { \ 830 head[i]->skip[i].ptr = cur; \ 831 head[i] = TAILQ_NEXT(head[i], entries); \ 832 } \ 833 } while (0) 834 835 static void 836 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules) 837 { 838 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT]; 839 int i; 840 841 cur = TAILQ_FIRST(rules); 842 prev = cur; 843 for (i = 0; i < PFE_SKIP_COUNT; ++i) 844 head[i] = cur; 845 while (cur != NULL) { 846 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 847 PF_SET_SKIP_STEPS(PFE_SKIP_IFP); 848 if (cur->direction != prev->direction) 849 PF_SET_SKIP_STEPS(PFE_SKIP_DIR); 850 if (cur->proto != prev->proto) 851 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO); 852 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0) 853 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR); 854 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0) 855 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR); 856 if (cur->ipsrc.neg != prev->ipsrc.neg || 857 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr)) 858 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR); 859 if (cur->ipdst.neg != prev->ipdst.neg || 860 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr)) 861 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR); 862 863 prev = cur; 864 cur = TAILQ_NEXT(cur, entries); 865 } 866 for (i = 0; i < PFE_SKIP_COUNT; ++i) 867 PF_SET_SKIP_STEPS(i); 868 } 869 870 static int 871 pf_commit_eth(uint32_t ticket, const char *anchor) 872 { 873 struct pf_keth_ruleq *rules; 874 struct pf_keth_ruleset *rs; 875 876 rs = pf_find_keth_ruleset(anchor); 877 if (rs == NULL) { 878 return (EINVAL); 879 } 880 881 if (!rs->inactive.open || 882 ticket != rs->inactive.ticket) 883 return (EBUSY); 884 885 PF_RULES_WASSERT(); 886 887 pf_eth_calc_skip_steps(rs->inactive.rules); 888 889 rules = rs->active.rules; 890 atomic_store_ptr(&rs->active.rules, rs->inactive.rules); 891 rs->inactive.rules = rules; 892 rs->inactive.ticket = rs->active.ticket; 893 894 return (pf_rollback_eth(rs->inactive.ticket, 895 rs->anchor ? rs->anchor->path : "")); 896 } 897 898 #ifdef ALTQ 899 static uint16_t 900 pf_qname2qid(const char *qname) 901 { 902 return (tagname2tag(&V_pf_qids, qname)); 903 } 904 905 static void 906 pf_qid_unref(uint16_t qid) 907 { 908 tag_unref(&V_pf_qids, qid); 909 } 910 911 static int 912 pf_begin_altq(u_int32_t *ticket) 913 { 914 struct pf_altq *altq, *tmp; 915 int error = 0; 916 917 PF_RULES_WASSERT(); 918 919 /* Purge the old altq lists */ 920 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 921 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 922 /* detach and destroy the discipline */ 923 error = altq_remove(altq); 924 } 925 free(altq, M_PFALTQ); 926 } 927 TAILQ_INIT(V_pf_altq_ifs_inactive); 928 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 929 pf_qid_unref(altq->qid); 930 free(altq, M_PFALTQ); 931 } 932 TAILQ_INIT(V_pf_altqs_inactive); 933 if (error) 934 return (error); 935 *ticket = ++V_ticket_altqs_inactive; 936 V_altqs_inactive_open = 1; 937 return (0); 938 } 939 940 static int 941 pf_rollback_altq(u_int32_t ticket) 942 { 943 struct pf_altq *altq, *tmp; 944 int error = 0; 945 946 PF_RULES_WASSERT(); 947 948 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 949 return (0); 950 /* Purge the old altq lists */ 951 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 952 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 953 /* detach and destroy the discipline */ 954 error = altq_remove(altq); 955 } 956 free(altq, M_PFALTQ); 957 } 958 TAILQ_INIT(V_pf_altq_ifs_inactive); 959 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 960 pf_qid_unref(altq->qid); 961 free(altq, M_PFALTQ); 962 } 963 TAILQ_INIT(V_pf_altqs_inactive); 964 V_altqs_inactive_open = 0; 965 return (error); 966 } 967 968 static int 969 pf_commit_altq(u_int32_t ticket) 970 { 971 struct pf_altqqueue *old_altqs, *old_altq_ifs; 972 struct pf_altq *altq, *tmp; 973 int err, error = 0; 974 975 PF_RULES_WASSERT(); 976 977 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 978 return (EBUSY); 979 980 /* swap altqs, keep the old. */ 981 old_altqs = V_pf_altqs_active; 982 old_altq_ifs = V_pf_altq_ifs_active; 983 V_pf_altqs_active = V_pf_altqs_inactive; 984 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive; 985 V_pf_altqs_inactive = old_altqs; 986 V_pf_altq_ifs_inactive = old_altq_ifs; 987 V_ticket_altqs_active = V_ticket_altqs_inactive; 988 989 /* Attach new disciplines */ 990 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 991 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 992 /* attach the discipline */ 993 error = altq_pfattach(altq); 994 if (error == 0 && V_pf_altq_running) 995 error = pf_enable_altq(altq); 996 if (error != 0) 997 return (error); 998 } 999 } 1000 1001 /* Purge the old altq lists */ 1002 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 1003 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 1004 /* detach and destroy the discipline */ 1005 if (V_pf_altq_running) 1006 error = pf_disable_altq(altq); 1007 err = altq_pfdetach(altq); 1008 if (err != 0 && error == 0) 1009 error = err; 1010 err = altq_remove(altq); 1011 if (err != 0 && error == 0) 1012 error = err; 1013 } 1014 free(altq, M_PFALTQ); 1015 } 1016 TAILQ_INIT(V_pf_altq_ifs_inactive); 1017 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1018 pf_qid_unref(altq->qid); 1019 free(altq, M_PFALTQ); 1020 } 1021 TAILQ_INIT(V_pf_altqs_inactive); 1022 1023 V_altqs_inactive_open = 0; 1024 return (error); 1025 } 1026 1027 static int 1028 pf_enable_altq(struct pf_altq *altq) 1029 { 1030 struct ifnet *ifp; 1031 struct tb_profile tb; 1032 int error = 0; 1033 1034 if ((ifp = ifunit(altq->ifname)) == NULL) 1035 return (EINVAL); 1036 1037 if (ifp->if_snd.altq_type != ALTQT_NONE) 1038 error = altq_enable(&ifp->if_snd); 1039 1040 /* set tokenbucket regulator */ 1041 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { 1042 tb.rate = altq->ifbandwidth; 1043 tb.depth = altq->tbrsize; 1044 error = tbr_set(&ifp->if_snd, &tb); 1045 } 1046 1047 return (error); 1048 } 1049 1050 static int 1051 pf_disable_altq(struct pf_altq *altq) 1052 { 1053 struct ifnet *ifp; 1054 struct tb_profile tb; 1055 int error; 1056 1057 if ((ifp = ifunit(altq->ifname)) == NULL) 1058 return (EINVAL); 1059 1060 /* 1061 * when the discipline is no longer referenced, it was overridden 1062 * by a new one. if so, just return. 1063 */ 1064 if (altq->altq_disc != ifp->if_snd.altq_disc) 1065 return (0); 1066 1067 error = altq_disable(&ifp->if_snd); 1068 1069 if (error == 0) { 1070 /* clear tokenbucket regulator */ 1071 tb.rate = 0; 1072 error = tbr_set(&ifp->if_snd, &tb); 1073 } 1074 1075 return (error); 1076 } 1077 1078 static int 1079 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket, 1080 struct pf_altq *altq) 1081 { 1082 struct ifnet *ifp1; 1083 int error = 0; 1084 1085 /* Deactivate the interface in question */ 1086 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; 1087 if ((ifp1 = ifunit(altq->ifname)) == NULL || 1088 (remove && ifp1 == ifp)) { 1089 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 1090 } else { 1091 error = altq_add(ifp1, altq); 1092 1093 if (ticket != V_ticket_altqs_inactive) 1094 error = EBUSY; 1095 1096 if (error) 1097 free(altq, M_PFALTQ); 1098 } 1099 1100 return (error); 1101 } 1102 1103 void 1104 pf_altq_ifnet_event(struct ifnet *ifp, int remove) 1105 { 1106 struct pf_altq *a1, *a2, *a3; 1107 u_int32_t ticket; 1108 int error = 0; 1109 1110 /* 1111 * No need to re-evaluate the configuration for events on interfaces 1112 * that do not support ALTQ, as it's not possible for such 1113 * interfaces to be part of the configuration. 1114 */ 1115 if (!ALTQ_IS_READY(&ifp->if_snd)) 1116 return; 1117 1118 /* Interrupt userland queue modifications */ 1119 if (V_altqs_inactive_open) 1120 pf_rollback_altq(V_ticket_altqs_inactive); 1121 1122 /* Start new altq ruleset */ 1123 if (pf_begin_altq(&ticket)) 1124 return; 1125 1126 /* Copy the current active set */ 1127 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) { 1128 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1129 if (a2 == NULL) { 1130 error = ENOMEM; 1131 break; 1132 } 1133 bcopy(a1, a2, sizeof(struct pf_altq)); 1134 1135 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1136 if (error) 1137 break; 1138 1139 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries); 1140 } 1141 if (error) 1142 goto out; 1143 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { 1144 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1145 if (a2 == NULL) { 1146 error = ENOMEM; 1147 break; 1148 } 1149 bcopy(a1, a2, sizeof(struct pf_altq)); 1150 1151 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) { 1152 error = EBUSY; 1153 free(a2, M_PFALTQ); 1154 break; 1155 } 1156 a2->altq_disc = NULL; 1157 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) { 1158 if (strncmp(a3->ifname, a2->ifname, 1159 IFNAMSIZ) == 0) { 1160 a2->altq_disc = a3->altq_disc; 1161 break; 1162 } 1163 } 1164 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1165 if (error) 1166 break; 1167 1168 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); 1169 } 1170 1171 out: 1172 if (error != 0) 1173 pf_rollback_altq(ticket); 1174 else 1175 pf_commit_altq(ticket); 1176 } 1177 #endif /* ALTQ */ 1178 1179 static struct pf_krule_global * 1180 pf_rule_tree_alloc(int flags) 1181 { 1182 struct pf_krule_global *tree; 1183 1184 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags); 1185 if (tree == NULL) 1186 return (NULL); 1187 RB_INIT(tree); 1188 return (tree); 1189 } 1190 1191 static void 1192 pf_rule_tree_free(struct pf_krule_global *tree) 1193 { 1194 1195 free(tree, M_TEMP); 1196 } 1197 1198 static int 1199 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) 1200 { 1201 struct pf_krule_global *tree; 1202 struct pf_kruleset *rs; 1203 struct pf_krule *rule; 1204 1205 PF_RULES_WASSERT(); 1206 1207 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1208 return (EINVAL); 1209 tree = pf_rule_tree_alloc(M_NOWAIT); 1210 if (tree == NULL) 1211 return (ENOMEM); 1212 rs = pf_find_or_create_kruleset(anchor); 1213 if (rs == NULL) { 1214 free(tree, M_TEMP); 1215 return (EINVAL); 1216 } 1217 pf_rule_tree_free(rs->rules[rs_num].inactive.tree); 1218 rs->rules[rs_num].inactive.tree = tree; 1219 1220 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1221 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1222 rs->rules[rs_num].inactive.rcount--; 1223 } 1224 *ticket = ++rs->rules[rs_num].inactive.ticket; 1225 rs->rules[rs_num].inactive.open = 1; 1226 return (0); 1227 } 1228 1229 static int 1230 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) 1231 { 1232 struct pf_kruleset *rs; 1233 struct pf_krule *rule; 1234 1235 PF_RULES_WASSERT(); 1236 1237 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1238 return (EINVAL); 1239 rs = pf_find_kruleset(anchor); 1240 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1241 rs->rules[rs_num].inactive.ticket != ticket) 1242 return (0); 1243 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1244 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1245 rs->rules[rs_num].inactive.rcount--; 1246 } 1247 rs->rules[rs_num].inactive.open = 0; 1248 return (0); 1249 } 1250 1251 #define PF_MD5_UPD(st, elm) \ 1252 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) 1253 1254 #define PF_MD5_UPD_STR(st, elm) \ 1255 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) 1256 1257 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ 1258 (stor) = htonl((st)->elm); \ 1259 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ 1260 } while (0) 1261 1262 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ 1263 (stor) = htons((st)->elm); \ 1264 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ 1265 } while (0) 1266 1267 static void 1268 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) 1269 { 1270 PF_MD5_UPD(pfr, addr.type); 1271 switch (pfr->addr.type) { 1272 case PF_ADDR_DYNIFTL: 1273 PF_MD5_UPD(pfr, addr.v.ifname); 1274 PF_MD5_UPD(pfr, addr.iflags); 1275 break; 1276 case PF_ADDR_TABLE: 1277 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX, 1278 strlen(PF_OPTIMIZER_TABLE_PFX))) 1279 PF_MD5_UPD(pfr, addr.v.tblname); 1280 break; 1281 case PF_ADDR_ADDRMASK: 1282 /* XXX ignore af? */ 1283 PF_MD5_UPD(pfr, addr.v.a.addr.addr32); 1284 PF_MD5_UPD(pfr, addr.v.a.mask.addr32); 1285 break; 1286 } 1287 1288 PF_MD5_UPD(pfr, port[0]); 1289 PF_MD5_UPD(pfr, port[1]); 1290 PF_MD5_UPD(pfr, neg); 1291 PF_MD5_UPD(pfr, port_op); 1292 } 1293 1294 static void 1295 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule) 1296 { 1297 u_int16_t x; 1298 u_int32_t y; 1299 1300 pf_hash_rule_addr(ctx, &rule->src); 1301 pf_hash_rule_addr(ctx, &rule->dst); 1302 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) 1303 PF_MD5_UPD_STR(rule, label[i]); 1304 PF_MD5_UPD_STR(rule, ifname); 1305 PF_MD5_UPD_STR(rule, rcv_ifname); 1306 PF_MD5_UPD_STR(rule, match_tagname); 1307 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ 1308 PF_MD5_UPD_HTONL(rule, os_fingerprint, y); 1309 PF_MD5_UPD_HTONL(rule, prob, y); 1310 PF_MD5_UPD_HTONL(rule, uid.uid[0], y); 1311 PF_MD5_UPD_HTONL(rule, uid.uid[1], y); 1312 PF_MD5_UPD(rule, uid.op); 1313 PF_MD5_UPD_HTONL(rule, gid.gid[0], y); 1314 PF_MD5_UPD_HTONL(rule, gid.gid[1], y); 1315 PF_MD5_UPD(rule, gid.op); 1316 PF_MD5_UPD_HTONL(rule, rule_flag, y); 1317 PF_MD5_UPD(rule, action); 1318 PF_MD5_UPD(rule, direction); 1319 PF_MD5_UPD(rule, af); 1320 PF_MD5_UPD(rule, quick); 1321 PF_MD5_UPD(rule, ifnot); 1322 PF_MD5_UPD(rule, rcvifnot); 1323 PF_MD5_UPD(rule, match_tag_not); 1324 PF_MD5_UPD(rule, natpass); 1325 PF_MD5_UPD(rule, keep_state); 1326 PF_MD5_UPD(rule, proto); 1327 PF_MD5_UPD(rule, type); 1328 PF_MD5_UPD(rule, code); 1329 PF_MD5_UPD(rule, flags); 1330 PF_MD5_UPD(rule, flagset); 1331 PF_MD5_UPD(rule, allow_opts); 1332 PF_MD5_UPD(rule, rt); 1333 PF_MD5_UPD(rule, tos); 1334 PF_MD5_UPD(rule, scrub_flags); 1335 PF_MD5_UPD(rule, min_ttl); 1336 PF_MD5_UPD(rule, set_tos); 1337 if (rule->anchor != NULL) 1338 PF_MD5_UPD_STR(rule, anchor->path); 1339 } 1340 1341 static void 1342 pf_hash_rule(struct pf_krule *rule) 1343 { 1344 MD5_CTX ctx; 1345 1346 MD5Init(&ctx); 1347 pf_hash_rule_rolling(&ctx, rule); 1348 MD5Final(rule->md5sum, &ctx); 1349 } 1350 1351 static int 1352 pf_krule_compare(struct pf_krule *a, struct pf_krule *b) 1353 { 1354 1355 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH)); 1356 } 1357 1358 static int 1359 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) 1360 { 1361 struct pf_kruleset *rs; 1362 struct pf_krule *rule, *old_rule; 1363 struct pf_krulequeue *old_rules; 1364 struct pf_krule_global *old_tree; 1365 int error; 1366 u_int32_t old_rcount; 1367 1368 PF_RULES_WASSERT(); 1369 1370 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1371 return (EINVAL); 1372 rs = pf_find_kruleset(anchor); 1373 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1374 ticket != rs->rules[rs_num].inactive.ticket) 1375 return (EBUSY); 1376 1377 /* Calculate checksum for the main ruleset */ 1378 if (rs == &pf_main_ruleset) { 1379 error = pf_setup_pfsync_matching(rs); 1380 if (error != 0) 1381 return (error); 1382 } 1383 1384 /* Swap rules, keep the old. */ 1385 old_rules = rs->rules[rs_num].active.ptr; 1386 old_rcount = rs->rules[rs_num].active.rcount; 1387 old_tree = rs->rules[rs_num].active.tree; 1388 1389 rs->rules[rs_num].active.ptr = 1390 rs->rules[rs_num].inactive.ptr; 1391 rs->rules[rs_num].active.tree = 1392 rs->rules[rs_num].inactive.tree; 1393 rs->rules[rs_num].active.rcount = 1394 rs->rules[rs_num].inactive.rcount; 1395 1396 /* Attempt to preserve counter information. */ 1397 if (V_pf_status.keep_counters && old_tree != NULL) { 1398 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr, 1399 entries) { 1400 old_rule = RB_FIND(pf_krule_global, old_tree, rule); 1401 if (old_rule == NULL) { 1402 continue; 1403 } 1404 pf_counter_u64_critical_enter(); 1405 pf_counter_u64_rollup_protected(&rule->evaluations, 1406 pf_counter_u64_fetch(&old_rule->evaluations)); 1407 pf_counter_u64_rollup_protected(&rule->packets[0], 1408 pf_counter_u64_fetch(&old_rule->packets[0])); 1409 pf_counter_u64_rollup_protected(&rule->packets[1], 1410 pf_counter_u64_fetch(&old_rule->packets[1])); 1411 pf_counter_u64_rollup_protected(&rule->bytes[0], 1412 pf_counter_u64_fetch(&old_rule->bytes[0])); 1413 pf_counter_u64_rollup_protected(&rule->bytes[1], 1414 pf_counter_u64_fetch(&old_rule->bytes[1])); 1415 pf_counter_u64_critical_exit(); 1416 } 1417 } 1418 1419 rs->rules[rs_num].inactive.ptr = old_rules; 1420 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ 1421 rs->rules[rs_num].inactive.rcount = old_rcount; 1422 1423 rs->rules[rs_num].active.ticket = 1424 rs->rules[rs_num].inactive.ticket; 1425 pf_calc_skip_steps(rs->rules[rs_num].active.ptr); 1426 1427 /* Purge the old rule list. */ 1428 PF_UNLNKDRULES_LOCK(); 1429 while ((rule = TAILQ_FIRST(old_rules)) != NULL) 1430 pf_unlink_rule_locked(old_rules, rule); 1431 PF_UNLNKDRULES_UNLOCK(); 1432 rs->rules[rs_num].inactive.rcount = 0; 1433 rs->rules[rs_num].inactive.open = 0; 1434 pf_remove_if_empty_kruleset(rs); 1435 free(old_tree, M_TEMP); 1436 1437 return (0); 1438 } 1439 1440 static int 1441 pf_setup_pfsync_matching(struct pf_kruleset *rs) 1442 { 1443 MD5_CTX ctx; 1444 struct pf_krule *rule; 1445 int rs_cnt; 1446 u_int8_t digest[PF_MD5_DIGEST_LENGTH]; 1447 1448 MD5Init(&ctx); 1449 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { 1450 /* XXX PF_RULESET_SCRUB as well? */ 1451 if (rs_cnt == PF_RULESET_SCRUB) 1452 continue; 1453 1454 if (rs->rules[rs_cnt].inactive.rcount) { 1455 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, 1456 entries) { 1457 pf_hash_rule_rolling(&ctx, rule); 1458 } 1459 } 1460 } 1461 1462 MD5Final(digest, &ctx); 1463 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); 1464 return (0); 1465 } 1466 1467 static int 1468 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr) 1469 { 1470 int error = 0; 1471 1472 switch (addr->type) { 1473 case PF_ADDR_TABLE: 1474 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname); 1475 if (addr->p.tbl == NULL) 1476 error = ENOMEM; 1477 break; 1478 default: 1479 error = EINVAL; 1480 } 1481 1482 return (error); 1483 } 1484 1485 static int 1486 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr, 1487 sa_family_t af) 1488 { 1489 int error = 0; 1490 1491 switch (addr->type) { 1492 case PF_ADDR_TABLE: 1493 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); 1494 if (addr->p.tbl == NULL) 1495 error = ENOMEM; 1496 break; 1497 case PF_ADDR_DYNIFTL: 1498 error = pfi_dynaddr_setup(addr, af); 1499 break; 1500 } 1501 1502 return (error); 1503 } 1504 1505 void 1506 pf_addr_copyout(struct pf_addr_wrap *addr) 1507 { 1508 1509 switch (addr->type) { 1510 case PF_ADDR_DYNIFTL: 1511 pfi_dynaddr_copyout(addr); 1512 break; 1513 case PF_ADDR_TABLE: 1514 pf_tbladdr_copyout(addr); 1515 break; 1516 } 1517 } 1518 1519 static void 1520 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out) 1521 { 1522 int secs = time_uptime; 1523 1524 bzero(out, sizeof(struct pf_src_node)); 1525 1526 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr)); 1527 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr)); 1528 1529 if (in->rule != NULL) 1530 out->rule.nr = in->rule->nr; 1531 1532 for (int i = 0; i < 2; i++) { 1533 out->bytes[i] = counter_u64_fetch(in->bytes[i]); 1534 out->packets[i] = counter_u64_fetch(in->packets[i]); 1535 } 1536 1537 out->states = in->states; 1538 out->conn = in->conn; 1539 out->af = in->af; 1540 out->ruletype = in->ruletype; 1541 1542 out->creation = secs - in->creation; 1543 if (out->expire > secs) 1544 out->expire -= secs; 1545 else 1546 out->expire = 0; 1547 1548 /* Adjust the connection rate estimate. */ 1549 out->conn_rate.limit = in->conn_rate.limit; 1550 out->conn_rate.seconds = in->conn_rate.seconds; 1551 /* If there's no limit there's no counter_rate. */ 1552 if (in->conn_rate.cr != NULL) 1553 out->conn_rate.count = counter_rate_get(in->conn_rate.cr); 1554 } 1555 1556 #ifdef ALTQ 1557 /* 1558 * Handle export of struct pf_kaltq to user binaries that may be using any 1559 * version of struct pf_altq. 1560 */ 1561 static int 1562 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size) 1563 { 1564 u_int32_t version; 1565 1566 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1567 version = 0; 1568 else 1569 version = pa->version; 1570 1571 if (version > PFIOC_ALTQ_VERSION) 1572 return (EINVAL); 1573 1574 #define ASSIGN(x) exported_q->x = q->x 1575 #define COPY(x) \ 1576 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x))) 1577 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX) 1578 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX) 1579 1580 switch (version) { 1581 case 0: { 1582 struct pf_altq_v0 *exported_q = 1583 &((struct pfioc_altq_v0 *)pa)->altq; 1584 1585 COPY(ifname); 1586 1587 ASSIGN(scheduler); 1588 ASSIGN(tbrsize); 1589 exported_q->tbrsize = SATU16(q->tbrsize); 1590 exported_q->ifbandwidth = SATU32(q->ifbandwidth); 1591 1592 COPY(qname); 1593 COPY(parent); 1594 ASSIGN(parent_qid); 1595 exported_q->bandwidth = SATU32(q->bandwidth); 1596 ASSIGN(priority); 1597 ASSIGN(local_flags); 1598 1599 ASSIGN(qlimit); 1600 ASSIGN(flags); 1601 1602 if (q->scheduler == ALTQT_HFSC) { 1603 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x 1604 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \ 1605 SATU32(q->pq_u.hfsc_opts.x) 1606 1607 ASSIGN_OPT_SATU32(rtsc_m1); 1608 ASSIGN_OPT(rtsc_d); 1609 ASSIGN_OPT_SATU32(rtsc_m2); 1610 1611 ASSIGN_OPT_SATU32(lssc_m1); 1612 ASSIGN_OPT(lssc_d); 1613 ASSIGN_OPT_SATU32(lssc_m2); 1614 1615 ASSIGN_OPT_SATU32(ulsc_m1); 1616 ASSIGN_OPT(ulsc_d); 1617 ASSIGN_OPT_SATU32(ulsc_m2); 1618 1619 ASSIGN_OPT(flags); 1620 1621 #undef ASSIGN_OPT 1622 #undef ASSIGN_OPT_SATU32 1623 } else 1624 COPY(pq_u); 1625 1626 ASSIGN(qid); 1627 break; 1628 } 1629 case 1: { 1630 struct pf_altq_v1 *exported_q = 1631 &((struct pfioc_altq_v1 *)pa)->altq; 1632 1633 COPY(ifname); 1634 1635 ASSIGN(scheduler); 1636 ASSIGN(tbrsize); 1637 ASSIGN(ifbandwidth); 1638 1639 COPY(qname); 1640 COPY(parent); 1641 ASSIGN(parent_qid); 1642 ASSIGN(bandwidth); 1643 ASSIGN(priority); 1644 ASSIGN(local_flags); 1645 1646 ASSIGN(qlimit); 1647 ASSIGN(flags); 1648 COPY(pq_u); 1649 1650 ASSIGN(qid); 1651 break; 1652 } 1653 default: 1654 panic("%s: unhandled struct pfioc_altq version", __func__); 1655 break; 1656 } 1657 1658 #undef ASSIGN 1659 #undef COPY 1660 #undef SATU16 1661 #undef SATU32 1662 1663 return (0); 1664 } 1665 1666 /* 1667 * Handle import to struct pf_kaltq of struct pf_altq from user binaries 1668 * that may be using any version of it. 1669 */ 1670 static int 1671 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size) 1672 { 1673 u_int32_t version; 1674 1675 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1676 version = 0; 1677 else 1678 version = pa->version; 1679 1680 if (version > PFIOC_ALTQ_VERSION) 1681 return (EINVAL); 1682 1683 #define ASSIGN(x) q->x = imported_q->x 1684 #define COPY(x) \ 1685 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x))) 1686 1687 switch (version) { 1688 case 0: { 1689 struct pf_altq_v0 *imported_q = 1690 &((struct pfioc_altq_v0 *)pa)->altq; 1691 1692 COPY(ifname); 1693 1694 ASSIGN(scheduler); 1695 ASSIGN(tbrsize); /* 16-bit -> 32-bit */ 1696 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */ 1697 1698 COPY(qname); 1699 COPY(parent); 1700 ASSIGN(parent_qid); 1701 ASSIGN(bandwidth); /* 32-bit -> 64-bit */ 1702 ASSIGN(priority); 1703 ASSIGN(local_flags); 1704 1705 ASSIGN(qlimit); 1706 ASSIGN(flags); 1707 1708 if (imported_q->scheduler == ALTQT_HFSC) { 1709 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x 1710 1711 /* 1712 * The m1 and m2 parameters are being copied from 1713 * 32-bit to 64-bit. 1714 */ 1715 ASSIGN_OPT(rtsc_m1); 1716 ASSIGN_OPT(rtsc_d); 1717 ASSIGN_OPT(rtsc_m2); 1718 1719 ASSIGN_OPT(lssc_m1); 1720 ASSIGN_OPT(lssc_d); 1721 ASSIGN_OPT(lssc_m2); 1722 1723 ASSIGN_OPT(ulsc_m1); 1724 ASSIGN_OPT(ulsc_d); 1725 ASSIGN_OPT(ulsc_m2); 1726 1727 ASSIGN_OPT(flags); 1728 1729 #undef ASSIGN_OPT 1730 } else 1731 COPY(pq_u); 1732 1733 ASSIGN(qid); 1734 break; 1735 } 1736 case 1: { 1737 struct pf_altq_v1 *imported_q = 1738 &((struct pfioc_altq_v1 *)pa)->altq; 1739 1740 COPY(ifname); 1741 1742 ASSIGN(scheduler); 1743 ASSIGN(tbrsize); 1744 ASSIGN(ifbandwidth); 1745 1746 COPY(qname); 1747 COPY(parent); 1748 ASSIGN(parent_qid); 1749 ASSIGN(bandwidth); 1750 ASSIGN(priority); 1751 ASSIGN(local_flags); 1752 1753 ASSIGN(qlimit); 1754 ASSIGN(flags); 1755 COPY(pq_u); 1756 1757 ASSIGN(qid); 1758 break; 1759 } 1760 default: 1761 panic("%s: unhandled struct pfioc_altq version", __func__); 1762 break; 1763 } 1764 1765 #undef ASSIGN 1766 #undef COPY 1767 1768 return (0); 1769 } 1770 1771 static struct pf_altq * 1772 pf_altq_get_nth_active(u_int32_t n) 1773 { 1774 struct pf_altq *altq; 1775 u_int32_t nr; 1776 1777 nr = 0; 1778 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1779 if (nr == n) 1780 return (altq); 1781 nr++; 1782 } 1783 1784 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { 1785 if (nr == n) 1786 return (altq); 1787 nr++; 1788 } 1789 1790 return (NULL); 1791 } 1792 #endif /* ALTQ */ 1793 1794 struct pf_krule * 1795 pf_krule_alloc(void) 1796 { 1797 struct pf_krule *rule; 1798 1799 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO); 1800 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF); 1801 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF); 1802 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF); 1803 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 1804 M_WAITOK | M_ZERO); 1805 return (rule); 1806 } 1807 1808 void 1809 pf_krule_free(struct pf_krule *rule) 1810 { 1811 #ifdef PF_WANT_32_TO_64_COUNTER 1812 bool wowned; 1813 #endif 1814 1815 if (rule == NULL) 1816 return; 1817 1818 #ifdef PF_WANT_32_TO_64_COUNTER 1819 if (rule->allrulelinked) { 1820 wowned = PF_RULES_WOWNED(); 1821 if (!wowned) 1822 PF_RULES_WLOCK(); 1823 LIST_REMOVE(rule, allrulelist); 1824 V_pf_allrulecount--; 1825 if (!wowned) 1826 PF_RULES_WUNLOCK(); 1827 } 1828 #endif 1829 1830 pf_counter_u64_deinit(&rule->evaluations); 1831 for (int i = 0; i < 2; i++) { 1832 pf_counter_u64_deinit(&rule->packets[i]); 1833 pf_counter_u64_deinit(&rule->bytes[i]); 1834 } 1835 counter_u64_free(rule->states_cur); 1836 counter_u64_free(rule->states_tot); 1837 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 1838 counter_u64_free(rule->src_nodes[sn_type]); 1839 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 1840 1841 mtx_destroy(&rule->nat.mtx); 1842 mtx_destroy(&rule->rdr.mtx); 1843 mtx_destroy(&rule->route.mtx); 1844 free(rule, M_PFRULE); 1845 } 1846 1847 void 1848 pf_krule_clear_counters(struct pf_krule *rule) 1849 { 1850 pf_counter_u64_zero(&rule->evaluations); 1851 for (int i = 0; i < 2; i++) { 1852 pf_counter_u64_zero(&rule->packets[i]); 1853 pf_counter_u64_zero(&rule->bytes[i]); 1854 } 1855 counter_u64_zero(rule->states_tot); 1856 } 1857 1858 static void 1859 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool, 1860 struct pf_pooladdr *pool) 1861 { 1862 1863 bzero(pool, sizeof(*pool)); 1864 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr)); 1865 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname)); 1866 } 1867 1868 static int 1869 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool, 1870 struct pf_kpooladdr *kpool) 1871 { 1872 int ret; 1873 1874 bzero(kpool, sizeof(*kpool)); 1875 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr)); 1876 ret = pf_user_strcpy(kpool->ifname, pool->ifname, 1877 sizeof(kpool->ifname)); 1878 return (ret); 1879 } 1880 1881 static void 1882 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool) 1883 { 1884 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), ""); 1885 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), ""); 1886 1887 bcopy(&pool->key, &kpool->key, sizeof(kpool->key)); 1888 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter)); 1889 1890 kpool->tblidx = pool->tblidx; 1891 kpool->proxy_port[0] = pool->proxy_port[0]; 1892 kpool->proxy_port[1] = pool->proxy_port[1]; 1893 kpool->opts = pool->opts; 1894 } 1895 1896 static int 1897 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule) 1898 { 1899 int ret; 1900 1901 #ifndef INET 1902 if (rule->af == AF_INET) { 1903 return (EAFNOSUPPORT); 1904 } 1905 #endif /* INET */ 1906 #ifndef INET6 1907 if (rule->af == AF_INET6) { 1908 return (EAFNOSUPPORT); 1909 } 1910 #endif /* INET6 */ 1911 1912 ret = pf_check_rule_addr(&rule->src); 1913 if (ret != 0) 1914 return (ret); 1915 ret = pf_check_rule_addr(&rule->dst); 1916 if (ret != 0) 1917 return (ret); 1918 1919 bcopy(&rule->src, &krule->src, sizeof(rule->src)); 1920 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst)); 1921 1922 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label)); 1923 if (ret != 0) 1924 return (ret); 1925 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname)); 1926 if (ret != 0) 1927 return (ret); 1928 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname)); 1929 if (ret != 0) 1930 return (ret); 1931 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname)); 1932 if (ret != 0) 1933 return (ret); 1934 ret = pf_user_strcpy(krule->tagname, rule->tagname, 1935 sizeof(rule->tagname)); 1936 if (ret != 0) 1937 return (ret); 1938 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname, 1939 sizeof(rule->match_tagname)); 1940 if (ret != 0) 1941 return (ret); 1942 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname, 1943 sizeof(rule->overload_tblname)); 1944 if (ret != 0) 1945 return (ret); 1946 1947 pf_pool_to_kpool(&rule->rpool, &krule->rdr); 1948 1949 /* Don't allow userspace to set evaluations, packets or bytes. */ 1950 /* kif, anchor, overload_tbl are not copied over. */ 1951 1952 krule->os_fingerprint = rule->os_fingerprint; 1953 1954 krule->rtableid = rule->rtableid; 1955 /* pf_rule->timeout is smaller than pf_krule->timeout */ 1956 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout)); 1957 krule->max_states = rule->max_states; 1958 krule->max_src_nodes = rule->max_src_nodes; 1959 krule->max_src_states = rule->max_src_states; 1960 krule->max_src_conn = rule->max_src_conn; 1961 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit; 1962 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds; 1963 krule->qid = rule->qid; 1964 krule->pqid = rule->pqid; 1965 krule->nr = rule->nr; 1966 krule->prob = rule->prob; 1967 krule->cuid = rule->cuid; 1968 krule->cpid = rule->cpid; 1969 1970 krule->return_icmp = rule->return_icmp; 1971 krule->return_icmp6 = rule->return_icmp6; 1972 krule->max_mss = rule->max_mss; 1973 krule->tag = rule->tag; 1974 krule->match_tag = rule->match_tag; 1975 krule->scrub_flags = rule->scrub_flags; 1976 1977 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid)); 1978 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid)); 1979 1980 krule->rule_flag = rule->rule_flag; 1981 krule->action = rule->action; 1982 krule->direction = rule->direction; 1983 krule->log = rule->log; 1984 krule->logif = rule->logif; 1985 krule->quick = rule->quick; 1986 krule->ifnot = rule->ifnot; 1987 krule->match_tag_not = rule->match_tag_not; 1988 krule->natpass = rule->natpass; 1989 1990 krule->keep_state = rule->keep_state; 1991 krule->af = rule->af; 1992 krule->proto = rule->proto; 1993 krule->type = rule->type; 1994 krule->code = rule->code; 1995 krule->flags = rule->flags; 1996 krule->flagset = rule->flagset; 1997 krule->min_ttl = rule->min_ttl; 1998 krule->allow_opts = rule->allow_opts; 1999 krule->rt = rule->rt; 2000 krule->return_ttl = rule->return_ttl; 2001 krule->tos = rule->tos; 2002 krule->set_tos = rule->set_tos; 2003 2004 krule->flush = rule->flush; 2005 krule->prio = rule->prio; 2006 krule->set_prio[0] = rule->set_prio[0]; 2007 krule->set_prio[1] = rule->set_prio[1]; 2008 2009 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert)); 2010 2011 return (0); 2012 } 2013 2014 int 2015 pf_ioctl_getrules(struct pfioc_rule *pr) 2016 { 2017 struct pf_kruleset *ruleset; 2018 struct pf_krule *tail; 2019 int rs_num; 2020 2021 PF_RULES_WLOCK(); 2022 ruleset = pf_find_kruleset(pr->anchor); 2023 if (ruleset == NULL) { 2024 PF_RULES_WUNLOCK(); 2025 return (EINVAL); 2026 } 2027 rs_num = pf_get_ruleset_number(pr->rule.action); 2028 if (rs_num >= PF_RULESET_MAX) { 2029 PF_RULES_WUNLOCK(); 2030 return (EINVAL); 2031 } 2032 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 2033 pf_krulequeue); 2034 if (tail) 2035 pr->nr = tail->nr + 1; 2036 else 2037 pr->nr = 0; 2038 pr->ticket = ruleset->rules[rs_num].active.ticket; 2039 PF_RULES_WUNLOCK(); 2040 2041 return (0); 2042 } 2043 2044 static int 2045 pf_rule_checkaf(struct pf_krule *r) 2046 { 2047 switch (r->af) { 2048 case 0: 2049 if (r->rule_flag & PFRULE_AFTO) 2050 return (EPFNOSUPPORT); 2051 break; 2052 case AF_INET: 2053 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6) 2054 return (EPFNOSUPPORT); 2055 break; 2056 #ifdef INET6 2057 case AF_INET6: 2058 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET) 2059 return (EPFNOSUPPORT); 2060 break; 2061 #endif /* INET6 */ 2062 default: 2063 return (EPFNOSUPPORT); 2064 } 2065 2066 if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0) 2067 return (EPFNOSUPPORT); 2068 2069 return (0); 2070 } 2071 2072 static int 2073 pf_validate_range(uint8_t op, uint16_t port[2]) 2074 { 2075 uint16_t a = ntohs(port[0]); 2076 uint16_t b = ntohs(port[1]); 2077 2078 if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */ 2079 (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */ 2080 (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */ 2081 return 1; 2082 return 0; 2083 } 2084 2085 int 2086 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, 2087 uint32_t pool_ticket, const char *anchor, const char *anchor_call, 2088 uid_t uid, pid_t pid) 2089 { 2090 struct pf_kruleset *ruleset; 2091 struct pf_krule *tail; 2092 struct pf_kpooladdr *pa; 2093 struct pfi_kkif *kif = NULL, *rcv_kif = NULL; 2094 int rs_num; 2095 int error = 0; 2096 2097 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 2098 #define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x) 2099 2100 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) 2101 ERROUT_UNLOCKED(EINVAL); 2102 2103 if ((error = pf_rule_checkaf(rule))) 2104 ERROUT_UNLOCKED(error); 2105 if (pf_validate_range(rule->src.port_op, rule->src.port)) 2106 ERROUT_UNLOCKED(EINVAL); 2107 if (pf_validate_range(rule->dst.port_op, rule->dst.port)) 2108 ERROUT_UNLOCKED(EINVAL); 2109 2110 if (rule->ifname[0]) 2111 kif = pf_kkif_create(M_WAITOK); 2112 if (rule->rcv_ifname[0]) 2113 rcv_kif = pf_kkif_create(M_WAITOK); 2114 pf_counter_u64_init(&rule->evaluations, M_WAITOK); 2115 for (int i = 0; i < 2; i++) { 2116 pf_counter_u64_init(&rule->packets[i], M_WAITOK); 2117 pf_counter_u64_init(&rule->bytes[i], M_WAITOK); 2118 } 2119 rule->states_cur = counter_u64_alloc(M_WAITOK); 2120 rule->states_tot = counter_u64_alloc(M_WAITOK); 2121 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 2122 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 2123 rule->cuid = uid; 2124 rule->cpid = pid; 2125 TAILQ_INIT(&rule->rdr.list); 2126 TAILQ_INIT(&rule->nat.list); 2127 TAILQ_INIT(&rule->route.list); 2128 2129 PF_CONFIG_LOCK(); 2130 PF_RULES_WLOCK(); 2131 #ifdef PF_WANT_32_TO_64_COUNTER 2132 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist); 2133 MPASS(!rule->allrulelinked); 2134 rule->allrulelinked = true; 2135 V_pf_allrulecount++; 2136 #endif 2137 ruleset = pf_find_kruleset(anchor); 2138 if (ruleset == NULL) 2139 ERROUT(EINVAL); 2140 rs_num = pf_get_ruleset_number(rule->action); 2141 if (rs_num >= PF_RULESET_MAX) 2142 ERROUT(EINVAL); 2143 if (ticket != ruleset->rules[rs_num].inactive.ticket) { 2144 DPFPRINTF(PF_DEBUG_MISC, 2145 "ticket: %d != [%d]%d", ticket, rs_num, 2146 ruleset->rules[rs_num].inactive.ticket); 2147 ERROUT(EBUSY); 2148 } 2149 if (pool_ticket != V_ticket_pabuf) { 2150 DPFPRINTF(PF_DEBUG_MISC, 2151 "pool_ticket: %d != %d", pool_ticket, 2152 V_ticket_pabuf); 2153 ERROUT(EBUSY); 2154 } 2155 /* 2156 * XXXMJG hack: there is no mechanism to ensure they started the 2157 * transaction. Ticket checked above may happen to match by accident, 2158 * even if nobody called DIOCXBEGIN, let alone this process. 2159 * Partially work around it by checking if the RB tree got allocated, 2160 * see pf_begin_rules. 2161 */ 2162 if (ruleset->rules[rs_num].inactive.tree == NULL) { 2163 ERROUT(EINVAL); 2164 } 2165 2166 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 2167 pf_krulequeue); 2168 if (tail) 2169 rule->nr = tail->nr + 1; 2170 else 2171 rule->nr = 0; 2172 if (rule->ifname[0]) { 2173 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2174 kif = NULL; 2175 pfi_kkif_ref(rule->kif); 2176 } else 2177 rule->kif = NULL; 2178 2179 if (rule->rcv_ifname[0]) { 2180 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname); 2181 rcv_kif = NULL; 2182 pfi_kkif_ref(rule->rcv_kif); 2183 } else 2184 rule->rcv_kif = NULL; 2185 2186 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) 2187 ERROUT(EBUSY); 2188 #ifdef ALTQ 2189 /* set queue IDs */ 2190 if (rule->qname[0] != 0) { 2191 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2192 ERROUT(EBUSY); 2193 else if (rule->pqname[0] != 0) { 2194 if ((rule->pqid = 2195 pf_qname2qid(rule->pqname)) == 0) 2196 ERROUT(EBUSY); 2197 } else 2198 rule->pqid = rule->qid; 2199 } 2200 #endif 2201 if (rule->tagname[0]) 2202 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2203 ERROUT(EBUSY); 2204 if (rule->match_tagname[0]) 2205 if ((rule->match_tag = 2206 pf_tagname2tag(rule->match_tagname)) == 0) 2207 ERROUT(EBUSY); 2208 if (rule->rt && !rule->direction) 2209 ERROUT(EINVAL); 2210 if (!rule->log) 2211 rule->logif = 0; 2212 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit, 2213 rule->pktrate.seconds)) 2214 ERROUT(ENOMEM); 2215 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) 2216 ERROUT(ENOMEM); 2217 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) 2218 ERROUT(ENOMEM); 2219 if (pf_kanchor_setup(rule, ruleset, anchor_call)) 2220 ERROUT(EINVAL); 2221 if (rule->scrub_flags & PFSTATE_SETPRIO && 2222 (rule->set_prio[0] > PF_PRIO_MAX || 2223 rule->set_prio[1] > PF_PRIO_MAX)) 2224 ERROUT(EINVAL); 2225 for (int i = 0; i < 3; i++) { 2226 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 2227 if (pa->addr.type == PF_ADDR_TABLE) { 2228 pa->addr.p.tbl = pfr_attach_table(ruleset, 2229 pa->addr.v.tblname); 2230 if (pa->addr.p.tbl == NULL) 2231 ERROUT(ENOMEM); 2232 } 2233 } 2234 2235 rule->overload_tbl = NULL; 2236 if (rule->overload_tblname[0]) { 2237 if ((rule->overload_tbl = pfr_attach_table(ruleset, 2238 rule->overload_tblname)) == NULL) 2239 ERROUT(EINVAL); 2240 else 2241 rule->overload_tbl->pfrkt_flags |= 2242 PFR_TFLAG_ACTIVE; 2243 } 2244 2245 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list); 2246 2247 /* 2248 * Old version of pfctl provide route redirection pools in single 2249 * common redirection pool rdr. New versions use rdr only for 2250 * rdr-to rules. 2251 */ 2252 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) { 2253 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list); 2254 } else { 2255 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list); 2256 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list); 2257 } 2258 2259 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) || 2260 (rule->action == PF_BINAT)) && rule->anchor == NULL && 2261 TAILQ_FIRST(&rule->rdr.list) == NULL) { 2262 ERROUT(EINVAL); 2263 } 2264 2265 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) { 2266 ERROUT(EINVAL); 2267 } 2268 2269 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR || 2270 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) { 2271 ERROUT(EINVAL); 2272 } 2273 2274 MPASS(error == 0); 2275 2276 rule->nat.cur = TAILQ_FIRST(&rule->nat.list); 2277 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list); 2278 rule->route.cur = TAILQ_FIRST(&rule->route.list); 2279 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, 2280 rule, entries); 2281 ruleset->rules[rs_num].inactive.rcount++; 2282 2283 PF_RULES_WUNLOCK(); 2284 pf_hash_rule(rule); 2285 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) { 2286 PF_RULES_WLOCK(); 2287 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries); 2288 ruleset->rules[rs_num].inactive.rcount--; 2289 pf_free_rule(rule); 2290 rule = NULL; 2291 ERROUT(EEXIST); 2292 } 2293 PF_CONFIG_UNLOCK(); 2294 2295 return (0); 2296 2297 #undef ERROUT 2298 #undef ERROUT_UNLOCKED 2299 errout: 2300 PF_RULES_WUNLOCK(); 2301 PF_CONFIG_UNLOCK(); 2302 errout_unlocked: 2303 pf_kkif_free(rcv_kif); 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_remove_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[psk->psk_nat ? PF_SK_STACK : 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_match_addr(psk->psk_src.neg, 2380 &psk->psk_src.addr.v.a.addr, 2381 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af)) 2382 continue; 2383 2384 if (! pf_match_addr(psk->psk_dst.neg, 2385 &psk->psk_dst.addr.v.a.addr, 2386 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af)) 2387 continue; 2388 2389 if (! pf_match_addr(psk->psk_rt_addr.neg, 2390 &psk->psk_rt_addr.addr.v.a.addr, 2391 &psk->psk_rt_addr.addr.v.a.mask, 2392 &s->act.rt_addr, sk->af)) 2393 continue; 2394 2395 if (psk->psk_src.port_op != 0 && 2396 ! pf_match_port(psk->psk_src.port_op, 2397 psk->psk_src.port[0], psk->psk_src.port[1], srcport)) 2398 continue; 2399 2400 if (psk->psk_dst.port_op != 0 && 2401 ! pf_match_port(psk->psk_dst.port_op, 2402 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) 2403 continue; 2404 2405 if (psk->psk_label[0] && 2406 ! pf_label_match(s->rule, psk->psk_label)) 2407 continue; 2408 2409 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname, 2410 kif->pfik_name)) 2411 continue; 2412 2413 if (psk->psk_kill_match) { 2414 /* Create the key to find matching states, with lock 2415 * held. */ 2416 2417 bzero(&match_key, sizeof(match_key)); 2418 2419 if (s->direction == PF_OUT) { 2420 dir = PF_IN; 2421 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK; 2422 } else { 2423 dir = PF_OUT; 2424 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE; 2425 } 2426 2427 match_key.af = s->key[idx]->af; 2428 match_key.proto = s->key[idx]->proto; 2429 pf_addrcpy(&match_key.addr[0], 2430 &s->key[idx]->addr[1], match_key.af); 2431 match_key.port[0] = s->key[idx]->port[1]; 2432 pf_addrcpy(&match_key.addr[1], 2433 &s->key[idx]->addr[0], match_key.af); 2434 match_key.port[1] = s->key[idx]->port[0]; 2435 } 2436 2437 pf_remove_state(s); 2438 killed++; 2439 2440 if (psk->psk_kill_match) 2441 killed += pf_kill_matching_state(&match_key, dir); 2442 2443 goto relock_DIOCKILLSTATES; 2444 } 2445 PF_HASHROW_UNLOCK(ih); 2446 2447 return (killed); 2448 } 2449 2450 void 2451 unhandled_af(int af) 2452 { 2453 panic("unhandled af %d", af); 2454 } 2455 2456 int 2457 pf_start(void) 2458 { 2459 int error = 0; 2460 2461 sx_xlock(&V_pf_ioctl_lock); 2462 if (V_pf_status.running) 2463 error = EEXIST; 2464 else { 2465 hook_pf(); 2466 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 2467 hook_pf_eth(); 2468 V_pf_status.running = 1; 2469 V_pf_status.since = time_uptime; 2470 new_unrhdr64(&V_pf_stateid, time_second); 2471 2472 DPFPRINTF(PF_DEBUG_MISC, "pf: started"); 2473 } 2474 sx_xunlock(&V_pf_ioctl_lock); 2475 2476 return (error); 2477 } 2478 2479 int 2480 pf_stop(void) 2481 { 2482 int error = 0; 2483 2484 sx_xlock(&V_pf_ioctl_lock); 2485 if (!V_pf_status.running) 2486 error = ENOENT; 2487 else { 2488 V_pf_status.running = 0; 2489 dehook_pf(); 2490 dehook_pf_eth(); 2491 V_pf_status.since = time_uptime; 2492 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped"); 2493 } 2494 sx_xunlock(&V_pf_ioctl_lock); 2495 2496 return (error); 2497 } 2498 2499 void 2500 pf_ioctl_clear_status(void) 2501 { 2502 PF_RULES_WLOCK(); 2503 for (int i = 0; i < PFRES_MAX; i++) 2504 counter_u64_zero(V_pf_status.counters[i]); 2505 for (int i = 0; i < FCNT_MAX; i++) 2506 pf_counter_u64_zero(&V_pf_status.fcounters[i]); 2507 for (int i = 0; i < SCNT_MAX; i++) 2508 counter_u64_zero(V_pf_status.scounters[i]); 2509 for (int i = 0; i < KLCNT_MAX; i++) 2510 counter_u64_zero(V_pf_status.lcounters[i]); 2511 V_pf_status.since = time_uptime; 2512 if (*V_pf_status.ifname) 2513 pfi_update_status(V_pf_status.ifname, NULL); 2514 PF_RULES_WUNLOCK(); 2515 } 2516 2517 int 2518 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds) 2519 { 2520 uint32_t old; 2521 2522 if (timeout < 0 || timeout >= PFTM_MAX || 2523 seconds < 0) 2524 return (EINVAL); 2525 2526 PF_RULES_WLOCK(); 2527 old = V_pf_default_rule.timeout[timeout]; 2528 if (timeout == PFTM_INTERVAL && seconds == 0) 2529 seconds = 1; 2530 V_pf_default_rule.timeout[timeout] = seconds; 2531 if (timeout == PFTM_INTERVAL && seconds < old) 2532 wakeup(pf_purge_thread); 2533 2534 if (prev_seconds != NULL) 2535 *prev_seconds = old; 2536 2537 PF_RULES_WUNLOCK(); 2538 2539 return (0); 2540 } 2541 2542 int 2543 pf_ioctl_get_timeout(int timeout, int *seconds) 2544 { 2545 PF_RULES_RLOCK_TRACKER; 2546 2547 if (timeout < 0 || timeout >= PFTM_MAX) 2548 return (EINVAL); 2549 2550 PF_RULES_RLOCK(); 2551 *seconds = V_pf_default_rule.timeout[timeout]; 2552 PF_RULES_RUNLOCK(); 2553 2554 return (0); 2555 } 2556 2557 int 2558 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit) 2559 { 2560 2561 PF_RULES_WLOCK(); 2562 if (index < 0 || index >= PF_LIMIT_MAX || 2563 V_pf_limits[index].zone == NULL) { 2564 PF_RULES_WUNLOCK(); 2565 return (EINVAL); 2566 } 2567 uma_zone_set_max(V_pf_limits[index].zone, 2568 limit == 0 ? INT_MAX : limit); 2569 if (old_limit != NULL) 2570 *old_limit = V_pf_limits[index].limit; 2571 V_pf_limits[index].limit = limit; 2572 PF_RULES_WUNLOCK(); 2573 2574 return (0); 2575 } 2576 2577 int 2578 pf_ioctl_get_limit(int index, unsigned int *limit) 2579 { 2580 PF_RULES_RLOCK_TRACKER; 2581 2582 if (index < 0 || index >= PF_LIMIT_MAX) 2583 return (EINVAL); 2584 2585 PF_RULES_RLOCK(); 2586 *limit = V_pf_limits[index].limit; 2587 PF_RULES_RUNLOCK(); 2588 2589 return (0); 2590 } 2591 2592 int 2593 pf_ioctl_begin_addrs(uint32_t *ticket) 2594 { 2595 PF_RULES_WLOCK(); 2596 pf_empty_kpool(&V_pf_pabuf[0]); 2597 pf_empty_kpool(&V_pf_pabuf[1]); 2598 pf_empty_kpool(&V_pf_pabuf[2]); 2599 *ticket = ++V_ticket_pabuf; 2600 PF_RULES_WUNLOCK(); 2601 2602 return (0); 2603 } 2604 2605 int 2606 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp) 2607 { 2608 struct pf_kpooladdr *pa = NULL; 2609 struct pfi_kkif *kif = NULL; 2610 int error; 2611 2612 if (pp->which != PF_RDR && pp->which != PF_NAT && 2613 pp->which != PF_RT) 2614 return (EINVAL); 2615 2616 switch (pp->af) { 2617 #ifdef INET 2618 case AF_INET: 2619 /* FALLTHROUGH */ 2620 #endif /* INET */ 2621 #ifdef INET6 2622 case AF_INET6: 2623 /* FALLTHROUGH */ 2624 #endif /* INET6 */ 2625 case AF_UNSPEC: 2626 break; 2627 default: 2628 return (EAFNOSUPPORT); 2629 } 2630 2631 if (pp->addr.addr.type != PF_ADDR_ADDRMASK && 2632 pp->addr.addr.type != PF_ADDR_DYNIFTL && 2633 pp->addr.addr.type != PF_ADDR_TABLE) 2634 return (EINVAL); 2635 2636 if (pp->addr.addr.p.dyn != NULL) 2637 return (EINVAL); 2638 2639 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); 2640 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa); 2641 if (error != 0) 2642 goto out; 2643 if (pa->ifname[0]) 2644 kif = pf_kkif_create(M_WAITOK); 2645 PF_RULES_WLOCK(); 2646 if (pp->ticket != V_ticket_pabuf) { 2647 PF_RULES_WUNLOCK(); 2648 if (pa->ifname[0]) 2649 pf_kkif_free(kif); 2650 error = EBUSY; 2651 goto out; 2652 } 2653 if (pa->ifname[0]) { 2654 pa->kif = pfi_kkif_attach(kif, pa->ifname); 2655 kif = NULL; 2656 pfi_kkif_ref(pa->kif); 2657 } else 2658 pa->kif = NULL; 2659 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = 2660 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { 2661 if (pa->ifname[0]) 2662 pfi_kkif_unref(pa->kif); 2663 PF_RULES_WUNLOCK(); 2664 goto out; 2665 } 2666 pa->af = pp->af; 2667 switch (pp->which) { 2668 case PF_NAT: 2669 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries); 2670 break; 2671 case PF_RDR: 2672 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries); 2673 break; 2674 case PF_RT: 2675 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries); 2676 break; 2677 } 2678 PF_RULES_WUNLOCK(); 2679 2680 return (0); 2681 2682 out: 2683 free(pa, M_PFRULE); 2684 return (error); 2685 } 2686 2687 int 2688 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp) 2689 { 2690 struct pf_kpool *pool; 2691 struct pf_kpooladdr *pa; 2692 2693 PF_RULES_RLOCK_TRACKER; 2694 2695 if (pp->which != PF_RDR && pp->which != PF_NAT && 2696 pp->which != PF_RT) 2697 return (EINVAL); 2698 2699 pp->anchor[sizeof(pp->anchor) - 1] = 0; 2700 pp->nr = 0; 2701 2702 PF_RULES_RLOCK(); 2703 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2704 pp->r_num, 0, 1, 0, pp->which); 2705 if (pool == NULL) { 2706 PF_RULES_RUNLOCK(); 2707 return (EBUSY); 2708 } 2709 TAILQ_FOREACH(pa, &pool->list, entries) 2710 pp->nr++; 2711 PF_RULES_RUNLOCK(); 2712 2713 return (0); 2714 } 2715 2716 int 2717 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp) 2718 { 2719 struct pf_kpool *pool; 2720 struct pf_kpooladdr *pa; 2721 u_int32_t nr = 0; 2722 2723 if (pp->which != PF_RDR && pp->which != PF_NAT && 2724 pp->which != PF_RT) 2725 return (EINVAL); 2726 2727 PF_RULES_RLOCK_TRACKER; 2728 2729 pp->anchor[sizeof(pp->anchor) - 1] = '\0'; 2730 2731 PF_RULES_RLOCK(); 2732 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2733 pp->r_num, 0, 1, 1, pp->which); 2734 if (pool == NULL) { 2735 PF_RULES_RUNLOCK(); 2736 return (EBUSY); 2737 } 2738 pa = TAILQ_FIRST(&pool->list); 2739 while ((pa != NULL) && (nr < pp->nr)) { 2740 pa = TAILQ_NEXT(pa, entries); 2741 nr++; 2742 } 2743 if (pa == NULL) { 2744 PF_RULES_RUNLOCK(); 2745 return (EBUSY); 2746 } 2747 pf_kpooladdr_to_pooladdr(pa, &pp->addr); 2748 pp->af = pa->af; 2749 pf_addr_copyout(&pp->addr.addr); 2750 PF_RULES_RUNLOCK(); 2751 2752 return (0); 2753 } 2754 2755 int 2756 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr) 2757 { 2758 struct pf_kruleset *ruleset; 2759 struct pf_kanchor *anchor; 2760 2761 PF_RULES_RLOCK_TRACKER; 2762 2763 pr->path[sizeof(pr->path) - 1] = '\0'; 2764 2765 PF_RULES_RLOCK(); 2766 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2767 PF_RULES_RUNLOCK(); 2768 return (ENOENT); 2769 } 2770 pr->nr = 0; 2771 if (ruleset == &pf_main_ruleset) { 2772 /* XXX kludge for pf_main_ruleset */ 2773 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2774 if (anchor->parent == NULL) 2775 pr->nr++; 2776 } else { 2777 RB_FOREACH(anchor, pf_kanchor_node, 2778 &ruleset->anchor->children) 2779 pr->nr++; 2780 } 2781 PF_RULES_RUNLOCK(); 2782 2783 return (0); 2784 } 2785 2786 int 2787 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr) 2788 { 2789 struct pf_kruleset *ruleset; 2790 struct pf_kanchor *anchor; 2791 u_int32_t nr = 0; 2792 int error = 0; 2793 2794 PF_RULES_RLOCK_TRACKER; 2795 2796 PF_RULES_RLOCK(); 2797 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2798 PF_RULES_RUNLOCK(); 2799 return (ENOENT); 2800 } 2801 2802 pr->name[0] = '\0'; 2803 if (ruleset == &pf_main_ruleset) { 2804 /* XXX kludge for pf_main_ruleset */ 2805 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2806 if (anchor->parent == NULL && nr++ == pr->nr) { 2807 strlcpy(pr->name, anchor->name, 2808 sizeof(pr->name)); 2809 break; 2810 } 2811 } else { 2812 RB_FOREACH(anchor, pf_kanchor_node, 2813 &ruleset->anchor->children) 2814 if (nr++ == pr->nr) { 2815 strlcpy(pr->name, anchor->name, 2816 sizeof(pr->name)); 2817 break; 2818 } 2819 } 2820 if (!pr->name[0]) 2821 error = EBUSY; 2822 PF_RULES_RUNLOCK(); 2823 2824 return (error); 2825 } 2826 2827 int 2828 pf_ioctl_natlook(struct pfioc_natlook *pnl) 2829 { 2830 struct pf_state_key *sk; 2831 struct pf_kstate *state; 2832 struct pf_state_key_cmp key; 2833 int m = 0, direction = pnl->direction; 2834 int sidx, didx; 2835 2836 /* NATLOOK src and dst are reversed, so reverse sidx/didx */ 2837 sidx = (direction == PF_IN) ? 1 : 0; 2838 didx = (direction == PF_IN) ? 0 : 1; 2839 2840 if (!pnl->proto || 2841 PF_AZERO(&pnl->saddr, pnl->af) || 2842 PF_AZERO(&pnl->daddr, pnl->af) || 2843 ((pnl->proto == IPPROTO_TCP || 2844 pnl->proto == IPPROTO_UDP) && 2845 (!pnl->dport || !pnl->sport))) 2846 return (EINVAL); 2847 2848 switch (pnl->direction) { 2849 case PF_IN: 2850 case PF_OUT: 2851 case PF_INOUT: 2852 break; 2853 default: 2854 return (EINVAL); 2855 } 2856 2857 switch (pnl->af) { 2858 #ifdef INET 2859 case AF_INET: 2860 break; 2861 #endif /* INET */ 2862 #ifdef INET6 2863 case AF_INET6: 2864 break; 2865 #endif /* INET6 */ 2866 default: 2867 return (EAFNOSUPPORT); 2868 } 2869 2870 bzero(&key, sizeof(key)); 2871 key.af = pnl->af; 2872 key.proto = pnl->proto; 2873 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af); 2874 key.port[sidx] = pnl->sport; 2875 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af); 2876 key.port[didx] = pnl->dport; 2877 2878 state = pf_find_state_all(&key, direction, &m); 2879 if (state == NULL) 2880 return (ENOENT); 2881 2882 if (m > 1) { 2883 PF_STATE_UNLOCK(state); 2884 return (E2BIG); /* more than one state */ 2885 } 2886 2887 sk = state->key[sidx]; 2888 pf_addrcpy(&pnl->rsaddr, 2889 &sk->addr[sidx], sk->af); 2890 pnl->rsport = sk->port[sidx]; 2891 pf_addrcpy(&pnl->rdaddr, 2892 &sk->addr[didx], sk->af); 2893 pnl->rdport = sk->port[didx]; 2894 PF_STATE_UNLOCK(state); 2895 2896 return (0); 2897 } 2898 2899 static int 2900 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) 2901 { 2902 int error = 0; 2903 PF_RULES_RLOCK_TRACKER; 2904 2905 #define ERROUT_IOCTL(target, x) \ 2906 do { \ 2907 error = (x); \ 2908 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \ 2909 goto target; \ 2910 } while (0) 2911 2912 2913 /* XXX keep in sync with switch() below */ 2914 if (securelevel_gt(td->td_ucred, 2)) 2915 switch (cmd) { 2916 case DIOCGETRULES: 2917 case DIOCGETRULENV: 2918 case DIOCGETADDRS: 2919 case DIOCGETADDR: 2920 case DIOCGETSTATE: 2921 case DIOCGETSTATENV: 2922 case DIOCSETSTATUSIF: 2923 case DIOCGETSTATUSNV: 2924 case DIOCCLRSTATUS: 2925 case DIOCNATLOOK: 2926 case DIOCSETDEBUG: 2927 #ifdef COMPAT_FREEBSD14 2928 case DIOCGETSTATES: 2929 case DIOCGETSTATESV2: 2930 #endif 2931 case DIOCGETTIMEOUT: 2932 case DIOCCLRRULECTRS: 2933 case DIOCGETLIMIT: 2934 case DIOCGETALTQSV0: 2935 case DIOCGETALTQSV1: 2936 case DIOCGETALTQV0: 2937 case DIOCGETALTQV1: 2938 case DIOCGETQSTATSV0: 2939 case DIOCGETQSTATSV1: 2940 case DIOCGETRULESETS: 2941 case DIOCGETRULESET: 2942 case DIOCRGETTABLES: 2943 case DIOCRGETTSTATS: 2944 case DIOCRCLRTSTATS: 2945 case DIOCRCLRADDRS: 2946 case DIOCRADDADDRS: 2947 case DIOCRDELADDRS: 2948 case DIOCRSETADDRS: 2949 case DIOCRGETADDRS: 2950 case DIOCRGETASTATS: 2951 case DIOCRCLRASTATS: 2952 case DIOCRTSTADDRS: 2953 case DIOCOSFPGET: 2954 case DIOCGETSRCNODES: 2955 case DIOCCLRSRCNODES: 2956 case DIOCGETSYNCOOKIES: 2957 case DIOCIGETIFACES: 2958 case DIOCGIFSPEEDV0: 2959 case DIOCGIFSPEEDV1: 2960 case DIOCSETIFFLAG: 2961 case DIOCCLRIFFLAG: 2962 case DIOCGETETHRULES: 2963 case DIOCGETETHRULE: 2964 case DIOCGETETHRULESETS: 2965 case DIOCGETETHRULESET: 2966 break; 2967 case DIOCRCLRTABLES: 2968 case DIOCRADDTABLES: 2969 case DIOCRDELTABLES: 2970 case DIOCRSETTFLAGS: 2971 if (((struct pfioc_table *)addr)->pfrio_flags & 2972 PFR_FLAG_DUMMY) 2973 break; /* dummy operation ok */ 2974 return (EPERM); 2975 default: 2976 return (EPERM); 2977 } 2978 2979 if (!(flags & FWRITE)) 2980 switch (cmd) { 2981 case DIOCGETRULES: 2982 case DIOCGETADDRS: 2983 case DIOCGETADDR: 2984 case DIOCGETSTATE: 2985 case DIOCGETSTATENV: 2986 case DIOCGETSTATUSNV: 2987 #ifdef COMPAT_FREEBSD14 2988 case DIOCGETSTATES: 2989 case DIOCGETSTATESV2: 2990 #endif 2991 case DIOCGETTIMEOUT: 2992 case DIOCGETLIMIT: 2993 case DIOCGETALTQSV0: 2994 case DIOCGETALTQSV1: 2995 case DIOCGETALTQV0: 2996 case DIOCGETALTQV1: 2997 case DIOCGETQSTATSV0: 2998 case DIOCGETQSTATSV1: 2999 case DIOCGETRULESETS: 3000 case DIOCGETRULESET: 3001 case DIOCNATLOOK: 3002 case DIOCRGETTABLES: 3003 case DIOCRGETTSTATS: 3004 case DIOCRGETADDRS: 3005 case DIOCRGETASTATS: 3006 case DIOCRTSTADDRS: 3007 case DIOCOSFPGET: 3008 case DIOCGETSRCNODES: 3009 case DIOCGETSYNCOOKIES: 3010 case DIOCIGETIFACES: 3011 case DIOCGIFSPEEDV1: 3012 case DIOCGIFSPEEDV0: 3013 case DIOCGETRULENV: 3014 case DIOCGETETHRULES: 3015 case DIOCGETETHRULE: 3016 case DIOCGETETHRULESETS: 3017 case DIOCGETETHRULESET: 3018 break; 3019 case DIOCRCLRTABLES: 3020 case DIOCRADDTABLES: 3021 case DIOCRDELTABLES: 3022 case DIOCRCLRTSTATS: 3023 case DIOCRCLRADDRS: 3024 case DIOCRADDADDRS: 3025 case DIOCRDELADDRS: 3026 case DIOCRSETADDRS: 3027 case DIOCRSETTFLAGS: 3028 if (((struct pfioc_table *)addr)->pfrio_flags & 3029 PFR_FLAG_DUMMY) { 3030 flags |= FWRITE; /* need write lock for dummy */ 3031 break; /* dummy operation ok */ 3032 } 3033 return (EACCES); 3034 default: 3035 return (EACCES); 3036 } 3037 3038 CURVNET_SET(TD_TO_VNET(td)); 3039 3040 switch (cmd) { 3041 #ifdef COMPAT_FREEBSD14 3042 case DIOCSTART: 3043 error = pf_start(); 3044 break; 3045 3046 case DIOCSTOP: 3047 error = pf_stop(); 3048 break; 3049 #endif 3050 3051 case DIOCGETETHRULES: { 3052 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3053 nvlist_t *nvl; 3054 void *packed; 3055 struct pf_keth_rule *tail; 3056 struct pf_keth_ruleset *rs; 3057 u_int32_t ticket, nr; 3058 const char *anchor = ""; 3059 3060 nvl = NULL; 3061 packed = NULL; 3062 3063 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x) 3064 3065 if (nv->len > pf_ioctl_maxcount) 3066 ERROUT(ENOMEM); 3067 3068 /* Copy the request in */ 3069 packed = malloc(nv->len, M_NVLIST, M_WAITOK); 3070 error = copyin(nv->data, packed, nv->len); 3071 if (error) 3072 ERROUT(error); 3073 3074 nvl = nvlist_unpack(packed, nv->len, 0); 3075 if (nvl == NULL) 3076 ERROUT(EBADMSG); 3077 3078 if (! nvlist_exists_string(nvl, "anchor")) 3079 ERROUT(EBADMSG); 3080 3081 anchor = nvlist_get_string(nvl, "anchor"); 3082 3083 rs = pf_find_keth_ruleset(anchor); 3084 3085 nvlist_destroy(nvl); 3086 nvl = NULL; 3087 free(packed, M_NVLIST); 3088 packed = NULL; 3089 3090 if (rs == NULL) 3091 ERROUT(ENOENT); 3092 3093 /* Reply */ 3094 nvl = nvlist_create(0); 3095 if (nvl == NULL) 3096 ERROUT(ENOMEM); 3097 3098 PF_RULES_RLOCK(); 3099 3100 ticket = rs->active.ticket; 3101 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq); 3102 if (tail) 3103 nr = tail->nr + 1; 3104 else 3105 nr = 0; 3106 3107 PF_RULES_RUNLOCK(); 3108 3109 nvlist_add_number(nvl, "ticket", ticket); 3110 nvlist_add_number(nvl, "nr", nr); 3111 3112 packed = nvlist_pack(nvl, &nv->len); 3113 if (packed == NULL) 3114 ERROUT(ENOMEM); 3115 3116 if (nv->size == 0) 3117 ERROUT(0); 3118 else if (nv->size < nv->len) 3119 ERROUT(ENOSPC); 3120 3121 error = copyout(packed, nv->data, nv->len); 3122 3123 #undef ERROUT 3124 DIOCGETETHRULES_error: 3125 free(packed, M_NVLIST); 3126 nvlist_destroy(nvl); 3127 break; 3128 } 3129 3130 case DIOCGETETHRULE: { 3131 struct epoch_tracker et; 3132 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3133 nvlist_t *nvl = NULL; 3134 void *nvlpacked = NULL; 3135 struct pf_keth_rule *rule = NULL; 3136 struct pf_keth_ruleset *rs; 3137 u_int32_t ticket, nr; 3138 bool clear = false; 3139 const char *anchor; 3140 3141 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x) 3142 3143 if (nv->len > pf_ioctl_maxcount) 3144 ERROUT(ENOMEM); 3145 3146 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3147 error = copyin(nv->data, nvlpacked, nv->len); 3148 if (error) 3149 ERROUT(error); 3150 3151 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3152 if (nvl == NULL) 3153 ERROUT(EBADMSG); 3154 if (! nvlist_exists_number(nvl, "ticket")) 3155 ERROUT(EBADMSG); 3156 ticket = nvlist_get_number(nvl, "ticket"); 3157 if (! nvlist_exists_string(nvl, "anchor")) 3158 ERROUT(EBADMSG); 3159 anchor = nvlist_get_string(nvl, "anchor"); 3160 3161 if (nvlist_exists_bool(nvl, "clear")) 3162 clear = nvlist_get_bool(nvl, "clear"); 3163 3164 if (clear && !(flags & FWRITE)) 3165 ERROUT(EACCES); 3166 3167 if (! nvlist_exists_number(nvl, "nr")) 3168 ERROUT(EBADMSG); 3169 nr = nvlist_get_number(nvl, "nr"); 3170 3171 PF_RULES_RLOCK(); 3172 rs = pf_find_keth_ruleset(anchor); 3173 if (rs == NULL) { 3174 PF_RULES_RUNLOCK(); 3175 ERROUT(ENOENT); 3176 } 3177 if (ticket != rs->active.ticket) { 3178 PF_RULES_RUNLOCK(); 3179 ERROUT(EBUSY); 3180 } 3181 3182 nvlist_destroy(nvl); 3183 nvl = NULL; 3184 free(nvlpacked, M_NVLIST); 3185 nvlpacked = NULL; 3186 3187 rule = TAILQ_FIRST(rs->active.rules); 3188 while ((rule != NULL) && (rule->nr != nr)) 3189 rule = TAILQ_NEXT(rule, entries); 3190 if (rule == NULL) { 3191 PF_RULES_RUNLOCK(); 3192 ERROUT(ENOENT); 3193 } 3194 /* Make sure rule can't go away. */ 3195 NET_EPOCH_ENTER(et); 3196 PF_RULES_RUNLOCK(); 3197 nvl = pf_keth_rule_to_nveth_rule(rule); 3198 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) { 3199 NET_EPOCH_EXIT(et); 3200 ERROUT(EBUSY); 3201 } 3202 NET_EPOCH_EXIT(et); 3203 if (nvl == NULL) 3204 ERROUT(ENOMEM); 3205 3206 nvlpacked = nvlist_pack(nvl, &nv->len); 3207 if (nvlpacked == NULL) 3208 ERROUT(ENOMEM); 3209 3210 if (nv->size == 0) 3211 ERROUT(0); 3212 else if (nv->size < nv->len) 3213 ERROUT(ENOSPC); 3214 3215 error = copyout(nvlpacked, nv->data, nv->len); 3216 if (error == 0 && clear) { 3217 counter_u64_zero(rule->evaluations); 3218 for (int i = 0; i < 2; i++) { 3219 counter_u64_zero(rule->packets[i]); 3220 counter_u64_zero(rule->bytes[i]); 3221 } 3222 } 3223 3224 #undef ERROUT 3225 DIOCGETETHRULE_error: 3226 free(nvlpacked, M_NVLIST); 3227 nvlist_destroy(nvl); 3228 break; 3229 } 3230 3231 case DIOCADDETHRULE: { 3232 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3233 nvlist_t *nvl = NULL; 3234 void *nvlpacked = NULL; 3235 struct pf_keth_rule *rule = NULL, *tail = NULL; 3236 struct pf_keth_ruleset *ruleset = NULL; 3237 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL; 3238 const char *anchor = "", *anchor_call = ""; 3239 3240 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x) 3241 3242 if (nv->len > pf_ioctl_maxcount) 3243 ERROUT(ENOMEM); 3244 3245 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3246 error = copyin(nv->data, nvlpacked, nv->len); 3247 if (error) 3248 ERROUT(error); 3249 3250 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3251 if (nvl == NULL) 3252 ERROUT(EBADMSG); 3253 3254 if (! nvlist_exists_number(nvl, "ticket")) 3255 ERROUT(EBADMSG); 3256 3257 if (nvlist_exists_string(nvl, "anchor")) 3258 anchor = nvlist_get_string(nvl, "anchor"); 3259 if (nvlist_exists_string(nvl, "anchor_call")) 3260 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3261 3262 ruleset = pf_find_keth_ruleset(anchor); 3263 if (ruleset == NULL) 3264 ERROUT(EINVAL); 3265 3266 if (nvlist_get_number(nvl, "ticket") != 3267 ruleset->inactive.ticket) { 3268 DPFPRINTF(PF_DEBUG_MISC, 3269 "ticket: %d != %d", 3270 (u_int32_t)nvlist_get_number(nvl, "ticket"), 3271 ruleset->inactive.ticket); 3272 ERROUT(EBUSY); 3273 } 3274 3275 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); 3276 rule->timestamp = NULL; 3277 3278 error = pf_nveth_rule_to_keth_rule(nvl, rule); 3279 if (error != 0) 3280 ERROUT(error); 3281 3282 if (rule->ifname[0]) 3283 kif = pf_kkif_create(M_WAITOK); 3284 if (rule->bridge_to_name[0]) 3285 bridge_to_kif = pf_kkif_create(M_WAITOK); 3286 rule->evaluations = counter_u64_alloc(M_WAITOK); 3287 for (int i = 0; i < 2; i++) { 3288 rule->packets[i] = counter_u64_alloc(M_WAITOK); 3289 rule->bytes[i] = counter_u64_alloc(M_WAITOK); 3290 } 3291 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 3292 M_WAITOK | M_ZERO); 3293 3294 PF_RULES_WLOCK(); 3295 3296 if (rule->ifname[0]) { 3297 rule->kif = pfi_kkif_attach(kif, rule->ifname); 3298 pfi_kkif_ref(rule->kif); 3299 } else 3300 rule->kif = NULL; 3301 if (rule->bridge_to_name[0]) { 3302 rule->bridge_to = pfi_kkif_attach(bridge_to_kif, 3303 rule->bridge_to_name); 3304 pfi_kkif_ref(rule->bridge_to); 3305 } else 3306 rule->bridge_to = NULL; 3307 3308 #ifdef ALTQ 3309 /* set queue IDs */ 3310 if (rule->qname[0] != 0) { 3311 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 3312 error = EBUSY; 3313 else 3314 rule->qid = rule->qid; 3315 } 3316 #endif 3317 if (rule->tagname[0]) 3318 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 3319 error = EBUSY; 3320 if (rule->match_tagname[0]) 3321 if ((rule->match_tag = pf_tagname2tag( 3322 rule->match_tagname)) == 0) 3323 error = EBUSY; 3324 3325 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE) 3326 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr); 3327 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE) 3328 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr); 3329 3330 if (error) { 3331 pf_free_eth_rule(rule); 3332 PF_RULES_WUNLOCK(); 3333 ERROUT(error); 3334 } 3335 3336 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) { 3337 pf_free_eth_rule(rule); 3338 PF_RULES_WUNLOCK(); 3339 ERROUT(EINVAL); 3340 } 3341 3342 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq); 3343 if (tail) 3344 rule->nr = tail->nr + 1; 3345 else 3346 rule->nr = 0; 3347 3348 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries); 3349 3350 PF_RULES_WUNLOCK(); 3351 3352 #undef ERROUT 3353 DIOCADDETHRULE_error: 3354 nvlist_destroy(nvl); 3355 free(nvlpacked, M_NVLIST); 3356 break; 3357 } 3358 3359 case DIOCGETETHRULESETS: { 3360 struct epoch_tracker et; 3361 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3362 nvlist_t *nvl = NULL; 3363 void *nvlpacked = NULL; 3364 struct pf_keth_ruleset *ruleset; 3365 struct pf_keth_anchor *anchor; 3366 int nr = 0; 3367 3368 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x) 3369 3370 if (nv->len > pf_ioctl_maxcount) 3371 ERROUT(ENOMEM); 3372 3373 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3374 error = copyin(nv->data, nvlpacked, nv->len); 3375 if (error) 3376 ERROUT(error); 3377 3378 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3379 if (nvl == NULL) 3380 ERROUT(EBADMSG); 3381 if (! nvlist_exists_string(nvl, "path")) 3382 ERROUT(EBADMSG); 3383 3384 NET_EPOCH_ENTER(et); 3385 3386 if ((ruleset = pf_find_keth_ruleset( 3387 nvlist_get_string(nvl, "path"))) == NULL) { 3388 NET_EPOCH_EXIT(et); 3389 ERROUT(ENOENT); 3390 } 3391 3392 if (ruleset->anchor == NULL) { 3393 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors) 3394 if (anchor->parent == NULL) 3395 nr++; 3396 } else { 3397 RB_FOREACH(anchor, pf_keth_anchor_node, 3398 &ruleset->anchor->children) 3399 nr++; 3400 } 3401 3402 NET_EPOCH_EXIT(et); 3403 3404 nvlist_destroy(nvl); 3405 nvl = NULL; 3406 free(nvlpacked, M_NVLIST); 3407 nvlpacked = NULL; 3408 3409 nvl = nvlist_create(0); 3410 if (nvl == NULL) 3411 ERROUT(ENOMEM); 3412 3413 nvlist_add_number(nvl, "nr", nr); 3414 3415 nvlpacked = nvlist_pack(nvl, &nv->len); 3416 if (nvlpacked == NULL) 3417 ERROUT(ENOMEM); 3418 3419 if (nv->size == 0) 3420 ERROUT(0); 3421 else if (nv->size < nv->len) 3422 ERROUT(ENOSPC); 3423 3424 error = copyout(nvlpacked, nv->data, nv->len); 3425 3426 #undef ERROUT 3427 DIOCGETETHRULESETS_error: 3428 free(nvlpacked, M_NVLIST); 3429 nvlist_destroy(nvl); 3430 break; 3431 } 3432 3433 case DIOCGETETHRULESET: { 3434 struct epoch_tracker et; 3435 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3436 nvlist_t *nvl = NULL; 3437 void *nvlpacked = NULL; 3438 struct pf_keth_ruleset *ruleset; 3439 struct pf_keth_anchor *anchor; 3440 int nr = 0, req_nr = 0; 3441 bool found = false; 3442 3443 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x) 3444 3445 if (nv->len > pf_ioctl_maxcount) 3446 ERROUT(ENOMEM); 3447 3448 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3449 error = copyin(nv->data, nvlpacked, nv->len); 3450 if (error) 3451 ERROUT(error); 3452 3453 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3454 if (nvl == NULL) 3455 ERROUT(EBADMSG); 3456 if (! nvlist_exists_string(nvl, "path")) 3457 ERROUT(EBADMSG); 3458 if (! nvlist_exists_number(nvl, "nr")) 3459 ERROUT(EBADMSG); 3460 3461 req_nr = nvlist_get_number(nvl, "nr"); 3462 3463 NET_EPOCH_ENTER(et); 3464 3465 if ((ruleset = pf_find_keth_ruleset( 3466 nvlist_get_string(nvl, "path"))) == NULL) { 3467 NET_EPOCH_EXIT(et); 3468 ERROUT(ENOENT); 3469 } 3470 3471 nvlist_destroy(nvl); 3472 nvl = NULL; 3473 free(nvlpacked, M_NVLIST); 3474 nvlpacked = NULL; 3475 3476 nvl = nvlist_create(0); 3477 if (nvl == NULL) { 3478 NET_EPOCH_EXIT(et); 3479 ERROUT(ENOMEM); 3480 } 3481 3482 if (ruleset->anchor == NULL) { 3483 RB_FOREACH(anchor, pf_keth_anchor_global, 3484 &V_pf_keth_anchors) { 3485 if (anchor->parent == NULL && nr++ == req_nr) { 3486 found = true; 3487 break; 3488 } 3489 } 3490 } else { 3491 RB_FOREACH(anchor, pf_keth_anchor_node, 3492 &ruleset->anchor->children) { 3493 if (nr++ == req_nr) { 3494 found = true; 3495 break; 3496 } 3497 } 3498 } 3499 3500 NET_EPOCH_EXIT(et); 3501 if (found) { 3502 nvlist_add_number(nvl, "nr", nr); 3503 nvlist_add_string(nvl, "name", anchor->name); 3504 if (ruleset->anchor) 3505 nvlist_add_string(nvl, "path", 3506 ruleset->anchor->path); 3507 else 3508 nvlist_add_string(nvl, "path", ""); 3509 } else { 3510 ERROUT(EBUSY); 3511 } 3512 3513 nvlpacked = nvlist_pack(nvl, &nv->len); 3514 if (nvlpacked == NULL) 3515 ERROUT(ENOMEM); 3516 3517 if (nv->size == 0) 3518 ERROUT(0); 3519 else if (nv->size < nv->len) 3520 ERROUT(ENOSPC); 3521 3522 error = copyout(nvlpacked, nv->data, nv->len); 3523 3524 #undef ERROUT 3525 DIOCGETETHRULESET_error: 3526 free(nvlpacked, M_NVLIST); 3527 nvlist_destroy(nvl); 3528 break; 3529 } 3530 3531 case DIOCADDRULENV: { 3532 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3533 nvlist_t *nvl = NULL; 3534 void *nvlpacked = NULL; 3535 struct pf_krule *rule = NULL; 3536 const char *anchor = "", *anchor_call = ""; 3537 uint32_t ticket = 0, pool_ticket = 0; 3538 3539 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x) 3540 3541 if (nv->len > pf_ioctl_maxcount) 3542 ERROUT(ENOMEM); 3543 3544 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3545 error = copyin(nv->data, nvlpacked, nv->len); 3546 if (error) 3547 ERROUT(error); 3548 3549 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3550 if (nvl == NULL) 3551 ERROUT(EBADMSG); 3552 3553 if (! nvlist_exists_number(nvl, "ticket")) 3554 ERROUT(EINVAL); 3555 ticket = nvlist_get_number(nvl, "ticket"); 3556 3557 if (! nvlist_exists_number(nvl, "pool_ticket")) 3558 ERROUT(EINVAL); 3559 pool_ticket = nvlist_get_number(nvl, "pool_ticket"); 3560 3561 if (! nvlist_exists_nvlist(nvl, "rule")) 3562 ERROUT(EINVAL); 3563 3564 rule = pf_krule_alloc(); 3565 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"), 3566 rule); 3567 if (error) 3568 ERROUT(error); 3569 3570 if (nvlist_exists_string(nvl, "anchor")) 3571 anchor = nvlist_get_string(nvl, "anchor"); 3572 if (nvlist_exists_string(nvl, "anchor_call")) 3573 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3574 3575 if ((error = nvlist_error(nvl))) 3576 ERROUT(error); 3577 3578 /* Frees rule on error */ 3579 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor, 3580 anchor_call, td->td_ucred->cr_ruid, 3581 td->td_proc ? td->td_proc->p_pid : 0); 3582 3583 nvlist_destroy(nvl); 3584 free(nvlpacked, M_NVLIST); 3585 break; 3586 #undef ERROUT 3587 DIOCADDRULENV_error: 3588 pf_krule_free(rule); 3589 nvlist_destroy(nvl); 3590 free(nvlpacked, M_NVLIST); 3591 3592 break; 3593 } 3594 case DIOCADDRULE: { 3595 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3596 struct pf_krule *rule; 3597 3598 rule = pf_krule_alloc(); 3599 error = pf_rule_to_krule(&pr->rule, rule); 3600 if (error != 0) { 3601 pf_krule_free(rule); 3602 goto fail; 3603 } 3604 3605 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3606 3607 /* Frees rule on error */ 3608 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, 3609 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid, 3610 td->td_proc ? td->td_proc->p_pid : 0); 3611 break; 3612 } 3613 3614 case DIOCGETRULES: { 3615 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3616 3617 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3618 3619 error = pf_ioctl_getrules(pr); 3620 3621 break; 3622 } 3623 3624 case DIOCGETRULENV: { 3625 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3626 nvlist_t *nvrule = NULL; 3627 nvlist_t *nvl = NULL; 3628 struct pf_kruleset *ruleset; 3629 struct pf_krule *rule; 3630 void *nvlpacked = NULL; 3631 int rs_num, nr; 3632 bool clear_counter = false; 3633 3634 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x) 3635 3636 if (nv->len > pf_ioctl_maxcount) 3637 ERROUT(ENOMEM); 3638 3639 /* Copy the request in */ 3640 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3641 error = copyin(nv->data, nvlpacked, nv->len); 3642 if (error) 3643 ERROUT(error); 3644 3645 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3646 if (nvl == NULL) 3647 ERROUT(EBADMSG); 3648 3649 if (! nvlist_exists_string(nvl, "anchor")) 3650 ERROUT(EBADMSG); 3651 if (! nvlist_exists_number(nvl, "ruleset")) 3652 ERROUT(EBADMSG); 3653 if (! nvlist_exists_number(nvl, "ticket")) 3654 ERROUT(EBADMSG); 3655 if (! nvlist_exists_number(nvl, "nr")) 3656 ERROUT(EBADMSG); 3657 3658 if (nvlist_exists_bool(nvl, "clear_counter")) 3659 clear_counter = nvlist_get_bool(nvl, "clear_counter"); 3660 3661 if (clear_counter && !(flags & FWRITE)) 3662 ERROUT(EACCES); 3663 3664 nr = nvlist_get_number(nvl, "nr"); 3665 3666 PF_RULES_WLOCK(); 3667 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor")); 3668 if (ruleset == NULL) { 3669 PF_RULES_WUNLOCK(); 3670 ERROUT(ENOENT); 3671 } 3672 3673 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset")); 3674 if (rs_num >= PF_RULESET_MAX) { 3675 PF_RULES_WUNLOCK(); 3676 ERROUT(EINVAL); 3677 } 3678 3679 if (nvlist_get_number(nvl, "ticket") != 3680 ruleset->rules[rs_num].active.ticket) { 3681 PF_RULES_WUNLOCK(); 3682 ERROUT(EBUSY); 3683 } 3684 3685 if ((error = nvlist_error(nvl))) { 3686 PF_RULES_WUNLOCK(); 3687 ERROUT(error); 3688 } 3689 3690 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3691 while ((rule != NULL) && (rule->nr != nr)) 3692 rule = TAILQ_NEXT(rule, entries); 3693 if (rule == NULL) { 3694 PF_RULES_WUNLOCK(); 3695 ERROUT(EBUSY); 3696 } 3697 3698 nvrule = pf_krule_to_nvrule(rule); 3699 3700 nvlist_destroy(nvl); 3701 nvl = nvlist_create(0); 3702 if (nvl == NULL) { 3703 PF_RULES_WUNLOCK(); 3704 ERROUT(ENOMEM); 3705 } 3706 nvlist_add_number(nvl, "nr", nr); 3707 nvlist_add_nvlist(nvl, "rule", nvrule); 3708 nvlist_destroy(nvrule); 3709 nvrule = NULL; 3710 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) { 3711 PF_RULES_WUNLOCK(); 3712 ERROUT(EBUSY); 3713 } 3714 3715 free(nvlpacked, M_NVLIST); 3716 nvlpacked = nvlist_pack(nvl, &nv->len); 3717 if (nvlpacked == NULL) { 3718 PF_RULES_WUNLOCK(); 3719 ERROUT(ENOMEM); 3720 } 3721 3722 if (nv->size == 0) { 3723 PF_RULES_WUNLOCK(); 3724 ERROUT(0); 3725 } 3726 else if (nv->size < nv->len) { 3727 PF_RULES_WUNLOCK(); 3728 ERROUT(ENOSPC); 3729 } 3730 3731 if (clear_counter) 3732 pf_krule_clear_counters(rule); 3733 3734 PF_RULES_WUNLOCK(); 3735 3736 error = copyout(nvlpacked, nv->data, nv->len); 3737 3738 #undef ERROUT 3739 DIOCGETRULENV_error: 3740 free(nvlpacked, M_NVLIST); 3741 nvlist_destroy(nvrule); 3742 nvlist_destroy(nvl); 3743 3744 break; 3745 } 3746 3747 case DIOCCHANGERULE: { 3748 struct pfioc_rule *pcr = (struct pfioc_rule *)addr; 3749 struct pf_kruleset *ruleset; 3750 struct pf_krule *oldrule = NULL, *newrule = NULL; 3751 struct pfi_kkif *kif = NULL; 3752 struct pf_kpooladdr *pa; 3753 u_int32_t nr = 0; 3754 int rs_num; 3755 3756 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0'; 3757 3758 if (pcr->action < PF_CHANGE_ADD_HEAD || 3759 pcr->action > PF_CHANGE_GET_TICKET) { 3760 error = EINVAL; 3761 goto fail; 3762 } 3763 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { 3764 error = EINVAL; 3765 goto fail; 3766 } 3767 3768 if (pcr->action != PF_CHANGE_REMOVE) { 3769 newrule = pf_krule_alloc(); 3770 error = pf_rule_to_krule(&pcr->rule, newrule); 3771 if (error != 0) { 3772 pf_krule_free(newrule); 3773 goto fail; 3774 } 3775 3776 if ((error = pf_rule_checkaf(newrule))) { 3777 pf_krule_free(newrule); 3778 goto fail; 3779 } 3780 if (newrule->ifname[0]) 3781 kif = pf_kkif_create(M_WAITOK); 3782 pf_counter_u64_init(&newrule->evaluations, M_WAITOK); 3783 for (int i = 0; i < 2; i++) { 3784 pf_counter_u64_init(&newrule->packets[i], M_WAITOK); 3785 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK); 3786 } 3787 newrule->states_cur = counter_u64_alloc(M_WAITOK); 3788 newrule->states_tot = counter_u64_alloc(M_WAITOK); 3789 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 3790 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 3791 newrule->cuid = td->td_ucred->cr_ruid; 3792 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 3793 TAILQ_INIT(&newrule->nat.list); 3794 TAILQ_INIT(&newrule->rdr.list); 3795 TAILQ_INIT(&newrule->route.list); 3796 } 3797 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x) 3798 3799 PF_CONFIG_LOCK(); 3800 PF_RULES_WLOCK(); 3801 #ifdef PF_WANT_32_TO_64_COUNTER 3802 if (newrule != NULL) { 3803 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist); 3804 newrule->allrulelinked = true; 3805 V_pf_allrulecount++; 3806 } 3807 #endif 3808 3809 if (!(pcr->action == PF_CHANGE_REMOVE || 3810 pcr->action == PF_CHANGE_GET_TICKET) && 3811 pcr->pool_ticket != V_ticket_pabuf) 3812 ERROUT(EBUSY); 3813 3814 ruleset = pf_find_kruleset(pcr->anchor); 3815 if (ruleset == NULL) 3816 ERROUT(EINVAL); 3817 3818 rs_num = pf_get_ruleset_number(pcr->rule.action); 3819 if (rs_num >= PF_RULESET_MAX) 3820 ERROUT(EINVAL); 3821 3822 /* 3823 * XXXMJG: there is no guarantee that the ruleset was 3824 * created by the usual route of calling DIOCXBEGIN. 3825 * As a result it is possible the rule tree will not 3826 * be allocated yet. Hack around it by doing it here. 3827 * Note it is fine to let the tree persist in case of 3828 * error as it will be freed down the road on future 3829 * updates (if need be). 3830 */ 3831 if (ruleset->rules[rs_num].active.tree == NULL) { 3832 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT); 3833 if (ruleset->rules[rs_num].active.tree == NULL) { 3834 ERROUT(ENOMEM); 3835 } 3836 } 3837 3838 if (pcr->action == PF_CHANGE_GET_TICKET) { 3839 pcr->ticket = ++ruleset->rules[rs_num].active.ticket; 3840 ERROUT(0); 3841 } else if (pcr->ticket != 3842 ruleset->rules[rs_num].active.ticket) 3843 ERROUT(EINVAL); 3844 3845 if (pcr->action != PF_CHANGE_REMOVE) { 3846 if (newrule->ifname[0]) { 3847 newrule->kif = pfi_kkif_attach(kif, 3848 newrule->ifname); 3849 kif = NULL; 3850 pfi_kkif_ref(newrule->kif); 3851 } else 3852 newrule->kif = NULL; 3853 3854 if (newrule->rtableid > 0 && 3855 newrule->rtableid >= rt_numfibs) 3856 error = EBUSY; 3857 3858 #ifdef ALTQ 3859 /* set queue IDs */ 3860 if (newrule->qname[0] != 0) { 3861 if ((newrule->qid = 3862 pf_qname2qid(newrule->qname)) == 0) 3863 error = EBUSY; 3864 else if (newrule->pqname[0] != 0) { 3865 if ((newrule->pqid = 3866 pf_qname2qid(newrule->pqname)) == 0) 3867 error = EBUSY; 3868 } else 3869 newrule->pqid = newrule->qid; 3870 } 3871 #endif /* ALTQ */ 3872 if (newrule->tagname[0]) 3873 if ((newrule->tag = 3874 pf_tagname2tag(newrule->tagname)) == 0) 3875 error = EBUSY; 3876 if (newrule->match_tagname[0]) 3877 if ((newrule->match_tag = pf_tagname2tag( 3878 newrule->match_tagname)) == 0) 3879 error = EBUSY; 3880 if (newrule->rt && !newrule->direction) 3881 error = EINVAL; 3882 if (!newrule->log) 3883 newrule->logif = 0; 3884 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) 3885 error = ENOMEM; 3886 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) 3887 error = ENOMEM; 3888 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call)) 3889 error = EINVAL; 3890 for (int i = 0; i < 3; i++) { 3891 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 3892 if (pa->addr.type == PF_ADDR_TABLE) { 3893 pa->addr.p.tbl = 3894 pfr_attach_table(ruleset, 3895 pa->addr.v.tblname); 3896 if (pa->addr.p.tbl == NULL) 3897 error = ENOMEM; 3898 } 3899 } 3900 3901 newrule->overload_tbl = NULL; 3902 if (newrule->overload_tblname[0]) { 3903 if ((newrule->overload_tbl = pfr_attach_table( 3904 ruleset, newrule->overload_tblname)) == 3905 NULL) 3906 error = EINVAL; 3907 else 3908 newrule->overload_tbl->pfrkt_flags |= 3909 PFR_TFLAG_ACTIVE; 3910 } 3911 3912 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list); 3913 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list); 3914 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list); 3915 if (((((newrule->action == PF_NAT) || 3916 (newrule->action == PF_RDR) || 3917 (newrule->action == PF_BINAT) || 3918 (newrule->rt > PF_NOPFROUTE)) && 3919 !newrule->anchor)) && 3920 (TAILQ_FIRST(&newrule->rdr.list) == NULL)) 3921 error = EINVAL; 3922 3923 if (error) { 3924 pf_free_rule(newrule); 3925 PF_RULES_WUNLOCK(); 3926 PF_CONFIG_UNLOCK(); 3927 goto fail; 3928 } 3929 3930 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list); 3931 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list); 3932 } 3933 pf_empty_kpool(&V_pf_pabuf[0]); 3934 pf_empty_kpool(&V_pf_pabuf[1]); 3935 pf_empty_kpool(&V_pf_pabuf[2]); 3936 3937 if (pcr->action == PF_CHANGE_ADD_HEAD) 3938 oldrule = TAILQ_FIRST( 3939 ruleset->rules[rs_num].active.ptr); 3940 else if (pcr->action == PF_CHANGE_ADD_TAIL) 3941 oldrule = TAILQ_LAST( 3942 ruleset->rules[rs_num].active.ptr, pf_krulequeue); 3943 else { 3944 oldrule = TAILQ_FIRST( 3945 ruleset->rules[rs_num].active.ptr); 3946 while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) 3947 oldrule = TAILQ_NEXT(oldrule, entries); 3948 if (oldrule == NULL) { 3949 if (newrule != NULL) 3950 pf_free_rule(newrule); 3951 PF_RULES_WUNLOCK(); 3952 PF_CONFIG_UNLOCK(); 3953 error = EINVAL; 3954 goto fail; 3955 } 3956 } 3957 3958 if (pcr->action == PF_CHANGE_REMOVE) { 3959 pf_unlink_rule(ruleset->rules[rs_num].active.ptr, 3960 oldrule); 3961 RB_REMOVE(pf_krule_global, 3962 ruleset->rules[rs_num].active.tree, oldrule); 3963 ruleset->rules[rs_num].active.rcount--; 3964 } else { 3965 pf_hash_rule(newrule); 3966 if (RB_INSERT(pf_krule_global, 3967 ruleset->rules[rs_num].active.tree, newrule) != NULL) { 3968 pf_free_rule(newrule); 3969 PF_RULES_WUNLOCK(); 3970 PF_CONFIG_UNLOCK(); 3971 error = EEXIST; 3972 goto fail; 3973 } 3974 3975 if (oldrule == NULL) 3976 TAILQ_INSERT_TAIL( 3977 ruleset->rules[rs_num].active.ptr, 3978 newrule, entries); 3979 else if (pcr->action == PF_CHANGE_ADD_HEAD || 3980 pcr->action == PF_CHANGE_ADD_BEFORE) 3981 TAILQ_INSERT_BEFORE(oldrule, newrule, entries); 3982 else 3983 TAILQ_INSERT_AFTER( 3984 ruleset->rules[rs_num].active.ptr, 3985 oldrule, newrule, entries); 3986 ruleset->rules[rs_num].active.rcount++; 3987 } 3988 3989 nr = 0; 3990 TAILQ_FOREACH(oldrule, 3991 ruleset->rules[rs_num].active.ptr, entries) 3992 oldrule->nr = nr++; 3993 3994 ruleset->rules[rs_num].active.ticket++; 3995 3996 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); 3997 pf_remove_if_empty_kruleset(ruleset); 3998 3999 PF_RULES_WUNLOCK(); 4000 PF_CONFIG_UNLOCK(); 4001 break; 4002 4003 #undef ERROUT 4004 DIOCCHANGERULE_error: 4005 PF_RULES_WUNLOCK(); 4006 PF_CONFIG_UNLOCK(); 4007 pf_krule_free(newrule); 4008 pf_kkif_free(kif); 4009 break; 4010 } 4011 4012 case DIOCCLRSTATESNV: { 4013 error = pf_clearstates_nv((struct pfioc_nv *)addr); 4014 break; 4015 } 4016 4017 case DIOCKILLSTATESNV: { 4018 error = pf_killstates_nv((struct pfioc_nv *)addr); 4019 break; 4020 } 4021 4022 case DIOCADDSTATE: { 4023 struct pfioc_state *ps = (struct pfioc_state *)addr; 4024 struct pfsync_state_1301 *sp = &ps->state; 4025 4026 if (sp->timeout >= PFTM_MAX) { 4027 error = EINVAL; 4028 goto fail; 4029 } 4030 if (V_pfsync_state_import_ptr != NULL) { 4031 PF_RULES_RLOCK(); 4032 error = V_pfsync_state_import_ptr( 4033 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL, 4034 PFSYNC_MSG_VERSION_1301); 4035 PF_RULES_RUNLOCK(); 4036 } else 4037 error = EOPNOTSUPP; 4038 break; 4039 } 4040 4041 case DIOCGETSTATE: { 4042 struct pfioc_state *ps = (struct pfioc_state *)addr; 4043 struct pf_kstate *s; 4044 4045 s = pf_find_state_byid(ps->state.id, ps->state.creatorid); 4046 if (s == NULL) { 4047 error = ENOENT; 4048 goto fail; 4049 } 4050 4051 pfsync_state_export((union pfsync_state_union*)&ps->state, 4052 s, PFSYNC_MSG_VERSION_1301); 4053 PF_STATE_UNLOCK(s); 4054 break; 4055 } 4056 4057 case DIOCGETSTATENV: { 4058 error = pf_getstate((struct pfioc_nv *)addr); 4059 break; 4060 } 4061 4062 #ifdef COMPAT_FREEBSD14 4063 case DIOCGETSTATES: { 4064 struct pfioc_states *ps = (struct pfioc_states *)addr; 4065 struct pf_kstate *s; 4066 struct pfsync_state_1301 *pstore, *p; 4067 int i, nr; 4068 size_t slice_count = 16, count; 4069 void *out; 4070 4071 if (ps->ps_len <= 0) { 4072 nr = uma_zone_get_cur(V_pf_state_z); 4073 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4074 break; 4075 } 4076 4077 out = ps->ps_states; 4078 pstore = mallocarray(slice_count, 4079 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO); 4080 nr = 0; 4081 4082 for (i = 0; i <= V_pf_hashmask; i++) { 4083 struct pf_idhash *ih = &V_pf_idhash[i]; 4084 4085 DIOCGETSTATES_retry: 4086 p = pstore; 4087 4088 if (LIST_EMPTY(&ih->states)) 4089 continue; 4090 4091 PF_HASHROW_LOCK(ih); 4092 count = 0; 4093 LIST_FOREACH(s, &ih->states, entry) { 4094 if (s->timeout == PFTM_UNLINKED) 4095 continue; 4096 count++; 4097 } 4098 4099 if (count > slice_count) { 4100 PF_HASHROW_UNLOCK(ih); 4101 free(pstore, M_TEMP); 4102 slice_count = count * 2; 4103 pstore = mallocarray(slice_count, 4104 sizeof(struct pfsync_state_1301), M_TEMP, 4105 M_WAITOK | M_ZERO); 4106 goto DIOCGETSTATES_retry; 4107 } 4108 4109 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4110 PF_HASHROW_UNLOCK(ih); 4111 goto DIOCGETSTATES_full; 4112 } 4113 4114 LIST_FOREACH(s, &ih->states, entry) { 4115 if (s->timeout == PFTM_UNLINKED) 4116 continue; 4117 4118 pfsync_state_export((union pfsync_state_union*)p, 4119 s, PFSYNC_MSG_VERSION_1301); 4120 p++; 4121 nr++; 4122 } 4123 PF_HASHROW_UNLOCK(ih); 4124 error = copyout(pstore, out, 4125 sizeof(struct pfsync_state_1301) * count); 4126 if (error) 4127 goto fail; 4128 out = ps->ps_states + nr; 4129 } 4130 DIOCGETSTATES_full: 4131 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4132 free(pstore, M_TEMP); 4133 4134 break; 4135 } 4136 4137 case DIOCGETSTATESV2: { 4138 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr; 4139 struct pf_kstate *s; 4140 struct pf_state_export *pstore, *p; 4141 int i, nr; 4142 size_t slice_count = 16, count; 4143 void *out; 4144 4145 if (ps->ps_req_version > PF_STATE_VERSION) { 4146 error = ENOTSUP; 4147 goto fail; 4148 } 4149 4150 if (ps->ps_len <= 0) { 4151 nr = uma_zone_get_cur(V_pf_state_z); 4152 ps->ps_len = sizeof(struct pf_state_export) * nr; 4153 break; 4154 } 4155 4156 out = ps->ps_states; 4157 pstore = mallocarray(slice_count, 4158 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO); 4159 nr = 0; 4160 4161 for (i = 0; i <= V_pf_hashmask; i++) { 4162 struct pf_idhash *ih = &V_pf_idhash[i]; 4163 4164 DIOCGETSTATESV2_retry: 4165 p = pstore; 4166 4167 if (LIST_EMPTY(&ih->states)) 4168 continue; 4169 4170 PF_HASHROW_LOCK(ih); 4171 count = 0; 4172 LIST_FOREACH(s, &ih->states, entry) { 4173 if (s->timeout == PFTM_UNLINKED) 4174 continue; 4175 count++; 4176 } 4177 4178 if (count > slice_count) { 4179 PF_HASHROW_UNLOCK(ih); 4180 free(pstore, M_TEMP); 4181 slice_count = count * 2; 4182 pstore = mallocarray(slice_count, 4183 sizeof(struct pf_state_export), M_TEMP, 4184 M_WAITOK | M_ZERO); 4185 goto DIOCGETSTATESV2_retry; 4186 } 4187 4188 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4189 PF_HASHROW_UNLOCK(ih); 4190 goto DIOCGETSTATESV2_full; 4191 } 4192 4193 LIST_FOREACH(s, &ih->states, entry) { 4194 if (s->timeout == PFTM_UNLINKED) 4195 continue; 4196 4197 pf_state_export(p, s); 4198 p++; 4199 nr++; 4200 } 4201 PF_HASHROW_UNLOCK(ih); 4202 error = copyout(pstore, out, 4203 sizeof(struct pf_state_export) * count); 4204 if (error) 4205 goto fail; 4206 out = ps->ps_states + nr; 4207 } 4208 DIOCGETSTATESV2_full: 4209 ps->ps_len = nr * sizeof(struct pf_state_export); 4210 free(pstore, M_TEMP); 4211 4212 break; 4213 } 4214 #endif 4215 case DIOCGETSTATUSNV: { 4216 error = pf_getstatus((struct pfioc_nv *)addr); 4217 break; 4218 } 4219 4220 case DIOCSETSTATUSIF: { 4221 struct pfioc_if *pi = (struct pfioc_if *)addr; 4222 4223 if (pi->ifname[0] == 0) { 4224 bzero(V_pf_status.ifname, IFNAMSIZ); 4225 break; 4226 } 4227 PF_RULES_WLOCK(); 4228 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); 4229 PF_RULES_WUNLOCK(); 4230 break; 4231 } 4232 4233 case DIOCCLRSTATUS: { 4234 pf_ioctl_clear_status(); 4235 break; 4236 } 4237 4238 case DIOCNATLOOK: { 4239 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; 4240 4241 error = pf_ioctl_natlook(pnl); 4242 break; 4243 } 4244 4245 case DIOCSETTIMEOUT: { 4246 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4247 4248 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds, 4249 &pt->seconds); 4250 break; 4251 } 4252 4253 case DIOCGETTIMEOUT: { 4254 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4255 4256 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds); 4257 break; 4258 } 4259 4260 case DIOCGETLIMIT: { 4261 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4262 4263 error = pf_ioctl_get_limit(pl->index, &pl->limit); 4264 break; 4265 } 4266 4267 case DIOCSETLIMIT: { 4268 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4269 unsigned int old_limit; 4270 4271 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit); 4272 pl->limit = old_limit; 4273 break; 4274 } 4275 4276 case DIOCSETDEBUG: { 4277 u_int32_t *level = (u_int32_t *)addr; 4278 4279 PF_RULES_WLOCK(); 4280 V_pf_status.debug = *level; 4281 PF_RULES_WUNLOCK(); 4282 break; 4283 } 4284 4285 case DIOCCLRRULECTRS: { 4286 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ 4287 struct pf_kruleset *ruleset = &pf_main_ruleset; 4288 struct pf_krule *rule; 4289 4290 PF_RULES_WLOCK(); 4291 TAILQ_FOREACH(rule, 4292 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { 4293 pf_counter_u64_zero(&rule->evaluations); 4294 for (int i = 0; i < 2; i++) { 4295 pf_counter_u64_zero(&rule->packets[i]); 4296 pf_counter_u64_zero(&rule->bytes[i]); 4297 } 4298 } 4299 PF_RULES_WUNLOCK(); 4300 break; 4301 } 4302 4303 case DIOCGIFSPEEDV0: 4304 case DIOCGIFSPEEDV1: { 4305 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr; 4306 struct pf_ifspeed_v1 ps; 4307 struct ifnet *ifp; 4308 4309 if (psp->ifname[0] == '\0') { 4310 error = EINVAL; 4311 goto fail; 4312 } 4313 4314 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); 4315 if (error != 0) 4316 goto fail; 4317 ifp = ifunit(ps.ifname); 4318 if (ifp != NULL) { 4319 psp->baudrate32 = 4320 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX); 4321 if (cmd == DIOCGIFSPEEDV1) 4322 psp->baudrate = ifp->if_baudrate; 4323 } else { 4324 error = EINVAL; 4325 } 4326 break; 4327 } 4328 4329 #ifdef ALTQ 4330 case DIOCSTARTALTQ: { 4331 struct pf_altq *altq; 4332 4333 PF_RULES_WLOCK(); 4334 /* enable all altq interfaces on active list */ 4335 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4336 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4337 error = pf_enable_altq(altq); 4338 if (error != 0) 4339 break; 4340 } 4341 } 4342 if (error == 0) 4343 V_pf_altq_running = 1; 4344 PF_RULES_WUNLOCK(); 4345 DPFPRINTF(PF_DEBUG_MISC, "altq: started"); 4346 break; 4347 } 4348 4349 case DIOCSTOPALTQ: { 4350 struct pf_altq *altq; 4351 4352 PF_RULES_WLOCK(); 4353 /* disable all altq interfaces on active list */ 4354 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4355 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4356 error = pf_disable_altq(altq); 4357 if (error != 0) 4358 break; 4359 } 4360 } 4361 if (error == 0) 4362 V_pf_altq_running = 0; 4363 PF_RULES_WUNLOCK(); 4364 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped"); 4365 break; 4366 } 4367 4368 case DIOCADDALTQV0: 4369 case DIOCADDALTQV1: { 4370 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4371 struct pf_altq *altq, *a; 4372 struct ifnet *ifp; 4373 4374 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); 4375 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); 4376 if (error) 4377 goto fail; 4378 altq->local_flags = 0; 4379 4380 PF_RULES_WLOCK(); 4381 if (pa->ticket != V_ticket_altqs_inactive) { 4382 PF_RULES_WUNLOCK(); 4383 free(altq, M_PFALTQ); 4384 error = EBUSY; 4385 goto fail; 4386 } 4387 4388 /* 4389 * if this is for a queue, find the discipline and 4390 * copy the necessary fields 4391 */ 4392 if (altq->qname[0] != 0) { 4393 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) { 4394 PF_RULES_WUNLOCK(); 4395 error = EBUSY; 4396 free(altq, M_PFALTQ); 4397 goto fail; 4398 } 4399 altq->altq_disc = NULL; 4400 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { 4401 if (strncmp(a->ifname, altq->ifname, 4402 IFNAMSIZ) == 0) { 4403 altq->altq_disc = a->altq_disc; 4404 break; 4405 } 4406 } 4407 } 4408 4409 if ((ifp = ifunit(altq->ifname)) == NULL) 4410 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 4411 else 4412 error = altq_add(ifp, altq); 4413 4414 if (error) { 4415 PF_RULES_WUNLOCK(); 4416 free(altq, M_PFALTQ); 4417 goto fail; 4418 } 4419 4420 if (altq->qname[0] != 0) 4421 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); 4422 else 4423 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries); 4424 /* version error check done on import above */ 4425 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4426 PF_RULES_WUNLOCK(); 4427 break; 4428 } 4429 4430 case DIOCGETALTQSV0: 4431 case DIOCGETALTQSV1: { 4432 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4433 struct pf_altq *altq; 4434 4435 PF_RULES_RLOCK(); 4436 pa->nr = 0; 4437 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) 4438 pa->nr++; 4439 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) 4440 pa->nr++; 4441 pa->ticket = V_ticket_altqs_active; 4442 PF_RULES_RUNLOCK(); 4443 break; 4444 } 4445 4446 case DIOCGETALTQV0: 4447 case DIOCGETALTQV1: { 4448 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4449 struct pf_altq *altq; 4450 4451 PF_RULES_RLOCK(); 4452 if (pa->ticket != V_ticket_altqs_active) { 4453 PF_RULES_RUNLOCK(); 4454 error = EBUSY; 4455 goto fail; 4456 } 4457 altq = pf_altq_get_nth_active(pa->nr); 4458 if (altq == NULL) { 4459 PF_RULES_RUNLOCK(); 4460 error = EBUSY; 4461 goto fail; 4462 } 4463 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4464 PF_RULES_RUNLOCK(); 4465 break; 4466 } 4467 4468 case DIOCCHANGEALTQV0: 4469 case DIOCCHANGEALTQV1: 4470 /* CHANGEALTQ not supported yet! */ 4471 error = ENODEV; 4472 break; 4473 4474 case DIOCGETQSTATSV0: 4475 case DIOCGETQSTATSV1: { 4476 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr; 4477 struct pf_altq *altq; 4478 int nbytes; 4479 u_int32_t version; 4480 4481 PF_RULES_RLOCK(); 4482 if (pq->ticket != V_ticket_altqs_active) { 4483 PF_RULES_RUNLOCK(); 4484 error = EBUSY; 4485 goto fail; 4486 } 4487 nbytes = pq->nbytes; 4488 altq = pf_altq_get_nth_active(pq->nr); 4489 if (altq == NULL) { 4490 PF_RULES_RUNLOCK(); 4491 error = EBUSY; 4492 goto fail; 4493 } 4494 4495 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { 4496 PF_RULES_RUNLOCK(); 4497 error = ENXIO; 4498 goto fail; 4499 } 4500 PF_RULES_RUNLOCK(); 4501 if (cmd == DIOCGETQSTATSV0) 4502 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */ 4503 else 4504 version = pq->version; 4505 error = altq_getqstats(altq, pq->buf, &nbytes, version); 4506 if (error == 0) { 4507 pq->scheduler = altq->scheduler; 4508 pq->nbytes = nbytes; 4509 } 4510 break; 4511 } 4512 #endif /* ALTQ */ 4513 4514 case DIOCBEGINADDRS: { 4515 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4516 4517 error = pf_ioctl_begin_addrs(&pp->ticket); 4518 break; 4519 } 4520 4521 case DIOCADDADDR: { 4522 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4523 struct pf_nl_pooladdr npp = {}; 4524 4525 npp.which = PF_RDR; 4526 memcpy(&npp, pp, sizeof(*pp)); 4527 error = pf_ioctl_add_addr(&npp); 4528 break; 4529 } 4530 4531 case DIOCGETADDRS: { 4532 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4533 struct pf_nl_pooladdr npp = {}; 4534 4535 npp.which = PF_RDR; 4536 memcpy(&npp, pp, sizeof(*pp)); 4537 error = pf_ioctl_get_addrs(&npp); 4538 memcpy(pp, &npp, sizeof(*pp)); 4539 4540 break; 4541 } 4542 4543 case DIOCGETADDR: { 4544 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4545 struct pf_nl_pooladdr npp = {}; 4546 4547 npp.which = PF_RDR; 4548 memcpy(&npp, pp, sizeof(*pp)); 4549 error = pf_ioctl_get_addr(&npp); 4550 memcpy(pp, &npp, sizeof(*pp)); 4551 4552 break; 4553 } 4554 4555 case DIOCCHANGEADDR: { 4556 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; 4557 struct pf_kpool *pool; 4558 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL; 4559 struct pf_kruleset *ruleset; 4560 struct pfi_kkif *kif = NULL; 4561 4562 pca->anchor[sizeof(pca->anchor) - 1] = '\0'; 4563 4564 if (pca->action < PF_CHANGE_ADD_HEAD || 4565 pca->action > PF_CHANGE_REMOVE) { 4566 error = EINVAL; 4567 goto fail; 4568 } 4569 if (pca->addr.addr.type != PF_ADDR_ADDRMASK && 4570 pca->addr.addr.type != PF_ADDR_DYNIFTL && 4571 pca->addr.addr.type != PF_ADDR_TABLE) { 4572 error = EINVAL; 4573 goto fail; 4574 } 4575 if (pca->addr.addr.p.dyn != NULL) { 4576 error = EINVAL; 4577 goto fail; 4578 } 4579 4580 if (pca->action != PF_CHANGE_REMOVE) { 4581 #ifndef INET 4582 if (pca->af == AF_INET) { 4583 error = EAFNOSUPPORT; 4584 goto fail; 4585 } 4586 #endif /* INET */ 4587 #ifndef INET6 4588 if (pca->af == AF_INET6) { 4589 error = EAFNOSUPPORT; 4590 goto fail; 4591 } 4592 #endif /* INET6 */ 4593 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); 4594 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); 4595 if (newpa->ifname[0]) 4596 kif = pf_kkif_create(M_WAITOK); 4597 newpa->kif = NULL; 4598 } 4599 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x) 4600 PF_RULES_WLOCK(); 4601 ruleset = pf_find_kruleset(pca->anchor); 4602 if (ruleset == NULL) 4603 ERROUT(EBUSY); 4604 4605 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action, 4606 pca->r_num, pca->r_last, 1, 1, PF_RDR); 4607 if (pool == NULL) 4608 ERROUT(EBUSY); 4609 4610 if (pca->action != PF_CHANGE_REMOVE) { 4611 if (newpa->ifname[0]) { 4612 newpa->kif = pfi_kkif_attach(kif, newpa->ifname); 4613 pfi_kkif_ref(newpa->kif); 4614 kif = NULL; 4615 } 4616 4617 switch (newpa->addr.type) { 4618 case PF_ADDR_DYNIFTL: 4619 error = pfi_dynaddr_setup(&newpa->addr, 4620 pca->af); 4621 break; 4622 case PF_ADDR_TABLE: 4623 newpa->addr.p.tbl = pfr_attach_table(ruleset, 4624 newpa->addr.v.tblname); 4625 if (newpa->addr.p.tbl == NULL) 4626 error = ENOMEM; 4627 break; 4628 } 4629 if (error) 4630 goto DIOCCHANGEADDR_error; 4631 } 4632 4633 switch (pca->action) { 4634 case PF_CHANGE_ADD_HEAD: 4635 oldpa = TAILQ_FIRST(&pool->list); 4636 break; 4637 case PF_CHANGE_ADD_TAIL: 4638 oldpa = TAILQ_LAST(&pool->list, pf_kpalist); 4639 break; 4640 default: 4641 oldpa = TAILQ_FIRST(&pool->list); 4642 for (int i = 0; oldpa && i < pca->nr; i++) 4643 oldpa = TAILQ_NEXT(oldpa, entries); 4644 4645 if (oldpa == NULL) 4646 ERROUT(EINVAL); 4647 } 4648 4649 if (pca->action == PF_CHANGE_REMOVE) { 4650 TAILQ_REMOVE(&pool->list, oldpa, entries); 4651 switch (oldpa->addr.type) { 4652 case PF_ADDR_DYNIFTL: 4653 pfi_dynaddr_remove(oldpa->addr.p.dyn); 4654 break; 4655 case PF_ADDR_TABLE: 4656 pfr_detach_table(oldpa->addr.p.tbl); 4657 break; 4658 } 4659 if (oldpa->kif) 4660 pfi_kkif_unref(oldpa->kif); 4661 free(oldpa, M_PFRULE); 4662 } else { 4663 if (oldpa == NULL) 4664 TAILQ_INSERT_TAIL(&pool->list, newpa, entries); 4665 else if (pca->action == PF_CHANGE_ADD_HEAD || 4666 pca->action == PF_CHANGE_ADD_BEFORE) 4667 TAILQ_INSERT_BEFORE(oldpa, newpa, entries); 4668 else 4669 TAILQ_INSERT_AFTER(&pool->list, oldpa, 4670 newpa, entries); 4671 } 4672 4673 pool->cur = TAILQ_FIRST(&pool->list); 4674 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); 4675 PF_RULES_WUNLOCK(); 4676 break; 4677 4678 #undef ERROUT 4679 DIOCCHANGEADDR_error: 4680 if (newpa != NULL) { 4681 if (newpa->kif) 4682 pfi_kkif_unref(newpa->kif); 4683 free(newpa, M_PFRULE); 4684 } 4685 PF_RULES_WUNLOCK(); 4686 pf_kkif_free(kif); 4687 break; 4688 } 4689 4690 case DIOCGETRULESETS: { 4691 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4692 4693 pr->path[sizeof(pr->path) - 1] = '\0'; 4694 4695 error = pf_ioctl_get_rulesets(pr); 4696 break; 4697 } 4698 4699 case DIOCGETRULESET: { 4700 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4701 4702 pr->path[sizeof(pr->path) - 1] = '\0'; 4703 4704 error = pf_ioctl_get_ruleset(pr); 4705 break; 4706 } 4707 4708 case DIOCRCLRTABLES: { 4709 struct pfioc_table *io = (struct pfioc_table *)addr; 4710 4711 if (io->pfrio_esize != 0) { 4712 error = ENODEV; 4713 goto fail; 4714 } 4715 PF_RULES_WLOCK(); 4716 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, 4717 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4718 PF_RULES_WUNLOCK(); 4719 break; 4720 } 4721 4722 case DIOCRADDTABLES: { 4723 struct pfioc_table *io = (struct pfioc_table *)addr; 4724 struct pfr_table *pfrts; 4725 size_t totlen; 4726 4727 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4728 error = ENODEV; 4729 goto fail; 4730 } 4731 4732 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4733 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4734 error = ENOMEM; 4735 goto fail; 4736 } 4737 4738 totlen = io->pfrio_size * sizeof(struct pfr_table); 4739 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4740 M_TEMP, M_WAITOK); 4741 error = copyin(io->pfrio_buffer, pfrts, totlen); 4742 if (error) { 4743 free(pfrts, M_TEMP); 4744 goto fail; 4745 } 4746 PF_RULES_WLOCK(); 4747 error = pfr_add_tables(pfrts, io->pfrio_size, 4748 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4749 PF_RULES_WUNLOCK(); 4750 free(pfrts, M_TEMP); 4751 break; 4752 } 4753 4754 case DIOCRDELTABLES: { 4755 struct pfioc_table *io = (struct pfioc_table *)addr; 4756 struct pfr_table *pfrts; 4757 size_t totlen; 4758 4759 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4760 error = ENODEV; 4761 goto fail; 4762 } 4763 4764 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4765 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4766 error = ENOMEM; 4767 goto fail; 4768 } 4769 4770 totlen = io->pfrio_size * sizeof(struct pfr_table); 4771 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4772 M_TEMP, M_WAITOK); 4773 error = copyin(io->pfrio_buffer, pfrts, totlen); 4774 if (error) { 4775 free(pfrts, M_TEMP); 4776 goto fail; 4777 } 4778 PF_RULES_WLOCK(); 4779 error = pfr_del_tables(pfrts, io->pfrio_size, 4780 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4781 PF_RULES_WUNLOCK(); 4782 free(pfrts, M_TEMP); 4783 break; 4784 } 4785 4786 case DIOCRGETTABLES: { 4787 struct pfioc_table *io = (struct pfioc_table *)addr; 4788 struct pfr_table *pfrts; 4789 size_t totlen; 4790 int n; 4791 4792 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4793 error = ENODEV; 4794 goto fail; 4795 } 4796 PF_RULES_RLOCK(); 4797 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4798 if (n < 0) { 4799 PF_RULES_RUNLOCK(); 4800 error = EINVAL; 4801 goto fail; 4802 } 4803 io->pfrio_size = min(io->pfrio_size, n); 4804 4805 totlen = io->pfrio_size * sizeof(struct pfr_table); 4806 4807 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4808 M_TEMP, M_NOWAIT | M_ZERO); 4809 if (pfrts == NULL) { 4810 error = ENOMEM; 4811 PF_RULES_RUNLOCK(); 4812 goto fail; 4813 } 4814 error = pfr_get_tables(&io->pfrio_table, pfrts, 4815 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4816 PF_RULES_RUNLOCK(); 4817 if (error == 0) 4818 error = copyout(pfrts, io->pfrio_buffer, totlen); 4819 free(pfrts, M_TEMP); 4820 break; 4821 } 4822 4823 case DIOCRGETTSTATS: { 4824 struct pfioc_table *io = (struct pfioc_table *)addr; 4825 struct pfr_tstats *pfrtstats; 4826 size_t totlen; 4827 int n; 4828 4829 if (io->pfrio_esize != sizeof(struct pfr_tstats)) { 4830 error = ENODEV; 4831 goto fail; 4832 } 4833 PF_TABLE_STATS_LOCK(); 4834 PF_RULES_RLOCK(); 4835 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4836 if (n < 0) { 4837 PF_RULES_RUNLOCK(); 4838 PF_TABLE_STATS_UNLOCK(); 4839 error = EINVAL; 4840 goto fail; 4841 } 4842 io->pfrio_size = min(io->pfrio_size, n); 4843 4844 totlen = io->pfrio_size * sizeof(struct pfr_tstats); 4845 pfrtstats = mallocarray(io->pfrio_size, 4846 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO); 4847 if (pfrtstats == NULL) { 4848 error = ENOMEM; 4849 PF_RULES_RUNLOCK(); 4850 PF_TABLE_STATS_UNLOCK(); 4851 goto fail; 4852 } 4853 error = pfr_get_tstats(&io->pfrio_table, pfrtstats, 4854 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4855 PF_RULES_RUNLOCK(); 4856 PF_TABLE_STATS_UNLOCK(); 4857 if (error == 0) 4858 error = copyout(pfrtstats, io->pfrio_buffer, totlen); 4859 free(pfrtstats, M_TEMP); 4860 break; 4861 } 4862 4863 case DIOCRCLRTSTATS: { 4864 struct pfioc_table *io = (struct pfioc_table *)addr; 4865 struct pfr_table *pfrts; 4866 size_t totlen; 4867 4868 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4869 error = ENODEV; 4870 goto fail; 4871 } 4872 4873 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4874 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4875 /* We used to count tables and use the minimum required 4876 * size, so we didn't fail on overly large requests. 4877 * Keep doing so. */ 4878 io->pfrio_size = pf_ioctl_maxcount; 4879 goto fail; 4880 } 4881 4882 totlen = io->pfrio_size * sizeof(struct pfr_table); 4883 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4884 M_TEMP, M_WAITOK); 4885 error = copyin(io->pfrio_buffer, pfrts, totlen); 4886 if (error) { 4887 free(pfrts, M_TEMP); 4888 goto fail; 4889 } 4890 4891 PF_TABLE_STATS_LOCK(); 4892 PF_RULES_RLOCK(); 4893 error = pfr_clr_tstats(pfrts, io->pfrio_size, 4894 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4895 PF_RULES_RUNLOCK(); 4896 PF_TABLE_STATS_UNLOCK(); 4897 free(pfrts, M_TEMP); 4898 break; 4899 } 4900 4901 case DIOCRSETTFLAGS: { 4902 struct pfioc_table *io = (struct pfioc_table *)addr; 4903 struct pfr_table *pfrts; 4904 size_t totlen; 4905 int n; 4906 4907 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4908 error = ENODEV; 4909 goto fail; 4910 } 4911 4912 PF_RULES_RLOCK(); 4913 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4914 if (n < 0) { 4915 PF_RULES_RUNLOCK(); 4916 error = EINVAL; 4917 goto fail; 4918 } 4919 4920 io->pfrio_size = min(io->pfrio_size, n); 4921 PF_RULES_RUNLOCK(); 4922 4923 totlen = io->pfrio_size * sizeof(struct pfr_table); 4924 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4925 M_TEMP, M_WAITOK); 4926 error = copyin(io->pfrio_buffer, pfrts, totlen); 4927 if (error) { 4928 free(pfrts, M_TEMP); 4929 goto fail; 4930 } 4931 PF_RULES_WLOCK(); 4932 error = pfr_set_tflags(pfrts, io->pfrio_size, 4933 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, 4934 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4935 PF_RULES_WUNLOCK(); 4936 free(pfrts, M_TEMP); 4937 break; 4938 } 4939 4940 case DIOCRCLRADDRS: { 4941 struct pfioc_table *io = (struct pfioc_table *)addr; 4942 4943 if (io->pfrio_esize != 0) { 4944 error = ENODEV; 4945 goto fail; 4946 } 4947 PF_RULES_WLOCK(); 4948 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, 4949 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4950 PF_RULES_WUNLOCK(); 4951 break; 4952 } 4953 4954 case DIOCRADDADDRS: { 4955 struct pfioc_table *io = (struct pfioc_table *)addr; 4956 struct pfr_addr *pfras; 4957 size_t totlen; 4958 4959 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4960 error = ENODEV; 4961 goto fail; 4962 } 4963 if (io->pfrio_size < 0 || 4964 io->pfrio_size > pf_ioctl_maxcount || 4965 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4966 error = EINVAL; 4967 goto fail; 4968 } 4969 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4970 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4971 M_TEMP, M_WAITOK); 4972 error = copyin(io->pfrio_buffer, pfras, totlen); 4973 if (error) { 4974 free(pfras, M_TEMP); 4975 goto fail; 4976 } 4977 PF_RULES_WLOCK(); 4978 io->pfrio_nadd = 0; 4979 error = pfr_add_addrs(&io->pfrio_table, pfras, 4980 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | 4981 PFR_FLAG_USERIOCTL); 4982 PF_RULES_WUNLOCK(); 4983 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4984 error = copyout(pfras, io->pfrio_buffer, totlen); 4985 free(pfras, M_TEMP); 4986 break; 4987 } 4988 4989 case DIOCRDELADDRS: { 4990 struct pfioc_table *io = (struct pfioc_table *)addr; 4991 struct pfr_addr *pfras; 4992 size_t totlen; 4993 4994 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4995 error = ENODEV; 4996 goto fail; 4997 } 4998 if (io->pfrio_size < 0 || 4999 io->pfrio_size > pf_ioctl_maxcount || 5000 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5001 error = EINVAL; 5002 goto fail; 5003 } 5004 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5005 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5006 M_TEMP, M_WAITOK); 5007 error = copyin(io->pfrio_buffer, pfras, totlen); 5008 if (error) { 5009 free(pfras, M_TEMP); 5010 goto fail; 5011 } 5012 PF_RULES_WLOCK(); 5013 error = pfr_del_addrs(&io->pfrio_table, pfras, 5014 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | 5015 PFR_FLAG_USERIOCTL); 5016 PF_RULES_WUNLOCK(); 5017 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5018 error = copyout(pfras, io->pfrio_buffer, totlen); 5019 free(pfras, M_TEMP); 5020 break; 5021 } 5022 5023 case DIOCRSETADDRS: { 5024 struct pfioc_table *io = (struct pfioc_table *)addr; 5025 struct pfr_addr *pfras; 5026 size_t totlen, count; 5027 5028 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5029 error = ENODEV; 5030 goto fail; 5031 } 5032 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { 5033 error = EINVAL; 5034 goto fail; 5035 } 5036 count = max(io->pfrio_size, io->pfrio_size2); 5037 if (count > pf_ioctl_maxcount || 5038 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { 5039 error = EINVAL; 5040 goto fail; 5041 } 5042 totlen = count * sizeof(struct pfr_addr); 5043 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP, 5044 M_WAITOK); 5045 error = copyin(io->pfrio_buffer, pfras, totlen); 5046 if (error) { 5047 free(pfras, M_TEMP); 5048 goto fail; 5049 } 5050 PF_RULES_WLOCK(); 5051 error = pfr_set_addrs(&io->pfrio_table, pfras, 5052 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, 5053 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | 5054 PFR_FLAG_USERIOCTL, 0); 5055 PF_RULES_WUNLOCK(); 5056 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5057 error = copyout(pfras, io->pfrio_buffer, totlen); 5058 free(pfras, M_TEMP); 5059 break; 5060 } 5061 5062 case DIOCRGETADDRS: { 5063 struct pfioc_table *io = (struct pfioc_table *)addr; 5064 struct pfr_addr *pfras; 5065 size_t totlen; 5066 5067 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5068 error = ENODEV; 5069 goto fail; 5070 } 5071 if (io->pfrio_size < 0 || 5072 io->pfrio_size > pf_ioctl_maxcount || 5073 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5074 error = EINVAL; 5075 goto fail; 5076 } 5077 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5078 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5079 M_TEMP, M_WAITOK | M_ZERO); 5080 PF_RULES_RLOCK(); 5081 error = pfr_get_addrs(&io->pfrio_table, pfras, 5082 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5083 PF_RULES_RUNLOCK(); 5084 if (error == 0) 5085 error = copyout(pfras, io->pfrio_buffer, totlen); 5086 free(pfras, M_TEMP); 5087 break; 5088 } 5089 5090 case DIOCRGETASTATS: { 5091 struct pfioc_table *io = (struct pfioc_table *)addr; 5092 struct pfr_astats *pfrastats; 5093 size_t totlen; 5094 5095 if (io->pfrio_esize != sizeof(struct pfr_astats)) { 5096 error = ENODEV; 5097 goto fail; 5098 } 5099 if (io->pfrio_size < 0 || 5100 io->pfrio_size > pf_ioctl_maxcount || 5101 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { 5102 error = EINVAL; 5103 goto fail; 5104 } 5105 totlen = io->pfrio_size * sizeof(struct pfr_astats); 5106 pfrastats = mallocarray(io->pfrio_size, 5107 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO); 5108 PF_RULES_RLOCK(); 5109 error = pfr_get_astats(&io->pfrio_table, pfrastats, 5110 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5111 PF_RULES_RUNLOCK(); 5112 if (error == 0) 5113 error = copyout(pfrastats, io->pfrio_buffer, totlen); 5114 free(pfrastats, M_TEMP); 5115 break; 5116 } 5117 5118 case DIOCRCLRASTATS: { 5119 struct pfioc_table *io = (struct pfioc_table *)addr; 5120 struct pfr_addr *pfras; 5121 size_t totlen; 5122 5123 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5124 error = ENODEV; 5125 goto fail; 5126 } 5127 if (io->pfrio_size < 0 || 5128 io->pfrio_size > pf_ioctl_maxcount || 5129 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5130 error = EINVAL; 5131 goto fail; 5132 } 5133 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5134 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5135 M_TEMP, M_WAITOK); 5136 error = copyin(io->pfrio_buffer, pfras, totlen); 5137 if (error) { 5138 free(pfras, M_TEMP); 5139 goto fail; 5140 } 5141 PF_RULES_WLOCK(); 5142 error = pfr_clr_astats(&io->pfrio_table, pfras, 5143 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | 5144 PFR_FLAG_USERIOCTL); 5145 PF_RULES_WUNLOCK(); 5146 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5147 error = copyout(pfras, io->pfrio_buffer, totlen); 5148 free(pfras, M_TEMP); 5149 break; 5150 } 5151 5152 case DIOCRTSTADDRS: { 5153 struct pfioc_table *io = (struct pfioc_table *)addr; 5154 struct pfr_addr *pfras; 5155 size_t totlen; 5156 5157 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5158 error = ENODEV; 5159 goto fail; 5160 } 5161 if (io->pfrio_size < 0 || 5162 io->pfrio_size > pf_ioctl_maxcount || 5163 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5164 error = EINVAL; 5165 goto fail; 5166 } 5167 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5168 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5169 M_TEMP, M_WAITOK); 5170 error = copyin(io->pfrio_buffer, pfras, totlen); 5171 if (error) { 5172 free(pfras, M_TEMP); 5173 goto fail; 5174 } 5175 PF_RULES_RLOCK(); 5176 error = pfr_tst_addrs(&io->pfrio_table, pfras, 5177 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | 5178 PFR_FLAG_USERIOCTL); 5179 PF_RULES_RUNLOCK(); 5180 if (error == 0) 5181 error = copyout(pfras, io->pfrio_buffer, totlen); 5182 free(pfras, M_TEMP); 5183 break; 5184 } 5185 5186 case DIOCRINADEFINE: { 5187 struct pfioc_table *io = (struct pfioc_table *)addr; 5188 struct pfr_addr *pfras; 5189 size_t totlen; 5190 5191 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5192 error = ENODEV; 5193 goto fail; 5194 } 5195 if (io->pfrio_size < 0 || 5196 io->pfrio_size > pf_ioctl_maxcount || 5197 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5198 error = EINVAL; 5199 goto fail; 5200 } 5201 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5202 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5203 M_TEMP, M_WAITOK); 5204 error = copyin(io->pfrio_buffer, pfras, totlen); 5205 if (error) { 5206 free(pfras, M_TEMP); 5207 goto fail; 5208 } 5209 PF_RULES_WLOCK(); 5210 error = pfr_ina_define(&io->pfrio_table, pfras, 5211 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, 5212 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5213 PF_RULES_WUNLOCK(); 5214 free(pfras, M_TEMP); 5215 break; 5216 } 5217 5218 case DIOCOSFPADD: { 5219 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5220 PF_RULES_WLOCK(); 5221 error = pf_osfp_add(io); 5222 PF_RULES_WUNLOCK(); 5223 break; 5224 } 5225 5226 case DIOCOSFPGET: { 5227 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5228 PF_RULES_RLOCK(); 5229 error = pf_osfp_get(io); 5230 PF_RULES_RUNLOCK(); 5231 break; 5232 } 5233 5234 case DIOCXBEGIN: { 5235 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5236 struct pfioc_trans_e *ioes, *ioe; 5237 size_t totlen; 5238 int i; 5239 5240 if (io->esize != sizeof(*ioe)) { 5241 error = ENODEV; 5242 goto fail; 5243 } 5244 if (io->size < 0 || 5245 io->size > pf_ioctl_maxcount || 5246 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5247 error = EINVAL; 5248 goto fail; 5249 } 5250 totlen = sizeof(struct pfioc_trans_e) * io->size; 5251 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5252 M_TEMP, M_WAITOK); 5253 error = copyin(io->array, ioes, totlen); 5254 if (error) { 5255 free(ioes, M_TEMP); 5256 goto fail; 5257 } 5258 PF_RULES_WLOCK(); 5259 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5260 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5261 switch (ioe->rs_num) { 5262 case PF_RULESET_ETH: 5263 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) { 5264 PF_RULES_WUNLOCK(); 5265 free(ioes, M_TEMP); 5266 goto fail; 5267 } 5268 break; 5269 #ifdef ALTQ 5270 case PF_RULESET_ALTQ: 5271 if (ioe->anchor[0]) { 5272 PF_RULES_WUNLOCK(); 5273 free(ioes, M_TEMP); 5274 error = EINVAL; 5275 goto fail; 5276 } 5277 if ((error = pf_begin_altq(&ioe->ticket))) { 5278 PF_RULES_WUNLOCK(); 5279 free(ioes, M_TEMP); 5280 goto fail; 5281 } 5282 break; 5283 #endif /* ALTQ */ 5284 case PF_RULESET_TABLE: 5285 { 5286 struct pfr_table table; 5287 5288 bzero(&table, sizeof(table)); 5289 strlcpy(table.pfrt_anchor, ioe->anchor, 5290 sizeof(table.pfrt_anchor)); 5291 if ((error = pfr_ina_begin(&table, 5292 &ioe->ticket, NULL, 0))) { 5293 PF_RULES_WUNLOCK(); 5294 free(ioes, M_TEMP); 5295 goto fail; 5296 } 5297 break; 5298 } 5299 default: 5300 if ((error = pf_begin_rules(&ioe->ticket, 5301 ioe->rs_num, ioe->anchor))) { 5302 PF_RULES_WUNLOCK(); 5303 free(ioes, M_TEMP); 5304 goto fail; 5305 } 5306 break; 5307 } 5308 } 5309 PF_RULES_WUNLOCK(); 5310 error = copyout(ioes, io->array, totlen); 5311 free(ioes, M_TEMP); 5312 break; 5313 } 5314 5315 case DIOCXROLLBACK: { 5316 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5317 struct pfioc_trans_e *ioe, *ioes; 5318 size_t totlen; 5319 int i; 5320 5321 if (io->esize != sizeof(*ioe)) { 5322 error = ENODEV; 5323 goto fail; 5324 } 5325 if (io->size < 0 || 5326 io->size > pf_ioctl_maxcount || 5327 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5328 error = EINVAL; 5329 goto fail; 5330 } 5331 totlen = sizeof(struct pfioc_trans_e) * io->size; 5332 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5333 M_TEMP, M_WAITOK); 5334 error = copyin(io->array, ioes, totlen); 5335 if (error) { 5336 free(ioes, M_TEMP); 5337 goto fail; 5338 } 5339 PF_RULES_WLOCK(); 5340 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5341 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5342 switch (ioe->rs_num) { 5343 case PF_RULESET_ETH: 5344 if ((error = pf_rollback_eth(ioe->ticket, 5345 ioe->anchor))) { 5346 PF_RULES_WUNLOCK(); 5347 free(ioes, M_TEMP); 5348 goto fail; /* really bad */ 5349 } 5350 break; 5351 #ifdef ALTQ 5352 case PF_RULESET_ALTQ: 5353 if (ioe->anchor[0]) { 5354 PF_RULES_WUNLOCK(); 5355 free(ioes, M_TEMP); 5356 error = EINVAL; 5357 goto fail; 5358 } 5359 if ((error = pf_rollback_altq(ioe->ticket))) { 5360 PF_RULES_WUNLOCK(); 5361 free(ioes, M_TEMP); 5362 goto fail; /* really bad */ 5363 } 5364 break; 5365 #endif /* ALTQ */ 5366 case PF_RULESET_TABLE: 5367 { 5368 struct pfr_table table; 5369 5370 bzero(&table, sizeof(table)); 5371 strlcpy(table.pfrt_anchor, ioe->anchor, 5372 sizeof(table.pfrt_anchor)); 5373 if ((error = pfr_ina_rollback(&table, 5374 ioe->ticket, NULL, 0))) { 5375 PF_RULES_WUNLOCK(); 5376 free(ioes, M_TEMP); 5377 goto fail; /* really bad */ 5378 } 5379 break; 5380 } 5381 default: 5382 if ((error = pf_rollback_rules(ioe->ticket, 5383 ioe->rs_num, ioe->anchor))) { 5384 PF_RULES_WUNLOCK(); 5385 free(ioes, M_TEMP); 5386 goto fail; /* really bad */ 5387 } 5388 break; 5389 } 5390 } 5391 PF_RULES_WUNLOCK(); 5392 free(ioes, M_TEMP); 5393 break; 5394 } 5395 5396 case DIOCXCOMMIT: { 5397 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5398 struct pfioc_trans_e *ioe, *ioes; 5399 struct pf_kruleset *rs; 5400 struct pf_keth_ruleset *ers; 5401 size_t totlen; 5402 int i; 5403 5404 if (io->esize != sizeof(*ioe)) { 5405 error = ENODEV; 5406 goto fail; 5407 } 5408 5409 if (io->size < 0 || 5410 io->size > pf_ioctl_maxcount || 5411 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5412 error = EINVAL; 5413 goto fail; 5414 } 5415 5416 totlen = sizeof(struct pfioc_trans_e) * io->size; 5417 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5418 M_TEMP, M_WAITOK); 5419 error = copyin(io->array, ioes, totlen); 5420 if (error) { 5421 free(ioes, M_TEMP); 5422 goto fail; 5423 } 5424 PF_RULES_WLOCK(); 5425 /* First makes sure everything will succeed. */ 5426 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5427 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5428 switch (ioe->rs_num) { 5429 case PF_RULESET_ETH: 5430 ers = pf_find_keth_ruleset(ioe->anchor); 5431 if (ers == NULL || ioe->ticket == 0 || 5432 ioe->ticket != ers->inactive.ticket) { 5433 PF_RULES_WUNLOCK(); 5434 free(ioes, M_TEMP); 5435 error = EINVAL; 5436 goto fail; 5437 } 5438 break; 5439 #ifdef ALTQ 5440 case PF_RULESET_ALTQ: 5441 if (ioe->anchor[0]) { 5442 PF_RULES_WUNLOCK(); 5443 free(ioes, M_TEMP); 5444 error = EINVAL; 5445 goto fail; 5446 } 5447 if (!V_altqs_inactive_open || ioe->ticket != 5448 V_ticket_altqs_inactive) { 5449 PF_RULES_WUNLOCK(); 5450 free(ioes, M_TEMP); 5451 error = EBUSY; 5452 goto fail; 5453 } 5454 break; 5455 #endif /* ALTQ */ 5456 case PF_RULESET_TABLE: 5457 rs = pf_find_kruleset(ioe->anchor); 5458 if (rs == NULL || !rs->topen || ioe->ticket != 5459 rs->tticket) { 5460 PF_RULES_WUNLOCK(); 5461 free(ioes, M_TEMP); 5462 error = EBUSY; 5463 goto fail; 5464 } 5465 break; 5466 default: 5467 if (ioe->rs_num < 0 || ioe->rs_num >= 5468 PF_RULESET_MAX) { 5469 PF_RULES_WUNLOCK(); 5470 free(ioes, M_TEMP); 5471 error = EINVAL; 5472 goto fail; 5473 } 5474 rs = pf_find_kruleset(ioe->anchor); 5475 if (rs == NULL || 5476 !rs->rules[ioe->rs_num].inactive.open || 5477 rs->rules[ioe->rs_num].inactive.ticket != 5478 ioe->ticket) { 5479 PF_RULES_WUNLOCK(); 5480 free(ioes, M_TEMP); 5481 error = EBUSY; 5482 goto fail; 5483 } 5484 break; 5485 } 5486 } 5487 /* Now do the commit - no errors should happen here. */ 5488 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5489 switch (ioe->rs_num) { 5490 case PF_RULESET_ETH: 5491 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) { 5492 PF_RULES_WUNLOCK(); 5493 free(ioes, M_TEMP); 5494 goto fail; /* really bad */ 5495 } 5496 break; 5497 #ifdef ALTQ 5498 case PF_RULESET_ALTQ: 5499 if ((error = pf_commit_altq(ioe->ticket))) { 5500 PF_RULES_WUNLOCK(); 5501 free(ioes, M_TEMP); 5502 goto fail; /* really bad */ 5503 } 5504 break; 5505 #endif /* ALTQ */ 5506 case PF_RULESET_TABLE: 5507 { 5508 struct pfr_table table; 5509 5510 bzero(&table, sizeof(table)); 5511 (void)strlcpy(table.pfrt_anchor, ioe->anchor, 5512 sizeof(table.pfrt_anchor)); 5513 if ((error = pfr_ina_commit(&table, 5514 ioe->ticket, NULL, NULL, 0))) { 5515 PF_RULES_WUNLOCK(); 5516 free(ioes, M_TEMP); 5517 goto fail; /* really bad */ 5518 } 5519 break; 5520 } 5521 default: 5522 if ((error = pf_commit_rules(ioe->ticket, 5523 ioe->rs_num, ioe->anchor))) { 5524 PF_RULES_WUNLOCK(); 5525 free(ioes, M_TEMP); 5526 goto fail; /* really bad */ 5527 } 5528 break; 5529 } 5530 } 5531 PF_RULES_WUNLOCK(); 5532 5533 /* Only hook into EtherNet taffic if we've got rules for it. */ 5534 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 5535 hook_pf_eth(); 5536 else 5537 dehook_pf_eth(); 5538 5539 free(ioes, M_TEMP); 5540 break; 5541 } 5542 5543 case DIOCGETSRCNODES: { 5544 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; 5545 struct pf_srchash *sh; 5546 struct pf_ksrc_node *n; 5547 struct pf_src_node *p, *pstore; 5548 uint32_t i, nr = 0; 5549 5550 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5551 i++, sh++) { 5552 PF_HASHROW_LOCK(sh); 5553 LIST_FOREACH(n, &sh->nodes, entry) 5554 nr++; 5555 PF_HASHROW_UNLOCK(sh); 5556 } 5557 5558 psn->psn_len = min(psn->psn_len, 5559 sizeof(struct pf_src_node) * nr); 5560 5561 if (psn->psn_len == 0) { 5562 psn->psn_len = sizeof(struct pf_src_node) * nr; 5563 goto fail; 5564 } 5565 5566 nr = 0; 5567 5568 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO); 5569 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5570 i++, sh++) { 5571 PF_HASHROW_LOCK(sh); 5572 LIST_FOREACH(n, &sh->nodes, entry) { 5573 5574 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) 5575 break; 5576 5577 pf_src_node_copy(n, p); 5578 5579 p++; 5580 nr++; 5581 } 5582 PF_HASHROW_UNLOCK(sh); 5583 } 5584 error = copyout(pstore, psn->psn_src_nodes, 5585 sizeof(struct pf_src_node) * nr); 5586 if (error) { 5587 free(pstore, M_TEMP); 5588 goto fail; 5589 } 5590 psn->psn_len = sizeof(struct pf_src_node) * nr; 5591 free(pstore, M_TEMP); 5592 break; 5593 } 5594 5595 case DIOCCLRSRCNODES: { 5596 pf_kill_srcnodes(NULL); 5597 break; 5598 } 5599 5600 case DIOCKILLSRCNODES: 5601 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); 5602 break; 5603 5604 #ifdef COMPAT_FREEBSD13 5605 case DIOCKEEPCOUNTERS_FREEBSD13: 5606 #endif 5607 case DIOCKEEPCOUNTERS: 5608 error = pf_keepcounters((struct pfioc_nv *)addr); 5609 break; 5610 5611 case DIOCGETSYNCOOKIES: 5612 error = pf_get_syncookies((struct pfioc_nv *)addr); 5613 break; 5614 5615 case DIOCSETSYNCOOKIES: 5616 error = pf_set_syncookies((struct pfioc_nv *)addr); 5617 break; 5618 5619 case DIOCSETHOSTID: { 5620 u_int32_t *hostid = (u_int32_t *)addr; 5621 5622 PF_RULES_WLOCK(); 5623 if (*hostid == 0) 5624 V_pf_status.hostid = arc4random(); 5625 else 5626 V_pf_status.hostid = *hostid; 5627 PF_RULES_WUNLOCK(); 5628 break; 5629 } 5630 5631 case DIOCOSFPFLUSH: 5632 PF_RULES_WLOCK(); 5633 pf_osfp_flush(); 5634 PF_RULES_WUNLOCK(); 5635 break; 5636 5637 case DIOCIGETIFACES: { 5638 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5639 struct pfi_kif *ifstore; 5640 size_t bufsiz; 5641 5642 if (io->pfiio_esize != sizeof(struct pfi_kif)) { 5643 error = ENODEV; 5644 goto fail; 5645 } 5646 5647 if (io->pfiio_size < 0 || 5648 io->pfiio_size > pf_ioctl_maxcount || 5649 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { 5650 error = EINVAL; 5651 goto fail; 5652 } 5653 5654 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5655 5656 bufsiz = io->pfiio_size * sizeof(struct pfi_kif); 5657 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif), 5658 M_TEMP, M_WAITOK | M_ZERO); 5659 5660 PF_RULES_RLOCK(); 5661 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); 5662 PF_RULES_RUNLOCK(); 5663 error = copyout(ifstore, io->pfiio_buffer, bufsiz); 5664 free(ifstore, M_TEMP); 5665 break; 5666 } 5667 5668 case DIOCSETIFFLAG: { 5669 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5670 5671 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5672 5673 PF_RULES_WLOCK(); 5674 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags); 5675 PF_RULES_WUNLOCK(); 5676 break; 5677 } 5678 5679 case DIOCCLRIFFLAG: { 5680 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5681 5682 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5683 5684 PF_RULES_WLOCK(); 5685 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); 5686 PF_RULES_WUNLOCK(); 5687 break; 5688 } 5689 5690 case DIOCSETREASS: { 5691 u_int32_t *reass = (u_int32_t *)addr; 5692 5693 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF); 5694 /* Removal of DF flag without reassembly enabled is not a 5695 * valid combination. Disable reassembly in such case. */ 5696 if (!(V_pf_status.reass & PF_REASS_ENABLED)) 5697 V_pf_status.reass = 0; 5698 break; 5699 } 5700 5701 default: 5702 error = ENODEV; 5703 break; 5704 } 5705 fail: 5706 CURVNET_RESTORE(); 5707 5708 #undef ERROUT_IOCTL 5709 5710 return (error); 5711 } 5712 5713 void 5714 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version) 5715 { 5716 bzero(sp, sizeof(union pfsync_state_union)); 5717 5718 /* copy from state key */ 5719 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5720 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5721 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5722 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5723 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5724 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5725 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5726 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5727 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto; 5728 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af; 5729 5730 /* copy from state */ 5731 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname)); 5732 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr)); 5733 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000)); 5734 sp->pfs_1301.expire = pf_state_expires(st); 5735 if (sp->pfs_1301.expire <= time_uptime) 5736 sp->pfs_1301.expire = htonl(0); 5737 else 5738 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime); 5739 5740 sp->pfs_1301.direction = st->direction; 5741 sp->pfs_1301.log = st->act.log; 5742 sp->pfs_1301.timeout = st->timeout; 5743 5744 switch (msg_version) { 5745 case PFSYNC_MSG_VERSION_1301: 5746 sp->pfs_1301.state_flags = st->state_flags; 5747 break; 5748 case PFSYNC_MSG_VERSION_1400: 5749 sp->pfs_1400.state_flags = htons(st->state_flags); 5750 sp->pfs_1400.qid = htons(st->act.qid); 5751 sp->pfs_1400.pqid = htons(st->act.pqid); 5752 sp->pfs_1400.dnpipe = htons(st->act.dnpipe); 5753 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe); 5754 sp->pfs_1400.rtableid = htonl(st->act.rtableid); 5755 sp->pfs_1400.min_ttl = st->act.min_ttl; 5756 sp->pfs_1400.set_tos = st->act.set_tos; 5757 sp->pfs_1400.max_mss = htons(st->act.max_mss); 5758 sp->pfs_1400.set_prio[0] = st->act.set_prio[0]; 5759 sp->pfs_1400.set_prio[1] = st->act.set_prio[1]; 5760 sp->pfs_1400.rt = st->act.rt; 5761 if (st->act.rt_kif) 5762 strlcpy(sp->pfs_1400.rt_ifname, 5763 st->act.rt_kif->pfik_name, 5764 sizeof(sp->pfs_1400.rt_ifname)); 5765 break; 5766 default: 5767 panic("%s: Unsupported pfsync_msg_version %d", 5768 __func__, msg_version); 5769 } 5770 5771 /* 5772 * XXX Why do we bother pfsyncing source node information if source 5773 * nodes are not synced? Showing users that there is source tracking 5774 * when there is none seems useless. 5775 */ 5776 if (st->sns[PF_SN_LIMIT] != NULL) 5777 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE; 5778 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE]) 5779 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5780 5781 sp->pfs_1301.id = st->id; 5782 sp->pfs_1301.creatorid = st->creatorid; 5783 pf_state_peer_hton(&st->src, &sp->pfs_1301.src); 5784 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst); 5785 5786 if (st->rule == NULL) 5787 sp->pfs_1301.rule = htonl(-1); 5788 else 5789 sp->pfs_1301.rule = htonl(st->rule->nr); 5790 if (st->anchor == NULL) 5791 sp->pfs_1301.anchor = htonl(-1); 5792 else 5793 sp->pfs_1301.anchor = htonl(st->anchor->nr); 5794 if (st->nat_rule == NULL) 5795 sp->pfs_1301.nat_rule = htonl(-1); 5796 else 5797 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr); 5798 5799 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]); 5800 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]); 5801 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]); 5802 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]); 5803 } 5804 5805 void 5806 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st) 5807 { 5808 bzero(sp, sizeof(*sp)); 5809 5810 sp->version = PF_STATE_VERSION; 5811 5812 /* copy from state key */ 5813 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5814 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5815 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5816 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5817 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5818 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5819 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5820 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5821 sp->proto = st->key[PF_SK_WIRE]->proto; 5822 sp->af = st->key[PF_SK_WIRE]->af; 5823 5824 /* copy from state */ 5825 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5826 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name, 5827 sizeof(sp->orig_ifname)); 5828 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr)); 5829 sp->creation = htonl(time_uptime - (st->creation / 1000)); 5830 sp->expire = pf_state_expires(st); 5831 if (sp->expire <= time_uptime) 5832 sp->expire = htonl(0); 5833 else 5834 sp->expire = htonl(sp->expire - time_uptime); 5835 5836 sp->direction = st->direction; 5837 sp->log = st->act.log; 5838 sp->timeout = st->timeout; 5839 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */ 5840 sp->state_flags_compat = st->state_flags; 5841 sp->state_flags = htons(st->state_flags); 5842 if (st->sns[PF_SN_LIMIT] != NULL) 5843 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5844 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL) 5845 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5846 sp->id = st->id; 5847 sp->creatorid = st->creatorid; 5848 pf_state_peer_hton(&st->src, &sp->src); 5849 pf_state_peer_hton(&st->dst, &sp->dst); 5850 5851 if (st->rule == NULL) 5852 sp->rule = htonl(-1); 5853 else 5854 sp->rule = htonl(st->rule->nr); 5855 if (st->anchor == NULL) 5856 sp->anchor = htonl(-1); 5857 else 5858 sp->anchor = htonl(st->anchor->nr); 5859 if (st->nat_rule == NULL) 5860 sp->nat_rule = htonl(-1); 5861 else 5862 sp->nat_rule = htonl(st->nat_rule->nr); 5863 5864 sp->packets[0] = st->packets[0]; 5865 sp->packets[1] = st->packets[1]; 5866 sp->bytes[0] = st->bytes[0]; 5867 sp->bytes[1] = st->bytes[1]; 5868 5869 sp->qid = htons(st->act.qid); 5870 sp->pqid = htons(st->act.pqid); 5871 sp->dnpipe = htons(st->act.dnpipe); 5872 sp->dnrpipe = htons(st->act.dnrpipe); 5873 sp->rtableid = htonl(st->act.rtableid); 5874 sp->min_ttl = st->act.min_ttl; 5875 sp->set_tos = st->act.set_tos; 5876 sp->max_mss = htons(st->act.max_mss); 5877 sp->rt = st->act.rt; 5878 if (st->act.rt_kif) 5879 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name, 5880 sizeof(sp->rt_ifname)); 5881 sp->set_prio[0] = st->act.set_prio[0]; 5882 sp->set_prio[1] = st->act.set_prio[1]; 5883 5884 } 5885 5886 static void 5887 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 5888 { 5889 struct pfr_ktable *kt; 5890 5891 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); 5892 5893 kt = aw->p.tbl; 5894 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 5895 kt = kt->pfrkt_root; 5896 aw->p.tbl = NULL; 5897 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 5898 kt->pfrkt_cnt : -1; 5899 } 5900 5901 static int 5902 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters, 5903 size_t number, char **names) 5904 { 5905 nvlist_t *nvc; 5906 5907 nvc = nvlist_create(0); 5908 if (nvc == NULL) 5909 return (ENOMEM); 5910 5911 for (int i = 0; i < number; i++) { 5912 nvlist_append_number_array(nvc, "counters", 5913 counter_u64_fetch(counters[i])); 5914 nvlist_append_string_array(nvc, "names", 5915 names[i]); 5916 nvlist_append_number_array(nvc, "ids", 5917 i); 5918 } 5919 nvlist_add_nvlist(nvl, name, nvc); 5920 nvlist_destroy(nvc); 5921 5922 return (0); 5923 } 5924 5925 static int 5926 pf_getstatus(struct pfioc_nv *nv) 5927 { 5928 nvlist_t *nvl = NULL, *nvc = NULL; 5929 void *nvlpacked = NULL; 5930 int error; 5931 struct pf_status s; 5932 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES; 5933 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES; 5934 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES; 5935 time_t since; 5936 5937 PF_RULES_RLOCK_TRACKER; 5938 5939 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 5940 5941 PF_RULES_RLOCK(); 5942 5943 nvl = nvlist_create(0); 5944 if (nvl == NULL) 5945 ERROUT(ENOMEM); 5946 5947 since = time_second - (time_uptime - V_pf_status.since); 5948 5949 nvlist_add_bool(nvl, "running", V_pf_status.running); 5950 nvlist_add_number(nvl, "since", since); 5951 nvlist_add_number(nvl, "debug", V_pf_status.debug); 5952 nvlist_add_number(nvl, "hostid", V_pf_status.hostid); 5953 nvlist_add_number(nvl, "states", V_pf_status.states); 5954 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes); 5955 nvlist_add_number(nvl, "reass", V_pf_status.reass); 5956 nvlist_add_bool(nvl, "syncookies_active", 5957 V_pf_status.syncookies_active); 5958 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen); 5959 5960 /* counters */ 5961 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters, 5962 PFRES_MAX, pf_reasons); 5963 if (error != 0) 5964 ERROUT(error); 5965 5966 /* lcounters */ 5967 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters, 5968 KLCNT_MAX, pf_lcounter); 5969 if (error != 0) 5970 ERROUT(error); 5971 5972 /* fcounters */ 5973 nvc = nvlist_create(0); 5974 if (nvc == NULL) 5975 ERROUT(ENOMEM); 5976 5977 for (int i = 0; i < FCNT_MAX; i++) { 5978 nvlist_append_number_array(nvc, "counters", 5979 pf_counter_u64_fetch(&V_pf_status.fcounters[i])); 5980 nvlist_append_string_array(nvc, "names", 5981 pf_fcounter[i]); 5982 nvlist_append_number_array(nvc, "ids", 5983 i); 5984 } 5985 nvlist_add_nvlist(nvl, "fcounters", nvc); 5986 nvlist_destroy(nvc); 5987 nvc = NULL; 5988 5989 /* scounters */ 5990 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters, 5991 SCNT_MAX, pf_fcounter); 5992 if (error != 0) 5993 ERROUT(error); 5994 5995 nvlist_add_string(nvl, "ifname", V_pf_status.ifname); 5996 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum, 5997 PF_MD5_DIGEST_LENGTH); 5998 5999 pfi_update_status(V_pf_status.ifname, &s); 6000 6001 /* pcounters / bcounters */ 6002 for (int i = 0; i < 2; i++) { 6003 for (int j = 0; j < 2; j++) { 6004 for (int k = 0; k < 2; k++) { 6005 nvlist_append_number_array(nvl, "pcounters", 6006 s.pcounters[i][j][k]); 6007 } 6008 nvlist_append_number_array(nvl, "bcounters", 6009 s.bcounters[i][j]); 6010 } 6011 } 6012 6013 nvlpacked = nvlist_pack(nvl, &nv->len); 6014 if (nvlpacked == NULL) 6015 ERROUT(ENOMEM); 6016 6017 if (nv->size == 0) 6018 ERROUT(0); 6019 else if (nv->size < nv->len) 6020 ERROUT(ENOSPC); 6021 6022 PF_RULES_RUNLOCK(); 6023 error = copyout(nvlpacked, nv->data, nv->len); 6024 goto done; 6025 6026 #undef ERROUT 6027 errout: 6028 PF_RULES_RUNLOCK(); 6029 done: 6030 free(nvlpacked, M_NVLIST); 6031 nvlist_destroy(nvc); 6032 nvlist_destroy(nvl); 6033 6034 return (error); 6035 } 6036 6037 /* 6038 * XXX - Check for version mismatch!!! 6039 */ 6040 static void 6041 pf_clear_all_states(void) 6042 { 6043 struct epoch_tracker et; 6044 struct pf_kstate *s; 6045 u_int i; 6046 6047 NET_EPOCH_ENTER(et); 6048 for (i = 0; i <= V_pf_hashmask; i++) { 6049 struct pf_idhash *ih = &V_pf_idhash[i]; 6050 relock: 6051 PF_HASHROW_LOCK(ih); 6052 LIST_FOREACH(s, &ih->states, entry) { 6053 s->timeout = PFTM_PURGE; 6054 /* Don't send out individual delete messages. */ 6055 s->state_flags |= PFSTATE_NOSYNC; 6056 pf_remove_state(s); 6057 goto relock; 6058 } 6059 PF_HASHROW_UNLOCK(ih); 6060 } 6061 NET_EPOCH_EXIT(et); 6062 } 6063 6064 static int 6065 pf_clear_tables(void) 6066 { 6067 struct pfioc_table io; 6068 int error; 6069 6070 bzero(&io, sizeof(io)); 6071 io.pfrio_flags |= PFR_FLAG_ALLRSETS; 6072 6073 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, 6074 io.pfrio_flags); 6075 6076 return (error); 6077 } 6078 6079 static void 6080 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) 6081 { 6082 struct pf_ksrc_node_list kill; 6083 u_int killed; 6084 6085 LIST_INIT(&kill); 6086 for (int i = 0; i <= V_pf_srchashmask; i++) { 6087 struct pf_srchash *sh = &V_pf_srchash[i]; 6088 struct pf_ksrc_node *sn, *tmp; 6089 6090 PF_HASHROW_LOCK(sh); 6091 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) 6092 if (psnk == NULL || 6093 (pf_match_addr(psnk->psnk_src.neg, 6094 &psnk->psnk_src.addr.v.a.addr, 6095 &psnk->psnk_src.addr.v.a.mask, 6096 &sn->addr, sn->af) && 6097 pf_match_addr(psnk->psnk_dst.neg, 6098 &psnk->psnk_dst.addr.v.a.addr, 6099 &psnk->psnk_dst.addr.v.a.mask, 6100 &sn->raddr, sn->af))) { 6101 pf_unlink_src_node(sn); 6102 LIST_INSERT_HEAD(&kill, sn, entry); 6103 sn->expire = 1; 6104 } 6105 PF_HASHROW_UNLOCK(sh); 6106 } 6107 6108 for (int i = 0; i <= V_pf_hashmask; i++) { 6109 struct pf_idhash *ih = &V_pf_idhash[i]; 6110 struct pf_kstate *s; 6111 6112 PF_HASHROW_LOCK(ih); 6113 LIST_FOREACH(s, &ih->states, entry) { 6114 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; 6115 sn_type++) { 6116 if (s->sns[sn_type] && 6117 s->sns[sn_type]->expire == 1) { 6118 s->sns[sn_type] = NULL; 6119 } 6120 } 6121 } 6122 PF_HASHROW_UNLOCK(ih); 6123 } 6124 6125 killed = pf_free_src_nodes(&kill); 6126 6127 if (psnk != NULL) 6128 psnk->psnk_killed = killed; 6129 } 6130 6131 static int 6132 pf_keepcounters(struct pfioc_nv *nv) 6133 { 6134 nvlist_t *nvl = NULL; 6135 void *nvlpacked = NULL; 6136 int error = 0; 6137 6138 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6139 6140 if (nv->len > pf_ioctl_maxcount) 6141 ERROUT(ENOMEM); 6142 6143 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6144 error = copyin(nv->data, nvlpacked, nv->len); 6145 if (error) 6146 ERROUT(error); 6147 6148 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6149 if (nvl == NULL) 6150 ERROUT(EBADMSG); 6151 6152 if (! nvlist_exists_bool(nvl, "keep_counters")) 6153 ERROUT(EBADMSG); 6154 6155 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters"); 6156 6157 on_error: 6158 nvlist_destroy(nvl); 6159 free(nvlpacked, M_NVLIST); 6160 return (error); 6161 } 6162 6163 unsigned int 6164 pf_clear_states(const struct pf_kstate_kill *kill) 6165 { 6166 struct pf_state_key_cmp match_key; 6167 struct pf_kstate *s; 6168 struct pfi_kkif *kif; 6169 int idx; 6170 unsigned int killed = 0, dir; 6171 6172 NET_EPOCH_ASSERT(); 6173 6174 for (unsigned int i = 0; i <= V_pf_hashmask; i++) { 6175 struct pf_idhash *ih = &V_pf_idhash[i]; 6176 6177 relock_DIOCCLRSTATES: 6178 PF_HASHROW_LOCK(ih); 6179 LIST_FOREACH(s, &ih->states, entry) { 6180 /* For floating states look at the original kif. */ 6181 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 6182 6183 if (kill->psk_ifname[0] && 6184 strcmp(kill->psk_ifname, 6185 kif->pfik_name)) 6186 continue; 6187 6188 if (kill->psk_kill_match) { 6189 bzero(&match_key, sizeof(match_key)); 6190 6191 if (s->direction == PF_OUT) { 6192 dir = PF_IN; 6193 idx = PF_SK_STACK; 6194 } else { 6195 dir = PF_OUT; 6196 idx = PF_SK_WIRE; 6197 } 6198 6199 match_key.af = s->key[idx]->af; 6200 match_key.proto = s->key[idx]->proto; 6201 pf_addrcpy(&match_key.addr[0], 6202 &s->key[idx]->addr[1], match_key.af); 6203 match_key.port[0] = s->key[idx]->port[1]; 6204 pf_addrcpy(&match_key.addr[1], 6205 &s->key[idx]->addr[0], match_key.af); 6206 match_key.port[1] = s->key[idx]->port[0]; 6207 } 6208 6209 /* 6210 * Don't send out individual 6211 * delete messages. 6212 */ 6213 s->state_flags |= PFSTATE_NOSYNC; 6214 pf_remove_state(s); 6215 killed++; 6216 6217 if (kill->psk_kill_match) 6218 killed += pf_kill_matching_state(&match_key, 6219 dir); 6220 6221 goto relock_DIOCCLRSTATES; 6222 } 6223 PF_HASHROW_UNLOCK(ih); 6224 } 6225 6226 if (V_pfsync_clear_states_ptr != NULL) 6227 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname); 6228 6229 return (killed); 6230 } 6231 6232 void 6233 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed) 6234 { 6235 struct pf_kstate *s; 6236 6237 NET_EPOCH_ASSERT(); 6238 if (kill->psk_pfcmp.id) { 6239 if (kill->psk_pfcmp.creatorid == 0) 6240 kill->psk_pfcmp.creatorid = V_pf_status.hostid; 6241 if ((s = pf_find_state_byid(kill->psk_pfcmp.id, 6242 kill->psk_pfcmp.creatorid))) { 6243 pf_remove_state(s); 6244 *killed = 1; 6245 } 6246 return; 6247 } 6248 6249 for (unsigned int i = 0; i <= V_pf_hashmask; i++) 6250 *killed += pf_killstates_row(kill, &V_pf_idhash[i]); 6251 } 6252 6253 static int 6254 pf_killstates_nv(struct pfioc_nv *nv) 6255 { 6256 struct pf_kstate_kill kill; 6257 struct epoch_tracker et; 6258 nvlist_t *nvl = NULL; 6259 void *nvlpacked = NULL; 6260 int error = 0; 6261 unsigned int killed = 0; 6262 6263 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6264 6265 if (nv->len > pf_ioctl_maxcount) 6266 ERROUT(ENOMEM); 6267 6268 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6269 error = copyin(nv->data, nvlpacked, nv->len); 6270 if (error) 6271 ERROUT(error); 6272 6273 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6274 if (nvl == NULL) 6275 ERROUT(EBADMSG); 6276 6277 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6278 if (error) 6279 ERROUT(error); 6280 6281 NET_EPOCH_ENTER(et); 6282 pf_killstates(&kill, &killed); 6283 NET_EPOCH_EXIT(et); 6284 6285 free(nvlpacked, M_NVLIST); 6286 nvlpacked = NULL; 6287 nvlist_destroy(nvl); 6288 nvl = nvlist_create(0); 6289 if (nvl == NULL) 6290 ERROUT(ENOMEM); 6291 6292 nvlist_add_number(nvl, "killed", killed); 6293 6294 nvlpacked = nvlist_pack(nvl, &nv->len); 6295 if (nvlpacked == NULL) 6296 ERROUT(ENOMEM); 6297 6298 if (nv->size == 0) 6299 ERROUT(0); 6300 else if (nv->size < nv->len) 6301 ERROUT(ENOSPC); 6302 6303 error = copyout(nvlpacked, nv->data, nv->len); 6304 6305 on_error: 6306 nvlist_destroy(nvl); 6307 free(nvlpacked, M_NVLIST); 6308 return (error); 6309 } 6310 6311 static int 6312 pf_clearstates_nv(struct pfioc_nv *nv) 6313 { 6314 struct pf_kstate_kill kill; 6315 struct epoch_tracker et; 6316 nvlist_t *nvl = NULL; 6317 void *nvlpacked = NULL; 6318 int error = 0; 6319 unsigned int killed; 6320 6321 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6322 6323 if (nv->len > pf_ioctl_maxcount) 6324 ERROUT(ENOMEM); 6325 6326 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6327 error = copyin(nv->data, nvlpacked, nv->len); 6328 if (error) 6329 ERROUT(error); 6330 6331 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6332 if (nvl == NULL) 6333 ERROUT(EBADMSG); 6334 6335 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6336 if (error) 6337 ERROUT(error); 6338 6339 NET_EPOCH_ENTER(et); 6340 killed = pf_clear_states(&kill); 6341 NET_EPOCH_EXIT(et); 6342 6343 free(nvlpacked, M_NVLIST); 6344 nvlpacked = NULL; 6345 nvlist_destroy(nvl); 6346 nvl = nvlist_create(0); 6347 if (nvl == NULL) 6348 ERROUT(ENOMEM); 6349 6350 nvlist_add_number(nvl, "killed", killed); 6351 6352 nvlpacked = nvlist_pack(nvl, &nv->len); 6353 if (nvlpacked == NULL) 6354 ERROUT(ENOMEM); 6355 6356 if (nv->size == 0) 6357 ERROUT(0); 6358 else if (nv->size < nv->len) 6359 ERROUT(ENOSPC); 6360 6361 error = copyout(nvlpacked, nv->data, nv->len); 6362 6363 #undef ERROUT 6364 on_error: 6365 nvlist_destroy(nvl); 6366 free(nvlpacked, M_NVLIST); 6367 return (error); 6368 } 6369 6370 static int 6371 pf_getstate(struct pfioc_nv *nv) 6372 { 6373 nvlist_t *nvl = NULL, *nvls; 6374 void *nvlpacked = NULL; 6375 struct pf_kstate *s = NULL; 6376 int error = 0; 6377 uint64_t id, creatorid; 6378 6379 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6380 6381 if (nv->len > pf_ioctl_maxcount) 6382 ERROUT(ENOMEM); 6383 6384 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6385 error = copyin(nv->data, nvlpacked, nv->len); 6386 if (error) 6387 ERROUT(error); 6388 6389 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6390 if (nvl == NULL) 6391 ERROUT(EBADMSG); 6392 6393 PFNV_CHK(pf_nvuint64(nvl, "id", &id)); 6394 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid)); 6395 6396 s = pf_find_state_byid(id, creatorid); 6397 if (s == NULL) 6398 ERROUT(ENOENT); 6399 6400 free(nvlpacked, M_NVLIST); 6401 nvlpacked = NULL; 6402 nvlist_destroy(nvl); 6403 nvl = nvlist_create(0); 6404 if (nvl == NULL) 6405 ERROUT(ENOMEM); 6406 6407 nvls = pf_state_to_nvstate(s); 6408 if (nvls == NULL) 6409 ERROUT(ENOMEM); 6410 6411 nvlist_add_nvlist(nvl, "state", nvls); 6412 nvlist_destroy(nvls); 6413 6414 nvlpacked = nvlist_pack(nvl, &nv->len); 6415 if (nvlpacked == NULL) 6416 ERROUT(ENOMEM); 6417 6418 if (nv->size == 0) 6419 ERROUT(0); 6420 else if (nv->size < nv->len) 6421 ERROUT(ENOSPC); 6422 6423 error = copyout(nvlpacked, nv->data, nv->len); 6424 6425 #undef ERROUT 6426 errout: 6427 if (s != NULL) 6428 PF_STATE_UNLOCK(s); 6429 free(nvlpacked, M_NVLIST); 6430 nvlist_destroy(nvl); 6431 return (error); 6432 } 6433 6434 /* 6435 * XXX - Check for version mismatch!!! 6436 */ 6437 6438 /* 6439 * Duplicate pfctl -Fa operation to get rid of as much as we can. 6440 */ 6441 static int 6442 shutdown_pf(void) 6443 { 6444 int error = 0; 6445 u_int32_t t[5]; 6446 char nn = '\0'; 6447 struct pf_kanchor *anchor; 6448 struct pf_keth_anchor *eth_anchor; 6449 int rs_num; 6450 6451 do { 6452 /* Unlink rules of all user defined anchors */ 6453 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) { 6454 /* Wildcard based anchors may not have a respective 6455 * explicit anchor rule or they may be left empty 6456 * without rules. It leads to anchor.refcnt=0, and the 6457 * rest of the logic does not expect it. */ 6458 if (anchor->refcnt == 0) 6459 anchor->refcnt = 1; 6460 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6461 if ((error = pf_begin_rules(&t[rs_num], rs_num, 6462 anchor->path)) != 0) { 6463 DPFPRINTF(PF_DEBUG_MISC, "%s: " 6464 "anchor.path=%s rs_num=%d", 6465 __func__, anchor->path, rs_num); 6466 goto error; /* XXX: rollback? */ 6467 } 6468 } 6469 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6470 error = pf_commit_rules(t[rs_num], rs_num, 6471 anchor->path); 6472 MPASS(error == 0); 6473 } 6474 } 6475 6476 /* Unlink rules of all user defined ether anchors */ 6477 RB_FOREACH(eth_anchor, pf_keth_anchor_global, 6478 &V_pf_keth_anchors) { 6479 /* Wildcard based anchors may not have a respective 6480 * explicit anchor rule or they may be left empty 6481 * without rules. It leads to anchor.refcnt=0, and the 6482 * rest of the logic does not expect it. */ 6483 if (eth_anchor->refcnt == 0) 6484 eth_anchor->refcnt = 1; 6485 if ((error = pf_begin_eth(&t[0], eth_anchor->path)) 6486 != 0) { 6487 DPFPRINTF(PF_DEBUG_MISC, "%s: eth " 6488 "anchor.path=%s", __func__, 6489 eth_anchor->path); 6490 goto error; 6491 } 6492 error = pf_commit_eth(t[0], eth_anchor->path); 6493 MPASS(error == 0); 6494 } 6495 6496 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) 6497 != 0) { 6498 DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__); 6499 break; 6500 } 6501 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) 6502 != 0) { 6503 DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__); 6504 break; /* XXX: rollback? */ 6505 } 6506 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) 6507 != 0) { 6508 DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__); 6509 break; /* XXX: rollback? */ 6510 } 6511 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) 6512 != 0) { 6513 DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__); 6514 break; /* XXX: rollback? */ 6515 } 6516 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) 6517 != 0) { 6518 DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__); 6519 break; /* XXX: rollback? */ 6520 } 6521 6522 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn); 6523 MPASS(error == 0); 6524 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn); 6525 MPASS(error == 0); 6526 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn); 6527 MPASS(error == 0); 6528 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn); 6529 MPASS(error == 0); 6530 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn); 6531 MPASS(error == 0); 6532 6533 if ((error = pf_clear_tables()) != 0) 6534 break; 6535 6536 if ((error = pf_begin_eth(&t[0], &nn)) != 0) { 6537 DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__); 6538 break; 6539 } 6540 error = pf_commit_eth(t[0], &nn); 6541 MPASS(error == 0); 6542 6543 #ifdef ALTQ 6544 if ((error = pf_begin_altq(&t[0])) != 0) { 6545 DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__); 6546 break; 6547 } 6548 pf_commit_altq(t[0]); 6549 #endif 6550 6551 pf_clear_all_states(); 6552 6553 pf_kill_srcnodes(NULL); 6554 6555 /* status does not use malloced mem so no need to cleanup */ 6556 /* fingerprints and interfaces have their own cleanup code */ 6557 } while(0); 6558 6559 error: 6560 return (error); 6561 } 6562 6563 static pfil_return_t 6564 pf_check_return(int chk, struct mbuf **m) 6565 { 6566 6567 switch (chk) { 6568 case PF_PASS: 6569 if (*m == NULL) 6570 return (PFIL_CONSUMED); 6571 else 6572 return (PFIL_PASS); 6573 break; 6574 default: 6575 if (*m != NULL) { 6576 m_freem(*m); 6577 *m = NULL; 6578 } 6579 return (PFIL_DROPPED); 6580 } 6581 } 6582 6583 static pfil_return_t 6584 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6585 void *ruleset __unused, struct inpcb *inp) 6586 { 6587 int chk; 6588 6589 CURVNET_ASSERT_SET(); 6590 6591 chk = pf_test_eth(PF_IN, flags, ifp, m, inp); 6592 6593 return (pf_check_return(chk, m)); 6594 } 6595 6596 static pfil_return_t 6597 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6598 void *ruleset __unused, struct inpcb *inp) 6599 { 6600 int chk; 6601 6602 CURVNET_ASSERT_SET(); 6603 6604 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp); 6605 6606 return (pf_check_return(chk, m)); 6607 } 6608 6609 #ifdef INET 6610 static pfil_return_t 6611 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6612 void *ruleset __unused, struct inpcb *inp) 6613 { 6614 int chk; 6615 6616 CURVNET_ASSERT_SET(); 6617 6618 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL); 6619 6620 return (pf_check_return(chk, m)); 6621 } 6622 6623 static pfil_return_t 6624 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6625 void *ruleset __unused, struct inpcb *inp) 6626 { 6627 int chk; 6628 6629 CURVNET_ASSERT_SET(); 6630 6631 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL); 6632 6633 return (pf_check_return(chk, m)); 6634 } 6635 #endif 6636 6637 #ifdef INET6 6638 static pfil_return_t 6639 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags, 6640 void *ruleset __unused, struct inpcb *inp) 6641 { 6642 int chk; 6643 6644 CURVNET_ASSERT_SET(); 6645 6646 /* 6647 * In case of loopback traffic IPv6 uses the real interface in 6648 * order to support scoped addresses. In order to support stateful 6649 * filtering we have change this to lo0 as it is the case in IPv4. 6650 */ 6651 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, 6652 m, inp, NULL); 6653 6654 return (pf_check_return(chk, m)); 6655 } 6656 6657 static pfil_return_t 6658 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags, 6659 void *ruleset __unused, struct inpcb *inp) 6660 { 6661 int chk; 6662 6663 CURVNET_ASSERT_SET(); 6664 6665 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL); 6666 6667 return (pf_check_return(chk, m)); 6668 } 6669 #endif /* INET6 */ 6670 6671 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook); 6672 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook); 6673 #define V_pf_eth_in_hook VNET(pf_eth_in_hook) 6674 #define V_pf_eth_out_hook VNET(pf_eth_out_hook) 6675 6676 #ifdef INET 6677 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook); 6678 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook); 6679 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook) 6680 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook) 6681 #endif 6682 #ifdef INET6 6683 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook); 6684 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook); 6685 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook) 6686 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook) 6687 #endif 6688 6689 static void 6690 hook_pf_eth(void) 6691 { 6692 struct pfil_hook_args pha = { 6693 .pa_version = PFIL_VERSION, 6694 .pa_modname = "pf", 6695 .pa_type = PFIL_TYPE_ETHERNET, 6696 }; 6697 struct pfil_link_args pla = { 6698 .pa_version = PFIL_VERSION, 6699 }; 6700 int ret __diagused; 6701 6702 if (atomic_load_bool(&V_pf_pfil_eth_hooked)) 6703 return; 6704 6705 pha.pa_mbuf_chk = pf_eth_check_in; 6706 pha.pa_flags = PFIL_IN; 6707 pha.pa_rulname = "eth-in"; 6708 V_pf_eth_in_hook = pfil_add_hook(&pha); 6709 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6710 pla.pa_head = V_link_pfil_head; 6711 pla.pa_hook = V_pf_eth_in_hook; 6712 ret = pfil_link(&pla); 6713 MPASS(ret == 0); 6714 pha.pa_mbuf_chk = pf_eth_check_out; 6715 pha.pa_flags = PFIL_OUT; 6716 pha.pa_rulname = "eth-out"; 6717 V_pf_eth_out_hook = pfil_add_hook(&pha); 6718 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6719 pla.pa_head = V_link_pfil_head; 6720 pla.pa_hook = V_pf_eth_out_hook; 6721 ret = pfil_link(&pla); 6722 MPASS(ret == 0); 6723 6724 atomic_store_bool(&V_pf_pfil_eth_hooked, true); 6725 } 6726 6727 static void 6728 hook_pf(void) 6729 { 6730 struct pfil_hook_args pha = { 6731 .pa_version = PFIL_VERSION, 6732 .pa_modname = "pf", 6733 }; 6734 struct pfil_link_args pla = { 6735 .pa_version = PFIL_VERSION, 6736 }; 6737 int ret __diagused; 6738 6739 if (atomic_load_bool(&V_pf_pfil_hooked)) 6740 return; 6741 6742 #ifdef INET 6743 pha.pa_type = PFIL_TYPE_IP4; 6744 pha.pa_mbuf_chk = pf_check_in; 6745 pha.pa_flags = PFIL_IN; 6746 pha.pa_rulname = "default-in"; 6747 V_pf_ip4_in_hook = pfil_add_hook(&pha); 6748 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6749 pla.pa_head = V_inet_pfil_head; 6750 pla.pa_hook = V_pf_ip4_in_hook; 6751 ret = pfil_link(&pla); 6752 MPASS(ret == 0); 6753 pha.pa_mbuf_chk = pf_check_out; 6754 pha.pa_flags = PFIL_OUT; 6755 pha.pa_rulname = "default-out"; 6756 V_pf_ip4_out_hook = pfil_add_hook(&pha); 6757 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6758 pla.pa_head = V_inet_pfil_head; 6759 pla.pa_hook = V_pf_ip4_out_hook; 6760 ret = pfil_link(&pla); 6761 MPASS(ret == 0); 6762 if (V_pf_filter_local) { 6763 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6764 pla.pa_head = V_inet_local_pfil_head; 6765 pla.pa_hook = V_pf_ip4_out_hook; 6766 ret = pfil_link(&pla); 6767 MPASS(ret == 0); 6768 } 6769 #endif 6770 #ifdef INET6 6771 pha.pa_type = PFIL_TYPE_IP6; 6772 pha.pa_mbuf_chk = pf_check6_in; 6773 pha.pa_flags = PFIL_IN; 6774 pha.pa_rulname = "default-in6"; 6775 V_pf_ip6_in_hook = pfil_add_hook(&pha); 6776 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6777 pla.pa_head = V_inet6_pfil_head; 6778 pla.pa_hook = V_pf_ip6_in_hook; 6779 ret = pfil_link(&pla); 6780 MPASS(ret == 0); 6781 pha.pa_mbuf_chk = pf_check6_out; 6782 pha.pa_rulname = "default-out6"; 6783 pha.pa_flags = PFIL_OUT; 6784 V_pf_ip6_out_hook = pfil_add_hook(&pha); 6785 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6786 pla.pa_head = V_inet6_pfil_head; 6787 pla.pa_hook = V_pf_ip6_out_hook; 6788 ret = pfil_link(&pla); 6789 MPASS(ret == 0); 6790 if (V_pf_filter_local) { 6791 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6792 pla.pa_head = V_inet6_local_pfil_head; 6793 pla.pa_hook = V_pf_ip6_out_hook; 6794 ret = pfil_link(&pla); 6795 MPASS(ret == 0); 6796 } 6797 #endif 6798 6799 atomic_store_bool(&V_pf_pfil_hooked, true); 6800 } 6801 6802 static void 6803 dehook_pf_eth(void) 6804 { 6805 6806 if (!atomic_load_bool(&V_pf_pfil_eth_hooked)) 6807 return; 6808 6809 pfil_remove_hook(V_pf_eth_in_hook); 6810 pfil_remove_hook(V_pf_eth_out_hook); 6811 6812 atomic_store_bool(&V_pf_pfil_eth_hooked, false); 6813 } 6814 6815 static void 6816 dehook_pf(void) 6817 { 6818 6819 if (!atomic_load_bool(&V_pf_pfil_hooked)) 6820 return; 6821 6822 #ifdef INET 6823 pfil_remove_hook(V_pf_ip4_in_hook); 6824 pfil_remove_hook(V_pf_ip4_out_hook); 6825 #endif 6826 #ifdef INET6 6827 pfil_remove_hook(V_pf_ip6_in_hook); 6828 pfil_remove_hook(V_pf_ip6_out_hook); 6829 #endif 6830 6831 atomic_store_bool(&V_pf_pfil_hooked, false); 6832 } 6833 6834 static void 6835 pf_load_vnet(void) 6836 { 6837 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname), 6838 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 6839 6840 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE); 6841 sx_init(&V_pf_ioctl_lock, "pf ioctl"); 6842 6843 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize, 6844 PF_RULE_TAG_HASH_SIZE_DEFAULT); 6845 #ifdef ALTQ 6846 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize, 6847 PF_QUEUE_TAG_HASH_SIZE_DEFAULT); 6848 #endif 6849 6850 V_pf_keth = &V_pf_main_keth_anchor.ruleset; 6851 6852 pfattach_vnet(); 6853 V_pf_vnet_active = 1; 6854 } 6855 6856 static int 6857 pf_load(void) 6858 { 6859 int error; 6860 6861 sx_init(&pf_end_lock, "pf end thread"); 6862 6863 pf_mtag_initialize(); 6864 6865 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME); 6866 if (pf_dev == NULL) 6867 return (ENOMEM); 6868 6869 pf_end_threads = 0; 6870 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge"); 6871 if (error != 0) 6872 return (error); 6873 6874 pfi_initialize(); 6875 6876 return (0); 6877 } 6878 6879 static void 6880 pf_unload_vnet(void) 6881 { 6882 int ret __diagused; 6883 6884 V_pf_vnet_active = 0; 6885 V_pf_status.running = 0; 6886 dehook_pf(); 6887 dehook_pf_eth(); 6888 6889 PF_RULES_WLOCK(); 6890 pf_syncookies_cleanup(); 6891 shutdown_pf(); 6892 PF_RULES_WUNLOCK(); 6893 6894 ret = swi_remove(V_pf_swi_cookie); 6895 MPASS(ret == 0); 6896 ret = intr_event_destroy(V_pf_swi_ie); 6897 MPASS(ret == 0); 6898 6899 pf_unload_vnet_purge(); 6900 6901 pf_normalize_cleanup(); 6902 PF_RULES_WLOCK(); 6903 pfi_cleanup_vnet(); 6904 PF_RULES_WUNLOCK(); 6905 pfr_cleanup(); 6906 pf_osfp_flush(); 6907 pf_cleanup(); 6908 if (IS_DEFAULT_VNET(curvnet)) 6909 pf_mtag_cleanup(); 6910 6911 pf_cleanup_tagset(&V_pf_tags); 6912 #ifdef ALTQ 6913 pf_cleanup_tagset(&V_pf_qids); 6914 #endif 6915 uma_zdestroy(V_pf_tag_z); 6916 6917 #ifdef PF_WANT_32_TO_64_COUNTER 6918 PF_RULES_WLOCK(); 6919 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); 6920 6921 MPASS(LIST_EMPTY(&V_pf_allkiflist)); 6922 MPASS(V_pf_allkifcount == 0); 6923 6924 LIST_REMOVE(&V_pf_default_rule, allrulelist); 6925 V_pf_allrulecount--; 6926 LIST_REMOVE(V_pf_rulemarker, allrulelist); 6927 6928 MPASS(LIST_EMPTY(&V_pf_allrulelist)); 6929 MPASS(V_pf_allrulecount == 0); 6930 6931 PF_RULES_WUNLOCK(); 6932 6933 free(V_pf_kifmarker, PFI_MTYPE); 6934 free(V_pf_rulemarker, M_PFRULE); 6935 #endif 6936 6937 /* Free counters last as we updated them during shutdown. */ 6938 pf_counter_u64_deinit(&V_pf_default_rule.evaluations); 6939 for (int i = 0; i < 2; i++) { 6940 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]); 6941 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]); 6942 } 6943 counter_u64_free(V_pf_default_rule.states_cur); 6944 counter_u64_free(V_pf_default_rule.states_tot); 6945 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 6946 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]); 6947 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp); 6948 6949 for (int i = 0; i < PFRES_MAX; i++) 6950 counter_u64_free(V_pf_status.counters[i]); 6951 for (int i = 0; i < KLCNT_MAX; i++) 6952 counter_u64_free(V_pf_status.lcounters[i]); 6953 for (int i = 0; i < FCNT_MAX; i++) 6954 pf_counter_u64_deinit(&V_pf_status.fcounters[i]); 6955 for (int i = 0; i < SCNT_MAX; i++) 6956 counter_u64_free(V_pf_status.scounters[i]); 6957 6958 rm_destroy(&V_pf_rules_lock); 6959 sx_destroy(&V_pf_ioctl_lock); 6960 } 6961 6962 static void 6963 pf_unload(void) 6964 { 6965 6966 sx_xlock(&pf_end_lock); 6967 pf_end_threads = 1; 6968 while (pf_end_threads < 2) { 6969 wakeup_one(pf_purge_thread); 6970 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0); 6971 } 6972 sx_xunlock(&pf_end_lock); 6973 6974 pf_nl_unregister(); 6975 6976 if (pf_dev != NULL) 6977 destroy_dev(pf_dev); 6978 6979 pfi_cleanup(); 6980 6981 sx_destroy(&pf_end_lock); 6982 } 6983 6984 static void 6985 vnet_pf_init(void *unused __unused) 6986 { 6987 6988 pf_load_vnet(); 6989 } 6990 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6991 vnet_pf_init, NULL); 6992 6993 static void 6994 vnet_pf_uninit(const void *unused __unused) 6995 { 6996 6997 pf_unload_vnet(); 6998 } 6999 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL); 7000 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 7001 vnet_pf_uninit, NULL); 7002 7003 static int 7004 pf_modevent(module_t mod, int type, void *data) 7005 { 7006 int error = 0; 7007 7008 switch(type) { 7009 case MOD_LOAD: 7010 error = pf_load(); 7011 pf_nl_register(); 7012 break; 7013 case MOD_UNLOAD: 7014 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after 7015 * the vnet_pf_uninit()s */ 7016 break; 7017 default: 7018 error = EINVAL; 7019 break; 7020 } 7021 7022 return (error); 7023 } 7024 7025 static moduledata_t pf_mod = { 7026 "pf", 7027 pf_modevent, 7028 0 7029 }; 7030 7031 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND); 7032 MODULE_DEPEND(pf, netlink, 1, 1, 1); 7033 MODULE_DEPEND(pf, crypto, 1, 1, 1); 7034 MODULE_VERSION(pf, PF_MODVER); 7035