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; 334 335 RB_INIT(&V_pf_anchors); 336 pf_init_kruleset(&pf_main_ruleset); 337 338 pf_init_keth(V_pf_keth); 339 340 /* default rule should never be garbage collected */ 341 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; 342 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS; 343 V_pf_default_rule.nr = (uint32_t)-1; 344 V_pf_default_rule.rtableid = -1; 345 346 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK); 347 for (int i = 0; i < 2; i++) { 348 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK); 349 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK); 350 } 351 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); 352 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); 353 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++) 354 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 355 356 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 357 M_WAITOK | M_ZERO); 358 359 #ifdef PF_WANT_32_TO_64_COUNTER 360 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO); 361 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO); 362 PF_RULES_WLOCK(); 363 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); 364 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist); 365 V_pf_allrulecount++; 366 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); 367 PF_RULES_WUNLOCK(); 368 #endif 369 370 /* initialize default timeouts */ 371 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 372 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; 373 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 374 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; 375 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; 376 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; 377 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 378 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL; 379 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 380 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL; 381 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL; 382 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; 383 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; 384 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; 385 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; 386 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; 387 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; 388 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; 389 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; 390 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; 391 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; 392 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; 393 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; 394 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; 395 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; 396 397 V_pf_status.debug = PF_DEBUG_URGENT; 398 /* 399 * XXX This is different than in OpenBSD where reassembly is enabled by 400 * defult. In FreeBSD we expect people to still use scrub rules and 401 * switch to the new syntax later. Only when they switch they must 402 * explicitly enable reassemle. We could change the default once the 403 * scrub rule functionality is hopefully removed some day in future. 404 */ 405 V_pf_status.reass = 0; 406 407 V_pf_pfil_hooked = false; 408 V_pf_pfil_eth_hooked = false; 409 410 /* XXX do our best to avoid a conflict */ 411 V_pf_status.hostid = arc4random(); 412 413 for (int i = 0; i < PFRES_MAX; i++) 414 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); 415 for (int i = 0; i < KLCNT_MAX; i++) 416 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); 417 for (int i = 0; i < FCNT_MAX; i++) 418 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK); 419 for (int i = 0; i < SCNT_MAX; i++) 420 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); 421 422 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET, 423 INTR_MPSAFE, &V_pf_swi_cookie) != 0) 424 /* XXXGL: leaked all above. */ 425 return; 426 } 427 428 static struct pf_kpool * 429 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action, 430 u_int32_t rule_number, u_int8_t r_last, u_int8_t active, 431 u_int8_t check_ticket, int which) 432 { 433 struct pf_kruleset *ruleset; 434 struct pf_krule *rule; 435 int rs_num; 436 437 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT); 438 439 ruleset = pf_find_kruleset(anchor); 440 if (ruleset == NULL) 441 return (NULL); 442 rs_num = pf_get_ruleset_number(rule_action); 443 if (rs_num >= PF_RULESET_MAX) 444 return (NULL); 445 if (active) { 446 if (check_ticket && ticket != 447 ruleset->rules[rs_num].active.ticket) 448 return (NULL); 449 if (r_last) 450 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 451 pf_krulequeue); 452 else 453 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 454 } else { 455 if (check_ticket && ticket != 456 ruleset->rules[rs_num].inactive.ticket) 457 return (NULL); 458 if (r_last) 459 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 460 pf_krulequeue); 461 else 462 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); 463 } 464 if (!r_last) { 465 while ((rule != NULL) && (rule->nr != rule_number)) 466 rule = TAILQ_NEXT(rule, entries); 467 } 468 if (rule == NULL) 469 return (NULL); 470 471 switch (which) { 472 case PF_RDR: 473 return (&rule->rdr); 474 case PF_NAT: 475 return (&rule->nat); 476 case PF_RT: 477 return (&rule->route); 478 default: 479 panic("Unknow pool type %d", which); 480 } 481 } 482 483 static void 484 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb) 485 { 486 struct pf_kpooladdr *mv_pool_pa; 487 488 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { 489 TAILQ_REMOVE(poola, mv_pool_pa, entries); 490 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); 491 } 492 } 493 494 static void 495 pf_empty_kpool(struct pf_kpalist *poola) 496 { 497 struct pf_kpooladdr *pa; 498 499 while ((pa = TAILQ_FIRST(poola)) != NULL) { 500 switch (pa->addr.type) { 501 case PF_ADDR_DYNIFTL: 502 pfi_dynaddr_remove(pa->addr.p.dyn); 503 break; 504 case PF_ADDR_TABLE: 505 /* XXX: this could be unfinished pooladdr on pabuf */ 506 if (pa->addr.p.tbl != NULL) 507 pfr_detach_table(pa->addr.p.tbl); 508 break; 509 } 510 if (pa->kif) 511 pfi_kkif_unref(pa->kif); 512 TAILQ_REMOVE(poola, pa, entries); 513 free(pa, M_PFRULE); 514 } 515 } 516 517 static void 518 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 519 { 520 521 PF_RULES_WASSERT(); 522 PF_UNLNKDRULES_ASSERT(); 523 524 TAILQ_REMOVE(rulequeue, rule, entries); 525 526 rule->rule_ref |= PFRULE_REFS; 527 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); 528 } 529 530 static void 531 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 532 { 533 534 PF_RULES_WASSERT(); 535 536 PF_UNLNKDRULES_LOCK(); 537 pf_unlink_rule_locked(rulequeue, rule); 538 PF_UNLNKDRULES_UNLOCK(); 539 } 540 541 static void 542 pf_free_eth_rule(struct pf_keth_rule *rule) 543 { 544 PF_RULES_WASSERT(); 545 546 if (rule == NULL) 547 return; 548 549 if (rule->tag) 550 tag_unref(&V_pf_tags, rule->tag); 551 if (rule->match_tag) 552 tag_unref(&V_pf_tags, rule->match_tag); 553 #ifdef ALTQ 554 pf_qid_unref(rule->qid); 555 #endif 556 557 if (rule->bridge_to) 558 pfi_kkif_unref(rule->bridge_to); 559 if (rule->kif) 560 pfi_kkif_unref(rule->kif); 561 562 if (rule->ipsrc.addr.type == PF_ADDR_TABLE) 563 pfr_detach_table(rule->ipsrc.addr.p.tbl); 564 if (rule->ipdst.addr.type == PF_ADDR_TABLE) 565 pfr_detach_table(rule->ipdst.addr.p.tbl); 566 567 counter_u64_free(rule->evaluations); 568 for (int i = 0; i < 2; i++) { 569 counter_u64_free(rule->packets[i]); 570 counter_u64_free(rule->bytes[i]); 571 } 572 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 573 pf_keth_anchor_remove(rule); 574 575 free(rule, M_PFRULE); 576 } 577 578 void 579 pf_free_rule(struct pf_krule *rule) 580 { 581 582 PF_RULES_WASSERT(); 583 PF_CONFIG_ASSERT(); 584 585 if (rule->tag) 586 tag_unref(&V_pf_tags, rule->tag); 587 if (rule->match_tag) 588 tag_unref(&V_pf_tags, rule->match_tag); 589 #ifdef ALTQ 590 if (rule->pqid != rule->qid) 591 pf_qid_unref(rule->pqid); 592 pf_qid_unref(rule->qid); 593 #endif 594 switch (rule->src.addr.type) { 595 case PF_ADDR_DYNIFTL: 596 pfi_dynaddr_remove(rule->src.addr.p.dyn); 597 break; 598 case PF_ADDR_TABLE: 599 pfr_detach_table(rule->src.addr.p.tbl); 600 break; 601 } 602 switch (rule->dst.addr.type) { 603 case PF_ADDR_DYNIFTL: 604 pfi_dynaddr_remove(rule->dst.addr.p.dyn); 605 break; 606 case PF_ADDR_TABLE: 607 pfr_detach_table(rule->dst.addr.p.tbl); 608 break; 609 } 610 if (rule->overload_tbl) 611 pfr_detach_table(rule->overload_tbl); 612 if (rule->kif) 613 pfi_kkif_unref(rule->kif); 614 if (rule->rcv_kif) 615 pfi_kkif_unref(rule->rcv_kif); 616 pf_remove_kanchor(rule); 617 pf_empty_kpool(&rule->rdr.list); 618 pf_empty_kpool(&rule->nat.list); 619 pf_empty_kpool(&rule->route.list); 620 621 pf_krule_free(rule); 622 } 623 624 static void 625 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size, 626 unsigned int default_size) 627 { 628 unsigned int i; 629 unsigned int hashsize; 630 631 if (*tunable_size == 0 || !powerof2(*tunable_size)) 632 *tunable_size = default_size; 633 634 hashsize = *tunable_size; 635 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH, 636 M_WAITOK); 637 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH, 638 M_WAITOK); 639 ts->mask = hashsize - 1; 640 ts->seed = arc4random(); 641 for (i = 0; i < hashsize; i++) { 642 TAILQ_INIT(&ts->namehash[i]); 643 TAILQ_INIT(&ts->taghash[i]); 644 } 645 BIT_FILL(TAGID_MAX, &ts->avail); 646 } 647 648 static void 649 pf_cleanup_tagset(struct pf_tagset *ts) 650 { 651 unsigned int i; 652 unsigned int hashsize; 653 struct pf_tagname *t, *tmp; 654 655 /* 656 * Only need to clean up one of the hashes as each tag is hashed 657 * into each table. 658 */ 659 hashsize = ts->mask + 1; 660 for (i = 0; i < hashsize; i++) 661 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp) 662 uma_zfree(V_pf_tag_z, t); 663 664 free(ts->namehash, M_PFHASH); 665 free(ts->taghash, M_PFHASH); 666 } 667 668 static uint16_t 669 tagname2hashindex(const struct pf_tagset *ts, const char *tagname) 670 { 671 size_t len; 672 673 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1); 674 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask); 675 } 676 677 static uint16_t 678 tag2hashindex(const struct pf_tagset *ts, uint16_t tag) 679 { 680 681 return (tag & ts->mask); 682 } 683 684 static u_int16_t 685 tagname2tag(struct pf_tagset *ts, const char *tagname) 686 { 687 struct pf_tagname *tag; 688 u_int32_t index; 689 u_int16_t new_tagid; 690 691 PF_RULES_WASSERT(); 692 693 index = tagname2hashindex(ts, tagname); 694 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 695 if (strcmp(tagname, tag->name) == 0) { 696 tag->ref++; 697 return (tag->tag); 698 } 699 700 /* 701 * new entry 702 * 703 * to avoid fragmentation, we do a linear search from the beginning 704 * and take the first free slot we find. 705 */ 706 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail); 707 /* 708 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX]. 709 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits 710 * set. It may also return a bit number greater than TAGID_MAX due 711 * to rounding of the number of bits in the vector up to a multiple 712 * of the vector word size at declaration/allocation time. 713 */ 714 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) 715 return (0); 716 717 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */ 718 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail); 719 720 /* allocate and fill new struct pf_tagname */ 721 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT); 722 if (tag == NULL) 723 return (0); 724 strlcpy(tag->name, tagname, sizeof(tag->name)); 725 tag->tag = new_tagid; 726 tag->ref = 1; 727 728 /* Insert into namehash */ 729 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries); 730 731 /* Insert into taghash */ 732 index = tag2hashindex(ts, new_tagid); 733 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries); 734 735 return (tag->tag); 736 } 737 738 static void 739 tag_unref(struct pf_tagset *ts, u_int16_t tag) 740 { 741 struct pf_tagname *t; 742 uint16_t index; 743 744 PF_RULES_WASSERT(); 745 746 index = tag2hashindex(ts, tag); 747 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 748 if (tag == t->tag) { 749 if (--t->ref == 0) { 750 TAILQ_REMOVE(&ts->taghash[index], t, 751 taghash_entries); 752 index = tagname2hashindex(ts, t->name); 753 TAILQ_REMOVE(&ts->namehash[index], t, 754 namehash_entries); 755 /* Bits are 0-based for BIT_SET() */ 756 BIT_SET(TAGID_MAX, tag - 1, &ts->avail); 757 uma_zfree(V_pf_tag_z, t); 758 } 759 break; 760 } 761 } 762 763 static uint16_t 764 pf_tagname2tag(const char *tagname) 765 { 766 return (tagname2tag(&V_pf_tags, tagname)); 767 } 768 769 static int 770 pf_begin_eth(uint32_t *ticket, const char *anchor) 771 { 772 struct pf_keth_rule *rule, *tmp; 773 struct pf_keth_ruleset *rs; 774 775 PF_RULES_WASSERT(); 776 777 rs = pf_find_or_create_keth_ruleset(anchor); 778 if (rs == NULL) 779 return (EINVAL); 780 781 /* Purge old inactive rules. */ 782 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 783 tmp) { 784 TAILQ_REMOVE(rs->inactive.rules, rule, 785 entries); 786 pf_free_eth_rule(rule); 787 } 788 789 *ticket = ++rs->inactive.ticket; 790 rs->inactive.open = 1; 791 792 return (0); 793 } 794 795 static int 796 pf_rollback_eth(uint32_t ticket, const char *anchor) 797 { 798 struct pf_keth_rule *rule, *tmp; 799 struct pf_keth_ruleset *rs; 800 801 PF_RULES_WASSERT(); 802 803 rs = pf_find_keth_ruleset(anchor); 804 if (rs == NULL) 805 return (EINVAL); 806 807 if (!rs->inactive.open || 808 ticket != rs->inactive.ticket) 809 return (0); 810 811 /* Purge old inactive rules. */ 812 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 813 tmp) { 814 TAILQ_REMOVE(rs->inactive.rules, rule, entries); 815 pf_free_eth_rule(rule); 816 } 817 818 rs->inactive.open = 0; 819 820 pf_remove_if_empty_keth_ruleset(rs); 821 822 return (0); 823 } 824 825 #define PF_SET_SKIP_STEPS(i) \ 826 do { \ 827 while (head[i] != cur) { \ 828 head[i]->skip[i].ptr = cur; \ 829 head[i] = TAILQ_NEXT(head[i], entries); \ 830 } \ 831 } while (0) 832 833 static void 834 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules) 835 { 836 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT]; 837 int i; 838 839 cur = TAILQ_FIRST(rules); 840 prev = cur; 841 for (i = 0; i < PFE_SKIP_COUNT; ++i) 842 head[i] = cur; 843 while (cur != NULL) { 844 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 845 PF_SET_SKIP_STEPS(PFE_SKIP_IFP); 846 if (cur->direction != prev->direction) 847 PF_SET_SKIP_STEPS(PFE_SKIP_DIR); 848 if (cur->proto != prev->proto) 849 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO); 850 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0) 851 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR); 852 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0) 853 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR); 854 if (cur->ipsrc.neg != prev->ipsrc.neg || 855 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr)) 856 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR); 857 if (cur->ipdst.neg != prev->ipdst.neg || 858 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr)) 859 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR); 860 861 prev = cur; 862 cur = TAILQ_NEXT(cur, entries); 863 } 864 for (i = 0; i < PFE_SKIP_COUNT; ++i) 865 PF_SET_SKIP_STEPS(i); 866 } 867 868 static int 869 pf_commit_eth(uint32_t ticket, const char *anchor) 870 { 871 struct pf_keth_ruleq *rules; 872 struct pf_keth_ruleset *rs; 873 874 rs = pf_find_keth_ruleset(anchor); 875 if (rs == NULL) { 876 return (EINVAL); 877 } 878 879 if (!rs->inactive.open || 880 ticket != rs->inactive.ticket) 881 return (EBUSY); 882 883 PF_RULES_WASSERT(); 884 885 pf_eth_calc_skip_steps(rs->inactive.rules); 886 887 rules = rs->active.rules; 888 atomic_store_ptr(&rs->active.rules, rs->inactive.rules); 889 rs->inactive.rules = rules; 890 rs->inactive.ticket = rs->active.ticket; 891 892 return (pf_rollback_eth(rs->inactive.ticket, 893 rs->anchor ? rs->anchor->path : "")); 894 } 895 896 #ifdef ALTQ 897 static uint16_t 898 pf_qname2qid(const char *qname) 899 { 900 return (tagname2tag(&V_pf_qids, qname)); 901 } 902 903 static void 904 pf_qid_unref(uint16_t qid) 905 { 906 tag_unref(&V_pf_qids, qid); 907 } 908 909 static int 910 pf_begin_altq(u_int32_t *ticket) 911 { 912 struct pf_altq *altq, *tmp; 913 int error = 0; 914 915 PF_RULES_WASSERT(); 916 917 /* Purge the old altq lists */ 918 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 919 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 920 /* detach and destroy the discipline */ 921 error = altq_remove(altq); 922 } 923 free(altq, M_PFALTQ); 924 } 925 TAILQ_INIT(V_pf_altq_ifs_inactive); 926 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 927 pf_qid_unref(altq->qid); 928 free(altq, M_PFALTQ); 929 } 930 TAILQ_INIT(V_pf_altqs_inactive); 931 if (error) 932 return (error); 933 *ticket = ++V_ticket_altqs_inactive; 934 V_altqs_inactive_open = 1; 935 return (0); 936 } 937 938 static int 939 pf_rollback_altq(u_int32_t ticket) 940 { 941 struct pf_altq *altq, *tmp; 942 int error = 0; 943 944 PF_RULES_WASSERT(); 945 946 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 947 return (0); 948 /* Purge the old altq lists */ 949 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 950 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 951 /* detach and destroy the discipline */ 952 error = altq_remove(altq); 953 } 954 free(altq, M_PFALTQ); 955 } 956 TAILQ_INIT(V_pf_altq_ifs_inactive); 957 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 958 pf_qid_unref(altq->qid); 959 free(altq, M_PFALTQ); 960 } 961 TAILQ_INIT(V_pf_altqs_inactive); 962 V_altqs_inactive_open = 0; 963 return (error); 964 } 965 966 static int 967 pf_commit_altq(u_int32_t ticket) 968 { 969 struct pf_altqqueue *old_altqs, *old_altq_ifs; 970 struct pf_altq *altq, *tmp; 971 int err, error = 0; 972 973 PF_RULES_WASSERT(); 974 975 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 976 return (EBUSY); 977 978 /* swap altqs, keep the old. */ 979 old_altqs = V_pf_altqs_active; 980 old_altq_ifs = V_pf_altq_ifs_active; 981 V_pf_altqs_active = V_pf_altqs_inactive; 982 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive; 983 V_pf_altqs_inactive = old_altqs; 984 V_pf_altq_ifs_inactive = old_altq_ifs; 985 V_ticket_altqs_active = V_ticket_altqs_inactive; 986 987 /* Attach new disciplines */ 988 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 989 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 990 /* attach the discipline */ 991 error = altq_pfattach(altq); 992 if (error == 0 && V_pf_altq_running) 993 error = pf_enable_altq(altq); 994 if (error != 0) 995 return (error); 996 } 997 } 998 999 /* Purge the old altq lists */ 1000 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 1001 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 1002 /* detach and destroy the discipline */ 1003 if (V_pf_altq_running) 1004 error = pf_disable_altq(altq); 1005 err = altq_pfdetach(altq); 1006 if (err != 0 && error == 0) 1007 error = err; 1008 err = altq_remove(altq); 1009 if (err != 0 && error == 0) 1010 error = err; 1011 } 1012 free(altq, M_PFALTQ); 1013 } 1014 TAILQ_INIT(V_pf_altq_ifs_inactive); 1015 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1016 pf_qid_unref(altq->qid); 1017 free(altq, M_PFALTQ); 1018 } 1019 TAILQ_INIT(V_pf_altqs_inactive); 1020 1021 V_altqs_inactive_open = 0; 1022 return (error); 1023 } 1024 1025 static int 1026 pf_enable_altq(struct pf_altq *altq) 1027 { 1028 struct ifnet *ifp; 1029 struct tb_profile tb; 1030 int error = 0; 1031 1032 if ((ifp = ifunit(altq->ifname)) == NULL) 1033 return (EINVAL); 1034 1035 if (ifp->if_snd.altq_type != ALTQT_NONE) 1036 error = altq_enable(&ifp->if_snd); 1037 1038 /* set tokenbucket regulator */ 1039 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { 1040 tb.rate = altq->ifbandwidth; 1041 tb.depth = altq->tbrsize; 1042 error = tbr_set(&ifp->if_snd, &tb); 1043 } 1044 1045 return (error); 1046 } 1047 1048 static int 1049 pf_disable_altq(struct pf_altq *altq) 1050 { 1051 struct ifnet *ifp; 1052 struct tb_profile tb; 1053 int error; 1054 1055 if ((ifp = ifunit(altq->ifname)) == NULL) 1056 return (EINVAL); 1057 1058 /* 1059 * when the discipline is no longer referenced, it was overridden 1060 * by a new one. if so, just return. 1061 */ 1062 if (altq->altq_disc != ifp->if_snd.altq_disc) 1063 return (0); 1064 1065 error = altq_disable(&ifp->if_snd); 1066 1067 if (error == 0) { 1068 /* clear tokenbucket regulator */ 1069 tb.rate = 0; 1070 error = tbr_set(&ifp->if_snd, &tb); 1071 } 1072 1073 return (error); 1074 } 1075 1076 static int 1077 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket, 1078 struct pf_altq *altq) 1079 { 1080 struct ifnet *ifp1; 1081 int error = 0; 1082 1083 /* Deactivate the interface in question */ 1084 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; 1085 if ((ifp1 = ifunit(altq->ifname)) == NULL || 1086 (remove && ifp1 == ifp)) { 1087 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 1088 } else { 1089 error = altq_add(ifp1, altq); 1090 1091 if (ticket != V_ticket_altqs_inactive) 1092 error = EBUSY; 1093 1094 if (error) 1095 free(altq, M_PFALTQ); 1096 } 1097 1098 return (error); 1099 } 1100 1101 void 1102 pf_altq_ifnet_event(struct ifnet *ifp, int remove) 1103 { 1104 struct pf_altq *a1, *a2, *a3; 1105 u_int32_t ticket; 1106 int error = 0; 1107 1108 /* 1109 * No need to re-evaluate the configuration for events on interfaces 1110 * that do not support ALTQ, as it's not possible for such 1111 * interfaces to be part of the configuration. 1112 */ 1113 if (!ALTQ_IS_READY(&ifp->if_snd)) 1114 return; 1115 1116 /* Interrupt userland queue modifications */ 1117 if (V_altqs_inactive_open) 1118 pf_rollback_altq(V_ticket_altqs_inactive); 1119 1120 /* Start new altq ruleset */ 1121 if (pf_begin_altq(&ticket)) 1122 return; 1123 1124 /* Copy the current active set */ 1125 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) { 1126 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1127 if (a2 == NULL) { 1128 error = ENOMEM; 1129 break; 1130 } 1131 bcopy(a1, a2, sizeof(struct pf_altq)); 1132 1133 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1134 if (error) 1135 break; 1136 1137 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries); 1138 } 1139 if (error) 1140 goto out; 1141 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { 1142 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1143 if (a2 == NULL) { 1144 error = ENOMEM; 1145 break; 1146 } 1147 bcopy(a1, a2, sizeof(struct pf_altq)); 1148 1149 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) { 1150 error = EBUSY; 1151 free(a2, M_PFALTQ); 1152 break; 1153 } 1154 a2->altq_disc = NULL; 1155 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) { 1156 if (strncmp(a3->ifname, a2->ifname, 1157 IFNAMSIZ) == 0) { 1158 a2->altq_disc = a3->altq_disc; 1159 break; 1160 } 1161 } 1162 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1163 if (error) 1164 break; 1165 1166 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); 1167 } 1168 1169 out: 1170 if (error != 0) 1171 pf_rollback_altq(ticket); 1172 else 1173 pf_commit_altq(ticket); 1174 } 1175 #endif /* ALTQ */ 1176 1177 static struct pf_krule_global * 1178 pf_rule_tree_alloc(int flags) 1179 { 1180 struct pf_krule_global *tree; 1181 1182 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags); 1183 if (tree == NULL) 1184 return (NULL); 1185 RB_INIT(tree); 1186 return (tree); 1187 } 1188 1189 static void 1190 pf_rule_tree_free(struct pf_krule_global *tree) 1191 { 1192 1193 free(tree, M_TEMP); 1194 } 1195 1196 static int 1197 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) 1198 { 1199 struct pf_krule_global *tree; 1200 struct pf_kruleset *rs; 1201 struct pf_krule *rule; 1202 1203 PF_RULES_WASSERT(); 1204 1205 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1206 return (EINVAL); 1207 tree = pf_rule_tree_alloc(M_NOWAIT); 1208 if (tree == NULL) 1209 return (ENOMEM); 1210 rs = pf_find_or_create_kruleset(anchor); 1211 if (rs == NULL) { 1212 free(tree, M_TEMP); 1213 return (EINVAL); 1214 } 1215 pf_rule_tree_free(rs->rules[rs_num].inactive.tree); 1216 rs->rules[rs_num].inactive.tree = tree; 1217 1218 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1219 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1220 rs->rules[rs_num].inactive.rcount--; 1221 } 1222 *ticket = ++rs->rules[rs_num].inactive.ticket; 1223 rs->rules[rs_num].inactive.open = 1; 1224 return (0); 1225 } 1226 1227 static int 1228 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) 1229 { 1230 struct pf_kruleset *rs; 1231 struct pf_krule *rule; 1232 1233 PF_RULES_WASSERT(); 1234 1235 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1236 return (EINVAL); 1237 rs = pf_find_kruleset(anchor); 1238 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1239 rs->rules[rs_num].inactive.ticket != ticket) 1240 return (0); 1241 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1242 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1243 rs->rules[rs_num].inactive.rcount--; 1244 } 1245 rs->rules[rs_num].inactive.open = 0; 1246 return (0); 1247 } 1248 1249 #define PF_MD5_UPD(st, elm) \ 1250 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) 1251 1252 #define PF_MD5_UPD_STR(st, elm) \ 1253 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) 1254 1255 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ 1256 (stor) = htonl((st)->elm); \ 1257 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ 1258 } while (0) 1259 1260 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ 1261 (stor) = htons((st)->elm); \ 1262 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ 1263 } while (0) 1264 1265 static void 1266 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) 1267 { 1268 PF_MD5_UPD(pfr, addr.type); 1269 switch (pfr->addr.type) { 1270 case PF_ADDR_DYNIFTL: 1271 PF_MD5_UPD(pfr, addr.v.ifname); 1272 PF_MD5_UPD(pfr, addr.iflags); 1273 break; 1274 case PF_ADDR_TABLE: 1275 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX, 1276 strlen(PF_OPTIMIZER_TABLE_PFX))) 1277 PF_MD5_UPD(pfr, addr.v.tblname); 1278 break; 1279 case PF_ADDR_ADDRMASK: 1280 /* XXX ignore af? */ 1281 PF_MD5_UPD(pfr, addr.v.a.addr.addr32); 1282 PF_MD5_UPD(pfr, addr.v.a.mask.addr32); 1283 break; 1284 } 1285 1286 PF_MD5_UPD(pfr, port[0]); 1287 PF_MD5_UPD(pfr, port[1]); 1288 PF_MD5_UPD(pfr, neg); 1289 PF_MD5_UPD(pfr, port_op); 1290 } 1291 1292 static void 1293 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule) 1294 { 1295 u_int16_t x; 1296 u_int32_t y; 1297 1298 pf_hash_rule_addr(ctx, &rule->src); 1299 pf_hash_rule_addr(ctx, &rule->dst); 1300 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) 1301 PF_MD5_UPD_STR(rule, label[i]); 1302 PF_MD5_UPD_STR(rule, ifname); 1303 PF_MD5_UPD_STR(rule, rcv_ifname); 1304 PF_MD5_UPD_STR(rule, match_tagname); 1305 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ 1306 PF_MD5_UPD_HTONL(rule, os_fingerprint, y); 1307 PF_MD5_UPD_HTONL(rule, prob, y); 1308 PF_MD5_UPD_HTONL(rule, uid.uid[0], y); 1309 PF_MD5_UPD_HTONL(rule, uid.uid[1], y); 1310 PF_MD5_UPD(rule, uid.op); 1311 PF_MD5_UPD_HTONL(rule, gid.gid[0], y); 1312 PF_MD5_UPD_HTONL(rule, gid.gid[1], y); 1313 PF_MD5_UPD(rule, gid.op); 1314 PF_MD5_UPD_HTONL(rule, rule_flag, y); 1315 PF_MD5_UPD(rule, action); 1316 PF_MD5_UPD(rule, direction); 1317 PF_MD5_UPD(rule, af); 1318 PF_MD5_UPD(rule, quick); 1319 PF_MD5_UPD(rule, ifnot); 1320 PF_MD5_UPD(rule, rcvifnot); 1321 PF_MD5_UPD(rule, match_tag_not); 1322 PF_MD5_UPD(rule, natpass); 1323 PF_MD5_UPD(rule, keep_state); 1324 PF_MD5_UPD(rule, proto); 1325 PF_MD5_UPD(rule, type); 1326 PF_MD5_UPD(rule, code); 1327 PF_MD5_UPD(rule, flags); 1328 PF_MD5_UPD(rule, flagset); 1329 PF_MD5_UPD(rule, allow_opts); 1330 PF_MD5_UPD(rule, rt); 1331 PF_MD5_UPD(rule, tos); 1332 PF_MD5_UPD(rule, scrub_flags); 1333 PF_MD5_UPD(rule, min_ttl); 1334 PF_MD5_UPD(rule, set_tos); 1335 if (rule->anchor != NULL) 1336 PF_MD5_UPD_STR(rule, anchor->path); 1337 } 1338 1339 static void 1340 pf_hash_rule(struct pf_krule *rule) 1341 { 1342 MD5_CTX ctx; 1343 1344 MD5Init(&ctx); 1345 pf_hash_rule_rolling(&ctx, rule); 1346 MD5Final(rule->md5sum, &ctx); 1347 } 1348 1349 static int 1350 pf_krule_compare(struct pf_krule *a, struct pf_krule *b) 1351 { 1352 1353 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH)); 1354 } 1355 1356 static int 1357 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) 1358 { 1359 struct pf_kruleset *rs; 1360 struct pf_krule *rule, *old_rule; 1361 struct pf_krulequeue *old_rules; 1362 struct pf_krule_global *old_tree; 1363 int error; 1364 u_int32_t old_rcount; 1365 1366 PF_RULES_WASSERT(); 1367 1368 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1369 return (EINVAL); 1370 rs = pf_find_kruleset(anchor); 1371 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1372 ticket != rs->rules[rs_num].inactive.ticket) 1373 return (EBUSY); 1374 1375 /* Calculate checksum for the main ruleset */ 1376 if (rs == &pf_main_ruleset) { 1377 error = pf_setup_pfsync_matching(rs); 1378 if (error != 0) 1379 return (error); 1380 } 1381 1382 /* Swap rules, keep the old. */ 1383 old_rules = rs->rules[rs_num].active.ptr; 1384 old_rcount = rs->rules[rs_num].active.rcount; 1385 old_tree = rs->rules[rs_num].active.tree; 1386 1387 rs->rules[rs_num].active.ptr = 1388 rs->rules[rs_num].inactive.ptr; 1389 rs->rules[rs_num].active.tree = 1390 rs->rules[rs_num].inactive.tree; 1391 rs->rules[rs_num].active.rcount = 1392 rs->rules[rs_num].inactive.rcount; 1393 1394 /* Attempt to preserve counter information. */ 1395 if (V_pf_status.keep_counters && old_tree != NULL) { 1396 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr, 1397 entries) { 1398 old_rule = RB_FIND(pf_krule_global, old_tree, rule); 1399 if (old_rule == NULL) { 1400 continue; 1401 } 1402 pf_counter_u64_critical_enter(); 1403 pf_counter_u64_rollup_protected(&rule->evaluations, 1404 pf_counter_u64_fetch(&old_rule->evaluations)); 1405 pf_counter_u64_rollup_protected(&rule->packets[0], 1406 pf_counter_u64_fetch(&old_rule->packets[0])); 1407 pf_counter_u64_rollup_protected(&rule->packets[1], 1408 pf_counter_u64_fetch(&old_rule->packets[1])); 1409 pf_counter_u64_rollup_protected(&rule->bytes[0], 1410 pf_counter_u64_fetch(&old_rule->bytes[0])); 1411 pf_counter_u64_rollup_protected(&rule->bytes[1], 1412 pf_counter_u64_fetch(&old_rule->bytes[1])); 1413 pf_counter_u64_critical_exit(); 1414 } 1415 } 1416 1417 rs->rules[rs_num].inactive.ptr = old_rules; 1418 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ 1419 rs->rules[rs_num].inactive.rcount = old_rcount; 1420 1421 rs->rules[rs_num].active.ticket = 1422 rs->rules[rs_num].inactive.ticket; 1423 pf_calc_skip_steps(rs->rules[rs_num].active.ptr); 1424 1425 /* Purge the old rule list. */ 1426 PF_UNLNKDRULES_LOCK(); 1427 while ((rule = TAILQ_FIRST(old_rules)) != NULL) 1428 pf_unlink_rule_locked(old_rules, rule); 1429 PF_UNLNKDRULES_UNLOCK(); 1430 rs->rules[rs_num].inactive.rcount = 0; 1431 rs->rules[rs_num].inactive.open = 0; 1432 pf_remove_if_empty_kruleset(rs); 1433 free(old_tree, M_TEMP); 1434 1435 return (0); 1436 } 1437 1438 static int 1439 pf_setup_pfsync_matching(struct pf_kruleset *rs) 1440 { 1441 MD5_CTX ctx; 1442 struct pf_krule *rule; 1443 int rs_cnt; 1444 u_int8_t digest[PF_MD5_DIGEST_LENGTH]; 1445 1446 MD5Init(&ctx); 1447 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { 1448 /* XXX PF_RULESET_SCRUB as well? */ 1449 if (rs_cnt == PF_RULESET_SCRUB) 1450 continue; 1451 1452 if (rs->rules[rs_cnt].inactive.rcount) { 1453 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, 1454 entries) { 1455 pf_hash_rule_rolling(&ctx, rule); 1456 } 1457 } 1458 } 1459 1460 MD5Final(digest, &ctx); 1461 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); 1462 return (0); 1463 } 1464 1465 static int 1466 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr) 1467 { 1468 int error = 0; 1469 1470 switch (addr->type) { 1471 case PF_ADDR_TABLE: 1472 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname); 1473 if (addr->p.tbl == NULL) 1474 error = ENOMEM; 1475 break; 1476 default: 1477 error = EINVAL; 1478 } 1479 1480 return (error); 1481 } 1482 1483 static int 1484 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr, 1485 sa_family_t af) 1486 { 1487 int error = 0; 1488 1489 switch (addr->type) { 1490 case PF_ADDR_TABLE: 1491 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); 1492 if (addr->p.tbl == NULL) 1493 error = ENOMEM; 1494 break; 1495 case PF_ADDR_DYNIFTL: 1496 error = pfi_dynaddr_setup(addr, af); 1497 break; 1498 } 1499 1500 return (error); 1501 } 1502 1503 void 1504 pf_addr_copyout(struct pf_addr_wrap *addr) 1505 { 1506 1507 switch (addr->type) { 1508 case PF_ADDR_DYNIFTL: 1509 pfi_dynaddr_copyout(addr); 1510 break; 1511 case PF_ADDR_TABLE: 1512 pf_tbladdr_copyout(addr); 1513 break; 1514 } 1515 } 1516 1517 static void 1518 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out) 1519 { 1520 int secs = time_uptime; 1521 1522 bzero(out, sizeof(struct pf_src_node)); 1523 1524 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr)); 1525 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr)); 1526 1527 if (in->rule != NULL) 1528 out->rule.nr = in->rule->nr; 1529 1530 for (int i = 0; i < 2; i++) { 1531 out->bytes[i] = counter_u64_fetch(in->bytes[i]); 1532 out->packets[i] = counter_u64_fetch(in->packets[i]); 1533 } 1534 1535 out->states = in->states; 1536 out->conn = in->conn; 1537 out->af = in->af; 1538 out->ruletype = in->ruletype; 1539 1540 out->creation = secs - in->creation; 1541 if (out->expire > secs) 1542 out->expire -= secs; 1543 else 1544 out->expire = 0; 1545 1546 /* Adjust the connection rate estimate. */ 1547 out->conn_rate.limit = in->conn_rate.limit; 1548 out->conn_rate.seconds = in->conn_rate.seconds; 1549 /* If there's no limit there's no counter_rate. */ 1550 if (in->conn_rate.cr != NULL) 1551 out->conn_rate.count = counter_rate_get(in->conn_rate.cr); 1552 } 1553 1554 #ifdef ALTQ 1555 /* 1556 * Handle export of struct pf_kaltq to user binaries that may be using any 1557 * version of struct pf_altq. 1558 */ 1559 static int 1560 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size) 1561 { 1562 u_int32_t version; 1563 1564 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1565 version = 0; 1566 else 1567 version = pa->version; 1568 1569 if (version > PFIOC_ALTQ_VERSION) 1570 return (EINVAL); 1571 1572 #define ASSIGN(x) exported_q->x = q->x 1573 #define COPY(x) \ 1574 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x))) 1575 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX) 1576 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX) 1577 1578 switch (version) { 1579 case 0: { 1580 struct pf_altq_v0 *exported_q = 1581 &((struct pfioc_altq_v0 *)pa)->altq; 1582 1583 COPY(ifname); 1584 1585 ASSIGN(scheduler); 1586 ASSIGN(tbrsize); 1587 exported_q->tbrsize = SATU16(q->tbrsize); 1588 exported_q->ifbandwidth = SATU32(q->ifbandwidth); 1589 1590 COPY(qname); 1591 COPY(parent); 1592 ASSIGN(parent_qid); 1593 exported_q->bandwidth = SATU32(q->bandwidth); 1594 ASSIGN(priority); 1595 ASSIGN(local_flags); 1596 1597 ASSIGN(qlimit); 1598 ASSIGN(flags); 1599 1600 if (q->scheduler == ALTQT_HFSC) { 1601 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x 1602 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \ 1603 SATU32(q->pq_u.hfsc_opts.x) 1604 1605 ASSIGN_OPT_SATU32(rtsc_m1); 1606 ASSIGN_OPT(rtsc_d); 1607 ASSIGN_OPT_SATU32(rtsc_m2); 1608 1609 ASSIGN_OPT_SATU32(lssc_m1); 1610 ASSIGN_OPT(lssc_d); 1611 ASSIGN_OPT_SATU32(lssc_m2); 1612 1613 ASSIGN_OPT_SATU32(ulsc_m1); 1614 ASSIGN_OPT(ulsc_d); 1615 ASSIGN_OPT_SATU32(ulsc_m2); 1616 1617 ASSIGN_OPT(flags); 1618 1619 #undef ASSIGN_OPT 1620 #undef ASSIGN_OPT_SATU32 1621 } else 1622 COPY(pq_u); 1623 1624 ASSIGN(qid); 1625 break; 1626 } 1627 case 1: { 1628 struct pf_altq_v1 *exported_q = 1629 &((struct pfioc_altq_v1 *)pa)->altq; 1630 1631 COPY(ifname); 1632 1633 ASSIGN(scheduler); 1634 ASSIGN(tbrsize); 1635 ASSIGN(ifbandwidth); 1636 1637 COPY(qname); 1638 COPY(parent); 1639 ASSIGN(parent_qid); 1640 ASSIGN(bandwidth); 1641 ASSIGN(priority); 1642 ASSIGN(local_flags); 1643 1644 ASSIGN(qlimit); 1645 ASSIGN(flags); 1646 COPY(pq_u); 1647 1648 ASSIGN(qid); 1649 break; 1650 } 1651 default: 1652 panic("%s: unhandled struct pfioc_altq version", __func__); 1653 break; 1654 } 1655 1656 #undef ASSIGN 1657 #undef COPY 1658 #undef SATU16 1659 #undef SATU32 1660 1661 return (0); 1662 } 1663 1664 /* 1665 * Handle import to struct pf_kaltq of struct pf_altq from user binaries 1666 * that may be using any version of it. 1667 */ 1668 static int 1669 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size) 1670 { 1671 u_int32_t version; 1672 1673 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1674 version = 0; 1675 else 1676 version = pa->version; 1677 1678 if (version > PFIOC_ALTQ_VERSION) 1679 return (EINVAL); 1680 1681 #define ASSIGN(x) q->x = imported_q->x 1682 #define COPY(x) \ 1683 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x))) 1684 1685 switch (version) { 1686 case 0: { 1687 struct pf_altq_v0 *imported_q = 1688 &((struct pfioc_altq_v0 *)pa)->altq; 1689 1690 COPY(ifname); 1691 1692 ASSIGN(scheduler); 1693 ASSIGN(tbrsize); /* 16-bit -> 32-bit */ 1694 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */ 1695 1696 COPY(qname); 1697 COPY(parent); 1698 ASSIGN(parent_qid); 1699 ASSIGN(bandwidth); /* 32-bit -> 64-bit */ 1700 ASSIGN(priority); 1701 ASSIGN(local_flags); 1702 1703 ASSIGN(qlimit); 1704 ASSIGN(flags); 1705 1706 if (imported_q->scheduler == ALTQT_HFSC) { 1707 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x 1708 1709 /* 1710 * The m1 and m2 parameters are being copied from 1711 * 32-bit to 64-bit. 1712 */ 1713 ASSIGN_OPT(rtsc_m1); 1714 ASSIGN_OPT(rtsc_d); 1715 ASSIGN_OPT(rtsc_m2); 1716 1717 ASSIGN_OPT(lssc_m1); 1718 ASSIGN_OPT(lssc_d); 1719 ASSIGN_OPT(lssc_m2); 1720 1721 ASSIGN_OPT(ulsc_m1); 1722 ASSIGN_OPT(ulsc_d); 1723 ASSIGN_OPT(ulsc_m2); 1724 1725 ASSIGN_OPT(flags); 1726 1727 #undef ASSIGN_OPT 1728 } else 1729 COPY(pq_u); 1730 1731 ASSIGN(qid); 1732 break; 1733 } 1734 case 1: { 1735 struct pf_altq_v1 *imported_q = 1736 &((struct pfioc_altq_v1 *)pa)->altq; 1737 1738 COPY(ifname); 1739 1740 ASSIGN(scheduler); 1741 ASSIGN(tbrsize); 1742 ASSIGN(ifbandwidth); 1743 1744 COPY(qname); 1745 COPY(parent); 1746 ASSIGN(parent_qid); 1747 ASSIGN(bandwidth); 1748 ASSIGN(priority); 1749 ASSIGN(local_flags); 1750 1751 ASSIGN(qlimit); 1752 ASSIGN(flags); 1753 COPY(pq_u); 1754 1755 ASSIGN(qid); 1756 break; 1757 } 1758 default: 1759 panic("%s: unhandled struct pfioc_altq version", __func__); 1760 break; 1761 } 1762 1763 #undef ASSIGN 1764 #undef COPY 1765 1766 return (0); 1767 } 1768 1769 static struct pf_altq * 1770 pf_altq_get_nth_active(u_int32_t n) 1771 { 1772 struct pf_altq *altq; 1773 u_int32_t nr; 1774 1775 nr = 0; 1776 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1777 if (nr == n) 1778 return (altq); 1779 nr++; 1780 } 1781 1782 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { 1783 if (nr == n) 1784 return (altq); 1785 nr++; 1786 } 1787 1788 return (NULL); 1789 } 1790 #endif /* ALTQ */ 1791 1792 struct pf_krule * 1793 pf_krule_alloc(void) 1794 { 1795 struct pf_krule *rule; 1796 1797 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO); 1798 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF); 1799 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF); 1800 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF); 1801 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 1802 M_WAITOK | M_ZERO); 1803 return (rule); 1804 } 1805 1806 void 1807 pf_krule_free(struct pf_krule *rule) 1808 { 1809 #ifdef PF_WANT_32_TO_64_COUNTER 1810 bool wowned; 1811 #endif 1812 1813 if (rule == NULL) 1814 return; 1815 1816 #ifdef PF_WANT_32_TO_64_COUNTER 1817 if (rule->allrulelinked) { 1818 wowned = PF_RULES_WOWNED(); 1819 if (!wowned) 1820 PF_RULES_WLOCK(); 1821 LIST_REMOVE(rule, allrulelist); 1822 V_pf_allrulecount--; 1823 if (!wowned) 1824 PF_RULES_WUNLOCK(); 1825 } 1826 #endif 1827 1828 pf_counter_u64_deinit(&rule->evaluations); 1829 for (int i = 0; i < 2; i++) { 1830 pf_counter_u64_deinit(&rule->packets[i]); 1831 pf_counter_u64_deinit(&rule->bytes[i]); 1832 } 1833 counter_u64_free(rule->states_cur); 1834 counter_u64_free(rule->states_tot); 1835 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 1836 counter_u64_free(rule->src_nodes[sn_type]); 1837 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 1838 1839 mtx_destroy(&rule->nat.mtx); 1840 mtx_destroy(&rule->rdr.mtx); 1841 mtx_destroy(&rule->route.mtx); 1842 free(rule, M_PFRULE); 1843 } 1844 1845 void 1846 pf_krule_clear_counters(struct pf_krule *rule) 1847 { 1848 pf_counter_u64_zero(&rule->evaluations); 1849 for (int i = 0; i < 2; i++) { 1850 pf_counter_u64_zero(&rule->packets[i]); 1851 pf_counter_u64_zero(&rule->bytes[i]); 1852 } 1853 counter_u64_zero(rule->states_tot); 1854 } 1855 1856 static void 1857 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool, 1858 struct pf_pooladdr *pool) 1859 { 1860 1861 bzero(pool, sizeof(*pool)); 1862 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr)); 1863 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname)); 1864 } 1865 1866 static int 1867 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool, 1868 struct pf_kpooladdr *kpool) 1869 { 1870 int ret; 1871 1872 bzero(kpool, sizeof(*kpool)); 1873 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr)); 1874 ret = pf_user_strcpy(kpool->ifname, pool->ifname, 1875 sizeof(kpool->ifname)); 1876 return (ret); 1877 } 1878 1879 static void 1880 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool) 1881 { 1882 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), ""); 1883 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), ""); 1884 1885 bcopy(&pool->key, &kpool->key, sizeof(kpool->key)); 1886 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter)); 1887 1888 kpool->tblidx = pool->tblidx; 1889 kpool->proxy_port[0] = pool->proxy_port[0]; 1890 kpool->proxy_port[1] = pool->proxy_port[1]; 1891 kpool->opts = pool->opts; 1892 } 1893 1894 static int 1895 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule) 1896 { 1897 int ret; 1898 1899 #ifndef INET 1900 if (rule->af == AF_INET) { 1901 return (EAFNOSUPPORT); 1902 } 1903 #endif /* INET */ 1904 #ifndef INET6 1905 if (rule->af == AF_INET6) { 1906 return (EAFNOSUPPORT); 1907 } 1908 #endif /* INET6 */ 1909 1910 ret = pf_check_rule_addr(&rule->src); 1911 if (ret != 0) 1912 return (ret); 1913 ret = pf_check_rule_addr(&rule->dst); 1914 if (ret != 0) 1915 return (ret); 1916 1917 bcopy(&rule->src, &krule->src, sizeof(rule->src)); 1918 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst)); 1919 1920 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label)); 1921 if (ret != 0) 1922 return (ret); 1923 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname)); 1924 if (ret != 0) 1925 return (ret); 1926 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname)); 1927 if (ret != 0) 1928 return (ret); 1929 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname)); 1930 if (ret != 0) 1931 return (ret); 1932 ret = pf_user_strcpy(krule->tagname, rule->tagname, 1933 sizeof(rule->tagname)); 1934 if (ret != 0) 1935 return (ret); 1936 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname, 1937 sizeof(rule->match_tagname)); 1938 if (ret != 0) 1939 return (ret); 1940 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname, 1941 sizeof(rule->overload_tblname)); 1942 if (ret != 0) 1943 return (ret); 1944 1945 pf_pool_to_kpool(&rule->rpool, &krule->rdr); 1946 1947 /* Don't allow userspace to set evaluations, packets or bytes. */ 1948 /* kif, anchor, overload_tbl are not copied over. */ 1949 1950 krule->os_fingerprint = rule->os_fingerprint; 1951 1952 krule->rtableid = rule->rtableid; 1953 /* pf_rule->timeout is smaller than pf_krule->timeout */ 1954 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout)); 1955 krule->max_states = rule->max_states; 1956 krule->max_src_nodes = rule->max_src_nodes; 1957 krule->max_src_states = rule->max_src_states; 1958 krule->max_src_conn = rule->max_src_conn; 1959 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit; 1960 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds; 1961 krule->qid = rule->qid; 1962 krule->pqid = rule->pqid; 1963 krule->nr = rule->nr; 1964 krule->prob = rule->prob; 1965 krule->cuid = rule->cuid; 1966 krule->cpid = rule->cpid; 1967 1968 krule->return_icmp = rule->return_icmp; 1969 krule->return_icmp6 = rule->return_icmp6; 1970 krule->max_mss = rule->max_mss; 1971 krule->tag = rule->tag; 1972 krule->match_tag = rule->match_tag; 1973 krule->scrub_flags = rule->scrub_flags; 1974 1975 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid)); 1976 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid)); 1977 1978 krule->rule_flag = rule->rule_flag; 1979 krule->action = rule->action; 1980 krule->direction = rule->direction; 1981 krule->log = rule->log; 1982 krule->logif = rule->logif; 1983 krule->quick = rule->quick; 1984 krule->ifnot = rule->ifnot; 1985 krule->match_tag_not = rule->match_tag_not; 1986 krule->natpass = rule->natpass; 1987 1988 krule->keep_state = rule->keep_state; 1989 krule->af = rule->af; 1990 krule->proto = rule->proto; 1991 krule->type = rule->type; 1992 krule->code = rule->code; 1993 krule->flags = rule->flags; 1994 krule->flagset = rule->flagset; 1995 krule->min_ttl = rule->min_ttl; 1996 krule->allow_opts = rule->allow_opts; 1997 krule->rt = rule->rt; 1998 krule->return_ttl = rule->return_ttl; 1999 krule->tos = rule->tos; 2000 krule->set_tos = rule->set_tos; 2001 2002 krule->flush = rule->flush; 2003 krule->prio = rule->prio; 2004 krule->set_prio[0] = rule->set_prio[0]; 2005 krule->set_prio[1] = rule->set_prio[1]; 2006 2007 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert)); 2008 2009 return (0); 2010 } 2011 2012 int 2013 pf_ioctl_getrules(struct pfioc_rule *pr) 2014 { 2015 struct pf_kruleset *ruleset; 2016 struct pf_krule *tail; 2017 int rs_num; 2018 2019 PF_RULES_WLOCK(); 2020 ruleset = pf_find_kruleset(pr->anchor); 2021 if (ruleset == NULL) { 2022 PF_RULES_WUNLOCK(); 2023 return (EINVAL); 2024 } 2025 rs_num = pf_get_ruleset_number(pr->rule.action); 2026 if (rs_num >= PF_RULESET_MAX) { 2027 PF_RULES_WUNLOCK(); 2028 return (EINVAL); 2029 } 2030 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 2031 pf_krulequeue); 2032 if (tail) 2033 pr->nr = tail->nr + 1; 2034 else 2035 pr->nr = 0; 2036 pr->ticket = ruleset->rules[rs_num].active.ticket; 2037 PF_RULES_WUNLOCK(); 2038 2039 return (0); 2040 } 2041 2042 static int 2043 pf_rule_checkaf(struct pf_krule *r) 2044 { 2045 switch (r->af) { 2046 case 0: 2047 if (r->rule_flag & PFRULE_AFTO) 2048 return (EPFNOSUPPORT); 2049 break; 2050 case AF_INET: 2051 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6) 2052 return (EPFNOSUPPORT); 2053 break; 2054 #ifdef INET6 2055 case AF_INET6: 2056 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET) 2057 return (EPFNOSUPPORT); 2058 break; 2059 #endif /* INET6 */ 2060 default: 2061 return (EPFNOSUPPORT); 2062 } 2063 2064 if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0) 2065 return (EPFNOSUPPORT); 2066 2067 return (0); 2068 } 2069 2070 static int 2071 pf_validate_range(uint8_t op, uint16_t port[2]) 2072 { 2073 uint16_t a = ntohs(port[0]); 2074 uint16_t b = ntohs(port[1]); 2075 2076 if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */ 2077 (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */ 2078 (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */ 2079 return 1; 2080 return 0; 2081 } 2082 2083 int 2084 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, 2085 uint32_t pool_ticket, const char *anchor, const char *anchor_call, 2086 uid_t uid, pid_t pid) 2087 { 2088 struct pf_kruleset *ruleset; 2089 struct pf_krule *tail; 2090 struct pf_kpooladdr *pa; 2091 struct pfi_kkif *kif = NULL, *rcv_kif = NULL; 2092 int rs_num; 2093 int error = 0; 2094 2095 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 2096 #define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x) 2097 2098 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) 2099 ERROUT_UNLOCKED(EINVAL); 2100 2101 if ((error = pf_rule_checkaf(rule))) 2102 ERROUT_UNLOCKED(error); 2103 if (pf_validate_range(rule->src.port_op, rule->src.port)) 2104 ERROUT_UNLOCKED(EINVAL); 2105 if (pf_validate_range(rule->dst.port_op, rule->dst.port)) 2106 ERROUT_UNLOCKED(EINVAL); 2107 2108 if (rule->ifname[0]) 2109 kif = pf_kkif_create(M_WAITOK); 2110 if (rule->rcv_ifname[0]) 2111 rcv_kif = pf_kkif_create(M_WAITOK); 2112 pf_counter_u64_init(&rule->evaluations, M_WAITOK); 2113 for (int i = 0; i < 2; i++) { 2114 pf_counter_u64_init(&rule->packets[i], M_WAITOK); 2115 pf_counter_u64_init(&rule->bytes[i], M_WAITOK); 2116 } 2117 rule->states_cur = counter_u64_alloc(M_WAITOK); 2118 rule->states_tot = counter_u64_alloc(M_WAITOK); 2119 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 2120 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 2121 rule->cuid = uid; 2122 rule->cpid = pid; 2123 TAILQ_INIT(&rule->rdr.list); 2124 TAILQ_INIT(&rule->nat.list); 2125 TAILQ_INIT(&rule->route.list); 2126 2127 PF_CONFIG_LOCK(); 2128 PF_RULES_WLOCK(); 2129 #ifdef PF_WANT_32_TO_64_COUNTER 2130 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist); 2131 MPASS(!rule->allrulelinked); 2132 rule->allrulelinked = true; 2133 V_pf_allrulecount++; 2134 #endif 2135 ruleset = pf_find_kruleset(anchor); 2136 if (ruleset == NULL) 2137 ERROUT(EINVAL); 2138 rs_num = pf_get_ruleset_number(rule->action); 2139 if (rs_num >= PF_RULESET_MAX) 2140 ERROUT(EINVAL); 2141 if (ticket != ruleset->rules[rs_num].inactive.ticket) { 2142 DPFPRINTF(PF_DEBUG_MISC, 2143 "ticket: %d != [%d]%d", ticket, rs_num, 2144 ruleset->rules[rs_num].inactive.ticket); 2145 ERROUT(EBUSY); 2146 } 2147 if (pool_ticket != V_ticket_pabuf) { 2148 DPFPRINTF(PF_DEBUG_MISC, 2149 "pool_ticket: %d != %d", pool_ticket, 2150 V_ticket_pabuf); 2151 ERROUT(EBUSY); 2152 } 2153 /* 2154 * XXXMJG hack: there is no mechanism to ensure they started the 2155 * transaction. Ticket checked above may happen to match by accident, 2156 * even if nobody called DIOCXBEGIN, let alone this process. 2157 * Partially work around it by checking if the RB tree got allocated, 2158 * see pf_begin_rules. 2159 */ 2160 if (ruleset->rules[rs_num].inactive.tree == NULL) { 2161 ERROUT(EINVAL); 2162 } 2163 2164 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 2165 pf_krulequeue); 2166 if (tail) 2167 rule->nr = tail->nr + 1; 2168 else 2169 rule->nr = 0; 2170 if (rule->ifname[0]) { 2171 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2172 kif = NULL; 2173 pfi_kkif_ref(rule->kif); 2174 } else 2175 rule->kif = NULL; 2176 2177 if (rule->rcv_ifname[0]) { 2178 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname); 2179 rcv_kif = NULL; 2180 pfi_kkif_ref(rule->rcv_kif); 2181 } else 2182 rule->rcv_kif = NULL; 2183 2184 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) 2185 ERROUT(EBUSY); 2186 #ifdef ALTQ 2187 /* set queue IDs */ 2188 if (rule->qname[0] != 0) { 2189 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2190 ERROUT(EBUSY); 2191 else if (rule->pqname[0] != 0) { 2192 if ((rule->pqid = 2193 pf_qname2qid(rule->pqname)) == 0) 2194 ERROUT(EBUSY); 2195 } else 2196 rule->pqid = rule->qid; 2197 } 2198 #endif 2199 if (rule->tagname[0]) 2200 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2201 ERROUT(EBUSY); 2202 if (rule->match_tagname[0]) 2203 if ((rule->match_tag = 2204 pf_tagname2tag(rule->match_tagname)) == 0) 2205 ERROUT(EBUSY); 2206 if (rule->rt && !rule->direction) 2207 ERROUT(EINVAL); 2208 if (!rule->log) 2209 rule->logif = 0; 2210 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit, 2211 rule->pktrate.seconds)) 2212 ERROUT(ENOMEM); 2213 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) 2214 ERROUT(ENOMEM); 2215 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) 2216 ERROUT(ENOMEM); 2217 if (pf_kanchor_setup(rule, ruleset, anchor_call)) 2218 ERROUT(EINVAL); 2219 if (rule->scrub_flags & PFSTATE_SETPRIO && 2220 (rule->set_prio[0] > PF_PRIO_MAX || 2221 rule->set_prio[1] > PF_PRIO_MAX)) 2222 ERROUT(EINVAL); 2223 for (int i = 0; i < 3; i++) { 2224 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 2225 if (pa->addr.type == PF_ADDR_TABLE) { 2226 pa->addr.p.tbl = pfr_attach_table(ruleset, 2227 pa->addr.v.tblname); 2228 if (pa->addr.p.tbl == NULL) 2229 ERROUT(ENOMEM); 2230 } 2231 } 2232 2233 rule->overload_tbl = NULL; 2234 if (rule->overload_tblname[0]) { 2235 if ((rule->overload_tbl = pfr_attach_table(ruleset, 2236 rule->overload_tblname)) == NULL) 2237 ERROUT(EINVAL); 2238 else 2239 rule->overload_tbl->pfrkt_flags |= 2240 PFR_TFLAG_ACTIVE; 2241 } 2242 2243 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list); 2244 2245 /* 2246 * Old version of pfctl provide route redirection pools in single 2247 * common redirection pool rdr. New versions use rdr only for 2248 * rdr-to rules. 2249 */ 2250 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) { 2251 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list); 2252 } else { 2253 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list); 2254 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list); 2255 } 2256 2257 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) || 2258 (rule->action == PF_BINAT)) && rule->anchor == NULL && 2259 TAILQ_FIRST(&rule->rdr.list) == NULL) { 2260 ERROUT(EINVAL); 2261 } 2262 2263 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) { 2264 ERROUT(EINVAL); 2265 } 2266 2267 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR || 2268 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) { 2269 ERROUT(EINVAL); 2270 } 2271 2272 MPASS(error == 0); 2273 2274 rule->nat.cur = TAILQ_FIRST(&rule->nat.list); 2275 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list); 2276 rule->route.cur = TAILQ_FIRST(&rule->route.list); 2277 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, 2278 rule, entries); 2279 ruleset->rules[rs_num].inactive.rcount++; 2280 2281 PF_RULES_WUNLOCK(); 2282 pf_hash_rule(rule); 2283 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) { 2284 PF_RULES_WLOCK(); 2285 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries); 2286 ruleset->rules[rs_num].inactive.rcount--; 2287 pf_free_rule(rule); 2288 rule = NULL; 2289 ERROUT(EEXIST); 2290 } 2291 PF_CONFIG_UNLOCK(); 2292 2293 return (0); 2294 2295 #undef ERROUT 2296 #undef ERROUT_UNLOCKED 2297 errout: 2298 PF_RULES_WUNLOCK(); 2299 PF_CONFIG_UNLOCK(); 2300 errout_unlocked: 2301 pf_kkif_free(rcv_kif); 2302 pf_kkif_free(kif); 2303 pf_krule_free(rule); 2304 return (error); 2305 } 2306 2307 static bool 2308 pf_label_match(const struct pf_krule *rule, const char *label) 2309 { 2310 int i = 0; 2311 2312 while (*rule->label[i]) { 2313 if (strcmp(rule->label[i], label) == 0) 2314 return (true); 2315 i++; 2316 } 2317 2318 return (false); 2319 } 2320 2321 static unsigned int 2322 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir) 2323 { 2324 struct pf_kstate *s; 2325 int more = 0; 2326 2327 s = pf_find_state_all(key, dir, &more); 2328 if (s == NULL) 2329 return (0); 2330 2331 if (more) { 2332 PF_STATE_UNLOCK(s); 2333 return (0); 2334 } 2335 2336 pf_remove_state(s); 2337 return (1); 2338 } 2339 2340 static int 2341 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih) 2342 { 2343 struct pf_kstate *s; 2344 struct pf_state_key *sk; 2345 struct pf_addr *srcaddr, *dstaddr; 2346 struct pf_state_key_cmp match_key; 2347 int idx, killed = 0; 2348 unsigned int dir; 2349 u_int16_t srcport, dstport; 2350 struct pfi_kkif *kif; 2351 2352 relock_DIOCKILLSTATES: 2353 PF_HASHROW_LOCK(ih); 2354 LIST_FOREACH(s, &ih->states, entry) { 2355 /* For floating states look at the original kif. */ 2356 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 2357 2358 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE]; 2359 if (s->direction == PF_OUT) { 2360 srcaddr = &sk->addr[1]; 2361 dstaddr = &sk->addr[0]; 2362 srcport = sk->port[1]; 2363 dstport = sk->port[0]; 2364 } else { 2365 srcaddr = &sk->addr[0]; 2366 dstaddr = &sk->addr[1]; 2367 srcport = sk->port[0]; 2368 dstport = sk->port[1]; 2369 } 2370 2371 if (psk->psk_af && sk->af != psk->psk_af) 2372 continue; 2373 2374 if (psk->psk_proto && psk->psk_proto != sk->proto) 2375 continue; 2376 2377 if (! pf_match_addr(psk->psk_src.neg, 2378 &psk->psk_src.addr.v.a.addr, 2379 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af)) 2380 continue; 2381 2382 if (! pf_match_addr(psk->psk_dst.neg, 2383 &psk->psk_dst.addr.v.a.addr, 2384 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af)) 2385 continue; 2386 2387 if (! pf_match_addr(psk->psk_rt_addr.neg, 2388 &psk->psk_rt_addr.addr.v.a.addr, 2389 &psk->psk_rt_addr.addr.v.a.mask, 2390 &s->act.rt_addr, sk->af)) 2391 continue; 2392 2393 if (psk->psk_src.port_op != 0 && 2394 ! pf_match_port(psk->psk_src.port_op, 2395 psk->psk_src.port[0], psk->psk_src.port[1], srcport)) 2396 continue; 2397 2398 if (psk->psk_dst.port_op != 0 && 2399 ! pf_match_port(psk->psk_dst.port_op, 2400 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) 2401 continue; 2402 2403 if (psk->psk_label[0] && 2404 ! pf_label_match(s->rule, psk->psk_label)) 2405 continue; 2406 2407 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname, 2408 kif->pfik_name)) 2409 continue; 2410 2411 if (psk->psk_kill_match) { 2412 /* Create the key to find matching states, with lock 2413 * held. */ 2414 2415 bzero(&match_key, sizeof(match_key)); 2416 2417 if (s->direction == PF_OUT) { 2418 dir = PF_IN; 2419 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK; 2420 } else { 2421 dir = PF_OUT; 2422 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE; 2423 } 2424 2425 match_key.af = s->key[idx]->af; 2426 match_key.proto = s->key[idx]->proto; 2427 pf_addrcpy(&match_key.addr[0], 2428 &s->key[idx]->addr[1], match_key.af); 2429 match_key.port[0] = s->key[idx]->port[1]; 2430 pf_addrcpy(&match_key.addr[1], 2431 &s->key[idx]->addr[0], match_key.af); 2432 match_key.port[1] = s->key[idx]->port[0]; 2433 } 2434 2435 pf_remove_state(s); 2436 killed++; 2437 2438 if (psk->psk_kill_match) 2439 killed += pf_kill_matching_state(&match_key, dir); 2440 2441 goto relock_DIOCKILLSTATES; 2442 } 2443 PF_HASHROW_UNLOCK(ih); 2444 2445 return (killed); 2446 } 2447 2448 void 2449 unhandled_af(int af) 2450 { 2451 panic("unhandled af %d", af); 2452 } 2453 2454 int 2455 pf_start(void) 2456 { 2457 int error = 0; 2458 2459 sx_xlock(&V_pf_ioctl_lock); 2460 if (V_pf_status.running) 2461 error = EEXIST; 2462 else { 2463 hook_pf(); 2464 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 2465 hook_pf_eth(); 2466 V_pf_status.running = 1; 2467 V_pf_status.since = time_uptime; 2468 new_unrhdr64(&V_pf_stateid, time_second); 2469 2470 DPFPRINTF(PF_DEBUG_MISC, "pf: started"); 2471 } 2472 sx_xunlock(&V_pf_ioctl_lock); 2473 2474 return (error); 2475 } 2476 2477 int 2478 pf_stop(void) 2479 { 2480 int error = 0; 2481 2482 sx_xlock(&V_pf_ioctl_lock); 2483 if (!V_pf_status.running) 2484 error = ENOENT; 2485 else { 2486 V_pf_status.running = 0; 2487 dehook_pf(); 2488 dehook_pf_eth(); 2489 V_pf_status.since = time_uptime; 2490 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped"); 2491 } 2492 sx_xunlock(&V_pf_ioctl_lock); 2493 2494 return (error); 2495 } 2496 2497 void 2498 pf_ioctl_clear_status(void) 2499 { 2500 PF_RULES_WLOCK(); 2501 for (int i = 0; i < PFRES_MAX; i++) 2502 counter_u64_zero(V_pf_status.counters[i]); 2503 for (int i = 0; i < FCNT_MAX; i++) 2504 pf_counter_u64_zero(&V_pf_status.fcounters[i]); 2505 for (int i = 0; i < SCNT_MAX; i++) 2506 counter_u64_zero(V_pf_status.scounters[i]); 2507 for (int i = 0; i < KLCNT_MAX; i++) 2508 counter_u64_zero(V_pf_status.lcounters[i]); 2509 V_pf_status.since = time_uptime; 2510 if (*V_pf_status.ifname) 2511 pfi_update_status(V_pf_status.ifname, NULL); 2512 PF_RULES_WUNLOCK(); 2513 } 2514 2515 int 2516 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds) 2517 { 2518 uint32_t old; 2519 2520 if (timeout < 0 || timeout >= PFTM_MAX || 2521 seconds < 0) 2522 return (EINVAL); 2523 2524 PF_RULES_WLOCK(); 2525 old = V_pf_default_rule.timeout[timeout]; 2526 if (timeout == PFTM_INTERVAL && seconds == 0) 2527 seconds = 1; 2528 V_pf_default_rule.timeout[timeout] = seconds; 2529 if (timeout == PFTM_INTERVAL && seconds < old) 2530 wakeup(pf_purge_thread); 2531 2532 if (prev_seconds != NULL) 2533 *prev_seconds = old; 2534 2535 PF_RULES_WUNLOCK(); 2536 2537 return (0); 2538 } 2539 2540 int 2541 pf_ioctl_get_timeout(int timeout, int *seconds) 2542 { 2543 PF_RULES_RLOCK_TRACKER; 2544 2545 if (timeout < 0 || timeout >= PFTM_MAX) 2546 return (EINVAL); 2547 2548 PF_RULES_RLOCK(); 2549 *seconds = V_pf_default_rule.timeout[timeout]; 2550 PF_RULES_RUNLOCK(); 2551 2552 return (0); 2553 } 2554 2555 int 2556 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit) 2557 { 2558 2559 PF_RULES_WLOCK(); 2560 if (index < 0 || index >= PF_LIMIT_MAX || 2561 V_pf_limits[index].zone == NULL) { 2562 PF_RULES_WUNLOCK(); 2563 return (EINVAL); 2564 } 2565 uma_zone_set_max(V_pf_limits[index].zone, 2566 limit == 0 ? INT_MAX : limit); 2567 if (old_limit != NULL) 2568 *old_limit = V_pf_limits[index].limit; 2569 V_pf_limits[index].limit = limit; 2570 PF_RULES_WUNLOCK(); 2571 2572 return (0); 2573 } 2574 2575 int 2576 pf_ioctl_get_limit(int index, unsigned int *limit) 2577 { 2578 PF_RULES_RLOCK_TRACKER; 2579 2580 if (index < 0 || index >= PF_LIMIT_MAX) 2581 return (EINVAL); 2582 2583 PF_RULES_RLOCK(); 2584 *limit = V_pf_limits[index].limit; 2585 PF_RULES_RUNLOCK(); 2586 2587 return (0); 2588 } 2589 2590 int 2591 pf_ioctl_begin_addrs(uint32_t *ticket) 2592 { 2593 PF_RULES_WLOCK(); 2594 pf_empty_kpool(&V_pf_pabuf[0]); 2595 pf_empty_kpool(&V_pf_pabuf[1]); 2596 pf_empty_kpool(&V_pf_pabuf[2]); 2597 *ticket = ++V_ticket_pabuf; 2598 PF_RULES_WUNLOCK(); 2599 2600 return (0); 2601 } 2602 2603 int 2604 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp) 2605 { 2606 struct pf_kpooladdr *pa = NULL; 2607 struct pfi_kkif *kif = NULL; 2608 int error; 2609 2610 if (pp->which != PF_RDR && pp->which != PF_NAT && 2611 pp->which != PF_RT) 2612 return (EINVAL); 2613 2614 switch (pp->af) { 2615 #ifdef INET 2616 case AF_INET: 2617 /* FALLTHROUGH */ 2618 #endif /* INET */ 2619 #ifdef INET6 2620 case AF_INET6: 2621 /* FALLTHROUGH */ 2622 #endif /* INET6 */ 2623 case AF_UNSPEC: 2624 break; 2625 default: 2626 return (EAFNOSUPPORT); 2627 } 2628 2629 if (pp->addr.addr.type != PF_ADDR_ADDRMASK && 2630 pp->addr.addr.type != PF_ADDR_DYNIFTL && 2631 pp->addr.addr.type != PF_ADDR_TABLE) 2632 return (EINVAL); 2633 2634 if (pp->addr.addr.p.dyn != NULL) 2635 return (EINVAL); 2636 2637 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); 2638 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa); 2639 if (error != 0) 2640 goto out; 2641 if (pa->ifname[0]) 2642 kif = pf_kkif_create(M_WAITOK); 2643 PF_RULES_WLOCK(); 2644 if (pp->ticket != V_ticket_pabuf) { 2645 PF_RULES_WUNLOCK(); 2646 if (pa->ifname[0]) 2647 pf_kkif_free(kif); 2648 error = EBUSY; 2649 goto out; 2650 } 2651 if (pa->ifname[0]) { 2652 pa->kif = pfi_kkif_attach(kif, pa->ifname); 2653 kif = NULL; 2654 pfi_kkif_ref(pa->kif); 2655 } else 2656 pa->kif = NULL; 2657 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = 2658 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { 2659 if (pa->ifname[0]) 2660 pfi_kkif_unref(pa->kif); 2661 PF_RULES_WUNLOCK(); 2662 goto out; 2663 } 2664 switch (pp->which) { 2665 case PF_NAT: 2666 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries); 2667 break; 2668 case PF_RDR: 2669 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries); 2670 break; 2671 case PF_RT: 2672 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries); 2673 break; 2674 } 2675 PF_RULES_WUNLOCK(); 2676 2677 return (0); 2678 2679 out: 2680 free(pa, M_PFRULE); 2681 return (error); 2682 } 2683 2684 int 2685 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp) 2686 { 2687 struct pf_kpool *pool; 2688 struct pf_kpooladdr *pa; 2689 2690 PF_RULES_RLOCK_TRACKER; 2691 2692 if (pp->which != PF_RDR && pp->which != PF_NAT && 2693 pp->which != PF_RT) 2694 return (EINVAL); 2695 2696 pp->anchor[sizeof(pp->anchor) - 1] = 0; 2697 pp->nr = 0; 2698 2699 PF_RULES_RLOCK(); 2700 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2701 pp->r_num, 0, 1, 0, pp->which); 2702 if (pool == NULL) { 2703 PF_RULES_RUNLOCK(); 2704 return (EBUSY); 2705 } 2706 TAILQ_FOREACH(pa, &pool->list, entries) 2707 pp->nr++; 2708 PF_RULES_RUNLOCK(); 2709 2710 return (0); 2711 } 2712 2713 int 2714 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp) 2715 { 2716 struct pf_kpool *pool; 2717 struct pf_kpooladdr *pa; 2718 u_int32_t nr = 0; 2719 2720 if (pp->which != PF_RDR && pp->which != PF_NAT && 2721 pp->which != PF_RT) 2722 return (EINVAL); 2723 2724 PF_RULES_RLOCK_TRACKER; 2725 2726 pp->anchor[sizeof(pp->anchor) - 1] = '\0'; 2727 2728 PF_RULES_RLOCK(); 2729 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2730 pp->r_num, 0, 1, 1, pp->which); 2731 if (pool == NULL) { 2732 PF_RULES_RUNLOCK(); 2733 return (EBUSY); 2734 } 2735 pa = TAILQ_FIRST(&pool->list); 2736 while ((pa != NULL) && (nr < pp->nr)) { 2737 pa = TAILQ_NEXT(pa, entries); 2738 nr++; 2739 } 2740 if (pa == NULL) { 2741 PF_RULES_RUNLOCK(); 2742 return (EBUSY); 2743 } 2744 pf_kpooladdr_to_pooladdr(pa, &pp->addr); 2745 pf_addr_copyout(&pp->addr.addr); 2746 PF_RULES_RUNLOCK(); 2747 2748 return (0); 2749 } 2750 2751 int 2752 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr) 2753 { 2754 struct pf_kruleset *ruleset; 2755 struct pf_kanchor *anchor; 2756 2757 PF_RULES_RLOCK_TRACKER; 2758 2759 pr->path[sizeof(pr->path) - 1] = '\0'; 2760 2761 PF_RULES_RLOCK(); 2762 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2763 PF_RULES_RUNLOCK(); 2764 return (ENOENT); 2765 } 2766 pr->nr = 0; 2767 if (ruleset == &pf_main_ruleset) { 2768 /* XXX kludge for pf_main_ruleset */ 2769 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2770 if (anchor->parent == NULL) 2771 pr->nr++; 2772 } else { 2773 RB_FOREACH(anchor, pf_kanchor_node, 2774 &ruleset->anchor->children) 2775 pr->nr++; 2776 } 2777 PF_RULES_RUNLOCK(); 2778 2779 return (0); 2780 } 2781 2782 int 2783 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr) 2784 { 2785 struct pf_kruleset *ruleset; 2786 struct pf_kanchor *anchor; 2787 u_int32_t nr = 0; 2788 int error = 0; 2789 2790 PF_RULES_RLOCK_TRACKER; 2791 2792 PF_RULES_RLOCK(); 2793 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2794 PF_RULES_RUNLOCK(); 2795 return (ENOENT); 2796 } 2797 2798 pr->name[0] = '\0'; 2799 if (ruleset == &pf_main_ruleset) { 2800 /* XXX kludge for pf_main_ruleset */ 2801 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2802 if (anchor->parent == NULL && nr++ == pr->nr) { 2803 strlcpy(pr->name, anchor->name, 2804 sizeof(pr->name)); 2805 break; 2806 } 2807 } else { 2808 RB_FOREACH(anchor, pf_kanchor_node, 2809 &ruleset->anchor->children) 2810 if (nr++ == pr->nr) { 2811 strlcpy(pr->name, anchor->name, 2812 sizeof(pr->name)); 2813 break; 2814 } 2815 } 2816 if (!pr->name[0]) 2817 error = EBUSY; 2818 PF_RULES_RUNLOCK(); 2819 2820 return (error); 2821 } 2822 2823 int 2824 pf_ioctl_natlook(struct pfioc_natlook *pnl) 2825 { 2826 struct pf_state_key *sk; 2827 struct pf_kstate *state; 2828 struct pf_state_key_cmp key; 2829 int m = 0, direction = pnl->direction; 2830 int sidx, didx; 2831 2832 /* NATLOOK src and dst are reversed, so reverse sidx/didx */ 2833 sidx = (direction == PF_IN) ? 1 : 0; 2834 didx = (direction == PF_IN) ? 0 : 1; 2835 2836 if (!pnl->proto || 2837 PF_AZERO(&pnl->saddr, pnl->af) || 2838 PF_AZERO(&pnl->daddr, pnl->af) || 2839 ((pnl->proto == IPPROTO_TCP || 2840 pnl->proto == IPPROTO_UDP) && 2841 (!pnl->dport || !pnl->sport))) 2842 return (EINVAL); 2843 2844 switch (pnl->direction) { 2845 case PF_IN: 2846 case PF_OUT: 2847 case PF_INOUT: 2848 break; 2849 default: 2850 return (EINVAL); 2851 } 2852 2853 switch (pnl->af) { 2854 #ifdef INET 2855 case AF_INET: 2856 break; 2857 #endif /* INET */ 2858 #ifdef INET6 2859 case AF_INET6: 2860 break; 2861 #endif /* INET6 */ 2862 default: 2863 return (EAFNOSUPPORT); 2864 } 2865 2866 bzero(&key, sizeof(key)); 2867 key.af = pnl->af; 2868 key.proto = pnl->proto; 2869 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af); 2870 key.port[sidx] = pnl->sport; 2871 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af); 2872 key.port[didx] = pnl->dport; 2873 2874 state = pf_find_state_all(&key, direction, &m); 2875 if (state == NULL) 2876 return (ENOENT); 2877 2878 if (m > 1) { 2879 PF_STATE_UNLOCK(state); 2880 return (E2BIG); /* more than one state */ 2881 } 2882 2883 sk = state->key[sidx]; 2884 pf_addrcpy(&pnl->rsaddr, 2885 &sk->addr[sidx], sk->af); 2886 pnl->rsport = sk->port[sidx]; 2887 pf_addrcpy(&pnl->rdaddr, 2888 &sk->addr[didx], sk->af); 2889 pnl->rdport = sk->port[didx]; 2890 PF_STATE_UNLOCK(state); 2891 2892 return (0); 2893 } 2894 2895 static int 2896 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) 2897 { 2898 int error = 0; 2899 PF_RULES_RLOCK_TRACKER; 2900 2901 #define ERROUT_IOCTL(target, x) \ 2902 do { \ 2903 error = (x); \ 2904 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \ 2905 goto target; \ 2906 } while (0) 2907 2908 2909 /* XXX keep in sync with switch() below */ 2910 if (securelevel_gt(td->td_ucred, 2)) 2911 switch (cmd) { 2912 case DIOCGETRULES: 2913 case DIOCGETRULENV: 2914 case DIOCGETADDRS: 2915 case DIOCGETADDR: 2916 case DIOCGETSTATE: 2917 case DIOCGETSTATENV: 2918 case DIOCSETSTATUSIF: 2919 case DIOCGETSTATUSNV: 2920 case DIOCCLRSTATUS: 2921 case DIOCNATLOOK: 2922 case DIOCSETDEBUG: 2923 #ifdef COMPAT_FREEBSD14 2924 case DIOCGETSTATES: 2925 case DIOCGETSTATESV2: 2926 #endif 2927 case DIOCGETTIMEOUT: 2928 case DIOCCLRRULECTRS: 2929 case DIOCGETLIMIT: 2930 case DIOCGETALTQSV0: 2931 case DIOCGETALTQSV1: 2932 case DIOCGETALTQV0: 2933 case DIOCGETALTQV1: 2934 case DIOCGETQSTATSV0: 2935 case DIOCGETQSTATSV1: 2936 case DIOCGETRULESETS: 2937 case DIOCGETRULESET: 2938 case DIOCRGETTABLES: 2939 case DIOCRGETTSTATS: 2940 case DIOCRCLRTSTATS: 2941 case DIOCRCLRADDRS: 2942 case DIOCRADDADDRS: 2943 case DIOCRDELADDRS: 2944 case DIOCRSETADDRS: 2945 case DIOCRGETADDRS: 2946 case DIOCRGETASTATS: 2947 case DIOCRCLRASTATS: 2948 case DIOCRTSTADDRS: 2949 case DIOCOSFPGET: 2950 case DIOCGETSRCNODES: 2951 case DIOCCLRSRCNODES: 2952 case DIOCGETSYNCOOKIES: 2953 case DIOCIGETIFACES: 2954 case DIOCGIFSPEEDV0: 2955 case DIOCGIFSPEEDV1: 2956 case DIOCSETIFFLAG: 2957 case DIOCCLRIFFLAG: 2958 case DIOCGETETHRULES: 2959 case DIOCGETETHRULE: 2960 case DIOCGETETHRULESETS: 2961 case DIOCGETETHRULESET: 2962 break; 2963 case DIOCRCLRTABLES: 2964 case DIOCRADDTABLES: 2965 case DIOCRDELTABLES: 2966 case DIOCRSETTFLAGS: 2967 if (((struct pfioc_table *)addr)->pfrio_flags & 2968 PFR_FLAG_DUMMY) 2969 break; /* dummy operation ok */ 2970 return (EPERM); 2971 default: 2972 return (EPERM); 2973 } 2974 2975 if (!(flags & FWRITE)) 2976 switch (cmd) { 2977 case DIOCGETRULES: 2978 case DIOCGETADDRS: 2979 case DIOCGETADDR: 2980 case DIOCGETSTATE: 2981 case DIOCGETSTATENV: 2982 case DIOCGETSTATUSNV: 2983 #ifdef COMPAT_FREEBSD14 2984 case DIOCGETSTATES: 2985 case DIOCGETSTATESV2: 2986 #endif 2987 case DIOCGETTIMEOUT: 2988 case DIOCGETLIMIT: 2989 case DIOCGETALTQSV0: 2990 case DIOCGETALTQSV1: 2991 case DIOCGETALTQV0: 2992 case DIOCGETALTQV1: 2993 case DIOCGETQSTATSV0: 2994 case DIOCGETQSTATSV1: 2995 case DIOCGETRULESETS: 2996 case DIOCGETRULESET: 2997 case DIOCNATLOOK: 2998 case DIOCRGETTABLES: 2999 case DIOCRGETTSTATS: 3000 case DIOCRGETADDRS: 3001 case DIOCRGETASTATS: 3002 case DIOCRTSTADDRS: 3003 case DIOCOSFPGET: 3004 case DIOCGETSRCNODES: 3005 case DIOCGETSYNCOOKIES: 3006 case DIOCIGETIFACES: 3007 case DIOCGIFSPEEDV1: 3008 case DIOCGIFSPEEDV0: 3009 case DIOCGETRULENV: 3010 case DIOCGETETHRULES: 3011 case DIOCGETETHRULE: 3012 case DIOCGETETHRULESETS: 3013 case DIOCGETETHRULESET: 3014 break; 3015 case DIOCRCLRTABLES: 3016 case DIOCRADDTABLES: 3017 case DIOCRDELTABLES: 3018 case DIOCRCLRTSTATS: 3019 case DIOCRCLRADDRS: 3020 case DIOCRADDADDRS: 3021 case DIOCRDELADDRS: 3022 case DIOCRSETADDRS: 3023 case DIOCRSETTFLAGS: 3024 if (((struct pfioc_table *)addr)->pfrio_flags & 3025 PFR_FLAG_DUMMY) { 3026 flags |= FWRITE; /* need write lock for dummy */ 3027 break; /* dummy operation ok */ 3028 } 3029 return (EACCES); 3030 default: 3031 return (EACCES); 3032 } 3033 3034 CURVNET_SET(TD_TO_VNET(td)); 3035 3036 switch (cmd) { 3037 #ifdef COMPAT_FREEBSD14 3038 case DIOCSTART: 3039 error = pf_start(); 3040 break; 3041 3042 case DIOCSTOP: 3043 error = pf_stop(); 3044 break; 3045 #endif 3046 3047 case DIOCGETETHRULES: { 3048 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3049 nvlist_t *nvl; 3050 void *packed; 3051 struct pf_keth_rule *tail; 3052 struct pf_keth_ruleset *rs; 3053 u_int32_t ticket, nr; 3054 const char *anchor = ""; 3055 3056 nvl = NULL; 3057 packed = NULL; 3058 3059 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x) 3060 3061 if (nv->len > pf_ioctl_maxcount) 3062 ERROUT(ENOMEM); 3063 3064 /* Copy the request in */ 3065 packed = malloc(nv->len, M_NVLIST, M_WAITOK); 3066 error = copyin(nv->data, packed, nv->len); 3067 if (error) 3068 ERROUT(error); 3069 3070 nvl = nvlist_unpack(packed, nv->len, 0); 3071 if (nvl == NULL) 3072 ERROUT(EBADMSG); 3073 3074 if (! nvlist_exists_string(nvl, "anchor")) 3075 ERROUT(EBADMSG); 3076 3077 anchor = nvlist_get_string(nvl, "anchor"); 3078 3079 rs = pf_find_keth_ruleset(anchor); 3080 3081 nvlist_destroy(nvl); 3082 nvl = NULL; 3083 free(packed, M_NVLIST); 3084 packed = NULL; 3085 3086 if (rs == NULL) 3087 ERROUT(ENOENT); 3088 3089 /* Reply */ 3090 nvl = nvlist_create(0); 3091 if (nvl == NULL) 3092 ERROUT(ENOMEM); 3093 3094 PF_RULES_RLOCK(); 3095 3096 ticket = rs->active.ticket; 3097 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq); 3098 if (tail) 3099 nr = tail->nr + 1; 3100 else 3101 nr = 0; 3102 3103 PF_RULES_RUNLOCK(); 3104 3105 nvlist_add_number(nvl, "ticket", ticket); 3106 nvlist_add_number(nvl, "nr", nr); 3107 3108 packed = nvlist_pack(nvl, &nv->len); 3109 if (packed == NULL) 3110 ERROUT(ENOMEM); 3111 3112 if (nv->size == 0) 3113 ERROUT(0); 3114 else if (nv->size < nv->len) 3115 ERROUT(ENOSPC); 3116 3117 error = copyout(packed, nv->data, nv->len); 3118 3119 #undef ERROUT 3120 DIOCGETETHRULES_error: 3121 free(packed, M_NVLIST); 3122 nvlist_destroy(nvl); 3123 break; 3124 } 3125 3126 case DIOCGETETHRULE: { 3127 struct epoch_tracker et; 3128 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3129 nvlist_t *nvl = NULL; 3130 void *nvlpacked = NULL; 3131 struct pf_keth_rule *rule = NULL; 3132 struct pf_keth_ruleset *rs; 3133 u_int32_t ticket, nr; 3134 bool clear = false; 3135 const char *anchor; 3136 3137 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x) 3138 3139 if (nv->len > pf_ioctl_maxcount) 3140 ERROUT(ENOMEM); 3141 3142 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3143 error = copyin(nv->data, nvlpacked, nv->len); 3144 if (error) 3145 ERROUT(error); 3146 3147 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3148 if (nvl == NULL) 3149 ERROUT(EBADMSG); 3150 if (! nvlist_exists_number(nvl, "ticket")) 3151 ERROUT(EBADMSG); 3152 ticket = nvlist_get_number(nvl, "ticket"); 3153 if (! nvlist_exists_string(nvl, "anchor")) 3154 ERROUT(EBADMSG); 3155 anchor = nvlist_get_string(nvl, "anchor"); 3156 3157 if (nvlist_exists_bool(nvl, "clear")) 3158 clear = nvlist_get_bool(nvl, "clear"); 3159 3160 if (clear && !(flags & FWRITE)) 3161 ERROUT(EACCES); 3162 3163 if (! nvlist_exists_number(nvl, "nr")) 3164 ERROUT(EBADMSG); 3165 nr = nvlist_get_number(nvl, "nr"); 3166 3167 PF_RULES_RLOCK(); 3168 rs = pf_find_keth_ruleset(anchor); 3169 if (rs == NULL) { 3170 PF_RULES_RUNLOCK(); 3171 ERROUT(ENOENT); 3172 } 3173 if (ticket != rs->active.ticket) { 3174 PF_RULES_RUNLOCK(); 3175 ERROUT(EBUSY); 3176 } 3177 3178 nvlist_destroy(nvl); 3179 nvl = NULL; 3180 free(nvlpacked, M_NVLIST); 3181 nvlpacked = NULL; 3182 3183 rule = TAILQ_FIRST(rs->active.rules); 3184 while ((rule != NULL) && (rule->nr != nr)) 3185 rule = TAILQ_NEXT(rule, entries); 3186 if (rule == NULL) { 3187 PF_RULES_RUNLOCK(); 3188 ERROUT(ENOENT); 3189 } 3190 /* Make sure rule can't go away. */ 3191 NET_EPOCH_ENTER(et); 3192 PF_RULES_RUNLOCK(); 3193 nvl = pf_keth_rule_to_nveth_rule(rule); 3194 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) { 3195 NET_EPOCH_EXIT(et); 3196 ERROUT(EBUSY); 3197 } 3198 NET_EPOCH_EXIT(et); 3199 if (nvl == NULL) 3200 ERROUT(ENOMEM); 3201 3202 nvlpacked = nvlist_pack(nvl, &nv->len); 3203 if (nvlpacked == NULL) 3204 ERROUT(ENOMEM); 3205 3206 if (nv->size == 0) 3207 ERROUT(0); 3208 else if (nv->size < nv->len) 3209 ERROUT(ENOSPC); 3210 3211 error = copyout(nvlpacked, nv->data, nv->len); 3212 if (error == 0 && clear) { 3213 counter_u64_zero(rule->evaluations); 3214 for (int i = 0; i < 2; i++) { 3215 counter_u64_zero(rule->packets[i]); 3216 counter_u64_zero(rule->bytes[i]); 3217 } 3218 } 3219 3220 #undef ERROUT 3221 DIOCGETETHRULE_error: 3222 free(nvlpacked, M_NVLIST); 3223 nvlist_destroy(nvl); 3224 break; 3225 } 3226 3227 case DIOCADDETHRULE: { 3228 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3229 nvlist_t *nvl = NULL; 3230 void *nvlpacked = NULL; 3231 struct pf_keth_rule *rule = NULL, *tail = NULL; 3232 struct pf_keth_ruleset *ruleset = NULL; 3233 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL; 3234 const char *anchor = "", *anchor_call = ""; 3235 3236 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x) 3237 3238 if (nv->len > pf_ioctl_maxcount) 3239 ERROUT(ENOMEM); 3240 3241 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3242 error = copyin(nv->data, nvlpacked, nv->len); 3243 if (error) 3244 ERROUT(error); 3245 3246 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3247 if (nvl == NULL) 3248 ERROUT(EBADMSG); 3249 3250 if (! nvlist_exists_number(nvl, "ticket")) 3251 ERROUT(EBADMSG); 3252 3253 if (nvlist_exists_string(nvl, "anchor")) 3254 anchor = nvlist_get_string(nvl, "anchor"); 3255 if (nvlist_exists_string(nvl, "anchor_call")) 3256 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3257 3258 ruleset = pf_find_keth_ruleset(anchor); 3259 if (ruleset == NULL) 3260 ERROUT(EINVAL); 3261 3262 if (nvlist_get_number(nvl, "ticket") != 3263 ruleset->inactive.ticket) { 3264 DPFPRINTF(PF_DEBUG_MISC, 3265 "ticket: %d != %d", 3266 (u_int32_t)nvlist_get_number(nvl, "ticket"), 3267 ruleset->inactive.ticket); 3268 ERROUT(EBUSY); 3269 } 3270 3271 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); 3272 rule->timestamp = NULL; 3273 3274 error = pf_nveth_rule_to_keth_rule(nvl, rule); 3275 if (error != 0) 3276 ERROUT(error); 3277 3278 if (rule->ifname[0]) 3279 kif = pf_kkif_create(M_WAITOK); 3280 if (rule->bridge_to_name[0]) 3281 bridge_to_kif = pf_kkif_create(M_WAITOK); 3282 rule->evaluations = counter_u64_alloc(M_WAITOK); 3283 for (int i = 0; i < 2; i++) { 3284 rule->packets[i] = counter_u64_alloc(M_WAITOK); 3285 rule->bytes[i] = counter_u64_alloc(M_WAITOK); 3286 } 3287 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 3288 M_WAITOK | M_ZERO); 3289 3290 PF_RULES_WLOCK(); 3291 3292 if (rule->ifname[0]) { 3293 rule->kif = pfi_kkif_attach(kif, rule->ifname); 3294 pfi_kkif_ref(rule->kif); 3295 } else 3296 rule->kif = NULL; 3297 if (rule->bridge_to_name[0]) { 3298 rule->bridge_to = pfi_kkif_attach(bridge_to_kif, 3299 rule->bridge_to_name); 3300 pfi_kkif_ref(rule->bridge_to); 3301 } else 3302 rule->bridge_to = NULL; 3303 3304 #ifdef ALTQ 3305 /* set queue IDs */ 3306 if (rule->qname[0] != 0) { 3307 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 3308 error = EBUSY; 3309 else 3310 rule->qid = rule->qid; 3311 } 3312 #endif 3313 if (rule->tagname[0]) 3314 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 3315 error = EBUSY; 3316 if (rule->match_tagname[0]) 3317 if ((rule->match_tag = pf_tagname2tag( 3318 rule->match_tagname)) == 0) 3319 error = EBUSY; 3320 3321 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE) 3322 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr); 3323 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE) 3324 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr); 3325 3326 if (error) { 3327 pf_free_eth_rule(rule); 3328 PF_RULES_WUNLOCK(); 3329 ERROUT(error); 3330 } 3331 3332 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) { 3333 pf_free_eth_rule(rule); 3334 PF_RULES_WUNLOCK(); 3335 ERROUT(EINVAL); 3336 } 3337 3338 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq); 3339 if (tail) 3340 rule->nr = tail->nr + 1; 3341 else 3342 rule->nr = 0; 3343 3344 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries); 3345 3346 PF_RULES_WUNLOCK(); 3347 3348 #undef ERROUT 3349 DIOCADDETHRULE_error: 3350 nvlist_destroy(nvl); 3351 free(nvlpacked, M_NVLIST); 3352 break; 3353 } 3354 3355 case DIOCGETETHRULESETS: { 3356 struct epoch_tracker et; 3357 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3358 nvlist_t *nvl = NULL; 3359 void *nvlpacked = NULL; 3360 struct pf_keth_ruleset *ruleset; 3361 struct pf_keth_anchor *anchor; 3362 int nr = 0; 3363 3364 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x) 3365 3366 if (nv->len > pf_ioctl_maxcount) 3367 ERROUT(ENOMEM); 3368 3369 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3370 error = copyin(nv->data, nvlpacked, nv->len); 3371 if (error) 3372 ERROUT(error); 3373 3374 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3375 if (nvl == NULL) 3376 ERROUT(EBADMSG); 3377 if (! nvlist_exists_string(nvl, "path")) 3378 ERROUT(EBADMSG); 3379 3380 NET_EPOCH_ENTER(et); 3381 3382 if ((ruleset = pf_find_keth_ruleset( 3383 nvlist_get_string(nvl, "path"))) == NULL) { 3384 NET_EPOCH_EXIT(et); 3385 ERROUT(ENOENT); 3386 } 3387 3388 if (ruleset->anchor == NULL) { 3389 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors) 3390 if (anchor->parent == NULL) 3391 nr++; 3392 } else { 3393 RB_FOREACH(anchor, pf_keth_anchor_node, 3394 &ruleset->anchor->children) 3395 nr++; 3396 } 3397 3398 NET_EPOCH_EXIT(et); 3399 3400 nvlist_destroy(nvl); 3401 nvl = NULL; 3402 free(nvlpacked, M_NVLIST); 3403 nvlpacked = NULL; 3404 3405 nvl = nvlist_create(0); 3406 if (nvl == NULL) 3407 ERROUT(ENOMEM); 3408 3409 nvlist_add_number(nvl, "nr", nr); 3410 3411 nvlpacked = nvlist_pack(nvl, &nv->len); 3412 if (nvlpacked == NULL) 3413 ERROUT(ENOMEM); 3414 3415 if (nv->size == 0) 3416 ERROUT(0); 3417 else if (nv->size < nv->len) 3418 ERROUT(ENOSPC); 3419 3420 error = copyout(nvlpacked, nv->data, nv->len); 3421 3422 #undef ERROUT 3423 DIOCGETETHRULESETS_error: 3424 free(nvlpacked, M_NVLIST); 3425 nvlist_destroy(nvl); 3426 break; 3427 } 3428 3429 case DIOCGETETHRULESET: { 3430 struct epoch_tracker et; 3431 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3432 nvlist_t *nvl = NULL; 3433 void *nvlpacked = NULL; 3434 struct pf_keth_ruleset *ruleset; 3435 struct pf_keth_anchor *anchor; 3436 int nr = 0, req_nr = 0; 3437 bool found = false; 3438 3439 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x) 3440 3441 if (nv->len > pf_ioctl_maxcount) 3442 ERROUT(ENOMEM); 3443 3444 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3445 error = copyin(nv->data, nvlpacked, nv->len); 3446 if (error) 3447 ERROUT(error); 3448 3449 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3450 if (nvl == NULL) 3451 ERROUT(EBADMSG); 3452 if (! nvlist_exists_string(nvl, "path")) 3453 ERROUT(EBADMSG); 3454 if (! nvlist_exists_number(nvl, "nr")) 3455 ERROUT(EBADMSG); 3456 3457 req_nr = nvlist_get_number(nvl, "nr"); 3458 3459 NET_EPOCH_ENTER(et); 3460 3461 if ((ruleset = pf_find_keth_ruleset( 3462 nvlist_get_string(nvl, "path"))) == NULL) { 3463 NET_EPOCH_EXIT(et); 3464 ERROUT(ENOENT); 3465 } 3466 3467 nvlist_destroy(nvl); 3468 nvl = NULL; 3469 free(nvlpacked, M_NVLIST); 3470 nvlpacked = NULL; 3471 3472 nvl = nvlist_create(0); 3473 if (nvl == NULL) { 3474 NET_EPOCH_EXIT(et); 3475 ERROUT(ENOMEM); 3476 } 3477 3478 if (ruleset->anchor == NULL) { 3479 RB_FOREACH(anchor, pf_keth_anchor_global, 3480 &V_pf_keth_anchors) { 3481 if (anchor->parent == NULL && nr++ == req_nr) { 3482 found = true; 3483 break; 3484 } 3485 } 3486 } else { 3487 RB_FOREACH(anchor, pf_keth_anchor_node, 3488 &ruleset->anchor->children) { 3489 if (nr++ == req_nr) { 3490 found = true; 3491 break; 3492 } 3493 } 3494 } 3495 3496 NET_EPOCH_EXIT(et); 3497 if (found) { 3498 nvlist_add_number(nvl, "nr", nr); 3499 nvlist_add_string(nvl, "name", anchor->name); 3500 if (ruleset->anchor) 3501 nvlist_add_string(nvl, "path", 3502 ruleset->anchor->path); 3503 else 3504 nvlist_add_string(nvl, "path", ""); 3505 } else { 3506 ERROUT(EBUSY); 3507 } 3508 3509 nvlpacked = nvlist_pack(nvl, &nv->len); 3510 if (nvlpacked == NULL) 3511 ERROUT(ENOMEM); 3512 3513 if (nv->size == 0) 3514 ERROUT(0); 3515 else if (nv->size < nv->len) 3516 ERROUT(ENOSPC); 3517 3518 error = copyout(nvlpacked, nv->data, nv->len); 3519 3520 #undef ERROUT 3521 DIOCGETETHRULESET_error: 3522 free(nvlpacked, M_NVLIST); 3523 nvlist_destroy(nvl); 3524 break; 3525 } 3526 3527 case DIOCADDRULENV: { 3528 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3529 nvlist_t *nvl = NULL; 3530 void *nvlpacked = NULL; 3531 struct pf_krule *rule = NULL; 3532 const char *anchor = "", *anchor_call = ""; 3533 uint32_t ticket = 0, pool_ticket = 0; 3534 3535 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x) 3536 3537 if (nv->len > pf_ioctl_maxcount) 3538 ERROUT(ENOMEM); 3539 3540 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3541 error = copyin(nv->data, nvlpacked, nv->len); 3542 if (error) 3543 ERROUT(error); 3544 3545 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3546 if (nvl == NULL) 3547 ERROUT(EBADMSG); 3548 3549 if (! nvlist_exists_number(nvl, "ticket")) 3550 ERROUT(EINVAL); 3551 ticket = nvlist_get_number(nvl, "ticket"); 3552 3553 if (! nvlist_exists_number(nvl, "pool_ticket")) 3554 ERROUT(EINVAL); 3555 pool_ticket = nvlist_get_number(nvl, "pool_ticket"); 3556 3557 if (! nvlist_exists_nvlist(nvl, "rule")) 3558 ERROUT(EINVAL); 3559 3560 rule = pf_krule_alloc(); 3561 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"), 3562 rule); 3563 if (error) 3564 ERROUT(error); 3565 3566 if (nvlist_exists_string(nvl, "anchor")) 3567 anchor = nvlist_get_string(nvl, "anchor"); 3568 if (nvlist_exists_string(nvl, "anchor_call")) 3569 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3570 3571 if ((error = nvlist_error(nvl))) 3572 ERROUT(error); 3573 3574 /* Frees rule on error */ 3575 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor, 3576 anchor_call, td->td_ucred->cr_ruid, 3577 td->td_proc ? td->td_proc->p_pid : 0); 3578 3579 nvlist_destroy(nvl); 3580 free(nvlpacked, M_NVLIST); 3581 break; 3582 #undef ERROUT 3583 DIOCADDRULENV_error: 3584 pf_krule_free(rule); 3585 nvlist_destroy(nvl); 3586 free(nvlpacked, M_NVLIST); 3587 3588 break; 3589 } 3590 case DIOCADDRULE: { 3591 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3592 struct pf_krule *rule; 3593 3594 rule = pf_krule_alloc(); 3595 error = pf_rule_to_krule(&pr->rule, rule); 3596 if (error != 0) { 3597 pf_krule_free(rule); 3598 goto fail; 3599 } 3600 3601 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3602 3603 /* Frees rule on error */ 3604 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, 3605 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid, 3606 td->td_proc ? td->td_proc->p_pid : 0); 3607 break; 3608 } 3609 3610 case DIOCGETRULES: { 3611 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3612 3613 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3614 3615 error = pf_ioctl_getrules(pr); 3616 3617 break; 3618 } 3619 3620 case DIOCGETRULENV: { 3621 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3622 nvlist_t *nvrule = NULL; 3623 nvlist_t *nvl = NULL; 3624 struct pf_kruleset *ruleset; 3625 struct pf_krule *rule; 3626 void *nvlpacked = NULL; 3627 int rs_num, nr; 3628 bool clear_counter = false; 3629 3630 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x) 3631 3632 if (nv->len > pf_ioctl_maxcount) 3633 ERROUT(ENOMEM); 3634 3635 /* Copy the request in */ 3636 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3637 error = copyin(nv->data, nvlpacked, nv->len); 3638 if (error) 3639 ERROUT(error); 3640 3641 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3642 if (nvl == NULL) 3643 ERROUT(EBADMSG); 3644 3645 if (! nvlist_exists_string(nvl, "anchor")) 3646 ERROUT(EBADMSG); 3647 if (! nvlist_exists_number(nvl, "ruleset")) 3648 ERROUT(EBADMSG); 3649 if (! nvlist_exists_number(nvl, "ticket")) 3650 ERROUT(EBADMSG); 3651 if (! nvlist_exists_number(nvl, "nr")) 3652 ERROUT(EBADMSG); 3653 3654 if (nvlist_exists_bool(nvl, "clear_counter")) 3655 clear_counter = nvlist_get_bool(nvl, "clear_counter"); 3656 3657 if (clear_counter && !(flags & FWRITE)) 3658 ERROUT(EACCES); 3659 3660 nr = nvlist_get_number(nvl, "nr"); 3661 3662 PF_RULES_WLOCK(); 3663 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor")); 3664 if (ruleset == NULL) { 3665 PF_RULES_WUNLOCK(); 3666 ERROUT(ENOENT); 3667 } 3668 3669 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset")); 3670 if (rs_num >= PF_RULESET_MAX) { 3671 PF_RULES_WUNLOCK(); 3672 ERROUT(EINVAL); 3673 } 3674 3675 if (nvlist_get_number(nvl, "ticket") != 3676 ruleset->rules[rs_num].active.ticket) { 3677 PF_RULES_WUNLOCK(); 3678 ERROUT(EBUSY); 3679 } 3680 3681 if ((error = nvlist_error(nvl))) { 3682 PF_RULES_WUNLOCK(); 3683 ERROUT(error); 3684 } 3685 3686 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3687 while ((rule != NULL) && (rule->nr != nr)) 3688 rule = TAILQ_NEXT(rule, entries); 3689 if (rule == NULL) { 3690 PF_RULES_WUNLOCK(); 3691 ERROUT(EBUSY); 3692 } 3693 3694 nvrule = pf_krule_to_nvrule(rule); 3695 3696 nvlist_destroy(nvl); 3697 nvl = nvlist_create(0); 3698 if (nvl == NULL) { 3699 PF_RULES_WUNLOCK(); 3700 ERROUT(ENOMEM); 3701 } 3702 nvlist_add_number(nvl, "nr", nr); 3703 nvlist_add_nvlist(nvl, "rule", nvrule); 3704 nvlist_destroy(nvrule); 3705 nvrule = NULL; 3706 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) { 3707 PF_RULES_WUNLOCK(); 3708 ERROUT(EBUSY); 3709 } 3710 3711 free(nvlpacked, M_NVLIST); 3712 nvlpacked = nvlist_pack(nvl, &nv->len); 3713 if (nvlpacked == NULL) { 3714 PF_RULES_WUNLOCK(); 3715 ERROUT(ENOMEM); 3716 } 3717 3718 if (nv->size == 0) { 3719 PF_RULES_WUNLOCK(); 3720 ERROUT(0); 3721 } 3722 else if (nv->size < nv->len) { 3723 PF_RULES_WUNLOCK(); 3724 ERROUT(ENOSPC); 3725 } 3726 3727 if (clear_counter) 3728 pf_krule_clear_counters(rule); 3729 3730 PF_RULES_WUNLOCK(); 3731 3732 error = copyout(nvlpacked, nv->data, nv->len); 3733 3734 #undef ERROUT 3735 DIOCGETRULENV_error: 3736 free(nvlpacked, M_NVLIST); 3737 nvlist_destroy(nvrule); 3738 nvlist_destroy(nvl); 3739 3740 break; 3741 } 3742 3743 case DIOCCHANGERULE: { 3744 struct pfioc_rule *pcr = (struct pfioc_rule *)addr; 3745 struct pf_kruleset *ruleset; 3746 struct pf_krule *oldrule = NULL, *newrule = NULL; 3747 struct pfi_kkif *kif = NULL; 3748 struct pf_kpooladdr *pa; 3749 u_int32_t nr = 0; 3750 int rs_num; 3751 3752 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0'; 3753 3754 if (pcr->action < PF_CHANGE_ADD_HEAD || 3755 pcr->action > PF_CHANGE_GET_TICKET) { 3756 error = EINVAL; 3757 goto fail; 3758 } 3759 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { 3760 error = EINVAL; 3761 goto fail; 3762 } 3763 3764 if (pcr->action != PF_CHANGE_REMOVE) { 3765 newrule = pf_krule_alloc(); 3766 error = pf_rule_to_krule(&pcr->rule, newrule); 3767 if (error != 0) { 3768 pf_krule_free(newrule); 3769 goto fail; 3770 } 3771 3772 if ((error = pf_rule_checkaf(newrule))) { 3773 pf_krule_free(newrule); 3774 goto fail; 3775 } 3776 if (newrule->ifname[0]) 3777 kif = pf_kkif_create(M_WAITOK); 3778 pf_counter_u64_init(&newrule->evaluations, M_WAITOK); 3779 for (int i = 0; i < 2; i++) { 3780 pf_counter_u64_init(&newrule->packets[i], M_WAITOK); 3781 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK); 3782 } 3783 newrule->states_cur = counter_u64_alloc(M_WAITOK); 3784 newrule->states_tot = counter_u64_alloc(M_WAITOK); 3785 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 3786 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 3787 newrule->cuid = td->td_ucred->cr_ruid; 3788 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 3789 TAILQ_INIT(&newrule->nat.list); 3790 TAILQ_INIT(&newrule->rdr.list); 3791 TAILQ_INIT(&newrule->route.list); 3792 } 3793 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x) 3794 3795 PF_CONFIG_LOCK(); 3796 PF_RULES_WLOCK(); 3797 #ifdef PF_WANT_32_TO_64_COUNTER 3798 if (newrule != NULL) { 3799 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist); 3800 newrule->allrulelinked = true; 3801 V_pf_allrulecount++; 3802 } 3803 #endif 3804 3805 if (!(pcr->action == PF_CHANGE_REMOVE || 3806 pcr->action == PF_CHANGE_GET_TICKET) && 3807 pcr->pool_ticket != V_ticket_pabuf) 3808 ERROUT(EBUSY); 3809 3810 ruleset = pf_find_kruleset(pcr->anchor); 3811 if (ruleset == NULL) 3812 ERROUT(EINVAL); 3813 3814 rs_num = pf_get_ruleset_number(pcr->rule.action); 3815 if (rs_num >= PF_RULESET_MAX) 3816 ERROUT(EINVAL); 3817 3818 /* 3819 * XXXMJG: there is no guarantee that the ruleset was 3820 * created by the usual route of calling DIOCXBEGIN. 3821 * As a result it is possible the rule tree will not 3822 * be allocated yet. Hack around it by doing it here. 3823 * Note it is fine to let the tree persist in case of 3824 * error as it will be freed down the road on future 3825 * updates (if need be). 3826 */ 3827 if (ruleset->rules[rs_num].active.tree == NULL) { 3828 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT); 3829 if (ruleset->rules[rs_num].active.tree == NULL) { 3830 ERROUT(ENOMEM); 3831 } 3832 } 3833 3834 if (pcr->action == PF_CHANGE_GET_TICKET) { 3835 pcr->ticket = ++ruleset->rules[rs_num].active.ticket; 3836 ERROUT(0); 3837 } else if (pcr->ticket != 3838 ruleset->rules[rs_num].active.ticket) 3839 ERROUT(EINVAL); 3840 3841 if (pcr->action != PF_CHANGE_REMOVE) { 3842 if (newrule->ifname[0]) { 3843 newrule->kif = pfi_kkif_attach(kif, 3844 newrule->ifname); 3845 kif = NULL; 3846 pfi_kkif_ref(newrule->kif); 3847 } else 3848 newrule->kif = NULL; 3849 3850 if (newrule->rtableid > 0 && 3851 newrule->rtableid >= rt_numfibs) 3852 error = EBUSY; 3853 3854 #ifdef ALTQ 3855 /* set queue IDs */ 3856 if (newrule->qname[0] != 0) { 3857 if ((newrule->qid = 3858 pf_qname2qid(newrule->qname)) == 0) 3859 error = EBUSY; 3860 else if (newrule->pqname[0] != 0) { 3861 if ((newrule->pqid = 3862 pf_qname2qid(newrule->pqname)) == 0) 3863 error = EBUSY; 3864 } else 3865 newrule->pqid = newrule->qid; 3866 } 3867 #endif /* ALTQ */ 3868 if (newrule->tagname[0]) 3869 if ((newrule->tag = 3870 pf_tagname2tag(newrule->tagname)) == 0) 3871 error = EBUSY; 3872 if (newrule->match_tagname[0]) 3873 if ((newrule->match_tag = pf_tagname2tag( 3874 newrule->match_tagname)) == 0) 3875 error = EBUSY; 3876 if (newrule->rt && !newrule->direction) 3877 error = EINVAL; 3878 if (!newrule->log) 3879 newrule->logif = 0; 3880 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) 3881 error = ENOMEM; 3882 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) 3883 error = ENOMEM; 3884 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call)) 3885 error = EINVAL; 3886 for (int i = 0; i < 3; i++) { 3887 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 3888 if (pa->addr.type == PF_ADDR_TABLE) { 3889 pa->addr.p.tbl = 3890 pfr_attach_table(ruleset, 3891 pa->addr.v.tblname); 3892 if (pa->addr.p.tbl == NULL) 3893 error = ENOMEM; 3894 } 3895 } 3896 3897 newrule->overload_tbl = NULL; 3898 if (newrule->overload_tblname[0]) { 3899 if ((newrule->overload_tbl = pfr_attach_table( 3900 ruleset, newrule->overload_tblname)) == 3901 NULL) 3902 error = EINVAL; 3903 else 3904 newrule->overload_tbl->pfrkt_flags |= 3905 PFR_TFLAG_ACTIVE; 3906 } 3907 3908 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list); 3909 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list); 3910 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list); 3911 if (((((newrule->action == PF_NAT) || 3912 (newrule->action == PF_RDR) || 3913 (newrule->action == PF_BINAT) || 3914 (newrule->rt > PF_NOPFROUTE)) && 3915 !newrule->anchor)) && 3916 (TAILQ_FIRST(&newrule->rdr.list) == NULL)) 3917 error = EINVAL; 3918 3919 if (error) { 3920 pf_free_rule(newrule); 3921 PF_RULES_WUNLOCK(); 3922 PF_CONFIG_UNLOCK(); 3923 goto fail; 3924 } 3925 3926 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list); 3927 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list); 3928 } 3929 pf_empty_kpool(&V_pf_pabuf[0]); 3930 pf_empty_kpool(&V_pf_pabuf[1]); 3931 pf_empty_kpool(&V_pf_pabuf[2]); 3932 3933 if (pcr->action == PF_CHANGE_ADD_HEAD) 3934 oldrule = TAILQ_FIRST( 3935 ruleset->rules[rs_num].active.ptr); 3936 else if (pcr->action == PF_CHANGE_ADD_TAIL) 3937 oldrule = TAILQ_LAST( 3938 ruleset->rules[rs_num].active.ptr, pf_krulequeue); 3939 else { 3940 oldrule = TAILQ_FIRST( 3941 ruleset->rules[rs_num].active.ptr); 3942 while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) 3943 oldrule = TAILQ_NEXT(oldrule, entries); 3944 if (oldrule == NULL) { 3945 if (newrule != NULL) 3946 pf_free_rule(newrule); 3947 PF_RULES_WUNLOCK(); 3948 PF_CONFIG_UNLOCK(); 3949 error = EINVAL; 3950 goto fail; 3951 } 3952 } 3953 3954 if (pcr->action == PF_CHANGE_REMOVE) { 3955 pf_unlink_rule(ruleset->rules[rs_num].active.ptr, 3956 oldrule); 3957 RB_REMOVE(pf_krule_global, 3958 ruleset->rules[rs_num].active.tree, oldrule); 3959 ruleset->rules[rs_num].active.rcount--; 3960 } else { 3961 pf_hash_rule(newrule); 3962 if (RB_INSERT(pf_krule_global, 3963 ruleset->rules[rs_num].active.tree, newrule) != NULL) { 3964 pf_free_rule(newrule); 3965 PF_RULES_WUNLOCK(); 3966 PF_CONFIG_UNLOCK(); 3967 error = EEXIST; 3968 goto fail; 3969 } 3970 3971 if (oldrule == NULL) 3972 TAILQ_INSERT_TAIL( 3973 ruleset->rules[rs_num].active.ptr, 3974 newrule, entries); 3975 else if (pcr->action == PF_CHANGE_ADD_HEAD || 3976 pcr->action == PF_CHANGE_ADD_BEFORE) 3977 TAILQ_INSERT_BEFORE(oldrule, newrule, entries); 3978 else 3979 TAILQ_INSERT_AFTER( 3980 ruleset->rules[rs_num].active.ptr, 3981 oldrule, newrule, entries); 3982 ruleset->rules[rs_num].active.rcount++; 3983 } 3984 3985 nr = 0; 3986 TAILQ_FOREACH(oldrule, 3987 ruleset->rules[rs_num].active.ptr, entries) 3988 oldrule->nr = nr++; 3989 3990 ruleset->rules[rs_num].active.ticket++; 3991 3992 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); 3993 pf_remove_if_empty_kruleset(ruleset); 3994 3995 PF_RULES_WUNLOCK(); 3996 PF_CONFIG_UNLOCK(); 3997 break; 3998 3999 #undef ERROUT 4000 DIOCCHANGERULE_error: 4001 PF_RULES_WUNLOCK(); 4002 PF_CONFIG_UNLOCK(); 4003 pf_krule_free(newrule); 4004 pf_kkif_free(kif); 4005 break; 4006 } 4007 4008 case DIOCCLRSTATESNV: { 4009 error = pf_clearstates_nv((struct pfioc_nv *)addr); 4010 break; 4011 } 4012 4013 case DIOCKILLSTATESNV: { 4014 error = pf_killstates_nv((struct pfioc_nv *)addr); 4015 break; 4016 } 4017 4018 case DIOCADDSTATE: { 4019 struct pfioc_state *ps = (struct pfioc_state *)addr; 4020 struct pfsync_state_1301 *sp = &ps->state; 4021 4022 if (sp->timeout >= PFTM_MAX) { 4023 error = EINVAL; 4024 goto fail; 4025 } 4026 if (V_pfsync_state_import_ptr != NULL) { 4027 PF_RULES_RLOCK(); 4028 error = V_pfsync_state_import_ptr( 4029 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL, 4030 PFSYNC_MSG_VERSION_1301); 4031 PF_RULES_RUNLOCK(); 4032 } else 4033 error = EOPNOTSUPP; 4034 break; 4035 } 4036 4037 case DIOCGETSTATE: { 4038 struct pfioc_state *ps = (struct pfioc_state *)addr; 4039 struct pf_kstate *s; 4040 4041 s = pf_find_state_byid(ps->state.id, ps->state.creatorid); 4042 if (s == NULL) { 4043 error = ENOENT; 4044 goto fail; 4045 } 4046 4047 pfsync_state_export((union pfsync_state_union*)&ps->state, 4048 s, PFSYNC_MSG_VERSION_1301); 4049 PF_STATE_UNLOCK(s); 4050 break; 4051 } 4052 4053 case DIOCGETSTATENV: { 4054 error = pf_getstate((struct pfioc_nv *)addr); 4055 break; 4056 } 4057 4058 #ifdef COMPAT_FREEBSD14 4059 case DIOCGETSTATES: { 4060 struct pfioc_states *ps = (struct pfioc_states *)addr; 4061 struct pf_kstate *s; 4062 struct pfsync_state_1301 *pstore, *p; 4063 int i, nr; 4064 size_t slice_count = 16, count; 4065 void *out; 4066 4067 if (ps->ps_len <= 0) { 4068 nr = uma_zone_get_cur(V_pf_state_z); 4069 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4070 break; 4071 } 4072 4073 out = ps->ps_states; 4074 pstore = mallocarray(slice_count, 4075 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO); 4076 nr = 0; 4077 4078 for (i = 0; i <= V_pf_hashmask; i++) { 4079 struct pf_idhash *ih = &V_pf_idhash[i]; 4080 4081 DIOCGETSTATES_retry: 4082 p = pstore; 4083 4084 if (LIST_EMPTY(&ih->states)) 4085 continue; 4086 4087 PF_HASHROW_LOCK(ih); 4088 count = 0; 4089 LIST_FOREACH(s, &ih->states, entry) { 4090 if (s->timeout == PFTM_UNLINKED) 4091 continue; 4092 count++; 4093 } 4094 4095 if (count > slice_count) { 4096 PF_HASHROW_UNLOCK(ih); 4097 free(pstore, M_TEMP); 4098 slice_count = count * 2; 4099 pstore = mallocarray(slice_count, 4100 sizeof(struct pfsync_state_1301), M_TEMP, 4101 M_WAITOK | M_ZERO); 4102 goto DIOCGETSTATES_retry; 4103 } 4104 4105 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4106 PF_HASHROW_UNLOCK(ih); 4107 goto DIOCGETSTATES_full; 4108 } 4109 4110 LIST_FOREACH(s, &ih->states, entry) { 4111 if (s->timeout == PFTM_UNLINKED) 4112 continue; 4113 4114 pfsync_state_export((union pfsync_state_union*)p, 4115 s, PFSYNC_MSG_VERSION_1301); 4116 p++; 4117 nr++; 4118 } 4119 PF_HASHROW_UNLOCK(ih); 4120 error = copyout(pstore, out, 4121 sizeof(struct pfsync_state_1301) * count); 4122 if (error) 4123 goto fail; 4124 out = ps->ps_states + nr; 4125 } 4126 DIOCGETSTATES_full: 4127 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4128 free(pstore, M_TEMP); 4129 4130 break; 4131 } 4132 4133 case DIOCGETSTATESV2: { 4134 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr; 4135 struct pf_kstate *s; 4136 struct pf_state_export *pstore, *p; 4137 int i, nr; 4138 size_t slice_count = 16, count; 4139 void *out; 4140 4141 if (ps->ps_req_version > PF_STATE_VERSION) { 4142 error = ENOTSUP; 4143 goto fail; 4144 } 4145 4146 if (ps->ps_len <= 0) { 4147 nr = uma_zone_get_cur(V_pf_state_z); 4148 ps->ps_len = sizeof(struct pf_state_export) * nr; 4149 break; 4150 } 4151 4152 out = ps->ps_states; 4153 pstore = mallocarray(slice_count, 4154 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO); 4155 nr = 0; 4156 4157 for (i = 0; i <= V_pf_hashmask; i++) { 4158 struct pf_idhash *ih = &V_pf_idhash[i]; 4159 4160 DIOCGETSTATESV2_retry: 4161 p = pstore; 4162 4163 if (LIST_EMPTY(&ih->states)) 4164 continue; 4165 4166 PF_HASHROW_LOCK(ih); 4167 count = 0; 4168 LIST_FOREACH(s, &ih->states, entry) { 4169 if (s->timeout == PFTM_UNLINKED) 4170 continue; 4171 count++; 4172 } 4173 4174 if (count > slice_count) { 4175 PF_HASHROW_UNLOCK(ih); 4176 free(pstore, M_TEMP); 4177 slice_count = count * 2; 4178 pstore = mallocarray(slice_count, 4179 sizeof(struct pf_state_export), M_TEMP, 4180 M_WAITOK | M_ZERO); 4181 goto DIOCGETSTATESV2_retry; 4182 } 4183 4184 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4185 PF_HASHROW_UNLOCK(ih); 4186 goto DIOCGETSTATESV2_full; 4187 } 4188 4189 LIST_FOREACH(s, &ih->states, entry) { 4190 if (s->timeout == PFTM_UNLINKED) 4191 continue; 4192 4193 pf_state_export(p, s); 4194 p++; 4195 nr++; 4196 } 4197 PF_HASHROW_UNLOCK(ih); 4198 error = copyout(pstore, out, 4199 sizeof(struct pf_state_export) * count); 4200 if (error) 4201 goto fail; 4202 out = ps->ps_states + nr; 4203 } 4204 DIOCGETSTATESV2_full: 4205 ps->ps_len = nr * sizeof(struct pf_state_export); 4206 free(pstore, M_TEMP); 4207 4208 break; 4209 } 4210 #endif 4211 case DIOCGETSTATUSNV: { 4212 error = pf_getstatus((struct pfioc_nv *)addr); 4213 break; 4214 } 4215 4216 case DIOCSETSTATUSIF: { 4217 struct pfioc_if *pi = (struct pfioc_if *)addr; 4218 4219 if (pi->ifname[0] == 0) { 4220 bzero(V_pf_status.ifname, IFNAMSIZ); 4221 break; 4222 } 4223 PF_RULES_WLOCK(); 4224 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); 4225 PF_RULES_WUNLOCK(); 4226 break; 4227 } 4228 4229 case DIOCCLRSTATUS: { 4230 pf_ioctl_clear_status(); 4231 break; 4232 } 4233 4234 case DIOCNATLOOK: { 4235 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; 4236 4237 error = pf_ioctl_natlook(pnl); 4238 break; 4239 } 4240 4241 case DIOCSETTIMEOUT: { 4242 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4243 4244 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds, 4245 &pt->seconds); 4246 break; 4247 } 4248 4249 case DIOCGETTIMEOUT: { 4250 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4251 4252 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds); 4253 break; 4254 } 4255 4256 case DIOCGETLIMIT: { 4257 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4258 4259 error = pf_ioctl_get_limit(pl->index, &pl->limit); 4260 break; 4261 } 4262 4263 case DIOCSETLIMIT: { 4264 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4265 unsigned int old_limit; 4266 4267 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit); 4268 pl->limit = old_limit; 4269 break; 4270 } 4271 4272 case DIOCSETDEBUG: { 4273 u_int32_t *level = (u_int32_t *)addr; 4274 4275 PF_RULES_WLOCK(); 4276 V_pf_status.debug = *level; 4277 PF_RULES_WUNLOCK(); 4278 break; 4279 } 4280 4281 case DIOCCLRRULECTRS: { 4282 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ 4283 struct pf_kruleset *ruleset = &pf_main_ruleset; 4284 struct pf_krule *rule; 4285 4286 PF_RULES_WLOCK(); 4287 TAILQ_FOREACH(rule, 4288 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { 4289 pf_counter_u64_zero(&rule->evaluations); 4290 for (int i = 0; i < 2; i++) { 4291 pf_counter_u64_zero(&rule->packets[i]); 4292 pf_counter_u64_zero(&rule->bytes[i]); 4293 } 4294 } 4295 PF_RULES_WUNLOCK(); 4296 break; 4297 } 4298 4299 case DIOCGIFSPEEDV0: 4300 case DIOCGIFSPEEDV1: { 4301 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr; 4302 struct pf_ifspeed_v1 ps; 4303 struct ifnet *ifp; 4304 4305 if (psp->ifname[0] == '\0') { 4306 error = EINVAL; 4307 goto fail; 4308 } 4309 4310 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); 4311 if (error != 0) 4312 goto fail; 4313 ifp = ifunit(ps.ifname); 4314 if (ifp != NULL) { 4315 psp->baudrate32 = 4316 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX); 4317 if (cmd == DIOCGIFSPEEDV1) 4318 psp->baudrate = ifp->if_baudrate; 4319 } else { 4320 error = EINVAL; 4321 } 4322 break; 4323 } 4324 4325 #ifdef ALTQ 4326 case DIOCSTARTALTQ: { 4327 struct pf_altq *altq; 4328 4329 PF_RULES_WLOCK(); 4330 /* enable all altq interfaces on active list */ 4331 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4332 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4333 error = pf_enable_altq(altq); 4334 if (error != 0) 4335 break; 4336 } 4337 } 4338 if (error == 0) 4339 V_pf_altq_running = 1; 4340 PF_RULES_WUNLOCK(); 4341 DPFPRINTF(PF_DEBUG_MISC, "altq: started"); 4342 break; 4343 } 4344 4345 case DIOCSTOPALTQ: { 4346 struct pf_altq *altq; 4347 4348 PF_RULES_WLOCK(); 4349 /* disable all altq interfaces on active list */ 4350 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4351 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4352 error = pf_disable_altq(altq); 4353 if (error != 0) 4354 break; 4355 } 4356 } 4357 if (error == 0) 4358 V_pf_altq_running = 0; 4359 PF_RULES_WUNLOCK(); 4360 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped"); 4361 break; 4362 } 4363 4364 case DIOCADDALTQV0: 4365 case DIOCADDALTQV1: { 4366 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4367 struct pf_altq *altq, *a; 4368 struct ifnet *ifp; 4369 4370 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); 4371 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); 4372 if (error) 4373 goto fail; 4374 altq->local_flags = 0; 4375 4376 PF_RULES_WLOCK(); 4377 if (pa->ticket != V_ticket_altqs_inactive) { 4378 PF_RULES_WUNLOCK(); 4379 free(altq, M_PFALTQ); 4380 error = EBUSY; 4381 goto fail; 4382 } 4383 4384 /* 4385 * if this is for a queue, find the discipline and 4386 * copy the necessary fields 4387 */ 4388 if (altq->qname[0] != 0) { 4389 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) { 4390 PF_RULES_WUNLOCK(); 4391 error = EBUSY; 4392 free(altq, M_PFALTQ); 4393 goto fail; 4394 } 4395 altq->altq_disc = NULL; 4396 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { 4397 if (strncmp(a->ifname, altq->ifname, 4398 IFNAMSIZ) == 0) { 4399 altq->altq_disc = a->altq_disc; 4400 break; 4401 } 4402 } 4403 } 4404 4405 if ((ifp = ifunit(altq->ifname)) == NULL) 4406 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 4407 else 4408 error = altq_add(ifp, altq); 4409 4410 if (error) { 4411 PF_RULES_WUNLOCK(); 4412 free(altq, M_PFALTQ); 4413 goto fail; 4414 } 4415 4416 if (altq->qname[0] != 0) 4417 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); 4418 else 4419 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries); 4420 /* version error check done on import above */ 4421 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4422 PF_RULES_WUNLOCK(); 4423 break; 4424 } 4425 4426 case DIOCGETALTQSV0: 4427 case DIOCGETALTQSV1: { 4428 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4429 struct pf_altq *altq; 4430 4431 PF_RULES_RLOCK(); 4432 pa->nr = 0; 4433 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) 4434 pa->nr++; 4435 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) 4436 pa->nr++; 4437 pa->ticket = V_ticket_altqs_active; 4438 PF_RULES_RUNLOCK(); 4439 break; 4440 } 4441 4442 case DIOCGETALTQV0: 4443 case DIOCGETALTQV1: { 4444 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4445 struct pf_altq *altq; 4446 4447 PF_RULES_RLOCK(); 4448 if (pa->ticket != V_ticket_altqs_active) { 4449 PF_RULES_RUNLOCK(); 4450 error = EBUSY; 4451 goto fail; 4452 } 4453 altq = pf_altq_get_nth_active(pa->nr); 4454 if (altq == NULL) { 4455 PF_RULES_RUNLOCK(); 4456 error = EBUSY; 4457 goto fail; 4458 } 4459 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4460 PF_RULES_RUNLOCK(); 4461 break; 4462 } 4463 4464 case DIOCCHANGEALTQV0: 4465 case DIOCCHANGEALTQV1: 4466 /* CHANGEALTQ not supported yet! */ 4467 error = ENODEV; 4468 break; 4469 4470 case DIOCGETQSTATSV0: 4471 case DIOCGETQSTATSV1: { 4472 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr; 4473 struct pf_altq *altq; 4474 int nbytes; 4475 u_int32_t version; 4476 4477 PF_RULES_RLOCK(); 4478 if (pq->ticket != V_ticket_altqs_active) { 4479 PF_RULES_RUNLOCK(); 4480 error = EBUSY; 4481 goto fail; 4482 } 4483 nbytes = pq->nbytes; 4484 altq = pf_altq_get_nth_active(pq->nr); 4485 if (altq == NULL) { 4486 PF_RULES_RUNLOCK(); 4487 error = EBUSY; 4488 goto fail; 4489 } 4490 4491 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { 4492 PF_RULES_RUNLOCK(); 4493 error = ENXIO; 4494 goto fail; 4495 } 4496 PF_RULES_RUNLOCK(); 4497 if (cmd == DIOCGETQSTATSV0) 4498 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */ 4499 else 4500 version = pq->version; 4501 error = altq_getqstats(altq, pq->buf, &nbytes, version); 4502 if (error == 0) { 4503 pq->scheduler = altq->scheduler; 4504 pq->nbytes = nbytes; 4505 } 4506 break; 4507 } 4508 #endif /* ALTQ */ 4509 4510 case DIOCBEGINADDRS: { 4511 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4512 4513 error = pf_ioctl_begin_addrs(&pp->ticket); 4514 break; 4515 } 4516 4517 case DIOCADDADDR: { 4518 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4519 struct pf_nl_pooladdr npp = {}; 4520 4521 npp.which = PF_RDR; 4522 memcpy(&npp, pp, sizeof(*pp)); 4523 error = pf_ioctl_add_addr(&npp); 4524 break; 4525 } 4526 4527 case DIOCGETADDRS: { 4528 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4529 struct pf_nl_pooladdr npp = {}; 4530 4531 npp.which = PF_RDR; 4532 memcpy(&npp, pp, sizeof(*pp)); 4533 error = pf_ioctl_get_addrs(&npp); 4534 memcpy(pp, &npp, sizeof(*pp)); 4535 4536 break; 4537 } 4538 4539 case DIOCGETADDR: { 4540 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4541 struct pf_nl_pooladdr npp = {}; 4542 4543 npp.which = PF_RDR; 4544 memcpy(&npp, pp, sizeof(*pp)); 4545 error = pf_ioctl_get_addr(&npp); 4546 memcpy(pp, &npp, sizeof(*pp)); 4547 4548 break; 4549 } 4550 4551 case DIOCCHANGEADDR: { 4552 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; 4553 struct pf_kpool *pool; 4554 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL; 4555 struct pf_kruleset *ruleset; 4556 struct pfi_kkif *kif = NULL; 4557 4558 pca->anchor[sizeof(pca->anchor) - 1] = '\0'; 4559 4560 if (pca->action < PF_CHANGE_ADD_HEAD || 4561 pca->action > PF_CHANGE_REMOVE) { 4562 error = EINVAL; 4563 goto fail; 4564 } 4565 if (pca->addr.addr.type != PF_ADDR_ADDRMASK && 4566 pca->addr.addr.type != PF_ADDR_DYNIFTL && 4567 pca->addr.addr.type != PF_ADDR_TABLE) { 4568 error = EINVAL; 4569 goto fail; 4570 } 4571 if (pca->addr.addr.p.dyn != NULL) { 4572 error = EINVAL; 4573 goto fail; 4574 } 4575 4576 if (pca->action != PF_CHANGE_REMOVE) { 4577 #ifndef INET 4578 if (pca->af == AF_INET) { 4579 error = EAFNOSUPPORT; 4580 goto fail; 4581 } 4582 #endif /* INET */ 4583 #ifndef INET6 4584 if (pca->af == AF_INET6) { 4585 error = EAFNOSUPPORT; 4586 goto fail; 4587 } 4588 #endif /* INET6 */ 4589 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); 4590 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); 4591 if (newpa->ifname[0]) 4592 kif = pf_kkif_create(M_WAITOK); 4593 newpa->kif = NULL; 4594 } 4595 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x) 4596 PF_RULES_WLOCK(); 4597 ruleset = pf_find_kruleset(pca->anchor); 4598 if (ruleset == NULL) 4599 ERROUT(EBUSY); 4600 4601 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action, 4602 pca->r_num, pca->r_last, 1, 1, PF_RDR); 4603 if (pool == NULL) 4604 ERROUT(EBUSY); 4605 4606 if (pca->action != PF_CHANGE_REMOVE) { 4607 if (newpa->ifname[0]) { 4608 newpa->kif = pfi_kkif_attach(kif, newpa->ifname); 4609 pfi_kkif_ref(newpa->kif); 4610 kif = NULL; 4611 } 4612 4613 switch (newpa->addr.type) { 4614 case PF_ADDR_DYNIFTL: 4615 error = pfi_dynaddr_setup(&newpa->addr, 4616 pca->af); 4617 break; 4618 case PF_ADDR_TABLE: 4619 newpa->addr.p.tbl = pfr_attach_table(ruleset, 4620 newpa->addr.v.tblname); 4621 if (newpa->addr.p.tbl == NULL) 4622 error = ENOMEM; 4623 break; 4624 } 4625 if (error) 4626 goto DIOCCHANGEADDR_error; 4627 } 4628 4629 switch (pca->action) { 4630 case PF_CHANGE_ADD_HEAD: 4631 oldpa = TAILQ_FIRST(&pool->list); 4632 break; 4633 case PF_CHANGE_ADD_TAIL: 4634 oldpa = TAILQ_LAST(&pool->list, pf_kpalist); 4635 break; 4636 default: 4637 oldpa = TAILQ_FIRST(&pool->list); 4638 for (int i = 0; oldpa && i < pca->nr; i++) 4639 oldpa = TAILQ_NEXT(oldpa, entries); 4640 4641 if (oldpa == NULL) 4642 ERROUT(EINVAL); 4643 } 4644 4645 if (pca->action == PF_CHANGE_REMOVE) { 4646 TAILQ_REMOVE(&pool->list, oldpa, entries); 4647 switch (oldpa->addr.type) { 4648 case PF_ADDR_DYNIFTL: 4649 pfi_dynaddr_remove(oldpa->addr.p.dyn); 4650 break; 4651 case PF_ADDR_TABLE: 4652 pfr_detach_table(oldpa->addr.p.tbl); 4653 break; 4654 } 4655 if (oldpa->kif) 4656 pfi_kkif_unref(oldpa->kif); 4657 free(oldpa, M_PFRULE); 4658 } else { 4659 if (oldpa == NULL) 4660 TAILQ_INSERT_TAIL(&pool->list, newpa, entries); 4661 else if (pca->action == PF_CHANGE_ADD_HEAD || 4662 pca->action == PF_CHANGE_ADD_BEFORE) 4663 TAILQ_INSERT_BEFORE(oldpa, newpa, entries); 4664 else 4665 TAILQ_INSERT_AFTER(&pool->list, oldpa, 4666 newpa, entries); 4667 } 4668 4669 pool->cur = TAILQ_FIRST(&pool->list); 4670 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); 4671 PF_RULES_WUNLOCK(); 4672 break; 4673 4674 #undef ERROUT 4675 DIOCCHANGEADDR_error: 4676 if (newpa != NULL) { 4677 if (newpa->kif) 4678 pfi_kkif_unref(newpa->kif); 4679 free(newpa, M_PFRULE); 4680 } 4681 PF_RULES_WUNLOCK(); 4682 pf_kkif_free(kif); 4683 break; 4684 } 4685 4686 case DIOCGETRULESETS: { 4687 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4688 4689 pr->path[sizeof(pr->path) - 1] = '\0'; 4690 4691 error = pf_ioctl_get_rulesets(pr); 4692 break; 4693 } 4694 4695 case DIOCGETRULESET: { 4696 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4697 4698 pr->path[sizeof(pr->path) - 1] = '\0'; 4699 4700 error = pf_ioctl_get_ruleset(pr); 4701 break; 4702 } 4703 4704 case DIOCRCLRTABLES: { 4705 struct pfioc_table *io = (struct pfioc_table *)addr; 4706 4707 if (io->pfrio_esize != 0) { 4708 error = ENODEV; 4709 goto fail; 4710 } 4711 PF_RULES_WLOCK(); 4712 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, 4713 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4714 PF_RULES_WUNLOCK(); 4715 break; 4716 } 4717 4718 case DIOCRADDTABLES: { 4719 struct pfioc_table *io = (struct pfioc_table *)addr; 4720 struct pfr_table *pfrts; 4721 size_t totlen; 4722 4723 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4724 error = ENODEV; 4725 goto fail; 4726 } 4727 4728 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4729 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4730 error = ENOMEM; 4731 goto fail; 4732 } 4733 4734 totlen = io->pfrio_size * sizeof(struct pfr_table); 4735 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4736 M_TEMP, M_WAITOK); 4737 error = copyin(io->pfrio_buffer, pfrts, totlen); 4738 if (error) { 4739 free(pfrts, M_TEMP); 4740 goto fail; 4741 } 4742 PF_RULES_WLOCK(); 4743 error = pfr_add_tables(pfrts, io->pfrio_size, 4744 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4745 PF_RULES_WUNLOCK(); 4746 free(pfrts, M_TEMP); 4747 break; 4748 } 4749 4750 case DIOCRDELTABLES: { 4751 struct pfioc_table *io = (struct pfioc_table *)addr; 4752 struct pfr_table *pfrts; 4753 size_t totlen; 4754 4755 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4756 error = ENODEV; 4757 goto fail; 4758 } 4759 4760 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4761 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4762 error = ENOMEM; 4763 goto fail; 4764 } 4765 4766 totlen = io->pfrio_size * sizeof(struct pfr_table); 4767 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4768 M_TEMP, M_WAITOK); 4769 error = copyin(io->pfrio_buffer, pfrts, totlen); 4770 if (error) { 4771 free(pfrts, M_TEMP); 4772 goto fail; 4773 } 4774 PF_RULES_WLOCK(); 4775 error = pfr_del_tables(pfrts, io->pfrio_size, 4776 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4777 PF_RULES_WUNLOCK(); 4778 free(pfrts, M_TEMP); 4779 break; 4780 } 4781 4782 case DIOCRGETTABLES: { 4783 struct pfioc_table *io = (struct pfioc_table *)addr; 4784 struct pfr_table *pfrts; 4785 size_t totlen; 4786 int n; 4787 4788 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4789 error = ENODEV; 4790 goto fail; 4791 } 4792 PF_RULES_RLOCK(); 4793 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4794 if (n < 0) { 4795 PF_RULES_RUNLOCK(); 4796 error = EINVAL; 4797 goto fail; 4798 } 4799 io->pfrio_size = min(io->pfrio_size, n); 4800 4801 totlen = io->pfrio_size * sizeof(struct pfr_table); 4802 4803 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4804 M_TEMP, M_NOWAIT | M_ZERO); 4805 if (pfrts == NULL) { 4806 error = ENOMEM; 4807 PF_RULES_RUNLOCK(); 4808 goto fail; 4809 } 4810 error = pfr_get_tables(&io->pfrio_table, pfrts, 4811 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4812 PF_RULES_RUNLOCK(); 4813 if (error == 0) 4814 error = copyout(pfrts, io->pfrio_buffer, totlen); 4815 free(pfrts, M_TEMP); 4816 break; 4817 } 4818 4819 case DIOCRGETTSTATS: { 4820 struct pfioc_table *io = (struct pfioc_table *)addr; 4821 struct pfr_tstats *pfrtstats; 4822 size_t totlen; 4823 int n; 4824 4825 if (io->pfrio_esize != sizeof(struct pfr_tstats)) { 4826 error = ENODEV; 4827 goto fail; 4828 } 4829 PF_TABLE_STATS_LOCK(); 4830 PF_RULES_RLOCK(); 4831 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4832 if (n < 0) { 4833 PF_RULES_RUNLOCK(); 4834 PF_TABLE_STATS_UNLOCK(); 4835 error = EINVAL; 4836 goto fail; 4837 } 4838 io->pfrio_size = min(io->pfrio_size, n); 4839 4840 totlen = io->pfrio_size * sizeof(struct pfr_tstats); 4841 pfrtstats = mallocarray(io->pfrio_size, 4842 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO); 4843 if (pfrtstats == NULL) { 4844 error = ENOMEM; 4845 PF_RULES_RUNLOCK(); 4846 PF_TABLE_STATS_UNLOCK(); 4847 goto fail; 4848 } 4849 error = pfr_get_tstats(&io->pfrio_table, pfrtstats, 4850 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4851 PF_RULES_RUNLOCK(); 4852 PF_TABLE_STATS_UNLOCK(); 4853 if (error == 0) 4854 error = copyout(pfrtstats, io->pfrio_buffer, totlen); 4855 free(pfrtstats, M_TEMP); 4856 break; 4857 } 4858 4859 case DIOCRCLRTSTATS: { 4860 struct pfioc_table *io = (struct pfioc_table *)addr; 4861 struct pfr_table *pfrts; 4862 size_t totlen; 4863 4864 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4865 error = ENODEV; 4866 goto fail; 4867 } 4868 4869 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4870 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4871 /* We used to count tables and use the minimum required 4872 * size, so we didn't fail on overly large requests. 4873 * Keep doing so. */ 4874 io->pfrio_size = pf_ioctl_maxcount; 4875 goto fail; 4876 } 4877 4878 totlen = io->pfrio_size * sizeof(struct pfr_table); 4879 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4880 M_TEMP, M_WAITOK); 4881 error = copyin(io->pfrio_buffer, pfrts, totlen); 4882 if (error) { 4883 free(pfrts, M_TEMP); 4884 goto fail; 4885 } 4886 4887 PF_TABLE_STATS_LOCK(); 4888 PF_RULES_RLOCK(); 4889 error = pfr_clr_tstats(pfrts, io->pfrio_size, 4890 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4891 PF_RULES_RUNLOCK(); 4892 PF_TABLE_STATS_UNLOCK(); 4893 free(pfrts, M_TEMP); 4894 break; 4895 } 4896 4897 case DIOCRSETTFLAGS: { 4898 struct pfioc_table *io = (struct pfioc_table *)addr; 4899 struct pfr_table *pfrts; 4900 size_t totlen; 4901 int n; 4902 4903 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4904 error = ENODEV; 4905 goto fail; 4906 } 4907 4908 PF_RULES_RLOCK(); 4909 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4910 if (n < 0) { 4911 PF_RULES_RUNLOCK(); 4912 error = EINVAL; 4913 goto fail; 4914 } 4915 4916 io->pfrio_size = min(io->pfrio_size, n); 4917 PF_RULES_RUNLOCK(); 4918 4919 totlen = io->pfrio_size * sizeof(struct pfr_table); 4920 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4921 M_TEMP, M_WAITOK); 4922 error = copyin(io->pfrio_buffer, pfrts, totlen); 4923 if (error) { 4924 free(pfrts, M_TEMP); 4925 goto fail; 4926 } 4927 PF_RULES_WLOCK(); 4928 error = pfr_set_tflags(pfrts, io->pfrio_size, 4929 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, 4930 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4931 PF_RULES_WUNLOCK(); 4932 free(pfrts, M_TEMP); 4933 break; 4934 } 4935 4936 case DIOCRCLRADDRS: { 4937 struct pfioc_table *io = (struct pfioc_table *)addr; 4938 4939 if (io->pfrio_esize != 0) { 4940 error = ENODEV; 4941 goto fail; 4942 } 4943 PF_RULES_WLOCK(); 4944 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, 4945 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4946 PF_RULES_WUNLOCK(); 4947 break; 4948 } 4949 4950 case DIOCRADDADDRS: { 4951 struct pfioc_table *io = (struct pfioc_table *)addr; 4952 struct pfr_addr *pfras; 4953 size_t totlen; 4954 4955 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4956 error = ENODEV; 4957 goto fail; 4958 } 4959 if (io->pfrio_size < 0 || 4960 io->pfrio_size > pf_ioctl_maxcount || 4961 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4962 error = EINVAL; 4963 goto fail; 4964 } 4965 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4966 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4967 M_TEMP, M_WAITOK); 4968 error = copyin(io->pfrio_buffer, pfras, totlen); 4969 if (error) { 4970 free(pfras, M_TEMP); 4971 goto fail; 4972 } 4973 PF_RULES_WLOCK(); 4974 error = pfr_add_addrs(&io->pfrio_table, pfras, 4975 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | 4976 PFR_FLAG_USERIOCTL); 4977 PF_RULES_WUNLOCK(); 4978 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4979 error = copyout(pfras, io->pfrio_buffer, totlen); 4980 free(pfras, M_TEMP); 4981 break; 4982 } 4983 4984 case DIOCRDELADDRS: { 4985 struct pfioc_table *io = (struct pfioc_table *)addr; 4986 struct pfr_addr *pfras; 4987 size_t totlen; 4988 4989 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4990 error = ENODEV; 4991 goto fail; 4992 } 4993 if (io->pfrio_size < 0 || 4994 io->pfrio_size > pf_ioctl_maxcount || 4995 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4996 error = EINVAL; 4997 goto fail; 4998 } 4999 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5000 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5001 M_TEMP, M_WAITOK); 5002 error = copyin(io->pfrio_buffer, pfras, totlen); 5003 if (error) { 5004 free(pfras, M_TEMP); 5005 goto fail; 5006 } 5007 PF_RULES_WLOCK(); 5008 error = pfr_del_addrs(&io->pfrio_table, pfras, 5009 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | 5010 PFR_FLAG_USERIOCTL); 5011 PF_RULES_WUNLOCK(); 5012 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5013 error = copyout(pfras, io->pfrio_buffer, totlen); 5014 free(pfras, M_TEMP); 5015 break; 5016 } 5017 5018 case DIOCRSETADDRS: { 5019 struct pfioc_table *io = (struct pfioc_table *)addr; 5020 struct pfr_addr *pfras; 5021 size_t totlen, count; 5022 5023 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5024 error = ENODEV; 5025 goto fail; 5026 } 5027 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { 5028 error = EINVAL; 5029 goto fail; 5030 } 5031 count = max(io->pfrio_size, io->pfrio_size2); 5032 if (count > pf_ioctl_maxcount || 5033 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { 5034 error = EINVAL; 5035 goto fail; 5036 } 5037 totlen = count * sizeof(struct pfr_addr); 5038 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP, 5039 M_WAITOK); 5040 error = copyin(io->pfrio_buffer, pfras, totlen); 5041 if (error) { 5042 free(pfras, M_TEMP); 5043 goto fail; 5044 } 5045 PF_RULES_WLOCK(); 5046 error = pfr_set_addrs(&io->pfrio_table, pfras, 5047 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, 5048 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | 5049 PFR_FLAG_USERIOCTL, 0); 5050 PF_RULES_WUNLOCK(); 5051 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5052 error = copyout(pfras, io->pfrio_buffer, totlen); 5053 free(pfras, M_TEMP); 5054 break; 5055 } 5056 5057 case DIOCRGETADDRS: { 5058 struct pfioc_table *io = (struct pfioc_table *)addr; 5059 struct pfr_addr *pfras; 5060 size_t totlen; 5061 5062 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5063 error = ENODEV; 5064 goto fail; 5065 } 5066 if (io->pfrio_size < 0 || 5067 io->pfrio_size > pf_ioctl_maxcount || 5068 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5069 error = EINVAL; 5070 goto fail; 5071 } 5072 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5073 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5074 M_TEMP, M_WAITOK | M_ZERO); 5075 PF_RULES_RLOCK(); 5076 error = pfr_get_addrs(&io->pfrio_table, pfras, 5077 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5078 PF_RULES_RUNLOCK(); 5079 if (error == 0) 5080 error = copyout(pfras, io->pfrio_buffer, totlen); 5081 free(pfras, M_TEMP); 5082 break; 5083 } 5084 5085 case DIOCRGETASTATS: { 5086 struct pfioc_table *io = (struct pfioc_table *)addr; 5087 struct pfr_astats *pfrastats; 5088 size_t totlen; 5089 5090 if (io->pfrio_esize != sizeof(struct pfr_astats)) { 5091 error = ENODEV; 5092 goto fail; 5093 } 5094 if (io->pfrio_size < 0 || 5095 io->pfrio_size > pf_ioctl_maxcount || 5096 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { 5097 error = EINVAL; 5098 goto fail; 5099 } 5100 totlen = io->pfrio_size * sizeof(struct pfr_astats); 5101 pfrastats = mallocarray(io->pfrio_size, 5102 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO); 5103 PF_RULES_RLOCK(); 5104 error = pfr_get_astats(&io->pfrio_table, pfrastats, 5105 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5106 PF_RULES_RUNLOCK(); 5107 if (error == 0) 5108 error = copyout(pfrastats, io->pfrio_buffer, totlen); 5109 free(pfrastats, M_TEMP); 5110 break; 5111 } 5112 5113 case DIOCRCLRASTATS: { 5114 struct pfioc_table *io = (struct pfioc_table *)addr; 5115 struct pfr_addr *pfras; 5116 size_t totlen; 5117 5118 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5119 error = ENODEV; 5120 goto fail; 5121 } 5122 if (io->pfrio_size < 0 || 5123 io->pfrio_size > pf_ioctl_maxcount || 5124 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5125 error = EINVAL; 5126 goto fail; 5127 } 5128 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5129 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5130 M_TEMP, M_WAITOK); 5131 error = copyin(io->pfrio_buffer, pfras, totlen); 5132 if (error) { 5133 free(pfras, M_TEMP); 5134 goto fail; 5135 } 5136 PF_RULES_WLOCK(); 5137 error = pfr_clr_astats(&io->pfrio_table, pfras, 5138 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | 5139 PFR_FLAG_USERIOCTL); 5140 PF_RULES_WUNLOCK(); 5141 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5142 error = copyout(pfras, io->pfrio_buffer, totlen); 5143 free(pfras, M_TEMP); 5144 break; 5145 } 5146 5147 case DIOCRTSTADDRS: { 5148 struct pfioc_table *io = (struct pfioc_table *)addr; 5149 struct pfr_addr *pfras; 5150 size_t totlen; 5151 5152 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5153 error = ENODEV; 5154 goto fail; 5155 } 5156 if (io->pfrio_size < 0 || 5157 io->pfrio_size > pf_ioctl_maxcount || 5158 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5159 error = EINVAL; 5160 goto fail; 5161 } 5162 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5163 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5164 M_TEMP, M_WAITOK); 5165 error = copyin(io->pfrio_buffer, pfras, totlen); 5166 if (error) { 5167 free(pfras, M_TEMP); 5168 goto fail; 5169 } 5170 PF_RULES_RLOCK(); 5171 error = pfr_tst_addrs(&io->pfrio_table, pfras, 5172 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | 5173 PFR_FLAG_USERIOCTL); 5174 PF_RULES_RUNLOCK(); 5175 if (error == 0) 5176 error = copyout(pfras, io->pfrio_buffer, totlen); 5177 free(pfras, M_TEMP); 5178 break; 5179 } 5180 5181 case DIOCRINADEFINE: { 5182 struct pfioc_table *io = (struct pfioc_table *)addr; 5183 struct pfr_addr *pfras; 5184 size_t totlen; 5185 5186 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5187 error = ENODEV; 5188 goto fail; 5189 } 5190 if (io->pfrio_size < 0 || 5191 io->pfrio_size > pf_ioctl_maxcount || 5192 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5193 error = EINVAL; 5194 goto fail; 5195 } 5196 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5197 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5198 M_TEMP, M_WAITOK); 5199 error = copyin(io->pfrio_buffer, pfras, totlen); 5200 if (error) { 5201 free(pfras, M_TEMP); 5202 goto fail; 5203 } 5204 PF_RULES_WLOCK(); 5205 error = pfr_ina_define(&io->pfrio_table, pfras, 5206 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, 5207 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5208 PF_RULES_WUNLOCK(); 5209 free(pfras, M_TEMP); 5210 break; 5211 } 5212 5213 case DIOCOSFPADD: { 5214 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5215 PF_RULES_WLOCK(); 5216 error = pf_osfp_add(io); 5217 PF_RULES_WUNLOCK(); 5218 break; 5219 } 5220 5221 case DIOCOSFPGET: { 5222 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5223 PF_RULES_RLOCK(); 5224 error = pf_osfp_get(io); 5225 PF_RULES_RUNLOCK(); 5226 break; 5227 } 5228 5229 case DIOCXBEGIN: { 5230 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5231 struct pfioc_trans_e *ioes, *ioe; 5232 size_t totlen; 5233 int i; 5234 5235 if (io->esize != sizeof(*ioe)) { 5236 error = ENODEV; 5237 goto fail; 5238 } 5239 if (io->size < 0 || 5240 io->size > pf_ioctl_maxcount || 5241 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5242 error = EINVAL; 5243 goto fail; 5244 } 5245 totlen = sizeof(struct pfioc_trans_e) * io->size; 5246 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5247 M_TEMP, M_WAITOK); 5248 error = copyin(io->array, ioes, totlen); 5249 if (error) { 5250 free(ioes, M_TEMP); 5251 goto fail; 5252 } 5253 PF_RULES_WLOCK(); 5254 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5255 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5256 switch (ioe->rs_num) { 5257 case PF_RULESET_ETH: 5258 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) { 5259 PF_RULES_WUNLOCK(); 5260 free(ioes, M_TEMP); 5261 goto fail; 5262 } 5263 break; 5264 #ifdef ALTQ 5265 case PF_RULESET_ALTQ: 5266 if (ioe->anchor[0]) { 5267 PF_RULES_WUNLOCK(); 5268 free(ioes, M_TEMP); 5269 error = EINVAL; 5270 goto fail; 5271 } 5272 if ((error = pf_begin_altq(&ioe->ticket))) { 5273 PF_RULES_WUNLOCK(); 5274 free(ioes, M_TEMP); 5275 goto fail; 5276 } 5277 break; 5278 #endif /* ALTQ */ 5279 case PF_RULESET_TABLE: 5280 { 5281 struct pfr_table table; 5282 5283 bzero(&table, sizeof(table)); 5284 strlcpy(table.pfrt_anchor, ioe->anchor, 5285 sizeof(table.pfrt_anchor)); 5286 if ((error = pfr_ina_begin(&table, 5287 &ioe->ticket, NULL, 0))) { 5288 PF_RULES_WUNLOCK(); 5289 free(ioes, M_TEMP); 5290 goto fail; 5291 } 5292 break; 5293 } 5294 default: 5295 if ((error = pf_begin_rules(&ioe->ticket, 5296 ioe->rs_num, ioe->anchor))) { 5297 PF_RULES_WUNLOCK(); 5298 free(ioes, M_TEMP); 5299 goto fail; 5300 } 5301 break; 5302 } 5303 } 5304 PF_RULES_WUNLOCK(); 5305 error = copyout(ioes, io->array, totlen); 5306 free(ioes, M_TEMP); 5307 break; 5308 } 5309 5310 case DIOCXROLLBACK: { 5311 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5312 struct pfioc_trans_e *ioe, *ioes; 5313 size_t totlen; 5314 int i; 5315 5316 if (io->esize != sizeof(*ioe)) { 5317 error = ENODEV; 5318 goto fail; 5319 } 5320 if (io->size < 0 || 5321 io->size > pf_ioctl_maxcount || 5322 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5323 error = EINVAL; 5324 goto fail; 5325 } 5326 totlen = sizeof(struct pfioc_trans_e) * io->size; 5327 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5328 M_TEMP, M_WAITOK); 5329 error = copyin(io->array, ioes, totlen); 5330 if (error) { 5331 free(ioes, M_TEMP); 5332 goto fail; 5333 } 5334 PF_RULES_WLOCK(); 5335 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5336 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5337 switch (ioe->rs_num) { 5338 case PF_RULESET_ETH: 5339 if ((error = pf_rollback_eth(ioe->ticket, 5340 ioe->anchor))) { 5341 PF_RULES_WUNLOCK(); 5342 free(ioes, M_TEMP); 5343 goto fail; /* really bad */ 5344 } 5345 break; 5346 #ifdef ALTQ 5347 case PF_RULESET_ALTQ: 5348 if (ioe->anchor[0]) { 5349 PF_RULES_WUNLOCK(); 5350 free(ioes, M_TEMP); 5351 error = EINVAL; 5352 goto fail; 5353 } 5354 if ((error = pf_rollback_altq(ioe->ticket))) { 5355 PF_RULES_WUNLOCK(); 5356 free(ioes, M_TEMP); 5357 goto fail; /* really bad */ 5358 } 5359 break; 5360 #endif /* ALTQ */ 5361 case PF_RULESET_TABLE: 5362 { 5363 struct pfr_table table; 5364 5365 bzero(&table, sizeof(table)); 5366 strlcpy(table.pfrt_anchor, ioe->anchor, 5367 sizeof(table.pfrt_anchor)); 5368 if ((error = pfr_ina_rollback(&table, 5369 ioe->ticket, NULL, 0))) { 5370 PF_RULES_WUNLOCK(); 5371 free(ioes, M_TEMP); 5372 goto fail; /* really bad */ 5373 } 5374 break; 5375 } 5376 default: 5377 if ((error = pf_rollback_rules(ioe->ticket, 5378 ioe->rs_num, ioe->anchor))) { 5379 PF_RULES_WUNLOCK(); 5380 free(ioes, M_TEMP); 5381 goto fail; /* really bad */ 5382 } 5383 break; 5384 } 5385 } 5386 PF_RULES_WUNLOCK(); 5387 free(ioes, M_TEMP); 5388 break; 5389 } 5390 5391 case DIOCXCOMMIT: { 5392 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5393 struct pfioc_trans_e *ioe, *ioes; 5394 struct pf_kruleset *rs; 5395 struct pf_keth_ruleset *ers; 5396 size_t totlen; 5397 int i; 5398 5399 if (io->esize != sizeof(*ioe)) { 5400 error = ENODEV; 5401 goto fail; 5402 } 5403 5404 if (io->size < 0 || 5405 io->size > pf_ioctl_maxcount || 5406 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5407 error = EINVAL; 5408 goto fail; 5409 } 5410 5411 totlen = sizeof(struct pfioc_trans_e) * io->size; 5412 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5413 M_TEMP, M_WAITOK); 5414 error = copyin(io->array, ioes, totlen); 5415 if (error) { 5416 free(ioes, M_TEMP); 5417 goto fail; 5418 } 5419 PF_RULES_WLOCK(); 5420 /* First makes sure everything will succeed. */ 5421 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5422 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5423 switch (ioe->rs_num) { 5424 case PF_RULESET_ETH: 5425 ers = pf_find_keth_ruleset(ioe->anchor); 5426 if (ers == NULL || ioe->ticket == 0 || 5427 ioe->ticket != ers->inactive.ticket) { 5428 PF_RULES_WUNLOCK(); 5429 free(ioes, M_TEMP); 5430 error = EINVAL; 5431 goto fail; 5432 } 5433 break; 5434 #ifdef ALTQ 5435 case PF_RULESET_ALTQ: 5436 if (ioe->anchor[0]) { 5437 PF_RULES_WUNLOCK(); 5438 free(ioes, M_TEMP); 5439 error = EINVAL; 5440 goto fail; 5441 } 5442 if (!V_altqs_inactive_open || ioe->ticket != 5443 V_ticket_altqs_inactive) { 5444 PF_RULES_WUNLOCK(); 5445 free(ioes, M_TEMP); 5446 error = EBUSY; 5447 goto fail; 5448 } 5449 break; 5450 #endif /* ALTQ */ 5451 case PF_RULESET_TABLE: 5452 rs = pf_find_kruleset(ioe->anchor); 5453 if (rs == NULL || !rs->topen || ioe->ticket != 5454 rs->tticket) { 5455 PF_RULES_WUNLOCK(); 5456 free(ioes, M_TEMP); 5457 error = EBUSY; 5458 goto fail; 5459 } 5460 break; 5461 default: 5462 if (ioe->rs_num < 0 || ioe->rs_num >= 5463 PF_RULESET_MAX) { 5464 PF_RULES_WUNLOCK(); 5465 free(ioes, M_TEMP); 5466 error = EINVAL; 5467 goto fail; 5468 } 5469 rs = pf_find_kruleset(ioe->anchor); 5470 if (rs == NULL || 5471 !rs->rules[ioe->rs_num].inactive.open || 5472 rs->rules[ioe->rs_num].inactive.ticket != 5473 ioe->ticket) { 5474 PF_RULES_WUNLOCK(); 5475 free(ioes, M_TEMP); 5476 error = EBUSY; 5477 goto fail; 5478 } 5479 break; 5480 } 5481 } 5482 /* Now do the commit - no errors should happen here. */ 5483 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5484 switch (ioe->rs_num) { 5485 case PF_RULESET_ETH: 5486 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) { 5487 PF_RULES_WUNLOCK(); 5488 free(ioes, M_TEMP); 5489 goto fail; /* really bad */ 5490 } 5491 break; 5492 #ifdef ALTQ 5493 case PF_RULESET_ALTQ: 5494 if ((error = pf_commit_altq(ioe->ticket))) { 5495 PF_RULES_WUNLOCK(); 5496 free(ioes, M_TEMP); 5497 goto fail; /* really bad */ 5498 } 5499 break; 5500 #endif /* ALTQ */ 5501 case PF_RULESET_TABLE: 5502 { 5503 struct pfr_table table; 5504 5505 bzero(&table, sizeof(table)); 5506 (void)strlcpy(table.pfrt_anchor, ioe->anchor, 5507 sizeof(table.pfrt_anchor)); 5508 if ((error = pfr_ina_commit(&table, 5509 ioe->ticket, NULL, NULL, 0))) { 5510 PF_RULES_WUNLOCK(); 5511 free(ioes, M_TEMP); 5512 goto fail; /* really bad */ 5513 } 5514 break; 5515 } 5516 default: 5517 if ((error = pf_commit_rules(ioe->ticket, 5518 ioe->rs_num, ioe->anchor))) { 5519 PF_RULES_WUNLOCK(); 5520 free(ioes, M_TEMP); 5521 goto fail; /* really bad */ 5522 } 5523 break; 5524 } 5525 } 5526 PF_RULES_WUNLOCK(); 5527 5528 /* Only hook into EtherNet taffic if we've got rules for it. */ 5529 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 5530 hook_pf_eth(); 5531 else 5532 dehook_pf_eth(); 5533 5534 free(ioes, M_TEMP); 5535 break; 5536 } 5537 5538 case DIOCGETSRCNODES: { 5539 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; 5540 struct pf_srchash *sh; 5541 struct pf_ksrc_node *n; 5542 struct pf_src_node *p, *pstore; 5543 uint32_t i, nr = 0; 5544 5545 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5546 i++, sh++) { 5547 PF_HASHROW_LOCK(sh); 5548 LIST_FOREACH(n, &sh->nodes, entry) 5549 nr++; 5550 PF_HASHROW_UNLOCK(sh); 5551 } 5552 5553 psn->psn_len = min(psn->psn_len, 5554 sizeof(struct pf_src_node) * nr); 5555 5556 if (psn->psn_len == 0) { 5557 psn->psn_len = sizeof(struct pf_src_node) * nr; 5558 goto fail; 5559 } 5560 5561 nr = 0; 5562 5563 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO); 5564 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5565 i++, sh++) { 5566 PF_HASHROW_LOCK(sh); 5567 LIST_FOREACH(n, &sh->nodes, entry) { 5568 5569 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) 5570 break; 5571 5572 pf_src_node_copy(n, p); 5573 5574 p++; 5575 nr++; 5576 } 5577 PF_HASHROW_UNLOCK(sh); 5578 } 5579 error = copyout(pstore, psn->psn_src_nodes, 5580 sizeof(struct pf_src_node) * nr); 5581 if (error) { 5582 free(pstore, M_TEMP); 5583 goto fail; 5584 } 5585 psn->psn_len = sizeof(struct pf_src_node) * nr; 5586 free(pstore, M_TEMP); 5587 break; 5588 } 5589 5590 case DIOCCLRSRCNODES: { 5591 pf_kill_srcnodes(NULL); 5592 break; 5593 } 5594 5595 case DIOCKILLSRCNODES: 5596 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); 5597 break; 5598 5599 #ifdef COMPAT_FREEBSD13 5600 case DIOCKEEPCOUNTERS_FREEBSD13: 5601 #endif 5602 case DIOCKEEPCOUNTERS: 5603 error = pf_keepcounters((struct pfioc_nv *)addr); 5604 break; 5605 5606 case DIOCGETSYNCOOKIES: 5607 error = pf_get_syncookies((struct pfioc_nv *)addr); 5608 break; 5609 5610 case DIOCSETSYNCOOKIES: 5611 error = pf_set_syncookies((struct pfioc_nv *)addr); 5612 break; 5613 5614 case DIOCSETHOSTID: { 5615 u_int32_t *hostid = (u_int32_t *)addr; 5616 5617 PF_RULES_WLOCK(); 5618 if (*hostid == 0) 5619 V_pf_status.hostid = arc4random(); 5620 else 5621 V_pf_status.hostid = *hostid; 5622 PF_RULES_WUNLOCK(); 5623 break; 5624 } 5625 5626 case DIOCOSFPFLUSH: 5627 PF_RULES_WLOCK(); 5628 pf_osfp_flush(); 5629 PF_RULES_WUNLOCK(); 5630 break; 5631 5632 case DIOCIGETIFACES: { 5633 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5634 struct pfi_kif *ifstore; 5635 size_t bufsiz; 5636 5637 if (io->pfiio_esize != sizeof(struct pfi_kif)) { 5638 error = ENODEV; 5639 goto fail; 5640 } 5641 5642 if (io->pfiio_size < 0 || 5643 io->pfiio_size > pf_ioctl_maxcount || 5644 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { 5645 error = EINVAL; 5646 goto fail; 5647 } 5648 5649 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5650 5651 bufsiz = io->pfiio_size * sizeof(struct pfi_kif); 5652 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif), 5653 M_TEMP, M_WAITOK | M_ZERO); 5654 5655 PF_RULES_RLOCK(); 5656 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); 5657 PF_RULES_RUNLOCK(); 5658 error = copyout(ifstore, io->pfiio_buffer, bufsiz); 5659 free(ifstore, M_TEMP); 5660 break; 5661 } 5662 5663 case DIOCSETIFFLAG: { 5664 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5665 5666 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5667 5668 PF_RULES_WLOCK(); 5669 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags); 5670 PF_RULES_WUNLOCK(); 5671 break; 5672 } 5673 5674 case DIOCCLRIFFLAG: { 5675 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5676 5677 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5678 5679 PF_RULES_WLOCK(); 5680 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); 5681 PF_RULES_WUNLOCK(); 5682 break; 5683 } 5684 5685 case DIOCSETREASS: { 5686 u_int32_t *reass = (u_int32_t *)addr; 5687 5688 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF); 5689 /* Removal of DF flag without reassembly enabled is not a 5690 * valid combination. Disable reassembly in such case. */ 5691 if (!(V_pf_status.reass & PF_REASS_ENABLED)) 5692 V_pf_status.reass = 0; 5693 break; 5694 } 5695 5696 default: 5697 error = ENODEV; 5698 break; 5699 } 5700 fail: 5701 CURVNET_RESTORE(); 5702 5703 #undef ERROUT_IOCTL 5704 5705 return (error); 5706 } 5707 5708 void 5709 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version) 5710 { 5711 bzero(sp, sizeof(union pfsync_state_union)); 5712 5713 /* copy from state key */ 5714 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5715 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5716 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5717 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5718 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5719 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5720 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5721 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5722 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto; 5723 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af; 5724 5725 /* copy from state */ 5726 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname)); 5727 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr)); 5728 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000)); 5729 sp->pfs_1301.expire = pf_state_expires(st); 5730 if (sp->pfs_1301.expire <= time_uptime) 5731 sp->pfs_1301.expire = htonl(0); 5732 else 5733 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime); 5734 5735 sp->pfs_1301.direction = st->direction; 5736 sp->pfs_1301.log = st->act.log; 5737 sp->pfs_1301.timeout = st->timeout; 5738 5739 switch (msg_version) { 5740 case PFSYNC_MSG_VERSION_1301: 5741 sp->pfs_1301.state_flags = st->state_flags; 5742 break; 5743 case PFSYNC_MSG_VERSION_1400: 5744 sp->pfs_1400.state_flags = htons(st->state_flags); 5745 sp->pfs_1400.qid = htons(st->act.qid); 5746 sp->pfs_1400.pqid = htons(st->act.pqid); 5747 sp->pfs_1400.dnpipe = htons(st->act.dnpipe); 5748 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe); 5749 sp->pfs_1400.rtableid = htonl(st->act.rtableid); 5750 sp->pfs_1400.min_ttl = st->act.min_ttl; 5751 sp->pfs_1400.set_tos = st->act.set_tos; 5752 sp->pfs_1400.max_mss = htons(st->act.max_mss); 5753 sp->pfs_1400.set_prio[0] = st->act.set_prio[0]; 5754 sp->pfs_1400.set_prio[1] = st->act.set_prio[1]; 5755 sp->pfs_1400.rt = st->act.rt; 5756 if (st->act.rt_kif) 5757 strlcpy(sp->pfs_1400.rt_ifname, 5758 st->act.rt_kif->pfik_name, 5759 sizeof(sp->pfs_1400.rt_ifname)); 5760 break; 5761 default: 5762 panic("%s: Unsupported pfsync_msg_version %d", 5763 __func__, msg_version); 5764 } 5765 5766 /* 5767 * XXX Why do we bother pfsyncing source node information if source 5768 * nodes are not synced? Showing users that there is source tracking 5769 * when there is none seems useless. 5770 */ 5771 if (st->sns[PF_SN_LIMIT] != NULL) 5772 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE; 5773 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE]) 5774 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5775 5776 sp->pfs_1301.id = st->id; 5777 sp->pfs_1301.creatorid = st->creatorid; 5778 pf_state_peer_hton(&st->src, &sp->pfs_1301.src); 5779 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst); 5780 5781 if (st->rule == NULL) 5782 sp->pfs_1301.rule = htonl(-1); 5783 else 5784 sp->pfs_1301.rule = htonl(st->rule->nr); 5785 if (st->anchor == NULL) 5786 sp->pfs_1301.anchor = htonl(-1); 5787 else 5788 sp->pfs_1301.anchor = htonl(st->anchor->nr); 5789 if (st->nat_rule == NULL) 5790 sp->pfs_1301.nat_rule = htonl(-1); 5791 else 5792 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr); 5793 5794 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]); 5795 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]); 5796 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]); 5797 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]); 5798 } 5799 5800 void 5801 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st) 5802 { 5803 bzero(sp, sizeof(*sp)); 5804 5805 sp->version = PF_STATE_VERSION; 5806 5807 /* copy from state key */ 5808 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5809 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5810 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5811 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5812 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5813 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5814 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5815 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5816 sp->proto = st->key[PF_SK_WIRE]->proto; 5817 sp->af = st->key[PF_SK_WIRE]->af; 5818 5819 /* copy from state */ 5820 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5821 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name, 5822 sizeof(sp->orig_ifname)); 5823 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr)); 5824 sp->creation = htonl(time_uptime - (st->creation / 1000)); 5825 sp->expire = pf_state_expires(st); 5826 if (sp->expire <= time_uptime) 5827 sp->expire = htonl(0); 5828 else 5829 sp->expire = htonl(sp->expire - time_uptime); 5830 5831 sp->direction = st->direction; 5832 sp->log = st->act.log; 5833 sp->timeout = st->timeout; 5834 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */ 5835 sp->state_flags_compat = st->state_flags; 5836 sp->state_flags = htons(st->state_flags); 5837 if (st->sns[PF_SN_LIMIT] != NULL) 5838 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5839 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL) 5840 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5841 sp->id = st->id; 5842 sp->creatorid = st->creatorid; 5843 pf_state_peer_hton(&st->src, &sp->src); 5844 pf_state_peer_hton(&st->dst, &sp->dst); 5845 5846 if (st->rule == NULL) 5847 sp->rule = htonl(-1); 5848 else 5849 sp->rule = htonl(st->rule->nr); 5850 if (st->anchor == NULL) 5851 sp->anchor = htonl(-1); 5852 else 5853 sp->anchor = htonl(st->anchor->nr); 5854 if (st->nat_rule == NULL) 5855 sp->nat_rule = htonl(-1); 5856 else 5857 sp->nat_rule = htonl(st->nat_rule->nr); 5858 5859 sp->packets[0] = st->packets[0]; 5860 sp->packets[1] = st->packets[1]; 5861 sp->bytes[0] = st->bytes[0]; 5862 sp->bytes[1] = st->bytes[1]; 5863 5864 sp->qid = htons(st->act.qid); 5865 sp->pqid = htons(st->act.pqid); 5866 sp->dnpipe = htons(st->act.dnpipe); 5867 sp->dnrpipe = htons(st->act.dnrpipe); 5868 sp->rtableid = htonl(st->act.rtableid); 5869 sp->min_ttl = st->act.min_ttl; 5870 sp->set_tos = st->act.set_tos; 5871 sp->max_mss = htons(st->act.max_mss); 5872 sp->rt = st->act.rt; 5873 if (st->act.rt_kif) 5874 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name, 5875 sizeof(sp->rt_ifname)); 5876 sp->set_prio[0] = st->act.set_prio[0]; 5877 sp->set_prio[1] = st->act.set_prio[1]; 5878 5879 } 5880 5881 static void 5882 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 5883 { 5884 struct pfr_ktable *kt; 5885 5886 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); 5887 5888 kt = aw->p.tbl; 5889 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 5890 kt = kt->pfrkt_root; 5891 aw->p.tbl = NULL; 5892 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 5893 kt->pfrkt_cnt : -1; 5894 } 5895 5896 static int 5897 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters, 5898 size_t number, char **names) 5899 { 5900 nvlist_t *nvc; 5901 5902 nvc = nvlist_create(0); 5903 if (nvc == NULL) 5904 return (ENOMEM); 5905 5906 for (int i = 0; i < number; i++) { 5907 nvlist_append_number_array(nvc, "counters", 5908 counter_u64_fetch(counters[i])); 5909 nvlist_append_string_array(nvc, "names", 5910 names[i]); 5911 nvlist_append_number_array(nvc, "ids", 5912 i); 5913 } 5914 nvlist_add_nvlist(nvl, name, nvc); 5915 nvlist_destroy(nvc); 5916 5917 return (0); 5918 } 5919 5920 static int 5921 pf_getstatus(struct pfioc_nv *nv) 5922 { 5923 nvlist_t *nvl = NULL, *nvc = NULL; 5924 void *nvlpacked = NULL; 5925 int error; 5926 struct pf_status s; 5927 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES; 5928 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES; 5929 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES; 5930 time_t since; 5931 5932 PF_RULES_RLOCK_TRACKER; 5933 5934 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 5935 5936 PF_RULES_RLOCK(); 5937 5938 nvl = nvlist_create(0); 5939 if (nvl == NULL) 5940 ERROUT(ENOMEM); 5941 5942 since = time_second - (time_uptime - V_pf_status.since); 5943 5944 nvlist_add_bool(nvl, "running", V_pf_status.running); 5945 nvlist_add_number(nvl, "since", since); 5946 nvlist_add_number(nvl, "debug", V_pf_status.debug); 5947 nvlist_add_number(nvl, "hostid", V_pf_status.hostid); 5948 nvlist_add_number(nvl, "states", V_pf_status.states); 5949 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes); 5950 nvlist_add_number(nvl, "reass", V_pf_status.reass); 5951 nvlist_add_bool(nvl, "syncookies_active", 5952 V_pf_status.syncookies_active); 5953 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen); 5954 5955 /* counters */ 5956 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters, 5957 PFRES_MAX, pf_reasons); 5958 if (error != 0) 5959 ERROUT(error); 5960 5961 /* lcounters */ 5962 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters, 5963 KLCNT_MAX, pf_lcounter); 5964 if (error != 0) 5965 ERROUT(error); 5966 5967 /* fcounters */ 5968 nvc = nvlist_create(0); 5969 if (nvc == NULL) 5970 ERROUT(ENOMEM); 5971 5972 for (int i = 0; i < FCNT_MAX; i++) { 5973 nvlist_append_number_array(nvc, "counters", 5974 pf_counter_u64_fetch(&V_pf_status.fcounters[i])); 5975 nvlist_append_string_array(nvc, "names", 5976 pf_fcounter[i]); 5977 nvlist_append_number_array(nvc, "ids", 5978 i); 5979 } 5980 nvlist_add_nvlist(nvl, "fcounters", nvc); 5981 nvlist_destroy(nvc); 5982 nvc = NULL; 5983 5984 /* scounters */ 5985 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters, 5986 SCNT_MAX, pf_fcounter); 5987 if (error != 0) 5988 ERROUT(error); 5989 5990 nvlist_add_string(nvl, "ifname", V_pf_status.ifname); 5991 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum, 5992 PF_MD5_DIGEST_LENGTH); 5993 5994 pfi_update_status(V_pf_status.ifname, &s); 5995 5996 /* pcounters / bcounters */ 5997 for (int i = 0; i < 2; i++) { 5998 for (int j = 0; j < 2; j++) { 5999 for (int k = 0; k < 2; k++) { 6000 nvlist_append_number_array(nvl, "pcounters", 6001 s.pcounters[i][j][k]); 6002 } 6003 nvlist_append_number_array(nvl, "bcounters", 6004 s.bcounters[i][j]); 6005 } 6006 } 6007 6008 nvlpacked = nvlist_pack(nvl, &nv->len); 6009 if (nvlpacked == NULL) 6010 ERROUT(ENOMEM); 6011 6012 if (nv->size == 0) 6013 ERROUT(0); 6014 else if (nv->size < nv->len) 6015 ERROUT(ENOSPC); 6016 6017 PF_RULES_RUNLOCK(); 6018 error = copyout(nvlpacked, nv->data, nv->len); 6019 goto done; 6020 6021 #undef ERROUT 6022 errout: 6023 PF_RULES_RUNLOCK(); 6024 done: 6025 free(nvlpacked, M_NVLIST); 6026 nvlist_destroy(nvc); 6027 nvlist_destroy(nvl); 6028 6029 return (error); 6030 } 6031 6032 /* 6033 * XXX - Check for version mismatch!!! 6034 */ 6035 static void 6036 pf_clear_all_states(void) 6037 { 6038 struct epoch_tracker et; 6039 struct pf_kstate *s; 6040 u_int i; 6041 6042 NET_EPOCH_ENTER(et); 6043 for (i = 0; i <= V_pf_hashmask; i++) { 6044 struct pf_idhash *ih = &V_pf_idhash[i]; 6045 relock: 6046 PF_HASHROW_LOCK(ih); 6047 LIST_FOREACH(s, &ih->states, entry) { 6048 s->timeout = PFTM_PURGE; 6049 /* Don't send out individual delete messages. */ 6050 s->state_flags |= PFSTATE_NOSYNC; 6051 pf_remove_state(s); 6052 goto relock; 6053 } 6054 PF_HASHROW_UNLOCK(ih); 6055 } 6056 NET_EPOCH_EXIT(et); 6057 } 6058 6059 static int 6060 pf_clear_tables(void) 6061 { 6062 struct pfioc_table io; 6063 int error; 6064 6065 bzero(&io, sizeof(io)); 6066 io.pfrio_flags |= PFR_FLAG_ALLRSETS; 6067 6068 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, 6069 io.pfrio_flags); 6070 6071 return (error); 6072 } 6073 6074 static void 6075 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) 6076 { 6077 struct pf_ksrc_node_list kill; 6078 u_int killed; 6079 6080 LIST_INIT(&kill); 6081 for (int i = 0; i <= V_pf_srchashmask; i++) { 6082 struct pf_srchash *sh = &V_pf_srchash[i]; 6083 struct pf_ksrc_node *sn, *tmp; 6084 6085 PF_HASHROW_LOCK(sh); 6086 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) 6087 if (psnk == NULL || 6088 (pf_match_addr(psnk->psnk_src.neg, 6089 &psnk->psnk_src.addr.v.a.addr, 6090 &psnk->psnk_src.addr.v.a.mask, 6091 &sn->addr, sn->af) && 6092 pf_match_addr(psnk->psnk_dst.neg, 6093 &psnk->psnk_dst.addr.v.a.addr, 6094 &psnk->psnk_dst.addr.v.a.mask, 6095 &sn->raddr, sn->af))) { 6096 pf_unlink_src_node(sn); 6097 LIST_INSERT_HEAD(&kill, sn, entry); 6098 sn->expire = 1; 6099 } 6100 PF_HASHROW_UNLOCK(sh); 6101 } 6102 6103 for (int i = 0; i <= V_pf_hashmask; i++) { 6104 struct pf_idhash *ih = &V_pf_idhash[i]; 6105 struct pf_kstate *s; 6106 6107 PF_HASHROW_LOCK(ih); 6108 LIST_FOREACH(s, &ih->states, entry) { 6109 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; 6110 sn_type++) { 6111 if (s->sns[sn_type] && 6112 s->sns[sn_type]->expire == 1) { 6113 s->sns[sn_type] = NULL; 6114 } 6115 } 6116 } 6117 PF_HASHROW_UNLOCK(ih); 6118 } 6119 6120 killed = pf_free_src_nodes(&kill); 6121 6122 if (psnk != NULL) 6123 psnk->psnk_killed = killed; 6124 } 6125 6126 static int 6127 pf_keepcounters(struct pfioc_nv *nv) 6128 { 6129 nvlist_t *nvl = NULL; 6130 void *nvlpacked = NULL; 6131 int error = 0; 6132 6133 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6134 6135 if (nv->len > pf_ioctl_maxcount) 6136 ERROUT(ENOMEM); 6137 6138 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6139 error = copyin(nv->data, nvlpacked, nv->len); 6140 if (error) 6141 ERROUT(error); 6142 6143 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6144 if (nvl == NULL) 6145 ERROUT(EBADMSG); 6146 6147 if (! nvlist_exists_bool(nvl, "keep_counters")) 6148 ERROUT(EBADMSG); 6149 6150 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters"); 6151 6152 on_error: 6153 nvlist_destroy(nvl); 6154 free(nvlpacked, M_NVLIST); 6155 return (error); 6156 } 6157 6158 unsigned int 6159 pf_clear_states(const struct pf_kstate_kill *kill) 6160 { 6161 struct pf_state_key_cmp match_key; 6162 struct pf_kstate *s; 6163 struct pfi_kkif *kif; 6164 int idx; 6165 unsigned int killed = 0, dir; 6166 6167 NET_EPOCH_ASSERT(); 6168 6169 for (unsigned int i = 0; i <= V_pf_hashmask; i++) { 6170 struct pf_idhash *ih = &V_pf_idhash[i]; 6171 6172 relock_DIOCCLRSTATES: 6173 PF_HASHROW_LOCK(ih); 6174 LIST_FOREACH(s, &ih->states, entry) { 6175 /* For floating states look at the original kif. */ 6176 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 6177 6178 if (kill->psk_ifname[0] && 6179 strcmp(kill->psk_ifname, 6180 kif->pfik_name)) 6181 continue; 6182 6183 if (kill->psk_kill_match) { 6184 bzero(&match_key, sizeof(match_key)); 6185 6186 if (s->direction == PF_OUT) { 6187 dir = PF_IN; 6188 idx = PF_SK_STACK; 6189 } else { 6190 dir = PF_OUT; 6191 idx = PF_SK_WIRE; 6192 } 6193 6194 match_key.af = s->key[idx]->af; 6195 match_key.proto = s->key[idx]->proto; 6196 pf_addrcpy(&match_key.addr[0], 6197 &s->key[idx]->addr[1], match_key.af); 6198 match_key.port[0] = s->key[idx]->port[1]; 6199 pf_addrcpy(&match_key.addr[1], 6200 &s->key[idx]->addr[0], match_key.af); 6201 match_key.port[1] = s->key[idx]->port[0]; 6202 } 6203 6204 /* 6205 * Don't send out individual 6206 * delete messages. 6207 */ 6208 s->state_flags |= PFSTATE_NOSYNC; 6209 pf_remove_state(s); 6210 killed++; 6211 6212 if (kill->psk_kill_match) 6213 killed += pf_kill_matching_state(&match_key, 6214 dir); 6215 6216 goto relock_DIOCCLRSTATES; 6217 } 6218 PF_HASHROW_UNLOCK(ih); 6219 } 6220 6221 if (V_pfsync_clear_states_ptr != NULL) 6222 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname); 6223 6224 return (killed); 6225 } 6226 6227 void 6228 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed) 6229 { 6230 struct pf_kstate *s; 6231 6232 NET_EPOCH_ASSERT(); 6233 if (kill->psk_pfcmp.id) { 6234 if (kill->psk_pfcmp.creatorid == 0) 6235 kill->psk_pfcmp.creatorid = V_pf_status.hostid; 6236 if ((s = pf_find_state_byid(kill->psk_pfcmp.id, 6237 kill->psk_pfcmp.creatorid))) { 6238 pf_remove_state(s); 6239 *killed = 1; 6240 } 6241 return; 6242 } 6243 6244 for (unsigned int i = 0; i <= V_pf_hashmask; i++) 6245 *killed += pf_killstates_row(kill, &V_pf_idhash[i]); 6246 } 6247 6248 static int 6249 pf_killstates_nv(struct pfioc_nv *nv) 6250 { 6251 struct pf_kstate_kill kill; 6252 struct epoch_tracker et; 6253 nvlist_t *nvl = NULL; 6254 void *nvlpacked = NULL; 6255 int error = 0; 6256 unsigned int killed = 0; 6257 6258 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6259 6260 if (nv->len > pf_ioctl_maxcount) 6261 ERROUT(ENOMEM); 6262 6263 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6264 error = copyin(nv->data, nvlpacked, nv->len); 6265 if (error) 6266 ERROUT(error); 6267 6268 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6269 if (nvl == NULL) 6270 ERROUT(EBADMSG); 6271 6272 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6273 if (error) 6274 ERROUT(error); 6275 6276 NET_EPOCH_ENTER(et); 6277 pf_killstates(&kill, &killed); 6278 NET_EPOCH_EXIT(et); 6279 6280 free(nvlpacked, M_NVLIST); 6281 nvlpacked = NULL; 6282 nvlist_destroy(nvl); 6283 nvl = nvlist_create(0); 6284 if (nvl == NULL) 6285 ERROUT(ENOMEM); 6286 6287 nvlist_add_number(nvl, "killed", killed); 6288 6289 nvlpacked = nvlist_pack(nvl, &nv->len); 6290 if (nvlpacked == NULL) 6291 ERROUT(ENOMEM); 6292 6293 if (nv->size == 0) 6294 ERROUT(0); 6295 else if (nv->size < nv->len) 6296 ERROUT(ENOSPC); 6297 6298 error = copyout(nvlpacked, nv->data, nv->len); 6299 6300 on_error: 6301 nvlist_destroy(nvl); 6302 free(nvlpacked, M_NVLIST); 6303 return (error); 6304 } 6305 6306 static int 6307 pf_clearstates_nv(struct pfioc_nv *nv) 6308 { 6309 struct pf_kstate_kill kill; 6310 struct epoch_tracker et; 6311 nvlist_t *nvl = NULL; 6312 void *nvlpacked = NULL; 6313 int error = 0; 6314 unsigned int killed; 6315 6316 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6317 6318 if (nv->len > pf_ioctl_maxcount) 6319 ERROUT(ENOMEM); 6320 6321 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6322 error = copyin(nv->data, nvlpacked, nv->len); 6323 if (error) 6324 ERROUT(error); 6325 6326 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6327 if (nvl == NULL) 6328 ERROUT(EBADMSG); 6329 6330 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6331 if (error) 6332 ERROUT(error); 6333 6334 NET_EPOCH_ENTER(et); 6335 killed = pf_clear_states(&kill); 6336 NET_EPOCH_EXIT(et); 6337 6338 free(nvlpacked, M_NVLIST); 6339 nvlpacked = NULL; 6340 nvlist_destroy(nvl); 6341 nvl = nvlist_create(0); 6342 if (nvl == NULL) 6343 ERROUT(ENOMEM); 6344 6345 nvlist_add_number(nvl, "killed", killed); 6346 6347 nvlpacked = nvlist_pack(nvl, &nv->len); 6348 if (nvlpacked == NULL) 6349 ERROUT(ENOMEM); 6350 6351 if (nv->size == 0) 6352 ERROUT(0); 6353 else if (nv->size < nv->len) 6354 ERROUT(ENOSPC); 6355 6356 error = copyout(nvlpacked, nv->data, nv->len); 6357 6358 #undef ERROUT 6359 on_error: 6360 nvlist_destroy(nvl); 6361 free(nvlpacked, M_NVLIST); 6362 return (error); 6363 } 6364 6365 static int 6366 pf_getstate(struct pfioc_nv *nv) 6367 { 6368 nvlist_t *nvl = NULL, *nvls; 6369 void *nvlpacked = NULL; 6370 struct pf_kstate *s = NULL; 6371 int error = 0; 6372 uint64_t id, creatorid; 6373 6374 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6375 6376 if (nv->len > pf_ioctl_maxcount) 6377 ERROUT(ENOMEM); 6378 6379 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6380 error = copyin(nv->data, nvlpacked, nv->len); 6381 if (error) 6382 ERROUT(error); 6383 6384 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6385 if (nvl == NULL) 6386 ERROUT(EBADMSG); 6387 6388 PFNV_CHK(pf_nvuint64(nvl, "id", &id)); 6389 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid)); 6390 6391 s = pf_find_state_byid(id, creatorid); 6392 if (s == NULL) 6393 ERROUT(ENOENT); 6394 6395 free(nvlpacked, M_NVLIST); 6396 nvlpacked = NULL; 6397 nvlist_destroy(nvl); 6398 nvl = nvlist_create(0); 6399 if (nvl == NULL) 6400 ERROUT(ENOMEM); 6401 6402 nvls = pf_state_to_nvstate(s); 6403 if (nvls == NULL) 6404 ERROUT(ENOMEM); 6405 6406 nvlist_add_nvlist(nvl, "state", nvls); 6407 nvlist_destroy(nvls); 6408 6409 nvlpacked = nvlist_pack(nvl, &nv->len); 6410 if (nvlpacked == NULL) 6411 ERROUT(ENOMEM); 6412 6413 if (nv->size == 0) 6414 ERROUT(0); 6415 else if (nv->size < nv->len) 6416 ERROUT(ENOSPC); 6417 6418 error = copyout(nvlpacked, nv->data, nv->len); 6419 6420 #undef ERROUT 6421 errout: 6422 if (s != NULL) 6423 PF_STATE_UNLOCK(s); 6424 free(nvlpacked, M_NVLIST); 6425 nvlist_destroy(nvl); 6426 return (error); 6427 } 6428 6429 /* 6430 * XXX - Check for version mismatch!!! 6431 */ 6432 6433 /* 6434 * Duplicate pfctl -Fa operation to get rid of as much as we can. 6435 */ 6436 static int 6437 shutdown_pf(void) 6438 { 6439 int error = 0; 6440 u_int32_t t[5]; 6441 char nn = '\0'; 6442 struct pf_kanchor *anchor; 6443 struct pf_keth_anchor *eth_anchor; 6444 int rs_num; 6445 6446 do { 6447 /* Unlink rules of all user defined anchors */ 6448 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) { 6449 /* Wildcard based anchors may not have a respective 6450 * explicit anchor rule or they may be left empty 6451 * without rules. It leads to anchor.refcnt=0, and the 6452 * rest of the logic does not expect it. */ 6453 if (anchor->refcnt == 0) 6454 anchor->refcnt = 1; 6455 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6456 if ((error = pf_begin_rules(&t[rs_num], rs_num, 6457 anchor->path)) != 0) { 6458 DPFPRINTF(PF_DEBUG_MISC, "%s: " 6459 "anchor.path=%s rs_num=%d", 6460 __func__, anchor->path, rs_num); 6461 goto error; /* XXX: rollback? */ 6462 } 6463 } 6464 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6465 error = pf_commit_rules(t[rs_num], rs_num, 6466 anchor->path); 6467 MPASS(error == 0); 6468 } 6469 } 6470 6471 /* Unlink rules of all user defined ether anchors */ 6472 RB_FOREACH(eth_anchor, pf_keth_anchor_global, 6473 &V_pf_keth_anchors) { 6474 /* Wildcard based anchors may not have a respective 6475 * explicit anchor rule or they may be left empty 6476 * without rules. It leads to anchor.refcnt=0, and the 6477 * rest of the logic does not expect it. */ 6478 if (eth_anchor->refcnt == 0) 6479 eth_anchor->refcnt = 1; 6480 if ((error = pf_begin_eth(&t[0], eth_anchor->path)) 6481 != 0) { 6482 DPFPRINTF(PF_DEBUG_MISC, "%s: eth " 6483 "anchor.path=%s", __func__, 6484 eth_anchor->path); 6485 goto error; 6486 } 6487 error = pf_commit_eth(t[0], eth_anchor->path); 6488 MPASS(error == 0); 6489 } 6490 6491 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) 6492 != 0) { 6493 DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__); 6494 break; 6495 } 6496 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) 6497 != 0) { 6498 DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__); 6499 break; /* XXX: rollback? */ 6500 } 6501 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) 6502 != 0) { 6503 DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__); 6504 break; /* XXX: rollback? */ 6505 } 6506 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) 6507 != 0) { 6508 DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__); 6509 break; /* XXX: rollback? */ 6510 } 6511 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) 6512 != 0) { 6513 DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__); 6514 break; /* XXX: rollback? */ 6515 } 6516 6517 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn); 6518 MPASS(error == 0); 6519 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn); 6520 MPASS(error == 0); 6521 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn); 6522 MPASS(error == 0); 6523 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn); 6524 MPASS(error == 0); 6525 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn); 6526 MPASS(error == 0); 6527 6528 if ((error = pf_clear_tables()) != 0) 6529 break; 6530 6531 if ((error = pf_begin_eth(&t[0], &nn)) != 0) { 6532 DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__); 6533 break; 6534 } 6535 error = pf_commit_eth(t[0], &nn); 6536 MPASS(error == 0); 6537 6538 #ifdef ALTQ 6539 if ((error = pf_begin_altq(&t[0])) != 0) { 6540 DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__); 6541 break; 6542 } 6543 pf_commit_altq(t[0]); 6544 #endif 6545 6546 pf_clear_all_states(); 6547 6548 pf_kill_srcnodes(NULL); 6549 6550 /* status does not use malloced mem so no need to cleanup */ 6551 /* fingerprints and interfaces have their own cleanup code */ 6552 } while(0); 6553 6554 error: 6555 return (error); 6556 } 6557 6558 static pfil_return_t 6559 pf_check_return(int chk, struct mbuf **m) 6560 { 6561 6562 switch (chk) { 6563 case PF_PASS: 6564 if (*m == NULL) 6565 return (PFIL_CONSUMED); 6566 else 6567 return (PFIL_PASS); 6568 break; 6569 default: 6570 if (*m != NULL) { 6571 m_freem(*m); 6572 *m = NULL; 6573 } 6574 return (PFIL_DROPPED); 6575 } 6576 } 6577 6578 static pfil_return_t 6579 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6580 void *ruleset __unused, struct inpcb *inp) 6581 { 6582 int chk; 6583 6584 CURVNET_ASSERT_SET(); 6585 6586 chk = pf_test_eth(PF_IN, flags, ifp, m, inp); 6587 6588 return (pf_check_return(chk, m)); 6589 } 6590 6591 static pfil_return_t 6592 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6593 void *ruleset __unused, struct inpcb *inp) 6594 { 6595 int chk; 6596 6597 CURVNET_ASSERT_SET(); 6598 6599 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp); 6600 6601 return (pf_check_return(chk, m)); 6602 } 6603 6604 #ifdef INET 6605 static pfil_return_t 6606 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6607 void *ruleset __unused, struct inpcb *inp) 6608 { 6609 int chk; 6610 6611 CURVNET_ASSERT_SET(); 6612 6613 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL); 6614 6615 return (pf_check_return(chk, m)); 6616 } 6617 6618 static pfil_return_t 6619 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6620 void *ruleset __unused, struct inpcb *inp) 6621 { 6622 int chk; 6623 6624 CURVNET_ASSERT_SET(); 6625 6626 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL); 6627 6628 return (pf_check_return(chk, m)); 6629 } 6630 #endif 6631 6632 #ifdef INET6 6633 static pfil_return_t 6634 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags, 6635 void *ruleset __unused, struct inpcb *inp) 6636 { 6637 int chk; 6638 6639 CURVNET_ASSERT_SET(); 6640 6641 /* 6642 * In case of loopback traffic IPv6 uses the real interface in 6643 * order to support scoped addresses. In order to support stateful 6644 * filtering we have change this to lo0 as it is the case in IPv4. 6645 */ 6646 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, 6647 m, inp, NULL); 6648 6649 return (pf_check_return(chk, m)); 6650 } 6651 6652 static pfil_return_t 6653 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags, 6654 void *ruleset __unused, struct inpcb *inp) 6655 { 6656 int chk; 6657 6658 CURVNET_ASSERT_SET(); 6659 6660 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL); 6661 6662 return (pf_check_return(chk, m)); 6663 } 6664 #endif /* INET6 */ 6665 6666 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook); 6667 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook); 6668 #define V_pf_eth_in_hook VNET(pf_eth_in_hook) 6669 #define V_pf_eth_out_hook VNET(pf_eth_out_hook) 6670 6671 #ifdef INET 6672 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook); 6673 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook); 6674 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook) 6675 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook) 6676 #endif 6677 #ifdef INET6 6678 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook); 6679 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook); 6680 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook) 6681 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook) 6682 #endif 6683 6684 static void 6685 hook_pf_eth(void) 6686 { 6687 struct pfil_hook_args pha = { 6688 .pa_version = PFIL_VERSION, 6689 .pa_modname = "pf", 6690 .pa_type = PFIL_TYPE_ETHERNET, 6691 }; 6692 struct pfil_link_args pla = { 6693 .pa_version = PFIL_VERSION, 6694 }; 6695 int ret __diagused; 6696 6697 if (atomic_load_bool(&V_pf_pfil_eth_hooked)) 6698 return; 6699 6700 pha.pa_mbuf_chk = pf_eth_check_in; 6701 pha.pa_flags = PFIL_IN; 6702 pha.pa_rulname = "eth-in"; 6703 V_pf_eth_in_hook = pfil_add_hook(&pha); 6704 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6705 pla.pa_head = V_link_pfil_head; 6706 pla.pa_hook = V_pf_eth_in_hook; 6707 ret = pfil_link(&pla); 6708 MPASS(ret == 0); 6709 pha.pa_mbuf_chk = pf_eth_check_out; 6710 pha.pa_flags = PFIL_OUT; 6711 pha.pa_rulname = "eth-out"; 6712 V_pf_eth_out_hook = pfil_add_hook(&pha); 6713 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6714 pla.pa_head = V_link_pfil_head; 6715 pla.pa_hook = V_pf_eth_out_hook; 6716 ret = pfil_link(&pla); 6717 MPASS(ret == 0); 6718 6719 atomic_store_bool(&V_pf_pfil_eth_hooked, true); 6720 } 6721 6722 static void 6723 hook_pf(void) 6724 { 6725 struct pfil_hook_args pha = { 6726 .pa_version = PFIL_VERSION, 6727 .pa_modname = "pf", 6728 }; 6729 struct pfil_link_args pla = { 6730 .pa_version = PFIL_VERSION, 6731 }; 6732 int ret __diagused; 6733 6734 if (atomic_load_bool(&V_pf_pfil_hooked)) 6735 return; 6736 6737 #ifdef INET 6738 pha.pa_type = PFIL_TYPE_IP4; 6739 pha.pa_mbuf_chk = pf_check_in; 6740 pha.pa_flags = PFIL_IN; 6741 pha.pa_rulname = "default-in"; 6742 V_pf_ip4_in_hook = pfil_add_hook(&pha); 6743 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6744 pla.pa_head = V_inet_pfil_head; 6745 pla.pa_hook = V_pf_ip4_in_hook; 6746 ret = pfil_link(&pla); 6747 MPASS(ret == 0); 6748 pha.pa_mbuf_chk = pf_check_out; 6749 pha.pa_flags = PFIL_OUT; 6750 pha.pa_rulname = "default-out"; 6751 V_pf_ip4_out_hook = pfil_add_hook(&pha); 6752 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6753 pla.pa_head = V_inet_pfil_head; 6754 pla.pa_hook = V_pf_ip4_out_hook; 6755 ret = pfil_link(&pla); 6756 MPASS(ret == 0); 6757 if (V_pf_filter_local) { 6758 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6759 pla.pa_head = V_inet_local_pfil_head; 6760 pla.pa_hook = V_pf_ip4_out_hook; 6761 ret = pfil_link(&pla); 6762 MPASS(ret == 0); 6763 } 6764 #endif 6765 #ifdef INET6 6766 pha.pa_type = PFIL_TYPE_IP6; 6767 pha.pa_mbuf_chk = pf_check6_in; 6768 pha.pa_flags = PFIL_IN; 6769 pha.pa_rulname = "default-in6"; 6770 V_pf_ip6_in_hook = pfil_add_hook(&pha); 6771 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6772 pla.pa_head = V_inet6_pfil_head; 6773 pla.pa_hook = V_pf_ip6_in_hook; 6774 ret = pfil_link(&pla); 6775 MPASS(ret == 0); 6776 pha.pa_mbuf_chk = pf_check6_out; 6777 pha.pa_rulname = "default-out6"; 6778 pha.pa_flags = PFIL_OUT; 6779 V_pf_ip6_out_hook = pfil_add_hook(&pha); 6780 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6781 pla.pa_head = V_inet6_pfil_head; 6782 pla.pa_hook = V_pf_ip6_out_hook; 6783 ret = pfil_link(&pla); 6784 MPASS(ret == 0); 6785 if (V_pf_filter_local) { 6786 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6787 pla.pa_head = V_inet6_local_pfil_head; 6788 pla.pa_hook = V_pf_ip6_out_hook; 6789 ret = pfil_link(&pla); 6790 MPASS(ret == 0); 6791 } 6792 #endif 6793 6794 atomic_store_bool(&V_pf_pfil_hooked, true); 6795 } 6796 6797 static void 6798 dehook_pf_eth(void) 6799 { 6800 6801 if (!atomic_load_bool(&V_pf_pfil_eth_hooked)) 6802 return; 6803 6804 pfil_remove_hook(V_pf_eth_in_hook); 6805 pfil_remove_hook(V_pf_eth_out_hook); 6806 6807 atomic_store_bool(&V_pf_pfil_eth_hooked, false); 6808 } 6809 6810 static void 6811 dehook_pf(void) 6812 { 6813 6814 if (!atomic_load_bool(&V_pf_pfil_hooked)) 6815 return; 6816 6817 #ifdef INET 6818 pfil_remove_hook(V_pf_ip4_in_hook); 6819 pfil_remove_hook(V_pf_ip4_out_hook); 6820 #endif 6821 #ifdef INET6 6822 pfil_remove_hook(V_pf_ip6_in_hook); 6823 pfil_remove_hook(V_pf_ip6_out_hook); 6824 #endif 6825 6826 atomic_store_bool(&V_pf_pfil_hooked, false); 6827 } 6828 6829 static void 6830 pf_load_vnet(void) 6831 { 6832 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname), 6833 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 6834 6835 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE); 6836 sx_init(&V_pf_ioctl_lock, "pf ioctl"); 6837 6838 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize, 6839 PF_RULE_TAG_HASH_SIZE_DEFAULT); 6840 #ifdef ALTQ 6841 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize, 6842 PF_QUEUE_TAG_HASH_SIZE_DEFAULT); 6843 #endif 6844 6845 V_pf_keth = &V_pf_main_keth_anchor.ruleset; 6846 6847 pfattach_vnet(); 6848 V_pf_vnet_active = 1; 6849 } 6850 6851 static int 6852 pf_load(void) 6853 { 6854 int error; 6855 6856 sx_init(&pf_end_lock, "pf end thread"); 6857 6858 pf_mtag_initialize(); 6859 6860 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME); 6861 if (pf_dev == NULL) 6862 return (ENOMEM); 6863 6864 pf_end_threads = 0; 6865 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge"); 6866 if (error != 0) 6867 return (error); 6868 6869 pfi_initialize(); 6870 6871 return (0); 6872 } 6873 6874 static void 6875 pf_unload_vnet(void) 6876 { 6877 int ret __diagused; 6878 6879 V_pf_vnet_active = 0; 6880 V_pf_status.running = 0; 6881 dehook_pf(); 6882 dehook_pf_eth(); 6883 6884 PF_RULES_WLOCK(); 6885 pf_syncookies_cleanup(); 6886 shutdown_pf(); 6887 PF_RULES_WUNLOCK(); 6888 6889 ret = swi_remove(V_pf_swi_cookie); 6890 MPASS(ret == 0); 6891 ret = intr_event_destroy(V_pf_swi_ie); 6892 MPASS(ret == 0); 6893 6894 pf_unload_vnet_purge(); 6895 6896 pf_normalize_cleanup(); 6897 PF_RULES_WLOCK(); 6898 pfi_cleanup_vnet(); 6899 PF_RULES_WUNLOCK(); 6900 pfr_cleanup(); 6901 pf_osfp_flush(); 6902 pf_cleanup(); 6903 if (IS_DEFAULT_VNET(curvnet)) 6904 pf_mtag_cleanup(); 6905 6906 pf_cleanup_tagset(&V_pf_tags); 6907 #ifdef ALTQ 6908 pf_cleanup_tagset(&V_pf_qids); 6909 #endif 6910 uma_zdestroy(V_pf_tag_z); 6911 6912 #ifdef PF_WANT_32_TO_64_COUNTER 6913 PF_RULES_WLOCK(); 6914 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); 6915 6916 MPASS(LIST_EMPTY(&V_pf_allkiflist)); 6917 MPASS(V_pf_allkifcount == 0); 6918 6919 LIST_REMOVE(&V_pf_default_rule, allrulelist); 6920 V_pf_allrulecount--; 6921 LIST_REMOVE(V_pf_rulemarker, allrulelist); 6922 6923 MPASS(LIST_EMPTY(&V_pf_allrulelist)); 6924 MPASS(V_pf_allrulecount == 0); 6925 6926 PF_RULES_WUNLOCK(); 6927 6928 free(V_pf_kifmarker, PFI_MTYPE); 6929 free(V_pf_rulemarker, M_PFRULE); 6930 #endif 6931 6932 /* Free counters last as we updated them during shutdown. */ 6933 pf_counter_u64_deinit(&V_pf_default_rule.evaluations); 6934 for (int i = 0; i < 2; i++) { 6935 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]); 6936 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]); 6937 } 6938 counter_u64_free(V_pf_default_rule.states_cur); 6939 counter_u64_free(V_pf_default_rule.states_tot); 6940 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 6941 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]); 6942 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp); 6943 6944 for (int i = 0; i < PFRES_MAX; i++) 6945 counter_u64_free(V_pf_status.counters[i]); 6946 for (int i = 0; i < KLCNT_MAX; i++) 6947 counter_u64_free(V_pf_status.lcounters[i]); 6948 for (int i = 0; i < FCNT_MAX; i++) 6949 pf_counter_u64_deinit(&V_pf_status.fcounters[i]); 6950 for (int i = 0; i < SCNT_MAX; i++) 6951 counter_u64_free(V_pf_status.scounters[i]); 6952 6953 rm_destroy(&V_pf_rules_lock); 6954 sx_destroy(&V_pf_ioctl_lock); 6955 } 6956 6957 static void 6958 pf_unload(void) 6959 { 6960 6961 sx_xlock(&pf_end_lock); 6962 pf_end_threads = 1; 6963 while (pf_end_threads < 2) { 6964 wakeup_one(pf_purge_thread); 6965 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0); 6966 } 6967 sx_xunlock(&pf_end_lock); 6968 6969 pf_nl_unregister(); 6970 6971 if (pf_dev != NULL) 6972 destroy_dev(pf_dev); 6973 6974 pfi_cleanup(); 6975 6976 sx_destroy(&pf_end_lock); 6977 } 6978 6979 static void 6980 vnet_pf_init(void *unused __unused) 6981 { 6982 6983 pf_load_vnet(); 6984 } 6985 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6986 vnet_pf_init, NULL); 6987 6988 static void 6989 vnet_pf_uninit(const void *unused __unused) 6990 { 6991 6992 pf_unload_vnet(); 6993 } 6994 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL); 6995 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 6996 vnet_pf_uninit, NULL); 6997 6998 static int 6999 pf_modevent(module_t mod, int type, void *data) 7000 { 7001 int error = 0; 7002 7003 switch(type) { 7004 case MOD_LOAD: 7005 error = pf_load(); 7006 pf_nl_register(); 7007 break; 7008 case MOD_UNLOAD: 7009 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after 7010 * the vnet_pf_uninit()s */ 7011 break; 7012 default: 7013 error = EINVAL; 7014 break; 7015 } 7016 7017 return (error); 7018 } 7019 7020 static moduledata_t pf_mod = { 7021 "pf", 7022 pf_modevent, 7023 0 7024 }; 7025 7026 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND); 7027 MODULE_DEPEND(pf, netlink, 1, 1, 1); 7028 MODULE_DEPEND(pf, crypto, 1, 1, 1); 7029 MODULE_VERSION(pf, PF_MODVER); 7030