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 MALLOC_DEFINE(M_PF, "pf", "pf(4)"); 191 192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE) 193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE 194 #endif 195 196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false; 197 #define V_pf_filter_local VNET(pf_filter_local) 198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW, 199 &VNET_NAME(pf_filter_local), false, 200 "Enable filtering for packets delivered to local network stack"); 201 202 #ifdef PF_DEFAULT_TO_DROP 203 VNET_DEFINE_STATIC(bool, default_to_drop) = true; 204 #else 205 VNET_DEFINE_STATIC(bool, default_to_drop); 206 #endif 207 #define V_default_to_drop VNET(default_to_drop) 208 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET, 209 &VNET_NAME(default_to_drop), false, 210 "Make the default rule drop all packets."); 211 212 static void pf_init_tagset(struct pf_tagset *, unsigned int *, 213 unsigned int); 214 static void pf_cleanup_tagset(struct pf_tagset *); 215 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *); 216 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t); 217 static u_int16_t tagname2tag(struct pf_tagset *, const char *); 218 static u_int16_t pf_tagname2tag(const char *); 219 static void tag_unref(struct pf_tagset *, u_int16_t); 220 221 struct cdev *pf_dev; 222 223 /* 224 * XXX - These are new and need to be checked when moveing to a new version 225 */ 226 static void pf_clear_all_states(void); 227 static int pf_killstates_row(struct pf_kstate_kill *, 228 struct pf_idhash *); 229 static int pf_killstates_nv(struct pfioc_nv *); 230 static int pf_clearstates_nv(struct pfioc_nv *); 231 static int pf_getstate(struct pfioc_nv *); 232 static int pf_getstatus(struct pfioc_nv *); 233 static int pf_clear_tables(void); 234 static void pf_kill_srcnodes(struct pfioc_src_node_kill *); 235 static int pf_keepcounters(struct pfioc_nv *); 236 static void pf_tbladdr_copyout(struct pf_addr_wrap *); 237 238 /* 239 * Wrapper functions for pfil(9) hooks 240 */ 241 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, 242 int flags, void *ruleset __unused, struct inpcb *inp); 243 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, 244 int flags, void *ruleset __unused, struct inpcb *inp); 245 #ifdef INET 246 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp, 247 int flags, void *ruleset __unused, struct inpcb *inp); 248 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp, 249 int flags, void *ruleset __unused, struct inpcb *inp); 250 #endif 251 #ifdef INET6 252 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp, 253 int flags, void *ruleset __unused, struct inpcb *inp); 254 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp, 255 int flags, void *ruleset __unused, struct inpcb *inp); 256 #endif 257 258 static void hook_pf_eth(void); 259 static void hook_pf(void); 260 static void dehook_pf_eth(void); 261 static void dehook_pf(void); 262 static int shutdown_pf(void); 263 static int pf_load(void); 264 static void pf_unload(void); 265 266 static struct cdevsw pf_cdevsw = { 267 .d_ioctl = pfioctl, 268 .d_name = PF_NAME, 269 .d_version = D_VERSION, 270 }; 271 272 VNET_DEFINE_STATIC(bool, pf_pfil_hooked); 273 #define V_pf_pfil_hooked VNET(pf_pfil_hooked) 274 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked); 275 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked) 276 277 /* 278 * We need a flag that is neither hooked nor running to know when 279 * the VNET is "valid". We primarily need this to control (global) 280 * external event, e.g., eventhandlers. 281 */ 282 VNET_DEFINE(int, pf_vnet_active); 283 #define V_pf_vnet_active VNET(pf_vnet_active) 284 285 int pf_end_threads; 286 struct proc *pf_purge_proc; 287 288 VNET_DEFINE(struct rmlock, pf_rules_lock); 289 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock); 290 #define V_pf_ioctl_lock VNET(pf_ioctl_lock) 291 struct sx pf_end_lock; 292 293 /* pfsync */ 294 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr); 295 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr); 296 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr); 297 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr); 298 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr); 299 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr); 300 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr); 301 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr; 302 303 /* pflog */ 304 pflog_packet_t *pflog_packet_ptr = NULL; 305 306 /* 307 * Copy a user-provided string, returning an error if truncation would occur. 308 * Avoid scanning past "sz" bytes in the source string since there's no 309 * guarantee that it's nul-terminated. 310 */ 311 static int 312 pf_user_strcpy(char *dst, const char *src, size_t sz) 313 { 314 if (strnlen(src, sz) == sz) 315 return (EINVAL); 316 (void)strlcpy(dst, src, sz); 317 return (0); 318 } 319 320 static void 321 pfattach_vnet(void) 322 { 323 u_int32_t *my_timeout = V_pf_default_rule.timeout; 324 325 bzero(&V_pf_status, sizeof(V_pf_status)); 326 327 pf_initialize(); 328 pfr_initialize(); 329 pfi_initialize_vnet(); 330 pf_normalize_init(); 331 pf_syncookies_init(); 332 333 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT; 334 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT; 335 V_pf_limits[PF_LIMIT_ANCHORS].limit = PF_ANCHOR_HIWAT; 336 V_pf_limits[PF_LIMIT_ETH_ANCHORS].limit = PF_ANCHOR_HIWAT; 337 338 RB_INIT(&V_pf_anchors); 339 pf_init_kruleset(&pf_main_ruleset); 340 341 pf_init_keth(V_pf_keth); 342 343 /* default rule should never be garbage collected */ 344 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; 345 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS; 346 V_pf_default_rule.nr = (uint32_t)-1; 347 V_pf_default_rule.rtableid = -1; 348 349 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK); 350 for (int i = 0; i < 2; i++) { 351 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK); 352 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK); 353 } 354 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); 355 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); 356 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++) 357 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 358 359 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 360 M_WAITOK | M_ZERO); 361 362 #ifdef PF_WANT_32_TO_64_COUNTER 363 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO); 364 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO); 365 PF_RULES_WLOCK(); 366 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); 367 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist); 368 V_pf_allrulecount++; 369 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); 370 PF_RULES_WUNLOCK(); 371 #endif 372 373 /* initialize default timeouts */ 374 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 375 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; 376 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 377 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; 378 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; 379 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; 380 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 381 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL; 382 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 383 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL; 384 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL; 385 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; 386 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; 387 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; 388 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; 389 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; 390 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; 391 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; 392 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; 393 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; 394 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; 395 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; 396 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; 397 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; 398 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; 399 400 V_pf_status.debug = PF_DEBUG_URGENT; 401 /* 402 * XXX This is different than in OpenBSD where reassembly is enabled by 403 * defult. In FreeBSD we expect people to still use scrub rules and 404 * switch to the new syntax later. Only when they switch they must 405 * explicitly enable reassemle. We could change the default once the 406 * scrub rule functionality is hopefully removed some day in future. 407 */ 408 V_pf_status.reass = 0; 409 410 V_pf_pfil_hooked = false; 411 V_pf_pfil_eth_hooked = false; 412 413 /* XXX do our best to avoid a conflict */ 414 V_pf_status.hostid = arc4random(); 415 416 for (int i = 0; i < PFRES_MAX; i++) 417 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); 418 for (int i = 0; i < KLCNT_MAX; i++) 419 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); 420 for (int i = 0; i < FCNT_MAX; i++) 421 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK); 422 for (int i = 0; i < SCNT_MAX; i++) 423 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); 424 425 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET, 426 INTR_MPSAFE, &V_pf_swi_cookie) != 0) 427 /* XXXGL: leaked all above. */ 428 return; 429 } 430 431 static struct pf_kpool * 432 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action, 433 u_int32_t rule_number, u_int8_t r_last, u_int8_t active, 434 u_int8_t check_ticket, int which) 435 { 436 struct pf_kruleset *ruleset; 437 struct pf_krule *rule; 438 int rs_num; 439 440 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT); 441 442 ruleset = pf_find_kruleset(anchor); 443 if (ruleset == NULL) 444 return (NULL); 445 rs_num = pf_get_ruleset_number(rule_action); 446 if (rs_num >= PF_RULESET_MAX) 447 return (NULL); 448 if (active) { 449 if (check_ticket && ticket != 450 ruleset->rules[rs_num].active.ticket) 451 return (NULL); 452 if (r_last) 453 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 454 pf_krulequeue); 455 else 456 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 457 } else { 458 if (check_ticket && ticket != 459 ruleset->rules[rs_num].inactive.ticket) 460 return (NULL); 461 if (r_last) 462 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 463 pf_krulequeue); 464 else 465 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); 466 } 467 if (!r_last) { 468 while ((rule != NULL) && (rule->nr != rule_number)) 469 rule = TAILQ_NEXT(rule, entries); 470 } 471 if (rule == NULL) 472 return (NULL); 473 474 switch (which) { 475 case PF_RDR: 476 return (&rule->rdr); 477 case PF_NAT: 478 return (&rule->nat); 479 case PF_RT: 480 return (&rule->route); 481 default: 482 panic("Unknow pool type %d", which); 483 } 484 } 485 486 static void 487 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb) 488 { 489 struct pf_kpooladdr *mv_pool_pa; 490 491 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { 492 TAILQ_REMOVE(poola, mv_pool_pa, entries); 493 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); 494 } 495 } 496 497 static void 498 pf_empty_kpool(struct pf_kpalist *poola) 499 { 500 struct pf_kpooladdr *pa; 501 502 while ((pa = TAILQ_FIRST(poola)) != NULL) { 503 switch (pa->addr.type) { 504 case PF_ADDR_DYNIFTL: 505 pfi_dynaddr_remove(pa->addr.p.dyn); 506 break; 507 case PF_ADDR_TABLE: 508 /* XXX: this could be unfinished pooladdr on pabuf */ 509 if (pa->addr.p.tbl != NULL) 510 pfr_detach_table(pa->addr.p.tbl); 511 break; 512 } 513 if (pa->kif) 514 pfi_kkif_unref(pa->kif); 515 TAILQ_REMOVE(poola, pa, entries); 516 free(pa, M_PFRULE); 517 } 518 } 519 520 static void 521 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 522 { 523 524 PF_RULES_WASSERT(); 525 PF_UNLNKDRULES_ASSERT(); 526 527 TAILQ_REMOVE(rulequeue, rule, entries); 528 529 rule->rule_ref |= PFRULE_REFS; 530 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); 531 } 532 533 static void 534 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 535 { 536 537 PF_RULES_WASSERT(); 538 539 PF_UNLNKDRULES_LOCK(); 540 pf_unlink_rule_locked(rulequeue, rule); 541 PF_UNLNKDRULES_UNLOCK(); 542 } 543 544 static void 545 pf_free_eth_rule(struct pf_keth_rule *rule) 546 { 547 PF_RULES_WASSERT(); 548 549 if (rule == NULL) 550 return; 551 552 if (rule->tag) 553 tag_unref(&V_pf_tags, rule->tag); 554 if (rule->match_tag) 555 tag_unref(&V_pf_tags, rule->match_tag); 556 #ifdef ALTQ 557 pf_qid_unref(rule->qid); 558 #endif 559 560 if (rule->bridge_to) 561 pfi_kkif_unref(rule->bridge_to); 562 if (rule->kif) 563 pfi_kkif_unref(rule->kif); 564 565 if (rule->ipsrc.addr.type == PF_ADDR_TABLE) 566 pfr_detach_table(rule->ipsrc.addr.p.tbl); 567 if (rule->ipdst.addr.type == PF_ADDR_TABLE) 568 pfr_detach_table(rule->ipdst.addr.p.tbl); 569 570 counter_u64_free(rule->evaluations); 571 for (int i = 0; i < 2; i++) { 572 counter_u64_free(rule->packets[i]); 573 counter_u64_free(rule->bytes[i]); 574 } 575 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 576 pf_keth_anchor_remove(rule); 577 578 free(rule, M_PFRULE); 579 } 580 581 void 582 pf_free_rule(struct pf_krule *rule) 583 { 584 585 PF_RULES_WASSERT(); 586 PF_CONFIG_ASSERT(); 587 588 if (rule->tag) 589 tag_unref(&V_pf_tags, rule->tag); 590 if (rule->match_tag) 591 tag_unref(&V_pf_tags, rule->match_tag); 592 #ifdef ALTQ 593 if (rule->pqid != rule->qid) 594 pf_qid_unref(rule->pqid); 595 pf_qid_unref(rule->qid); 596 #endif 597 switch (rule->src.addr.type) { 598 case PF_ADDR_DYNIFTL: 599 pfi_dynaddr_remove(rule->src.addr.p.dyn); 600 break; 601 case PF_ADDR_TABLE: 602 pfr_detach_table(rule->src.addr.p.tbl); 603 break; 604 } 605 switch (rule->dst.addr.type) { 606 case PF_ADDR_DYNIFTL: 607 pfi_dynaddr_remove(rule->dst.addr.p.dyn); 608 break; 609 case PF_ADDR_TABLE: 610 pfr_detach_table(rule->dst.addr.p.tbl); 611 break; 612 } 613 if (rule->overload_tbl) 614 pfr_detach_table(rule->overload_tbl); 615 if (rule->kif) 616 pfi_kkif_unref(rule->kif); 617 if (rule->rcv_kif) 618 pfi_kkif_unref(rule->rcv_kif); 619 pf_remove_kanchor(rule); 620 pf_empty_kpool(&rule->rdr.list); 621 pf_empty_kpool(&rule->nat.list); 622 pf_empty_kpool(&rule->route.list); 623 624 pf_krule_free(rule); 625 } 626 627 static void 628 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size, 629 unsigned int default_size) 630 { 631 unsigned int i; 632 unsigned int hashsize; 633 634 if (*tunable_size == 0 || !powerof2(*tunable_size)) 635 *tunable_size = default_size; 636 637 hashsize = *tunable_size; 638 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH, 639 M_WAITOK); 640 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH, 641 M_WAITOK); 642 ts->mask = hashsize - 1; 643 ts->seed = arc4random(); 644 for (i = 0; i < hashsize; i++) { 645 TAILQ_INIT(&ts->namehash[i]); 646 TAILQ_INIT(&ts->taghash[i]); 647 } 648 BIT_FILL(TAGID_MAX, &ts->avail); 649 } 650 651 static void 652 pf_cleanup_tagset(struct pf_tagset *ts) 653 { 654 unsigned int i; 655 unsigned int hashsize; 656 struct pf_tagname *t, *tmp; 657 658 /* 659 * Only need to clean up one of the hashes as each tag is hashed 660 * into each table. 661 */ 662 hashsize = ts->mask + 1; 663 for (i = 0; i < hashsize; i++) 664 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp) 665 uma_zfree(V_pf_tag_z, t); 666 667 free(ts->namehash, M_PFHASH); 668 free(ts->taghash, M_PFHASH); 669 } 670 671 static uint16_t 672 tagname2hashindex(const struct pf_tagset *ts, const char *tagname) 673 { 674 size_t len; 675 676 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1); 677 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask); 678 } 679 680 static uint16_t 681 tag2hashindex(const struct pf_tagset *ts, uint16_t tag) 682 { 683 684 return (tag & ts->mask); 685 } 686 687 static u_int16_t 688 tagname2tag(struct pf_tagset *ts, const char *tagname) 689 { 690 struct pf_tagname *tag; 691 u_int32_t index; 692 u_int16_t new_tagid; 693 694 PF_RULES_WASSERT(); 695 696 index = tagname2hashindex(ts, tagname); 697 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 698 if (strcmp(tagname, tag->name) == 0) { 699 tag->ref++; 700 return (tag->tag); 701 } 702 703 /* 704 * new entry 705 * 706 * to avoid fragmentation, we do a linear search from the beginning 707 * and take the first free slot we find. 708 */ 709 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail); 710 /* 711 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX]. 712 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits 713 * set. It may also return a bit number greater than TAGID_MAX due 714 * to rounding of the number of bits in the vector up to a multiple 715 * of the vector word size at declaration/allocation time. 716 */ 717 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) 718 return (0); 719 720 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */ 721 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail); 722 723 /* allocate and fill new struct pf_tagname */ 724 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT); 725 if (tag == NULL) 726 return (0); 727 strlcpy(tag->name, tagname, sizeof(tag->name)); 728 tag->tag = new_tagid; 729 tag->ref = 1; 730 731 /* Insert into namehash */ 732 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries); 733 734 /* Insert into taghash */ 735 index = tag2hashindex(ts, new_tagid); 736 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries); 737 738 return (tag->tag); 739 } 740 741 static void 742 tag_unref(struct pf_tagset *ts, u_int16_t tag) 743 { 744 struct pf_tagname *t; 745 uint16_t index; 746 747 PF_RULES_WASSERT(); 748 749 index = tag2hashindex(ts, tag); 750 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 751 if (tag == t->tag) { 752 if (--t->ref == 0) { 753 TAILQ_REMOVE(&ts->taghash[index], t, 754 taghash_entries); 755 index = tagname2hashindex(ts, t->name); 756 TAILQ_REMOVE(&ts->namehash[index], t, 757 namehash_entries); 758 /* Bits are 0-based for BIT_SET() */ 759 BIT_SET(TAGID_MAX, tag - 1, &ts->avail); 760 uma_zfree(V_pf_tag_z, t); 761 } 762 break; 763 } 764 } 765 766 static uint16_t 767 pf_tagname2tag(const char *tagname) 768 { 769 return (tagname2tag(&V_pf_tags, tagname)); 770 } 771 772 static int 773 pf_begin_eth(uint32_t *ticket, const char *anchor) 774 { 775 struct pf_keth_rule *rule, *tmp; 776 struct pf_keth_ruleset *rs; 777 778 PF_RULES_WASSERT(); 779 780 rs = pf_find_or_create_keth_ruleset(anchor); 781 if (rs == NULL) 782 return (EINVAL); 783 784 /* Purge old inactive rules. */ 785 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 786 tmp) { 787 TAILQ_REMOVE(rs->inactive.rules, rule, 788 entries); 789 pf_free_eth_rule(rule); 790 } 791 792 *ticket = ++rs->inactive.ticket; 793 rs->inactive.open = 1; 794 795 return (0); 796 } 797 798 static int 799 pf_rollback_eth(uint32_t ticket, const char *anchor) 800 { 801 struct pf_keth_rule *rule, *tmp; 802 struct pf_keth_ruleset *rs; 803 804 PF_RULES_WASSERT(); 805 806 rs = pf_find_keth_ruleset(anchor); 807 if (rs == NULL) 808 return (EINVAL); 809 810 if (!rs->inactive.open || 811 ticket != rs->inactive.ticket) 812 return (0); 813 814 /* Purge old inactive rules. */ 815 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 816 tmp) { 817 TAILQ_REMOVE(rs->inactive.rules, rule, entries); 818 pf_free_eth_rule(rule); 819 } 820 821 rs->inactive.open = 0; 822 823 pf_remove_if_empty_keth_ruleset(rs); 824 825 return (0); 826 } 827 828 #define PF_SET_SKIP_STEPS(i) \ 829 do { \ 830 while (head[i] != cur) { \ 831 head[i]->skip[i].ptr = cur; \ 832 head[i] = TAILQ_NEXT(head[i], entries); \ 833 } \ 834 } while (0) 835 836 static void 837 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules) 838 { 839 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT]; 840 int i; 841 842 cur = TAILQ_FIRST(rules); 843 prev = cur; 844 for (i = 0; i < PFE_SKIP_COUNT; ++i) 845 head[i] = cur; 846 while (cur != NULL) { 847 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 848 PF_SET_SKIP_STEPS(PFE_SKIP_IFP); 849 if (cur->direction != prev->direction) 850 PF_SET_SKIP_STEPS(PFE_SKIP_DIR); 851 if (cur->proto != prev->proto) 852 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO); 853 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0) 854 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR); 855 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0) 856 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR); 857 if (cur->ipsrc.neg != prev->ipsrc.neg || 858 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr)) 859 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR); 860 if (cur->ipdst.neg != prev->ipdst.neg || 861 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr)) 862 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR); 863 864 prev = cur; 865 cur = TAILQ_NEXT(cur, entries); 866 } 867 for (i = 0; i < PFE_SKIP_COUNT; ++i) 868 PF_SET_SKIP_STEPS(i); 869 } 870 871 static int 872 pf_commit_eth(uint32_t ticket, const char *anchor) 873 { 874 struct pf_keth_ruleq *rules; 875 struct pf_keth_ruleset *rs; 876 877 rs = pf_find_keth_ruleset(anchor); 878 if (rs == NULL) { 879 return (EINVAL); 880 } 881 882 if (!rs->inactive.open || 883 ticket != rs->inactive.ticket) 884 return (EBUSY); 885 886 PF_RULES_WASSERT(); 887 888 pf_eth_calc_skip_steps(rs->inactive.rules); 889 890 rules = rs->active.rules; 891 atomic_store_ptr(&rs->active.rules, rs->inactive.rules); 892 rs->inactive.rules = rules; 893 rs->inactive.ticket = rs->active.ticket; 894 895 return (pf_rollback_eth(rs->inactive.ticket, 896 rs->anchor ? rs->anchor->path : "")); 897 } 898 899 #ifdef ALTQ 900 static uint16_t 901 pf_qname2qid(const char *qname) 902 { 903 return (tagname2tag(&V_pf_qids, qname)); 904 } 905 906 static void 907 pf_qid_unref(uint16_t qid) 908 { 909 tag_unref(&V_pf_qids, qid); 910 } 911 912 static int 913 pf_begin_altq(u_int32_t *ticket) 914 { 915 struct pf_altq *altq, *tmp; 916 int error = 0; 917 918 PF_RULES_WASSERT(); 919 920 /* Purge the old altq lists */ 921 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 922 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 923 /* detach and destroy the discipline */ 924 error = altq_remove(altq); 925 } 926 free(altq, M_PFALTQ); 927 } 928 TAILQ_INIT(V_pf_altq_ifs_inactive); 929 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 930 pf_qid_unref(altq->qid); 931 free(altq, M_PFALTQ); 932 } 933 TAILQ_INIT(V_pf_altqs_inactive); 934 if (error) 935 return (error); 936 *ticket = ++V_ticket_altqs_inactive; 937 V_altqs_inactive_open = 1; 938 return (0); 939 } 940 941 static int 942 pf_rollback_altq(u_int32_t ticket) 943 { 944 struct pf_altq *altq, *tmp; 945 int error = 0; 946 947 PF_RULES_WASSERT(); 948 949 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 950 return (0); 951 /* Purge the old altq lists */ 952 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 953 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 954 /* detach and destroy the discipline */ 955 error = altq_remove(altq); 956 } 957 free(altq, M_PFALTQ); 958 } 959 TAILQ_INIT(V_pf_altq_ifs_inactive); 960 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 961 pf_qid_unref(altq->qid); 962 free(altq, M_PFALTQ); 963 } 964 TAILQ_INIT(V_pf_altqs_inactive); 965 V_altqs_inactive_open = 0; 966 return (error); 967 } 968 969 static int 970 pf_commit_altq(u_int32_t ticket) 971 { 972 struct pf_altqqueue *old_altqs, *old_altq_ifs; 973 struct pf_altq *altq, *tmp; 974 int err, error = 0; 975 976 PF_RULES_WASSERT(); 977 978 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 979 return (EBUSY); 980 981 /* swap altqs, keep the old. */ 982 old_altqs = V_pf_altqs_active; 983 old_altq_ifs = V_pf_altq_ifs_active; 984 V_pf_altqs_active = V_pf_altqs_inactive; 985 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive; 986 V_pf_altqs_inactive = old_altqs; 987 V_pf_altq_ifs_inactive = old_altq_ifs; 988 V_ticket_altqs_active = V_ticket_altqs_inactive; 989 990 /* Attach new disciplines */ 991 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 992 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 993 /* attach the discipline */ 994 error = altq_pfattach(altq); 995 if (error == 0 && V_pf_altq_running) 996 error = pf_enable_altq(altq); 997 if (error != 0) 998 return (error); 999 } 1000 } 1001 1002 /* Purge the old altq lists */ 1003 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 1004 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 1005 /* detach and destroy the discipline */ 1006 if (V_pf_altq_running) 1007 error = pf_disable_altq(altq); 1008 err = altq_pfdetach(altq); 1009 if (err != 0 && error == 0) 1010 error = err; 1011 err = altq_remove(altq); 1012 if (err != 0 && error == 0) 1013 error = err; 1014 } 1015 free(altq, M_PFALTQ); 1016 } 1017 TAILQ_INIT(V_pf_altq_ifs_inactive); 1018 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1019 pf_qid_unref(altq->qid); 1020 free(altq, M_PFALTQ); 1021 } 1022 TAILQ_INIT(V_pf_altqs_inactive); 1023 1024 V_altqs_inactive_open = 0; 1025 return (error); 1026 } 1027 1028 static int 1029 pf_enable_altq(struct pf_altq *altq) 1030 { 1031 struct ifnet *ifp; 1032 struct tb_profile tb; 1033 int error = 0; 1034 1035 if ((ifp = ifunit(altq->ifname)) == NULL) 1036 return (EINVAL); 1037 1038 if (ifp->if_snd.altq_type != ALTQT_NONE) 1039 error = altq_enable(&ifp->if_snd); 1040 1041 /* set tokenbucket regulator */ 1042 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { 1043 tb.rate = altq->ifbandwidth; 1044 tb.depth = altq->tbrsize; 1045 error = tbr_set(&ifp->if_snd, &tb); 1046 } 1047 1048 return (error); 1049 } 1050 1051 static int 1052 pf_disable_altq(struct pf_altq *altq) 1053 { 1054 struct ifnet *ifp; 1055 struct tb_profile tb; 1056 int error; 1057 1058 if ((ifp = ifunit(altq->ifname)) == NULL) 1059 return (EINVAL); 1060 1061 /* 1062 * when the discipline is no longer referenced, it was overridden 1063 * by a new one. if so, just return. 1064 */ 1065 if (altq->altq_disc != ifp->if_snd.altq_disc) 1066 return (0); 1067 1068 error = altq_disable(&ifp->if_snd); 1069 1070 if (error == 0) { 1071 /* clear tokenbucket regulator */ 1072 tb.rate = 0; 1073 error = tbr_set(&ifp->if_snd, &tb); 1074 } 1075 1076 return (error); 1077 } 1078 1079 static int 1080 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket, 1081 struct pf_altq *altq) 1082 { 1083 struct ifnet *ifp1; 1084 int error = 0; 1085 1086 /* Deactivate the interface in question */ 1087 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; 1088 if ((ifp1 = ifunit(altq->ifname)) == NULL || 1089 (remove && ifp1 == ifp)) { 1090 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 1091 } else { 1092 error = altq_add(ifp1, altq); 1093 1094 if (ticket != V_ticket_altqs_inactive) 1095 error = EBUSY; 1096 1097 if (error) 1098 free(altq, M_PFALTQ); 1099 } 1100 1101 return (error); 1102 } 1103 1104 void 1105 pf_altq_ifnet_event(struct ifnet *ifp, int remove) 1106 { 1107 struct pf_altq *a1, *a2, *a3; 1108 u_int32_t ticket; 1109 int error = 0; 1110 1111 /* 1112 * No need to re-evaluate the configuration for events on interfaces 1113 * that do not support ALTQ, as it's not possible for such 1114 * interfaces to be part of the configuration. 1115 */ 1116 if (!ALTQ_IS_READY(&ifp->if_snd)) 1117 return; 1118 1119 /* Interrupt userland queue modifications */ 1120 if (V_altqs_inactive_open) 1121 pf_rollback_altq(V_ticket_altqs_inactive); 1122 1123 /* Start new altq ruleset */ 1124 if (pf_begin_altq(&ticket)) 1125 return; 1126 1127 /* Copy the current active set */ 1128 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) { 1129 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1130 if (a2 == NULL) { 1131 error = ENOMEM; 1132 break; 1133 } 1134 bcopy(a1, a2, sizeof(struct pf_altq)); 1135 1136 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1137 if (error) 1138 break; 1139 1140 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries); 1141 } 1142 if (error) 1143 goto out; 1144 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { 1145 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1146 if (a2 == NULL) { 1147 error = ENOMEM; 1148 break; 1149 } 1150 bcopy(a1, a2, sizeof(struct pf_altq)); 1151 1152 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) { 1153 error = EBUSY; 1154 free(a2, M_PFALTQ); 1155 break; 1156 } 1157 a2->altq_disc = NULL; 1158 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) { 1159 if (strncmp(a3->ifname, a2->ifname, 1160 IFNAMSIZ) == 0) { 1161 a2->altq_disc = a3->altq_disc; 1162 break; 1163 } 1164 } 1165 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1166 if (error) 1167 break; 1168 1169 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); 1170 } 1171 1172 out: 1173 if (error != 0) 1174 pf_rollback_altq(ticket); 1175 else 1176 pf_commit_altq(ticket); 1177 } 1178 #endif /* ALTQ */ 1179 1180 static struct pf_krule_global * 1181 pf_rule_tree_alloc(int flags) 1182 { 1183 struct pf_krule_global *tree; 1184 1185 tree = malloc(sizeof(struct pf_krule_global), M_PF, flags); 1186 if (tree == NULL) 1187 return (NULL); 1188 RB_INIT(tree); 1189 return (tree); 1190 } 1191 1192 void 1193 pf_rule_tree_free(struct pf_krule_global *tree) 1194 { 1195 1196 free(tree, M_PF); 1197 } 1198 1199 static int 1200 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) 1201 { 1202 struct pf_krule_global *tree; 1203 struct pf_kruleset *rs; 1204 struct pf_krule *rule; 1205 1206 PF_RULES_WASSERT(); 1207 1208 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1209 return (EINVAL); 1210 tree = pf_rule_tree_alloc(M_NOWAIT); 1211 if (tree == NULL) 1212 return (ENOMEM); 1213 rs = pf_find_or_create_kruleset(anchor); 1214 if (rs == NULL) { 1215 pf_rule_tree_free(tree); 1216 return (EINVAL); 1217 } 1218 pf_rule_tree_free(rs->rules[rs_num].inactive.tree); 1219 rs->rules[rs_num].inactive.tree = tree; 1220 1221 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1222 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1223 rs->rules[rs_num].inactive.rcount--; 1224 } 1225 *ticket = ++rs->rules[rs_num].inactive.ticket; 1226 rs->rules[rs_num].inactive.open = 1; 1227 return (0); 1228 } 1229 1230 static int 1231 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) 1232 { 1233 struct pf_kruleset *rs; 1234 struct pf_krule *rule; 1235 1236 PF_RULES_WASSERT(); 1237 1238 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1239 return (EINVAL); 1240 rs = pf_find_kruleset(anchor); 1241 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1242 rs->rules[rs_num].inactive.ticket != ticket) 1243 return (0); 1244 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1245 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1246 rs->rules[rs_num].inactive.rcount--; 1247 } 1248 rs->rules[rs_num].inactive.open = 0; 1249 return (0); 1250 } 1251 1252 #define PF_MD5_UPD(st, elm) \ 1253 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) 1254 1255 #define PF_MD5_UPD_STR(st, elm) \ 1256 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) 1257 1258 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ 1259 (stor) = htonl((st)->elm); \ 1260 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ 1261 } while (0) 1262 1263 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ 1264 (stor) = htons((st)->elm); \ 1265 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ 1266 } while (0) 1267 1268 static void 1269 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) 1270 { 1271 PF_MD5_UPD(pfr, addr.type); 1272 switch (pfr->addr.type) { 1273 case PF_ADDR_DYNIFTL: 1274 PF_MD5_UPD(pfr, addr.v.ifname); 1275 PF_MD5_UPD(pfr, addr.iflags); 1276 break; 1277 case PF_ADDR_TABLE: 1278 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX, 1279 strlen(PF_OPTIMIZER_TABLE_PFX))) 1280 PF_MD5_UPD(pfr, addr.v.tblname); 1281 break; 1282 case PF_ADDR_ADDRMASK: 1283 /* XXX ignore af? */ 1284 PF_MD5_UPD(pfr, addr.v.a.addr.addr32); 1285 PF_MD5_UPD(pfr, addr.v.a.mask.addr32); 1286 break; 1287 } 1288 1289 PF_MD5_UPD(pfr, port[0]); 1290 PF_MD5_UPD(pfr, port[1]); 1291 PF_MD5_UPD(pfr, neg); 1292 PF_MD5_UPD(pfr, port_op); 1293 } 1294 1295 static void 1296 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule) 1297 { 1298 u_int16_t x; 1299 u_int32_t y; 1300 1301 pf_hash_rule_addr(ctx, &rule->src); 1302 pf_hash_rule_addr(ctx, &rule->dst); 1303 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) 1304 PF_MD5_UPD_STR(rule, label[i]); 1305 PF_MD5_UPD_STR(rule, ifname); 1306 PF_MD5_UPD_STR(rule, rcv_ifname); 1307 PF_MD5_UPD_STR(rule, match_tagname); 1308 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ 1309 PF_MD5_UPD_HTONL(rule, os_fingerprint, y); 1310 PF_MD5_UPD_HTONL(rule, prob, y); 1311 PF_MD5_UPD_HTONL(rule, uid.uid[0], y); 1312 PF_MD5_UPD_HTONL(rule, uid.uid[1], y); 1313 PF_MD5_UPD(rule, uid.op); 1314 PF_MD5_UPD_HTONL(rule, gid.gid[0], y); 1315 PF_MD5_UPD_HTONL(rule, gid.gid[1], y); 1316 PF_MD5_UPD(rule, gid.op); 1317 PF_MD5_UPD_HTONL(rule, rule_flag, y); 1318 PF_MD5_UPD(rule, action); 1319 PF_MD5_UPD(rule, direction); 1320 PF_MD5_UPD(rule, af); 1321 PF_MD5_UPD(rule, quick); 1322 PF_MD5_UPD(rule, ifnot); 1323 PF_MD5_UPD(rule, rcvifnot); 1324 PF_MD5_UPD(rule, match_tag_not); 1325 PF_MD5_UPD(rule, natpass); 1326 PF_MD5_UPD(rule, keep_state); 1327 PF_MD5_UPD(rule, proto); 1328 PF_MD5_UPD(rule, type); 1329 PF_MD5_UPD(rule, code); 1330 PF_MD5_UPD(rule, flags); 1331 PF_MD5_UPD(rule, flagset); 1332 PF_MD5_UPD(rule, allow_opts); 1333 PF_MD5_UPD(rule, rt); 1334 PF_MD5_UPD(rule, tos); 1335 PF_MD5_UPD(rule, scrub_flags); 1336 PF_MD5_UPD(rule, min_ttl); 1337 PF_MD5_UPD(rule, set_tos); 1338 if (rule->anchor != NULL) 1339 PF_MD5_UPD_STR(rule, anchor->path); 1340 } 1341 1342 static void 1343 pf_hash_rule(struct pf_krule *rule) 1344 { 1345 MD5_CTX ctx; 1346 1347 MD5Init(&ctx); 1348 pf_hash_rule_rolling(&ctx, rule); 1349 MD5Final(rule->md5sum, &ctx); 1350 } 1351 1352 static int 1353 pf_krule_compare(struct pf_krule *a, struct pf_krule *b) 1354 { 1355 1356 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH)); 1357 } 1358 1359 static int 1360 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) 1361 { 1362 struct pf_kruleset *rs; 1363 struct pf_krule *rule, *old_rule; 1364 struct pf_krulequeue *old_rules; 1365 struct pf_krule_global *old_tree; 1366 int error; 1367 u_int32_t old_rcount; 1368 1369 PF_RULES_WASSERT(); 1370 1371 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1372 return (EINVAL); 1373 rs = pf_find_kruleset(anchor); 1374 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1375 ticket != rs->rules[rs_num].inactive.ticket) 1376 return (EBUSY); 1377 1378 /* Calculate checksum for the main ruleset */ 1379 if (rs == &pf_main_ruleset) { 1380 error = pf_setup_pfsync_matching(rs); 1381 if (error != 0) 1382 return (error); 1383 } 1384 1385 /* Swap rules, keep the old. */ 1386 old_rules = rs->rules[rs_num].active.ptr; 1387 old_rcount = rs->rules[rs_num].active.rcount; 1388 old_tree = rs->rules[rs_num].active.tree; 1389 1390 rs->rules[rs_num].active.ptr = 1391 rs->rules[rs_num].inactive.ptr; 1392 rs->rules[rs_num].active.tree = 1393 rs->rules[rs_num].inactive.tree; 1394 rs->rules[rs_num].active.rcount = 1395 rs->rules[rs_num].inactive.rcount; 1396 1397 /* Attempt to preserve counter information. */ 1398 if (V_pf_status.keep_counters && old_tree != NULL) { 1399 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr, 1400 entries) { 1401 old_rule = RB_FIND(pf_krule_global, old_tree, rule); 1402 if (old_rule == NULL) { 1403 continue; 1404 } 1405 pf_counter_u64_critical_enter(); 1406 pf_counter_u64_rollup_protected(&rule->evaluations, 1407 pf_counter_u64_fetch(&old_rule->evaluations)); 1408 pf_counter_u64_rollup_protected(&rule->packets[0], 1409 pf_counter_u64_fetch(&old_rule->packets[0])); 1410 pf_counter_u64_rollup_protected(&rule->packets[1], 1411 pf_counter_u64_fetch(&old_rule->packets[1])); 1412 pf_counter_u64_rollup_protected(&rule->bytes[0], 1413 pf_counter_u64_fetch(&old_rule->bytes[0])); 1414 pf_counter_u64_rollup_protected(&rule->bytes[1], 1415 pf_counter_u64_fetch(&old_rule->bytes[1])); 1416 pf_counter_u64_critical_exit(); 1417 } 1418 } 1419 1420 rs->rules[rs_num].inactive.ptr = old_rules; 1421 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ 1422 rs->rules[rs_num].inactive.rcount = old_rcount; 1423 1424 rs->rules[rs_num].active.ticket = 1425 rs->rules[rs_num].inactive.ticket; 1426 pf_calc_skip_steps(rs->rules[rs_num].active.ptr); 1427 1428 /* Purge the old rule list. */ 1429 PF_UNLNKDRULES_LOCK(); 1430 while ((rule = TAILQ_FIRST(old_rules)) != NULL) 1431 pf_unlink_rule_locked(old_rules, rule); 1432 PF_UNLNKDRULES_UNLOCK(); 1433 rs->rules[rs_num].inactive.rcount = 0; 1434 rs->rules[rs_num].inactive.open = 0; 1435 pf_remove_if_empty_kruleset(rs); 1436 pf_rule_tree_free(old_tree); 1437 1438 return (0); 1439 } 1440 1441 static int 1442 pf_setup_pfsync_matching(struct pf_kruleset *rs) 1443 { 1444 MD5_CTX ctx; 1445 struct pf_krule *rule; 1446 int rs_cnt; 1447 u_int8_t digest[PF_MD5_DIGEST_LENGTH]; 1448 1449 MD5Init(&ctx); 1450 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { 1451 /* XXX PF_RULESET_SCRUB as well? */ 1452 if (rs_cnt == PF_RULESET_SCRUB) 1453 continue; 1454 1455 if (rs->rules[rs_cnt].inactive.rcount) { 1456 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, 1457 entries) { 1458 pf_hash_rule_rolling(&ctx, rule); 1459 } 1460 } 1461 } 1462 1463 MD5Final(digest, &ctx); 1464 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); 1465 return (0); 1466 } 1467 1468 static int 1469 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr) 1470 { 1471 int error = 0; 1472 1473 switch (addr->type) { 1474 case PF_ADDR_TABLE: 1475 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname); 1476 if (addr->p.tbl == NULL) 1477 error = ENOMEM; 1478 break; 1479 default: 1480 error = EINVAL; 1481 } 1482 1483 return (error); 1484 } 1485 1486 static int 1487 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr, 1488 sa_family_t af) 1489 { 1490 int error = 0; 1491 1492 switch (addr->type) { 1493 case PF_ADDR_TABLE: 1494 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); 1495 if (addr->p.tbl == NULL) 1496 error = ENOMEM; 1497 break; 1498 case PF_ADDR_DYNIFTL: 1499 error = pfi_dynaddr_setup(addr, af); 1500 break; 1501 } 1502 1503 return (error); 1504 } 1505 1506 void 1507 pf_addr_copyout(struct pf_addr_wrap *addr) 1508 { 1509 1510 switch (addr->type) { 1511 case PF_ADDR_DYNIFTL: 1512 pfi_dynaddr_copyout(addr); 1513 break; 1514 case PF_ADDR_TABLE: 1515 pf_tbladdr_copyout(addr); 1516 break; 1517 } 1518 } 1519 1520 static void 1521 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out) 1522 { 1523 int secs = time_uptime; 1524 1525 bzero(out, sizeof(struct pf_src_node)); 1526 1527 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr)); 1528 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr)); 1529 1530 if (in->rule != NULL) 1531 out->rule.nr = in->rule->nr; 1532 1533 for (int i = 0; i < 2; i++) { 1534 out->bytes[i] = counter_u64_fetch(in->bytes[i]); 1535 out->packets[i] = counter_u64_fetch(in->packets[i]); 1536 } 1537 1538 out->states = in->states; 1539 out->conn = in->conn; 1540 out->af = in->af; 1541 out->ruletype = in->ruletype; 1542 1543 out->creation = secs - in->creation; 1544 if (out->expire > secs) 1545 out->expire -= secs; 1546 else 1547 out->expire = 0; 1548 1549 /* Adjust the connection rate estimate. */ 1550 out->conn_rate.limit = in->conn_rate.limit; 1551 out->conn_rate.seconds = in->conn_rate.seconds; 1552 /* If there's no limit there's no counter_rate. */ 1553 if (in->conn_rate.cr != NULL) 1554 out->conn_rate.count = counter_rate_get(in->conn_rate.cr); 1555 } 1556 1557 #ifdef ALTQ 1558 /* 1559 * Handle export of struct pf_kaltq to user binaries that may be using any 1560 * version of struct pf_altq. 1561 */ 1562 static int 1563 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size) 1564 { 1565 u_int32_t version; 1566 1567 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1568 version = 0; 1569 else 1570 version = pa->version; 1571 1572 if (version > PFIOC_ALTQ_VERSION) 1573 return (EINVAL); 1574 1575 #define ASSIGN(x) exported_q->x = q->x 1576 #define COPY(x) \ 1577 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x))) 1578 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX) 1579 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX) 1580 1581 switch (version) { 1582 case 0: { 1583 struct pf_altq_v0 *exported_q = 1584 &((struct pfioc_altq_v0 *)pa)->altq; 1585 1586 COPY(ifname); 1587 1588 ASSIGN(scheduler); 1589 ASSIGN(tbrsize); 1590 exported_q->tbrsize = SATU16(q->tbrsize); 1591 exported_q->ifbandwidth = SATU32(q->ifbandwidth); 1592 1593 COPY(qname); 1594 COPY(parent); 1595 ASSIGN(parent_qid); 1596 exported_q->bandwidth = SATU32(q->bandwidth); 1597 ASSIGN(priority); 1598 ASSIGN(local_flags); 1599 1600 ASSIGN(qlimit); 1601 ASSIGN(flags); 1602 1603 if (q->scheduler == ALTQT_HFSC) { 1604 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x 1605 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \ 1606 SATU32(q->pq_u.hfsc_opts.x) 1607 1608 ASSIGN_OPT_SATU32(rtsc_m1); 1609 ASSIGN_OPT(rtsc_d); 1610 ASSIGN_OPT_SATU32(rtsc_m2); 1611 1612 ASSIGN_OPT_SATU32(lssc_m1); 1613 ASSIGN_OPT(lssc_d); 1614 ASSIGN_OPT_SATU32(lssc_m2); 1615 1616 ASSIGN_OPT_SATU32(ulsc_m1); 1617 ASSIGN_OPT(ulsc_d); 1618 ASSIGN_OPT_SATU32(ulsc_m2); 1619 1620 ASSIGN_OPT(flags); 1621 1622 #undef ASSIGN_OPT 1623 #undef ASSIGN_OPT_SATU32 1624 } else 1625 COPY(pq_u); 1626 1627 ASSIGN(qid); 1628 break; 1629 } 1630 case 1: { 1631 struct pf_altq_v1 *exported_q = 1632 &((struct pfioc_altq_v1 *)pa)->altq; 1633 1634 COPY(ifname); 1635 1636 ASSIGN(scheduler); 1637 ASSIGN(tbrsize); 1638 ASSIGN(ifbandwidth); 1639 1640 COPY(qname); 1641 COPY(parent); 1642 ASSIGN(parent_qid); 1643 ASSIGN(bandwidth); 1644 ASSIGN(priority); 1645 ASSIGN(local_flags); 1646 1647 ASSIGN(qlimit); 1648 ASSIGN(flags); 1649 COPY(pq_u); 1650 1651 ASSIGN(qid); 1652 break; 1653 } 1654 default: 1655 panic("%s: unhandled struct pfioc_altq version", __func__); 1656 break; 1657 } 1658 1659 #undef ASSIGN 1660 #undef COPY 1661 #undef SATU16 1662 #undef SATU32 1663 1664 return (0); 1665 } 1666 1667 /* 1668 * Handle import to struct pf_kaltq of struct pf_altq from user binaries 1669 * that may be using any version of it. 1670 */ 1671 static int 1672 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size) 1673 { 1674 u_int32_t version; 1675 1676 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1677 version = 0; 1678 else 1679 version = pa->version; 1680 1681 if (version > PFIOC_ALTQ_VERSION) 1682 return (EINVAL); 1683 1684 #define ASSIGN(x) q->x = imported_q->x 1685 #define COPY(x) \ 1686 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x))) 1687 1688 switch (version) { 1689 case 0: { 1690 struct pf_altq_v0 *imported_q = 1691 &((struct pfioc_altq_v0 *)pa)->altq; 1692 1693 COPY(ifname); 1694 1695 ASSIGN(scheduler); 1696 ASSIGN(tbrsize); /* 16-bit -> 32-bit */ 1697 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */ 1698 1699 COPY(qname); 1700 COPY(parent); 1701 ASSIGN(parent_qid); 1702 ASSIGN(bandwidth); /* 32-bit -> 64-bit */ 1703 ASSIGN(priority); 1704 ASSIGN(local_flags); 1705 1706 ASSIGN(qlimit); 1707 ASSIGN(flags); 1708 1709 if (imported_q->scheduler == ALTQT_HFSC) { 1710 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x 1711 1712 /* 1713 * The m1 and m2 parameters are being copied from 1714 * 32-bit to 64-bit. 1715 */ 1716 ASSIGN_OPT(rtsc_m1); 1717 ASSIGN_OPT(rtsc_d); 1718 ASSIGN_OPT(rtsc_m2); 1719 1720 ASSIGN_OPT(lssc_m1); 1721 ASSIGN_OPT(lssc_d); 1722 ASSIGN_OPT(lssc_m2); 1723 1724 ASSIGN_OPT(ulsc_m1); 1725 ASSIGN_OPT(ulsc_d); 1726 ASSIGN_OPT(ulsc_m2); 1727 1728 ASSIGN_OPT(flags); 1729 1730 #undef ASSIGN_OPT 1731 } else 1732 COPY(pq_u); 1733 1734 ASSIGN(qid); 1735 break; 1736 } 1737 case 1: { 1738 struct pf_altq_v1 *imported_q = 1739 &((struct pfioc_altq_v1 *)pa)->altq; 1740 1741 COPY(ifname); 1742 1743 ASSIGN(scheduler); 1744 ASSIGN(tbrsize); 1745 ASSIGN(ifbandwidth); 1746 1747 COPY(qname); 1748 COPY(parent); 1749 ASSIGN(parent_qid); 1750 ASSIGN(bandwidth); 1751 ASSIGN(priority); 1752 ASSIGN(local_flags); 1753 1754 ASSIGN(qlimit); 1755 ASSIGN(flags); 1756 COPY(pq_u); 1757 1758 ASSIGN(qid); 1759 break; 1760 } 1761 default: 1762 panic("%s: unhandled struct pfioc_altq version", __func__); 1763 break; 1764 } 1765 1766 #undef ASSIGN 1767 #undef COPY 1768 1769 return (0); 1770 } 1771 1772 static struct pf_altq * 1773 pf_altq_get_nth_active(u_int32_t n) 1774 { 1775 struct pf_altq *altq; 1776 u_int32_t nr; 1777 1778 nr = 0; 1779 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1780 if (nr == n) 1781 return (altq); 1782 nr++; 1783 } 1784 1785 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { 1786 if (nr == n) 1787 return (altq); 1788 nr++; 1789 } 1790 1791 return (NULL); 1792 } 1793 #endif /* ALTQ */ 1794 1795 struct pf_krule * 1796 pf_krule_alloc(void) 1797 { 1798 struct pf_krule *rule; 1799 1800 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO); 1801 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF); 1802 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF); 1803 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF); 1804 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 1805 M_WAITOK | M_ZERO); 1806 return (rule); 1807 } 1808 1809 void 1810 pf_krule_free(struct pf_krule *rule) 1811 { 1812 #ifdef PF_WANT_32_TO_64_COUNTER 1813 bool wowned; 1814 #endif 1815 1816 if (rule == NULL) 1817 return; 1818 1819 #ifdef PF_WANT_32_TO_64_COUNTER 1820 if (rule->allrulelinked) { 1821 wowned = PF_RULES_WOWNED(); 1822 if (!wowned) 1823 PF_RULES_WLOCK(); 1824 LIST_REMOVE(rule, allrulelist); 1825 V_pf_allrulecount--; 1826 if (!wowned) 1827 PF_RULES_WUNLOCK(); 1828 } 1829 #endif 1830 1831 pf_counter_u64_deinit(&rule->evaluations); 1832 for (int i = 0; i < 2; i++) { 1833 pf_counter_u64_deinit(&rule->packets[i]); 1834 pf_counter_u64_deinit(&rule->bytes[i]); 1835 } 1836 counter_u64_free(rule->states_cur); 1837 counter_u64_free(rule->states_tot); 1838 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 1839 counter_u64_free(rule->src_nodes[sn_type]); 1840 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 1841 1842 mtx_destroy(&rule->nat.mtx); 1843 mtx_destroy(&rule->rdr.mtx); 1844 mtx_destroy(&rule->route.mtx); 1845 free(rule, M_PFRULE); 1846 } 1847 1848 void 1849 pf_krule_clear_counters(struct pf_krule *rule) 1850 { 1851 pf_counter_u64_zero(&rule->evaluations); 1852 for (int i = 0; i < 2; i++) { 1853 pf_counter_u64_zero(&rule->packets[i]); 1854 pf_counter_u64_zero(&rule->bytes[i]); 1855 } 1856 counter_u64_zero(rule->states_tot); 1857 } 1858 1859 static void 1860 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool, 1861 struct pf_pooladdr *pool) 1862 { 1863 1864 bzero(pool, sizeof(*pool)); 1865 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr)); 1866 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname)); 1867 } 1868 1869 static int 1870 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool, 1871 struct pf_kpooladdr *kpool) 1872 { 1873 int ret; 1874 1875 bzero(kpool, sizeof(*kpool)); 1876 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr)); 1877 ret = pf_user_strcpy(kpool->ifname, pool->ifname, 1878 sizeof(kpool->ifname)); 1879 return (ret); 1880 } 1881 1882 static void 1883 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool) 1884 { 1885 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), ""); 1886 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), ""); 1887 1888 bcopy(&pool->key, &kpool->key, sizeof(kpool->key)); 1889 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter)); 1890 1891 kpool->tblidx = pool->tblidx; 1892 kpool->proxy_port[0] = pool->proxy_port[0]; 1893 kpool->proxy_port[1] = pool->proxy_port[1]; 1894 kpool->opts = pool->opts; 1895 } 1896 1897 static int 1898 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule) 1899 { 1900 int ret; 1901 1902 #ifndef INET 1903 if (rule->af == AF_INET) { 1904 return (EAFNOSUPPORT); 1905 } 1906 #endif /* INET */ 1907 #ifndef INET6 1908 if (rule->af == AF_INET6) { 1909 return (EAFNOSUPPORT); 1910 } 1911 #endif /* INET6 */ 1912 1913 ret = pf_check_rule_addr(&rule->src); 1914 if (ret != 0) 1915 return (ret); 1916 ret = pf_check_rule_addr(&rule->dst); 1917 if (ret != 0) 1918 return (ret); 1919 1920 bcopy(&rule->src, &krule->src, sizeof(rule->src)); 1921 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst)); 1922 1923 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label)); 1924 if (ret != 0) 1925 return (ret); 1926 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname)); 1927 if (ret != 0) 1928 return (ret); 1929 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname)); 1930 if (ret != 0) 1931 return (ret); 1932 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname)); 1933 if (ret != 0) 1934 return (ret); 1935 ret = pf_user_strcpy(krule->tagname, rule->tagname, 1936 sizeof(rule->tagname)); 1937 if (ret != 0) 1938 return (ret); 1939 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname, 1940 sizeof(rule->match_tagname)); 1941 if (ret != 0) 1942 return (ret); 1943 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname, 1944 sizeof(rule->overload_tblname)); 1945 if (ret != 0) 1946 return (ret); 1947 1948 pf_pool_to_kpool(&rule->rpool, &krule->rdr); 1949 1950 /* Don't allow userspace to set evaluations, packets or bytes. */ 1951 /* kif, anchor, overload_tbl are not copied over. */ 1952 1953 krule->os_fingerprint = rule->os_fingerprint; 1954 1955 krule->rtableid = rule->rtableid; 1956 /* pf_rule->timeout is smaller than pf_krule->timeout */ 1957 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout)); 1958 krule->max_states = rule->max_states; 1959 krule->max_src_nodes = rule->max_src_nodes; 1960 krule->max_src_states = rule->max_src_states; 1961 krule->max_src_conn = rule->max_src_conn; 1962 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit; 1963 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds; 1964 krule->qid = rule->qid; 1965 krule->pqid = rule->pqid; 1966 krule->nr = rule->nr; 1967 krule->prob = rule->prob; 1968 krule->cuid = rule->cuid; 1969 krule->cpid = rule->cpid; 1970 1971 krule->return_icmp = rule->return_icmp; 1972 krule->return_icmp6 = rule->return_icmp6; 1973 krule->max_mss = rule->max_mss; 1974 krule->tag = rule->tag; 1975 krule->match_tag = rule->match_tag; 1976 krule->scrub_flags = rule->scrub_flags; 1977 1978 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid)); 1979 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid)); 1980 1981 krule->rule_flag = rule->rule_flag; 1982 krule->action = rule->action; 1983 krule->direction = rule->direction; 1984 krule->log = rule->log; 1985 krule->logif = rule->logif; 1986 krule->quick = rule->quick; 1987 krule->ifnot = rule->ifnot; 1988 krule->match_tag_not = rule->match_tag_not; 1989 krule->natpass = rule->natpass; 1990 1991 krule->keep_state = rule->keep_state; 1992 krule->af = rule->af; 1993 krule->proto = rule->proto; 1994 krule->type = rule->type; 1995 krule->code = rule->code; 1996 krule->flags = rule->flags; 1997 krule->flagset = rule->flagset; 1998 krule->min_ttl = rule->min_ttl; 1999 krule->allow_opts = rule->allow_opts; 2000 krule->rt = rule->rt; 2001 krule->return_ttl = rule->return_ttl; 2002 krule->tos = rule->tos; 2003 krule->set_tos = rule->set_tos; 2004 2005 krule->flush = rule->flush; 2006 krule->prio = rule->prio; 2007 krule->set_prio[0] = rule->set_prio[0]; 2008 krule->set_prio[1] = rule->set_prio[1]; 2009 2010 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert)); 2011 2012 return (0); 2013 } 2014 2015 int 2016 pf_ioctl_getrules(struct pfioc_rule *pr) 2017 { 2018 struct pf_kruleset *ruleset; 2019 struct pf_krule *tail; 2020 int rs_num; 2021 2022 PF_RULES_WLOCK(); 2023 ruleset = pf_find_kruleset(pr->anchor); 2024 if (ruleset == NULL) { 2025 PF_RULES_WUNLOCK(); 2026 return (EINVAL); 2027 } 2028 rs_num = pf_get_ruleset_number(pr->rule.action); 2029 if (rs_num >= PF_RULESET_MAX) { 2030 PF_RULES_WUNLOCK(); 2031 return (EINVAL); 2032 } 2033 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 2034 pf_krulequeue); 2035 if (tail) 2036 pr->nr = tail->nr + 1; 2037 else 2038 pr->nr = 0; 2039 pr->ticket = ruleset->rules[rs_num].active.ticket; 2040 PF_RULES_WUNLOCK(); 2041 2042 return (0); 2043 } 2044 2045 static int 2046 pf_rule_checkaf(struct pf_krule *r) 2047 { 2048 switch (r->af) { 2049 case 0: 2050 if (r->rule_flag & PFRULE_AFTO) 2051 return (EPFNOSUPPORT); 2052 break; 2053 case AF_INET: 2054 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6) 2055 return (EPFNOSUPPORT); 2056 break; 2057 #ifdef INET6 2058 case AF_INET6: 2059 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET) 2060 return (EPFNOSUPPORT); 2061 break; 2062 #endif /* INET6 */ 2063 default: 2064 return (EPFNOSUPPORT); 2065 } 2066 2067 if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0) 2068 return (EPFNOSUPPORT); 2069 2070 return (0); 2071 } 2072 2073 static int 2074 pf_validate_range(uint8_t op, uint16_t port[2]) 2075 { 2076 uint16_t a = ntohs(port[0]); 2077 uint16_t b = ntohs(port[1]); 2078 2079 if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */ 2080 (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */ 2081 (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */ 2082 return 1; 2083 return 0; 2084 } 2085 2086 int 2087 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, 2088 uint32_t pool_ticket, const char *anchor, const char *anchor_call, 2089 uid_t uid, pid_t pid) 2090 { 2091 struct pf_kruleset *ruleset; 2092 struct pf_krule *tail; 2093 struct pf_kpooladdr *pa; 2094 struct pfi_kkif *kif = NULL, *rcv_kif = NULL; 2095 int rs_num; 2096 int error = 0; 2097 2098 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 2099 #define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x) 2100 2101 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) 2102 ERROUT_UNLOCKED(EINVAL); 2103 2104 if ((error = pf_rule_checkaf(rule))) 2105 ERROUT_UNLOCKED(error); 2106 if (pf_validate_range(rule->src.port_op, rule->src.port)) 2107 ERROUT_UNLOCKED(EINVAL); 2108 if (pf_validate_range(rule->dst.port_op, rule->dst.port)) 2109 ERROUT_UNLOCKED(EINVAL); 2110 2111 if (rule->ifname[0]) 2112 kif = pf_kkif_create(M_WAITOK); 2113 if (rule->rcv_ifname[0]) 2114 rcv_kif = pf_kkif_create(M_WAITOK); 2115 pf_counter_u64_init(&rule->evaluations, M_WAITOK); 2116 for (int i = 0; i < 2; i++) { 2117 pf_counter_u64_init(&rule->packets[i], M_WAITOK); 2118 pf_counter_u64_init(&rule->bytes[i], M_WAITOK); 2119 } 2120 rule->states_cur = counter_u64_alloc(M_WAITOK); 2121 rule->states_tot = counter_u64_alloc(M_WAITOK); 2122 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 2123 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 2124 rule->cuid = uid; 2125 rule->cpid = pid; 2126 TAILQ_INIT(&rule->rdr.list); 2127 TAILQ_INIT(&rule->nat.list); 2128 TAILQ_INIT(&rule->route.list); 2129 2130 PF_CONFIG_LOCK(); 2131 PF_RULES_WLOCK(); 2132 #ifdef PF_WANT_32_TO_64_COUNTER 2133 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist); 2134 MPASS(!rule->allrulelinked); 2135 rule->allrulelinked = true; 2136 V_pf_allrulecount++; 2137 #endif 2138 ruleset = pf_find_kruleset(anchor); 2139 if (ruleset == NULL) 2140 ERROUT(EINVAL); 2141 rs_num = pf_get_ruleset_number(rule->action); 2142 if (rs_num >= PF_RULESET_MAX) 2143 ERROUT(EINVAL); 2144 if (ticket != ruleset->rules[rs_num].inactive.ticket) { 2145 DPFPRINTF(PF_DEBUG_MISC, 2146 "ticket: %d != [%d]%d", ticket, rs_num, 2147 ruleset->rules[rs_num].inactive.ticket); 2148 ERROUT(EBUSY); 2149 } 2150 if (pool_ticket != V_ticket_pabuf) { 2151 DPFPRINTF(PF_DEBUG_MISC, 2152 "pool_ticket: %d != %d", pool_ticket, 2153 V_ticket_pabuf); 2154 ERROUT(EBUSY); 2155 } 2156 /* 2157 * XXXMJG hack: there is no mechanism to ensure they started the 2158 * transaction. Ticket checked above may happen to match by accident, 2159 * even if nobody called DIOCXBEGIN, let alone this process. 2160 * Partially work around it by checking if the RB tree got allocated, 2161 * see pf_begin_rules. 2162 */ 2163 if (ruleset->rules[rs_num].inactive.tree == NULL) { 2164 ERROUT(EINVAL); 2165 } 2166 2167 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 2168 pf_krulequeue); 2169 if (tail) 2170 rule->nr = tail->nr + 1; 2171 else 2172 rule->nr = 0; 2173 if (rule->ifname[0]) { 2174 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2175 kif = NULL; 2176 pfi_kkif_ref(rule->kif); 2177 } else 2178 rule->kif = NULL; 2179 2180 if (rule->rcv_ifname[0]) { 2181 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname); 2182 rcv_kif = NULL; 2183 pfi_kkif_ref(rule->rcv_kif); 2184 } else 2185 rule->rcv_kif = NULL; 2186 2187 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) 2188 ERROUT(EBUSY); 2189 #ifdef ALTQ 2190 /* set queue IDs */ 2191 if (rule->qname[0] != 0) { 2192 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 2193 ERROUT(EBUSY); 2194 else if (rule->pqname[0] != 0) { 2195 if ((rule->pqid = 2196 pf_qname2qid(rule->pqname)) == 0) 2197 ERROUT(EBUSY); 2198 } else 2199 rule->pqid = rule->qid; 2200 } 2201 #endif 2202 if (rule->tagname[0]) 2203 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2204 ERROUT(EBUSY); 2205 if (rule->match_tagname[0]) 2206 if ((rule->match_tag = 2207 pf_tagname2tag(rule->match_tagname)) == 0) 2208 ERROUT(EBUSY); 2209 if (rule->rt && !rule->direction) 2210 ERROUT(EINVAL); 2211 if (!rule->log) 2212 rule->logif = 0; 2213 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit, 2214 rule->pktrate.seconds)) 2215 ERROUT(ENOMEM); 2216 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) 2217 ERROUT(ENOMEM); 2218 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) 2219 ERROUT(ENOMEM); 2220 if (pf_kanchor_setup(rule, ruleset, anchor_call)) 2221 ERROUT(EINVAL); 2222 if (rule->scrub_flags & PFSTATE_SETPRIO && 2223 (rule->set_prio[0] > PF_PRIO_MAX || 2224 rule->set_prio[1] > PF_PRIO_MAX)) 2225 ERROUT(EINVAL); 2226 for (int i = 0; i < 3; i++) { 2227 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 2228 if (pa->addr.type == PF_ADDR_TABLE) { 2229 pa->addr.p.tbl = pfr_attach_table(ruleset, 2230 pa->addr.v.tblname); 2231 if (pa->addr.p.tbl == NULL) 2232 ERROUT(ENOMEM); 2233 } 2234 } 2235 2236 rule->overload_tbl = NULL; 2237 if (rule->overload_tblname[0]) { 2238 if ((rule->overload_tbl = pfr_attach_table(ruleset, 2239 rule->overload_tblname)) == NULL) 2240 ERROUT(EINVAL); 2241 else 2242 rule->overload_tbl->pfrkt_flags |= 2243 PFR_TFLAG_ACTIVE; 2244 } 2245 2246 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list); 2247 2248 /* 2249 * Old version of pfctl provide route redirection pools in single 2250 * common redirection pool rdr. New versions use rdr only for 2251 * rdr-to rules. 2252 */ 2253 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) { 2254 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list); 2255 } else { 2256 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list); 2257 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list); 2258 } 2259 2260 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) || 2261 (rule->action == PF_BINAT)) && rule->anchor == NULL && 2262 TAILQ_FIRST(&rule->rdr.list) == NULL) { 2263 ERROUT(EINVAL); 2264 } 2265 2266 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) { 2267 ERROUT(EINVAL); 2268 } 2269 2270 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR || 2271 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) { 2272 ERROUT(EINVAL); 2273 } 2274 2275 MPASS(error == 0); 2276 2277 rule->nat.cur = TAILQ_FIRST(&rule->nat.list); 2278 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list); 2279 rule->route.cur = TAILQ_FIRST(&rule->route.list); 2280 rule->route.ipv6_nexthop_af = AF_INET6; 2281 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, 2282 rule, entries); 2283 ruleset->rules[rs_num].inactive.rcount++; 2284 2285 PF_RULES_WUNLOCK(); 2286 pf_hash_rule(rule); 2287 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) { 2288 PF_RULES_WLOCK(); 2289 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries); 2290 ruleset->rules[rs_num].inactive.rcount--; 2291 pf_free_rule(rule); 2292 rule = NULL; 2293 ERROUT(EEXIST); 2294 } 2295 PF_CONFIG_UNLOCK(); 2296 2297 return (0); 2298 2299 #undef ERROUT 2300 #undef ERROUT_UNLOCKED 2301 errout: 2302 PF_RULES_WUNLOCK(); 2303 PF_CONFIG_UNLOCK(); 2304 errout_unlocked: 2305 pf_kkif_free(rcv_kif); 2306 pf_kkif_free(kif); 2307 pf_krule_free(rule); 2308 return (error); 2309 } 2310 2311 static bool 2312 pf_label_match(const struct pf_krule *rule, const char *label) 2313 { 2314 int i = 0; 2315 2316 while (*rule->label[i]) { 2317 if (strcmp(rule->label[i], label) == 0) 2318 return (true); 2319 i++; 2320 } 2321 2322 return (false); 2323 } 2324 2325 static unsigned int 2326 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir) 2327 { 2328 struct pf_kstate *s; 2329 int more = 0; 2330 2331 s = pf_find_state_all(key, dir, &more); 2332 if (s == NULL) 2333 return (0); 2334 2335 if (more) { 2336 PF_STATE_UNLOCK(s); 2337 return (0); 2338 } 2339 2340 pf_remove_state(s); 2341 return (1); 2342 } 2343 2344 static int 2345 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih) 2346 { 2347 struct pf_kstate *s; 2348 struct pf_state_key *sk; 2349 struct pf_addr *srcaddr, *dstaddr; 2350 struct pf_state_key_cmp match_key; 2351 int idx, killed = 0; 2352 unsigned int dir; 2353 u_int16_t srcport, dstport; 2354 struct pfi_kkif *kif; 2355 2356 relock_DIOCKILLSTATES: 2357 PF_HASHROW_LOCK(ih); 2358 LIST_FOREACH(s, &ih->states, entry) { 2359 /* For floating states look at the original kif. */ 2360 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 2361 2362 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE]; 2363 if (s->direction == PF_OUT) { 2364 srcaddr = &sk->addr[1]; 2365 dstaddr = &sk->addr[0]; 2366 srcport = sk->port[1]; 2367 dstport = sk->port[0]; 2368 } else { 2369 srcaddr = &sk->addr[0]; 2370 dstaddr = &sk->addr[1]; 2371 srcport = sk->port[0]; 2372 dstport = sk->port[1]; 2373 } 2374 2375 if (psk->psk_af && sk->af != psk->psk_af) 2376 continue; 2377 2378 if (psk->psk_proto && psk->psk_proto != sk->proto) 2379 continue; 2380 2381 if (! pf_match_addr(psk->psk_src.neg, 2382 &psk->psk_src.addr.v.a.addr, 2383 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af)) 2384 continue; 2385 2386 if (! pf_match_addr(psk->psk_dst.neg, 2387 &psk->psk_dst.addr.v.a.addr, 2388 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af)) 2389 continue; 2390 2391 if (! pf_match_addr(psk->psk_rt_addr.neg, 2392 &psk->psk_rt_addr.addr.v.a.addr, 2393 &psk->psk_rt_addr.addr.v.a.mask, 2394 &s->act.rt_addr, sk->af)) 2395 continue; 2396 2397 if (psk->psk_src.port_op != 0 && 2398 ! pf_match_port(psk->psk_src.port_op, 2399 psk->psk_src.port[0], psk->psk_src.port[1], srcport)) 2400 continue; 2401 2402 if (psk->psk_dst.port_op != 0 && 2403 ! pf_match_port(psk->psk_dst.port_op, 2404 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) 2405 continue; 2406 2407 if (psk->psk_label[0] && 2408 ! pf_label_match(s->rule, psk->psk_label)) 2409 continue; 2410 2411 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname, 2412 kif->pfik_name)) 2413 continue; 2414 2415 if (psk->psk_kill_match) { 2416 /* Create the key to find matching states, with lock 2417 * held. */ 2418 2419 bzero(&match_key, sizeof(match_key)); 2420 2421 if (s->direction == PF_OUT) { 2422 dir = PF_IN; 2423 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK; 2424 } else { 2425 dir = PF_OUT; 2426 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE; 2427 } 2428 2429 match_key.af = s->key[idx]->af; 2430 match_key.proto = s->key[idx]->proto; 2431 pf_addrcpy(&match_key.addr[0], 2432 &s->key[idx]->addr[1], match_key.af); 2433 match_key.port[0] = s->key[idx]->port[1]; 2434 pf_addrcpy(&match_key.addr[1], 2435 &s->key[idx]->addr[0], match_key.af); 2436 match_key.port[1] = s->key[idx]->port[0]; 2437 } 2438 2439 pf_remove_state(s); 2440 killed++; 2441 2442 if (psk->psk_kill_match) 2443 killed += pf_kill_matching_state(&match_key, dir); 2444 2445 goto relock_DIOCKILLSTATES; 2446 } 2447 PF_HASHROW_UNLOCK(ih); 2448 2449 return (killed); 2450 } 2451 2452 void 2453 unhandled_af(int af) 2454 { 2455 panic("unhandled af %d", af); 2456 } 2457 2458 int 2459 pf_start(void) 2460 { 2461 int error = 0; 2462 2463 sx_xlock(&V_pf_ioctl_lock); 2464 if (V_pf_status.running) 2465 error = EEXIST; 2466 else { 2467 hook_pf(); 2468 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 2469 hook_pf_eth(); 2470 V_pf_status.running = 1; 2471 V_pf_status.since = time_uptime; 2472 new_unrhdr64(&V_pf_stateid, time_second); 2473 2474 DPFPRINTF(PF_DEBUG_MISC, "pf: started"); 2475 } 2476 sx_xunlock(&V_pf_ioctl_lock); 2477 2478 return (error); 2479 } 2480 2481 int 2482 pf_stop(void) 2483 { 2484 int error = 0; 2485 2486 sx_xlock(&V_pf_ioctl_lock); 2487 if (!V_pf_status.running) 2488 error = ENOENT; 2489 else { 2490 V_pf_status.running = 0; 2491 dehook_pf(); 2492 dehook_pf_eth(); 2493 V_pf_status.since = time_uptime; 2494 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped"); 2495 } 2496 sx_xunlock(&V_pf_ioctl_lock); 2497 2498 return (error); 2499 } 2500 2501 void 2502 pf_ioctl_clear_status(void) 2503 { 2504 PF_RULES_WLOCK(); 2505 for (int i = 0; i < PFRES_MAX; i++) 2506 counter_u64_zero(V_pf_status.counters[i]); 2507 for (int i = 0; i < FCNT_MAX; i++) 2508 pf_counter_u64_zero(&V_pf_status.fcounters[i]); 2509 for (int i = 0; i < SCNT_MAX; i++) 2510 counter_u64_zero(V_pf_status.scounters[i]); 2511 for (int i = 0; i < KLCNT_MAX; i++) 2512 counter_u64_zero(V_pf_status.lcounters[i]); 2513 V_pf_status.since = time_uptime; 2514 if (*V_pf_status.ifname) 2515 pfi_update_status(V_pf_status.ifname, NULL); 2516 PF_RULES_WUNLOCK(); 2517 } 2518 2519 int 2520 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds) 2521 { 2522 uint32_t old; 2523 2524 if (timeout < 0 || timeout >= PFTM_MAX || 2525 seconds < 0) 2526 return (EINVAL); 2527 2528 PF_RULES_WLOCK(); 2529 old = V_pf_default_rule.timeout[timeout]; 2530 if (timeout == PFTM_INTERVAL && seconds == 0) 2531 seconds = 1; 2532 V_pf_default_rule.timeout[timeout] = seconds; 2533 if (timeout == PFTM_INTERVAL && seconds < old) 2534 wakeup(pf_purge_thread); 2535 2536 if (prev_seconds != NULL) 2537 *prev_seconds = old; 2538 2539 PF_RULES_WUNLOCK(); 2540 2541 return (0); 2542 } 2543 2544 int 2545 pf_ioctl_get_timeout(int timeout, int *seconds) 2546 { 2547 PF_RULES_RLOCK_TRACKER; 2548 2549 if (timeout < 0 || timeout >= PFTM_MAX) 2550 return (EINVAL); 2551 2552 PF_RULES_RLOCK(); 2553 *seconds = V_pf_default_rule.timeout[timeout]; 2554 PF_RULES_RUNLOCK(); 2555 2556 return (0); 2557 } 2558 2559 int 2560 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit) 2561 { 2562 2563 PF_RULES_WLOCK(); 2564 if (index < 0 || index >= PF_LIMIT_MAX || 2565 V_pf_limits[index].zone == NULL) { 2566 PF_RULES_WUNLOCK(); 2567 return (EINVAL); 2568 } 2569 uma_zone_set_max(V_pf_limits[index].zone, 2570 limit == 0 ? INT_MAX : limit); 2571 if (old_limit != NULL) 2572 *old_limit = V_pf_limits[index].limit; 2573 V_pf_limits[index].limit = limit; 2574 PF_RULES_WUNLOCK(); 2575 2576 return (0); 2577 } 2578 2579 int 2580 pf_ioctl_get_limit(int index, unsigned int *limit) 2581 { 2582 PF_RULES_RLOCK_TRACKER; 2583 2584 if (index < 0 || index >= PF_LIMIT_MAX) 2585 return (EINVAL); 2586 2587 PF_RULES_RLOCK(); 2588 *limit = V_pf_limits[index].limit; 2589 PF_RULES_RUNLOCK(); 2590 2591 return (0); 2592 } 2593 2594 int 2595 pf_ioctl_begin_addrs(uint32_t *ticket) 2596 { 2597 PF_RULES_WLOCK(); 2598 pf_empty_kpool(&V_pf_pabuf[0]); 2599 pf_empty_kpool(&V_pf_pabuf[1]); 2600 pf_empty_kpool(&V_pf_pabuf[2]); 2601 *ticket = ++V_ticket_pabuf; 2602 PF_RULES_WUNLOCK(); 2603 2604 return (0); 2605 } 2606 2607 int 2608 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp) 2609 { 2610 struct pf_kpooladdr *pa = NULL; 2611 struct pfi_kkif *kif = NULL; 2612 int error; 2613 2614 if (pp->which != PF_RDR && pp->which != PF_NAT && 2615 pp->which != PF_RT) 2616 return (EINVAL); 2617 2618 switch (pp->af) { 2619 #ifdef INET 2620 case AF_INET: 2621 /* FALLTHROUGH */ 2622 #endif /* INET */ 2623 #ifdef INET6 2624 case AF_INET6: 2625 /* FALLTHROUGH */ 2626 #endif /* INET6 */ 2627 case AF_UNSPEC: 2628 break; 2629 default: 2630 return (EAFNOSUPPORT); 2631 } 2632 2633 if (pp->addr.addr.type != PF_ADDR_ADDRMASK && 2634 pp->addr.addr.type != PF_ADDR_DYNIFTL && 2635 pp->addr.addr.type != PF_ADDR_TABLE) 2636 return (EINVAL); 2637 2638 if (pp->addr.addr.p.dyn != NULL) 2639 return (EINVAL); 2640 2641 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); 2642 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa); 2643 if (error != 0) 2644 goto out; 2645 if (pa->ifname[0]) 2646 kif = pf_kkif_create(M_WAITOK); 2647 PF_RULES_WLOCK(); 2648 if (pp->ticket != V_ticket_pabuf) { 2649 PF_RULES_WUNLOCK(); 2650 if (pa->ifname[0]) 2651 pf_kkif_free(kif); 2652 error = EBUSY; 2653 goto out; 2654 } 2655 if (pa->ifname[0]) { 2656 pa->kif = pfi_kkif_attach(kif, pa->ifname); 2657 kif = NULL; 2658 pfi_kkif_ref(pa->kif); 2659 } else 2660 pa->kif = NULL; 2661 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = 2662 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { 2663 if (pa->ifname[0]) 2664 pfi_kkif_unref(pa->kif); 2665 PF_RULES_WUNLOCK(); 2666 goto out; 2667 } 2668 pa->af = pp->af; 2669 switch (pp->which) { 2670 case PF_NAT: 2671 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries); 2672 break; 2673 case PF_RDR: 2674 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries); 2675 break; 2676 case PF_RT: 2677 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries); 2678 break; 2679 } 2680 PF_RULES_WUNLOCK(); 2681 2682 return (0); 2683 2684 out: 2685 free(pa, M_PFRULE); 2686 return (error); 2687 } 2688 2689 int 2690 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp) 2691 { 2692 struct pf_kpool *pool; 2693 struct pf_kpooladdr *pa; 2694 2695 PF_RULES_RLOCK_TRACKER; 2696 2697 if (pp->which != PF_RDR && pp->which != PF_NAT && 2698 pp->which != PF_RT) 2699 return (EINVAL); 2700 2701 pp->anchor[sizeof(pp->anchor) - 1] = 0; 2702 pp->nr = 0; 2703 2704 PF_RULES_RLOCK(); 2705 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2706 pp->r_num, 0, 1, 0, pp->which); 2707 if (pool == NULL) { 2708 PF_RULES_RUNLOCK(); 2709 return (EBUSY); 2710 } 2711 TAILQ_FOREACH(pa, &pool->list, entries) 2712 pp->nr++; 2713 PF_RULES_RUNLOCK(); 2714 2715 return (0); 2716 } 2717 2718 int 2719 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp) 2720 { 2721 struct pf_kpool *pool; 2722 struct pf_kpooladdr *pa; 2723 u_int32_t nr = 0; 2724 2725 if (pp->which != PF_RDR && pp->which != PF_NAT && 2726 pp->which != PF_RT) 2727 return (EINVAL); 2728 2729 PF_RULES_RLOCK_TRACKER; 2730 2731 pp->anchor[sizeof(pp->anchor) - 1] = '\0'; 2732 2733 PF_RULES_RLOCK(); 2734 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2735 pp->r_num, 0, 1, 1, pp->which); 2736 if (pool == NULL) { 2737 PF_RULES_RUNLOCK(); 2738 return (EBUSY); 2739 } 2740 pa = TAILQ_FIRST(&pool->list); 2741 while ((pa != NULL) && (nr < pp->nr)) { 2742 pa = TAILQ_NEXT(pa, entries); 2743 nr++; 2744 } 2745 if (pa == NULL) { 2746 PF_RULES_RUNLOCK(); 2747 return (EBUSY); 2748 } 2749 pf_kpooladdr_to_pooladdr(pa, &pp->addr); 2750 pp->af = pa->af; 2751 pf_addr_copyout(&pp->addr.addr); 2752 PF_RULES_RUNLOCK(); 2753 2754 return (0); 2755 } 2756 2757 int 2758 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr) 2759 { 2760 struct pf_kruleset *ruleset; 2761 struct pf_kanchor *anchor; 2762 2763 PF_RULES_RLOCK_TRACKER; 2764 2765 pr->path[sizeof(pr->path) - 1] = '\0'; 2766 2767 PF_RULES_RLOCK(); 2768 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2769 PF_RULES_RUNLOCK(); 2770 return (ENOENT); 2771 } 2772 pr->nr = 0; 2773 if (ruleset == &pf_main_ruleset) { 2774 /* XXX kludge for pf_main_ruleset */ 2775 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2776 if (anchor->parent == NULL) 2777 pr->nr++; 2778 } else { 2779 RB_FOREACH(anchor, pf_kanchor_node, 2780 &ruleset->anchor->children) 2781 pr->nr++; 2782 } 2783 PF_RULES_RUNLOCK(); 2784 2785 return (0); 2786 } 2787 2788 int 2789 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr) 2790 { 2791 struct pf_kruleset *ruleset; 2792 struct pf_kanchor *anchor; 2793 u_int32_t nr = 0; 2794 int error = 0; 2795 2796 PF_RULES_RLOCK_TRACKER; 2797 2798 PF_RULES_RLOCK(); 2799 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2800 PF_RULES_RUNLOCK(); 2801 return (ENOENT); 2802 } 2803 2804 pr->name[0] = '\0'; 2805 if (ruleset == &pf_main_ruleset) { 2806 /* XXX kludge for pf_main_ruleset */ 2807 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2808 if (anchor->parent == NULL && nr++ == pr->nr) { 2809 strlcpy(pr->name, anchor->name, 2810 sizeof(pr->name)); 2811 break; 2812 } 2813 } else { 2814 RB_FOREACH(anchor, pf_kanchor_node, 2815 &ruleset->anchor->children) 2816 if (nr++ == pr->nr) { 2817 strlcpy(pr->name, anchor->name, 2818 sizeof(pr->name)); 2819 break; 2820 } 2821 } 2822 if (!pr->name[0]) 2823 error = EBUSY; 2824 PF_RULES_RUNLOCK(); 2825 2826 return (error); 2827 } 2828 2829 int 2830 pf_ioctl_natlook(struct pfioc_natlook *pnl) 2831 { 2832 struct pf_state_key *sk; 2833 struct pf_kstate *state; 2834 struct pf_state_key_cmp key; 2835 int m = 0, direction = pnl->direction; 2836 int sidx, didx; 2837 2838 /* NATLOOK src and dst are reversed, so reverse sidx/didx */ 2839 sidx = (direction == PF_IN) ? 1 : 0; 2840 didx = (direction == PF_IN) ? 0 : 1; 2841 2842 if (!pnl->proto || 2843 PF_AZERO(&pnl->saddr, pnl->af) || 2844 PF_AZERO(&pnl->daddr, pnl->af) || 2845 ((pnl->proto == IPPROTO_TCP || 2846 pnl->proto == IPPROTO_UDP) && 2847 (!pnl->dport || !pnl->sport))) 2848 return (EINVAL); 2849 2850 switch (pnl->direction) { 2851 case PF_IN: 2852 case PF_OUT: 2853 case PF_INOUT: 2854 break; 2855 default: 2856 return (EINVAL); 2857 } 2858 2859 switch (pnl->af) { 2860 #ifdef INET 2861 case AF_INET: 2862 break; 2863 #endif /* INET */ 2864 #ifdef INET6 2865 case AF_INET6: 2866 break; 2867 #endif /* INET6 */ 2868 default: 2869 return (EAFNOSUPPORT); 2870 } 2871 2872 bzero(&key, sizeof(key)); 2873 key.af = pnl->af; 2874 key.proto = pnl->proto; 2875 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af); 2876 key.port[sidx] = pnl->sport; 2877 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af); 2878 key.port[didx] = pnl->dport; 2879 2880 state = pf_find_state_all(&key, direction, &m); 2881 if (state == NULL) 2882 return (ENOENT); 2883 2884 if (m > 1) { 2885 PF_STATE_UNLOCK(state); 2886 return (E2BIG); /* more than one state */ 2887 } 2888 2889 sk = state->key[sidx]; 2890 pf_addrcpy(&pnl->rsaddr, 2891 &sk->addr[sidx], sk->af); 2892 pnl->rsport = sk->port[sidx]; 2893 pf_addrcpy(&pnl->rdaddr, 2894 &sk->addr[didx], sk->af); 2895 pnl->rdport = sk->port[didx]; 2896 PF_STATE_UNLOCK(state); 2897 2898 return (0); 2899 } 2900 2901 static int 2902 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) 2903 { 2904 int error = 0; 2905 PF_RULES_RLOCK_TRACKER; 2906 2907 #define ERROUT_IOCTL(target, x) \ 2908 do { \ 2909 error = (x); \ 2910 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \ 2911 goto target; \ 2912 } while (0) 2913 2914 2915 /* XXX keep in sync with switch() below */ 2916 if (securelevel_gt(td->td_ucred, 2)) 2917 switch (cmd) { 2918 case DIOCGETRULES: 2919 case DIOCGETRULENV: 2920 case DIOCGETADDRS: 2921 case DIOCGETADDR: 2922 case DIOCGETSTATE: 2923 case DIOCGETSTATENV: 2924 case DIOCSETSTATUSIF: 2925 case DIOCGETSTATUSNV: 2926 case DIOCCLRSTATUS: 2927 case DIOCNATLOOK: 2928 case DIOCSETDEBUG: 2929 #ifdef COMPAT_FREEBSD14 2930 case DIOCGETSTATES: 2931 case DIOCGETSTATESV2: 2932 #endif 2933 case DIOCGETTIMEOUT: 2934 case DIOCCLRRULECTRS: 2935 case DIOCGETLIMIT: 2936 case DIOCGETALTQSV0: 2937 case DIOCGETALTQSV1: 2938 case DIOCGETALTQV0: 2939 case DIOCGETALTQV1: 2940 case DIOCGETQSTATSV0: 2941 case DIOCGETQSTATSV1: 2942 case DIOCGETRULESETS: 2943 case DIOCGETRULESET: 2944 case DIOCRGETTABLES: 2945 case DIOCRGETTSTATS: 2946 case DIOCRCLRTSTATS: 2947 case DIOCRCLRADDRS: 2948 case DIOCRADDADDRS: 2949 case DIOCRDELADDRS: 2950 case DIOCRSETADDRS: 2951 case DIOCRGETADDRS: 2952 case DIOCRGETASTATS: 2953 case DIOCRCLRASTATS: 2954 case DIOCRTSTADDRS: 2955 case DIOCOSFPGET: 2956 case DIOCGETSRCNODES: 2957 case DIOCCLRSRCNODES: 2958 case DIOCGETSYNCOOKIES: 2959 case DIOCIGETIFACES: 2960 case DIOCGIFSPEEDV0: 2961 case DIOCGIFSPEEDV1: 2962 case DIOCSETIFFLAG: 2963 case DIOCCLRIFFLAG: 2964 case DIOCGETETHRULES: 2965 case DIOCGETETHRULE: 2966 case DIOCGETETHRULESETS: 2967 case DIOCGETETHRULESET: 2968 break; 2969 case DIOCRCLRTABLES: 2970 case DIOCRADDTABLES: 2971 case DIOCRDELTABLES: 2972 case DIOCRSETTFLAGS: 2973 if (((struct pfioc_table *)addr)->pfrio_flags & 2974 PFR_FLAG_DUMMY) 2975 break; /* dummy operation ok */ 2976 return (EPERM); 2977 default: 2978 return (EPERM); 2979 } 2980 2981 if (!(flags & FWRITE)) 2982 switch (cmd) { 2983 case DIOCGETRULES: 2984 case DIOCGETADDRS: 2985 case DIOCGETADDR: 2986 case DIOCGETSTATE: 2987 case DIOCGETSTATENV: 2988 case DIOCGETSTATUSNV: 2989 #ifdef COMPAT_FREEBSD14 2990 case DIOCGETSTATES: 2991 case DIOCGETSTATESV2: 2992 #endif 2993 case DIOCGETTIMEOUT: 2994 case DIOCGETLIMIT: 2995 case DIOCGETALTQSV0: 2996 case DIOCGETALTQSV1: 2997 case DIOCGETALTQV0: 2998 case DIOCGETALTQV1: 2999 case DIOCGETQSTATSV0: 3000 case DIOCGETQSTATSV1: 3001 case DIOCGETRULESETS: 3002 case DIOCGETRULESET: 3003 case DIOCNATLOOK: 3004 case DIOCRGETTABLES: 3005 case DIOCRGETTSTATS: 3006 case DIOCRGETADDRS: 3007 case DIOCRGETASTATS: 3008 case DIOCRTSTADDRS: 3009 case DIOCOSFPGET: 3010 case DIOCGETSRCNODES: 3011 case DIOCGETSYNCOOKIES: 3012 case DIOCIGETIFACES: 3013 case DIOCGIFSPEEDV1: 3014 case DIOCGIFSPEEDV0: 3015 case DIOCGETRULENV: 3016 case DIOCGETETHRULES: 3017 case DIOCGETETHRULE: 3018 case DIOCGETETHRULESETS: 3019 case DIOCGETETHRULESET: 3020 break; 3021 case DIOCRCLRTABLES: 3022 case DIOCRADDTABLES: 3023 case DIOCRDELTABLES: 3024 case DIOCRCLRTSTATS: 3025 case DIOCRCLRADDRS: 3026 case DIOCRADDADDRS: 3027 case DIOCRDELADDRS: 3028 case DIOCRSETADDRS: 3029 case DIOCRSETTFLAGS: 3030 if (((struct pfioc_table *)addr)->pfrio_flags & 3031 PFR_FLAG_DUMMY) { 3032 flags |= FWRITE; /* need write lock for dummy */ 3033 break; /* dummy operation ok */ 3034 } 3035 return (EACCES); 3036 default: 3037 return (EACCES); 3038 } 3039 3040 CURVNET_SET(TD_TO_VNET(td)); 3041 3042 switch (cmd) { 3043 #ifdef COMPAT_FREEBSD14 3044 case DIOCSTART: 3045 error = pf_start(); 3046 break; 3047 3048 case DIOCSTOP: 3049 error = pf_stop(); 3050 break; 3051 #endif 3052 3053 case DIOCGETETHRULES: { 3054 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3055 nvlist_t *nvl; 3056 void *packed; 3057 struct pf_keth_rule *tail; 3058 struct pf_keth_ruleset *rs; 3059 u_int32_t ticket, nr; 3060 const char *anchor = ""; 3061 3062 nvl = NULL; 3063 packed = NULL; 3064 3065 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x) 3066 3067 if (nv->len > pf_ioctl_maxcount) 3068 ERROUT(ENOMEM); 3069 3070 /* Copy the request in */ 3071 packed = malloc(nv->len, M_NVLIST, M_WAITOK); 3072 error = copyin(nv->data, packed, nv->len); 3073 if (error) 3074 ERROUT(error); 3075 3076 nvl = nvlist_unpack(packed, nv->len, 0); 3077 if (nvl == NULL) 3078 ERROUT(EBADMSG); 3079 3080 if (! nvlist_exists_string(nvl, "anchor")) 3081 ERROUT(EBADMSG); 3082 3083 anchor = nvlist_get_string(nvl, "anchor"); 3084 3085 rs = pf_find_keth_ruleset(anchor); 3086 3087 nvlist_destroy(nvl); 3088 nvl = NULL; 3089 free(packed, M_NVLIST); 3090 packed = NULL; 3091 3092 if (rs == NULL) 3093 ERROUT(ENOENT); 3094 3095 /* Reply */ 3096 nvl = nvlist_create(0); 3097 if (nvl == NULL) 3098 ERROUT(ENOMEM); 3099 3100 PF_RULES_RLOCK(); 3101 3102 ticket = rs->active.ticket; 3103 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq); 3104 if (tail) 3105 nr = tail->nr + 1; 3106 else 3107 nr = 0; 3108 3109 PF_RULES_RUNLOCK(); 3110 3111 nvlist_add_number(nvl, "ticket", ticket); 3112 nvlist_add_number(nvl, "nr", nr); 3113 3114 packed = nvlist_pack(nvl, &nv->len); 3115 if (packed == NULL) 3116 ERROUT(ENOMEM); 3117 3118 if (nv->size == 0) 3119 ERROUT(0); 3120 else if (nv->size < nv->len) 3121 ERROUT(ENOSPC); 3122 3123 error = copyout(packed, nv->data, nv->len); 3124 3125 #undef ERROUT 3126 DIOCGETETHRULES_error: 3127 free(packed, M_NVLIST); 3128 nvlist_destroy(nvl); 3129 break; 3130 } 3131 3132 case DIOCGETETHRULE: { 3133 struct epoch_tracker et; 3134 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3135 nvlist_t *nvl = NULL; 3136 void *nvlpacked = NULL; 3137 struct pf_keth_rule *rule = NULL; 3138 struct pf_keth_ruleset *rs; 3139 u_int32_t ticket, nr; 3140 bool clear = false; 3141 const char *anchor; 3142 3143 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x) 3144 3145 if (nv->len > pf_ioctl_maxcount) 3146 ERROUT(ENOMEM); 3147 3148 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3149 error = copyin(nv->data, nvlpacked, nv->len); 3150 if (error) 3151 ERROUT(error); 3152 3153 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3154 if (nvl == NULL) 3155 ERROUT(EBADMSG); 3156 if (! nvlist_exists_number(nvl, "ticket")) 3157 ERROUT(EBADMSG); 3158 ticket = nvlist_get_number(nvl, "ticket"); 3159 if (! nvlist_exists_string(nvl, "anchor")) 3160 ERROUT(EBADMSG); 3161 anchor = nvlist_get_string(nvl, "anchor"); 3162 3163 if (nvlist_exists_bool(nvl, "clear")) 3164 clear = nvlist_get_bool(nvl, "clear"); 3165 3166 if (clear && !(flags & FWRITE)) 3167 ERROUT(EACCES); 3168 3169 if (! nvlist_exists_number(nvl, "nr")) 3170 ERROUT(EBADMSG); 3171 nr = nvlist_get_number(nvl, "nr"); 3172 3173 PF_RULES_RLOCK(); 3174 rs = pf_find_keth_ruleset(anchor); 3175 if (rs == NULL) { 3176 PF_RULES_RUNLOCK(); 3177 ERROUT(ENOENT); 3178 } 3179 if (ticket != rs->active.ticket) { 3180 PF_RULES_RUNLOCK(); 3181 ERROUT(EBUSY); 3182 } 3183 3184 nvlist_destroy(nvl); 3185 nvl = NULL; 3186 free(nvlpacked, M_NVLIST); 3187 nvlpacked = NULL; 3188 3189 rule = TAILQ_FIRST(rs->active.rules); 3190 while ((rule != NULL) && (rule->nr != nr)) 3191 rule = TAILQ_NEXT(rule, entries); 3192 if (rule == NULL) { 3193 PF_RULES_RUNLOCK(); 3194 ERROUT(ENOENT); 3195 } 3196 /* Make sure rule can't go away. */ 3197 NET_EPOCH_ENTER(et); 3198 PF_RULES_RUNLOCK(); 3199 nvl = pf_keth_rule_to_nveth_rule(rule); 3200 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) { 3201 NET_EPOCH_EXIT(et); 3202 ERROUT(EBUSY); 3203 } 3204 NET_EPOCH_EXIT(et); 3205 if (nvl == NULL) 3206 ERROUT(ENOMEM); 3207 3208 nvlpacked = nvlist_pack(nvl, &nv->len); 3209 if (nvlpacked == NULL) 3210 ERROUT(ENOMEM); 3211 3212 if (nv->size == 0) 3213 ERROUT(0); 3214 else if (nv->size < nv->len) 3215 ERROUT(ENOSPC); 3216 3217 error = copyout(nvlpacked, nv->data, nv->len); 3218 if (error == 0 && clear) { 3219 counter_u64_zero(rule->evaluations); 3220 for (int i = 0; i < 2; i++) { 3221 counter_u64_zero(rule->packets[i]); 3222 counter_u64_zero(rule->bytes[i]); 3223 } 3224 } 3225 3226 #undef ERROUT 3227 DIOCGETETHRULE_error: 3228 free(nvlpacked, M_NVLIST); 3229 nvlist_destroy(nvl); 3230 break; 3231 } 3232 3233 case DIOCADDETHRULE: { 3234 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3235 nvlist_t *nvl = NULL; 3236 void *nvlpacked = NULL; 3237 struct pf_keth_rule *rule = NULL, *tail = NULL; 3238 struct pf_keth_ruleset *ruleset = NULL; 3239 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL; 3240 const char *anchor = "", *anchor_call = ""; 3241 3242 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x) 3243 3244 if (nv->len > pf_ioctl_maxcount) 3245 ERROUT(ENOMEM); 3246 3247 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3248 error = copyin(nv->data, nvlpacked, nv->len); 3249 if (error) 3250 ERROUT(error); 3251 3252 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3253 if (nvl == NULL) 3254 ERROUT(EBADMSG); 3255 3256 if (! nvlist_exists_number(nvl, "ticket")) 3257 ERROUT(EBADMSG); 3258 3259 if (nvlist_exists_string(nvl, "anchor")) 3260 anchor = nvlist_get_string(nvl, "anchor"); 3261 if (nvlist_exists_string(nvl, "anchor_call")) 3262 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3263 3264 ruleset = pf_find_keth_ruleset(anchor); 3265 if (ruleset == NULL) 3266 ERROUT(EINVAL); 3267 3268 if (nvlist_get_number(nvl, "ticket") != 3269 ruleset->inactive.ticket) { 3270 DPFPRINTF(PF_DEBUG_MISC, 3271 "ticket: %d != %d", 3272 (u_int32_t)nvlist_get_number(nvl, "ticket"), 3273 ruleset->inactive.ticket); 3274 ERROUT(EBUSY); 3275 } 3276 3277 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); 3278 rule->timestamp = NULL; 3279 3280 error = pf_nveth_rule_to_keth_rule(nvl, rule); 3281 if (error != 0) 3282 ERROUT(error); 3283 3284 if (rule->ifname[0]) 3285 kif = pf_kkif_create(M_WAITOK); 3286 if (rule->bridge_to_name[0]) 3287 bridge_to_kif = pf_kkif_create(M_WAITOK); 3288 rule->evaluations = counter_u64_alloc(M_WAITOK); 3289 for (int i = 0; i < 2; i++) { 3290 rule->packets[i] = counter_u64_alloc(M_WAITOK); 3291 rule->bytes[i] = counter_u64_alloc(M_WAITOK); 3292 } 3293 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 3294 M_WAITOK | M_ZERO); 3295 3296 PF_RULES_WLOCK(); 3297 3298 if (rule->ifname[0]) { 3299 rule->kif = pfi_kkif_attach(kif, rule->ifname); 3300 pfi_kkif_ref(rule->kif); 3301 } else 3302 rule->kif = NULL; 3303 if (rule->bridge_to_name[0]) { 3304 rule->bridge_to = pfi_kkif_attach(bridge_to_kif, 3305 rule->bridge_to_name); 3306 pfi_kkif_ref(rule->bridge_to); 3307 } else 3308 rule->bridge_to = NULL; 3309 3310 #ifdef ALTQ 3311 /* set queue IDs */ 3312 if (rule->qname[0] != 0) { 3313 if ((rule->qid = pf_qname2qid(rule->qname)) == 0) 3314 error = EBUSY; 3315 else 3316 rule->qid = rule->qid; 3317 } 3318 #endif 3319 if (rule->tagname[0]) 3320 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 3321 error = EBUSY; 3322 if (rule->match_tagname[0]) 3323 if ((rule->match_tag = pf_tagname2tag( 3324 rule->match_tagname)) == 0) 3325 error = EBUSY; 3326 3327 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE) 3328 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr); 3329 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE) 3330 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr); 3331 3332 if (error) { 3333 pf_free_eth_rule(rule); 3334 PF_RULES_WUNLOCK(); 3335 ERROUT(error); 3336 } 3337 3338 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) { 3339 pf_free_eth_rule(rule); 3340 PF_RULES_WUNLOCK(); 3341 ERROUT(EINVAL); 3342 } 3343 3344 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq); 3345 if (tail) 3346 rule->nr = tail->nr + 1; 3347 else 3348 rule->nr = 0; 3349 3350 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries); 3351 3352 PF_RULES_WUNLOCK(); 3353 3354 #undef ERROUT 3355 DIOCADDETHRULE_error: 3356 nvlist_destroy(nvl); 3357 free(nvlpacked, M_NVLIST); 3358 break; 3359 } 3360 3361 case DIOCGETETHRULESETS: { 3362 struct epoch_tracker et; 3363 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3364 nvlist_t *nvl = NULL; 3365 void *nvlpacked = NULL; 3366 struct pf_keth_ruleset *ruleset; 3367 struct pf_keth_anchor *anchor; 3368 int nr = 0; 3369 3370 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x) 3371 3372 if (nv->len > pf_ioctl_maxcount) 3373 ERROUT(ENOMEM); 3374 3375 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3376 error = copyin(nv->data, nvlpacked, nv->len); 3377 if (error) 3378 ERROUT(error); 3379 3380 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3381 if (nvl == NULL) 3382 ERROUT(EBADMSG); 3383 if (! nvlist_exists_string(nvl, "path")) 3384 ERROUT(EBADMSG); 3385 3386 NET_EPOCH_ENTER(et); 3387 3388 if ((ruleset = pf_find_keth_ruleset( 3389 nvlist_get_string(nvl, "path"))) == NULL) { 3390 NET_EPOCH_EXIT(et); 3391 ERROUT(ENOENT); 3392 } 3393 3394 if (ruleset->anchor == NULL) { 3395 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors) 3396 if (anchor->parent == NULL) 3397 nr++; 3398 } else { 3399 RB_FOREACH(anchor, pf_keth_anchor_node, 3400 &ruleset->anchor->children) 3401 nr++; 3402 } 3403 3404 NET_EPOCH_EXIT(et); 3405 3406 nvlist_destroy(nvl); 3407 nvl = NULL; 3408 free(nvlpacked, M_NVLIST); 3409 nvlpacked = NULL; 3410 3411 nvl = nvlist_create(0); 3412 if (nvl == NULL) 3413 ERROUT(ENOMEM); 3414 3415 nvlist_add_number(nvl, "nr", nr); 3416 3417 nvlpacked = nvlist_pack(nvl, &nv->len); 3418 if (nvlpacked == NULL) 3419 ERROUT(ENOMEM); 3420 3421 if (nv->size == 0) 3422 ERROUT(0); 3423 else if (nv->size < nv->len) 3424 ERROUT(ENOSPC); 3425 3426 error = copyout(nvlpacked, nv->data, nv->len); 3427 3428 #undef ERROUT 3429 DIOCGETETHRULESETS_error: 3430 free(nvlpacked, M_NVLIST); 3431 nvlist_destroy(nvl); 3432 break; 3433 } 3434 3435 case DIOCGETETHRULESET: { 3436 struct epoch_tracker et; 3437 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3438 nvlist_t *nvl = NULL; 3439 void *nvlpacked = NULL; 3440 struct pf_keth_ruleset *ruleset; 3441 struct pf_keth_anchor *anchor; 3442 int nr = 0, req_nr = 0; 3443 bool found = false; 3444 3445 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x) 3446 3447 if (nv->len > pf_ioctl_maxcount) 3448 ERROUT(ENOMEM); 3449 3450 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3451 error = copyin(nv->data, nvlpacked, nv->len); 3452 if (error) 3453 ERROUT(error); 3454 3455 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3456 if (nvl == NULL) 3457 ERROUT(EBADMSG); 3458 if (! nvlist_exists_string(nvl, "path")) 3459 ERROUT(EBADMSG); 3460 if (! nvlist_exists_number(nvl, "nr")) 3461 ERROUT(EBADMSG); 3462 3463 req_nr = nvlist_get_number(nvl, "nr"); 3464 3465 NET_EPOCH_ENTER(et); 3466 3467 if ((ruleset = pf_find_keth_ruleset( 3468 nvlist_get_string(nvl, "path"))) == NULL) { 3469 NET_EPOCH_EXIT(et); 3470 ERROUT(ENOENT); 3471 } 3472 3473 nvlist_destroy(nvl); 3474 nvl = NULL; 3475 free(nvlpacked, M_NVLIST); 3476 nvlpacked = NULL; 3477 3478 nvl = nvlist_create(0); 3479 if (nvl == NULL) { 3480 NET_EPOCH_EXIT(et); 3481 ERROUT(ENOMEM); 3482 } 3483 3484 if (ruleset->anchor == NULL) { 3485 RB_FOREACH(anchor, pf_keth_anchor_global, 3486 &V_pf_keth_anchors) { 3487 if (anchor->parent == NULL && nr++ == req_nr) { 3488 found = true; 3489 break; 3490 } 3491 } 3492 } else { 3493 RB_FOREACH(anchor, pf_keth_anchor_node, 3494 &ruleset->anchor->children) { 3495 if (nr++ == req_nr) { 3496 found = true; 3497 break; 3498 } 3499 } 3500 } 3501 3502 NET_EPOCH_EXIT(et); 3503 if (found) { 3504 nvlist_add_number(nvl, "nr", nr); 3505 nvlist_add_string(nvl, "name", anchor->name); 3506 if (ruleset->anchor) 3507 nvlist_add_string(nvl, "path", 3508 ruleset->anchor->path); 3509 else 3510 nvlist_add_string(nvl, "path", ""); 3511 } else { 3512 ERROUT(EBUSY); 3513 } 3514 3515 nvlpacked = nvlist_pack(nvl, &nv->len); 3516 if (nvlpacked == NULL) 3517 ERROUT(ENOMEM); 3518 3519 if (nv->size == 0) 3520 ERROUT(0); 3521 else if (nv->size < nv->len) 3522 ERROUT(ENOSPC); 3523 3524 error = copyout(nvlpacked, nv->data, nv->len); 3525 3526 #undef ERROUT 3527 DIOCGETETHRULESET_error: 3528 free(nvlpacked, M_NVLIST); 3529 nvlist_destroy(nvl); 3530 break; 3531 } 3532 3533 case DIOCADDRULENV: { 3534 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3535 nvlist_t *nvl = NULL; 3536 void *nvlpacked = NULL; 3537 struct pf_krule *rule = NULL; 3538 const char *anchor = "", *anchor_call = ""; 3539 uint32_t ticket = 0, pool_ticket = 0; 3540 3541 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x) 3542 3543 if (nv->len > pf_ioctl_maxcount) 3544 ERROUT(ENOMEM); 3545 3546 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3547 error = copyin(nv->data, nvlpacked, nv->len); 3548 if (error) 3549 ERROUT(error); 3550 3551 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3552 if (nvl == NULL) 3553 ERROUT(EBADMSG); 3554 3555 if (! nvlist_exists_number(nvl, "ticket")) 3556 ERROUT(EINVAL); 3557 ticket = nvlist_get_number(nvl, "ticket"); 3558 3559 if (! nvlist_exists_number(nvl, "pool_ticket")) 3560 ERROUT(EINVAL); 3561 pool_ticket = nvlist_get_number(nvl, "pool_ticket"); 3562 3563 if (! nvlist_exists_nvlist(nvl, "rule")) 3564 ERROUT(EINVAL); 3565 3566 rule = pf_krule_alloc(); 3567 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"), 3568 rule); 3569 if (error) 3570 ERROUT(error); 3571 3572 if (nvlist_exists_string(nvl, "anchor")) 3573 anchor = nvlist_get_string(nvl, "anchor"); 3574 if (nvlist_exists_string(nvl, "anchor_call")) 3575 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3576 3577 if ((error = nvlist_error(nvl))) 3578 ERROUT(error); 3579 3580 /* Frees rule on error */ 3581 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor, 3582 anchor_call, td->td_ucred->cr_ruid, 3583 td->td_proc ? td->td_proc->p_pid : 0); 3584 3585 nvlist_destroy(nvl); 3586 free(nvlpacked, M_NVLIST); 3587 break; 3588 #undef ERROUT 3589 DIOCADDRULENV_error: 3590 pf_krule_free(rule); 3591 nvlist_destroy(nvl); 3592 free(nvlpacked, M_NVLIST); 3593 3594 break; 3595 } 3596 case DIOCADDRULE: { 3597 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3598 struct pf_krule *rule; 3599 3600 rule = pf_krule_alloc(); 3601 error = pf_rule_to_krule(&pr->rule, rule); 3602 if (error != 0) { 3603 pf_krule_free(rule); 3604 goto fail; 3605 } 3606 3607 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3608 3609 /* Frees rule on error */ 3610 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, 3611 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid, 3612 td->td_proc ? td->td_proc->p_pid : 0); 3613 break; 3614 } 3615 3616 case DIOCGETRULES: { 3617 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3618 3619 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3620 3621 error = pf_ioctl_getrules(pr); 3622 3623 break; 3624 } 3625 3626 case DIOCGETRULENV: { 3627 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3628 nvlist_t *nvrule = NULL; 3629 nvlist_t *nvl = NULL; 3630 struct pf_kruleset *ruleset; 3631 struct pf_krule *rule; 3632 void *nvlpacked = NULL; 3633 int rs_num, nr; 3634 bool clear_counter = false; 3635 3636 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x) 3637 3638 if (nv->len > pf_ioctl_maxcount) 3639 ERROUT(ENOMEM); 3640 3641 /* Copy the request in */ 3642 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3643 error = copyin(nv->data, nvlpacked, nv->len); 3644 if (error) 3645 ERROUT(error); 3646 3647 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3648 if (nvl == NULL) 3649 ERROUT(EBADMSG); 3650 3651 if (! nvlist_exists_string(nvl, "anchor")) 3652 ERROUT(EBADMSG); 3653 if (! nvlist_exists_number(nvl, "ruleset")) 3654 ERROUT(EBADMSG); 3655 if (! nvlist_exists_number(nvl, "ticket")) 3656 ERROUT(EBADMSG); 3657 if (! nvlist_exists_number(nvl, "nr")) 3658 ERROUT(EBADMSG); 3659 3660 if (nvlist_exists_bool(nvl, "clear_counter")) 3661 clear_counter = nvlist_get_bool(nvl, "clear_counter"); 3662 3663 if (clear_counter && !(flags & FWRITE)) 3664 ERROUT(EACCES); 3665 3666 nr = nvlist_get_number(nvl, "nr"); 3667 3668 PF_RULES_WLOCK(); 3669 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor")); 3670 if (ruleset == NULL) { 3671 PF_RULES_WUNLOCK(); 3672 ERROUT(ENOENT); 3673 } 3674 3675 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset")); 3676 if (rs_num >= PF_RULESET_MAX) { 3677 PF_RULES_WUNLOCK(); 3678 ERROUT(EINVAL); 3679 } 3680 3681 if (nvlist_get_number(nvl, "ticket") != 3682 ruleset->rules[rs_num].active.ticket) { 3683 PF_RULES_WUNLOCK(); 3684 ERROUT(EBUSY); 3685 } 3686 3687 if ((error = nvlist_error(nvl))) { 3688 PF_RULES_WUNLOCK(); 3689 ERROUT(error); 3690 } 3691 3692 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3693 while ((rule != NULL) && (rule->nr != nr)) 3694 rule = TAILQ_NEXT(rule, entries); 3695 if (rule == NULL) { 3696 PF_RULES_WUNLOCK(); 3697 ERROUT(EBUSY); 3698 } 3699 3700 nvrule = pf_krule_to_nvrule(rule); 3701 3702 nvlist_destroy(nvl); 3703 nvl = nvlist_create(0); 3704 if (nvl == NULL) { 3705 PF_RULES_WUNLOCK(); 3706 ERROUT(ENOMEM); 3707 } 3708 nvlist_add_number(nvl, "nr", nr); 3709 nvlist_add_nvlist(nvl, "rule", nvrule); 3710 nvlist_destroy(nvrule); 3711 nvrule = NULL; 3712 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) { 3713 PF_RULES_WUNLOCK(); 3714 ERROUT(EBUSY); 3715 } 3716 3717 free(nvlpacked, M_NVLIST); 3718 nvlpacked = nvlist_pack(nvl, &nv->len); 3719 if (nvlpacked == NULL) { 3720 PF_RULES_WUNLOCK(); 3721 ERROUT(ENOMEM); 3722 } 3723 3724 if (nv->size == 0) { 3725 PF_RULES_WUNLOCK(); 3726 ERROUT(0); 3727 } 3728 else if (nv->size < nv->len) { 3729 PF_RULES_WUNLOCK(); 3730 ERROUT(ENOSPC); 3731 } 3732 3733 if (clear_counter) 3734 pf_krule_clear_counters(rule); 3735 3736 PF_RULES_WUNLOCK(); 3737 3738 error = copyout(nvlpacked, nv->data, nv->len); 3739 3740 #undef ERROUT 3741 DIOCGETRULENV_error: 3742 free(nvlpacked, M_NVLIST); 3743 nvlist_destroy(nvrule); 3744 nvlist_destroy(nvl); 3745 3746 break; 3747 } 3748 3749 case DIOCCHANGERULE: { 3750 struct pfioc_rule *pcr = (struct pfioc_rule *)addr; 3751 struct pf_kruleset *ruleset; 3752 struct pf_krule *oldrule = NULL, *newrule = NULL; 3753 struct pfi_kkif *kif = NULL; 3754 struct pf_kpooladdr *pa; 3755 u_int32_t nr = 0; 3756 int rs_num; 3757 3758 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0'; 3759 3760 if (pcr->action < PF_CHANGE_ADD_HEAD || 3761 pcr->action > PF_CHANGE_GET_TICKET) { 3762 error = EINVAL; 3763 goto fail; 3764 } 3765 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { 3766 error = EINVAL; 3767 goto fail; 3768 } 3769 3770 if (pcr->action != PF_CHANGE_REMOVE) { 3771 newrule = pf_krule_alloc(); 3772 error = pf_rule_to_krule(&pcr->rule, newrule); 3773 if (error != 0) { 3774 pf_krule_free(newrule); 3775 goto fail; 3776 } 3777 3778 if ((error = pf_rule_checkaf(newrule))) { 3779 pf_krule_free(newrule); 3780 goto fail; 3781 } 3782 if (newrule->ifname[0]) 3783 kif = pf_kkif_create(M_WAITOK); 3784 pf_counter_u64_init(&newrule->evaluations, M_WAITOK); 3785 for (int i = 0; i < 2; i++) { 3786 pf_counter_u64_init(&newrule->packets[i], M_WAITOK); 3787 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK); 3788 } 3789 newrule->states_cur = counter_u64_alloc(M_WAITOK); 3790 newrule->states_tot = counter_u64_alloc(M_WAITOK); 3791 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 3792 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 3793 newrule->cuid = td->td_ucred->cr_ruid; 3794 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 3795 TAILQ_INIT(&newrule->nat.list); 3796 TAILQ_INIT(&newrule->rdr.list); 3797 TAILQ_INIT(&newrule->route.list); 3798 } 3799 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x) 3800 3801 PF_CONFIG_LOCK(); 3802 PF_RULES_WLOCK(); 3803 #ifdef PF_WANT_32_TO_64_COUNTER 3804 if (newrule != NULL) { 3805 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist); 3806 newrule->allrulelinked = true; 3807 V_pf_allrulecount++; 3808 } 3809 #endif 3810 3811 if (!(pcr->action == PF_CHANGE_REMOVE || 3812 pcr->action == PF_CHANGE_GET_TICKET) && 3813 pcr->pool_ticket != V_ticket_pabuf) 3814 ERROUT(EBUSY); 3815 3816 ruleset = pf_find_kruleset(pcr->anchor); 3817 if (ruleset == NULL) 3818 ERROUT(EINVAL); 3819 3820 rs_num = pf_get_ruleset_number(pcr->rule.action); 3821 if (rs_num >= PF_RULESET_MAX) 3822 ERROUT(EINVAL); 3823 3824 /* 3825 * XXXMJG: there is no guarantee that the ruleset was 3826 * created by the usual route of calling DIOCXBEGIN. 3827 * As a result it is possible the rule tree will not 3828 * be allocated yet. Hack around it by doing it here. 3829 * Note it is fine to let the tree persist in case of 3830 * error as it will be freed down the road on future 3831 * updates (if need be). 3832 */ 3833 if (ruleset->rules[rs_num].active.tree == NULL) { 3834 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT); 3835 if (ruleset->rules[rs_num].active.tree == NULL) { 3836 ERROUT(ENOMEM); 3837 } 3838 } 3839 3840 if (pcr->action == PF_CHANGE_GET_TICKET) { 3841 pcr->ticket = ++ruleset->rules[rs_num].active.ticket; 3842 ERROUT(0); 3843 } else if (pcr->ticket != 3844 ruleset->rules[rs_num].active.ticket) 3845 ERROUT(EINVAL); 3846 3847 if (pcr->action != PF_CHANGE_REMOVE) { 3848 if (newrule->ifname[0]) { 3849 newrule->kif = pfi_kkif_attach(kif, 3850 newrule->ifname); 3851 kif = NULL; 3852 pfi_kkif_ref(newrule->kif); 3853 } else 3854 newrule->kif = NULL; 3855 3856 if (newrule->rtableid > 0 && 3857 newrule->rtableid >= rt_numfibs) 3858 error = EBUSY; 3859 3860 #ifdef ALTQ 3861 /* set queue IDs */ 3862 if (newrule->qname[0] != 0) { 3863 if ((newrule->qid = 3864 pf_qname2qid(newrule->qname)) == 0) 3865 error = EBUSY; 3866 else if (newrule->pqname[0] != 0) { 3867 if ((newrule->pqid = 3868 pf_qname2qid(newrule->pqname)) == 0) 3869 error = EBUSY; 3870 } else 3871 newrule->pqid = newrule->qid; 3872 } 3873 #endif /* ALTQ */ 3874 if (newrule->tagname[0]) 3875 if ((newrule->tag = 3876 pf_tagname2tag(newrule->tagname)) == 0) 3877 error = EBUSY; 3878 if (newrule->match_tagname[0]) 3879 if ((newrule->match_tag = pf_tagname2tag( 3880 newrule->match_tagname)) == 0) 3881 error = EBUSY; 3882 if (newrule->rt && !newrule->direction) 3883 error = EINVAL; 3884 if (!newrule->log) 3885 newrule->logif = 0; 3886 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) 3887 error = ENOMEM; 3888 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) 3889 error = ENOMEM; 3890 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call)) 3891 error = EINVAL; 3892 for (int i = 0; i < 3; i++) { 3893 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 3894 if (pa->addr.type == PF_ADDR_TABLE) { 3895 pa->addr.p.tbl = 3896 pfr_attach_table(ruleset, 3897 pa->addr.v.tblname); 3898 if (pa->addr.p.tbl == NULL) 3899 error = ENOMEM; 3900 } 3901 } 3902 3903 newrule->overload_tbl = NULL; 3904 if (newrule->overload_tblname[0]) { 3905 if ((newrule->overload_tbl = pfr_attach_table( 3906 ruleset, newrule->overload_tblname)) == 3907 NULL) 3908 error = EINVAL; 3909 else 3910 newrule->overload_tbl->pfrkt_flags |= 3911 PFR_TFLAG_ACTIVE; 3912 } 3913 3914 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list); 3915 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list); 3916 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list); 3917 if (((((newrule->action == PF_NAT) || 3918 (newrule->action == PF_RDR) || 3919 (newrule->action == PF_BINAT) || 3920 (newrule->rt > PF_NOPFROUTE)) && 3921 !newrule->anchor)) && 3922 (TAILQ_FIRST(&newrule->rdr.list) == NULL)) 3923 error = EINVAL; 3924 3925 if (error) { 3926 pf_free_rule(newrule); 3927 PF_RULES_WUNLOCK(); 3928 PF_CONFIG_UNLOCK(); 3929 goto fail; 3930 } 3931 3932 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list); 3933 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list); 3934 } 3935 pf_empty_kpool(&V_pf_pabuf[0]); 3936 pf_empty_kpool(&V_pf_pabuf[1]); 3937 pf_empty_kpool(&V_pf_pabuf[2]); 3938 3939 if (pcr->action == PF_CHANGE_ADD_HEAD) 3940 oldrule = TAILQ_FIRST( 3941 ruleset->rules[rs_num].active.ptr); 3942 else if (pcr->action == PF_CHANGE_ADD_TAIL) 3943 oldrule = TAILQ_LAST( 3944 ruleset->rules[rs_num].active.ptr, pf_krulequeue); 3945 else { 3946 oldrule = TAILQ_FIRST( 3947 ruleset->rules[rs_num].active.ptr); 3948 while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) 3949 oldrule = TAILQ_NEXT(oldrule, entries); 3950 if (oldrule == NULL) { 3951 if (newrule != NULL) 3952 pf_free_rule(newrule); 3953 PF_RULES_WUNLOCK(); 3954 PF_CONFIG_UNLOCK(); 3955 error = EINVAL; 3956 goto fail; 3957 } 3958 } 3959 3960 if (pcr->action == PF_CHANGE_REMOVE) { 3961 pf_unlink_rule(ruleset->rules[rs_num].active.ptr, 3962 oldrule); 3963 RB_REMOVE(pf_krule_global, 3964 ruleset->rules[rs_num].active.tree, oldrule); 3965 ruleset->rules[rs_num].active.rcount--; 3966 } else { 3967 pf_hash_rule(newrule); 3968 if (RB_INSERT(pf_krule_global, 3969 ruleset->rules[rs_num].active.tree, newrule) != NULL) { 3970 pf_free_rule(newrule); 3971 PF_RULES_WUNLOCK(); 3972 PF_CONFIG_UNLOCK(); 3973 error = EEXIST; 3974 goto fail; 3975 } 3976 3977 if (oldrule == NULL) 3978 TAILQ_INSERT_TAIL( 3979 ruleset->rules[rs_num].active.ptr, 3980 newrule, entries); 3981 else if (pcr->action == PF_CHANGE_ADD_HEAD || 3982 pcr->action == PF_CHANGE_ADD_BEFORE) 3983 TAILQ_INSERT_BEFORE(oldrule, newrule, entries); 3984 else 3985 TAILQ_INSERT_AFTER( 3986 ruleset->rules[rs_num].active.ptr, 3987 oldrule, newrule, entries); 3988 ruleset->rules[rs_num].active.rcount++; 3989 } 3990 3991 nr = 0; 3992 TAILQ_FOREACH(oldrule, 3993 ruleset->rules[rs_num].active.ptr, entries) 3994 oldrule->nr = nr++; 3995 3996 ruleset->rules[rs_num].active.ticket++; 3997 3998 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); 3999 pf_remove_if_empty_kruleset(ruleset); 4000 4001 PF_RULES_WUNLOCK(); 4002 PF_CONFIG_UNLOCK(); 4003 break; 4004 4005 #undef ERROUT 4006 DIOCCHANGERULE_error: 4007 PF_RULES_WUNLOCK(); 4008 PF_CONFIG_UNLOCK(); 4009 pf_krule_free(newrule); 4010 pf_kkif_free(kif); 4011 break; 4012 } 4013 4014 case DIOCCLRSTATESNV: { 4015 error = pf_clearstates_nv((struct pfioc_nv *)addr); 4016 break; 4017 } 4018 4019 case DIOCKILLSTATESNV: { 4020 error = pf_killstates_nv((struct pfioc_nv *)addr); 4021 break; 4022 } 4023 4024 case DIOCADDSTATE: { 4025 struct pfioc_state *ps = (struct pfioc_state *)addr; 4026 struct pfsync_state_1301 *sp = &ps->state; 4027 4028 if (sp->timeout >= PFTM_MAX) { 4029 error = EINVAL; 4030 goto fail; 4031 } 4032 if (V_pfsync_state_import_ptr != NULL) { 4033 PF_RULES_RLOCK(); 4034 error = V_pfsync_state_import_ptr( 4035 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL, 4036 PFSYNC_MSG_VERSION_1301); 4037 PF_RULES_RUNLOCK(); 4038 } else 4039 error = EOPNOTSUPP; 4040 break; 4041 } 4042 4043 case DIOCGETSTATE: { 4044 struct pfioc_state *ps = (struct pfioc_state *)addr; 4045 struct pf_kstate *s; 4046 4047 s = pf_find_state_byid(ps->state.id, ps->state.creatorid); 4048 if (s == NULL) { 4049 error = ENOENT; 4050 goto fail; 4051 } 4052 4053 pfsync_state_export((union pfsync_state_union*)&ps->state, 4054 s, PFSYNC_MSG_VERSION_1301); 4055 PF_STATE_UNLOCK(s); 4056 break; 4057 } 4058 4059 case DIOCGETSTATENV: { 4060 error = pf_getstate((struct pfioc_nv *)addr); 4061 break; 4062 } 4063 4064 #ifdef COMPAT_FREEBSD14 4065 case DIOCGETSTATES: { 4066 struct pfioc_states *ps = (struct pfioc_states *)addr; 4067 struct pf_kstate *s; 4068 struct pfsync_state_1301 *pstore, *p; 4069 int i, nr; 4070 size_t slice_count = 16, count; 4071 void *out; 4072 4073 if (ps->ps_len <= 0) { 4074 nr = uma_zone_get_cur(V_pf_state_z); 4075 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4076 break; 4077 } 4078 4079 out = ps->ps_states; 4080 pstore = mallocarray(slice_count, 4081 sizeof(struct pfsync_state_1301), M_PF, M_WAITOK | M_ZERO); 4082 nr = 0; 4083 4084 for (i = 0; i <= V_pf_hashmask; i++) { 4085 struct pf_idhash *ih = &V_pf_idhash[i]; 4086 4087 DIOCGETSTATES_retry: 4088 p = pstore; 4089 4090 if (LIST_EMPTY(&ih->states)) 4091 continue; 4092 4093 PF_HASHROW_LOCK(ih); 4094 count = 0; 4095 LIST_FOREACH(s, &ih->states, entry) { 4096 if (s->timeout == PFTM_UNLINKED) 4097 continue; 4098 count++; 4099 } 4100 4101 if (count > slice_count) { 4102 PF_HASHROW_UNLOCK(ih); 4103 free(pstore, M_PF); 4104 slice_count = count * 2; 4105 pstore = mallocarray(slice_count, 4106 sizeof(struct pfsync_state_1301), M_PF, 4107 M_WAITOK | M_ZERO); 4108 goto DIOCGETSTATES_retry; 4109 } 4110 4111 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4112 PF_HASHROW_UNLOCK(ih); 4113 goto DIOCGETSTATES_full; 4114 } 4115 4116 LIST_FOREACH(s, &ih->states, entry) { 4117 if (s->timeout == PFTM_UNLINKED) 4118 continue; 4119 4120 pfsync_state_export((union pfsync_state_union*)p, 4121 s, PFSYNC_MSG_VERSION_1301); 4122 p++; 4123 nr++; 4124 } 4125 PF_HASHROW_UNLOCK(ih); 4126 error = copyout(pstore, out, 4127 sizeof(struct pfsync_state_1301) * count); 4128 if (error) { 4129 free(pstore, M_PF); 4130 goto fail; 4131 } 4132 out = ps->ps_states + nr; 4133 } 4134 DIOCGETSTATES_full: 4135 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4136 free(pstore, M_PF); 4137 4138 break; 4139 } 4140 4141 case DIOCGETSTATESV2: { 4142 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr; 4143 struct pf_kstate *s; 4144 struct pf_state_export *pstore, *p; 4145 int i, nr; 4146 size_t slice_count = 16, count; 4147 void *out; 4148 4149 if (ps->ps_req_version > PF_STATE_VERSION) { 4150 error = ENOTSUP; 4151 goto fail; 4152 } 4153 4154 if (ps->ps_len <= 0) { 4155 nr = uma_zone_get_cur(V_pf_state_z); 4156 ps->ps_len = sizeof(struct pf_state_export) * nr; 4157 break; 4158 } 4159 4160 out = ps->ps_states; 4161 pstore = mallocarray(slice_count, 4162 sizeof(struct pf_state_export), M_PF, M_WAITOK | M_ZERO); 4163 nr = 0; 4164 4165 for (i = 0; i <= V_pf_hashmask; i++) { 4166 struct pf_idhash *ih = &V_pf_idhash[i]; 4167 4168 DIOCGETSTATESV2_retry: 4169 p = pstore; 4170 4171 if (LIST_EMPTY(&ih->states)) 4172 continue; 4173 4174 PF_HASHROW_LOCK(ih); 4175 count = 0; 4176 LIST_FOREACH(s, &ih->states, entry) { 4177 if (s->timeout == PFTM_UNLINKED) 4178 continue; 4179 count++; 4180 } 4181 4182 if (count > slice_count) { 4183 PF_HASHROW_UNLOCK(ih); 4184 free(pstore, M_PF); 4185 slice_count = count * 2; 4186 pstore = mallocarray(slice_count, 4187 sizeof(struct pf_state_export), M_PF, 4188 M_WAITOK | M_ZERO); 4189 goto DIOCGETSTATESV2_retry; 4190 } 4191 4192 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4193 PF_HASHROW_UNLOCK(ih); 4194 goto DIOCGETSTATESV2_full; 4195 } 4196 4197 LIST_FOREACH(s, &ih->states, entry) { 4198 if (s->timeout == PFTM_UNLINKED) 4199 continue; 4200 4201 pf_state_export(p, s); 4202 p++; 4203 nr++; 4204 } 4205 PF_HASHROW_UNLOCK(ih); 4206 error = copyout(pstore, out, 4207 sizeof(struct pf_state_export) * count); 4208 if (error) { 4209 free(pstore, M_PF); 4210 goto fail; 4211 } 4212 out = ps->ps_states + nr; 4213 } 4214 DIOCGETSTATESV2_full: 4215 ps->ps_len = nr * sizeof(struct pf_state_export); 4216 free(pstore, M_PF); 4217 4218 break; 4219 } 4220 #endif 4221 case DIOCGETSTATUSNV: { 4222 error = pf_getstatus((struct pfioc_nv *)addr); 4223 break; 4224 } 4225 4226 case DIOCSETSTATUSIF: { 4227 struct pfioc_if *pi = (struct pfioc_if *)addr; 4228 4229 if (pi->ifname[0] == 0) { 4230 bzero(V_pf_status.ifname, IFNAMSIZ); 4231 break; 4232 } 4233 PF_RULES_WLOCK(); 4234 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); 4235 PF_RULES_WUNLOCK(); 4236 break; 4237 } 4238 4239 case DIOCCLRSTATUS: { 4240 pf_ioctl_clear_status(); 4241 break; 4242 } 4243 4244 case DIOCNATLOOK: { 4245 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; 4246 4247 error = pf_ioctl_natlook(pnl); 4248 break; 4249 } 4250 4251 case DIOCSETTIMEOUT: { 4252 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4253 4254 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds, 4255 &pt->seconds); 4256 break; 4257 } 4258 4259 case DIOCGETTIMEOUT: { 4260 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4261 4262 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds); 4263 break; 4264 } 4265 4266 case DIOCGETLIMIT: { 4267 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4268 4269 error = pf_ioctl_get_limit(pl->index, &pl->limit); 4270 break; 4271 } 4272 4273 case DIOCSETLIMIT: { 4274 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4275 unsigned int old_limit; 4276 4277 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit); 4278 pl->limit = old_limit; 4279 break; 4280 } 4281 4282 case DIOCSETDEBUG: { 4283 u_int32_t *level = (u_int32_t *)addr; 4284 4285 PF_RULES_WLOCK(); 4286 V_pf_status.debug = *level; 4287 PF_RULES_WUNLOCK(); 4288 break; 4289 } 4290 4291 case DIOCCLRRULECTRS: { 4292 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ 4293 struct pf_kruleset *ruleset = &pf_main_ruleset; 4294 struct pf_krule *rule; 4295 4296 PF_RULES_WLOCK(); 4297 TAILQ_FOREACH(rule, 4298 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { 4299 pf_counter_u64_zero(&rule->evaluations); 4300 for (int i = 0; i < 2; i++) { 4301 pf_counter_u64_zero(&rule->packets[i]); 4302 pf_counter_u64_zero(&rule->bytes[i]); 4303 } 4304 } 4305 PF_RULES_WUNLOCK(); 4306 break; 4307 } 4308 4309 case DIOCGIFSPEEDV0: 4310 case DIOCGIFSPEEDV1: { 4311 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr; 4312 struct pf_ifspeed_v1 ps; 4313 struct ifnet *ifp; 4314 4315 if (psp->ifname[0] == '\0') { 4316 error = EINVAL; 4317 goto fail; 4318 } 4319 4320 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); 4321 if (error != 0) 4322 goto fail; 4323 ifp = ifunit(ps.ifname); 4324 if (ifp != NULL) { 4325 psp->baudrate32 = 4326 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX); 4327 if (cmd == DIOCGIFSPEEDV1) 4328 psp->baudrate = ifp->if_baudrate; 4329 } else { 4330 error = EINVAL; 4331 } 4332 break; 4333 } 4334 4335 #ifdef ALTQ 4336 case DIOCSTARTALTQ: { 4337 struct pf_altq *altq; 4338 4339 PF_RULES_WLOCK(); 4340 /* enable all altq interfaces on active list */ 4341 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4342 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4343 error = pf_enable_altq(altq); 4344 if (error != 0) 4345 break; 4346 } 4347 } 4348 if (error == 0) 4349 V_pf_altq_running = 1; 4350 PF_RULES_WUNLOCK(); 4351 DPFPRINTF(PF_DEBUG_MISC, "altq: started"); 4352 break; 4353 } 4354 4355 case DIOCSTOPALTQ: { 4356 struct pf_altq *altq; 4357 4358 PF_RULES_WLOCK(); 4359 /* disable all altq interfaces on active list */ 4360 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4361 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4362 error = pf_disable_altq(altq); 4363 if (error != 0) 4364 break; 4365 } 4366 } 4367 if (error == 0) 4368 V_pf_altq_running = 0; 4369 PF_RULES_WUNLOCK(); 4370 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped"); 4371 break; 4372 } 4373 4374 case DIOCADDALTQV0: 4375 case DIOCADDALTQV1: { 4376 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4377 struct pf_altq *altq, *a; 4378 struct ifnet *ifp; 4379 4380 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); 4381 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); 4382 if (error) 4383 goto fail; 4384 altq->local_flags = 0; 4385 4386 PF_RULES_WLOCK(); 4387 if (pa->ticket != V_ticket_altqs_inactive) { 4388 PF_RULES_WUNLOCK(); 4389 free(altq, M_PFALTQ); 4390 error = EBUSY; 4391 goto fail; 4392 } 4393 4394 /* 4395 * if this is for a queue, find the discipline and 4396 * copy the necessary fields 4397 */ 4398 if (altq->qname[0] != 0) { 4399 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) { 4400 PF_RULES_WUNLOCK(); 4401 error = EBUSY; 4402 free(altq, M_PFALTQ); 4403 goto fail; 4404 } 4405 altq->altq_disc = NULL; 4406 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { 4407 if (strncmp(a->ifname, altq->ifname, 4408 IFNAMSIZ) == 0) { 4409 altq->altq_disc = a->altq_disc; 4410 break; 4411 } 4412 } 4413 } 4414 4415 if ((ifp = ifunit(altq->ifname)) == NULL) 4416 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 4417 else 4418 error = altq_add(ifp, altq); 4419 4420 if (error) { 4421 PF_RULES_WUNLOCK(); 4422 free(altq, M_PFALTQ); 4423 goto fail; 4424 } 4425 4426 if (altq->qname[0] != 0) 4427 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); 4428 else 4429 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries); 4430 /* version error check done on import above */ 4431 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4432 PF_RULES_WUNLOCK(); 4433 break; 4434 } 4435 4436 case DIOCGETALTQSV0: 4437 case DIOCGETALTQSV1: { 4438 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4439 struct pf_altq *altq; 4440 4441 PF_RULES_RLOCK(); 4442 pa->nr = 0; 4443 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) 4444 pa->nr++; 4445 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) 4446 pa->nr++; 4447 pa->ticket = V_ticket_altqs_active; 4448 PF_RULES_RUNLOCK(); 4449 break; 4450 } 4451 4452 case DIOCGETALTQV0: 4453 case DIOCGETALTQV1: { 4454 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4455 struct pf_altq *altq; 4456 4457 PF_RULES_RLOCK(); 4458 if (pa->ticket != V_ticket_altqs_active) { 4459 PF_RULES_RUNLOCK(); 4460 error = EBUSY; 4461 goto fail; 4462 } 4463 altq = pf_altq_get_nth_active(pa->nr); 4464 if (altq == NULL) { 4465 PF_RULES_RUNLOCK(); 4466 error = EBUSY; 4467 goto fail; 4468 } 4469 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4470 PF_RULES_RUNLOCK(); 4471 break; 4472 } 4473 4474 case DIOCCHANGEALTQV0: 4475 case DIOCCHANGEALTQV1: 4476 /* CHANGEALTQ not supported yet! */ 4477 error = ENODEV; 4478 break; 4479 4480 case DIOCGETQSTATSV0: 4481 case DIOCGETQSTATSV1: { 4482 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr; 4483 struct pf_altq *altq; 4484 int nbytes; 4485 u_int32_t version; 4486 4487 PF_RULES_RLOCK(); 4488 if (pq->ticket != V_ticket_altqs_active) { 4489 PF_RULES_RUNLOCK(); 4490 error = EBUSY; 4491 goto fail; 4492 } 4493 nbytes = pq->nbytes; 4494 altq = pf_altq_get_nth_active(pq->nr); 4495 if (altq == NULL) { 4496 PF_RULES_RUNLOCK(); 4497 error = EBUSY; 4498 goto fail; 4499 } 4500 4501 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { 4502 PF_RULES_RUNLOCK(); 4503 error = ENXIO; 4504 goto fail; 4505 } 4506 PF_RULES_RUNLOCK(); 4507 if (cmd == DIOCGETQSTATSV0) 4508 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */ 4509 else 4510 version = pq->version; 4511 error = altq_getqstats(altq, pq->buf, &nbytes, version); 4512 if (error == 0) { 4513 pq->scheduler = altq->scheduler; 4514 pq->nbytes = nbytes; 4515 } 4516 break; 4517 } 4518 #endif /* ALTQ */ 4519 4520 case DIOCBEGINADDRS: { 4521 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4522 4523 error = pf_ioctl_begin_addrs(&pp->ticket); 4524 break; 4525 } 4526 4527 case DIOCADDADDR: { 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_add_addr(&npp); 4534 break; 4535 } 4536 4537 case DIOCGETADDRS: { 4538 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4539 struct pf_nl_pooladdr npp = {}; 4540 4541 npp.which = PF_RDR; 4542 memcpy(&npp, pp, sizeof(*pp)); 4543 error = pf_ioctl_get_addrs(&npp); 4544 memcpy(pp, &npp, sizeof(*pp)); 4545 4546 break; 4547 } 4548 4549 case DIOCGETADDR: { 4550 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4551 struct pf_nl_pooladdr npp = {}; 4552 4553 npp.which = PF_RDR; 4554 memcpy(&npp, pp, sizeof(*pp)); 4555 error = pf_ioctl_get_addr(&npp); 4556 memcpy(pp, &npp, sizeof(*pp)); 4557 4558 break; 4559 } 4560 4561 case DIOCCHANGEADDR: { 4562 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; 4563 struct pf_kpool *pool; 4564 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL; 4565 struct pf_kruleset *ruleset; 4566 struct pfi_kkif *kif = NULL; 4567 4568 pca->anchor[sizeof(pca->anchor) - 1] = '\0'; 4569 4570 if (pca->action < PF_CHANGE_ADD_HEAD || 4571 pca->action > PF_CHANGE_REMOVE) { 4572 error = EINVAL; 4573 goto fail; 4574 } 4575 if (pca->addr.addr.type != PF_ADDR_ADDRMASK && 4576 pca->addr.addr.type != PF_ADDR_DYNIFTL && 4577 pca->addr.addr.type != PF_ADDR_TABLE) { 4578 error = EINVAL; 4579 goto fail; 4580 } 4581 if (pca->addr.addr.p.dyn != NULL) { 4582 error = EINVAL; 4583 goto fail; 4584 } 4585 4586 if (pca->action != PF_CHANGE_REMOVE) { 4587 #ifndef INET 4588 if (pca->af == AF_INET) { 4589 error = EAFNOSUPPORT; 4590 goto fail; 4591 } 4592 #endif /* INET */ 4593 #ifndef INET6 4594 if (pca->af == AF_INET6) { 4595 error = EAFNOSUPPORT; 4596 goto fail; 4597 } 4598 #endif /* INET6 */ 4599 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); 4600 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); 4601 if (newpa->ifname[0]) 4602 kif = pf_kkif_create(M_WAITOK); 4603 newpa->kif = NULL; 4604 } 4605 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x) 4606 PF_RULES_WLOCK(); 4607 ruleset = pf_find_kruleset(pca->anchor); 4608 if (ruleset == NULL) 4609 ERROUT(EBUSY); 4610 4611 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action, 4612 pca->r_num, pca->r_last, 1, 1, PF_RDR); 4613 if (pool == NULL) 4614 ERROUT(EBUSY); 4615 4616 if (pca->action != PF_CHANGE_REMOVE) { 4617 if (newpa->ifname[0]) { 4618 newpa->kif = pfi_kkif_attach(kif, newpa->ifname); 4619 pfi_kkif_ref(newpa->kif); 4620 kif = NULL; 4621 } 4622 4623 switch (newpa->addr.type) { 4624 case PF_ADDR_DYNIFTL: 4625 error = pfi_dynaddr_setup(&newpa->addr, 4626 pca->af); 4627 break; 4628 case PF_ADDR_TABLE: 4629 newpa->addr.p.tbl = pfr_attach_table(ruleset, 4630 newpa->addr.v.tblname); 4631 if (newpa->addr.p.tbl == NULL) 4632 error = ENOMEM; 4633 break; 4634 } 4635 if (error) 4636 goto DIOCCHANGEADDR_error; 4637 } 4638 4639 switch (pca->action) { 4640 case PF_CHANGE_ADD_HEAD: 4641 oldpa = TAILQ_FIRST(&pool->list); 4642 break; 4643 case PF_CHANGE_ADD_TAIL: 4644 oldpa = TAILQ_LAST(&pool->list, pf_kpalist); 4645 break; 4646 default: 4647 oldpa = TAILQ_FIRST(&pool->list); 4648 for (int i = 0; oldpa && i < pca->nr; i++) 4649 oldpa = TAILQ_NEXT(oldpa, entries); 4650 4651 if (oldpa == NULL) 4652 ERROUT(EINVAL); 4653 } 4654 4655 if (pca->action == PF_CHANGE_REMOVE) { 4656 TAILQ_REMOVE(&pool->list, oldpa, entries); 4657 switch (oldpa->addr.type) { 4658 case PF_ADDR_DYNIFTL: 4659 pfi_dynaddr_remove(oldpa->addr.p.dyn); 4660 break; 4661 case PF_ADDR_TABLE: 4662 pfr_detach_table(oldpa->addr.p.tbl); 4663 break; 4664 } 4665 if (oldpa->kif) 4666 pfi_kkif_unref(oldpa->kif); 4667 free(oldpa, M_PFRULE); 4668 } else { 4669 if (oldpa == NULL) 4670 TAILQ_INSERT_TAIL(&pool->list, newpa, entries); 4671 else if (pca->action == PF_CHANGE_ADD_HEAD || 4672 pca->action == PF_CHANGE_ADD_BEFORE) 4673 TAILQ_INSERT_BEFORE(oldpa, newpa, entries); 4674 else 4675 TAILQ_INSERT_AFTER(&pool->list, oldpa, 4676 newpa, entries); 4677 } 4678 4679 pool->cur = TAILQ_FIRST(&pool->list); 4680 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); 4681 PF_RULES_WUNLOCK(); 4682 break; 4683 4684 #undef ERROUT 4685 DIOCCHANGEADDR_error: 4686 if (newpa != NULL) { 4687 if (newpa->kif) 4688 pfi_kkif_unref(newpa->kif); 4689 free(newpa, M_PFRULE); 4690 } 4691 PF_RULES_WUNLOCK(); 4692 pf_kkif_free(kif); 4693 break; 4694 } 4695 4696 case DIOCGETRULESETS: { 4697 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4698 4699 pr->path[sizeof(pr->path) - 1] = '\0'; 4700 4701 error = pf_ioctl_get_rulesets(pr); 4702 break; 4703 } 4704 4705 case DIOCGETRULESET: { 4706 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4707 4708 pr->path[sizeof(pr->path) - 1] = '\0'; 4709 4710 error = pf_ioctl_get_ruleset(pr); 4711 break; 4712 } 4713 4714 case DIOCRCLRTABLES: { 4715 struct pfioc_table *io = (struct pfioc_table *)addr; 4716 4717 if (io->pfrio_esize != 0) { 4718 error = ENODEV; 4719 goto fail; 4720 } 4721 PF_RULES_WLOCK(); 4722 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, 4723 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4724 PF_RULES_WUNLOCK(); 4725 break; 4726 } 4727 4728 case DIOCRADDTABLES: { 4729 struct pfioc_table *io = (struct pfioc_table *)addr; 4730 struct pfr_table *pfrts; 4731 size_t totlen; 4732 4733 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4734 error = ENODEV; 4735 goto fail; 4736 } 4737 4738 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4739 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4740 error = ENOMEM; 4741 goto fail; 4742 } 4743 4744 totlen = io->pfrio_size * sizeof(struct pfr_table); 4745 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4746 M_PF, M_WAITOK); 4747 error = copyin(io->pfrio_buffer, pfrts, totlen); 4748 if (error) { 4749 free(pfrts, M_PF); 4750 goto fail; 4751 } 4752 PF_RULES_WLOCK(); 4753 error = pfr_add_tables(pfrts, io->pfrio_size, 4754 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4755 PF_RULES_WUNLOCK(); 4756 free(pfrts, M_PF); 4757 break; 4758 } 4759 4760 case DIOCRDELTABLES: { 4761 struct pfioc_table *io = (struct pfioc_table *)addr; 4762 struct pfr_table *pfrts; 4763 size_t totlen; 4764 4765 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4766 error = ENODEV; 4767 goto fail; 4768 } 4769 4770 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4771 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4772 error = ENOMEM; 4773 goto fail; 4774 } 4775 4776 totlen = io->pfrio_size * sizeof(struct pfr_table); 4777 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4778 M_PF, M_WAITOK); 4779 error = copyin(io->pfrio_buffer, pfrts, totlen); 4780 if (error) { 4781 free(pfrts, M_PF); 4782 goto fail; 4783 } 4784 PF_RULES_WLOCK(); 4785 error = pfr_del_tables(pfrts, io->pfrio_size, 4786 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4787 PF_RULES_WUNLOCK(); 4788 free(pfrts, M_PF); 4789 break; 4790 } 4791 4792 case DIOCRGETTABLES: { 4793 struct pfioc_table *io = (struct pfioc_table *)addr; 4794 struct pfr_table *pfrts; 4795 size_t totlen; 4796 int n; 4797 4798 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4799 error = ENODEV; 4800 goto fail; 4801 } 4802 PF_RULES_RLOCK(); 4803 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4804 if (n < 0) { 4805 PF_RULES_RUNLOCK(); 4806 error = EINVAL; 4807 goto fail; 4808 } 4809 io->pfrio_size = min(io->pfrio_size, n); 4810 4811 totlen = io->pfrio_size * sizeof(struct pfr_table); 4812 4813 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4814 M_PF, M_NOWAIT | M_ZERO); 4815 if (pfrts == NULL) { 4816 error = ENOMEM; 4817 PF_RULES_RUNLOCK(); 4818 goto fail; 4819 } 4820 error = pfr_get_tables(&io->pfrio_table, pfrts, 4821 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4822 PF_RULES_RUNLOCK(); 4823 if (error == 0) 4824 error = copyout(pfrts, io->pfrio_buffer, totlen); 4825 free(pfrts, M_PF); 4826 break; 4827 } 4828 4829 case DIOCRGETTSTATS: { 4830 struct pfioc_table *io = (struct pfioc_table *)addr; 4831 struct pfr_tstats *pfrtstats; 4832 size_t totlen; 4833 int n; 4834 4835 if (io->pfrio_esize != sizeof(struct pfr_tstats)) { 4836 error = ENODEV; 4837 goto fail; 4838 } 4839 PF_TABLE_STATS_LOCK(); 4840 PF_RULES_RLOCK(); 4841 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4842 if (n < 0) { 4843 PF_RULES_RUNLOCK(); 4844 PF_TABLE_STATS_UNLOCK(); 4845 error = EINVAL; 4846 goto fail; 4847 } 4848 io->pfrio_size = min(io->pfrio_size, n); 4849 4850 totlen = io->pfrio_size * sizeof(struct pfr_tstats); 4851 pfrtstats = mallocarray(io->pfrio_size, 4852 sizeof(struct pfr_tstats), M_PF, M_NOWAIT | M_ZERO); 4853 if (pfrtstats == NULL) { 4854 error = ENOMEM; 4855 PF_RULES_RUNLOCK(); 4856 PF_TABLE_STATS_UNLOCK(); 4857 goto fail; 4858 } 4859 error = pfr_get_tstats(&io->pfrio_table, pfrtstats, 4860 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4861 PF_RULES_RUNLOCK(); 4862 PF_TABLE_STATS_UNLOCK(); 4863 if (error == 0) 4864 error = copyout(pfrtstats, io->pfrio_buffer, totlen); 4865 free(pfrtstats, M_PF); 4866 break; 4867 } 4868 4869 case DIOCRCLRTSTATS: { 4870 struct pfioc_table *io = (struct pfioc_table *)addr; 4871 struct pfr_table *pfrts; 4872 size_t totlen; 4873 4874 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4875 error = ENODEV; 4876 goto fail; 4877 } 4878 4879 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4880 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4881 /* We used to count tables and use the minimum required 4882 * size, so we didn't fail on overly large requests. 4883 * Keep doing so. */ 4884 io->pfrio_size = pf_ioctl_maxcount; 4885 goto fail; 4886 } 4887 4888 totlen = io->pfrio_size * sizeof(struct pfr_table); 4889 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4890 M_PF, M_WAITOK); 4891 error = copyin(io->pfrio_buffer, pfrts, totlen); 4892 if (error) { 4893 free(pfrts, M_PF); 4894 goto fail; 4895 } 4896 4897 PF_TABLE_STATS_LOCK(); 4898 PF_RULES_RLOCK(); 4899 error = pfr_clr_tstats(pfrts, io->pfrio_size, 4900 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4901 PF_RULES_RUNLOCK(); 4902 PF_TABLE_STATS_UNLOCK(); 4903 free(pfrts, M_PF); 4904 break; 4905 } 4906 4907 case DIOCRSETTFLAGS: { 4908 struct pfioc_table *io = (struct pfioc_table *)addr; 4909 struct pfr_table *pfrts; 4910 size_t totlen; 4911 int n; 4912 4913 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4914 error = ENODEV; 4915 goto fail; 4916 } 4917 4918 PF_RULES_RLOCK(); 4919 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4920 if (n < 0) { 4921 PF_RULES_RUNLOCK(); 4922 error = EINVAL; 4923 goto fail; 4924 } 4925 4926 io->pfrio_size = min(io->pfrio_size, n); 4927 PF_RULES_RUNLOCK(); 4928 4929 totlen = io->pfrio_size * sizeof(struct pfr_table); 4930 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4931 M_PF, M_WAITOK); 4932 error = copyin(io->pfrio_buffer, pfrts, totlen); 4933 if (error) { 4934 free(pfrts, M_PF); 4935 goto fail; 4936 } 4937 PF_RULES_WLOCK(); 4938 error = pfr_set_tflags(pfrts, io->pfrio_size, 4939 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, 4940 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4941 PF_RULES_WUNLOCK(); 4942 free(pfrts, M_PF); 4943 break; 4944 } 4945 4946 case DIOCRCLRADDRS: { 4947 struct pfioc_table *io = (struct pfioc_table *)addr; 4948 4949 if (io->pfrio_esize != 0) { 4950 error = ENODEV; 4951 goto fail; 4952 } 4953 PF_RULES_WLOCK(); 4954 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, 4955 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4956 PF_RULES_WUNLOCK(); 4957 break; 4958 } 4959 4960 case DIOCRADDADDRS: { 4961 struct pfioc_table *io = (struct pfioc_table *)addr; 4962 struct pfr_addr *pfras; 4963 size_t totlen; 4964 4965 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 4966 error = ENODEV; 4967 goto fail; 4968 } 4969 if (io->pfrio_size < 0 || 4970 io->pfrio_size > pf_ioctl_maxcount || 4971 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 4972 error = EINVAL; 4973 goto fail; 4974 } 4975 totlen = io->pfrio_size * sizeof(struct pfr_addr); 4976 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 4977 M_PF, M_WAITOK); 4978 error = copyin(io->pfrio_buffer, pfras, totlen); 4979 if (error) { 4980 free(pfras, M_PF); 4981 goto fail; 4982 } 4983 PF_RULES_WLOCK(); 4984 io->pfrio_nadd = 0; 4985 error = pfr_add_addrs(&io->pfrio_table, pfras, 4986 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | 4987 PFR_FLAG_USERIOCTL); 4988 PF_RULES_WUNLOCK(); 4989 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 4990 error = copyout(pfras, io->pfrio_buffer, totlen); 4991 free(pfras, M_PF); 4992 break; 4993 } 4994 4995 case DIOCRDELADDRS: { 4996 struct pfioc_table *io = (struct pfioc_table *)addr; 4997 struct pfr_addr *pfras; 4998 size_t totlen; 4999 5000 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5001 error = ENODEV; 5002 goto fail; 5003 } 5004 if (io->pfrio_size < 0 || 5005 io->pfrio_size > pf_ioctl_maxcount || 5006 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5007 error = EINVAL; 5008 goto fail; 5009 } 5010 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5011 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5012 M_PF, M_WAITOK); 5013 error = copyin(io->pfrio_buffer, pfras, totlen); 5014 if (error) { 5015 free(pfras, M_PF); 5016 goto fail; 5017 } 5018 PF_RULES_WLOCK(); 5019 error = pfr_del_addrs(&io->pfrio_table, pfras, 5020 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | 5021 PFR_FLAG_USERIOCTL); 5022 PF_RULES_WUNLOCK(); 5023 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5024 error = copyout(pfras, io->pfrio_buffer, totlen); 5025 free(pfras, M_PF); 5026 break; 5027 } 5028 5029 case DIOCRSETADDRS: { 5030 struct pfioc_table *io = (struct pfioc_table *)addr; 5031 struct pfr_addr *pfras; 5032 size_t totlen, count; 5033 5034 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5035 error = ENODEV; 5036 goto fail; 5037 } 5038 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { 5039 error = EINVAL; 5040 goto fail; 5041 } 5042 count = max(io->pfrio_size, io->pfrio_size2); 5043 if (count > pf_ioctl_maxcount || 5044 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { 5045 error = EINVAL; 5046 goto fail; 5047 } 5048 totlen = count * sizeof(struct pfr_addr); 5049 pfras = mallocarray(count, sizeof(struct pfr_addr), M_PF, 5050 M_WAITOK); 5051 error = copyin(io->pfrio_buffer, pfras, totlen); 5052 if (error) { 5053 free(pfras, M_PF); 5054 goto fail; 5055 } 5056 PF_RULES_WLOCK(); 5057 error = pfr_set_addrs(&io->pfrio_table, pfras, 5058 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, 5059 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | 5060 PFR_FLAG_USERIOCTL, 0); 5061 PF_RULES_WUNLOCK(); 5062 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5063 error = copyout(pfras, io->pfrio_buffer, totlen); 5064 free(pfras, M_PF); 5065 break; 5066 } 5067 5068 case DIOCRGETADDRS: { 5069 struct pfioc_table *io = (struct pfioc_table *)addr; 5070 struct pfr_addr *pfras; 5071 size_t totlen; 5072 5073 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5074 error = ENODEV; 5075 goto fail; 5076 } 5077 if (io->pfrio_size < 0 || 5078 io->pfrio_size > pf_ioctl_maxcount || 5079 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5080 error = EINVAL; 5081 goto fail; 5082 } 5083 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5084 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5085 M_PF, M_WAITOK | M_ZERO); 5086 PF_RULES_RLOCK(); 5087 error = pfr_get_addrs(&io->pfrio_table, pfras, 5088 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5089 PF_RULES_RUNLOCK(); 5090 if (error == 0) 5091 error = copyout(pfras, io->pfrio_buffer, totlen); 5092 free(pfras, M_PF); 5093 break; 5094 } 5095 5096 case DIOCRGETASTATS: { 5097 struct pfioc_table *io = (struct pfioc_table *)addr; 5098 struct pfr_astats *pfrastats; 5099 size_t totlen; 5100 5101 if (io->pfrio_esize != sizeof(struct pfr_astats)) { 5102 error = ENODEV; 5103 goto fail; 5104 } 5105 if (io->pfrio_size < 0 || 5106 io->pfrio_size > pf_ioctl_maxcount || 5107 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { 5108 error = EINVAL; 5109 goto fail; 5110 } 5111 totlen = io->pfrio_size * sizeof(struct pfr_astats); 5112 pfrastats = mallocarray(io->pfrio_size, 5113 sizeof(struct pfr_astats), M_PF, M_WAITOK | M_ZERO); 5114 PF_RULES_RLOCK(); 5115 error = pfr_get_astats(&io->pfrio_table, pfrastats, 5116 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5117 PF_RULES_RUNLOCK(); 5118 if (error == 0) 5119 error = copyout(pfrastats, io->pfrio_buffer, totlen); 5120 free(pfrastats, M_PF); 5121 break; 5122 } 5123 5124 case DIOCRCLRASTATS: { 5125 struct pfioc_table *io = (struct pfioc_table *)addr; 5126 struct pfr_addr *pfras; 5127 size_t totlen; 5128 5129 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5130 error = ENODEV; 5131 goto fail; 5132 } 5133 if (io->pfrio_size < 0 || 5134 io->pfrio_size > pf_ioctl_maxcount || 5135 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5136 error = EINVAL; 5137 goto fail; 5138 } 5139 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5140 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5141 M_PF, M_WAITOK); 5142 error = copyin(io->pfrio_buffer, pfras, totlen); 5143 if (error) { 5144 free(pfras, M_PF); 5145 goto fail; 5146 } 5147 PF_RULES_WLOCK(); 5148 error = pfr_clr_astats(&io->pfrio_table, pfras, 5149 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | 5150 PFR_FLAG_USERIOCTL); 5151 PF_RULES_WUNLOCK(); 5152 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5153 error = copyout(pfras, io->pfrio_buffer, totlen); 5154 free(pfras, M_PF); 5155 break; 5156 } 5157 5158 case DIOCRTSTADDRS: { 5159 struct pfioc_table *io = (struct pfioc_table *)addr; 5160 struct pfr_addr *pfras; 5161 size_t totlen; 5162 5163 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5164 error = ENODEV; 5165 goto fail; 5166 } 5167 if (io->pfrio_size < 0 || 5168 io->pfrio_size > pf_ioctl_maxcount || 5169 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5170 error = EINVAL; 5171 goto fail; 5172 } 5173 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5174 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5175 M_PF, M_WAITOK); 5176 error = copyin(io->pfrio_buffer, pfras, totlen); 5177 if (error) { 5178 free(pfras, M_PF); 5179 goto fail; 5180 } 5181 PF_RULES_RLOCK(); 5182 error = pfr_tst_addrs(&io->pfrio_table, pfras, 5183 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | 5184 PFR_FLAG_USERIOCTL); 5185 PF_RULES_RUNLOCK(); 5186 if (error == 0) 5187 error = copyout(pfras, io->pfrio_buffer, totlen); 5188 free(pfras, M_PF); 5189 break; 5190 } 5191 5192 case DIOCRINADEFINE: { 5193 struct pfioc_table *io = (struct pfioc_table *)addr; 5194 struct pfr_addr *pfras; 5195 size_t totlen; 5196 5197 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5198 error = ENODEV; 5199 goto fail; 5200 } 5201 if (io->pfrio_size < 0 || 5202 io->pfrio_size > pf_ioctl_maxcount || 5203 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5204 error = EINVAL; 5205 goto fail; 5206 } 5207 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5208 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5209 M_PF, M_WAITOK); 5210 error = copyin(io->pfrio_buffer, pfras, totlen); 5211 if (error) { 5212 free(pfras, M_PF); 5213 goto fail; 5214 } 5215 PF_RULES_WLOCK(); 5216 error = pfr_ina_define(&io->pfrio_table, pfras, 5217 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, 5218 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5219 PF_RULES_WUNLOCK(); 5220 free(pfras, M_PF); 5221 break; 5222 } 5223 5224 case DIOCOSFPADD: { 5225 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5226 PF_RULES_WLOCK(); 5227 error = pf_osfp_add(io); 5228 PF_RULES_WUNLOCK(); 5229 break; 5230 } 5231 5232 case DIOCOSFPGET: { 5233 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5234 PF_RULES_RLOCK(); 5235 error = pf_osfp_get(io); 5236 PF_RULES_RUNLOCK(); 5237 break; 5238 } 5239 5240 case DIOCXBEGIN: { 5241 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5242 struct pfioc_trans_e *ioes, *ioe; 5243 size_t totlen; 5244 int i; 5245 5246 if (io->esize != sizeof(*ioe)) { 5247 error = ENODEV; 5248 goto fail; 5249 } 5250 if (io->size < 0 || 5251 io->size > pf_ioctl_maxcount || 5252 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5253 error = EINVAL; 5254 goto fail; 5255 } 5256 totlen = sizeof(struct pfioc_trans_e) * io->size; 5257 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5258 M_PF, M_WAITOK); 5259 error = copyin(io->array, ioes, totlen); 5260 if (error) { 5261 free(ioes, M_PF); 5262 goto fail; 5263 } 5264 PF_RULES_WLOCK(); 5265 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5266 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5267 switch (ioe->rs_num) { 5268 case PF_RULESET_ETH: 5269 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) { 5270 PF_RULES_WUNLOCK(); 5271 free(ioes, M_PF); 5272 goto fail; 5273 } 5274 break; 5275 #ifdef ALTQ 5276 case PF_RULESET_ALTQ: 5277 if (ioe->anchor[0]) { 5278 PF_RULES_WUNLOCK(); 5279 free(ioes, M_PF); 5280 error = EINVAL; 5281 goto fail; 5282 } 5283 if ((error = pf_begin_altq(&ioe->ticket))) { 5284 PF_RULES_WUNLOCK(); 5285 free(ioes, M_PF); 5286 goto fail; 5287 } 5288 break; 5289 #endif /* ALTQ */ 5290 case PF_RULESET_TABLE: 5291 { 5292 struct pfr_table table; 5293 5294 bzero(&table, sizeof(table)); 5295 strlcpy(table.pfrt_anchor, ioe->anchor, 5296 sizeof(table.pfrt_anchor)); 5297 if ((error = pfr_ina_begin(&table, 5298 &ioe->ticket, NULL, 0))) { 5299 PF_RULES_WUNLOCK(); 5300 free(ioes, M_PF); 5301 goto fail; 5302 } 5303 break; 5304 } 5305 default: 5306 if ((error = pf_begin_rules(&ioe->ticket, 5307 ioe->rs_num, ioe->anchor))) { 5308 PF_RULES_WUNLOCK(); 5309 free(ioes, M_PF); 5310 goto fail; 5311 } 5312 break; 5313 } 5314 } 5315 PF_RULES_WUNLOCK(); 5316 error = copyout(ioes, io->array, totlen); 5317 free(ioes, M_PF); 5318 break; 5319 } 5320 5321 case DIOCXROLLBACK: { 5322 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5323 struct pfioc_trans_e *ioe, *ioes; 5324 size_t totlen; 5325 int i; 5326 5327 if (io->esize != sizeof(*ioe)) { 5328 error = ENODEV; 5329 goto fail; 5330 } 5331 if (io->size < 0 || 5332 io->size > pf_ioctl_maxcount || 5333 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5334 error = EINVAL; 5335 goto fail; 5336 } 5337 totlen = sizeof(struct pfioc_trans_e) * io->size; 5338 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5339 M_PF, M_WAITOK); 5340 error = copyin(io->array, ioes, totlen); 5341 if (error) { 5342 free(ioes, M_PF); 5343 goto fail; 5344 } 5345 PF_RULES_WLOCK(); 5346 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5347 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5348 switch (ioe->rs_num) { 5349 case PF_RULESET_ETH: 5350 if ((error = pf_rollback_eth(ioe->ticket, 5351 ioe->anchor))) { 5352 PF_RULES_WUNLOCK(); 5353 free(ioes, M_PF); 5354 goto fail; /* really bad */ 5355 } 5356 break; 5357 #ifdef ALTQ 5358 case PF_RULESET_ALTQ: 5359 if (ioe->anchor[0]) { 5360 PF_RULES_WUNLOCK(); 5361 free(ioes, M_PF); 5362 error = EINVAL; 5363 goto fail; 5364 } 5365 if ((error = pf_rollback_altq(ioe->ticket))) { 5366 PF_RULES_WUNLOCK(); 5367 free(ioes, M_PF); 5368 goto fail; /* really bad */ 5369 } 5370 break; 5371 #endif /* ALTQ */ 5372 case PF_RULESET_TABLE: 5373 { 5374 struct pfr_table table; 5375 5376 bzero(&table, sizeof(table)); 5377 strlcpy(table.pfrt_anchor, ioe->anchor, 5378 sizeof(table.pfrt_anchor)); 5379 if ((error = pfr_ina_rollback(&table, 5380 ioe->ticket, NULL, 0))) { 5381 PF_RULES_WUNLOCK(); 5382 free(ioes, M_PF); 5383 goto fail; /* really bad */ 5384 } 5385 break; 5386 } 5387 default: 5388 if ((error = pf_rollback_rules(ioe->ticket, 5389 ioe->rs_num, ioe->anchor))) { 5390 PF_RULES_WUNLOCK(); 5391 free(ioes, M_PF); 5392 goto fail; /* really bad */ 5393 } 5394 break; 5395 } 5396 } 5397 PF_RULES_WUNLOCK(); 5398 free(ioes, M_PF); 5399 break; 5400 } 5401 5402 case DIOCXCOMMIT: { 5403 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5404 struct pfioc_trans_e *ioe, *ioes; 5405 struct pf_kruleset *rs; 5406 struct pf_keth_ruleset *ers; 5407 size_t totlen; 5408 int i; 5409 5410 if (io->esize != sizeof(*ioe)) { 5411 error = ENODEV; 5412 goto fail; 5413 } 5414 5415 if (io->size < 0 || 5416 io->size > pf_ioctl_maxcount || 5417 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5418 error = EINVAL; 5419 goto fail; 5420 } 5421 5422 totlen = sizeof(struct pfioc_trans_e) * io->size; 5423 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5424 M_PF, M_WAITOK); 5425 error = copyin(io->array, ioes, totlen); 5426 if (error) { 5427 free(ioes, M_PF); 5428 goto fail; 5429 } 5430 PF_RULES_WLOCK(); 5431 /* First makes sure everything will succeed. */ 5432 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5433 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5434 switch (ioe->rs_num) { 5435 case PF_RULESET_ETH: 5436 ers = pf_find_keth_ruleset(ioe->anchor); 5437 if (ers == NULL || ioe->ticket == 0 || 5438 ioe->ticket != ers->inactive.ticket) { 5439 PF_RULES_WUNLOCK(); 5440 free(ioes, M_PF); 5441 error = EINVAL; 5442 goto fail; 5443 } 5444 break; 5445 #ifdef ALTQ 5446 case PF_RULESET_ALTQ: 5447 if (ioe->anchor[0]) { 5448 PF_RULES_WUNLOCK(); 5449 free(ioes, M_PF); 5450 error = EINVAL; 5451 goto fail; 5452 } 5453 if (!V_altqs_inactive_open || ioe->ticket != 5454 V_ticket_altqs_inactive) { 5455 PF_RULES_WUNLOCK(); 5456 free(ioes, M_PF); 5457 error = EBUSY; 5458 goto fail; 5459 } 5460 break; 5461 #endif /* ALTQ */ 5462 case PF_RULESET_TABLE: 5463 rs = pf_find_kruleset(ioe->anchor); 5464 if (rs == NULL || !rs->topen || ioe->ticket != 5465 rs->tticket) { 5466 PF_RULES_WUNLOCK(); 5467 free(ioes, M_PF); 5468 error = EBUSY; 5469 goto fail; 5470 } 5471 break; 5472 default: 5473 if (ioe->rs_num < 0 || ioe->rs_num >= 5474 PF_RULESET_MAX) { 5475 PF_RULES_WUNLOCK(); 5476 free(ioes, M_PF); 5477 error = EINVAL; 5478 goto fail; 5479 } 5480 rs = pf_find_kruleset(ioe->anchor); 5481 if (rs == NULL || 5482 !rs->rules[ioe->rs_num].inactive.open || 5483 rs->rules[ioe->rs_num].inactive.ticket != 5484 ioe->ticket) { 5485 PF_RULES_WUNLOCK(); 5486 free(ioes, M_PF); 5487 error = EBUSY; 5488 goto fail; 5489 } 5490 break; 5491 } 5492 } 5493 /* Now do the commit - no errors should happen here. */ 5494 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5495 switch (ioe->rs_num) { 5496 case PF_RULESET_ETH: 5497 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) { 5498 PF_RULES_WUNLOCK(); 5499 free(ioes, M_PF); 5500 goto fail; /* really bad */ 5501 } 5502 break; 5503 #ifdef ALTQ 5504 case PF_RULESET_ALTQ: 5505 if ((error = pf_commit_altq(ioe->ticket))) { 5506 PF_RULES_WUNLOCK(); 5507 free(ioes, M_PF); 5508 goto fail; /* really bad */ 5509 } 5510 break; 5511 #endif /* ALTQ */ 5512 case PF_RULESET_TABLE: 5513 { 5514 struct pfr_table table; 5515 5516 bzero(&table, sizeof(table)); 5517 (void)strlcpy(table.pfrt_anchor, ioe->anchor, 5518 sizeof(table.pfrt_anchor)); 5519 if ((error = pfr_ina_commit(&table, 5520 ioe->ticket, NULL, NULL, 0))) { 5521 PF_RULES_WUNLOCK(); 5522 free(ioes, M_PF); 5523 goto fail; /* really bad */ 5524 } 5525 break; 5526 } 5527 default: 5528 if ((error = pf_commit_rules(ioe->ticket, 5529 ioe->rs_num, ioe->anchor))) { 5530 PF_RULES_WUNLOCK(); 5531 free(ioes, M_PF); 5532 goto fail; /* really bad */ 5533 } 5534 break; 5535 } 5536 } 5537 PF_RULES_WUNLOCK(); 5538 5539 /* Only hook into EtherNet taffic if we've got rules for it. */ 5540 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 5541 hook_pf_eth(); 5542 else 5543 dehook_pf_eth(); 5544 5545 free(ioes, M_PF); 5546 break; 5547 } 5548 5549 case DIOCGETSRCNODES: { 5550 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; 5551 struct pf_srchash *sh; 5552 struct pf_ksrc_node *n; 5553 struct pf_src_node *p, *pstore; 5554 uint32_t i, nr = 0; 5555 5556 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5557 i++, sh++) { 5558 PF_HASHROW_LOCK(sh); 5559 LIST_FOREACH(n, &sh->nodes, entry) 5560 nr++; 5561 PF_HASHROW_UNLOCK(sh); 5562 } 5563 5564 psn->psn_len = min(psn->psn_len, 5565 sizeof(struct pf_src_node) * nr); 5566 5567 if (psn->psn_len == 0) { 5568 psn->psn_len = sizeof(struct pf_src_node) * nr; 5569 goto fail; 5570 } 5571 5572 nr = 0; 5573 5574 p = pstore = malloc(psn->psn_len, M_PF, M_WAITOK | M_ZERO); 5575 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5576 i++, sh++) { 5577 PF_HASHROW_LOCK(sh); 5578 LIST_FOREACH(n, &sh->nodes, entry) { 5579 5580 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) 5581 break; 5582 5583 pf_src_node_copy(n, p); 5584 5585 p++; 5586 nr++; 5587 } 5588 PF_HASHROW_UNLOCK(sh); 5589 } 5590 error = copyout(pstore, psn->psn_src_nodes, 5591 sizeof(struct pf_src_node) * nr); 5592 if (error) { 5593 free(pstore, M_PF); 5594 goto fail; 5595 } 5596 psn->psn_len = sizeof(struct pf_src_node) * nr; 5597 free(pstore, M_PF); 5598 break; 5599 } 5600 5601 case DIOCCLRSRCNODES: { 5602 pf_kill_srcnodes(NULL); 5603 break; 5604 } 5605 5606 case DIOCKILLSRCNODES: 5607 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); 5608 break; 5609 5610 #ifdef COMPAT_FREEBSD13 5611 case DIOCKEEPCOUNTERS_FREEBSD13: 5612 #endif 5613 case DIOCKEEPCOUNTERS: 5614 error = pf_keepcounters((struct pfioc_nv *)addr); 5615 break; 5616 5617 case DIOCGETSYNCOOKIES: 5618 error = pf_get_syncookies((struct pfioc_nv *)addr); 5619 break; 5620 5621 case DIOCSETSYNCOOKIES: 5622 error = pf_set_syncookies((struct pfioc_nv *)addr); 5623 break; 5624 5625 case DIOCSETHOSTID: { 5626 u_int32_t *hostid = (u_int32_t *)addr; 5627 5628 PF_RULES_WLOCK(); 5629 if (*hostid == 0) 5630 V_pf_status.hostid = arc4random(); 5631 else 5632 V_pf_status.hostid = *hostid; 5633 PF_RULES_WUNLOCK(); 5634 break; 5635 } 5636 5637 case DIOCOSFPFLUSH: 5638 PF_RULES_WLOCK(); 5639 pf_osfp_flush(); 5640 PF_RULES_WUNLOCK(); 5641 break; 5642 5643 case DIOCIGETIFACES: { 5644 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5645 struct pfi_kif *ifstore; 5646 size_t bufsiz; 5647 5648 if (io->pfiio_esize != sizeof(struct pfi_kif)) { 5649 error = ENODEV; 5650 goto fail; 5651 } 5652 5653 if (io->pfiio_size < 0 || 5654 io->pfiio_size > pf_ioctl_maxcount || 5655 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { 5656 error = EINVAL; 5657 goto fail; 5658 } 5659 5660 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5661 5662 bufsiz = io->pfiio_size * sizeof(struct pfi_kif); 5663 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif), 5664 M_PF, M_WAITOK | M_ZERO); 5665 5666 PF_RULES_RLOCK(); 5667 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); 5668 PF_RULES_RUNLOCK(); 5669 error = copyout(ifstore, io->pfiio_buffer, bufsiz); 5670 free(ifstore, M_PF); 5671 break; 5672 } 5673 5674 case DIOCSETIFFLAG: { 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_set_flags(io->pfiio_name, io->pfiio_flags); 5681 PF_RULES_WUNLOCK(); 5682 break; 5683 } 5684 5685 case DIOCCLRIFFLAG: { 5686 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5687 5688 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5689 5690 PF_RULES_WLOCK(); 5691 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); 5692 PF_RULES_WUNLOCK(); 5693 break; 5694 } 5695 5696 case DIOCSETREASS: { 5697 u_int32_t *reass = (u_int32_t *)addr; 5698 5699 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF); 5700 /* Removal of DF flag without reassembly enabled is not a 5701 * valid combination. Disable reassembly in such case. */ 5702 if (!(V_pf_status.reass & PF_REASS_ENABLED)) 5703 V_pf_status.reass = 0; 5704 break; 5705 } 5706 5707 default: 5708 error = ENODEV; 5709 break; 5710 } 5711 fail: 5712 CURVNET_RESTORE(); 5713 5714 #undef ERROUT_IOCTL 5715 5716 return (error); 5717 } 5718 5719 void 5720 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version) 5721 { 5722 bzero(sp, sizeof(union pfsync_state_union)); 5723 5724 /* copy from state key */ 5725 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5726 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5727 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5728 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5729 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5730 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5731 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5732 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5733 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto; 5734 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af; 5735 5736 /* copy from state */ 5737 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname)); 5738 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr)); 5739 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000)); 5740 sp->pfs_1301.expire = pf_state_expires(st); 5741 if (sp->pfs_1301.expire <= time_uptime) 5742 sp->pfs_1301.expire = htonl(0); 5743 else 5744 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime); 5745 5746 sp->pfs_1301.direction = st->direction; 5747 sp->pfs_1301.log = st->act.log; 5748 sp->pfs_1301.timeout = st->timeout; 5749 5750 switch (msg_version) { 5751 case PFSYNC_MSG_VERSION_1301: 5752 sp->pfs_1301.state_flags = st->state_flags; 5753 break; 5754 case PFSYNC_MSG_VERSION_1400: 5755 sp->pfs_1400.state_flags = htons(st->state_flags); 5756 sp->pfs_1400.qid = htons(st->act.qid); 5757 sp->pfs_1400.pqid = htons(st->act.pqid); 5758 sp->pfs_1400.dnpipe = htons(st->act.dnpipe); 5759 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe); 5760 sp->pfs_1400.rtableid = htonl(st->act.rtableid); 5761 sp->pfs_1400.min_ttl = st->act.min_ttl; 5762 sp->pfs_1400.set_tos = st->act.set_tos; 5763 sp->pfs_1400.max_mss = htons(st->act.max_mss); 5764 sp->pfs_1400.set_prio[0] = st->act.set_prio[0]; 5765 sp->pfs_1400.set_prio[1] = st->act.set_prio[1]; 5766 sp->pfs_1400.rt = st->act.rt; 5767 if (st->act.rt_kif) 5768 strlcpy(sp->pfs_1400.rt_ifname, 5769 st->act.rt_kif->pfik_name, 5770 sizeof(sp->pfs_1400.rt_ifname)); 5771 break; 5772 default: 5773 panic("%s: Unsupported pfsync_msg_version %d", 5774 __func__, msg_version); 5775 } 5776 5777 /* 5778 * XXX Why do we bother pfsyncing source node information if source 5779 * nodes are not synced? Showing users that there is source tracking 5780 * when there is none seems useless. 5781 */ 5782 if (st->sns[PF_SN_LIMIT] != NULL) 5783 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE; 5784 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE]) 5785 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5786 5787 sp->pfs_1301.id = st->id; 5788 sp->pfs_1301.creatorid = st->creatorid; 5789 pf_state_peer_hton(&st->src, &sp->pfs_1301.src); 5790 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst); 5791 5792 if (st->rule == NULL) 5793 sp->pfs_1301.rule = htonl(-1); 5794 else 5795 sp->pfs_1301.rule = htonl(st->rule->nr); 5796 if (st->anchor == NULL) 5797 sp->pfs_1301.anchor = htonl(-1); 5798 else 5799 sp->pfs_1301.anchor = htonl(st->anchor->nr); 5800 if (st->nat_rule == NULL) 5801 sp->pfs_1301.nat_rule = htonl(-1); 5802 else 5803 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr); 5804 5805 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]); 5806 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]); 5807 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]); 5808 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]); 5809 } 5810 5811 void 5812 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st) 5813 { 5814 bzero(sp, sizeof(*sp)); 5815 5816 sp->version = PF_STATE_VERSION; 5817 5818 /* copy from state key */ 5819 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5820 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5821 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5822 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5823 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5824 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5825 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5826 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5827 sp->proto = st->key[PF_SK_WIRE]->proto; 5828 sp->af = st->key[PF_SK_WIRE]->af; 5829 5830 /* copy from state */ 5831 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5832 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name, 5833 sizeof(sp->orig_ifname)); 5834 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr)); 5835 sp->creation = htonl(time_uptime - (st->creation / 1000)); 5836 sp->expire = pf_state_expires(st); 5837 if (sp->expire <= time_uptime) 5838 sp->expire = htonl(0); 5839 else 5840 sp->expire = htonl(sp->expire - time_uptime); 5841 5842 sp->direction = st->direction; 5843 sp->log = st->act.log; 5844 sp->timeout = st->timeout; 5845 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */ 5846 sp->state_flags_compat = st->state_flags; 5847 sp->state_flags = htons(st->state_flags); 5848 if (st->sns[PF_SN_LIMIT] != NULL) 5849 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5850 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL) 5851 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5852 sp->id = st->id; 5853 sp->creatorid = st->creatorid; 5854 pf_state_peer_hton(&st->src, &sp->src); 5855 pf_state_peer_hton(&st->dst, &sp->dst); 5856 5857 if (st->rule == NULL) 5858 sp->rule = htonl(-1); 5859 else 5860 sp->rule = htonl(st->rule->nr); 5861 if (st->anchor == NULL) 5862 sp->anchor = htonl(-1); 5863 else 5864 sp->anchor = htonl(st->anchor->nr); 5865 if (st->nat_rule == NULL) 5866 sp->nat_rule = htonl(-1); 5867 else 5868 sp->nat_rule = htonl(st->nat_rule->nr); 5869 5870 sp->packets[0] = st->packets[0]; 5871 sp->packets[1] = st->packets[1]; 5872 sp->bytes[0] = st->bytes[0]; 5873 sp->bytes[1] = st->bytes[1]; 5874 5875 sp->qid = htons(st->act.qid); 5876 sp->pqid = htons(st->act.pqid); 5877 sp->dnpipe = htons(st->act.dnpipe); 5878 sp->dnrpipe = htons(st->act.dnrpipe); 5879 sp->rtableid = htonl(st->act.rtableid); 5880 sp->min_ttl = st->act.min_ttl; 5881 sp->set_tos = st->act.set_tos; 5882 sp->max_mss = htons(st->act.max_mss); 5883 sp->rt = st->act.rt; 5884 if (st->act.rt_kif) 5885 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name, 5886 sizeof(sp->rt_ifname)); 5887 sp->set_prio[0] = st->act.set_prio[0]; 5888 sp->set_prio[1] = st->act.set_prio[1]; 5889 5890 } 5891 5892 static void 5893 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 5894 { 5895 struct pfr_ktable *kt; 5896 5897 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); 5898 5899 kt = aw->p.tbl; 5900 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 5901 kt = kt->pfrkt_root; 5902 aw->p.tbl = NULL; 5903 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 5904 kt->pfrkt_cnt : -1; 5905 } 5906 5907 static int 5908 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters, 5909 size_t number, char **names) 5910 { 5911 nvlist_t *nvc; 5912 5913 nvc = nvlist_create(0); 5914 if (nvc == NULL) 5915 return (ENOMEM); 5916 5917 for (int i = 0; i < number; i++) { 5918 nvlist_append_number_array(nvc, "counters", 5919 counter_u64_fetch(counters[i])); 5920 nvlist_append_string_array(nvc, "names", 5921 names[i]); 5922 nvlist_append_number_array(nvc, "ids", 5923 i); 5924 } 5925 nvlist_add_nvlist(nvl, name, nvc); 5926 nvlist_destroy(nvc); 5927 5928 return (0); 5929 } 5930 5931 static int 5932 pf_getstatus(struct pfioc_nv *nv) 5933 { 5934 nvlist_t *nvl = NULL, *nvc = NULL; 5935 void *nvlpacked = NULL; 5936 int error; 5937 struct pf_status s; 5938 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES; 5939 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES; 5940 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES; 5941 time_t since; 5942 5943 PF_RULES_RLOCK_TRACKER; 5944 5945 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 5946 5947 PF_RULES_RLOCK(); 5948 5949 nvl = nvlist_create(0); 5950 if (nvl == NULL) 5951 ERROUT(ENOMEM); 5952 5953 since = time_second - (time_uptime - V_pf_status.since); 5954 5955 nvlist_add_bool(nvl, "running", V_pf_status.running); 5956 nvlist_add_number(nvl, "since", since); 5957 nvlist_add_number(nvl, "debug", V_pf_status.debug); 5958 nvlist_add_number(nvl, "hostid", V_pf_status.hostid); 5959 nvlist_add_number(nvl, "states", V_pf_status.states); 5960 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes); 5961 nvlist_add_number(nvl, "reass", V_pf_status.reass); 5962 nvlist_add_bool(nvl, "syncookies_active", 5963 V_pf_status.syncookies_active); 5964 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen); 5965 5966 /* counters */ 5967 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters, 5968 PFRES_MAX, pf_reasons); 5969 if (error != 0) 5970 ERROUT(error); 5971 5972 /* lcounters */ 5973 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters, 5974 KLCNT_MAX, pf_lcounter); 5975 if (error != 0) 5976 ERROUT(error); 5977 5978 /* fcounters */ 5979 nvc = nvlist_create(0); 5980 if (nvc == NULL) 5981 ERROUT(ENOMEM); 5982 5983 for (int i = 0; i < FCNT_MAX; i++) { 5984 nvlist_append_number_array(nvc, "counters", 5985 pf_counter_u64_fetch(&V_pf_status.fcounters[i])); 5986 nvlist_append_string_array(nvc, "names", 5987 pf_fcounter[i]); 5988 nvlist_append_number_array(nvc, "ids", 5989 i); 5990 } 5991 nvlist_add_nvlist(nvl, "fcounters", nvc); 5992 nvlist_destroy(nvc); 5993 nvc = NULL; 5994 5995 /* scounters */ 5996 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters, 5997 SCNT_MAX, pf_fcounter); 5998 if (error != 0) 5999 ERROUT(error); 6000 6001 nvlist_add_string(nvl, "ifname", V_pf_status.ifname); 6002 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum, 6003 PF_MD5_DIGEST_LENGTH); 6004 6005 pfi_update_status(V_pf_status.ifname, &s); 6006 6007 /* pcounters / bcounters */ 6008 for (int i = 0; i < 2; i++) { 6009 for (int j = 0; j < 2; j++) { 6010 for (int k = 0; k < 2; k++) { 6011 nvlist_append_number_array(nvl, "pcounters", 6012 s.pcounters[i][j][k]); 6013 } 6014 nvlist_append_number_array(nvl, "bcounters", 6015 s.bcounters[i][j]); 6016 } 6017 } 6018 6019 nvlpacked = nvlist_pack(nvl, &nv->len); 6020 if (nvlpacked == NULL) 6021 ERROUT(ENOMEM); 6022 6023 if (nv->size == 0) 6024 ERROUT(0); 6025 else if (nv->size < nv->len) 6026 ERROUT(ENOSPC); 6027 6028 PF_RULES_RUNLOCK(); 6029 error = copyout(nvlpacked, nv->data, nv->len); 6030 goto done; 6031 6032 #undef ERROUT 6033 errout: 6034 PF_RULES_RUNLOCK(); 6035 done: 6036 free(nvlpacked, M_NVLIST); 6037 nvlist_destroy(nvc); 6038 nvlist_destroy(nvl); 6039 6040 return (error); 6041 } 6042 6043 /* 6044 * XXX - Check for version mismatch!!! 6045 */ 6046 static void 6047 pf_clear_all_states(void) 6048 { 6049 struct epoch_tracker et; 6050 struct pf_kstate *s; 6051 u_int i; 6052 6053 NET_EPOCH_ENTER(et); 6054 for (i = 0; i <= V_pf_hashmask; i++) { 6055 struct pf_idhash *ih = &V_pf_idhash[i]; 6056 relock: 6057 PF_HASHROW_LOCK(ih); 6058 LIST_FOREACH(s, &ih->states, entry) { 6059 s->timeout = PFTM_PURGE; 6060 /* Don't send out individual delete messages. */ 6061 s->state_flags |= PFSTATE_NOSYNC; 6062 pf_remove_state(s); 6063 goto relock; 6064 } 6065 PF_HASHROW_UNLOCK(ih); 6066 } 6067 NET_EPOCH_EXIT(et); 6068 } 6069 6070 static int 6071 pf_clear_tables(void) 6072 { 6073 struct pfioc_table io; 6074 int error; 6075 6076 bzero(&io, sizeof(io)); 6077 io.pfrio_flags |= PFR_FLAG_ALLRSETS; 6078 6079 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, 6080 io.pfrio_flags); 6081 6082 return (error); 6083 } 6084 6085 static void 6086 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) 6087 { 6088 struct pf_ksrc_node_list kill; 6089 u_int killed; 6090 6091 LIST_INIT(&kill); 6092 for (int i = 0; i <= V_pf_srchashmask; i++) { 6093 struct pf_srchash *sh = &V_pf_srchash[i]; 6094 struct pf_ksrc_node *sn, *tmp; 6095 6096 PF_HASHROW_LOCK(sh); 6097 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) 6098 if (psnk == NULL || 6099 (pf_match_addr(psnk->psnk_src.neg, 6100 &psnk->psnk_src.addr.v.a.addr, 6101 &psnk->psnk_src.addr.v.a.mask, 6102 &sn->addr, sn->af) && 6103 pf_match_addr(psnk->psnk_dst.neg, 6104 &psnk->psnk_dst.addr.v.a.addr, 6105 &psnk->psnk_dst.addr.v.a.mask, 6106 &sn->raddr, sn->af))) { 6107 pf_unlink_src_node(sn); 6108 LIST_INSERT_HEAD(&kill, sn, entry); 6109 sn->expire = 1; 6110 } 6111 PF_HASHROW_UNLOCK(sh); 6112 } 6113 6114 for (int i = 0; i <= V_pf_hashmask; i++) { 6115 struct pf_idhash *ih = &V_pf_idhash[i]; 6116 struct pf_kstate *s; 6117 6118 PF_HASHROW_LOCK(ih); 6119 LIST_FOREACH(s, &ih->states, entry) { 6120 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; 6121 sn_type++) { 6122 if (s->sns[sn_type] && 6123 s->sns[sn_type]->expire == 1) { 6124 s->sns[sn_type] = NULL; 6125 } 6126 } 6127 } 6128 PF_HASHROW_UNLOCK(ih); 6129 } 6130 6131 killed = pf_free_src_nodes(&kill); 6132 6133 if (psnk != NULL) 6134 psnk->psnk_killed = killed; 6135 } 6136 6137 static int 6138 pf_keepcounters(struct pfioc_nv *nv) 6139 { 6140 nvlist_t *nvl = NULL; 6141 void *nvlpacked = NULL; 6142 int error = 0; 6143 6144 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6145 6146 if (nv->len > pf_ioctl_maxcount) 6147 ERROUT(ENOMEM); 6148 6149 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6150 error = copyin(nv->data, nvlpacked, nv->len); 6151 if (error) 6152 ERROUT(error); 6153 6154 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6155 if (nvl == NULL) 6156 ERROUT(EBADMSG); 6157 6158 if (! nvlist_exists_bool(nvl, "keep_counters")) 6159 ERROUT(EBADMSG); 6160 6161 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters"); 6162 6163 on_error: 6164 nvlist_destroy(nvl); 6165 free(nvlpacked, M_NVLIST); 6166 return (error); 6167 } 6168 6169 unsigned int 6170 pf_clear_states(const struct pf_kstate_kill *kill) 6171 { 6172 struct pf_state_key_cmp match_key; 6173 struct pf_kstate *s; 6174 struct pfi_kkif *kif; 6175 int idx; 6176 unsigned int killed = 0, dir; 6177 6178 NET_EPOCH_ASSERT(); 6179 6180 for (unsigned int i = 0; i <= V_pf_hashmask; i++) { 6181 struct pf_idhash *ih = &V_pf_idhash[i]; 6182 6183 relock_DIOCCLRSTATES: 6184 PF_HASHROW_LOCK(ih); 6185 LIST_FOREACH(s, &ih->states, entry) { 6186 /* For floating states look at the original kif. */ 6187 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 6188 6189 if (kill->psk_ifname[0] && 6190 strcmp(kill->psk_ifname, 6191 kif->pfik_name)) 6192 continue; 6193 6194 if (kill->psk_kill_match) { 6195 bzero(&match_key, sizeof(match_key)); 6196 6197 if (s->direction == PF_OUT) { 6198 dir = PF_IN; 6199 idx = PF_SK_STACK; 6200 } else { 6201 dir = PF_OUT; 6202 idx = PF_SK_WIRE; 6203 } 6204 6205 match_key.af = s->key[idx]->af; 6206 match_key.proto = s->key[idx]->proto; 6207 pf_addrcpy(&match_key.addr[0], 6208 &s->key[idx]->addr[1], match_key.af); 6209 match_key.port[0] = s->key[idx]->port[1]; 6210 pf_addrcpy(&match_key.addr[1], 6211 &s->key[idx]->addr[0], match_key.af); 6212 match_key.port[1] = s->key[idx]->port[0]; 6213 } 6214 6215 /* 6216 * Don't send out individual 6217 * delete messages. 6218 */ 6219 s->state_flags |= PFSTATE_NOSYNC; 6220 pf_remove_state(s); 6221 killed++; 6222 6223 if (kill->psk_kill_match) 6224 killed += pf_kill_matching_state(&match_key, 6225 dir); 6226 6227 goto relock_DIOCCLRSTATES; 6228 } 6229 PF_HASHROW_UNLOCK(ih); 6230 } 6231 6232 if (V_pfsync_clear_states_ptr != NULL) 6233 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname); 6234 6235 return (killed); 6236 } 6237 6238 void 6239 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed) 6240 { 6241 struct pf_kstate *s; 6242 6243 NET_EPOCH_ASSERT(); 6244 if (kill->psk_pfcmp.id) { 6245 if (kill->psk_pfcmp.creatorid == 0) 6246 kill->psk_pfcmp.creatorid = V_pf_status.hostid; 6247 if ((s = pf_find_state_byid(kill->psk_pfcmp.id, 6248 kill->psk_pfcmp.creatorid))) { 6249 pf_remove_state(s); 6250 *killed = 1; 6251 } 6252 return; 6253 } 6254 6255 for (unsigned int i = 0; i <= V_pf_hashmask; i++) 6256 *killed += pf_killstates_row(kill, &V_pf_idhash[i]); 6257 } 6258 6259 static int 6260 pf_killstates_nv(struct pfioc_nv *nv) 6261 { 6262 struct pf_kstate_kill kill; 6263 struct epoch_tracker et; 6264 nvlist_t *nvl = NULL; 6265 void *nvlpacked = NULL; 6266 int error = 0; 6267 unsigned int killed = 0; 6268 6269 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6270 6271 if (nv->len > pf_ioctl_maxcount) 6272 ERROUT(ENOMEM); 6273 6274 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6275 error = copyin(nv->data, nvlpacked, nv->len); 6276 if (error) 6277 ERROUT(error); 6278 6279 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6280 if (nvl == NULL) 6281 ERROUT(EBADMSG); 6282 6283 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6284 if (error) 6285 ERROUT(error); 6286 6287 NET_EPOCH_ENTER(et); 6288 pf_killstates(&kill, &killed); 6289 NET_EPOCH_EXIT(et); 6290 6291 free(nvlpacked, M_NVLIST); 6292 nvlpacked = NULL; 6293 nvlist_destroy(nvl); 6294 nvl = nvlist_create(0); 6295 if (nvl == NULL) 6296 ERROUT(ENOMEM); 6297 6298 nvlist_add_number(nvl, "killed", killed); 6299 6300 nvlpacked = nvlist_pack(nvl, &nv->len); 6301 if (nvlpacked == NULL) 6302 ERROUT(ENOMEM); 6303 6304 if (nv->size == 0) 6305 ERROUT(0); 6306 else if (nv->size < nv->len) 6307 ERROUT(ENOSPC); 6308 6309 error = copyout(nvlpacked, nv->data, nv->len); 6310 6311 on_error: 6312 nvlist_destroy(nvl); 6313 free(nvlpacked, M_NVLIST); 6314 return (error); 6315 } 6316 6317 static int 6318 pf_clearstates_nv(struct pfioc_nv *nv) 6319 { 6320 struct pf_kstate_kill kill; 6321 struct epoch_tracker et; 6322 nvlist_t *nvl = NULL; 6323 void *nvlpacked = NULL; 6324 int error = 0; 6325 unsigned int killed; 6326 6327 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6328 6329 if (nv->len > pf_ioctl_maxcount) 6330 ERROUT(ENOMEM); 6331 6332 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6333 error = copyin(nv->data, nvlpacked, nv->len); 6334 if (error) 6335 ERROUT(error); 6336 6337 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6338 if (nvl == NULL) 6339 ERROUT(EBADMSG); 6340 6341 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6342 if (error) 6343 ERROUT(error); 6344 6345 NET_EPOCH_ENTER(et); 6346 killed = pf_clear_states(&kill); 6347 NET_EPOCH_EXIT(et); 6348 6349 free(nvlpacked, M_NVLIST); 6350 nvlpacked = NULL; 6351 nvlist_destroy(nvl); 6352 nvl = nvlist_create(0); 6353 if (nvl == NULL) 6354 ERROUT(ENOMEM); 6355 6356 nvlist_add_number(nvl, "killed", killed); 6357 6358 nvlpacked = nvlist_pack(nvl, &nv->len); 6359 if (nvlpacked == NULL) 6360 ERROUT(ENOMEM); 6361 6362 if (nv->size == 0) 6363 ERROUT(0); 6364 else if (nv->size < nv->len) 6365 ERROUT(ENOSPC); 6366 6367 error = copyout(nvlpacked, nv->data, nv->len); 6368 6369 #undef ERROUT 6370 on_error: 6371 nvlist_destroy(nvl); 6372 free(nvlpacked, M_NVLIST); 6373 return (error); 6374 } 6375 6376 static int 6377 pf_getstate(struct pfioc_nv *nv) 6378 { 6379 nvlist_t *nvl = NULL, *nvls; 6380 void *nvlpacked = NULL; 6381 struct pf_kstate *s = NULL; 6382 int error = 0; 6383 uint64_t id, creatorid; 6384 6385 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6386 6387 if (nv->len > pf_ioctl_maxcount) 6388 ERROUT(ENOMEM); 6389 6390 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6391 error = copyin(nv->data, nvlpacked, nv->len); 6392 if (error) 6393 ERROUT(error); 6394 6395 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6396 if (nvl == NULL) 6397 ERROUT(EBADMSG); 6398 6399 PFNV_CHK(pf_nvuint64(nvl, "id", &id)); 6400 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid)); 6401 6402 s = pf_find_state_byid(id, creatorid); 6403 if (s == NULL) 6404 ERROUT(ENOENT); 6405 6406 free(nvlpacked, M_NVLIST); 6407 nvlpacked = NULL; 6408 nvlist_destroy(nvl); 6409 nvl = nvlist_create(0); 6410 if (nvl == NULL) 6411 ERROUT(ENOMEM); 6412 6413 nvls = pf_state_to_nvstate(s); 6414 if (nvls == NULL) 6415 ERROUT(ENOMEM); 6416 6417 nvlist_add_nvlist(nvl, "state", nvls); 6418 nvlist_destroy(nvls); 6419 6420 nvlpacked = nvlist_pack(nvl, &nv->len); 6421 if (nvlpacked == NULL) 6422 ERROUT(ENOMEM); 6423 6424 if (nv->size == 0) 6425 ERROUT(0); 6426 else if (nv->size < nv->len) 6427 ERROUT(ENOSPC); 6428 6429 error = copyout(nvlpacked, nv->data, nv->len); 6430 6431 #undef ERROUT 6432 errout: 6433 if (s != NULL) 6434 PF_STATE_UNLOCK(s); 6435 free(nvlpacked, M_NVLIST); 6436 nvlist_destroy(nvl); 6437 return (error); 6438 } 6439 6440 /* 6441 * XXX - Check for version mismatch!!! 6442 */ 6443 6444 /* 6445 * Duplicate pfctl -Fa operation to get rid of as much as we can. 6446 */ 6447 static int 6448 shutdown_pf(void) 6449 { 6450 int error = 0; 6451 u_int32_t t[5]; 6452 char nn = '\0'; 6453 struct pf_kanchor *anchor, *tmp_anchor; 6454 struct pf_keth_anchor *eth_anchor, *tmp_eth_anchor; 6455 int rs_num; 6456 6457 do { 6458 /* Unlink rules of all user defined anchors */ 6459 RB_FOREACH_SAFE(anchor, pf_kanchor_global, &V_pf_anchors, 6460 tmp_anchor) { 6461 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6462 if ((error = pf_begin_rules(&t[rs_num], rs_num, 6463 anchor->path)) != 0) { 6464 DPFPRINTF(PF_DEBUG_MISC, "%s: " 6465 "anchor.path=%s rs_num=%d", 6466 __func__, anchor->path, rs_num); 6467 goto error; /* XXX: rollback? */ 6468 } 6469 } 6470 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6471 error = pf_commit_rules(t[rs_num], rs_num, 6472 anchor->path); 6473 MPASS(error == 0); 6474 } 6475 } 6476 6477 /* Unlink rules of all user defined ether anchors */ 6478 RB_FOREACH_SAFE(eth_anchor, pf_keth_anchor_global, 6479 &V_pf_keth_anchors, tmp_eth_anchor) { 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