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 void pf_qid_unref(uint16_t); 120 #endif /* ALTQ */ 121 static int pf_begin_rules(u_int32_t *, int, const char *); 122 static int pf_rollback_rules(u_int32_t, int, char *); 123 static int pf_setup_pfsync_matching(struct pf_kruleset *); 124 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *); 125 static void pf_hash_rule(struct pf_krule *); 126 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *); 127 static int pf_commit_rules(u_int32_t, int, char *); 128 static int pf_addr_setup(struct pf_kruleset *, 129 struct pf_addr_wrap *, sa_family_t); 130 static void pf_src_node_copy(const struct pf_ksrc_node *, 131 struct pf_src_node *); 132 #ifdef ALTQ 133 static int pf_export_kaltq(struct pf_altq *, 134 struct pfioc_altq_v1 *, size_t); 135 static int pf_import_kaltq(struct pfioc_altq_v1 *, 136 struct pf_altq *, size_t); 137 #endif /* ALTQ */ 138 139 VNET_DEFINE(struct pf_krule, pf_default_rule); 140 141 static __inline int pf_krule_compare(struct pf_krule *, 142 struct pf_krule *); 143 144 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare); 145 146 #ifdef ALTQ 147 VNET_DEFINE_STATIC(int, pf_altq_running); 148 #define V_pf_altq_running VNET(pf_altq_running) 149 #endif 150 151 #define TAGID_MAX 50000 152 struct pf_tagname { 153 TAILQ_ENTRY(pf_tagname) namehash_entries; 154 TAILQ_ENTRY(pf_tagname) taghash_entries; 155 char name[PF_TAG_NAME_SIZE]; 156 uint16_t tag; 157 int ref; 158 }; 159 160 struct pf_tagset { 161 TAILQ_HEAD(, pf_tagname) *namehash; 162 TAILQ_HEAD(, pf_tagname) *taghash; 163 unsigned int mask; 164 uint32_t seed; 165 BITSET_DEFINE(, TAGID_MAX) avail; 166 }; 167 168 VNET_DEFINE(struct pf_tagset, pf_tags); 169 #define V_pf_tags VNET(pf_tags) 170 static unsigned int pf_rule_tag_hashsize; 171 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128 172 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN, 173 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT, 174 "Size of pf(4) rule tag hashtable"); 175 176 #ifdef ALTQ 177 VNET_DEFINE(struct pf_tagset, pf_qids); 178 #define V_pf_qids VNET(pf_qids) 179 static unsigned int pf_queue_tag_hashsize; 180 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128 181 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN, 182 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT, 183 "Size of pf(4) queue tag hashtable"); 184 #endif 185 VNET_DEFINE(uma_zone_t, pf_tag_z); 186 #define V_pf_tag_z VNET(pf_tag_z) 187 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db"); 188 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules"); 189 MALLOC_DEFINE(M_PF, "pf", "pf(4)"); 190 191 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE) 192 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE 193 #endif 194 195 VNET_DEFINE_STATIC(bool, pf_filter_local) = false; 196 #define V_pf_filter_local VNET(pf_filter_local) 197 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW, 198 &VNET_NAME(pf_filter_local), false, 199 "Enable filtering for packets delivered to local network stack"); 200 201 #ifdef PF_DEFAULT_TO_DROP 202 VNET_DEFINE_STATIC(bool, default_to_drop) = true; 203 #else 204 VNET_DEFINE_STATIC(bool, default_to_drop); 205 #endif 206 #define V_default_to_drop VNET(default_to_drop) 207 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET, 208 &VNET_NAME(default_to_drop), false, 209 "Make the default rule drop all packets."); 210 211 static void pf_init_tagset(struct pf_tagset *, unsigned int *, 212 unsigned int); 213 static void pf_cleanup_tagset(struct pf_tagset *); 214 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *); 215 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t); 216 static u_int16_t tagname2tag(struct pf_tagset *, const char *, bool); 217 static void tag_unref(struct pf_tagset *, u_int16_t); 218 219 struct cdev *pf_dev; 220 221 /* 222 * XXX - These are new and need to be checked when moveing to a new version 223 */ 224 static void pf_clear_all_states(void); 225 static int pf_killstates_row(struct pf_kstate_kill *, 226 struct pf_idhash *); 227 static int pf_killstates_nv(struct pfioc_nv *); 228 static int pf_clearstates_nv(struct pfioc_nv *); 229 static int pf_getstate(struct pfioc_nv *); 230 static int pf_getstatus(struct pfioc_nv *); 231 static int pf_clear_tables(void); 232 static void pf_kill_srcnodes(struct pfioc_src_node_kill *); 233 static int pf_keepcounters(struct pfioc_nv *); 234 static void pf_tbladdr_copyout(struct pf_addr_wrap *); 235 236 /* 237 * Wrapper functions for pfil(9) hooks 238 */ 239 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, 240 int flags, void *ruleset __unused, struct inpcb *inp); 241 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, 242 int flags, void *ruleset __unused, struct inpcb *inp); 243 #ifdef INET 244 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp, 245 int flags, void *ruleset __unused, struct inpcb *inp); 246 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp, 247 int flags, void *ruleset __unused, struct inpcb *inp); 248 #endif 249 #ifdef INET6 250 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp, 251 int flags, void *ruleset __unused, struct inpcb *inp); 252 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp, 253 int flags, void *ruleset __unused, struct inpcb *inp); 254 #endif 255 256 static void hook_pf_eth(void); 257 static void hook_pf(void); 258 static void dehook_pf_eth(void); 259 static void dehook_pf(void); 260 static int shutdown_pf(void); 261 static int pf_load(void); 262 static void pf_unload(void *); 263 264 static struct cdevsw pf_cdevsw = { 265 .d_ioctl = pfioctl, 266 .d_name = PF_NAME, 267 .d_version = D_VERSION, 268 }; 269 270 VNET_DEFINE_STATIC(bool, pf_pfil_hooked); 271 #define V_pf_pfil_hooked VNET(pf_pfil_hooked) 272 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked); 273 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked) 274 275 /* 276 * We need a flag that is neither hooked nor running to know when 277 * the VNET is "valid". We primarily need this to control (global) 278 * external event, e.g., eventhandlers. 279 */ 280 VNET_DEFINE(int, pf_vnet_active); 281 #define V_pf_vnet_active VNET(pf_vnet_active) 282 283 int pf_end_threads; 284 struct proc *pf_purge_proc; 285 286 VNET_DEFINE(struct rmlock, pf_rules_lock); 287 VNET_DEFINE(struct rmlock, pf_tags_lock); 288 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock); 289 #define V_pf_ioctl_lock VNET(pf_ioctl_lock) 290 struct sx pf_end_lock; 291 292 /* pfsync */ 293 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr); 294 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr); 295 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr); 296 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr); 297 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr); 298 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr); 299 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr); 300 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr; 301 302 /* pflog */ 303 pflog_packet_t *pflog_packet_ptr = NULL; 304 305 /* 306 * Copy a user-provided string, returning an error if truncation would occur. 307 * Avoid scanning past "sz" bytes in the source string since there's no 308 * guarantee that it's nul-terminated. 309 */ 310 static int 311 pf_user_strcpy(char *dst, const char *src, size_t sz) 312 { 313 if (strnlen(src, sz) == sz) 314 return (EINVAL); 315 (void)strlcpy(dst, src, sz); 316 return (0); 317 } 318 319 static void 320 pfattach_vnet(void) 321 { 322 u_int32_t *my_timeout = V_pf_default_rule.timeout; 323 324 bzero(&V_pf_status, sizeof(V_pf_status)); 325 326 pf_initialize(); 327 pfr_initialize(); 328 pfi_initialize_vnet(); 329 pf_normalize_init(); 330 pf_syncookies_init(); 331 332 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT; 333 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT; 334 V_pf_limits[PF_LIMIT_ANCHORS].limit = PF_ANCHOR_HIWAT; 335 V_pf_limits[PF_LIMIT_ETH_ANCHORS].limit = PF_ANCHOR_HIWAT; 336 337 RB_INIT(&V_pf_anchors); 338 pf_init_kruleset(&pf_main_ruleset); 339 340 pf_init_keth(V_pf_keth); 341 342 /* default rule should never be garbage collected */ 343 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; 344 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS; 345 V_pf_default_rule.nr = (uint32_t)-1; 346 V_pf_default_rule.rtableid = -1; 347 348 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK); 349 for (int i = 0; i < 2; i++) { 350 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK); 351 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK); 352 } 353 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); 354 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); 355 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++) 356 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 357 358 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 359 M_WAITOK | M_ZERO); 360 361 #ifdef PF_WANT_32_TO_64_COUNTER 362 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO); 363 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO); 364 PF_RULES_WLOCK(); 365 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); 366 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist); 367 V_pf_allrulecount++; 368 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); 369 PF_RULES_WUNLOCK(); 370 #endif 371 372 /* initialize default timeouts */ 373 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 374 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; 375 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 376 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; 377 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; 378 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; 379 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; 380 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL; 381 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; 382 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL; 383 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL; 384 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; 385 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; 386 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; 387 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; 388 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; 389 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; 390 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; 391 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; 392 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; 393 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; 394 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; 395 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; 396 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; 397 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; 398 399 V_pf_status.debug = PF_DEBUG_URGENT; 400 /* 401 * XXX This is different than in OpenBSD where reassembly is enabled by 402 * defult. In FreeBSD we expect people to still use scrub rules and 403 * switch to the new syntax later. Only when they switch they must 404 * explicitly enable reassemle. We could change the default once the 405 * scrub rule functionality is hopefully removed some day in future. 406 */ 407 V_pf_status.reass = 0; 408 409 V_pf_pfil_hooked = false; 410 V_pf_pfil_eth_hooked = false; 411 412 /* XXX do our best to avoid a conflict */ 413 V_pf_status.hostid = arc4random(); 414 415 for (int i = 0; i < PFRES_MAX; i++) 416 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); 417 for (int i = 0; i < KLCNT_MAX; i++) 418 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); 419 for (int i = 0; i < FCNT_MAX; i++) 420 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK); 421 for (int i = 0; i < SCNT_MAX; i++) 422 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); 423 for (int i = 0; i < NCNT_MAX; i++) 424 V_pf_status.ncounters[i] = counter_u64_alloc(M_WAITOK); 425 426 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET, 427 INTR_MPSAFE, &V_pf_swi_cookie) != 0) 428 /* XXXGL: leaked all above. */ 429 return; 430 } 431 432 static struct pf_kpool * 433 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action, 434 u_int32_t rule_number, u_int8_t r_last, u_int8_t active, 435 u_int8_t check_ticket, int which) 436 { 437 struct pf_kruleset *ruleset; 438 struct pf_krule *rule; 439 int rs_num; 440 441 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT); 442 443 ruleset = pf_find_kruleset(anchor); 444 if (ruleset == NULL) 445 return (NULL); 446 rs_num = pf_get_ruleset_number(rule_action); 447 if (rs_num >= PF_RULESET_MAX) 448 return (NULL); 449 if (active) { 450 if (check_ticket && ticket != 451 ruleset->rules[rs_num].active.ticket) 452 return (NULL); 453 if (r_last) 454 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 455 pf_krulequeue); 456 else 457 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 458 } else { 459 if (check_ticket && ticket != 460 ruleset->rules[rs_num].inactive.ticket) 461 return (NULL); 462 if (r_last) 463 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 464 pf_krulequeue); 465 else 466 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); 467 } 468 if (!r_last) { 469 while ((rule != NULL) && (rule->nr != rule_number)) 470 rule = TAILQ_NEXT(rule, entries); 471 } 472 if (rule == NULL) 473 return (NULL); 474 475 switch (which) { 476 case PF_RDR: 477 return (&rule->rdr); 478 case PF_NAT: 479 return (&rule->nat); 480 case PF_RT: 481 return (&rule->route); 482 default: 483 panic("Unknow pool type %d", which); 484 } 485 } 486 487 static void 488 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb) 489 { 490 struct pf_kpooladdr *mv_pool_pa; 491 492 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { 493 TAILQ_REMOVE(poola, mv_pool_pa, entries); 494 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); 495 } 496 } 497 498 static void 499 pf_empty_kpool(struct pf_kpalist *poola) 500 { 501 struct pf_kpooladdr *pa; 502 503 while ((pa = TAILQ_FIRST(poola)) != NULL) { 504 switch (pa->addr.type) { 505 case PF_ADDR_DYNIFTL: 506 pfi_dynaddr_remove(pa->addr.p.dyn); 507 break; 508 case PF_ADDR_TABLE: 509 /* XXX: this could be unfinished pooladdr on pabuf */ 510 if (pa->addr.p.tbl != NULL) 511 pfr_detach_table(pa->addr.p.tbl); 512 break; 513 } 514 if (pa->kif) 515 pfi_kkif_unref(pa->kif); 516 TAILQ_REMOVE(poola, pa, entries); 517 free(pa, M_PFRULE); 518 } 519 } 520 521 static void 522 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 523 { 524 525 PF_RULES_WASSERT(); 526 PF_UNLNKDRULES_ASSERT(); 527 528 TAILQ_REMOVE(rulequeue, rule, entries); 529 530 rule->rule_ref |= PFRULE_REFS; 531 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); 532 } 533 534 static void 535 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule) 536 { 537 538 PF_RULES_WASSERT(); 539 540 PF_UNLNKDRULES_LOCK(); 541 pf_unlink_rule_locked(rulequeue, rule); 542 PF_UNLNKDRULES_UNLOCK(); 543 } 544 545 static void 546 pf_free_eth_rule(struct pf_keth_rule *rule) 547 { 548 PF_RULES_WASSERT(); 549 550 if (rule == NULL) 551 return; 552 553 if (rule->tag) 554 tag_unref(&V_pf_tags, rule->tag); 555 if (rule->match_tag) 556 tag_unref(&V_pf_tags, rule->match_tag); 557 #ifdef ALTQ 558 pf_qid_unref(rule->qid); 559 #endif 560 561 if (rule->bridge_to) 562 pfi_kkif_unref(rule->bridge_to); 563 if (rule->kif) 564 pfi_kkif_unref(rule->kif); 565 566 if (rule->ipsrc.addr.type == PF_ADDR_TABLE) 567 pfr_detach_table(rule->ipsrc.addr.p.tbl); 568 if (rule->ipdst.addr.type == PF_ADDR_TABLE) 569 pfr_detach_table(rule->ipdst.addr.p.tbl); 570 571 counter_u64_free(rule->evaluations); 572 for (int i = 0; i < 2; i++) { 573 counter_u64_free(rule->packets[i]); 574 counter_u64_free(rule->bytes[i]); 575 } 576 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 577 pf_keth_anchor_remove(rule); 578 579 free(rule, M_PFRULE); 580 } 581 582 void 583 pf_free_rule(struct pf_krule *rule) 584 { 585 586 PF_RULES_WASSERT(); 587 PF_CONFIG_ASSERT(); 588 589 if (rule->tag) 590 tag_unref(&V_pf_tags, rule->tag); 591 if (rule->match_tag) 592 tag_unref(&V_pf_tags, rule->match_tag); 593 #ifdef ALTQ 594 if (rule->pqid != rule->qid) 595 pf_qid_unref(rule->pqid); 596 pf_qid_unref(rule->qid); 597 #endif 598 switch (rule->src.addr.type) { 599 case PF_ADDR_DYNIFTL: 600 pfi_dynaddr_remove(rule->src.addr.p.dyn); 601 break; 602 case PF_ADDR_TABLE: 603 pfr_detach_table(rule->src.addr.p.tbl); 604 break; 605 } 606 switch (rule->dst.addr.type) { 607 case PF_ADDR_DYNIFTL: 608 pfi_dynaddr_remove(rule->dst.addr.p.dyn); 609 break; 610 case PF_ADDR_TABLE: 611 pfr_detach_table(rule->dst.addr.p.tbl); 612 break; 613 } 614 if (rule->overload_tbl) 615 pfr_detach_table(rule->overload_tbl); 616 if (rule->kif) 617 pfi_kkif_unref(rule->kif); 618 if (rule->rcv_kif) 619 pfi_kkif_unref(rule->rcv_kif); 620 pf_remove_kanchor(rule); 621 pf_empty_kpool(&rule->rdr.list); 622 pf_empty_kpool(&rule->nat.list); 623 pf_empty_kpool(&rule->route.list); 624 625 pf_krule_free(rule); 626 } 627 628 static void 629 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size, 630 unsigned int default_size) 631 { 632 unsigned int i; 633 unsigned int hashsize; 634 635 if (*tunable_size == 0 || !powerof2(*tunable_size)) 636 *tunable_size = default_size; 637 638 hashsize = *tunable_size; 639 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH, 640 M_WAITOK); 641 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH, 642 M_WAITOK); 643 ts->mask = hashsize - 1; 644 ts->seed = arc4random(); 645 for (i = 0; i < hashsize; i++) { 646 TAILQ_INIT(&ts->namehash[i]); 647 TAILQ_INIT(&ts->taghash[i]); 648 } 649 BIT_FILL(TAGID_MAX, &ts->avail); 650 } 651 652 static void 653 pf_cleanup_tagset(struct pf_tagset *ts) 654 { 655 unsigned int i; 656 unsigned int hashsize; 657 struct pf_tagname *t, *tmp; 658 659 /* 660 * Only need to clean up one of the hashes as each tag is hashed 661 * into each table. 662 */ 663 hashsize = ts->mask + 1; 664 for (i = 0; i < hashsize; i++) 665 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp) 666 uma_zfree(V_pf_tag_z, t); 667 668 free(ts->namehash, M_PFHASH); 669 free(ts->taghash, M_PFHASH); 670 } 671 672 static uint16_t 673 tagname2hashindex(const struct pf_tagset *ts, const char *tagname) 674 { 675 size_t len; 676 677 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1); 678 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask); 679 } 680 681 static uint16_t 682 tag2hashindex(const struct pf_tagset *ts, uint16_t tag) 683 { 684 685 return (tag & ts->mask); 686 } 687 688 static u_int16_t 689 tagname2tag(struct pf_tagset *ts, const char *tagname, bool add_new) 690 { 691 struct pf_tagname *tag; 692 u_int32_t index; 693 u_int16_t new_tagid; 694 695 PF_TAGS_RLOCK_TRACKER; 696 697 PF_TAGS_RLOCK(); 698 699 index = tagname2hashindex(ts, tagname); 700 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 701 if (strcmp(tagname, tag->name) == 0) { 702 tag->ref++; 703 new_tagid = tag->tag; 704 PF_TAGS_RUNLOCK(); 705 return (new_tagid); 706 } 707 708 /* 709 * When used for pfsync with queues we must not create new entries. 710 * Pf tags can be created just fine by this function, but queues 711 * require additional configuration. If they are missing on the target 712 * system we just ignore them 713 */ 714 if (add_new == false) { 715 printf("%s: Not creating a new tag\n", __func__); 716 PF_TAGS_RUNLOCK(); 717 return (0); 718 } 719 720 /* 721 * If a new entry must be created do it under a write lock. 722 * But first search again, somebody could have created the tag 723 * between unlocking the read lock and locking the write lock. 724 */ 725 PF_TAGS_RUNLOCK(); 726 PF_TAGS_WLOCK(); 727 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries) 728 if (strcmp(tagname, tag->name) == 0) { 729 tag->ref++; 730 new_tagid = tag->tag; 731 PF_TAGS_WUNLOCK(); 732 return (new_tagid); 733 } 734 735 /* 736 * new entry 737 * 738 * to avoid fragmentation, we do a linear search from the beginning 739 * and take the first free slot we find. 740 */ 741 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail); 742 /* 743 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX]. 744 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits 745 * set. It may also return a bit number greater than TAGID_MAX due 746 * to rounding of the number of bits in the vector up to a multiple 747 * of the vector word size at declaration/allocation time. 748 */ 749 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) { 750 PF_TAGS_WUNLOCK(); 751 return (0); 752 } 753 754 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */ 755 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail); 756 757 /* allocate and fill new struct pf_tagname */ 758 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT); 759 if (tag == NULL) { 760 PF_TAGS_WUNLOCK(); 761 return (0); 762 } 763 strlcpy(tag->name, tagname, sizeof(tag->name)); 764 tag->tag = new_tagid; 765 tag->ref = 1; 766 767 /* Insert into namehash */ 768 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries); 769 770 /* Insert into taghash */ 771 index = tag2hashindex(ts, new_tagid); 772 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries); 773 774 PF_TAGS_WUNLOCK(); 775 return (new_tagid); 776 } 777 778 static char * 779 tag2tagname(struct pf_tagset *ts, u_int16_t tag) 780 { 781 struct pf_tagname *t; 782 uint16_t index; 783 784 PF_TAGS_RLOCK_TRACKER; 785 786 PF_TAGS_RLOCK(); 787 788 index = tag2hashindex(ts, tag); 789 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 790 if (tag == t->tag) { 791 PF_TAGS_RUNLOCK(); 792 return (t->name); 793 } 794 795 PF_TAGS_RUNLOCK(); 796 return (NULL); 797 } 798 799 static void 800 tag_unref(struct pf_tagset *ts, u_int16_t tag) 801 { 802 struct pf_tagname *t; 803 uint16_t index; 804 805 PF_TAGS_WLOCK(); 806 807 index = tag2hashindex(ts, tag); 808 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries) 809 if (tag == t->tag) { 810 if (--t->ref == 0) { 811 TAILQ_REMOVE(&ts->taghash[index], t, 812 taghash_entries); 813 index = tagname2hashindex(ts, t->name); 814 TAILQ_REMOVE(&ts->namehash[index], t, 815 namehash_entries); 816 /* Bits are 0-based for BIT_SET() */ 817 BIT_SET(TAGID_MAX, tag - 1, &ts->avail); 818 uma_zfree(V_pf_tag_z, t); 819 } 820 break; 821 } 822 823 PF_TAGS_WUNLOCK(); 824 } 825 826 uint16_t 827 pf_tagname2tag(const char *tagname) 828 { 829 return (tagname2tag(&V_pf_tags, tagname, true)); 830 } 831 832 static const char * 833 pf_tag2tagname(uint16_t tag) 834 { 835 return (tag2tagname(&V_pf_tags, tag)); 836 } 837 838 static int 839 pf_begin_eth(uint32_t *ticket, const char *anchor) 840 { 841 struct pf_keth_rule *rule, *tmp; 842 struct pf_keth_ruleset *rs; 843 844 PF_RULES_WASSERT(); 845 846 rs = pf_find_or_create_keth_ruleset(anchor); 847 if (rs == NULL) 848 return (EINVAL); 849 850 /* Purge old inactive rules. */ 851 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 852 tmp) { 853 TAILQ_REMOVE(rs->inactive.rules, rule, 854 entries); 855 pf_free_eth_rule(rule); 856 } 857 858 *ticket = ++rs->inactive.ticket; 859 rs->inactive.open = 1; 860 861 return (0); 862 } 863 864 static int 865 pf_rollback_eth(uint32_t ticket, const char *anchor) 866 { 867 struct pf_keth_rule *rule, *tmp; 868 struct pf_keth_ruleset *rs; 869 870 PF_RULES_WASSERT(); 871 872 rs = pf_find_keth_ruleset(anchor); 873 if (rs == NULL) 874 return (EINVAL); 875 876 if (!rs->inactive.open || 877 ticket != rs->inactive.ticket) 878 return (0); 879 880 /* Purge old inactive rules. */ 881 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries, 882 tmp) { 883 TAILQ_REMOVE(rs->inactive.rules, rule, entries); 884 pf_free_eth_rule(rule); 885 } 886 887 rs->inactive.open = 0; 888 889 pf_remove_if_empty_keth_ruleset(rs); 890 891 return (0); 892 } 893 894 #define PF_SET_SKIP_STEPS(i) \ 895 do { \ 896 while (head[i] != cur) { \ 897 head[i]->skip[i].ptr = cur; \ 898 head[i] = TAILQ_NEXT(head[i], entries); \ 899 } \ 900 } while (0) 901 902 static void 903 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules) 904 { 905 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT]; 906 int i; 907 908 cur = TAILQ_FIRST(rules); 909 prev = cur; 910 for (i = 0; i < PFE_SKIP_COUNT; ++i) 911 head[i] = cur; 912 while (cur != NULL) { 913 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 914 PF_SET_SKIP_STEPS(PFE_SKIP_IFP); 915 if (cur->direction != prev->direction) 916 PF_SET_SKIP_STEPS(PFE_SKIP_DIR); 917 if (cur->proto != prev->proto) 918 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO); 919 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0) 920 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR); 921 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0) 922 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR); 923 if (cur->ipsrc.neg != prev->ipsrc.neg || 924 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr)) 925 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR); 926 if (cur->ipdst.neg != prev->ipdst.neg || 927 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr)) 928 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR); 929 930 prev = cur; 931 cur = TAILQ_NEXT(cur, entries); 932 } 933 for (i = 0; i < PFE_SKIP_COUNT; ++i) 934 PF_SET_SKIP_STEPS(i); 935 } 936 937 static int 938 pf_commit_eth(uint32_t ticket, const char *anchor) 939 { 940 struct pf_keth_ruleq *rules; 941 struct pf_keth_ruleset *rs; 942 943 rs = pf_find_keth_ruleset(anchor); 944 if (rs == NULL) { 945 return (EINVAL); 946 } 947 948 if (!rs->inactive.open || 949 ticket != rs->inactive.ticket) 950 return (EBUSY); 951 952 PF_RULES_WASSERT(); 953 954 pf_eth_calc_skip_steps(rs->inactive.rules); 955 956 rules = rs->active.rules; 957 atomic_store_ptr(&rs->active.rules, rs->inactive.rules); 958 rs->inactive.rules = rules; 959 rs->inactive.ticket = rs->active.ticket; 960 961 return (pf_rollback_eth(rs->inactive.ticket, 962 rs->anchor ? rs->anchor->path : "")); 963 } 964 965 #ifdef ALTQ 966 uint16_t 967 pf_qname2qid(const char *qname, bool add_new) 968 { 969 return (tagname2tag(&V_pf_qids, qname, add_new)); 970 } 971 972 static const char * 973 pf_qid2qname(uint16_t qid) 974 { 975 return (tag2tagname(&V_pf_qids, qid)); 976 } 977 978 static void 979 pf_qid_unref(uint16_t qid) 980 { 981 tag_unref(&V_pf_qids, qid); 982 } 983 984 static int 985 pf_begin_altq(u_int32_t *ticket) 986 { 987 struct pf_altq *altq, *tmp; 988 int error = 0; 989 990 PF_RULES_WASSERT(); 991 992 /* Purge the old altq lists */ 993 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 994 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 995 /* detach and destroy the discipline */ 996 error = altq_remove(altq); 997 } 998 free(altq, M_PFALTQ); 999 } 1000 TAILQ_INIT(V_pf_altq_ifs_inactive); 1001 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1002 pf_qid_unref(altq->qid); 1003 free(altq, M_PFALTQ); 1004 } 1005 TAILQ_INIT(V_pf_altqs_inactive); 1006 if (error) 1007 return (error); 1008 *ticket = ++V_ticket_altqs_inactive; 1009 V_altqs_inactive_open = 1; 1010 return (0); 1011 } 1012 1013 static int 1014 pf_rollback_altq(u_int32_t ticket) 1015 { 1016 struct pf_altq *altq, *tmp; 1017 int error = 0; 1018 1019 PF_RULES_WASSERT(); 1020 1021 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 1022 return (0); 1023 /* Purge the old altq lists */ 1024 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 1025 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 1026 /* detach and destroy the discipline */ 1027 error = altq_remove(altq); 1028 } 1029 free(altq, M_PFALTQ); 1030 } 1031 TAILQ_INIT(V_pf_altq_ifs_inactive); 1032 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1033 pf_qid_unref(altq->qid); 1034 free(altq, M_PFALTQ); 1035 } 1036 TAILQ_INIT(V_pf_altqs_inactive); 1037 V_altqs_inactive_open = 0; 1038 return (error); 1039 } 1040 1041 static int 1042 pf_commit_altq(u_int32_t ticket) 1043 { 1044 struct pf_altqqueue *old_altqs, *old_altq_ifs; 1045 struct pf_altq *altq, *tmp; 1046 int err, error = 0; 1047 1048 PF_RULES_WASSERT(); 1049 1050 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) 1051 return (EBUSY); 1052 1053 /* swap altqs, keep the old. */ 1054 old_altqs = V_pf_altqs_active; 1055 old_altq_ifs = V_pf_altq_ifs_active; 1056 V_pf_altqs_active = V_pf_altqs_inactive; 1057 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive; 1058 V_pf_altqs_inactive = old_altqs; 1059 V_pf_altq_ifs_inactive = old_altq_ifs; 1060 V_ticket_altqs_active = V_ticket_altqs_inactive; 1061 1062 /* Attach new disciplines */ 1063 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1064 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 1065 /* attach the discipline */ 1066 error = altq_pfattach(altq); 1067 if (error == 0 && V_pf_altq_running) 1068 error = pf_enable_altq(altq); 1069 if (error != 0) 1070 return (error); 1071 } 1072 } 1073 1074 /* Purge the old altq lists */ 1075 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) { 1076 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 1077 /* detach and destroy the discipline */ 1078 if (V_pf_altq_running) 1079 error = pf_disable_altq(altq); 1080 err = altq_pfdetach(altq); 1081 if (err != 0 && error == 0) 1082 error = err; 1083 err = altq_remove(altq); 1084 if (err != 0 && error == 0) 1085 error = err; 1086 } 1087 free(altq, M_PFALTQ); 1088 } 1089 TAILQ_INIT(V_pf_altq_ifs_inactive); 1090 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) { 1091 pf_qid_unref(altq->qid); 1092 free(altq, M_PFALTQ); 1093 } 1094 TAILQ_INIT(V_pf_altqs_inactive); 1095 1096 V_altqs_inactive_open = 0; 1097 return (error); 1098 } 1099 1100 static int 1101 pf_enable_altq(struct pf_altq *altq) 1102 { 1103 struct ifnet *ifp; 1104 struct tb_profile tb; 1105 int error = 0; 1106 1107 if ((ifp = ifunit(altq->ifname)) == NULL) 1108 return (EINVAL); 1109 1110 if (ifp->if_snd.altq_type != ALTQT_NONE) 1111 error = altq_enable(&ifp->if_snd); 1112 1113 /* set tokenbucket regulator */ 1114 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { 1115 tb.rate = altq->ifbandwidth; 1116 tb.depth = altq->tbrsize; 1117 error = tbr_set(&ifp->if_snd, &tb); 1118 } 1119 1120 return (error); 1121 } 1122 1123 static int 1124 pf_disable_altq(struct pf_altq *altq) 1125 { 1126 struct ifnet *ifp; 1127 struct tb_profile tb; 1128 int error; 1129 1130 if ((ifp = ifunit(altq->ifname)) == NULL) 1131 return (EINVAL); 1132 1133 /* 1134 * when the discipline is no longer referenced, it was overridden 1135 * by a new one. if so, just return. 1136 */ 1137 if (altq->altq_disc != ifp->if_snd.altq_disc) 1138 return (0); 1139 1140 error = altq_disable(&ifp->if_snd); 1141 1142 if (error == 0) { 1143 /* clear tokenbucket regulator */ 1144 tb.rate = 0; 1145 error = tbr_set(&ifp->if_snd, &tb); 1146 } 1147 1148 return (error); 1149 } 1150 1151 static int 1152 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket, 1153 struct pf_altq *altq) 1154 { 1155 struct ifnet *ifp1; 1156 int error = 0; 1157 1158 /* Deactivate the interface in question */ 1159 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; 1160 if ((ifp1 = ifunit(altq->ifname)) == NULL || 1161 (remove && ifp1 == ifp)) { 1162 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 1163 } else { 1164 error = altq_add(ifp1, altq); 1165 1166 if (ticket != V_ticket_altqs_inactive) 1167 error = EBUSY; 1168 1169 if (error) 1170 free(altq, M_PFALTQ); 1171 } 1172 1173 return (error); 1174 } 1175 1176 void 1177 pf_altq_ifnet_event(struct ifnet *ifp, int remove) 1178 { 1179 struct pf_altq *a1, *a2, *a3; 1180 u_int32_t ticket; 1181 int error = 0; 1182 1183 /* 1184 * No need to re-evaluate the configuration for events on interfaces 1185 * that do not support ALTQ, as it's not possible for such 1186 * interfaces to be part of the configuration. 1187 */ 1188 if (!ALTQ_IS_READY(&ifp->if_snd)) 1189 return; 1190 1191 /* Interrupt userland queue modifications */ 1192 if (V_altqs_inactive_open) 1193 pf_rollback_altq(V_ticket_altqs_inactive); 1194 1195 /* Start new altq ruleset */ 1196 if (pf_begin_altq(&ticket)) 1197 return; 1198 1199 /* Copy the current active set */ 1200 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) { 1201 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1202 if (a2 == NULL) { 1203 error = ENOMEM; 1204 break; 1205 } 1206 bcopy(a1, a2, sizeof(struct pf_altq)); 1207 1208 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1209 if (error) 1210 break; 1211 1212 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries); 1213 } 1214 if (error) 1215 goto out; 1216 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { 1217 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); 1218 if (a2 == NULL) { 1219 error = ENOMEM; 1220 break; 1221 } 1222 bcopy(a1, a2, sizeof(struct pf_altq)); 1223 1224 if ((a2->qid = pf_qname2qid(a2->qname, true)) == 0) { 1225 error = EBUSY; 1226 free(a2, M_PFALTQ); 1227 break; 1228 } 1229 a2->altq_disc = NULL; 1230 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) { 1231 if (strncmp(a3->ifname, a2->ifname, 1232 IFNAMSIZ) == 0) { 1233 a2->altq_disc = a3->altq_disc; 1234 break; 1235 } 1236 } 1237 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2); 1238 if (error) 1239 break; 1240 1241 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); 1242 } 1243 1244 out: 1245 if (error != 0) 1246 pf_rollback_altq(ticket); 1247 else 1248 pf_commit_altq(ticket); 1249 } 1250 #endif /* ALTQ */ 1251 1252 static struct pf_krule_global * 1253 pf_rule_tree_alloc(int flags) 1254 { 1255 struct pf_krule_global *tree; 1256 1257 tree = malloc(sizeof(struct pf_krule_global), M_PF, flags); 1258 if (tree == NULL) 1259 return (NULL); 1260 RB_INIT(tree); 1261 return (tree); 1262 } 1263 1264 void 1265 pf_rule_tree_free(struct pf_krule_global *tree) 1266 { 1267 1268 free(tree, M_PF); 1269 } 1270 1271 static int 1272 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) 1273 { 1274 struct pf_krule_global *tree; 1275 struct pf_kruleset *rs; 1276 struct pf_krule *rule; 1277 1278 PF_RULES_WASSERT(); 1279 1280 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1281 return (EINVAL); 1282 tree = pf_rule_tree_alloc(M_NOWAIT); 1283 if (tree == NULL) 1284 return (ENOMEM); 1285 rs = pf_find_or_create_kruleset(anchor); 1286 if (rs == NULL) { 1287 pf_rule_tree_free(tree); 1288 return (EINVAL); 1289 } 1290 pf_rule_tree_free(rs->rules[rs_num].inactive.tree); 1291 rs->rules[rs_num].inactive.tree = tree; 1292 1293 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1294 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1295 rs->rules[rs_num].inactive.rcount--; 1296 } 1297 *ticket = ++rs->rules[rs_num].inactive.ticket; 1298 rs->rules[rs_num].inactive.open = 1; 1299 return (0); 1300 } 1301 1302 static int 1303 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) 1304 { 1305 struct pf_kruleset *rs; 1306 struct pf_krule *rule; 1307 1308 PF_RULES_WASSERT(); 1309 1310 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1311 return (EINVAL); 1312 rs = pf_find_kruleset(anchor); 1313 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1314 rs->rules[rs_num].inactive.ticket != ticket) 1315 return (0); 1316 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { 1317 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); 1318 rs->rules[rs_num].inactive.rcount--; 1319 } 1320 rs->rules[rs_num].inactive.open = 0; 1321 return (0); 1322 } 1323 1324 #define PF_MD5_UPD(st, elm) \ 1325 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) 1326 1327 #define PF_MD5_UPD_STR(st, elm) \ 1328 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) 1329 1330 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ 1331 (stor) = htonl((st)->elm); \ 1332 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ 1333 } while (0) 1334 1335 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ 1336 (stor) = htons((st)->elm); \ 1337 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ 1338 } while (0) 1339 1340 static void 1341 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) 1342 { 1343 PF_MD5_UPD(pfr, addr.type); 1344 switch (pfr->addr.type) { 1345 case PF_ADDR_DYNIFTL: 1346 PF_MD5_UPD(pfr, addr.v.ifname); 1347 PF_MD5_UPD(pfr, addr.iflags); 1348 break; 1349 case PF_ADDR_TABLE: 1350 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX, 1351 strlen(PF_OPTIMIZER_TABLE_PFX))) 1352 PF_MD5_UPD(pfr, addr.v.tblname); 1353 break; 1354 case PF_ADDR_ADDRMASK: 1355 /* XXX ignore af? */ 1356 PF_MD5_UPD(pfr, addr.v.a.addr.addr32); 1357 PF_MD5_UPD(pfr, addr.v.a.mask.addr32); 1358 break; 1359 } 1360 1361 PF_MD5_UPD(pfr, port[0]); 1362 PF_MD5_UPD(pfr, port[1]); 1363 PF_MD5_UPD(pfr, neg); 1364 PF_MD5_UPD(pfr, port_op); 1365 } 1366 1367 static void 1368 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule) 1369 { 1370 u_int16_t x; 1371 u_int32_t y; 1372 1373 pf_hash_rule_addr(ctx, &rule->src); 1374 pf_hash_rule_addr(ctx, &rule->dst); 1375 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) 1376 PF_MD5_UPD_STR(rule, label[i]); 1377 PF_MD5_UPD_STR(rule, ifname); 1378 PF_MD5_UPD_STR(rule, rcv_ifname); 1379 PF_MD5_UPD_STR(rule, match_tagname); 1380 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ 1381 PF_MD5_UPD_HTONL(rule, os_fingerprint, y); 1382 PF_MD5_UPD_HTONL(rule, prob, y); 1383 PF_MD5_UPD_HTONL(rule, uid.uid[0], y); 1384 PF_MD5_UPD_HTONL(rule, uid.uid[1], y); 1385 PF_MD5_UPD(rule, uid.op); 1386 PF_MD5_UPD_HTONL(rule, gid.gid[0], y); 1387 PF_MD5_UPD_HTONL(rule, gid.gid[1], y); 1388 PF_MD5_UPD(rule, gid.op); 1389 PF_MD5_UPD_HTONL(rule, rule_flag, y); 1390 PF_MD5_UPD(rule, action); 1391 PF_MD5_UPD(rule, direction); 1392 PF_MD5_UPD(rule, af); 1393 PF_MD5_UPD(rule, quick); 1394 PF_MD5_UPD(rule, ifnot); 1395 PF_MD5_UPD(rule, rcvifnot); 1396 PF_MD5_UPD(rule, match_tag_not); 1397 PF_MD5_UPD(rule, natpass); 1398 PF_MD5_UPD(rule, keep_state); 1399 PF_MD5_UPD(rule, proto); 1400 PF_MD5_UPD(rule, type); 1401 PF_MD5_UPD(rule, code); 1402 PF_MD5_UPD(rule, flags); 1403 PF_MD5_UPD(rule, flagset); 1404 PF_MD5_UPD(rule, allow_opts); 1405 PF_MD5_UPD(rule, rt); 1406 PF_MD5_UPD(rule, tos); 1407 PF_MD5_UPD(rule, scrub_flags); 1408 PF_MD5_UPD(rule, min_ttl); 1409 PF_MD5_UPD(rule, set_tos); 1410 if (rule->anchor != NULL) 1411 PF_MD5_UPD_STR(rule, anchor->path); 1412 } 1413 1414 static void 1415 pf_hash_rule(struct pf_krule *rule) 1416 { 1417 MD5_CTX ctx; 1418 1419 MD5Init(&ctx); 1420 pf_hash_rule_rolling(&ctx, rule); 1421 MD5Final(rule->md5sum, &ctx); 1422 } 1423 1424 static int 1425 pf_krule_compare(struct pf_krule *a, struct pf_krule *b) 1426 { 1427 1428 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH)); 1429 } 1430 1431 static int 1432 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) 1433 { 1434 struct pf_kruleset *rs; 1435 struct pf_krule *rule, *old_rule; 1436 struct pf_krulequeue *old_rules; 1437 struct pf_krule_global *old_tree; 1438 int error; 1439 u_int32_t old_rcount; 1440 1441 PF_RULES_WASSERT(); 1442 1443 if (rs_num < 0 || rs_num >= PF_RULESET_MAX) 1444 return (EINVAL); 1445 rs = pf_find_kruleset(anchor); 1446 if (rs == NULL || !rs->rules[rs_num].inactive.open || 1447 ticket != rs->rules[rs_num].inactive.ticket) 1448 return (EBUSY); 1449 1450 /* Calculate checksum for the main ruleset */ 1451 if (rs == &pf_main_ruleset) { 1452 error = pf_setup_pfsync_matching(rs); 1453 if (error != 0) 1454 return (error); 1455 } 1456 1457 /* Swap rules, keep the old. */ 1458 old_rules = rs->rules[rs_num].active.ptr; 1459 old_rcount = rs->rules[rs_num].active.rcount; 1460 old_tree = rs->rules[rs_num].active.tree; 1461 1462 rs->rules[rs_num].active.ptr = 1463 rs->rules[rs_num].inactive.ptr; 1464 rs->rules[rs_num].active.tree = 1465 rs->rules[rs_num].inactive.tree; 1466 rs->rules[rs_num].active.rcount = 1467 rs->rules[rs_num].inactive.rcount; 1468 1469 /* Attempt to preserve counter information. */ 1470 if (V_pf_status.keep_counters && old_tree != NULL) { 1471 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr, 1472 entries) { 1473 old_rule = RB_FIND(pf_krule_global, old_tree, rule); 1474 if (old_rule == NULL) { 1475 continue; 1476 } 1477 pf_counter_u64_critical_enter(); 1478 pf_counter_u64_rollup_protected(&rule->evaluations, 1479 pf_counter_u64_fetch(&old_rule->evaluations)); 1480 pf_counter_u64_rollup_protected(&rule->packets[0], 1481 pf_counter_u64_fetch(&old_rule->packets[0])); 1482 pf_counter_u64_rollup_protected(&rule->packets[1], 1483 pf_counter_u64_fetch(&old_rule->packets[1])); 1484 pf_counter_u64_rollup_protected(&rule->bytes[0], 1485 pf_counter_u64_fetch(&old_rule->bytes[0])); 1486 pf_counter_u64_rollup_protected(&rule->bytes[1], 1487 pf_counter_u64_fetch(&old_rule->bytes[1])); 1488 pf_counter_u64_critical_exit(); 1489 } 1490 } 1491 1492 rs->rules[rs_num].inactive.ptr = old_rules; 1493 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ 1494 rs->rules[rs_num].inactive.rcount = old_rcount; 1495 1496 rs->rules[rs_num].active.ticket = 1497 rs->rules[rs_num].inactive.ticket; 1498 pf_calc_skip_steps(rs->rules[rs_num].active.ptr); 1499 1500 /* Purge the old rule list. */ 1501 PF_UNLNKDRULES_LOCK(); 1502 while ((rule = TAILQ_FIRST(old_rules)) != NULL) 1503 pf_unlink_rule_locked(old_rules, rule); 1504 PF_UNLNKDRULES_UNLOCK(); 1505 rs->rules[rs_num].inactive.rcount = 0; 1506 rs->rules[rs_num].inactive.open = 0; 1507 pf_remove_if_empty_kruleset(rs); 1508 pf_rule_tree_free(old_tree); 1509 1510 return (0); 1511 } 1512 1513 static int 1514 pf_setup_pfsync_matching(struct pf_kruleset *rs) 1515 { 1516 MD5_CTX ctx; 1517 struct pf_krule *rule; 1518 int rs_cnt; 1519 u_int8_t digest[PF_MD5_DIGEST_LENGTH]; 1520 1521 MD5Init(&ctx); 1522 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { 1523 /* XXX PF_RULESET_SCRUB as well? */ 1524 if (rs_cnt == PF_RULESET_SCRUB) 1525 continue; 1526 1527 if (rs->rules[rs_cnt].inactive.rcount) { 1528 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, 1529 entries) { 1530 pf_hash_rule_rolling(&ctx, rule); 1531 } 1532 } 1533 } 1534 1535 MD5Final(digest, &ctx); 1536 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); 1537 return (0); 1538 } 1539 1540 static int 1541 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr) 1542 { 1543 int error = 0; 1544 1545 switch (addr->type) { 1546 case PF_ADDR_TABLE: 1547 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname); 1548 if (addr->p.tbl == NULL) 1549 error = ENOMEM; 1550 break; 1551 default: 1552 error = EINVAL; 1553 } 1554 1555 return (error); 1556 } 1557 1558 static int 1559 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr, 1560 sa_family_t af) 1561 { 1562 int error = 0; 1563 1564 switch (addr->type) { 1565 case PF_ADDR_TABLE: 1566 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); 1567 if (addr->p.tbl == NULL) 1568 error = ENOMEM; 1569 break; 1570 case PF_ADDR_DYNIFTL: 1571 error = pfi_dynaddr_setup(addr, af); 1572 break; 1573 } 1574 1575 return (error); 1576 } 1577 1578 void 1579 pf_addr_copyout(struct pf_addr_wrap *addr) 1580 { 1581 1582 switch (addr->type) { 1583 case PF_ADDR_DYNIFTL: 1584 pfi_dynaddr_copyout(addr); 1585 break; 1586 case PF_ADDR_TABLE: 1587 pf_tbladdr_copyout(addr); 1588 break; 1589 } 1590 } 1591 1592 static void 1593 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out) 1594 { 1595 int secs = time_uptime; 1596 1597 bzero(out, sizeof(struct pf_src_node)); 1598 1599 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr)); 1600 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr)); 1601 1602 if (in->rule != NULL) 1603 out->rule.nr = in->rule->nr; 1604 1605 for (int i = 0; i < 2; i++) { 1606 out->bytes[i] = counter_u64_fetch(in->bytes[i]); 1607 out->packets[i] = counter_u64_fetch(in->packets[i]); 1608 } 1609 1610 out->states = in->states; 1611 out->conn = in->conn; 1612 out->af = in->af; 1613 out->ruletype = in->ruletype; 1614 1615 out->creation = secs - in->creation; 1616 if (out->expire > secs) 1617 out->expire -= secs; 1618 else 1619 out->expire = 0; 1620 1621 /* Adjust the connection rate estimate. */ 1622 out->conn_rate.limit = in->conn_rate.limit; 1623 out->conn_rate.seconds = in->conn_rate.seconds; 1624 /* If there's no limit there's no counter_rate. */ 1625 if (in->conn_rate.cr != NULL) 1626 out->conn_rate.count = counter_rate_get(in->conn_rate.cr); 1627 } 1628 1629 #ifdef ALTQ 1630 /* 1631 * Handle export of struct pf_kaltq to user binaries that may be using any 1632 * version of struct pf_altq. 1633 */ 1634 static int 1635 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size) 1636 { 1637 u_int32_t version; 1638 1639 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1640 version = 0; 1641 else 1642 version = pa->version; 1643 1644 if (version > PFIOC_ALTQ_VERSION) 1645 return (EINVAL); 1646 1647 #define ASSIGN(x) exported_q->x = q->x 1648 #define COPY(x) \ 1649 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x))) 1650 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX) 1651 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX) 1652 1653 switch (version) { 1654 case 0: { 1655 struct pf_altq_v0 *exported_q = 1656 &((struct pfioc_altq_v0 *)pa)->altq; 1657 1658 COPY(ifname); 1659 1660 ASSIGN(scheduler); 1661 ASSIGN(tbrsize); 1662 exported_q->tbrsize = SATU16(q->tbrsize); 1663 exported_q->ifbandwidth = SATU32(q->ifbandwidth); 1664 1665 COPY(qname); 1666 COPY(parent); 1667 ASSIGN(parent_qid); 1668 exported_q->bandwidth = SATU32(q->bandwidth); 1669 ASSIGN(priority); 1670 ASSIGN(local_flags); 1671 1672 ASSIGN(qlimit); 1673 ASSIGN(flags); 1674 1675 if (q->scheduler == ALTQT_HFSC) { 1676 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x 1677 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \ 1678 SATU32(q->pq_u.hfsc_opts.x) 1679 1680 ASSIGN_OPT_SATU32(rtsc_m1); 1681 ASSIGN_OPT(rtsc_d); 1682 ASSIGN_OPT_SATU32(rtsc_m2); 1683 1684 ASSIGN_OPT_SATU32(lssc_m1); 1685 ASSIGN_OPT(lssc_d); 1686 ASSIGN_OPT_SATU32(lssc_m2); 1687 1688 ASSIGN_OPT_SATU32(ulsc_m1); 1689 ASSIGN_OPT(ulsc_d); 1690 ASSIGN_OPT_SATU32(ulsc_m2); 1691 1692 ASSIGN_OPT(flags); 1693 1694 #undef ASSIGN_OPT 1695 #undef ASSIGN_OPT_SATU32 1696 } else 1697 COPY(pq_u); 1698 1699 ASSIGN(qid); 1700 break; 1701 } 1702 case 1: { 1703 struct pf_altq_v1 *exported_q = 1704 &((struct pfioc_altq_v1 *)pa)->altq; 1705 1706 COPY(ifname); 1707 1708 ASSIGN(scheduler); 1709 ASSIGN(tbrsize); 1710 ASSIGN(ifbandwidth); 1711 1712 COPY(qname); 1713 COPY(parent); 1714 ASSIGN(parent_qid); 1715 ASSIGN(bandwidth); 1716 ASSIGN(priority); 1717 ASSIGN(local_flags); 1718 1719 ASSIGN(qlimit); 1720 ASSIGN(flags); 1721 COPY(pq_u); 1722 1723 ASSIGN(qid); 1724 break; 1725 } 1726 default: 1727 panic("%s: unhandled struct pfioc_altq version", __func__); 1728 break; 1729 } 1730 1731 #undef ASSIGN 1732 #undef COPY 1733 #undef SATU16 1734 #undef SATU32 1735 1736 return (0); 1737 } 1738 1739 /* 1740 * Handle import to struct pf_kaltq of struct pf_altq from user binaries 1741 * that may be using any version of it. 1742 */ 1743 static int 1744 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size) 1745 { 1746 u_int32_t version; 1747 1748 if (ioc_size == sizeof(struct pfioc_altq_v0)) 1749 version = 0; 1750 else 1751 version = pa->version; 1752 1753 if (version > PFIOC_ALTQ_VERSION) 1754 return (EINVAL); 1755 1756 #define ASSIGN(x) q->x = imported_q->x 1757 #define COPY(x) \ 1758 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x))) 1759 1760 switch (version) { 1761 case 0: { 1762 struct pf_altq_v0 *imported_q = 1763 &((struct pfioc_altq_v0 *)pa)->altq; 1764 1765 COPY(ifname); 1766 1767 ASSIGN(scheduler); 1768 ASSIGN(tbrsize); /* 16-bit -> 32-bit */ 1769 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */ 1770 1771 COPY(qname); 1772 COPY(parent); 1773 ASSIGN(parent_qid); 1774 ASSIGN(bandwidth); /* 32-bit -> 64-bit */ 1775 ASSIGN(priority); 1776 ASSIGN(local_flags); 1777 1778 ASSIGN(qlimit); 1779 ASSIGN(flags); 1780 1781 if (imported_q->scheduler == ALTQT_HFSC) { 1782 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x 1783 1784 /* 1785 * The m1 and m2 parameters are being copied from 1786 * 32-bit to 64-bit. 1787 */ 1788 ASSIGN_OPT(rtsc_m1); 1789 ASSIGN_OPT(rtsc_d); 1790 ASSIGN_OPT(rtsc_m2); 1791 1792 ASSIGN_OPT(lssc_m1); 1793 ASSIGN_OPT(lssc_d); 1794 ASSIGN_OPT(lssc_m2); 1795 1796 ASSIGN_OPT(ulsc_m1); 1797 ASSIGN_OPT(ulsc_d); 1798 ASSIGN_OPT(ulsc_m2); 1799 1800 ASSIGN_OPT(flags); 1801 1802 #undef ASSIGN_OPT 1803 } else 1804 COPY(pq_u); 1805 1806 ASSIGN(qid); 1807 break; 1808 } 1809 case 1: { 1810 struct pf_altq_v1 *imported_q = 1811 &((struct pfioc_altq_v1 *)pa)->altq; 1812 1813 COPY(ifname); 1814 1815 ASSIGN(scheduler); 1816 ASSIGN(tbrsize); 1817 ASSIGN(ifbandwidth); 1818 1819 COPY(qname); 1820 COPY(parent); 1821 ASSIGN(parent_qid); 1822 ASSIGN(bandwidth); 1823 ASSIGN(priority); 1824 ASSIGN(local_flags); 1825 1826 ASSIGN(qlimit); 1827 ASSIGN(flags); 1828 COPY(pq_u); 1829 1830 ASSIGN(qid); 1831 break; 1832 } 1833 default: 1834 panic("%s: unhandled struct pfioc_altq version", __func__); 1835 break; 1836 } 1837 1838 #undef ASSIGN 1839 #undef COPY 1840 1841 return (0); 1842 } 1843 1844 static struct pf_altq * 1845 pf_altq_get_nth_active(u_int32_t n) 1846 { 1847 struct pf_altq *altq; 1848 u_int32_t nr; 1849 1850 nr = 0; 1851 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 1852 if (nr == n) 1853 return (altq); 1854 nr++; 1855 } 1856 1857 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { 1858 if (nr == n) 1859 return (altq); 1860 nr++; 1861 } 1862 1863 return (NULL); 1864 } 1865 #endif /* ALTQ */ 1866 1867 struct pf_krule * 1868 pf_krule_alloc(void) 1869 { 1870 struct pf_krule *rule; 1871 1872 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO); 1873 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF); 1874 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF); 1875 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF); 1876 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 1877 M_WAITOK | M_ZERO); 1878 return (rule); 1879 } 1880 1881 void 1882 pf_krule_free(struct pf_krule *rule) 1883 { 1884 #ifdef PF_WANT_32_TO_64_COUNTER 1885 bool wowned; 1886 #endif 1887 1888 if (rule == NULL) 1889 return; 1890 1891 #ifdef PF_WANT_32_TO_64_COUNTER 1892 if (rule->allrulelinked) { 1893 wowned = PF_RULES_WOWNED(); 1894 if (!wowned) 1895 PF_RULES_WLOCK(); 1896 LIST_REMOVE(rule, allrulelist); 1897 V_pf_allrulecount--; 1898 if (!wowned) 1899 PF_RULES_WUNLOCK(); 1900 } 1901 #endif 1902 1903 pf_counter_u64_deinit(&rule->evaluations); 1904 for (int i = 0; i < 2; i++) { 1905 pf_counter_u64_deinit(&rule->packets[i]); 1906 pf_counter_u64_deinit(&rule->bytes[i]); 1907 } 1908 counter_u64_free(rule->states_cur); 1909 counter_u64_free(rule->states_tot); 1910 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 1911 counter_u64_free(rule->src_nodes[sn_type]); 1912 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp); 1913 1914 mtx_destroy(&rule->nat.mtx); 1915 mtx_destroy(&rule->rdr.mtx); 1916 mtx_destroy(&rule->route.mtx); 1917 free(rule, M_PFRULE); 1918 } 1919 1920 void 1921 pf_krule_clear_counters(struct pf_krule *rule) 1922 { 1923 pf_counter_u64_zero(&rule->evaluations); 1924 for (int i = 0; i < 2; i++) { 1925 pf_counter_u64_zero(&rule->packets[i]); 1926 pf_counter_u64_zero(&rule->bytes[i]); 1927 } 1928 counter_u64_zero(rule->states_tot); 1929 } 1930 1931 static void 1932 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool, 1933 struct pf_pooladdr *pool) 1934 { 1935 1936 bzero(pool, sizeof(*pool)); 1937 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr)); 1938 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname)); 1939 } 1940 1941 static int 1942 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool, 1943 struct pf_kpooladdr *kpool) 1944 { 1945 int ret; 1946 1947 bzero(kpool, sizeof(*kpool)); 1948 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr)); 1949 ret = pf_user_strcpy(kpool->ifname, pool->ifname, 1950 sizeof(kpool->ifname)); 1951 return (ret); 1952 } 1953 1954 static void 1955 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool) 1956 { 1957 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), ""); 1958 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), ""); 1959 1960 bcopy(&pool->key, &kpool->key, sizeof(kpool->key)); 1961 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter)); 1962 1963 kpool->tblidx = pool->tblidx; 1964 kpool->proxy_port[0] = pool->proxy_port[0]; 1965 kpool->proxy_port[1] = pool->proxy_port[1]; 1966 kpool->opts = pool->opts; 1967 } 1968 1969 static int 1970 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule) 1971 { 1972 int ret; 1973 1974 #ifndef INET 1975 if (rule->af == AF_INET) { 1976 return (EAFNOSUPPORT); 1977 } 1978 #endif /* INET */ 1979 #ifndef INET6 1980 if (rule->af == AF_INET6) { 1981 return (EAFNOSUPPORT); 1982 } 1983 #endif /* INET6 */ 1984 1985 ret = pf_check_rule_addr(&rule->src); 1986 if (ret != 0) 1987 return (ret); 1988 ret = pf_check_rule_addr(&rule->dst); 1989 if (ret != 0) 1990 return (ret); 1991 1992 bcopy(&rule->src, &krule->src, sizeof(rule->src)); 1993 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst)); 1994 1995 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label)); 1996 if (ret != 0) 1997 return (ret); 1998 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname)); 1999 if (ret != 0) 2000 return (ret); 2001 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname)); 2002 if (ret != 0) 2003 return (ret); 2004 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname)); 2005 if (ret != 0) 2006 return (ret); 2007 ret = pf_user_strcpy(krule->tagname, rule->tagname, 2008 sizeof(rule->tagname)); 2009 if (ret != 0) 2010 return (ret); 2011 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname, 2012 sizeof(rule->match_tagname)); 2013 if (ret != 0) 2014 return (ret); 2015 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname, 2016 sizeof(rule->overload_tblname)); 2017 if (ret != 0) 2018 return (ret); 2019 2020 pf_pool_to_kpool(&rule->rpool, &krule->rdr); 2021 2022 /* Don't allow userspace to set evaluations, packets or bytes. */ 2023 /* kif, anchor, overload_tbl are not copied over. */ 2024 2025 krule->os_fingerprint = rule->os_fingerprint; 2026 2027 krule->rtableid = rule->rtableid; 2028 /* pf_rule->timeout is smaller than pf_krule->timeout */ 2029 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout)); 2030 krule->max_states = rule->max_states; 2031 krule->max_src_nodes = rule->max_src_nodes; 2032 krule->max_src_states = rule->max_src_states; 2033 krule->max_src_conn = rule->max_src_conn; 2034 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit; 2035 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds; 2036 krule->qid = rule->qid; 2037 krule->pqid = rule->pqid; 2038 krule->nr = rule->nr; 2039 krule->prob = rule->prob; 2040 krule->cuid = rule->cuid; 2041 krule->cpid = rule->cpid; 2042 2043 krule->return_icmp = rule->return_icmp; 2044 krule->return_icmp6 = rule->return_icmp6; 2045 krule->max_mss = rule->max_mss; 2046 krule->tag = rule->tag; 2047 krule->match_tag = rule->match_tag; 2048 krule->scrub_flags = rule->scrub_flags; 2049 2050 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid)); 2051 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid)); 2052 2053 krule->rule_flag = rule->rule_flag; 2054 krule->action = rule->action; 2055 krule->direction = rule->direction; 2056 krule->log = rule->log; 2057 krule->logif = rule->logif; 2058 krule->quick = rule->quick; 2059 krule->ifnot = rule->ifnot; 2060 krule->match_tag_not = rule->match_tag_not; 2061 krule->natpass = rule->natpass; 2062 2063 krule->keep_state = rule->keep_state; 2064 krule->af = rule->af; 2065 krule->proto = rule->proto; 2066 krule->type = rule->type; 2067 krule->code = rule->code; 2068 krule->flags = rule->flags; 2069 krule->flagset = rule->flagset; 2070 krule->min_ttl = rule->min_ttl; 2071 krule->allow_opts = rule->allow_opts; 2072 krule->rt = rule->rt; 2073 krule->return_ttl = rule->return_ttl; 2074 krule->tos = rule->tos; 2075 krule->set_tos = rule->set_tos; 2076 2077 krule->flush = rule->flush; 2078 krule->prio = rule->prio; 2079 krule->set_prio[0] = rule->set_prio[0]; 2080 krule->set_prio[1] = rule->set_prio[1]; 2081 2082 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert)); 2083 2084 return (0); 2085 } 2086 2087 int 2088 pf_ioctl_getrules(struct pfioc_rule *pr) 2089 { 2090 struct pf_kruleset *ruleset; 2091 struct pf_krule *tail; 2092 int rs_num; 2093 2094 PF_RULES_WLOCK(); 2095 ruleset = pf_find_kruleset(pr->anchor); 2096 if (ruleset == NULL) { 2097 PF_RULES_WUNLOCK(); 2098 return (EINVAL); 2099 } 2100 rs_num = pf_get_ruleset_number(pr->rule.action); 2101 if (rs_num >= PF_RULESET_MAX) { 2102 PF_RULES_WUNLOCK(); 2103 return (EINVAL); 2104 } 2105 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, 2106 pf_krulequeue); 2107 if (tail) 2108 pr->nr = tail->nr + 1; 2109 else 2110 pr->nr = 0; 2111 pr->ticket = ruleset->rules[rs_num].active.ticket; 2112 PF_RULES_WUNLOCK(); 2113 2114 return (0); 2115 } 2116 2117 static int 2118 pf_rule_checkaf(struct pf_krule *r) 2119 { 2120 switch (r->af) { 2121 case 0: 2122 if (r->rule_flag & PFRULE_AFTO) 2123 return (EPFNOSUPPORT); 2124 break; 2125 case AF_INET: 2126 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6) 2127 return (EPFNOSUPPORT); 2128 break; 2129 #ifdef INET6 2130 case AF_INET6: 2131 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET) 2132 return (EPFNOSUPPORT); 2133 break; 2134 #endif /* INET6 */ 2135 default: 2136 return (EPFNOSUPPORT); 2137 } 2138 2139 if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0) 2140 return (EPFNOSUPPORT); 2141 2142 return (0); 2143 } 2144 2145 static int 2146 pf_validate_range(uint8_t op, uint16_t port[2]) 2147 { 2148 uint16_t a = ntohs(port[0]); 2149 uint16_t b = ntohs(port[1]); 2150 2151 if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */ 2152 (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */ 2153 (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */ 2154 return 1; 2155 return 0; 2156 } 2157 2158 int 2159 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, 2160 uint32_t pool_ticket, const char *anchor, const char *anchor_call, 2161 uid_t uid, pid_t pid) 2162 { 2163 struct pf_kruleset *ruleset; 2164 struct pf_krule *tail; 2165 struct pf_kpooladdr *pa; 2166 struct pfi_kkif *kif = NULL, *rcv_kif = NULL; 2167 int rs_num; 2168 int error = 0; 2169 2170 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 2171 #define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x) 2172 2173 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) 2174 ERROUT_UNLOCKED(EINVAL); 2175 2176 if ((error = pf_rule_checkaf(rule))) 2177 ERROUT_UNLOCKED(error); 2178 if (pf_validate_range(rule->src.port_op, rule->src.port)) 2179 ERROUT_UNLOCKED(EINVAL); 2180 if (pf_validate_range(rule->dst.port_op, rule->dst.port)) 2181 ERROUT_UNLOCKED(EINVAL); 2182 2183 if (rule->ifname[0]) 2184 kif = pf_kkif_create(M_WAITOK); 2185 if (rule->rcv_ifname[0]) 2186 rcv_kif = pf_kkif_create(M_WAITOK); 2187 pf_counter_u64_init(&rule->evaluations, M_WAITOK); 2188 for (int i = 0; i < 2; i++) { 2189 pf_counter_u64_init(&rule->packets[i], M_WAITOK); 2190 pf_counter_u64_init(&rule->bytes[i], M_WAITOK); 2191 } 2192 rule->states_cur = counter_u64_alloc(M_WAITOK); 2193 rule->states_tot = counter_u64_alloc(M_WAITOK); 2194 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 2195 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 2196 rule->cuid = uid; 2197 rule->cpid = pid; 2198 TAILQ_INIT(&rule->rdr.list); 2199 TAILQ_INIT(&rule->nat.list); 2200 TAILQ_INIT(&rule->route.list); 2201 2202 PF_CONFIG_LOCK(); 2203 PF_RULES_WLOCK(); 2204 #ifdef PF_WANT_32_TO_64_COUNTER 2205 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist); 2206 MPASS(!rule->allrulelinked); 2207 rule->allrulelinked = true; 2208 V_pf_allrulecount++; 2209 #endif 2210 ruleset = pf_find_kruleset(anchor); 2211 if (ruleset == NULL) 2212 ERROUT(EINVAL); 2213 rs_num = pf_get_ruleset_number(rule->action); 2214 if (rs_num >= PF_RULESET_MAX) 2215 ERROUT(EINVAL); 2216 if (ticket != ruleset->rules[rs_num].inactive.ticket) { 2217 DPFPRINTF(PF_DEBUG_MISC, 2218 "ticket: %d != [%d]%d", ticket, rs_num, 2219 ruleset->rules[rs_num].inactive.ticket); 2220 ERROUT(EBUSY); 2221 } 2222 if (pool_ticket != V_ticket_pabuf) { 2223 DPFPRINTF(PF_DEBUG_MISC, 2224 "pool_ticket: %d != %d", pool_ticket, 2225 V_ticket_pabuf); 2226 ERROUT(EBUSY); 2227 } 2228 /* 2229 * XXXMJG hack: there is no mechanism to ensure they started the 2230 * transaction. Ticket checked above may happen to match by accident, 2231 * even if nobody called DIOCXBEGIN, let alone this process. 2232 * Partially work around it by checking if the RB tree got allocated, 2233 * see pf_begin_rules. 2234 */ 2235 if (ruleset->rules[rs_num].inactive.tree == NULL) { 2236 ERROUT(EINVAL); 2237 } 2238 2239 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, 2240 pf_krulequeue); 2241 if (tail) 2242 rule->nr = tail->nr + 1; 2243 else 2244 rule->nr = 0; 2245 if (rule->ifname[0]) { 2246 rule->kif = pfi_kkif_attach(kif, rule->ifname); 2247 kif = NULL; 2248 pfi_kkif_ref(rule->kif); 2249 } else 2250 rule->kif = NULL; 2251 2252 if (rule->rcv_ifname[0]) { 2253 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname); 2254 rcv_kif = NULL; 2255 pfi_kkif_ref(rule->rcv_kif); 2256 } else 2257 rule->rcv_kif = NULL; 2258 2259 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) 2260 ERROUT(EBUSY); 2261 #ifdef ALTQ 2262 /* set queue IDs */ 2263 if (rule->qname[0] != 0) { 2264 if ((rule->qid = pf_qname2qid(rule->qname, true)) == 0) 2265 ERROUT(EBUSY); 2266 else if (rule->pqname[0] != 0) { 2267 if ((rule->pqid = 2268 pf_qname2qid(rule->pqname, true)) == 0) 2269 ERROUT(EBUSY); 2270 } else 2271 rule->pqid = rule->qid; 2272 } 2273 #endif 2274 if (rule->tagname[0]) 2275 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 2276 ERROUT(EBUSY); 2277 if (rule->match_tagname[0]) 2278 if ((rule->match_tag = 2279 pf_tagname2tag(rule->match_tagname)) == 0) 2280 ERROUT(EBUSY); 2281 if (rule->rt && !rule->direction) 2282 ERROUT(EINVAL); 2283 if (!rule->log) 2284 rule->logif = 0; 2285 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit, 2286 rule->pktrate.seconds)) 2287 ERROUT(ENOMEM); 2288 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) 2289 ERROUT(ENOMEM); 2290 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) 2291 ERROUT(ENOMEM); 2292 if (pf_kanchor_setup(rule, ruleset, anchor_call)) 2293 ERROUT(EINVAL); 2294 if (rule->scrub_flags & PFSTATE_SETPRIO && 2295 (rule->set_prio[0] > PF_PRIO_MAX || 2296 rule->set_prio[1] > PF_PRIO_MAX)) 2297 ERROUT(EINVAL); 2298 for (int i = 0; i < 3; i++) { 2299 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 2300 if (pa->addr.type == PF_ADDR_TABLE) { 2301 pa->addr.p.tbl = pfr_attach_table(ruleset, 2302 pa->addr.v.tblname); 2303 if (pa->addr.p.tbl == NULL) 2304 ERROUT(ENOMEM); 2305 } 2306 } 2307 2308 rule->overload_tbl = NULL; 2309 if (rule->overload_tblname[0]) { 2310 if ((rule->overload_tbl = pfr_attach_table(ruleset, 2311 rule->overload_tblname)) == NULL) 2312 ERROUT(EINVAL); 2313 else 2314 rule->overload_tbl->pfrkt_flags |= 2315 PFR_TFLAG_ACTIVE; 2316 } 2317 2318 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list); 2319 2320 /* 2321 * Old version of pfctl provide route redirection pools in single 2322 * common redirection pool rdr. New versions use rdr only for 2323 * rdr-to rules. 2324 */ 2325 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) { 2326 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list); 2327 } else { 2328 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list); 2329 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list); 2330 } 2331 2332 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) || 2333 (rule->action == PF_BINAT)) && rule->anchor == NULL && 2334 TAILQ_FIRST(&rule->rdr.list) == NULL) { 2335 ERROUT(EINVAL); 2336 } 2337 2338 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) { 2339 ERROUT(EINVAL); 2340 } 2341 2342 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR || 2343 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) { 2344 ERROUT(EINVAL); 2345 } 2346 2347 MPASS(error == 0); 2348 2349 rule->nat.cur = TAILQ_FIRST(&rule->nat.list); 2350 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list); 2351 rule->route.cur = TAILQ_FIRST(&rule->route.list); 2352 rule->route.ipv6_nexthop_af = AF_INET6; 2353 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, 2354 rule, entries); 2355 ruleset->rules[rs_num].inactive.rcount++; 2356 2357 PF_RULES_WUNLOCK(); 2358 pf_hash_rule(rule); 2359 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) { 2360 PF_RULES_WLOCK(); 2361 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries); 2362 ruleset->rules[rs_num].inactive.rcount--; 2363 pf_free_rule(rule); 2364 rule = NULL; 2365 ERROUT(EEXIST); 2366 } 2367 PF_CONFIG_UNLOCK(); 2368 2369 return (0); 2370 2371 #undef ERROUT 2372 #undef ERROUT_UNLOCKED 2373 errout: 2374 PF_RULES_WUNLOCK(); 2375 PF_CONFIG_UNLOCK(); 2376 errout_unlocked: 2377 pf_kkif_free(rcv_kif); 2378 pf_kkif_free(kif); 2379 pf_krule_free(rule); 2380 return (error); 2381 } 2382 2383 static bool 2384 pf_label_match(const struct pf_krule *rule, const char *label) 2385 { 2386 int i = 0; 2387 2388 while (*rule->label[i]) { 2389 if (strcmp(rule->label[i], label) == 0) 2390 return (true); 2391 i++; 2392 } 2393 2394 return (false); 2395 } 2396 2397 static unsigned int 2398 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir) 2399 { 2400 struct pf_kstate *s; 2401 int more = 0; 2402 2403 s = pf_find_state_all(key, dir, &more); 2404 if (s == NULL) 2405 return (0); 2406 2407 if (more) { 2408 PF_STATE_UNLOCK(s); 2409 return (0); 2410 } 2411 2412 pf_remove_state(s); 2413 return (1); 2414 } 2415 2416 static int 2417 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih) 2418 { 2419 struct pf_kstate *s; 2420 struct pf_state_key *sk; 2421 struct pf_addr *srcaddr, *dstaddr; 2422 struct pf_state_key_cmp match_key; 2423 int idx, killed = 0; 2424 unsigned int dir; 2425 u_int16_t srcport, dstport; 2426 struct pfi_kkif *kif; 2427 2428 relock_DIOCKILLSTATES: 2429 PF_HASHROW_LOCK(ih); 2430 LIST_FOREACH(s, &ih->states, entry) { 2431 /* For floating states look at the original kif. */ 2432 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 2433 2434 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE]; 2435 if (s->direction == PF_OUT) { 2436 srcaddr = &sk->addr[1]; 2437 dstaddr = &sk->addr[0]; 2438 srcport = sk->port[1]; 2439 dstport = sk->port[0]; 2440 } else { 2441 srcaddr = &sk->addr[0]; 2442 dstaddr = &sk->addr[1]; 2443 srcport = sk->port[0]; 2444 dstport = sk->port[1]; 2445 } 2446 2447 if (psk->psk_af && sk->af != psk->psk_af) 2448 continue; 2449 2450 if (psk->psk_proto && psk->psk_proto != sk->proto) 2451 continue; 2452 2453 if (! pf_match_addr(psk->psk_src.neg, 2454 &psk->psk_src.addr.v.a.addr, 2455 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af)) 2456 continue; 2457 2458 if (! pf_match_addr(psk->psk_dst.neg, 2459 &psk->psk_dst.addr.v.a.addr, 2460 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af)) 2461 continue; 2462 2463 if (! pf_match_addr(psk->psk_rt_addr.neg, 2464 &psk->psk_rt_addr.addr.v.a.addr, 2465 &psk->psk_rt_addr.addr.v.a.mask, 2466 &s->act.rt_addr, sk->af)) 2467 continue; 2468 2469 if (psk->psk_src.port_op != 0 && 2470 ! pf_match_port(psk->psk_src.port_op, 2471 psk->psk_src.port[0], psk->psk_src.port[1], srcport)) 2472 continue; 2473 2474 if (psk->psk_dst.port_op != 0 && 2475 ! pf_match_port(psk->psk_dst.port_op, 2476 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) 2477 continue; 2478 2479 if (psk->psk_label[0] && 2480 ! pf_label_match(s->rule, psk->psk_label)) 2481 continue; 2482 2483 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname, 2484 kif->pfik_name)) 2485 continue; 2486 2487 if (psk->psk_kill_match) { 2488 /* Create the key to find matching states, with lock 2489 * held. */ 2490 2491 bzero(&match_key, sizeof(match_key)); 2492 2493 if (s->direction == PF_OUT) { 2494 dir = PF_IN; 2495 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK; 2496 } else { 2497 dir = PF_OUT; 2498 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE; 2499 } 2500 2501 match_key.af = s->key[idx]->af; 2502 match_key.proto = s->key[idx]->proto; 2503 pf_addrcpy(&match_key.addr[0], 2504 &s->key[idx]->addr[1], match_key.af); 2505 match_key.port[0] = s->key[idx]->port[1]; 2506 pf_addrcpy(&match_key.addr[1], 2507 &s->key[idx]->addr[0], match_key.af); 2508 match_key.port[1] = s->key[idx]->port[0]; 2509 } 2510 2511 pf_remove_state(s); 2512 killed++; 2513 2514 if (psk->psk_kill_match) 2515 killed += pf_kill_matching_state(&match_key, dir); 2516 2517 goto relock_DIOCKILLSTATES; 2518 } 2519 PF_HASHROW_UNLOCK(ih); 2520 2521 return (killed); 2522 } 2523 2524 void 2525 unhandled_af(int af) 2526 { 2527 panic("unhandled af %d", af); 2528 } 2529 2530 int 2531 pf_start(void) 2532 { 2533 int error = 0; 2534 2535 sx_xlock(&V_pf_ioctl_lock); 2536 if (V_pf_status.running) 2537 error = EEXIST; 2538 else { 2539 hook_pf(); 2540 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 2541 hook_pf_eth(); 2542 V_pf_status.running = 1; 2543 V_pf_status.since = time_uptime; 2544 new_unrhdr64(&V_pf_stateid, time_second); 2545 2546 DPFPRINTF(PF_DEBUG_MISC, "pf: started"); 2547 } 2548 sx_xunlock(&V_pf_ioctl_lock); 2549 2550 return (error); 2551 } 2552 2553 int 2554 pf_stop(void) 2555 { 2556 int error = 0; 2557 2558 sx_xlock(&V_pf_ioctl_lock); 2559 if (!V_pf_status.running) 2560 error = ENOENT; 2561 else { 2562 V_pf_status.running = 0; 2563 dehook_pf(); 2564 dehook_pf_eth(); 2565 V_pf_status.since = time_uptime; 2566 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped"); 2567 } 2568 sx_xunlock(&V_pf_ioctl_lock); 2569 2570 return (error); 2571 } 2572 2573 void 2574 pf_ioctl_clear_status(void) 2575 { 2576 PF_RULES_WLOCK(); 2577 for (int i = 0; i < PFRES_MAX; i++) 2578 counter_u64_zero(V_pf_status.counters[i]); 2579 for (int i = 0; i < FCNT_MAX; i++) 2580 pf_counter_u64_zero(&V_pf_status.fcounters[i]); 2581 for (int i = 0; i < SCNT_MAX; i++) 2582 counter_u64_zero(V_pf_status.scounters[i]); 2583 for (int i = 0; i < NCNT_MAX; i++) 2584 counter_u64_zero(V_pf_status.ncounters[i]); 2585 for (int i = 0; i < KLCNT_MAX; i++) 2586 counter_u64_zero(V_pf_status.lcounters[i]); 2587 V_pf_status.since = time_uptime; 2588 if (*V_pf_status.ifname) 2589 pfi_update_status(V_pf_status.ifname, NULL); 2590 PF_RULES_WUNLOCK(); 2591 } 2592 2593 int 2594 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds) 2595 { 2596 uint32_t old; 2597 2598 if (timeout < 0 || timeout >= PFTM_MAX || 2599 seconds < 0) 2600 return (EINVAL); 2601 2602 PF_RULES_WLOCK(); 2603 old = V_pf_default_rule.timeout[timeout]; 2604 if (timeout == PFTM_INTERVAL && seconds == 0) 2605 seconds = 1; 2606 V_pf_default_rule.timeout[timeout] = seconds; 2607 if (timeout == PFTM_INTERVAL && seconds < old) 2608 wakeup(pf_purge_thread); 2609 2610 if (prev_seconds != NULL) 2611 *prev_seconds = old; 2612 2613 PF_RULES_WUNLOCK(); 2614 2615 return (0); 2616 } 2617 2618 int 2619 pf_ioctl_get_timeout(int timeout, int *seconds) 2620 { 2621 PF_RULES_RLOCK_TRACKER; 2622 2623 if (timeout < 0 || timeout >= PFTM_MAX) 2624 return (EINVAL); 2625 2626 PF_RULES_RLOCK(); 2627 *seconds = V_pf_default_rule.timeout[timeout]; 2628 PF_RULES_RUNLOCK(); 2629 2630 return (0); 2631 } 2632 2633 int 2634 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit) 2635 { 2636 2637 PF_RULES_WLOCK(); 2638 if (index < 0 || index >= PF_LIMIT_MAX || 2639 V_pf_limits[index].zone == NULL) { 2640 PF_RULES_WUNLOCK(); 2641 return (EINVAL); 2642 } 2643 uma_zone_set_max(V_pf_limits[index].zone, 2644 limit == 0 ? INT_MAX : limit); 2645 if (old_limit != NULL) 2646 *old_limit = V_pf_limits[index].limit; 2647 V_pf_limits[index].limit = limit; 2648 PF_RULES_WUNLOCK(); 2649 2650 return (0); 2651 } 2652 2653 int 2654 pf_ioctl_get_limit(int index, unsigned int *limit) 2655 { 2656 PF_RULES_RLOCK_TRACKER; 2657 2658 if (index < 0 || index >= PF_LIMIT_MAX) 2659 return (EINVAL); 2660 2661 PF_RULES_RLOCK(); 2662 *limit = V_pf_limits[index].limit; 2663 PF_RULES_RUNLOCK(); 2664 2665 return (0); 2666 } 2667 2668 int 2669 pf_ioctl_begin_addrs(uint32_t *ticket) 2670 { 2671 PF_RULES_WLOCK(); 2672 pf_empty_kpool(&V_pf_pabuf[0]); 2673 pf_empty_kpool(&V_pf_pabuf[1]); 2674 pf_empty_kpool(&V_pf_pabuf[2]); 2675 *ticket = ++V_ticket_pabuf; 2676 PF_RULES_WUNLOCK(); 2677 2678 return (0); 2679 } 2680 2681 int 2682 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp) 2683 { 2684 struct pf_kpooladdr *pa = NULL; 2685 struct pfi_kkif *kif = NULL; 2686 int error; 2687 2688 if (pp->which != PF_RDR && pp->which != PF_NAT && 2689 pp->which != PF_RT) 2690 return (EINVAL); 2691 2692 switch (pp->af) { 2693 #ifdef INET 2694 case AF_INET: 2695 /* FALLTHROUGH */ 2696 #endif /* INET */ 2697 #ifdef INET6 2698 case AF_INET6: 2699 /* FALLTHROUGH */ 2700 #endif /* INET6 */ 2701 case AF_UNSPEC: 2702 break; 2703 default: 2704 return (EAFNOSUPPORT); 2705 } 2706 2707 if (pp->addr.addr.type != PF_ADDR_ADDRMASK && 2708 pp->addr.addr.type != PF_ADDR_DYNIFTL && 2709 pp->addr.addr.type != PF_ADDR_TABLE) 2710 return (EINVAL); 2711 2712 if (pp->addr.addr.p.dyn != NULL) 2713 return (EINVAL); 2714 2715 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); 2716 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa); 2717 if (error != 0) 2718 goto out; 2719 if (pa->ifname[0]) 2720 kif = pf_kkif_create(M_WAITOK); 2721 PF_RULES_WLOCK(); 2722 if (pp->ticket != V_ticket_pabuf) { 2723 PF_RULES_WUNLOCK(); 2724 if (pa->ifname[0]) 2725 pf_kkif_free(kif); 2726 error = EBUSY; 2727 goto out; 2728 } 2729 if (pa->ifname[0]) { 2730 pa->kif = pfi_kkif_attach(kif, pa->ifname); 2731 kif = NULL; 2732 pfi_kkif_ref(pa->kif); 2733 } else 2734 pa->kif = NULL; 2735 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = 2736 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { 2737 if (pa->ifname[0]) 2738 pfi_kkif_unref(pa->kif); 2739 PF_RULES_WUNLOCK(); 2740 goto out; 2741 } 2742 pa->af = pp->af; 2743 switch (pp->which) { 2744 case PF_NAT: 2745 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries); 2746 break; 2747 case PF_RDR: 2748 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries); 2749 break; 2750 case PF_RT: 2751 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries); 2752 break; 2753 } 2754 PF_RULES_WUNLOCK(); 2755 2756 return (0); 2757 2758 out: 2759 free(pa, M_PFRULE); 2760 return (error); 2761 } 2762 2763 int 2764 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp) 2765 { 2766 struct pf_kpool *pool; 2767 struct pf_kpooladdr *pa; 2768 2769 PF_RULES_RLOCK_TRACKER; 2770 2771 if (pp->which != PF_RDR && pp->which != PF_NAT && 2772 pp->which != PF_RT) 2773 return (EINVAL); 2774 2775 pp->anchor[sizeof(pp->anchor) - 1] = 0; 2776 pp->nr = 0; 2777 2778 PF_RULES_RLOCK(); 2779 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2780 pp->r_num, 0, 1, 0, pp->which); 2781 if (pool == NULL) { 2782 PF_RULES_RUNLOCK(); 2783 return (EBUSY); 2784 } 2785 TAILQ_FOREACH(pa, &pool->list, entries) 2786 pp->nr++; 2787 PF_RULES_RUNLOCK(); 2788 2789 return (0); 2790 } 2791 2792 int 2793 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp) 2794 { 2795 struct pf_kpool *pool; 2796 struct pf_kpooladdr *pa; 2797 u_int32_t nr = 0; 2798 2799 if (pp->which != PF_RDR && pp->which != PF_NAT && 2800 pp->which != PF_RT) 2801 return (EINVAL); 2802 2803 PF_RULES_RLOCK_TRACKER; 2804 2805 pp->anchor[sizeof(pp->anchor) - 1] = '\0'; 2806 2807 PF_RULES_RLOCK(); 2808 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, 2809 pp->r_num, 0, 1, 1, pp->which); 2810 if (pool == NULL) { 2811 PF_RULES_RUNLOCK(); 2812 return (EBUSY); 2813 } 2814 pa = TAILQ_FIRST(&pool->list); 2815 while ((pa != NULL) && (nr < pp->nr)) { 2816 pa = TAILQ_NEXT(pa, entries); 2817 nr++; 2818 } 2819 if (pa == NULL) { 2820 PF_RULES_RUNLOCK(); 2821 return (EBUSY); 2822 } 2823 pf_kpooladdr_to_pooladdr(pa, &pp->addr); 2824 pp->af = pa->af; 2825 pf_addr_copyout(&pp->addr.addr); 2826 PF_RULES_RUNLOCK(); 2827 2828 return (0); 2829 } 2830 2831 int 2832 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr) 2833 { 2834 struct pf_kruleset *ruleset; 2835 struct pf_kanchor *anchor; 2836 2837 PF_RULES_RLOCK_TRACKER; 2838 2839 pr->path[sizeof(pr->path) - 1] = '\0'; 2840 2841 PF_RULES_RLOCK(); 2842 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2843 PF_RULES_RUNLOCK(); 2844 return (ENOENT); 2845 } 2846 pr->nr = 0; 2847 if (ruleset == &pf_main_ruleset) { 2848 /* XXX kludge for pf_main_ruleset */ 2849 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2850 if (anchor->parent == NULL) 2851 pr->nr++; 2852 } else { 2853 RB_FOREACH(anchor, pf_kanchor_node, 2854 &ruleset->anchor->children) 2855 pr->nr++; 2856 } 2857 PF_RULES_RUNLOCK(); 2858 2859 return (0); 2860 } 2861 2862 int 2863 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr) 2864 { 2865 struct pf_kruleset *ruleset; 2866 struct pf_kanchor *anchor; 2867 u_int32_t nr = 0; 2868 int error = 0; 2869 2870 PF_RULES_RLOCK_TRACKER; 2871 2872 PF_RULES_RLOCK(); 2873 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { 2874 PF_RULES_RUNLOCK(); 2875 return (ENOENT); 2876 } 2877 2878 pr->name[0] = '\0'; 2879 if (ruleset == &pf_main_ruleset) { 2880 /* XXX kludge for pf_main_ruleset */ 2881 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) 2882 if (anchor->parent == NULL && nr++ == pr->nr) { 2883 strlcpy(pr->name, anchor->name, 2884 sizeof(pr->name)); 2885 break; 2886 } 2887 } else { 2888 RB_FOREACH(anchor, pf_kanchor_node, 2889 &ruleset->anchor->children) 2890 if (nr++ == pr->nr) { 2891 strlcpy(pr->name, anchor->name, 2892 sizeof(pr->name)); 2893 break; 2894 } 2895 } 2896 if (!pr->name[0]) 2897 error = EBUSY; 2898 PF_RULES_RUNLOCK(); 2899 2900 return (error); 2901 } 2902 2903 int 2904 pf_ioctl_natlook(struct pfioc_natlook *pnl) 2905 { 2906 struct pf_state_key *sk; 2907 struct pf_kstate *state; 2908 struct pf_state_key_cmp key; 2909 int m = 0, direction = pnl->direction; 2910 int sidx, didx; 2911 2912 /* NATLOOK src and dst are reversed, so reverse sidx/didx */ 2913 sidx = (direction == PF_IN) ? 1 : 0; 2914 didx = (direction == PF_IN) ? 0 : 1; 2915 2916 if (!pnl->proto || 2917 PF_AZERO(&pnl->saddr, pnl->af) || 2918 PF_AZERO(&pnl->daddr, pnl->af) || 2919 ((pnl->proto == IPPROTO_TCP || 2920 pnl->proto == IPPROTO_UDP) && 2921 (!pnl->dport || !pnl->sport))) 2922 return (EINVAL); 2923 2924 switch (pnl->direction) { 2925 case PF_IN: 2926 case PF_OUT: 2927 case PF_INOUT: 2928 break; 2929 default: 2930 return (EINVAL); 2931 } 2932 2933 switch (pnl->af) { 2934 #ifdef INET 2935 case AF_INET: 2936 break; 2937 #endif /* INET */ 2938 #ifdef INET6 2939 case AF_INET6: 2940 break; 2941 #endif /* INET6 */ 2942 default: 2943 return (EAFNOSUPPORT); 2944 } 2945 2946 bzero(&key, sizeof(key)); 2947 key.af = pnl->af; 2948 key.proto = pnl->proto; 2949 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af); 2950 key.port[sidx] = pnl->sport; 2951 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af); 2952 key.port[didx] = pnl->dport; 2953 2954 state = pf_find_state_all(&key, direction, &m); 2955 if (state == NULL) 2956 return (ENOENT); 2957 2958 if (m > 1) { 2959 PF_STATE_UNLOCK(state); 2960 return (E2BIG); /* more than one state */ 2961 } 2962 2963 sk = state->key[sidx]; 2964 pf_addrcpy(&pnl->rsaddr, 2965 &sk->addr[sidx], sk->af); 2966 pnl->rsport = sk->port[sidx]; 2967 pf_addrcpy(&pnl->rdaddr, 2968 &sk->addr[didx], sk->af); 2969 pnl->rdport = sk->port[didx]; 2970 PF_STATE_UNLOCK(state); 2971 2972 return (0); 2973 } 2974 2975 static int 2976 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) 2977 { 2978 int error = 0; 2979 PF_RULES_RLOCK_TRACKER; 2980 2981 #define ERROUT_IOCTL(target, x) \ 2982 do { \ 2983 error = (x); \ 2984 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \ 2985 goto target; \ 2986 } while (0) 2987 2988 2989 /* XXX keep in sync with switch() below */ 2990 if (securelevel_gt(td->td_ucred, 2)) 2991 switch (cmd) { 2992 case DIOCGETRULES: 2993 case DIOCGETRULENV: 2994 case DIOCGETADDRS: 2995 case DIOCGETADDR: 2996 case DIOCGETSTATE: 2997 case DIOCGETSTATENV: 2998 case DIOCSETSTATUSIF: 2999 case DIOCGETSTATUSNV: 3000 case DIOCCLRSTATUS: 3001 case DIOCNATLOOK: 3002 case DIOCSETDEBUG: 3003 #ifdef COMPAT_FREEBSD14 3004 case DIOCGETSTATES: 3005 case DIOCGETSTATESV2: 3006 #endif 3007 case DIOCGETTIMEOUT: 3008 case DIOCCLRRULECTRS: 3009 case DIOCGETLIMIT: 3010 case DIOCGETALTQSV0: 3011 case DIOCGETALTQSV1: 3012 case DIOCGETALTQV0: 3013 case DIOCGETALTQV1: 3014 case DIOCGETQSTATSV0: 3015 case DIOCGETQSTATSV1: 3016 case DIOCGETRULESETS: 3017 case DIOCGETRULESET: 3018 case DIOCRGETTABLES: 3019 case DIOCRGETTSTATS: 3020 case DIOCRCLRTSTATS: 3021 case DIOCRCLRADDRS: 3022 case DIOCRADDADDRS: 3023 case DIOCRDELADDRS: 3024 case DIOCRSETADDRS: 3025 case DIOCRGETADDRS: 3026 case DIOCRGETASTATS: 3027 case DIOCRCLRASTATS: 3028 case DIOCRTSTADDRS: 3029 case DIOCOSFPGET: 3030 case DIOCGETSRCNODES: 3031 case DIOCCLRSRCNODES: 3032 case DIOCGETSYNCOOKIES: 3033 case DIOCIGETIFACES: 3034 case DIOCGIFSPEEDV0: 3035 case DIOCGIFSPEEDV1: 3036 case DIOCSETIFFLAG: 3037 case DIOCCLRIFFLAG: 3038 case DIOCGETETHRULES: 3039 case DIOCGETETHRULE: 3040 case DIOCGETETHRULESETS: 3041 case DIOCGETETHRULESET: 3042 break; 3043 case DIOCRCLRTABLES: 3044 case DIOCRADDTABLES: 3045 case DIOCRDELTABLES: 3046 case DIOCRSETTFLAGS: 3047 if (((struct pfioc_table *)addr)->pfrio_flags & 3048 PFR_FLAG_DUMMY) 3049 break; /* dummy operation ok */ 3050 return (EPERM); 3051 default: 3052 return (EPERM); 3053 } 3054 3055 if (!(flags & FWRITE)) 3056 switch (cmd) { 3057 case DIOCGETRULES: 3058 case DIOCGETADDRS: 3059 case DIOCGETADDR: 3060 case DIOCGETSTATE: 3061 case DIOCGETSTATENV: 3062 case DIOCGETSTATUSNV: 3063 #ifdef COMPAT_FREEBSD14 3064 case DIOCGETSTATES: 3065 case DIOCGETSTATESV2: 3066 #endif 3067 case DIOCGETTIMEOUT: 3068 case DIOCGETLIMIT: 3069 case DIOCGETALTQSV0: 3070 case DIOCGETALTQSV1: 3071 case DIOCGETALTQV0: 3072 case DIOCGETALTQV1: 3073 case DIOCGETQSTATSV0: 3074 case DIOCGETQSTATSV1: 3075 case DIOCGETRULESETS: 3076 case DIOCGETRULESET: 3077 case DIOCNATLOOK: 3078 case DIOCRGETTABLES: 3079 case DIOCRGETTSTATS: 3080 case DIOCRGETADDRS: 3081 case DIOCRGETASTATS: 3082 case DIOCRTSTADDRS: 3083 case DIOCOSFPGET: 3084 case DIOCGETSRCNODES: 3085 case DIOCGETSYNCOOKIES: 3086 case DIOCIGETIFACES: 3087 case DIOCGIFSPEEDV1: 3088 case DIOCGIFSPEEDV0: 3089 case DIOCGETRULENV: 3090 case DIOCGETETHRULES: 3091 case DIOCGETETHRULE: 3092 case DIOCGETETHRULESETS: 3093 case DIOCGETETHRULESET: 3094 break; 3095 case DIOCRCLRTABLES: 3096 case DIOCRADDTABLES: 3097 case DIOCRDELTABLES: 3098 case DIOCRCLRTSTATS: 3099 case DIOCRCLRADDRS: 3100 case DIOCRADDADDRS: 3101 case DIOCRDELADDRS: 3102 case DIOCRSETADDRS: 3103 case DIOCRSETTFLAGS: 3104 if (((struct pfioc_table *)addr)->pfrio_flags & 3105 PFR_FLAG_DUMMY) { 3106 flags |= FWRITE; /* need write lock for dummy */ 3107 break; /* dummy operation ok */ 3108 } 3109 return (EACCES); 3110 default: 3111 return (EACCES); 3112 } 3113 3114 CURVNET_SET(TD_TO_VNET(td)); 3115 3116 switch (cmd) { 3117 #ifdef COMPAT_FREEBSD14 3118 case DIOCSTART: 3119 error = pf_start(); 3120 break; 3121 3122 case DIOCSTOP: 3123 error = pf_stop(); 3124 break; 3125 #endif 3126 3127 case DIOCGETETHRULES: { 3128 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3129 nvlist_t *nvl; 3130 void *packed; 3131 struct pf_keth_rule *tail; 3132 struct pf_keth_ruleset *rs; 3133 u_int32_t ticket, nr; 3134 const char *anchor = ""; 3135 3136 nvl = NULL; 3137 packed = NULL; 3138 3139 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x) 3140 3141 if (nv->len > pf_ioctl_maxcount) 3142 ERROUT(ENOMEM); 3143 3144 /* Copy the request in */ 3145 packed = malloc(nv->len, M_NVLIST, M_WAITOK); 3146 error = copyin(nv->data, packed, nv->len); 3147 if (error) 3148 ERROUT(error); 3149 3150 nvl = nvlist_unpack(packed, nv->len, 0); 3151 if (nvl == NULL) 3152 ERROUT(EBADMSG); 3153 3154 if (! nvlist_exists_string(nvl, "anchor")) 3155 ERROUT(EBADMSG); 3156 3157 anchor = nvlist_get_string(nvl, "anchor"); 3158 3159 rs = pf_find_keth_ruleset(anchor); 3160 3161 nvlist_destroy(nvl); 3162 nvl = NULL; 3163 free(packed, M_NVLIST); 3164 packed = NULL; 3165 3166 if (rs == NULL) 3167 ERROUT(ENOENT); 3168 3169 /* Reply */ 3170 nvl = nvlist_create(0); 3171 if (nvl == NULL) 3172 ERROUT(ENOMEM); 3173 3174 PF_RULES_RLOCK(); 3175 3176 ticket = rs->active.ticket; 3177 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq); 3178 if (tail) 3179 nr = tail->nr + 1; 3180 else 3181 nr = 0; 3182 3183 PF_RULES_RUNLOCK(); 3184 3185 nvlist_add_number(nvl, "ticket", ticket); 3186 nvlist_add_number(nvl, "nr", nr); 3187 3188 packed = nvlist_pack(nvl, &nv->len); 3189 if (packed == NULL) 3190 ERROUT(ENOMEM); 3191 3192 if (nv->size == 0) 3193 ERROUT(0); 3194 else if (nv->size < nv->len) 3195 ERROUT(ENOSPC); 3196 3197 error = copyout(packed, nv->data, nv->len); 3198 3199 #undef ERROUT 3200 DIOCGETETHRULES_error: 3201 free(packed, M_NVLIST); 3202 nvlist_destroy(nvl); 3203 break; 3204 } 3205 3206 case DIOCGETETHRULE: { 3207 struct epoch_tracker et; 3208 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3209 nvlist_t *nvl = NULL; 3210 void *nvlpacked = NULL; 3211 struct pf_keth_rule *rule = NULL; 3212 struct pf_keth_ruleset *rs; 3213 u_int32_t ticket, nr; 3214 bool clear = false; 3215 const char *anchor; 3216 3217 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x) 3218 3219 if (nv->len > pf_ioctl_maxcount) 3220 ERROUT(ENOMEM); 3221 3222 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3223 error = copyin(nv->data, nvlpacked, nv->len); 3224 if (error) 3225 ERROUT(error); 3226 3227 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3228 if (nvl == NULL) 3229 ERROUT(EBADMSG); 3230 if (! nvlist_exists_number(nvl, "ticket")) 3231 ERROUT(EBADMSG); 3232 ticket = nvlist_get_number(nvl, "ticket"); 3233 if (! nvlist_exists_string(nvl, "anchor")) 3234 ERROUT(EBADMSG); 3235 anchor = nvlist_get_string(nvl, "anchor"); 3236 3237 if (nvlist_exists_bool(nvl, "clear")) 3238 clear = nvlist_get_bool(nvl, "clear"); 3239 3240 if (clear && !(flags & FWRITE)) 3241 ERROUT(EACCES); 3242 3243 if (! nvlist_exists_number(nvl, "nr")) 3244 ERROUT(EBADMSG); 3245 nr = nvlist_get_number(nvl, "nr"); 3246 3247 PF_RULES_RLOCK(); 3248 rs = pf_find_keth_ruleset(anchor); 3249 if (rs == NULL) { 3250 PF_RULES_RUNLOCK(); 3251 ERROUT(ENOENT); 3252 } 3253 if (ticket != rs->active.ticket) { 3254 PF_RULES_RUNLOCK(); 3255 ERROUT(EBUSY); 3256 } 3257 3258 nvlist_destroy(nvl); 3259 nvl = NULL; 3260 free(nvlpacked, M_NVLIST); 3261 nvlpacked = NULL; 3262 3263 rule = TAILQ_FIRST(rs->active.rules); 3264 while ((rule != NULL) && (rule->nr != nr)) 3265 rule = TAILQ_NEXT(rule, entries); 3266 if (rule == NULL) { 3267 PF_RULES_RUNLOCK(); 3268 ERROUT(ENOENT); 3269 } 3270 /* Make sure rule can't go away. */ 3271 NET_EPOCH_ENTER(et); 3272 PF_RULES_RUNLOCK(); 3273 nvl = pf_keth_rule_to_nveth_rule(rule); 3274 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) { 3275 NET_EPOCH_EXIT(et); 3276 ERROUT(EBUSY); 3277 } 3278 NET_EPOCH_EXIT(et); 3279 if (nvl == NULL) 3280 ERROUT(ENOMEM); 3281 3282 nvlpacked = nvlist_pack(nvl, &nv->len); 3283 if (nvlpacked == NULL) 3284 ERROUT(ENOMEM); 3285 3286 if (nv->size == 0) 3287 ERROUT(0); 3288 else if (nv->size < nv->len) 3289 ERROUT(ENOSPC); 3290 3291 error = copyout(nvlpacked, nv->data, nv->len); 3292 if (error == 0 && clear) { 3293 counter_u64_zero(rule->evaluations); 3294 for (int i = 0; i < 2; i++) { 3295 counter_u64_zero(rule->packets[i]); 3296 counter_u64_zero(rule->bytes[i]); 3297 } 3298 } 3299 3300 #undef ERROUT 3301 DIOCGETETHRULE_error: 3302 free(nvlpacked, M_NVLIST); 3303 nvlist_destroy(nvl); 3304 break; 3305 } 3306 3307 case DIOCADDETHRULE: { 3308 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3309 nvlist_t *nvl = NULL; 3310 void *nvlpacked = NULL; 3311 struct pf_keth_rule *rule = NULL, *tail = NULL; 3312 struct pf_keth_ruleset *ruleset = NULL; 3313 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL; 3314 const char *anchor = "", *anchor_call = ""; 3315 3316 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x) 3317 3318 if (nv->len > pf_ioctl_maxcount) 3319 ERROUT(ENOMEM); 3320 3321 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3322 error = copyin(nv->data, nvlpacked, nv->len); 3323 if (error) 3324 ERROUT(error); 3325 3326 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3327 if (nvl == NULL) 3328 ERROUT(EBADMSG); 3329 3330 if (! nvlist_exists_number(nvl, "ticket")) 3331 ERROUT(EBADMSG); 3332 3333 if (nvlist_exists_string(nvl, "anchor")) 3334 anchor = nvlist_get_string(nvl, "anchor"); 3335 if (nvlist_exists_string(nvl, "anchor_call")) 3336 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3337 3338 ruleset = pf_find_keth_ruleset(anchor); 3339 if (ruleset == NULL) 3340 ERROUT(EINVAL); 3341 3342 if (nvlist_get_number(nvl, "ticket") != 3343 ruleset->inactive.ticket) { 3344 DPFPRINTF(PF_DEBUG_MISC, 3345 "ticket: %d != %d", 3346 (u_int32_t)nvlist_get_number(nvl, "ticket"), 3347 ruleset->inactive.ticket); 3348 ERROUT(EBUSY); 3349 } 3350 3351 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); 3352 rule->timestamp = NULL; 3353 3354 error = pf_nveth_rule_to_keth_rule(nvl, rule); 3355 if (error != 0) 3356 ERROUT(error); 3357 3358 if (rule->ifname[0]) 3359 kif = pf_kkif_create(M_WAITOK); 3360 if (rule->bridge_to_name[0]) 3361 bridge_to_kif = pf_kkif_create(M_WAITOK); 3362 rule->evaluations = counter_u64_alloc(M_WAITOK); 3363 for (int i = 0; i < 2; i++) { 3364 rule->packets[i] = counter_u64_alloc(M_WAITOK); 3365 rule->bytes[i] = counter_u64_alloc(M_WAITOK); 3366 } 3367 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone, 3368 M_WAITOK | M_ZERO); 3369 3370 PF_RULES_WLOCK(); 3371 3372 if (rule->ifname[0]) { 3373 rule->kif = pfi_kkif_attach(kif, rule->ifname); 3374 pfi_kkif_ref(rule->kif); 3375 } else 3376 rule->kif = NULL; 3377 if (rule->bridge_to_name[0]) { 3378 rule->bridge_to = pfi_kkif_attach(bridge_to_kif, 3379 rule->bridge_to_name); 3380 pfi_kkif_ref(rule->bridge_to); 3381 } else 3382 rule->bridge_to = NULL; 3383 3384 #ifdef ALTQ 3385 /* set queue IDs */ 3386 if (rule->qname[0] != 0) { 3387 if ((rule->qid = pf_qname2qid(rule->qname, true)) == 0) 3388 error = EBUSY; 3389 else 3390 rule->qid = rule->qid; 3391 } 3392 #endif 3393 if (rule->tagname[0]) 3394 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) 3395 error = EBUSY; 3396 if (rule->match_tagname[0]) 3397 if ((rule->match_tag = pf_tagname2tag( 3398 rule->match_tagname)) == 0) 3399 error = EBUSY; 3400 3401 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE) 3402 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr); 3403 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE) 3404 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr); 3405 3406 if (error) { 3407 pf_free_eth_rule(rule); 3408 PF_RULES_WUNLOCK(); 3409 ERROUT(error); 3410 } 3411 3412 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) { 3413 pf_free_eth_rule(rule); 3414 PF_RULES_WUNLOCK(); 3415 ERROUT(EINVAL); 3416 } 3417 3418 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq); 3419 if (tail) 3420 rule->nr = tail->nr + 1; 3421 else 3422 rule->nr = 0; 3423 3424 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries); 3425 3426 PF_RULES_WUNLOCK(); 3427 3428 #undef ERROUT 3429 DIOCADDETHRULE_error: 3430 nvlist_destroy(nvl); 3431 free(nvlpacked, M_NVLIST); 3432 break; 3433 } 3434 3435 case DIOCGETETHRULESETS: { 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; 3443 3444 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x) 3445 3446 if (nv->len > pf_ioctl_maxcount) 3447 ERROUT(ENOMEM); 3448 3449 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3450 error = copyin(nv->data, nvlpacked, nv->len); 3451 if (error) 3452 ERROUT(error); 3453 3454 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3455 if (nvl == NULL) 3456 ERROUT(EBADMSG); 3457 if (! nvlist_exists_string(nvl, "path")) 3458 ERROUT(EBADMSG); 3459 3460 NET_EPOCH_ENTER(et); 3461 3462 if ((ruleset = pf_find_keth_ruleset( 3463 nvlist_get_string(nvl, "path"))) == NULL) { 3464 NET_EPOCH_EXIT(et); 3465 ERROUT(ENOENT); 3466 } 3467 3468 if (ruleset->anchor == NULL) { 3469 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors) 3470 if (anchor->parent == NULL) 3471 nr++; 3472 } else { 3473 RB_FOREACH(anchor, pf_keth_anchor_node, 3474 &ruleset->anchor->children) 3475 nr++; 3476 } 3477 3478 NET_EPOCH_EXIT(et); 3479 3480 nvlist_destroy(nvl); 3481 nvl = NULL; 3482 free(nvlpacked, M_NVLIST); 3483 nvlpacked = NULL; 3484 3485 nvl = nvlist_create(0); 3486 if (nvl == NULL) 3487 ERROUT(ENOMEM); 3488 3489 nvlist_add_number(nvl, "nr", nr); 3490 3491 nvlpacked = nvlist_pack(nvl, &nv->len); 3492 if (nvlpacked == NULL) 3493 ERROUT(ENOMEM); 3494 3495 if (nv->size == 0) 3496 ERROUT(0); 3497 else if (nv->size < nv->len) 3498 ERROUT(ENOSPC); 3499 3500 error = copyout(nvlpacked, nv->data, nv->len); 3501 3502 #undef ERROUT 3503 DIOCGETETHRULESETS_error: 3504 free(nvlpacked, M_NVLIST); 3505 nvlist_destroy(nvl); 3506 break; 3507 } 3508 3509 case DIOCGETETHRULESET: { 3510 struct epoch_tracker et; 3511 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3512 nvlist_t *nvl = NULL; 3513 void *nvlpacked = NULL; 3514 struct pf_keth_ruleset *ruleset; 3515 struct pf_keth_anchor *anchor; 3516 int nr = 0, req_nr = 0; 3517 bool found = false; 3518 3519 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x) 3520 3521 if (nv->len > pf_ioctl_maxcount) 3522 ERROUT(ENOMEM); 3523 3524 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3525 error = copyin(nv->data, nvlpacked, nv->len); 3526 if (error) 3527 ERROUT(error); 3528 3529 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3530 if (nvl == NULL) 3531 ERROUT(EBADMSG); 3532 if (! nvlist_exists_string(nvl, "path")) 3533 ERROUT(EBADMSG); 3534 if (! nvlist_exists_number(nvl, "nr")) 3535 ERROUT(EBADMSG); 3536 3537 req_nr = nvlist_get_number(nvl, "nr"); 3538 3539 NET_EPOCH_ENTER(et); 3540 3541 if ((ruleset = pf_find_keth_ruleset( 3542 nvlist_get_string(nvl, "path"))) == NULL) { 3543 NET_EPOCH_EXIT(et); 3544 ERROUT(ENOENT); 3545 } 3546 3547 nvlist_destroy(nvl); 3548 nvl = NULL; 3549 free(nvlpacked, M_NVLIST); 3550 nvlpacked = NULL; 3551 3552 nvl = nvlist_create(0); 3553 if (nvl == NULL) { 3554 NET_EPOCH_EXIT(et); 3555 ERROUT(ENOMEM); 3556 } 3557 3558 if (ruleset->anchor == NULL) { 3559 RB_FOREACH(anchor, pf_keth_anchor_global, 3560 &V_pf_keth_anchors) { 3561 if (anchor->parent == NULL && nr++ == req_nr) { 3562 found = true; 3563 break; 3564 } 3565 } 3566 } else { 3567 RB_FOREACH(anchor, pf_keth_anchor_node, 3568 &ruleset->anchor->children) { 3569 if (nr++ == req_nr) { 3570 found = true; 3571 break; 3572 } 3573 } 3574 } 3575 3576 NET_EPOCH_EXIT(et); 3577 if (found) { 3578 nvlist_add_number(nvl, "nr", nr); 3579 nvlist_add_string(nvl, "name", anchor->name); 3580 if (ruleset->anchor) 3581 nvlist_add_string(nvl, "path", 3582 ruleset->anchor->path); 3583 else 3584 nvlist_add_string(nvl, "path", ""); 3585 } else { 3586 ERROUT(EBUSY); 3587 } 3588 3589 nvlpacked = nvlist_pack(nvl, &nv->len); 3590 if (nvlpacked == NULL) 3591 ERROUT(ENOMEM); 3592 3593 if (nv->size == 0) 3594 ERROUT(0); 3595 else if (nv->size < nv->len) 3596 ERROUT(ENOSPC); 3597 3598 error = copyout(nvlpacked, nv->data, nv->len); 3599 3600 #undef ERROUT 3601 DIOCGETETHRULESET_error: 3602 free(nvlpacked, M_NVLIST); 3603 nvlist_destroy(nvl); 3604 break; 3605 } 3606 3607 case DIOCADDRULENV: { 3608 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3609 nvlist_t *nvl = NULL; 3610 void *nvlpacked = NULL; 3611 struct pf_krule *rule = NULL; 3612 const char *anchor = "", *anchor_call = ""; 3613 uint32_t ticket = 0, pool_ticket = 0; 3614 3615 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x) 3616 3617 if (nv->len > pf_ioctl_maxcount) 3618 ERROUT(ENOMEM); 3619 3620 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3621 error = copyin(nv->data, nvlpacked, nv->len); 3622 if (error) 3623 ERROUT(error); 3624 3625 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3626 if (nvl == NULL) 3627 ERROUT(EBADMSG); 3628 3629 if (! nvlist_exists_number(nvl, "ticket")) 3630 ERROUT(EINVAL); 3631 ticket = nvlist_get_number(nvl, "ticket"); 3632 3633 if (! nvlist_exists_number(nvl, "pool_ticket")) 3634 ERROUT(EINVAL); 3635 pool_ticket = nvlist_get_number(nvl, "pool_ticket"); 3636 3637 if (! nvlist_exists_nvlist(nvl, "rule")) 3638 ERROUT(EINVAL); 3639 3640 rule = pf_krule_alloc(); 3641 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"), 3642 rule); 3643 if (error) 3644 ERROUT(error); 3645 3646 if (nvlist_exists_string(nvl, "anchor")) 3647 anchor = nvlist_get_string(nvl, "anchor"); 3648 if (nvlist_exists_string(nvl, "anchor_call")) 3649 anchor_call = nvlist_get_string(nvl, "anchor_call"); 3650 3651 if ((error = nvlist_error(nvl))) 3652 ERROUT(error); 3653 3654 /* Frees rule on error */ 3655 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor, 3656 anchor_call, td->td_ucred->cr_ruid, 3657 td->td_proc ? td->td_proc->p_pid : 0); 3658 3659 nvlist_destroy(nvl); 3660 free(nvlpacked, M_NVLIST); 3661 break; 3662 #undef ERROUT 3663 DIOCADDRULENV_error: 3664 pf_krule_free(rule); 3665 nvlist_destroy(nvl); 3666 free(nvlpacked, M_NVLIST); 3667 3668 break; 3669 } 3670 case DIOCADDRULE: { 3671 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3672 struct pf_krule *rule; 3673 3674 rule = pf_krule_alloc(); 3675 error = pf_rule_to_krule(&pr->rule, rule); 3676 if (error != 0) { 3677 pf_krule_free(rule); 3678 goto fail; 3679 } 3680 3681 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3682 3683 /* Frees rule on error */ 3684 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, 3685 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid, 3686 td->td_proc ? td->td_proc->p_pid : 0); 3687 break; 3688 } 3689 3690 case DIOCGETRULES: { 3691 struct pfioc_rule *pr = (struct pfioc_rule *)addr; 3692 3693 pr->anchor[sizeof(pr->anchor) - 1] = '\0'; 3694 3695 error = pf_ioctl_getrules(pr); 3696 3697 break; 3698 } 3699 3700 case DIOCGETRULENV: { 3701 struct pfioc_nv *nv = (struct pfioc_nv *)addr; 3702 nvlist_t *nvrule = NULL; 3703 nvlist_t *nvl = NULL; 3704 struct pf_kruleset *ruleset; 3705 struct pf_krule *rule; 3706 void *nvlpacked = NULL; 3707 int rs_num, nr; 3708 bool clear_counter = false; 3709 3710 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x) 3711 3712 if (nv->len > pf_ioctl_maxcount) 3713 ERROUT(ENOMEM); 3714 3715 /* Copy the request in */ 3716 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 3717 error = copyin(nv->data, nvlpacked, nv->len); 3718 if (error) 3719 ERROUT(error); 3720 3721 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 3722 if (nvl == NULL) 3723 ERROUT(EBADMSG); 3724 3725 if (! nvlist_exists_string(nvl, "anchor")) 3726 ERROUT(EBADMSG); 3727 if (! nvlist_exists_number(nvl, "ruleset")) 3728 ERROUT(EBADMSG); 3729 if (! nvlist_exists_number(nvl, "ticket")) 3730 ERROUT(EBADMSG); 3731 if (! nvlist_exists_number(nvl, "nr")) 3732 ERROUT(EBADMSG); 3733 3734 if (nvlist_exists_bool(nvl, "clear_counter")) 3735 clear_counter = nvlist_get_bool(nvl, "clear_counter"); 3736 3737 if (clear_counter && !(flags & FWRITE)) 3738 ERROUT(EACCES); 3739 3740 nr = nvlist_get_number(nvl, "nr"); 3741 3742 PF_RULES_WLOCK(); 3743 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor")); 3744 if (ruleset == NULL) { 3745 PF_RULES_WUNLOCK(); 3746 ERROUT(ENOENT); 3747 } 3748 3749 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset")); 3750 if (rs_num >= PF_RULESET_MAX) { 3751 PF_RULES_WUNLOCK(); 3752 ERROUT(EINVAL); 3753 } 3754 3755 if (nvlist_get_number(nvl, "ticket") != 3756 ruleset->rules[rs_num].active.ticket) { 3757 PF_RULES_WUNLOCK(); 3758 ERROUT(EBUSY); 3759 } 3760 3761 if ((error = nvlist_error(nvl))) { 3762 PF_RULES_WUNLOCK(); 3763 ERROUT(error); 3764 } 3765 3766 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); 3767 while ((rule != NULL) && (rule->nr != nr)) 3768 rule = TAILQ_NEXT(rule, entries); 3769 if (rule == NULL) { 3770 PF_RULES_WUNLOCK(); 3771 ERROUT(EBUSY); 3772 } 3773 3774 nvrule = pf_krule_to_nvrule(rule); 3775 3776 nvlist_destroy(nvl); 3777 nvl = nvlist_create(0); 3778 if (nvl == NULL) { 3779 PF_RULES_WUNLOCK(); 3780 ERROUT(ENOMEM); 3781 } 3782 nvlist_add_number(nvl, "nr", nr); 3783 nvlist_add_nvlist(nvl, "rule", nvrule); 3784 nvlist_destroy(nvrule); 3785 nvrule = NULL; 3786 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) { 3787 PF_RULES_WUNLOCK(); 3788 ERROUT(EBUSY); 3789 } 3790 3791 free(nvlpacked, M_NVLIST); 3792 nvlpacked = nvlist_pack(nvl, &nv->len); 3793 if (nvlpacked == NULL) { 3794 PF_RULES_WUNLOCK(); 3795 ERROUT(ENOMEM); 3796 } 3797 3798 if (nv->size == 0) { 3799 PF_RULES_WUNLOCK(); 3800 ERROUT(0); 3801 } 3802 else if (nv->size < nv->len) { 3803 PF_RULES_WUNLOCK(); 3804 ERROUT(ENOSPC); 3805 } 3806 3807 if (clear_counter) 3808 pf_krule_clear_counters(rule); 3809 3810 PF_RULES_WUNLOCK(); 3811 3812 error = copyout(nvlpacked, nv->data, nv->len); 3813 3814 #undef ERROUT 3815 DIOCGETRULENV_error: 3816 free(nvlpacked, M_NVLIST); 3817 nvlist_destroy(nvrule); 3818 nvlist_destroy(nvl); 3819 3820 break; 3821 } 3822 3823 case DIOCCHANGERULE: { 3824 struct pfioc_rule *pcr = (struct pfioc_rule *)addr; 3825 struct pf_kruleset *ruleset; 3826 struct pf_krule *oldrule = NULL, *newrule = NULL; 3827 struct pfi_kkif *kif = NULL; 3828 struct pf_kpooladdr *pa; 3829 u_int32_t nr = 0; 3830 int rs_num; 3831 3832 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0'; 3833 3834 if (pcr->action < PF_CHANGE_ADD_HEAD || 3835 pcr->action > PF_CHANGE_GET_TICKET) { 3836 error = EINVAL; 3837 goto fail; 3838 } 3839 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { 3840 error = EINVAL; 3841 goto fail; 3842 } 3843 3844 if (pcr->action != PF_CHANGE_REMOVE) { 3845 newrule = pf_krule_alloc(); 3846 error = pf_rule_to_krule(&pcr->rule, newrule); 3847 if (error != 0) { 3848 pf_krule_free(newrule); 3849 goto fail; 3850 } 3851 3852 if ((error = pf_rule_checkaf(newrule))) { 3853 pf_krule_free(newrule); 3854 goto fail; 3855 } 3856 if (newrule->ifname[0]) 3857 kif = pf_kkif_create(M_WAITOK); 3858 pf_counter_u64_init(&newrule->evaluations, M_WAITOK); 3859 for (int i = 0; i < 2; i++) { 3860 pf_counter_u64_init(&newrule->packets[i], M_WAITOK); 3861 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK); 3862 } 3863 newrule->states_cur = counter_u64_alloc(M_WAITOK); 3864 newrule->states_tot = counter_u64_alloc(M_WAITOK); 3865 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 3866 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK); 3867 newrule->cuid = td->td_ucred->cr_ruid; 3868 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; 3869 TAILQ_INIT(&newrule->nat.list); 3870 TAILQ_INIT(&newrule->rdr.list); 3871 TAILQ_INIT(&newrule->route.list); 3872 } 3873 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x) 3874 3875 PF_CONFIG_LOCK(); 3876 PF_RULES_WLOCK(); 3877 #ifdef PF_WANT_32_TO_64_COUNTER 3878 if (newrule != NULL) { 3879 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist); 3880 newrule->allrulelinked = true; 3881 V_pf_allrulecount++; 3882 } 3883 #endif 3884 3885 if (!(pcr->action == PF_CHANGE_REMOVE || 3886 pcr->action == PF_CHANGE_GET_TICKET) && 3887 pcr->pool_ticket != V_ticket_pabuf) 3888 ERROUT(EBUSY); 3889 3890 ruleset = pf_find_kruleset(pcr->anchor); 3891 if (ruleset == NULL) 3892 ERROUT(EINVAL); 3893 3894 rs_num = pf_get_ruleset_number(pcr->rule.action); 3895 if (rs_num >= PF_RULESET_MAX) 3896 ERROUT(EINVAL); 3897 3898 /* 3899 * XXXMJG: there is no guarantee that the ruleset was 3900 * created by the usual route of calling DIOCXBEGIN. 3901 * As a result it is possible the rule tree will not 3902 * be allocated yet. Hack around it by doing it here. 3903 * Note it is fine to let the tree persist in case of 3904 * error as it will be freed down the road on future 3905 * updates (if need be). 3906 */ 3907 if (ruleset->rules[rs_num].active.tree == NULL) { 3908 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT); 3909 if (ruleset->rules[rs_num].active.tree == NULL) { 3910 ERROUT(ENOMEM); 3911 } 3912 } 3913 3914 if (pcr->action == PF_CHANGE_GET_TICKET) { 3915 pcr->ticket = ++ruleset->rules[rs_num].active.ticket; 3916 ERROUT(0); 3917 } else if (pcr->ticket != 3918 ruleset->rules[rs_num].active.ticket) 3919 ERROUT(EINVAL); 3920 3921 if (pcr->action != PF_CHANGE_REMOVE) { 3922 if (newrule->ifname[0]) { 3923 newrule->kif = pfi_kkif_attach(kif, 3924 newrule->ifname); 3925 kif = NULL; 3926 pfi_kkif_ref(newrule->kif); 3927 } else 3928 newrule->kif = NULL; 3929 3930 if (newrule->rtableid > 0 && 3931 newrule->rtableid >= rt_numfibs) 3932 error = EBUSY; 3933 3934 #ifdef ALTQ 3935 /* set queue IDs */ 3936 if (newrule->qname[0] != 0) { 3937 if ((newrule->qid = 3938 pf_qname2qid(newrule->qname, true)) == 0) 3939 error = EBUSY; 3940 else if (newrule->pqname[0] != 0) { 3941 if ((newrule->pqid = 3942 pf_qname2qid(newrule->pqname, true)) == 0) 3943 error = EBUSY; 3944 } else 3945 newrule->pqid = newrule->qid; 3946 } 3947 #endif /* ALTQ */ 3948 if (newrule->tagname[0]) 3949 if ((newrule->tag = 3950 pf_tagname2tag(newrule->tagname)) == 0) 3951 error = EBUSY; 3952 if (newrule->match_tagname[0]) 3953 if ((newrule->match_tag = pf_tagname2tag( 3954 newrule->match_tagname)) == 0) 3955 error = EBUSY; 3956 if (newrule->rt && !newrule->direction) 3957 error = EINVAL; 3958 if (!newrule->log) 3959 newrule->logif = 0; 3960 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) 3961 error = ENOMEM; 3962 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) 3963 error = ENOMEM; 3964 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call)) 3965 error = EINVAL; 3966 for (int i = 0; i < 3; i++) { 3967 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) 3968 if (pa->addr.type == PF_ADDR_TABLE) { 3969 pa->addr.p.tbl = 3970 pfr_attach_table(ruleset, 3971 pa->addr.v.tblname); 3972 if (pa->addr.p.tbl == NULL) 3973 error = ENOMEM; 3974 } 3975 } 3976 3977 newrule->overload_tbl = NULL; 3978 if (newrule->overload_tblname[0]) { 3979 if ((newrule->overload_tbl = pfr_attach_table( 3980 ruleset, newrule->overload_tblname)) == 3981 NULL) 3982 error = EINVAL; 3983 else 3984 newrule->overload_tbl->pfrkt_flags |= 3985 PFR_TFLAG_ACTIVE; 3986 } 3987 3988 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list); 3989 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list); 3990 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list); 3991 if (((((newrule->action == PF_NAT) || 3992 (newrule->action == PF_RDR) || 3993 (newrule->action == PF_BINAT) || 3994 (newrule->rt > PF_NOPFROUTE)) && 3995 !newrule->anchor)) && 3996 (TAILQ_FIRST(&newrule->rdr.list) == NULL)) 3997 error = EINVAL; 3998 3999 if (error) { 4000 pf_free_rule(newrule); 4001 PF_RULES_WUNLOCK(); 4002 PF_CONFIG_UNLOCK(); 4003 goto fail; 4004 } 4005 4006 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list); 4007 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list); 4008 } 4009 pf_empty_kpool(&V_pf_pabuf[0]); 4010 pf_empty_kpool(&V_pf_pabuf[1]); 4011 pf_empty_kpool(&V_pf_pabuf[2]); 4012 4013 if (pcr->action == PF_CHANGE_ADD_HEAD) 4014 oldrule = TAILQ_FIRST( 4015 ruleset->rules[rs_num].active.ptr); 4016 else if (pcr->action == PF_CHANGE_ADD_TAIL) 4017 oldrule = TAILQ_LAST( 4018 ruleset->rules[rs_num].active.ptr, pf_krulequeue); 4019 else { 4020 oldrule = TAILQ_FIRST( 4021 ruleset->rules[rs_num].active.ptr); 4022 while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) 4023 oldrule = TAILQ_NEXT(oldrule, entries); 4024 if (oldrule == NULL) { 4025 if (newrule != NULL) 4026 pf_free_rule(newrule); 4027 PF_RULES_WUNLOCK(); 4028 PF_CONFIG_UNLOCK(); 4029 error = EINVAL; 4030 goto fail; 4031 } 4032 } 4033 4034 if (pcr->action == PF_CHANGE_REMOVE) { 4035 pf_unlink_rule(ruleset->rules[rs_num].active.ptr, 4036 oldrule); 4037 RB_REMOVE(pf_krule_global, 4038 ruleset->rules[rs_num].active.tree, oldrule); 4039 ruleset->rules[rs_num].active.rcount--; 4040 } else { 4041 pf_hash_rule(newrule); 4042 if (RB_INSERT(pf_krule_global, 4043 ruleset->rules[rs_num].active.tree, newrule) != NULL) { 4044 pf_free_rule(newrule); 4045 PF_RULES_WUNLOCK(); 4046 PF_CONFIG_UNLOCK(); 4047 error = EEXIST; 4048 goto fail; 4049 } 4050 4051 if (oldrule == NULL) 4052 TAILQ_INSERT_TAIL( 4053 ruleset->rules[rs_num].active.ptr, 4054 newrule, entries); 4055 else if (pcr->action == PF_CHANGE_ADD_HEAD || 4056 pcr->action == PF_CHANGE_ADD_BEFORE) 4057 TAILQ_INSERT_BEFORE(oldrule, newrule, entries); 4058 else 4059 TAILQ_INSERT_AFTER( 4060 ruleset->rules[rs_num].active.ptr, 4061 oldrule, newrule, entries); 4062 ruleset->rules[rs_num].active.rcount++; 4063 } 4064 4065 nr = 0; 4066 TAILQ_FOREACH(oldrule, 4067 ruleset->rules[rs_num].active.ptr, entries) 4068 oldrule->nr = nr++; 4069 4070 ruleset->rules[rs_num].active.ticket++; 4071 4072 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); 4073 pf_remove_if_empty_kruleset(ruleset); 4074 4075 PF_RULES_WUNLOCK(); 4076 PF_CONFIG_UNLOCK(); 4077 break; 4078 4079 #undef ERROUT 4080 DIOCCHANGERULE_error: 4081 PF_RULES_WUNLOCK(); 4082 PF_CONFIG_UNLOCK(); 4083 pf_krule_free(newrule); 4084 pf_kkif_free(kif); 4085 break; 4086 } 4087 4088 case DIOCCLRSTATESNV: { 4089 error = pf_clearstates_nv((struct pfioc_nv *)addr); 4090 break; 4091 } 4092 4093 case DIOCKILLSTATESNV: { 4094 error = pf_killstates_nv((struct pfioc_nv *)addr); 4095 break; 4096 } 4097 4098 case DIOCADDSTATE: { 4099 struct pfioc_state *ps = (struct pfioc_state *)addr; 4100 struct pfsync_state_1301 *sp = &ps->state; 4101 4102 if (sp->timeout >= PFTM_MAX) { 4103 error = EINVAL; 4104 goto fail; 4105 } 4106 if (V_pfsync_state_import_ptr != NULL) { 4107 PF_RULES_RLOCK(); 4108 error = V_pfsync_state_import_ptr( 4109 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL, 4110 PFSYNC_MSG_VERSION_1301); 4111 PF_RULES_RUNLOCK(); 4112 } else 4113 error = EOPNOTSUPP; 4114 break; 4115 } 4116 4117 case DIOCGETSTATE: { 4118 struct pfioc_state *ps = (struct pfioc_state *)addr; 4119 struct pf_kstate *s; 4120 4121 s = pf_find_state_byid(ps->state.id, ps->state.creatorid); 4122 if (s == NULL) { 4123 error = ENOENT; 4124 goto fail; 4125 } 4126 4127 pfsync_state_export((union pfsync_state_union*)&ps->state, 4128 s, PFSYNC_MSG_VERSION_1301); 4129 PF_STATE_UNLOCK(s); 4130 break; 4131 } 4132 4133 case DIOCGETSTATENV: { 4134 error = pf_getstate((struct pfioc_nv *)addr); 4135 break; 4136 } 4137 4138 #ifdef COMPAT_FREEBSD14 4139 case DIOCGETSTATES: { 4140 struct pfioc_states *ps = (struct pfioc_states *)addr; 4141 struct pf_kstate *s; 4142 struct pfsync_state_1301 *pstore, *p; 4143 int i, nr; 4144 size_t slice_count = 16, count; 4145 void *out; 4146 4147 if (ps->ps_len <= 0) { 4148 nr = uma_zone_get_cur(V_pf_state_z); 4149 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4150 break; 4151 } 4152 4153 out = ps->ps_states; 4154 pstore = mallocarray(slice_count, 4155 sizeof(struct pfsync_state_1301), M_PF, M_WAITOK | M_ZERO); 4156 nr = 0; 4157 4158 for (i = 0; i <= V_pf_hashmask; i++) { 4159 struct pf_idhash *ih = &V_pf_idhash[i]; 4160 4161 DIOCGETSTATES_retry: 4162 p = pstore; 4163 4164 if (LIST_EMPTY(&ih->states)) 4165 continue; 4166 4167 PF_HASHROW_LOCK(ih); 4168 count = 0; 4169 LIST_FOREACH(s, &ih->states, entry) { 4170 if (s->timeout == PFTM_UNLINKED) 4171 continue; 4172 count++; 4173 } 4174 4175 if (count > slice_count) { 4176 PF_HASHROW_UNLOCK(ih); 4177 free(pstore, M_PF); 4178 slice_count = count * 2; 4179 pstore = mallocarray(slice_count, 4180 sizeof(struct pfsync_state_1301), M_PF, 4181 M_WAITOK | M_ZERO); 4182 goto DIOCGETSTATES_retry; 4183 } 4184 4185 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4186 PF_HASHROW_UNLOCK(ih); 4187 goto DIOCGETSTATES_full; 4188 } 4189 4190 LIST_FOREACH(s, &ih->states, entry) { 4191 if (s->timeout == PFTM_UNLINKED) 4192 continue; 4193 4194 pfsync_state_export((union pfsync_state_union*)p, 4195 s, PFSYNC_MSG_VERSION_1301); 4196 p++; 4197 nr++; 4198 } 4199 PF_HASHROW_UNLOCK(ih); 4200 error = copyout(pstore, out, 4201 sizeof(struct pfsync_state_1301) * count); 4202 if (error) { 4203 free(pstore, M_PF); 4204 goto fail; 4205 } 4206 out = ps->ps_states + nr; 4207 } 4208 DIOCGETSTATES_full: 4209 ps->ps_len = sizeof(struct pfsync_state_1301) * nr; 4210 free(pstore, M_PF); 4211 4212 break; 4213 } 4214 4215 case DIOCGETSTATESV2: { 4216 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr; 4217 struct pf_kstate *s; 4218 struct pf_state_export *pstore, *p; 4219 int i, nr; 4220 size_t slice_count = 16, count; 4221 void *out; 4222 4223 if (ps->ps_req_version > PF_STATE_VERSION) { 4224 error = ENOTSUP; 4225 goto fail; 4226 } 4227 4228 if (ps->ps_len <= 0) { 4229 nr = uma_zone_get_cur(V_pf_state_z); 4230 ps->ps_len = sizeof(struct pf_state_export) * nr; 4231 break; 4232 } 4233 4234 out = ps->ps_states; 4235 pstore = mallocarray(slice_count, 4236 sizeof(struct pf_state_export), M_PF, M_WAITOK | M_ZERO); 4237 nr = 0; 4238 4239 for (i = 0; i <= V_pf_hashmask; i++) { 4240 struct pf_idhash *ih = &V_pf_idhash[i]; 4241 4242 DIOCGETSTATESV2_retry: 4243 p = pstore; 4244 4245 if (LIST_EMPTY(&ih->states)) 4246 continue; 4247 4248 PF_HASHROW_LOCK(ih); 4249 count = 0; 4250 LIST_FOREACH(s, &ih->states, entry) { 4251 if (s->timeout == PFTM_UNLINKED) 4252 continue; 4253 count++; 4254 } 4255 4256 if (count > slice_count) { 4257 PF_HASHROW_UNLOCK(ih); 4258 free(pstore, M_PF); 4259 slice_count = count * 2; 4260 pstore = mallocarray(slice_count, 4261 sizeof(struct pf_state_export), M_PF, 4262 M_WAITOK | M_ZERO); 4263 goto DIOCGETSTATESV2_retry; 4264 } 4265 4266 if ((nr+count) * sizeof(*p) > ps->ps_len) { 4267 PF_HASHROW_UNLOCK(ih); 4268 goto DIOCGETSTATESV2_full; 4269 } 4270 4271 LIST_FOREACH(s, &ih->states, entry) { 4272 if (s->timeout == PFTM_UNLINKED) 4273 continue; 4274 4275 pf_state_export(p, s); 4276 p++; 4277 nr++; 4278 } 4279 PF_HASHROW_UNLOCK(ih); 4280 error = copyout(pstore, out, 4281 sizeof(struct pf_state_export) * count); 4282 if (error) { 4283 free(pstore, M_PF); 4284 goto fail; 4285 } 4286 out = ps->ps_states + nr; 4287 } 4288 DIOCGETSTATESV2_full: 4289 ps->ps_len = nr * sizeof(struct pf_state_export); 4290 free(pstore, M_PF); 4291 4292 break; 4293 } 4294 #endif 4295 case DIOCGETSTATUSNV: { 4296 error = pf_getstatus((struct pfioc_nv *)addr); 4297 break; 4298 } 4299 4300 case DIOCSETSTATUSIF: { 4301 struct pfioc_if *pi = (struct pfioc_if *)addr; 4302 4303 if (pi->ifname[0] == 0) { 4304 bzero(V_pf_status.ifname, IFNAMSIZ); 4305 break; 4306 } 4307 PF_RULES_WLOCK(); 4308 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); 4309 PF_RULES_WUNLOCK(); 4310 break; 4311 } 4312 4313 case DIOCCLRSTATUS: { 4314 pf_ioctl_clear_status(); 4315 break; 4316 } 4317 4318 case DIOCNATLOOK: { 4319 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; 4320 4321 error = pf_ioctl_natlook(pnl); 4322 break; 4323 } 4324 4325 case DIOCSETTIMEOUT: { 4326 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4327 4328 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds, 4329 &pt->seconds); 4330 break; 4331 } 4332 4333 case DIOCGETTIMEOUT: { 4334 struct pfioc_tm *pt = (struct pfioc_tm *)addr; 4335 4336 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds); 4337 break; 4338 } 4339 4340 case DIOCGETLIMIT: { 4341 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4342 4343 error = pf_ioctl_get_limit(pl->index, &pl->limit); 4344 break; 4345 } 4346 4347 case DIOCSETLIMIT: { 4348 struct pfioc_limit *pl = (struct pfioc_limit *)addr; 4349 unsigned int old_limit; 4350 4351 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit); 4352 pl->limit = old_limit; 4353 break; 4354 } 4355 4356 case DIOCSETDEBUG: { 4357 u_int32_t *level = (u_int32_t *)addr; 4358 4359 PF_RULES_WLOCK(); 4360 V_pf_status.debug = *level; 4361 PF_RULES_WUNLOCK(); 4362 break; 4363 } 4364 4365 case DIOCCLRRULECTRS: { 4366 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ 4367 struct pf_kruleset *ruleset = &pf_main_ruleset; 4368 struct pf_krule *rule; 4369 4370 PF_RULES_WLOCK(); 4371 TAILQ_FOREACH(rule, 4372 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { 4373 pf_counter_u64_zero(&rule->evaluations); 4374 for (int i = 0; i < 2; i++) { 4375 pf_counter_u64_zero(&rule->packets[i]); 4376 pf_counter_u64_zero(&rule->bytes[i]); 4377 } 4378 } 4379 PF_RULES_WUNLOCK(); 4380 break; 4381 } 4382 4383 case DIOCGIFSPEEDV0: 4384 case DIOCGIFSPEEDV1: { 4385 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr; 4386 struct pf_ifspeed_v1 ps; 4387 struct ifnet *ifp; 4388 4389 if (psp->ifname[0] == '\0') { 4390 error = EINVAL; 4391 goto fail; 4392 } 4393 4394 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); 4395 if (error != 0) 4396 goto fail; 4397 ifp = ifunit(ps.ifname); 4398 if (ifp != NULL) { 4399 psp->baudrate32 = 4400 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX); 4401 if (cmd == DIOCGIFSPEEDV1) 4402 psp->baudrate = ifp->if_baudrate; 4403 } else { 4404 error = EINVAL; 4405 } 4406 break; 4407 } 4408 4409 #ifdef ALTQ 4410 case DIOCSTARTALTQ: { 4411 struct pf_altq *altq; 4412 4413 PF_RULES_WLOCK(); 4414 /* enable all altq interfaces on active list */ 4415 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4416 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4417 error = pf_enable_altq(altq); 4418 if (error != 0) 4419 break; 4420 } 4421 } 4422 if (error == 0) 4423 V_pf_altq_running = 1; 4424 PF_RULES_WUNLOCK(); 4425 DPFPRINTF(PF_DEBUG_MISC, "altq: started"); 4426 break; 4427 } 4428 4429 case DIOCSTOPALTQ: { 4430 struct pf_altq *altq; 4431 4432 PF_RULES_WLOCK(); 4433 /* disable all altq interfaces on active list */ 4434 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) { 4435 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { 4436 error = pf_disable_altq(altq); 4437 if (error != 0) 4438 break; 4439 } 4440 } 4441 if (error == 0) 4442 V_pf_altq_running = 0; 4443 PF_RULES_WUNLOCK(); 4444 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped"); 4445 break; 4446 } 4447 4448 case DIOCADDALTQV0: 4449 case DIOCADDALTQV1: { 4450 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4451 struct pf_altq *altq, *a; 4452 struct ifnet *ifp; 4453 4454 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); 4455 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); 4456 if (error) 4457 goto fail; 4458 altq->local_flags = 0; 4459 4460 PF_RULES_WLOCK(); 4461 if (pa->ticket != V_ticket_altqs_inactive) { 4462 PF_RULES_WUNLOCK(); 4463 free(altq, M_PFALTQ); 4464 error = EBUSY; 4465 goto fail; 4466 } 4467 4468 /* 4469 * if this is for a queue, find the discipline and 4470 * copy the necessary fields 4471 */ 4472 if (altq->qname[0] != 0) { 4473 if ((altq->qid = pf_qname2qid(altq->qname, true)) == 0) { 4474 PF_RULES_WUNLOCK(); 4475 error = EBUSY; 4476 free(altq, M_PFALTQ); 4477 goto fail; 4478 } 4479 altq->altq_disc = NULL; 4480 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { 4481 if (strncmp(a->ifname, altq->ifname, 4482 IFNAMSIZ) == 0) { 4483 altq->altq_disc = a->altq_disc; 4484 break; 4485 } 4486 } 4487 } 4488 4489 if ((ifp = ifunit(altq->ifname)) == NULL) 4490 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; 4491 else 4492 error = altq_add(ifp, altq); 4493 4494 if (error) { 4495 PF_RULES_WUNLOCK(); 4496 free(altq, M_PFALTQ); 4497 goto fail; 4498 } 4499 4500 if (altq->qname[0] != 0) 4501 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); 4502 else 4503 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries); 4504 /* version error check done on import above */ 4505 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4506 PF_RULES_WUNLOCK(); 4507 break; 4508 } 4509 4510 case DIOCGETALTQSV0: 4511 case DIOCGETALTQSV1: { 4512 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4513 struct pf_altq *altq; 4514 4515 PF_RULES_RLOCK(); 4516 pa->nr = 0; 4517 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) 4518 pa->nr++; 4519 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) 4520 pa->nr++; 4521 pa->ticket = V_ticket_altqs_active; 4522 PF_RULES_RUNLOCK(); 4523 break; 4524 } 4525 4526 case DIOCGETALTQV0: 4527 case DIOCGETALTQV1: { 4528 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr; 4529 struct pf_altq *altq; 4530 4531 PF_RULES_RLOCK(); 4532 if (pa->ticket != V_ticket_altqs_active) { 4533 PF_RULES_RUNLOCK(); 4534 error = EBUSY; 4535 goto fail; 4536 } 4537 altq = pf_altq_get_nth_active(pa->nr); 4538 if (altq == NULL) { 4539 PF_RULES_RUNLOCK(); 4540 error = EBUSY; 4541 goto fail; 4542 } 4543 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); 4544 PF_RULES_RUNLOCK(); 4545 break; 4546 } 4547 4548 case DIOCCHANGEALTQV0: 4549 case DIOCCHANGEALTQV1: 4550 /* CHANGEALTQ not supported yet! */ 4551 error = ENODEV; 4552 break; 4553 4554 case DIOCGETQSTATSV0: 4555 case DIOCGETQSTATSV1: { 4556 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr; 4557 struct pf_altq *altq; 4558 int nbytes; 4559 u_int32_t version; 4560 4561 PF_RULES_RLOCK(); 4562 if (pq->ticket != V_ticket_altqs_active) { 4563 PF_RULES_RUNLOCK(); 4564 error = EBUSY; 4565 goto fail; 4566 } 4567 nbytes = pq->nbytes; 4568 altq = pf_altq_get_nth_active(pq->nr); 4569 if (altq == NULL) { 4570 PF_RULES_RUNLOCK(); 4571 error = EBUSY; 4572 goto fail; 4573 } 4574 4575 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { 4576 PF_RULES_RUNLOCK(); 4577 error = ENXIO; 4578 goto fail; 4579 } 4580 PF_RULES_RUNLOCK(); 4581 if (cmd == DIOCGETQSTATSV0) 4582 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */ 4583 else 4584 version = pq->version; 4585 error = altq_getqstats(altq, pq->buf, &nbytes, version); 4586 if (error == 0) { 4587 pq->scheduler = altq->scheduler; 4588 pq->nbytes = nbytes; 4589 } 4590 break; 4591 } 4592 #endif /* ALTQ */ 4593 4594 case DIOCBEGINADDRS: { 4595 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4596 4597 error = pf_ioctl_begin_addrs(&pp->ticket); 4598 break; 4599 } 4600 4601 case DIOCADDADDR: { 4602 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4603 struct pf_nl_pooladdr npp = {}; 4604 4605 npp.which = PF_RDR; 4606 memcpy(&npp, pp, sizeof(*pp)); 4607 error = pf_ioctl_add_addr(&npp); 4608 break; 4609 } 4610 4611 case DIOCGETADDRS: { 4612 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4613 struct pf_nl_pooladdr npp = {}; 4614 4615 npp.which = PF_RDR; 4616 memcpy(&npp, pp, sizeof(*pp)); 4617 error = pf_ioctl_get_addrs(&npp); 4618 memcpy(pp, &npp, sizeof(*pp)); 4619 4620 break; 4621 } 4622 4623 case DIOCGETADDR: { 4624 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; 4625 struct pf_nl_pooladdr npp = {}; 4626 4627 npp.which = PF_RDR; 4628 memcpy(&npp, pp, sizeof(*pp)); 4629 error = pf_ioctl_get_addr(&npp); 4630 memcpy(pp, &npp, sizeof(*pp)); 4631 4632 break; 4633 } 4634 4635 case DIOCCHANGEADDR: { 4636 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; 4637 struct pf_kpool *pool; 4638 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL; 4639 struct pf_kruleset *ruleset; 4640 struct pfi_kkif *kif = NULL; 4641 4642 pca->anchor[sizeof(pca->anchor) - 1] = '\0'; 4643 4644 if (pca->action < PF_CHANGE_ADD_HEAD || 4645 pca->action > PF_CHANGE_REMOVE) { 4646 error = EINVAL; 4647 goto fail; 4648 } 4649 if (pca->addr.addr.type != PF_ADDR_ADDRMASK && 4650 pca->addr.addr.type != PF_ADDR_DYNIFTL && 4651 pca->addr.addr.type != PF_ADDR_TABLE) { 4652 error = EINVAL; 4653 goto fail; 4654 } 4655 if (pca->addr.addr.p.dyn != NULL) { 4656 error = EINVAL; 4657 goto fail; 4658 } 4659 4660 if (pca->action != PF_CHANGE_REMOVE) { 4661 #ifndef INET 4662 if (pca->af == AF_INET) { 4663 error = EAFNOSUPPORT; 4664 goto fail; 4665 } 4666 #endif /* INET */ 4667 #ifndef INET6 4668 if (pca->af == AF_INET6) { 4669 error = EAFNOSUPPORT; 4670 goto fail; 4671 } 4672 #endif /* INET6 */ 4673 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); 4674 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); 4675 if (newpa->ifname[0]) 4676 kif = pf_kkif_create(M_WAITOK); 4677 newpa->kif = NULL; 4678 } 4679 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x) 4680 PF_RULES_WLOCK(); 4681 ruleset = pf_find_kruleset(pca->anchor); 4682 if (ruleset == NULL) 4683 ERROUT(EBUSY); 4684 4685 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action, 4686 pca->r_num, pca->r_last, 1, 1, PF_RDR); 4687 if (pool == NULL) 4688 ERROUT(EBUSY); 4689 4690 if (pca->action != PF_CHANGE_REMOVE) { 4691 if (newpa->ifname[0]) { 4692 newpa->kif = pfi_kkif_attach(kif, newpa->ifname); 4693 pfi_kkif_ref(newpa->kif); 4694 kif = NULL; 4695 } 4696 4697 switch (newpa->addr.type) { 4698 case PF_ADDR_DYNIFTL: 4699 error = pfi_dynaddr_setup(&newpa->addr, 4700 pca->af); 4701 break; 4702 case PF_ADDR_TABLE: 4703 newpa->addr.p.tbl = pfr_attach_table(ruleset, 4704 newpa->addr.v.tblname); 4705 if (newpa->addr.p.tbl == NULL) 4706 error = ENOMEM; 4707 break; 4708 } 4709 if (error) 4710 goto DIOCCHANGEADDR_error; 4711 } 4712 4713 switch (pca->action) { 4714 case PF_CHANGE_ADD_HEAD: 4715 oldpa = TAILQ_FIRST(&pool->list); 4716 break; 4717 case PF_CHANGE_ADD_TAIL: 4718 oldpa = TAILQ_LAST(&pool->list, pf_kpalist); 4719 break; 4720 default: 4721 oldpa = TAILQ_FIRST(&pool->list); 4722 for (int i = 0; oldpa && i < pca->nr; i++) 4723 oldpa = TAILQ_NEXT(oldpa, entries); 4724 4725 if (oldpa == NULL) 4726 ERROUT(EINVAL); 4727 } 4728 4729 if (pca->action == PF_CHANGE_REMOVE) { 4730 TAILQ_REMOVE(&pool->list, oldpa, entries); 4731 switch (oldpa->addr.type) { 4732 case PF_ADDR_DYNIFTL: 4733 pfi_dynaddr_remove(oldpa->addr.p.dyn); 4734 break; 4735 case PF_ADDR_TABLE: 4736 pfr_detach_table(oldpa->addr.p.tbl); 4737 break; 4738 } 4739 if (oldpa->kif) 4740 pfi_kkif_unref(oldpa->kif); 4741 free(oldpa, M_PFRULE); 4742 } else { 4743 if (oldpa == NULL) 4744 TAILQ_INSERT_TAIL(&pool->list, newpa, entries); 4745 else if (pca->action == PF_CHANGE_ADD_HEAD || 4746 pca->action == PF_CHANGE_ADD_BEFORE) 4747 TAILQ_INSERT_BEFORE(oldpa, newpa, entries); 4748 else 4749 TAILQ_INSERT_AFTER(&pool->list, oldpa, 4750 newpa, entries); 4751 } 4752 4753 pool->cur = TAILQ_FIRST(&pool->list); 4754 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); 4755 PF_RULES_WUNLOCK(); 4756 break; 4757 4758 #undef ERROUT 4759 DIOCCHANGEADDR_error: 4760 if (newpa != NULL) { 4761 if (newpa->kif) 4762 pfi_kkif_unref(newpa->kif); 4763 free(newpa, M_PFRULE); 4764 } 4765 PF_RULES_WUNLOCK(); 4766 pf_kkif_free(kif); 4767 break; 4768 } 4769 4770 case DIOCGETRULESETS: { 4771 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4772 4773 pr->path[sizeof(pr->path) - 1] = '\0'; 4774 4775 error = pf_ioctl_get_rulesets(pr); 4776 break; 4777 } 4778 4779 case DIOCGETRULESET: { 4780 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; 4781 4782 pr->path[sizeof(pr->path) - 1] = '\0'; 4783 4784 error = pf_ioctl_get_ruleset(pr); 4785 break; 4786 } 4787 4788 case DIOCRCLRTABLES: { 4789 struct pfioc_table *io = (struct pfioc_table *)addr; 4790 4791 if (io->pfrio_esize != 0) { 4792 error = ENODEV; 4793 goto fail; 4794 } 4795 if (strnlen(io->pfrio_table.pfrt_anchor, MAXPATHLEN) 4796 == MAXPATHLEN) { 4797 error = EINVAL; 4798 goto fail; 4799 } 4800 if (strnlen(io->pfrio_table.pfrt_name, PF_TABLE_NAME_SIZE) 4801 == PF_TABLE_NAME_SIZE) { 4802 error = EINVAL; 4803 goto fail; 4804 } 4805 4806 PF_RULES_WLOCK(); 4807 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, 4808 io->pfrio_flags | PFR_FLAG_USERIOCTL); 4809 PF_RULES_WUNLOCK(); 4810 break; 4811 } 4812 4813 case DIOCRADDTABLES: { 4814 struct pfioc_table *io = (struct pfioc_table *)addr; 4815 struct pfr_table *pfrts; 4816 size_t totlen; 4817 4818 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4819 error = ENODEV; 4820 goto fail; 4821 } 4822 4823 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4824 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4825 error = ENOMEM; 4826 goto fail; 4827 } 4828 4829 totlen = io->pfrio_size * sizeof(struct pfr_table); 4830 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4831 M_PF, M_WAITOK); 4832 error = copyin(io->pfrio_buffer, pfrts, totlen); 4833 if (error) { 4834 free(pfrts, M_PF); 4835 goto fail; 4836 } 4837 PF_RULES_WLOCK(); 4838 error = pfr_add_tables(pfrts, io->pfrio_size, 4839 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4840 PF_RULES_WUNLOCK(); 4841 free(pfrts, M_PF); 4842 break; 4843 } 4844 4845 case DIOCRDELTABLES: { 4846 struct pfioc_table *io = (struct pfioc_table *)addr; 4847 struct pfr_table *pfrts; 4848 size_t totlen; 4849 4850 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4851 error = ENODEV; 4852 goto fail; 4853 } 4854 4855 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4856 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4857 error = ENOMEM; 4858 goto fail; 4859 } 4860 4861 totlen = io->pfrio_size * sizeof(struct pfr_table); 4862 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4863 M_PF, M_WAITOK); 4864 error = copyin(io->pfrio_buffer, pfrts, totlen); 4865 if (error) { 4866 free(pfrts, M_PF); 4867 goto fail; 4868 } 4869 PF_RULES_WLOCK(); 4870 error = pfr_del_tables(pfrts, io->pfrio_size, 4871 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4872 PF_RULES_WUNLOCK(); 4873 free(pfrts, M_PF); 4874 break; 4875 } 4876 4877 case DIOCRGETTABLES: { 4878 struct pfioc_table *io = (struct pfioc_table *)addr; 4879 struct pfr_table *pfrts; 4880 size_t totlen; 4881 int n; 4882 4883 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4884 error = ENODEV; 4885 goto fail; 4886 } 4887 PF_RULES_RLOCK(); 4888 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4889 if (n < 0) { 4890 PF_RULES_RUNLOCK(); 4891 error = EINVAL; 4892 goto fail; 4893 } 4894 io->pfrio_size = min(io->pfrio_size, n); 4895 4896 totlen = io->pfrio_size * sizeof(struct pfr_table); 4897 4898 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4899 M_PF, M_NOWAIT | M_ZERO); 4900 if (pfrts == NULL) { 4901 error = ENOMEM; 4902 PF_RULES_RUNLOCK(); 4903 goto fail; 4904 } 4905 error = pfr_get_tables(&io->pfrio_table, pfrts, 4906 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4907 PF_RULES_RUNLOCK(); 4908 if (error == 0) 4909 error = copyout(pfrts, io->pfrio_buffer, totlen); 4910 free(pfrts, M_PF); 4911 break; 4912 } 4913 4914 case DIOCRGETTSTATS: { 4915 struct pfioc_table *io = (struct pfioc_table *)addr; 4916 struct pfr_tstats *pfrtstats; 4917 size_t totlen; 4918 int n; 4919 4920 if (io->pfrio_esize != sizeof(struct pfr_tstats)) { 4921 error = ENODEV; 4922 goto fail; 4923 } 4924 PF_TABLE_STATS_LOCK(); 4925 PF_RULES_RLOCK(); 4926 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 4927 if (n < 0) { 4928 PF_RULES_RUNLOCK(); 4929 PF_TABLE_STATS_UNLOCK(); 4930 error = EINVAL; 4931 goto fail; 4932 } 4933 io->pfrio_size = min(io->pfrio_size, n); 4934 4935 totlen = io->pfrio_size * sizeof(struct pfr_tstats); 4936 pfrtstats = mallocarray(io->pfrio_size, 4937 sizeof(struct pfr_tstats), M_PF, M_NOWAIT | M_ZERO); 4938 if (pfrtstats == NULL) { 4939 error = ENOMEM; 4940 PF_RULES_RUNLOCK(); 4941 PF_TABLE_STATS_UNLOCK(); 4942 goto fail; 4943 } 4944 error = pfr_get_tstats(&io->pfrio_table, pfrtstats, 4945 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4946 PF_RULES_RUNLOCK(); 4947 PF_TABLE_STATS_UNLOCK(); 4948 if (error == 0) 4949 error = copyout(pfrtstats, io->pfrio_buffer, totlen); 4950 free(pfrtstats, M_PF); 4951 break; 4952 } 4953 4954 case DIOCRCLRTSTATS: { 4955 struct pfioc_table *io = (struct pfioc_table *)addr; 4956 struct pfr_table *pfrts; 4957 size_t totlen; 4958 4959 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4960 error = ENODEV; 4961 goto fail; 4962 } 4963 4964 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || 4965 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { 4966 /* We used to count tables and use the minimum required 4967 * size, so we didn't fail on overly large requests. 4968 * Keep doing so. */ 4969 io->pfrio_size = pf_ioctl_maxcount; 4970 goto fail; 4971 } 4972 4973 totlen = io->pfrio_size * sizeof(struct pfr_table); 4974 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 4975 M_PF, M_WAITOK); 4976 error = copyin(io->pfrio_buffer, pfrts, totlen); 4977 if (error) { 4978 free(pfrts, M_PF); 4979 goto fail; 4980 } 4981 4982 PF_TABLE_STATS_LOCK(); 4983 PF_RULES_RLOCK(); 4984 error = pfr_clr_tstats(pfrts, io->pfrio_size, 4985 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); 4986 PF_RULES_RUNLOCK(); 4987 PF_TABLE_STATS_UNLOCK(); 4988 free(pfrts, M_PF); 4989 break; 4990 } 4991 4992 case DIOCRSETTFLAGS: { 4993 struct pfioc_table *io = (struct pfioc_table *)addr; 4994 struct pfr_table *pfrts; 4995 size_t totlen; 4996 int n; 4997 4998 if (io->pfrio_esize != sizeof(struct pfr_table)) { 4999 error = ENODEV; 5000 goto fail; 5001 } 5002 5003 PF_RULES_RLOCK(); 5004 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); 5005 if (n < 0) { 5006 PF_RULES_RUNLOCK(); 5007 error = EINVAL; 5008 goto fail; 5009 } 5010 5011 io->pfrio_size = min(io->pfrio_size, n); 5012 PF_RULES_RUNLOCK(); 5013 5014 totlen = io->pfrio_size * sizeof(struct pfr_table); 5015 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table), 5016 M_PF, M_WAITOK); 5017 error = copyin(io->pfrio_buffer, pfrts, totlen); 5018 if (error) { 5019 free(pfrts, M_PF); 5020 goto fail; 5021 } 5022 PF_RULES_WLOCK(); 5023 error = pfr_set_tflags(pfrts, io->pfrio_size, 5024 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, 5025 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5026 PF_RULES_WUNLOCK(); 5027 free(pfrts, M_PF); 5028 break; 5029 } 5030 5031 case DIOCRCLRADDRS: { 5032 struct pfioc_table *io = (struct pfioc_table *)addr; 5033 5034 if (io->pfrio_esize != 0) { 5035 error = ENODEV; 5036 goto fail; 5037 } 5038 PF_RULES_WLOCK(); 5039 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, 5040 io->pfrio_flags | PFR_FLAG_USERIOCTL); 5041 PF_RULES_WUNLOCK(); 5042 break; 5043 } 5044 5045 case DIOCRADDADDRS: { 5046 struct pfioc_table *io = (struct pfioc_table *)addr; 5047 struct pfr_addr *pfras; 5048 size_t totlen; 5049 5050 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5051 error = ENODEV; 5052 goto fail; 5053 } 5054 if (io->pfrio_size < 0 || 5055 io->pfrio_size > pf_ioctl_maxcount || 5056 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5057 error = EINVAL; 5058 goto fail; 5059 } 5060 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5061 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5062 M_PF, M_WAITOK); 5063 error = copyin(io->pfrio_buffer, pfras, totlen); 5064 if (error) { 5065 free(pfras, M_PF); 5066 goto fail; 5067 } 5068 PF_RULES_WLOCK(); 5069 io->pfrio_nadd = 0; 5070 error = pfr_add_addrs(&io->pfrio_table, pfras, 5071 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | 5072 PFR_FLAG_USERIOCTL); 5073 PF_RULES_WUNLOCK(); 5074 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5075 error = copyout(pfras, io->pfrio_buffer, totlen); 5076 free(pfras, M_PF); 5077 break; 5078 } 5079 5080 case DIOCRDELADDRS: { 5081 struct pfioc_table *io = (struct pfioc_table *)addr; 5082 struct pfr_addr *pfras; 5083 size_t totlen; 5084 5085 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5086 error = ENODEV; 5087 goto fail; 5088 } 5089 if (io->pfrio_size < 0 || 5090 io->pfrio_size > pf_ioctl_maxcount || 5091 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5092 error = EINVAL; 5093 goto fail; 5094 } 5095 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5096 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5097 M_PF, M_WAITOK); 5098 error = copyin(io->pfrio_buffer, pfras, totlen); 5099 if (error) { 5100 free(pfras, M_PF); 5101 goto fail; 5102 } 5103 PF_RULES_WLOCK(); 5104 error = pfr_del_addrs(&io->pfrio_table, pfras, 5105 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | 5106 PFR_FLAG_USERIOCTL); 5107 PF_RULES_WUNLOCK(); 5108 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5109 error = copyout(pfras, io->pfrio_buffer, totlen); 5110 free(pfras, M_PF); 5111 break; 5112 } 5113 5114 case DIOCRSETADDRS: { 5115 struct pfioc_table *io = (struct pfioc_table *)addr; 5116 struct pfr_addr *pfras; 5117 size_t totlen, count; 5118 5119 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5120 error = ENODEV; 5121 goto fail; 5122 } 5123 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { 5124 error = EINVAL; 5125 goto fail; 5126 } 5127 count = max(io->pfrio_size, io->pfrio_size2); 5128 if (count > pf_ioctl_maxcount || 5129 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { 5130 error = EINVAL; 5131 goto fail; 5132 } 5133 totlen = count * sizeof(struct pfr_addr); 5134 pfras = mallocarray(count, sizeof(struct pfr_addr), M_PF, 5135 M_WAITOK); 5136 error = copyin(io->pfrio_buffer, pfras, totlen); 5137 if (error) { 5138 free(pfras, M_PF); 5139 goto fail; 5140 } 5141 PF_RULES_WLOCK(); 5142 error = pfr_set_addrs(&io->pfrio_table, pfras, 5143 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, 5144 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | 5145 PFR_FLAG_USERIOCTL, 0); 5146 PF_RULES_WUNLOCK(); 5147 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5148 error = copyout(pfras, io->pfrio_buffer, totlen); 5149 free(pfras, M_PF); 5150 break; 5151 } 5152 5153 case DIOCRGETADDRS: { 5154 struct pfioc_table *io = (struct pfioc_table *)addr; 5155 struct pfr_addr *pfras; 5156 size_t totlen; 5157 5158 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5159 error = ENODEV; 5160 goto fail; 5161 } 5162 if (io->pfrio_size < 0 || 5163 io->pfrio_size > pf_ioctl_maxcount || 5164 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5165 error = EINVAL; 5166 goto fail; 5167 } 5168 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5169 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5170 M_PF, M_WAITOK | M_ZERO); 5171 PF_RULES_RLOCK(); 5172 error = pfr_get_addrs(&io->pfrio_table, pfras, 5173 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5174 PF_RULES_RUNLOCK(); 5175 if (error == 0) 5176 error = copyout(pfras, io->pfrio_buffer, totlen); 5177 free(pfras, M_PF); 5178 break; 5179 } 5180 5181 case DIOCRGETASTATS: { 5182 struct pfioc_table *io = (struct pfioc_table *)addr; 5183 struct pfr_astats *pfrastats; 5184 size_t totlen; 5185 5186 if (io->pfrio_esize != sizeof(struct pfr_astats)) { 5187 error = ENODEV; 5188 goto fail; 5189 } 5190 if (io->pfrio_size < 0 || 5191 io->pfrio_size > pf_ioctl_maxcount || 5192 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { 5193 error = EINVAL; 5194 goto fail; 5195 } 5196 totlen = io->pfrio_size * sizeof(struct pfr_astats); 5197 pfrastats = mallocarray(io->pfrio_size, 5198 sizeof(struct pfr_astats), M_PF, M_WAITOK | M_ZERO); 5199 PF_RULES_RLOCK(); 5200 error = pfr_get_astats(&io->pfrio_table, pfrastats, 5201 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5202 PF_RULES_RUNLOCK(); 5203 if (error == 0) 5204 error = copyout(pfrastats, io->pfrio_buffer, totlen); 5205 free(pfrastats, M_PF); 5206 break; 5207 } 5208 5209 case DIOCRCLRASTATS: { 5210 struct pfioc_table *io = (struct pfioc_table *)addr; 5211 struct pfr_addr *pfras; 5212 size_t totlen; 5213 5214 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5215 error = ENODEV; 5216 goto fail; 5217 } 5218 if (io->pfrio_size < 0 || 5219 io->pfrio_size > pf_ioctl_maxcount || 5220 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5221 error = EINVAL; 5222 goto fail; 5223 } 5224 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5225 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5226 M_PF, M_WAITOK); 5227 error = copyin(io->pfrio_buffer, pfras, totlen); 5228 if (error) { 5229 free(pfras, M_PF); 5230 goto fail; 5231 } 5232 PF_RULES_WLOCK(); 5233 error = pfr_clr_astats(&io->pfrio_table, pfras, 5234 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | 5235 PFR_FLAG_USERIOCTL); 5236 PF_RULES_WUNLOCK(); 5237 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) 5238 error = copyout(pfras, io->pfrio_buffer, totlen); 5239 free(pfras, M_PF); 5240 break; 5241 } 5242 5243 case DIOCRTSTADDRS: { 5244 struct pfioc_table *io = (struct pfioc_table *)addr; 5245 struct pfr_addr *pfras; 5246 size_t totlen; 5247 5248 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5249 error = ENODEV; 5250 goto fail; 5251 } 5252 if (io->pfrio_size < 0 || 5253 io->pfrio_size > pf_ioctl_maxcount || 5254 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5255 error = EINVAL; 5256 goto fail; 5257 } 5258 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5259 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5260 M_PF, M_WAITOK); 5261 error = copyin(io->pfrio_buffer, pfras, totlen); 5262 if (error) { 5263 free(pfras, M_PF); 5264 goto fail; 5265 } 5266 PF_RULES_RLOCK(); 5267 error = pfr_tst_addrs(&io->pfrio_table, pfras, 5268 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | 5269 PFR_FLAG_USERIOCTL); 5270 PF_RULES_RUNLOCK(); 5271 if (error == 0) 5272 error = copyout(pfras, io->pfrio_buffer, totlen); 5273 free(pfras, M_PF); 5274 break; 5275 } 5276 5277 case DIOCRINADEFINE: { 5278 struct pfioc_table *io = (struct pfioc_table *)addr; 5279 struct pfr_addr *pfras; 5280 size_t totlen; 5281 5282 if (io->pfrio_esize != sizeof(struct pfr_addr)) { 5283 error = ENODEV; 5284 goto fail; 5285 } 5286 if (io->pfrio_size < 0 || 5287 io->pfrio_size > pf_ioctl_maxcount || 5288 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { 5289 error = EINVAL; 5290 goto fail; 5291 } 5292 totlen = io->pfrio_size * sizeof(struct pfr_addr); 5293 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), 5294 M_PF, M_WAITOK); 5295 error = copyin(io->pfrio_buffer, pfras, totlen); 5296 if (error) { 5297 free(pfras, M_PF); 5298 goto fail; 5299 } 5300 PF_RULES_WLOCK(); 5301 error = pfr_ina_define(&io->pfrio_table, pfras, 5302 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, 5303 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); 5304 PF_RULES_WUNLOCK(); 5305 free(pfras, M_PF); 5306 break; 5307 } 5308 5309 case DIOCOSFPADD: { 5310 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5311 PF_RULES_WLOCK(); 5312 error = pf_osfp_add(io); 5313 PF_RULES_WUNLOCK(); 5314 break; 5315 } 5316 5317 case DIOCOSFPGET: { 5318 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; 5319 PF_RULES_RLOCK(); 5320 error = pf_osfp_get(io); 5321 PF_RULES_RUNLOCK(); 5322 break; 5323 } 5324 5325 case DIOCXBEGIN: { 5326 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5327 struct pfioc_trans_e *ioes, *ioe; 5328 size_t totlen; 5329 int i; 5330 5331 if (io->esize != sizeof(*ioe)) { 5332 error = ENODEV; 5333 goto fail; 5334 } 5335 if (io->size < 0 || 5336 io->size > pf_ioctl_maxcount || 5337 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5338 error = EINVAL; 5339 goto fail; 5340 } 5341 totlen = sizeof(struct pfioc_trans_e) * io->size; 5342 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5343 M_PF, M_WAITOK); 5344 error = copyin(io->array, ioes, totlen); 5345 if (error) { 5346 free(ioes, M_PF); 5347 goto fail; 5348 } 5349 PF_RULES_WLOCK(); 5350 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5351 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5352 switch (ioe->rs_num) { 5353 case PF_RULESET_ETH: 5354 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) { 5355 PF_RULES_WUNLOCK(); 5356 free(ioes, M_PF); 5357 goto fail; 5358 } 5359 break; 5360 #ifdef ALTQ 5361 case PF_RULESET_ALTQ: 5362 if (ioe->anchor[0]) { 5363 PF_RULES_WUNLOCK(); 5364 free(ioes, M_PF); 5365 error = EINVAL; 5366 goto fail; 5367 } 5368 if ((error = pf_begin_altq(&ioe->ticket))) { 5369 PF_RULES_WUNLOCK(); 5370 free(ioes, M_PF); 5371 goto fail; 5372 } 5373 break; 5374 #endif /* ALTQ */ 5375 case PF_RULESET_TABLE: 5376 { 5377 struct pfr_table table; 5378 5379 bzero(&table, sizeof(table)); 5380 strlcpy(table.pfrt_anchor, ioe->anchor, 5381 sizeof(table.pfrt_anchor)); 5382 if ((error = pfr_ina_begin(&table, 5383 &ioe->ticket, NULL, 0))) { 5384 PF_RULES_WUNLOCK(); 5385 free(ioes, M_PF); 5386 goto fail; 5387 } 5388 break; 5389 } 5390 default: 5391 if ((error = pf_begin_rules(&ioe->ticket, 5392 ioe->rs_num, ioe->anchor))) { 5393 PF_RULES_WUNLOCK(); 5394 free(ioes, M_PF); 5395 goto fail; 5396 } 5397 break; 5398 } 5399 } 5400 PF_RULES_WUNLOCK(); 5401 error = copyout(ioes, io->array, totlen); 5402 free(ioes, M_PF); 5403 break; 5404 } 5405 5406 case DIOCXROLLBACK: { 5407 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5408 struct pfioc_trans_e *ioe, *ioes; 5409 size_t totlen; 5410 int i; 5411 5412 if (io->esize != sizeof(*ioe)) { 5413 error = ENODEV; 5414 goto fail; 5415 } 5416 if (io->size < 0 || 5417 io->size > pf_ioctl_maxcount || 5418 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5419 error = EINVAL; 5420 goto fail; 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 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5432 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5433 switch (ioe->rs_num) { 5434 case PF_RULESET_ETH: 5435 if ((error = pf_rollback_eth(ioe->ticket, 5436 ioe->anchor))) { 5437 PF_RULES_WUNLOCK(); 5438 free(ioes, M_PF); 5439 goto fail; /* really bad */ 5440 } 5441 break; 5442 #ifdef ALTQ 5443 case PF_RULESET_ALTQ: 5444 if (ioe->anchor[0]) { 5445 PF_RULES_WUNLOCK(); 5446 free(ioes, M_PF); 5447 error = EINVAL; 5448 goto fail; 5449 } 5450 if ((error = pf_rollback_altq(ioe->ticket))) { 5451 PF_RULES_WUNLOCK(); 5452 free(ioes, M_PF); 5453 goto fail; /* really bad */ 5454 } 5455 break; 5456 #endif /* ALTQ */ 5457 case PF_RULESET_TABLE: 5458 { 5459 struct pfr_table table; 5460 5461 bzero(&table, sizeof(table)); 5462 strlcpy(table.pfrt_anchor, ioe->anchor, 5463 sizeof(table.pfrt_anchor)); 5464 if ((error = pfr_ina_rollback(&table, 5465 ioe->ticket, NULL, 0))) { 5466 PF_RULES_WUNLOCK(); 5467 free(ioes, M_PF); 5468 goto fail; /* really bad */ 5469 } 5470 break; 5471 } 5472 default: 5473 if ((error = pf_rollback_rules(ioe->ticket, 5474 ioe->rs_num, ioe->anchor))) { 5475 PF_RULES_WUNLOCK(); 5476 free(ioes, M_PF); 5477 goto fail; /* really bad */ 5478 } 5479 break; 5480 } 5481 } 5482 PF_RULES_WUNLOCK(); 5483 free(ioes, M_PF); 5484 break; 5485 } 5486 5487 case DIOCXCOMMIT: { 5488 struct pfioc_trans *io = (struct pfioc_trans *)addr; 5489 struct pfioc_trans_e *ioe, *ioes; 5490 struct pf_kruleset *rs; 5491 struct pf_keth_ruleset *ers; 5492 size_t totlen; 5493 int i; 5494 5495 if (io->esize != sizeof(*ioe)) { 5496 error = ENODEV; 5497 goto fail; 5498 } 5499 5500 if (io->size < 0 || 5501 io->size > pf_ioctl_maxcount || 5502 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { 5503 error = EINVAL; 5504 goto fail; 5505 } 5506 5507 totlen = sizeof(struct pfioc_trans_e) * io->size; 5508 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), 5509 M_PF, M_WAITOK); 5510 error = copyin(io->array, ioes, totlen); 5511 if (error) { 5512 free(ioes, M_PF); 5513 goto fail; 5514 } 5515 PF_RULES_WLOCK(); 5516 /* First makes sure everything will succeed. */ 5517 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5518 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; 5519 switch (ioe->rs_num) { 5520 case PF_RULESET_ETH: 5521 ers = pf_find_keth_ruleset(ioe->anchor); 5522 if (ers == NULL || ioe->ticket == 0 || 5523 ioe->ticket != ers->inactive.ticket) { 5524 PF_RULES_WUNLOCK(); 5525 free(ioes, M_PF); 5526 error = EINVAL; 5527 goto fail; 5528 } 5529 break; 5530 #ifdef ALTQ 5531 case PF_RULESET_ALTQ: 5532 if (ioe->anchor[0]) { 5533 PF_RULES_WUNLOCK(); 5534 free(ioes, M_PF); 5535 error = EINVAL; 5536 goto fail; 5537 } 5538 if (!V_altqs_inactive_open || ioe->ticket != 5539 V_ticket_altqs_inactive) { 5540 PF_RULES_WUNLOCK(); 5541 free(ioes, M_PF); 5542 error = EBUSY; 5543 goto fail; 5544 } 5545 break; 5546 #endif /* ALTQ */ 5547 case PF_RULESET_TABLE: 5548 rs = pf_find_kruleset(ioe->anchor); 5549 if (rs == NULL || !rs->topen || ioe->ticket != 5550 rs->tticket) { 5551 PF_RULES_WUNLOCK(); 5552 free(ioes, M_PF); 5553 error = EBUSY; 5554 goto fail; 5555 } 5556 break; 5557 default: 5558 if (ioe->rs_num < 0 || ioe->rs_num >= 5559 PF_RULESET_MAX) { 5560 PF_RULES_WUNLOCK(); 5561 free(ioes, M_PF); 5562 error = EINVAL; 5563 goto fail; 5564 } 5565 rs = pf_find_kruleset(ioe->anchor); 5566 if (rs == NULL || 5567 !rs->rules[ioe->rs_num].inactive.open || 5568 rs->rules[ioe->rs_num].inactive.ticket != 5569 ioe->ticket) { 5570 PF_RULES_WUNLOCK(); 5571 free(ioes, M_PF); 5572 error = EBUSY; 5573 goto fail; 5574 } 5575 break; 5576 } 5577 } 5578 /* Now do the commit - no errors should happen here. */ 5579 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { 5580 switch (ioe->rs_num) { 5581 case PF_RULESET_ETH: 5582 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) { 5583 PF_RULES_WUNLOCK(); 5584 free(ioes, M_PF); 5585 goto fail; /* really bad */ 5586 } 5587 break; 5588 #ifdef ALTQ 5589 case PF_RULESET_ALTQ: 5590 if ((error = pf_commit_altq(ioe->ticket))) { 5591 PF_RULES_WUNLOCK(); 5592 free(ioes, M_PF); 5593 goto fail; /* really bad */ 5594 } 5595 break; 5596 #endif /* ALTQ */ 5597 case PF_RULESET_TABLE: 5598 { 5599 struct pfr_table table; 5600 5601 bzero(&table, sizeof(table)); 5602 (void)strlcpy(table.pfrt_anchor, ioe->anchor, 5603 sizeof(table.pfrt_anchor)); 5604 if ((error = pfr_ina_commit(&table, 5605 ioe->ticket, NULL, NULL, 0))) { 5606 PF_RULES_WUNLOCK(); 5607 free(ioes, M_PF); 5608 goto fail; /* really bad */ 5609 } 5610 break; 5611 } 5612 default: 5613 if ((error = pf_commit_rules(ioe->ticket, 5614 ioe->rs_num, ioe->anchor))) { 5615 PF_RULES_WUNLOCK(); 5616 free(ioes, M_PF); 5617 goto fail; /* really bad */ 5618 } 5619 break; 5620 } 5621 } 5622 PF_RULES_WUNLOCK(); 5623 5624 /* Only hook into EtherNet taffic if we've got rules for it. */ 5625 if (! TAILQ_EMPTY(V_pf_keth->active.rules)) 5626 hook_pf_eth(); 5627 else 5628 dehook_pf_eth(); 5629 5630 free(ioes, M_PF); 5631 break; 5632 } 5633 5634 case DIOCGETSRCNODES: { 5635 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; 5636 struct pf_srchash *sh; 5637 struct pf_ksrc_node *n; 5638 struct pf_src_node *p, *pstore; 5639 uint32_t i, nr = 0; 5640 5641 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5642 i++, sh++) { 5643 PF_HASHROW_LOCK(sh); 5644 LIST_FOREACH(n, &sh->nodes, entry) 5645 nr++; 5646 PF_HASHROW_UNLOCK(sh); 5647 } 5648 5649 psn->psn_len = min(psn->psn_len, 5650 sizeof(struct pf_src_node) * nr); 5651 5652 if (psn->psn_len == 0) { 5653 psn->psn_len = sizeof(struct pf_src_node) * nr; 5654 goto fail; 5655 } 5656 5657 nr = 0; 5658 5659 p = pstore = malloc(psn->psn_len, M_PF, M_WAITOK | M_ZERO); 5660 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; 5661 i++, sh++) { 5662 PF_HASHROW_LOCK(sh); 5663 LIST_FOREACH(n, &sh->nodes, entry) { 5664 5665 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) 5666 break; 5667 5668 pf_src_node_copy(n, p); 5669 5670 p++; 5671 nr++; 5672 } 5673 PF_HASHROW_UNLOCK(sh); 5674 } 5675 error = copyout(pstore, psn->psn_src_nodes, 5676 sizeof(struct pf_src_node) * nr); 5677 if (error) { 5678 free(pstore, M_PF); 5679 goto fail; 5680 } 5681 psn->psn_len = sizeof(struct pf_src_node) * nr; 5682 free(pstore, M_PF); 5683 break; 5684 } 5685 5686 case DIOCCLRSRCNODES: { 5687 pf_kill_srcnodes(NULL); 5688 break; 5689 } 5690 5691 case DIOCKILLSRCNODES: 5692 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); 5693 break; 5694 5695 #ifdef COMPAT_FREEBSD13 5696 case DIOCKEEPCOUNTERS_FREEBSD13: 5697 #endif 5698 case DIOCKEEPCOUNTERS: 5699 error = pf_keepcounters((struct pfioc_nv *)addr); 5700 break; 5701 5702 case DIOCGETSYNCOOKIES: 5703 error = pf_get_syncookies((struct pfioc_nv *)addr); 5704 break; 5705 5706 case DIOCSETSYNCOOKIES: 5707 error = pf_set_syncookies((struct pfioc_nv *)addr); 5708 break; 5709 5710 case DIOCSETHOSTID: { 5711 u_int32_t *hostid = (u_int32_t *)addr; 5712 5713 PF_RULES_WLOCK(); 5714 if (*hostid == 0) 5715 V_pf_status.hostid = arc4random(); 5716 else 5717 V_pf_status.hostid = *hostid; 5718 PF_RULES_WUNLOCK(); 5719 break; 5720 } 5721 5722 case DIOCOSFPFLUSH: 5723 PF_RULES_WLOCK(); 5724 pf_osfp_flush(); 5725 PF_RULES_WUNLOCK(); 5726 break; 5727 5728 case DIOCIGETIFACES: { 5729 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5730 struct pfi_kif *ifstore; 5731 size_t bufsiz; 5732 5733 if (io->pfiio_esize != sizeof(struct pfi_kif)) { 5734 error = ENODEV; 5735 goto fail; 5736 } 5737 5738 if (io->pfiio_size < 0 || 5739 io->pfiio_size > pf_ioctl_maxcount || 5740 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { 5741 error = EINVAL; 5742 goto fail; 5743 } 5744 5745 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5746 5747 bufsiz = io->pfiio_size * sizeof(struct pfi_kif); 5748 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif), 5749 M_PF, M_WAITOK | M_ZERO); 5750 5751 PF_RULES_RLOCK(); 5752 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); 5753 PF_RULES_RUNLOCK(); 5754 error = copyout(ifstore, io->pfiio_buffer, bufsiz); 5755 free(ifstore, M_PF); 5756 break; 5757 } 5758 5759 case DIOCSETIFFLAG: { 5760 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5761 5762 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5763 5764 PF_RULES_WLOCK(); 5765 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags); 5766 PF_RULES_WUNLOCK(); 5767 break; 5768 } 5769 5770 case DIOCCLRIFFLAG: { 5771 struct pfioc_iface *io = (struct pfioc_iface *)addr; 5772 5773 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; 5774 5775 PF_RULES_WLOCK(); 5776 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); 5777 PF_RULES_WUNLOCK(); 5778 break; 5779 } 5780 5781 case DIOCSETREASS: { 5782 u_int32_t *reass = (u_int32_t *)addr; 5783 5784 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF); 5785 /* Removal of DF flag without reassembly enabled is not a 5786 * valid combination. Disable reassembly in such case. */ 5787 if (!(V_pf_status.reass & PF_REASS_ENABLED)) 5788 V_pf_status.reass = 0; 5789 break; 5790 } 5791 5792 default: 5793 error = ENODEV; 5794 break; 5795 } 5796 fail: 5797 CURVNET_RESTORE(); 5798 5799 #undef ERROUT_IOCTL 5800 5801 return (error); 5802 } 5803 5804 void 5805 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version) 5806 { 5807 const char *tagname; 5808 bzero(sp, sizeof(union pfsync_state_union)); 5809 5810 /* copy from state key */ 5811 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5812 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5813 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5814 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5815 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5816 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5817 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5818 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5819 5820 /* copy from state */ 5821 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname)); 5822 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr)); 5823 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000)); 5824 sp->pfs_1301.expire = pf_state_expires(st); 5825 if (sp->pfs_1301.expire <= time_uptime) 5826 sp->pfs_1301.expire = htonl(0); 5827 else 5828 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime); 5829 5830 switch (msg_version) { 5831 case PFSYNC_MSG_VERSION_1301: 5832 sp->pfs_1301.state_flags = st->state_flags; 5833 sp->pfs_1301.direction = st->direction; 5834 sp->pfs_1301.log = st->act.log; 5835 sp->pfs_1301.timeout = st->timeout; 5836 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto; 5837 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af; 5838 /* 5839 * XXX Why do we bother pfsyncing source node information if source 5840 * nodes are not synced? Showing users that there is source tracking 5841 * when there is none seems useless. 5842 */ 5843 if (st->sns[PF_SN_LIMIT] != NULL) 5844 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE; 5845 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE]) 5846 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5847 break; 5848 case PFSYNC_MSG_VERSION_1400: 5849 sp->pfs_1400.state_flags = htons(st->state_flags); 5850 sp->pfs_1400.direction = st->direction; 5851 sp->pfs_1400.log = st->act.log; 5852 sp->pfs_1400.timeout = st->timeout; 5853 sp->pfs_1400.proto = st->key[PF_SK_WIRE]->proto; 5854 sp->pfs_1400.af = st->key[PF_SK_WIRE]->af; 5855 sp->pfs_1400.qid = htons(st->act.qid); 5856 sp->pfs_1400.pqid = htons(st->act.pqid); 5857 sp->pfs_1400.dnpipe = htons(st->act.dnpipe); 5858 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe); 5859 sp->pfs_1400.rtableid = htonl(st->act.rtableid); 5860 sp->pfs_1400.min_ttl = st->act.min_ttl; 5861 sp->pfs_1400.set_tos = st->act.set_tos; 5862 sp->pfs_1400.max_mss = htons(st->act.max_mss); 5863 sp->pfs_1400.set_prio[0] = st->act.set_prio[0]; 5864 sp->pfs_1400.set_prio[1] = st->act.set_prio[1]; 5865 sp->pfs_1400.rt = st->act.rt; 5866 if (st->act.rt_kif) 5867 strlcpy(sp->pfs_1400.rt_ifname, 5868 st->act.rt_kif->pfik_name, 5869 sizeof(sp->pfs_1400.rt_ifname)); 5870 /* 5871 * XXX Why do we bother pfsyncing source node information if source 5872 * nodes are not synced? Showing users that there is source tracking 5873 * when there is none seems useless. 5874 */ 5875 if (st->sns[PF_SN_LIMIT] != NULL) 5876 sp->pfs_1400.sync_flags |= PFSYNC_FLAG_SRCNODE; 5877 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE]) 5878 sp->pfs_1400.sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5879 break; 5880 case PFSYNC_MSG_VERSION_1500: 5881 sp->pfs_1500.state_flags = htons(st->state_flags); 5882 sp->pfs_1500.direction = st->direction; 5883 sp->pfs_1500.log = st->act.log; 5884 sp->pfs_1500.timeout = st->timeout; 5885 sp->pfs_1500.wire_proto = st->key[PF_SK_WIRE]->proto; 5886 sp->pfs_1500.wire_af = st->key[PF_SK_WIRE]->af; 5887 sp->pfs_1500.stack_proto = st->key[PF_SK_STACK]->proto; 5888 sp->pfs_1500.stack_af = st->key[PF_SK_STACK]->af; 5889 sp->pfs_1500.qid = htons(st->act.qid); 5890 sp->pfs_1500.pqid = htons(st->act.pqid); 5891 sp->pfs_1500.dnpipe = htons(st->act.dnpipe); 5892 sp->pfs_1500.dnrpipe = htons(st->act.dnrpipe); 5893 sp->pfs_1500.rtableid = htonl(st->act.rtableid); 5894 sp->pfs_1500.min_ttl = st->act.min_ttl; 5895 sp->pfs_1500.set_tos = st->act.set_tos; 5896 sp->pfs_1500.max_mss = htons(st->act.max_mss); 5897 sp->pfs_1500.set_prio[0] = st->act.set_prio[0]; 5898 sp->pfs_1500.set_prio[1] = st->act.set_prio[1]; 5899 sp->pfs_1500.rt = st->act.rt; 5900 sp->pfs_1500.rt_af = st->act.rt_af; 5901 if (st->act.rt_kif) 5902 strlcpy(sp->pfs_1500.rt_ifname, 5903 st->act.rt_kif->pfik_name, 5904 sizeof(sp->pfs_1500.rt_ifname)); 5905 strlcpy(sp->pfs_1500.orig_ifname, 5906 st->orig_kif->pfik_name, 5907 sizeof(sp->pfs_1500.orig_ifname)); 5908 if ((tagname = pf_tag2tagname(st->tag)) != NULL) 5909 strlcpy(sp->pfs_1500.tagname, tagname, 5910 sizeof(sp->pfs_1500.tagname)); 5911 break; 5912 default: 5913 panic("%s: Unsupported pfsync_msg_version %d", 5914 __func__, msg_version); 5915 } 5916 5917 sp->pfs_1301.id = st->id; 5918 sp->pfs_1301.creatorid = st->creatorid; 5919 pf_state_peer_hton(&st->src, &sp->pfs_1301.src); 5920 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst); 5921 5922 if (st->rule == NULL) 5923 sp->pfs_1301.rule = htonl(-1); 5924 else 5925 sp->pfs_1301.rule = htonl(st->rule->nr); 5926 if (st->anchor == NULL) 5927 sp->pfs_1301.anchor = htonl(-1); 5928 else 5929 sp->pfs_1301.anchor = htonl(st->anchor->nr); 5930 if (st->nat_rule == NULL) 5931 sp->pfs_1301.nat_rule = htonl(-1); 5932 else 5933 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr); 5934 5935 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]); 5936 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]); 5937 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]); 5938 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]); 5939 } 5940 5941 void 5942 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st) 5943 { 5944 bzero(sp, sizeof(*sp)); 5945 5946 sp->version = PF_STATE_VERSION; 5947 5948 /* copy from state key */ 5949 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; 5950 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; 5951 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; 5952 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; 5953 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; 5954 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; 5955 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; 5956 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; 5957 sp->proto = st->key[PF_SK_WIRE]->proto; 5958 sp->af = st->key[PF_SK_WIRE]->af; 5959 5960 /* copy from state */ 5961 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); 5962 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name, 5963 sizeof(sp->orig_ifname)); 5964 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr)); 5965 sp->creation = htonl(time_uptime - (st->creation / 1000)); 5966 sp->expire = pf_state_expires(st); 5967 if (sp->expire <= time_uptime) 5968 sp->expire = htonl(0); 5969 else 5970 sp->expire = htonl(sp->expire - time_uptime); 5971 5972 sp->direction = st->direction; 5973 sp->log = st->act.log; 5974 sp->timeout = st->timeout; 5975 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */ 5976 sp->state_flags_compat = st->state_flags; 5977 sp->state_flags = htons(st->state_flags); 5978 if (st->sns[PF_SN_LIMIT] != NULL) 5979 sp->sync_flags |= PFSYNC_FLAG_SRCNODE; 5980 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL) 5981 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; 5982 sp->id = st->id; 5983 sp->creatorid = st->creatorid; 5984 pf_state_peer_hton(&st->src, &sp->src); 5985 pf_state_peer_hton(&st->dst, &sp->dst); 5986 5987 if (st->rule == NULL) 5988 sp->rule = htonl(-1); 5989 else 5990 sp->rule = htonl(st->rule->nr); 5991 if (st->anchor == NULL) 5992 sp->anchor = htonl(-1); 5993 else 5994 sp->anchor = htonl(st->anchor->nr); 5995 if (st->nat_rule == NULL) 5996 sp->nat_rule = htonl(-1); 5997 else 5998 sp->nat_rule = htonl(st->nat_rule->nr); 5999 6000 sp->packets[0] = st->packets[0]; 6001 sp->packets[1] = st->packets[1]; 6002 sp->bytes[0] = st->bytes[0]; 6003 sp->bytes[1] = st->bytes[1]; 6004 6005 sp->qid = htons(st->act.qid); 6006 sp->pqid = htons(st->act.pqid); 6007 sp->dnpipe = htons(st->act.dnpipe); 6008 sp->dnrpipe = htons(st->act.dnrpipe); 6009 sp->rtableid = htonl(st->act.rtableid); 6010 sp->min_ttl = st->act.min_ttl; 6011 sp->set_tos = st->act.set_tos; 6012 sp->max_mss = htons(st->act.max_mss); 6013 sp->rt = st->act.rt; 6014 if (st->act.rt_kif) 6015 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name, 6016 sizeof(sp->rt_ifname)); 6017 sp->set_prio[0] = st->act.set_prio[0]; 6018 sp->set_prio[1] = st->act.set_prio[1]; 6019 6020 } 6021 6022 static void 6023 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 6024 { 6025 struct pfr_ktable *kt; 6026 6027 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); 6028 6029 kt = aw->p.tbl; 6030 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 6031 kt = kt->pfrkt_root; 6032 aw->p.tbl = NULL; 6033 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 6034 kt->pfrkt_cnt : -1; 6035 } 6036 6037 static int 6038 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters, 6039 size_t number, char **names) 6040 { 6041 nvlist_t *nvc; 6042 6043 nvc = nvlist_create(0); 6044 if (nvc == NULL) 6045 return (ENOMEM); 6046 6047 for (int i = 0; i < number; i++) { 6048 nvlist_append_number_array(nvc, "counters", 6049 counter_u64_fetch(counters[i])); 6050 nvlist_append_string_array(nvc, "names", 6051 names[i]); 6052 nvlist_append_number_array(nvc, "ids", 6053 i); 6054 } 6055 nvlist_add_nvlist(nvl, name, nvc); 6056 nvlist_destroy(nvc); 6057 6058 return (0); 6059 } 6060 6061 static int 6062 pf_getstatus(struct pfioc_nv *nv) 6063 { 6064 nvlist_t *nvl = NULL, *nvc = NULL; 6065 void *nvlpacked = NULL; 6066 int error; 6067 struct pf_status s; 6068 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES; 6069 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES; 6070 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES; 6071 time_t since; 6072 6073 PF_RULES_RLOCK_TRACKER; 6074 6075 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6076 6077 PF_RULES_RLOCK(); 6078 6079 nvl = nvlist_create(0); 6080 if (nvl == NULL) 6081 ERROUT(ENOMEM); 6082 6083 since = time_second - (time_uptime - V_pf_status.since); 6084 6085 nvlist_add_bool(nvl, "running", V_pf_status.running); 6086 nvlist_add_number(nvl, "since", since); 6087 nvlist_add_number(nvl, "debug", V_pf_status.debug); 6088 nvlist_add_number(nvl, "hostid", V_pf_status.hostid); 6089 nvlist_add_number(nvl, "states", V_pf_status.states); 6090 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes); 6091 nvlist_add_number(nvl, "reass", V_pf_status.reass); 6092 nvlist_add_bool(nvl, "syncookies_active", 6093 V_pf_status.syncookies_active); 6094 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen); 6095 6096 /* counters */ 6097 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters, 6098 PFRES_MAX, pf_reasons); 6099 if (error != 0) 6100 ERROUT(error); 6101 6102 /* lcounters */ 6103 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters, 6104 KLCNT_MAX, pf_lcounter); 6105 if (error != 0) 6106 ERROUT(error); 6107 6108 /* fcounters */ 6109 nvc = nvlist_create(0); 6110 if (nvc == NULL) 6111 ERROUT(ENOMEM); 6112 6113 for (int i = 0; i < FCNT_MAX; i++) { 6114 nvlist_append_number_array(nvc, "counters", 6115 pf_counter_u64_fetch(&V_pf_status.fcounters[i])); 6116 nvlist_append_string_array(nvc, "names", 6117 pf_fcounter[i]); 6118 nvlist_append_number_array(nvc, "ids", 6119 i); 6120 } 6121 nvlist_add_nvlist(nvl, "fcounters", nvc); 6122 nvlist_destroy(nvc); 6123 nvc = NULL; 6124 6125 /* scounters */ 6126 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters, 6127 SCNT_MAX, pf_fcounter); 6128 if (error != 0) 6129 ERROUT(error); 6130 6131 nvlist_add_string(nvl, "ifname", V_pf_status.ifname); 6132 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum, 6133 PF_MD5_DIGEST_LENGTH); 6134 6135 pfi_update_status(V_pf_status.ifname, &s); 6136 6137 /* pcounters / bcounters */ 6138 for (int i = 0; i < 2; i++) { 6139 for (int j = 0; j < 2; j++) { 6140 for (int k = 0; k < 2; k++) { 6141 nvlist_append_number_array(nvl, "pcounters", 6142 s.pcounters[i][j][k]); 6143 } 6144 nvlist_append_number_array(nvl, "bcounters", 6145 s.bcounters[i][j]); 6146 } 6147 } 6148 6149 nvlpacked = nvlist_pack(nvl, &nv->len); 6150 if (nvlpacked == NULL) 6151 ERROUT(ENOMEM); 6152 6153 if (nv->size == 0) 6154 ERROUT(0); 6155 else if (nv->size < nv->len) 6156 ERROUT(ENOSPC); 6157 6158 PF_RULES_RUNLOCK(); 6159 error = copyout(nvlpacked, nv->data, nv->len); 6160 goto done; 6161 6162 #undef ERROUT 6163 errout: 6164 PF_RULES_RUNLOCK(); 6165 done: 6166 free(nvlpacked, M_NVLIST); 6167 nvlist_destroy(nvc); 6168 nvlist_destroy(nvl); 6169 6170 return (error); 6171 } 6172 6173 /* 6174 * XXX - Check for version mismatch!!! 6175 */ 6176 static void 6177 pf_clear_all_states(void) 6178 { 6179 struct epoch_tracker et; 6180 struct pf_kstate *s; 6181 u_int i; 6182 6183 NET_EPOCH_ENTER(et); 6184 for (i = 0; i <= V_pf_hashmask; i++) { 6185 struct pf_idhash *ih = &V_pf_idhash[i]; 6186 relock: 6187 PF_HASHROW_LOCK(ih); 6188 LIST_FOREACH(s, &ih->states, entry) { 6189 s->timeout = PFTM_PURGE; 6190 /* Don't send out individual delete messages. */ 6191 s->state_flags |= PFSTATE_NOSYNC; 6192 pf_remove_state(s); 6193 goto relock; 6194 } 6195 PF_HASHROW_UNLOCK(ih); 6196 } 6197 NET_EPOCH_EXIT(et); 6198 } 6199 6200 static int 6201 pf_clear_tables(void) 6202 { 6203 struct pfioc_table io; 6204 int error; 6205 6206 bzero(&io, sizeof(io)); 6207 io.pfrio_flags |= PFR_FLAG_ALLRSETS; 6208 6209 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, 6210 io.pfrio_flags); 6211 6212 return (error); 6213 } 6214 6215 static void 6216 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) 6217 { 6218 struct pf_ksrc_node_list kill; 6219 u_int killed; 6220 6221 LIST_INIT(&kill); 6222 for (int i = 0; i <= V_pf_srchashmask; i++) { 6223 struct pf_srchash *sh = &V_pf_srchash[i]; 6224 struct pf_ksrc_node *sn, *tmp; 6225 6226 PF_HASHROW_LOCK(sh); 6227 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) 6228 if (psnk == NULL || 6229 (pf_match_addr(psnk->psnk_src.neg, 6230 &psnk->psnk_src.addr.v.a.addr, 6231 &psnk->psnk_src.addr.v.a.mask, 6232 &sn->addr, sn->af) && 6233 pf_match_addr(psnk->psnk_dst.neg, 6234 &psnk->psnk_dst.addr.v.a.addr, 6235 &psnk->psnk_dst.addr.v.a.mask, 6236 &sn->raddr, sn->af))) { 6237 pf_unlink_src_node(sn); 6238 LIST_INSERT_HEAD(&kill, sn, entry); 6239 sn->expire = 1; 6240 } 6241 PF_HASHROW_UNLOCK(sh); 6242 } 6243 6244 for (int i = 0; i <= V_pf_hashmask; i++) { 6245 struct pf_idhash *ih = &V_pf_idhash[i]; 6246 struct pf_kstate *s; 6247 6248 PF_HASHROW_LOCK(ih); 6249 LIST_FOREACH(s, &ih->states, entry) { 6250 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; 6251 sn_type++) { 6252 if (s->sns[sn_type] && 6253 s->sns[sn_type]->expire == 1) { 6254 s->sns[sn_type] = NULL; 6255 } 6256 } 6257 } 6258 PF_HASHROW_UNLOCK(ih); 6259 } 6260 6261 killed = pf_free_src_nodes(&kill); 6262 6263 if (psnk != NULL) 6264 psnk->psnk_killed = killed; 6265 } 6266 6267 static int 6268 pf_keepcounters(struct pfioc_nv *nv) 6269 { 6270 nvlist_t *nvl = NULL; 6271 void *nvlpacked = NULL; 6272 int error = 0; 6273 6274 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6275 6276 if (nv->len > pf_ioctl_maxcount) 6277 ERROUT(ENOMEM); 6278 6279 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6280 error = copyin(nv->data, nvlpacked, nv->len); 6281 if (error) 6282 ERROUT(error); 6283 6284 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6285 if (nvl == NULL) 6286 ERROUT(EBADMSG); 6287 6288 if (! nvlist_exists_bool(nvl, "keep_counters")) 6289 ERROUT(EBADMSG); 6290 6291 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters"); 6292 6293 on_error: 6294 nvlist_destroy(nvl); 6295 free(nvlpacked, M_NVLIST); 6296 return (error); 6297 } 6298 6299 unsigned int 6300 pf_clear_states(const struct pf_kstate_kill *kill) 6301 { 6302 struct pf_state_key_cmp match_key; 6303 struct pf_kstate *s; 6304 struct pfi_kkif *kif; 6305 int idx; 6306 unsigned int killed = 0, dir; 6307 6308 NET_EPOCH_ASSERT(); 6309 6310 for (unsigned int i = 0; i <= V_pf_hashmask; i++) { 6311 struct pf_idhash *ih = &V_pf_idhash[i]; 6312 6313 relock_DIOCCLRSTATES: 6314 PF_HASHROW_LOCK(ih); 6315 LIST_FOREACH(s, &ih->states, entry) { 6316 /* For floating states look at the original kif. */ 6317 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif; 6318 6319 if (kill->psk_ifname[0] && 6320 strcmp(kill->psk_ifname, 6321 kif->pfik_name)) 6322 continue; 6323 6324 if (kill->psk_kill_match) { 6325 bzero(&match_key, sizeof(match_key)); 6326 6327 if (s->direction == PF_OUT) { 6328 dir = PF_IN; 6329 idx = PF_SK_STACK; 6330 } else { 6331 dir = PF_OUT; 6332 idx = PF_SK_WIRE; 6333 } 6334 6335 match_key.af = s->key[idx]->af; 6336 match_key.proto = s->key[idx]->proto; 6337 pf_addrcpy(&match_key.addr[0], 6338 &s->key[idx]->addr[1], match_key.af); 6339 match_key.port[0] = s->key[idx]->port[1]; 6340 pf_addrcpy(&match_key.addr[1], 6341 &s->key[idx]->addr[0], match_key.af); 6342 match_key.port[1] = s->key[idx]->port[0]; 6343 } 6344 6345 /* 6346 * Don't send out individual 6347 * delete messages. 6348 */ 6349 s->state_flags |= PFSTATE_NOSYNC; 6350 pf_remove_state(s); 6351 killed++; 6352 6353 if (kill->psk_kill_match) 6354 killed += pf_kill_matching_state(&match_key, 6355 dir); 6356 6357 goto relock_DIOCCLRSTATES; 6358 } 6359 PF_HASHROW_UNLOCK(ih); 6360 } 6361 6362 if (V_pfsync_clear_states_ptr != NULL) 6363 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname); 6364 6365 return (killed); 6366 } 6367 6368 void 6369 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed) 6370 { 6371 struct pf_kstate *s; 6372 6373 NET_EPOCH_ASSERT(); 6374 if (kill->psk_pfcmp.id) { 6375 if (kill->psk_pfcmp.creatorid == 0) 6376 kill->psk_pfcmp.creatorid = V_pf_status.hostid; 6377 if ((s = pf_find_state_byid(kill->psk_pfcmp.id, 6378 kill->psk_pfcmp.creatorid))) { 6379 pf_remove_state(s); 6380 *killed = 1; 6381 } 6382 return; 6383 } 6384 6385 for (unsigned int i = 0; i <= V_pf_hashmask; i++) 6386 *killed += pf_killstates_row(kill, &V_pf_idhash[i]); 6387 } 6388 6389 static int 6390 pf_killstates_nv(struct pfioc_nv *nv) 6391 { 6392 struct pf_kstate_kill kill; 6393 struct epoch_tracker et; 6394 nvlist_t *nvl = NULL; 6395 void *nvlpacked = NULL; 6396 int error = 0; 6397 unsigned int killed = 0; 6398 6399 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6400 6401 if (nv->len > pf_ioctl_maxcount) 6402 ERROUT(ENOMEM); 6403 6404 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6405 error = copyin(nv->data, nvlpacked, nv->len); 6406 if (error) 6407 ERROUT(error); 6408 6409 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6410 if (nvl == NULL) 6411 ERROUT(EBADMSG); 6412 6413 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6414 if (error) 6415 ERROUT(error); 6416 6417 NET_EPOCH_ENTER(et); 6418 pf_killstates(&kill, &killed); 6419 NET_EPOCH_EXIT(et); 6420 6421 free(nvlpacked, M_NVLIST); 6422 nvlpacked = NULL; 6423 nvlist_destroy(nvl); 6424 nvl = nvlist_create(0); 6425 if (nvl == NULL) 6426 ERROUT(ENOMEM); 6427 6428 nvlist_add_number(nvl, "killed", killed); 6429 6430 nvlpacked = nvlist_pack(nvl, &nv->len); 6431 if (nvlpacked == NULL) 6432 ERROUT(ENOMEM); 6433 6434 if (nv->size == 0) 6435 ERROUT(0); 6436 else if (nv->size < nv->len) 6437 ERROUT(ENOSPC); 6438 6439 error = copyout(nvlpacked, nv->data, nv->len); 6440 6441 on_error: 6442 nvlist_destroy(nvl); 6443 free(nvlpacked, M_NVLIST); 6444 return (error); 6445 } 6446 6447 static int 6448 pf_clearstates_nv(struct pfioc_nv *nv) 6449 { 6450 struct pf_kstate_kill kill; 6451 struct epoch_tracker et; 6452 nvlist_t *nvl = NULL; 6453 void *nvlpacked = NULL; 6454 int error = 0; 6455 unsigned int killed; 6456 6457 #define ERROUT(x) ERROUT_FUNCTION(on_error, x) 6458 6459 if (nv->len > pf_ioctl_maxcount) 6460 ERROUT(ENOMEM); 6461 6462 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6463 error = copyin(nv->data, nvlpacked, nv->len); 6464 if (error) 6465 ERROUT(error); 6466 6467 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6468 if (nvl == NULL) 6469 ERROUT(EBADMSG); 6470 6471 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill); 6472 if (error) 6473 ERROUT(error); 6474 6475 NET_EPOCH_ENTER(et); 6476 killed = pf_clear_states(&kill); 6477 NET_EPOCH_EXIT(et); 6478 6479 free(nvlpacked, M_NVLIST); 6480 nvlpacked = NULL; 6481 nvlist_destroy(nvl); 6482 nvl = nvlist_create(0); 6483 if (nvl == NULL) 6484 ERROUT(ENOMEM); 6485 6486 nvlist_add_number(nvl, "killed", killed); 6487 6488 nvlpacked = nvlist_pack(nvl, &nv->len); 6489 if (nvlpacked == NULL) 6490 ERROUT(ENOMEM); 6491 6492 if (nv->size == 0) 6493 ERROUT(0); 6494 else if (nv->size < nv->len) 6495 ERROUT(ENOSPC); 6496 6497 error = copyout(nvlpacked, nv->data, nv->len); 6498 6499 #undef ERROUT 6500 on_error: 6501 nvlist_destroy(nvl); 6502 free(nvlpacked, M_NVLIST); 6503 return (error); 6504 } 6505 6506 static int 6507 pf_getstate(struct pfioc_nv *nv) 6508 { 6509 nvlist_t *nvl = NULL, *nvls; 6510 void *nvlpacked = NULL; 6511 struct pf_kstate *s = NULL; 6512 int error = 0; 6513 uint64_t id, creatorid; 6514 6515 #define ERROUT(x) ERROUT_FUNCTION(errout, x) 6516 6517 if (nv->len > pf_ioctl_maxcount) 6518 ERROUT(ENOMEM); 6519 6520 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK); 6521 error = copyin(nv->data, nvlpacked, nv->len); 6522 if (error) 6523 ERROUT(error); 6524 6525 nvl = nvlist_unpack(nvlpacked, nv->len, 0); 6526 if (nvl == NULL) 6527 ERROUT(EBADMSG); 6528 6529 PFNV_CHK(pf_nvuint64(nvl, "id", &id)); 6530 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid)); 6531 6532 s = pf_find_state_byid(id, creatorid); 6533 if (s == NULL) 6534 ERROUT(ENOENT); 6535 6536 free(nvlpacked, M_NVLIST); 6537 nvlpacked = NULL; 6538 nvlist_destroy(nvl); 6539 nvl = nvlist_create(0); 6540 if (nvl == NULL) 6541 ERROUT(ENOMEM); 6542 6543 nvls = pf_state_to_nvstate(s); 6544 if (nvls == NULL) 6545 ERROUT(ENOMEM); 6546 6547 nvlist_add_nvlist(nvl, "state", nvls); 6548 nvlist_destroy(nvls); 6549 6550 nvlpacked = nvlist_pack(nvl, &nv->len); 6551 if (nvlpacked == NULL) 6552 ERROUT(ENOMEM); 6553 6554 if (nv->size == 0) 6555 ERROUT(0); 6556 else if (nv->size < nv->len) 6557 ERROUT(ENOSPC); 6558 6559 error = copyout(nvlpacked, nv->data, nv->len); 6560 6561 #undef ERROUT 6562 errout: 6563 if (s != NULL) 6564 PF_STATE_UNLOCK(s); 6565 free(nvlpacked, M_NVLIST); 6566 nvlist_destroy(nvl); 6567 return (error); 6568 } 6569 6570 /* 6571 * XXX - Check for version mismatch!!! 6572 */ 6573 6574 /* 6575 * Duplicate pfctl -Fa operation to get rid of as much as we can. 6576 */ 6577 static int 6578 shutdown_pf(void) 6579 { 6580 int error = 0; 6581 u_int32_t t[5]; 6582 char nn = '\0'; 6583 struct pf_kanchor *anchor, *tmp_anchor; 6584 struct pf_keth_anchor *eth_anchor, *tmp_eth_anchor; 6585 int rs_num; 6586 6587 do { 6588 /* Unlink rules of all user defined anchors */ 6589 RB_FOREACH_SAFE(anchor, pf_kanchor_global, &V_pf_anchors, 6590 tmp_anchor) { 6591 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6592 if ((error = pf_begin_rules(&t[rs_num], rs_num, 6593 anchor->path)) != 0) { 6594 DPFPRINTF(PF_DEBUG_MISC, "%s: " 6595 "anchor.path=%s rs_num=%d", 6596 __func__, anchor->path, rs_num); 6597 goto error; /* XXX: rollback? */ 6598 } 6599 } 6600 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { 6601 error = pf_commit_rules(t[rs_num], rs_num, 6602 anchor->path); 6603 MPASS(error == 0); 6604 } 6605 } 6606 6607 /* Unlink rules of all user defined ether anchors */ 6608 RB_FOREACH_SAFE(eth_anchor, pf_keth_anchor_global, 6609 &V_pf_keth_anchors, tmp_eth_anchor) { 6610 if ((error = pf_begin_eth(&t[0], eth_anchor->path)) 6611 != 0) { 6612 DPFPRINTF(PF_DEBUG_MISC, "%s: eth " 6613 "anchor.path=%s", __func__, 6614 eth_anchor->path); 6615 goto error; 6616 } 6617 error = pf_commit_eth(t[0], eth_anchor->path); 6618 MPASS(error == 0); 6619 } 6620 6621 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) 6622 != 0) { 6623 DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__); 6624 break; 6625 } 6626 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) 6627 != 0) { 6628 DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__); 6629 break; /* XXX: rollback? */ 6630 } 6631 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) 6632 != 0) { 6633 DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__); 6634 break; /* XXX: rollback? */ 6635 } 6636 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) 6637 != 0) { 6638 DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__); 6639 break; /* XXX: rollback? */ 6640 } 6641 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) 6642 != 0) { 6643 DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__); 6644 break; /* XXX: rollback? */ 6645 } 6646 6647 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn); 6648 MPASS(error == 0); 6649 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn); 6650 MPASS(error == 0); 6651 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn); 6652 MPASS(error == 0); 6653 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn); 6654 MPASS(error == 0); 6655 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn); 6656 MPASS(error == 0); 6657 6658 if ((error = pf_clear_tables()) != 0) 6659 break; 6660 6661 if ((error = pf_begin_eth(&t[0], &nn)) != 0) { 6662 DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__); 6663 break; 6664 } 6665 error = pf_commit_eth(t[0], &nn); 6666 MPASS(error == 0); 6667 6668 #ifdef ALTQ 6669 if ((error = pf_begin_altq(&t[0])) != 0) { 6670 DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__); 6671 break; 6672 } 6673 pf_commit_altq(t[0]); 6674 #endif 6675 6676 pf_clear_all_states(); 6677 6678 pf_kill_srcnodes(NULL); 6679 6680 for (int i = 0; i < PF_RULESET_MAX; i++) { 6681 pf_rule_tree_free(pf_main_ruleset.rules[i].active.tree); 6682 pf_rule_tree_free(pf_main_ruleset.rules[i].inactive.tree); 6683 } 6684 6685 /* status does not use malloced mem so no need to cleanup */ 6686 /* fingerprints and interfaces have their own cleanup code */ 6687 } while(0); 6688 6689 error: 6690 return (error); 6691 } 6692 6693 static pfil_return_t 6694 pf_check_return(int chk, struct mbuf **m) 6695 { 6696 6697 switch (chk) { 6698 case PF_PASS: 6699 if (*m == NULL) 6700 return (PFIL_CONSUMED); 6701 else 6702 return (PFIL_PASS); 6703 break; 6704 default: 6705 if (*m != NULL) { 6706 m_freem(*m); 6707 *m = NULL; 6708 } 6709 return (PFIL_DROPPED); 6710 } 6711 } 6712 6713 static pfil_return_t 6714 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6715 void *ruleset __unused, struct inpcb *inp) 6716 { 6717 int chk; 6718 6719 CURVNET_ASSERT_SET(); 6720 6721 chk = pf_test_eth(PF_IN, flags, ifp, m, inp); 6722 6723 return (pf_check_return(chk, m)); 6724 } 6725 6726 static pfil_return_t 6727 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6728 void *ruleset __unused, struct inpcb *inp) 6729 { 6730 int chk; 6731 6732 CURVNET_ASSERT_SET(); 6733 6734 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp); 6735 6736 return (pf_check_return(chk, m)); 6737 } 6738 6739 #ifdef INET 6740 static pfil_return_t 6741 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags, 6742 void *ruleset __unused, struct inpcb *inp) 6743 { 6744 int chk; 6745 6746 CURVNET_ASSERT_SET(); 6747 6748 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL); 6749 6750 return (pf_check_return(chk, m)); 6751 } 6752 6753 static pfil_return_t 6754 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags, 6755 void *ruleset __unused, struct inpcb *inp) 6756 { 6757 int chk; 6758 6759 CURVNET_ASSERT_SET(); 6760 6761 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL); 6762 6763 return (pf_check_return(chk, m)); 6764 } 6765 #endif 6766 6767 #ifdef INET6 6768 static pfil_return_t 6769 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags, 6770 void *ruleset __unused, struct inpcb *inp) 6771 { 6772 int chk; 6773 6774 CURVNET_ASSERT_SET(); 6775 6776 /* 6777 * In case of loopback traffic IPv6 uses the real interface in 6778 * order to support scoped addresses. In order to support stateful 6779 * filtering we have change this to lo0 as it is the case in IPv4. 6780 */ 6781 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, 6782 m, inp, NULL); 6783 6784 return (pf_check_return(chk, m)); 6785 } 6786 6787 static pfil_return_t 6788 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags, 6789 void *ruleset __unused, struct inpcb *inp) 6790 { 6791 int chk; 6792 6793 CURVNET_ASSERT_SET(); 6794 6795 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL); 6796 6797 return (pf_check_return(chk, m)); 6798 } 6799 #endif /* INET6 */ 6800 6801 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook); 6802 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook); 6803 #define V_pf_eth_in_hook VNET(pf_eth_in_hook) 6804 #define V_pf_eth_out_hook VNET(pf_eth_out_hook) 6805 6806 #ifdef INET 6807 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook); 6808 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook); 6809 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook) 6810 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook) 6811 #endif 6812 #ifdef INET6 6813 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook); 6814 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook); 6815 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook) 6816 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook) 6817 #endif 6818 6819 static void 6820 hook_pf_eth(void) 6821 { 6822 struct pfil_hook_args pha = { 6823 .pa_version = PFIL_VERSION, 6824 .pa_modname = "pf", 6825 .pa_type = PFIL_TYPE_ETHERNET, 6826 }; 6827 struct pfil_link_args pla = { 6828 .pa_version = PFIL_VERSION, 6829 }; 6830 int ret __diagused; 6831 6832 if (atomic_load_bool(&V_pf_pfil_eth_hooked)) 6833 return; 6834 6835 pha.pa_mbuf_chk = pf_eth_check_in; 6836 pha.pa_flags = PFIL_IN; 6837 pha.pa_rulname = "eth-in"; 6838 V_pf_eth_in_hook = pfil_add_hook(&pha); 6839 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6840 pla.pa_head = V_link_pfil_head; 6841 pla.pa_hook = V_pf_eth_in_hook; 6842 ret = pfil_link(&pla); 6843 MPASS(ret == 0); 6844 pha.pa_mbuf_chk = pf_eth_check_out; 6845 pha.pa_flags = PFIL_OUT; 6846 pha.pa_rulname = "eth-out"; 6847 V_pf_eth_out_hook = pfil_add_hook(&pha); 6848 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6849 pla.pa_head = V_link_pfil_head; 6850 pla.pa_hook = V_pf_eth_out_hook; 6851 ret = pfil_link(&pla); 6852 MPASS(ret == 0); 6853 6854 atomic_store_bool(&V_pf_pfil_eth_hooked, true); 6855 } 6856 6857 static void 6858 hook_pf(void) 6859 { 6860 struct pfil_hook_args pha = { 6861 .pa_version = PFIL_VERSION, 6862 .pa_modname = "pf", 6863 }; 6864 struct pfil_link_args pla = { 6865 .pa_version = PFIL_VERSION, 6866 }; 6867 int ret __diagused; 6868 6869 if (atomic_load_bool(&V_pf_pfil_hooked)) 6870 return; 6871 6872 #ifdef INET 6873 pha.pa_type = PFIL_TYPE_IP4; 6874 pha.pa_mbuf_chk = pf_check_in; 6875 pha.pa_flags = PFIL_IN; 6876 pha.pa_rulname = "default-in"; 6877 V_pf_ip4_in_hook = pfil_add_hook(&pha); 6878 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6879 pla.pa_head = V_inet_pfil_head; 6880 pla.pa_hook = V_pf_ip4_in_hook; 6881 ret = pfil_link(&pla); 6882 MPASS(ret == 0); 6883 pha.pa_mbuf_chk = pf_check_out; 6884 pha.pa_flags = PFIL_OUT; 6885 pha.pa_rulname = "default-out"; 6886 V_pf_ip4_out_hook = pfil_add_hook(&pha); 6887 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6888 pla.pa_head = V_inet_pfil_head; 6889 pla.pa_hook = V_pf_ip4_out_hook; 6890 ret = pfil_link(&pla); 6891 MPASS(ret == 0); 6892 if (V_pf_filter_local) { 6893 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6894 pla.pa_head = V_inet_local_pfil_head; 6895 pla.pa_hook = V_pf_ip4_out_hook; 6896 ret = pfil_link(&pla); 6897 MPASS(ret == 0); 6898 } 6899 #endif 6900 #ifdef INET6 6901 pha.pa_type = PFIL_TYPE_IP6; 6902 pha.pa_mbuf_chk = pf_check6_in; 6903 pha.pa_flags = PFIL_IN; 6904 pha.pa_rulname = "default-in6"; 6905 V_pf_ip6_in_hook = pfil_add_hook(&pha); 6906 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR; 6907 pla.pa_head = V_inet6_pfil_head; 6908 pla.pa_hook = V_pf_ip6_in_hook; 6909 ret = pfil_link(&pla); 6910 MPASS(ret == 0); 6911 pha.pa_mbuf_chk = pf_check6_out; 6912 pha.pa_rulname = "default-out6"; 6913 pha.pa_flags = PFIL_OUT; 6914 V_pf_ip6_out_hook = pfil_add_hook(&pha); 6915 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6916 pla.pa_head = V_inet6_pfil_head; 6917 pla.pa_hook = V_pf_ip6_out_hook; 6918 ret = pfil_link(&pla); 6919 MPASS(ret == 0); 6920 if (V_pf_filter_local) { 6921 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR; 6922 pla.pa_head = V_inet6_local_pfil_head; 6923 pla.pa_hook = V_pf_ip6_out_hook; 6924 ret = pfil_link(&pla); 6925 MPASS(ret == 0); 6926 } 6927 #endif 6928 6929 atomic_store_bool(&V_pf_pfil_hooked, true); 6930 } 6931 6932 static void 6933 dehook_pf_eth(void) 6934 { 6935 6936 if (!atomic_load_bool(&V_pf_pfil_eth_hooked)) 6937 return; 6938 6939 pfil_remove_hook(V_pf_eth_in_hook); 6940 pfil_remove_hook(V_pf_eth_out_hook); 6941 6942 atomic_store_bool(&V_pf_pfil_eth_hooked, false); 6943 } 6944 6945 static void 6946 dehook_pf(void) 6947 { 6948 6949 if (!atomic_load_bool(&V_pf_pfil_hooked)) 6950 return; 6951 6952 #ifdef INET 6953 pfil_remove_hook(V_pf_ip4_in_hook); 6954 pfil_remove_hook(V_pf_ip4_out_hook); 6955 #endif 6956 #ifdef INET6 6957 pfil_remove_hook(V_pf_ip6_in_hook); 6958 pfil_remove_hook(V_pf_ip6_out_hook); 6959 #endif 6960 6961 atomic_store_bool(&V_pf_pfil_hooked, false); 6962 } 6963 6964 static void 6965 pf_load_vnet(void) 6966 { 6967 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname), 6968 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 6969 6970 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE); 6971 rm_init_flags(&V_pf_tags_lock, "pf tags and queues", RM_RECURSE); 6972 sx_init(&V_pf_ioctl_lock, "pf ioctl"); 6973 6974 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize, 6975 PF_RULE_TAG_HASH_SIZE_DEFAULT); 6976 #ifdef ALTQ 6977 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize, 6978 PF_QUEUE_TAG_HASH_SIZE_DEFAULT); 6979 #endif 6980 6981 V_pf_keth = &V_pf_main_keth_anchor.ruleset; 6982 6983 pfattach_vnet(); 6984 V_pf_vnet_active = 1; 6985 } 6986 6987 static int 6988 pf_load(void) 6989 { 6990 int error; 6991 6992 sx_init(&pf_end_lock, "pf end thread"); 6993 6994 pf_mtag_initialize(); 6995 6996 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME); 6997 if (pf_dev == NULL) 6998 return (ENOMEM); 6999 7000 pf_end_threads = 0; 7001 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge"); 7002 if (error != 0) 7003 return (error); 7004 7005 pfi_initialize(); 7006 7007 return (0); 7008 } 7009 7010 static void 7011 pf_unload_vnet(void) 7012 { 7013 int ret __diagused; 7014 7015 V_pf_vnet_active = 0; 7016 V_pf_status.running = 0; 7017 dehook_pf(); 7018 dehook_pf_eth(); 7019 7020 PF_RULES_WLOCK(); 7021 pf_syncookies_cleanup(); 7022 shutdown_pf(); 7023 PF_RULES_WUNLOCK(); 7024 7025 ret = swi_remove(V_pf_swi_cookie); 7026 MPASS(ret == 0); 7027 ret = intr_event_destroy(V_pf_swi_ie); 7028 MPASS(ret == 0); 7029 7030 pf_unload_vnet_purge(); 7031 7032 pf_normalize_cleanup(); 7033 PF_RULES_WLOCK(); 7034 pfi_cleanup_vnet(); 7035 PF_RULES_WUNLOCK(); 7036 pfr_cleanup(); 7037 pf_osfp_flush(); 7038 pf_cleanup(); 7039 if (IS_DEFAULT_VNET(curvnet)) 7040 pf_mtag_cleanup(); 7041 7042 pf_cleanup_tagset(&V_pf_tags); 7043 #ifdef ALTQ 7044 pf_cleanup_tagset(&V_pf_qids); 7045 #endif 7046 uma_zdestroy(V_pf_tag_z); 7047 7048 #ifdef PF_WANT_32_TO_64_COUNTER 7049 PF_RULES_WLOCK(); 7050 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); 7051 7052 MPASS(LIST_EMPTY(&V_pf_allkiflist)); 7053 MPASS(V_pf_allkifcount == 0); 7054 7055 LIST_REMOVE(&V_pf_default_rule, allrulelist); 7056 V_pf_allrulecount--; 7057 LIST_REMOVE(V_pf_rulemarker, allrulelist); 7058 7059 MPASS(LIST_EMPTY(&V_pf_allrulelist)); 7060 MPASS(V_pf_allrulecount == 0); 7061 7062 PF_RULES_WUNLOCK(); 7063 7064 free(V_pf_kifmarker, PFI_MTYPE); 7065 free(V_pf_rulemarker, M_PFRULE); 7066 #endif 7067 7068 /* Free counters last as we updated them during shutdown. */ 7069 pf_counter_u64_deinit(&V_pf_default_rule.evaluations); 7070 for (int i = 0; i < 2; i++) { 7071 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]); 7072 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]); 7073 } 7074 counter_u64_free(V_pf_default_rule.states_cur); 7075 counter_u64_free(V_pf_default_rule.states_tot); 7076 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++) 7077 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]); 7078 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp); 7079 7080 for (int i = 0; i < PFRES_MAX; i++) 7081 counter_u64_free(V_pf_status.counters[i]); 7082 for (int i = 0; i < KLCNT_MAX; i++) 7083 counter_u64_free(V_pf_status.lcounters[i]); 7084 for (int i = 0; i < FCNT_MAX; i++) 7085 pf_counter_u64_deinit(&V_pf_status.fcounters[i]); 7086 for (int i = 0; i < SCNT_MAX; i++) 7087 counter_u64_free(V_pf_status.scounters[i]); 7088 for (int i = 0; i < NCNT_MAX; i++) 7089 counter_u64_free(V_pf_status.ncounters[i]); 7090 7091 rm_destroy(&V_pf_rules_lock); 7092 sx_destroy(&V_pf_ioctl_lock); 7093 } 7094 7095 static void 7096 pf_unload(void *dummy __unused) 7097 { 7098 7099 sx_xlock(&pf_end_lock); 7100 pf_end_threads = 1; 7101 while (pf_end_threads < 2) { 7102 wakeup_one(pf_purge_thread); 7103 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0); 7104 } 7105 sx_xunlock(&pf_end_lock); 7106 7107 pf_nl_unregister(); 7108 7109 if (pf_dev != NULL) 7110 destroy_dev(pf_dev); 7111 7112 pfi_cleanup(); 7113 7114 sx_destroy(&pf_end_lock); 7115 } 7116 7117 static void 7118 vnet_pf_init(void *unused __unused) 7119 { 7120 7121 pf_load_vnet(); 7122 } 7123 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 7124 vnet_pf_init, NULL); 7125 7126 static void 7127 vnet_pf_uninit(const void *unused __unused) 7128 { 7129 7130 pf_unload_vnet(); 7131 } 7132 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL); 7133 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD, 7134 vnet_pf_uninit, NULL); 7135 7136 static int 7137 pf_modevent(module_t mod, int type, void *data) 7138 { 7139 int error = 0; 7140 7141 switch(type) { 7142 case MOD_LOAD: 7143 error = pf_load(); 7144 pf_nl_register(); 7145 break; 7146 case MOD_UNLOAD: 7147 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after 7148 * the vnet_pf_uninit()s */ 7149 break; 7150 default: 7151 error = EINVAL; 7152 break; 7153 } 7154 7155 return (error); 7156 } 7157 7158 static moduledata_t pf_mod = { 7159 "pf", 7160 pf_modevent, 7161 0 7162 }; 7163 7164 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND); 7165 MODULE_DEPEND(pf, netlink, 1, 1, 1); 7166 MODULE_DEPEND(pf, crypto, 1, 1, 1); 7167 MODULE_VERSION(pf, PF_MODVER); 7168