1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 2001 Daniel Hartmeier 5 * Copyright (c) 2003 Cedric Berger 6 * Copyright (c) 2005 Henning Brauer <henning@openbsd.org> 7 * Copyright (c) 2005 Ryan McBride <mcbride@openbsd.org> 8 * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org> 9 * All rights reserved. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 15 * - Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * - Redistributions in binary form must reproduce the above 18 * copyright notice, this list of conditions and the following 19 * disclaimer in the documentation and/or other materials provided 20 * with the distribution. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 27 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 30 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 32 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 33 * POSSIBILITY OF SUCH DAMAGE. 34 * 35 * $OpenBSD: pf_if.c,v 1.54 2008/06/14 16:55:28 mk Exp $ 36 */ 37 38 #include <sys/cdefs.h> 39 #include "opt_inet.h" 40 #include "opt_inet6.h" 41 42 #include <sys/param.h> 43 #include <sys/kernel.h> 44 #include <sys/eventhandler.h> 45 #include <sys/lock.h> 46 #include <sys/mbuf.h> 47 #include <sys/socket.h> 48 49 #include <net/if.h> 50 #include <net/if_var.h> 51 #include <net/if_private.h> 52 #include <net/vnet.h> 53 #include <net/pfvar.h> 54 #include <net/route.h> 55 56 VNET_DEFINE(struct pfi_kkif *, pfi_all); 57 VNET_DEFINE_STATIC(long, pfi_update); 58 #define V_pfi_update VNET(pfi_update) 59 #define PFI_BUFFER_MAX 0x10000 60 61 VNET_DECLARE(int, pf_vnet_active); 62 #define V_pf_vnet_active VNET(pf_vnet_active) 63 64 VNET_DEFINE_STATIC(struct pfr_addr *, pfi_buffer); 65 VNET_DEFINE_STATIC(int, pfi_buffer_cnt); 66 VNET_DEFINE_STATIC(int, pfi_buffer_max); 67 #define V_pfi_buffer VNET(pfi_buffer) 68 #define V_pfi_buffer_cnt VNET(pfi_buffer_cnt) 69 #define V_pfi_buffer_max VNET(pfi_buffer_max) 70 71 #ifdef PF_WANT_32_TO_64_COUNTER 72 VNET_DEFINE(struct allkiflist_head, pf_allkiflist); 73 VNET_DEFINE(size_t, pf_allkifcount); 74 VNET_DEFINE(struct pfi_kkif *, pf_kifmarker); 75 #endif 76 77 eventhandler_tag pfi_attach_cookie; 78 eventhandler_tag pfi_detach_cookie; 79 eventhandler_tag pfi_attach_group_cookie; 80 eventhandler_tag pfi_change_group_cookie; 81 eventhandler_tag pfi_detach_group_cookie; 82 eventhandler_tag pfi_ifaddr_event_cookie; 83 84 static void pfi_attach_ifnet(struct ifnet *, struct pfi_kkif *); 85 static void pfi_attach_ifgroup(struct ifg_group *, struct pfi_kkif *); 86 87 static void pfi_kkif_update(struct pfi_kkif *); 88 static void pfi_dynaddr_update(struct pfi_dynaddr *dyn); 89 static void pfi_table_update(struct pfr_ktable *, struct pfi_kkif *, int, 90 int); 91 static void pfi_instance_add(struct ifnet *, int, int); 92 static void pfi_address_add(struct sockaddr *, int, int); 93 static int pfi_kkif_compare(struct pfi_kkif *, struct pfi_kkif *); 94 static int pfi_skip_if(const char *, struct pfi_kkif *); 95 static int pfi_unmask(void *); 96 static void pfi_attach_ifnet_event(void * __unused, struct ifnet *); 97 static void pfi_detach_ifnet_event(void * __unused, struct ifnet *); 98 static void pfi_attach_group_event(void * __unused, struct ifg_group *); 99 static void pfi_change_group_event(void * __unused, char *); 100 static void pfi_detach_group_event(void * __unused, struct ifg_group *); 101 static void pfi_ifaddr_event(void * __unused, struct ifnet *); 102 103 RB_HEAD(pfi_ifhead, pfi_kkif); 104 static RB_PROTOTYPE(pfi_ifhead, pfi_kkif, pfik_tree, pfi_kkif_compare); 105 static RB_GENERATE(pfi_ifhead, pfi_kkif, pfik_tree, pfi_kkif_compare); 106 VNET_DEFINE_STATIC(struct pfi_ifhead, pfi_ifs); 107 #define V_pfi_ifs VNET(pfi_ifs) 108 109 #define PFI_BUFFER_MAX 0x10000 110 MALLOC_DEFINE(PFI_MTYPE, "pf_ifnet", "pf(4) interface database"); 111 112 LIST_HEAD(pfi_list, pfi_kkif); 113 VNET_DEFINE_STATIC(struct pfi_list, pfi_unlinked_kifs); 114 #define V_pfi_unlinked_kifs VNET(pfi_unlinked_kifs) 115 static struct mtx pfi_unlnkdkifs_mtx; 116 MTX_SYSINIT(pfi_unlnkdkifs_mtx, &pfi_unlnkdkifs_mtx, "pf unlinked interfaces", 117 MTX_DEF); 118 119 void 120 pfi_initialize_vnet(void) 121 { 122 struct pfi_list kifs = LIST_HEAD_INITIALIZER(); 123 struct epoch_tracker et; 124 struct pfi_kkif *kif; 125 struct ifg_group *ifg; 126 struct ifnet *ifp; 127 int nkifs; 128 129 V_pfi_buffer_max = 64; 130 V_pfi_buffer = malloc(V_pfi_buffer_max * sizeof(*V_pfi_buffer), 131 PFI_MTYPE, M_WAITOK); 132 133 nkifs = 1; /* one for V_pfi_all */ 134 IFNET_RLOCK(); 135 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 136 nkifs++; 137 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 138 nkifs++; 139 140 for (int n = 0; n < nkifs; n++) { 141 kif = pf_kkif_create(M_WAITOK); 142 LIST_INSERT_HEAD(&kifs, kif, pfik_list); 143 } 144 145 NET_EPOCH_ENTER(et); 146 PF_RULES_WLOCK(); 147 kif = LIST_FIRST(&kifs); 148 LIST_REMOVE(kif, pfik_list); 149 V_pfi_all = pfi_kkif_attach(kif, IFG_ALL); 150 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) { 151 kif = LIST_FIRST(&kifs); 152 LIST_REMOVE(kif, pfik_list); 153 pfi_attach_ifgroup(ifg, kif); 154 } 155 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 156 kif = LIST_FIRST(&kifs); 157 LIST_REMOVE(kif, pfik_list); 158 pfi_attach_ifnet(ifp, kif); 159 } 160 PF_RULES_WUNLOCK(); 161 NET_EPOCH_EXIT(et); 162 IFNET_RUNLOCK(); 163 164 MPASS(LIST_EMPTY(&kifs)); 165 } 166 167 void 168 pfi_initialize(void) 169 { 170 171 pfi_attach_cookie = EVENTHANDLER_REGISTER(ifnet_arrival_event, 172 pfi_attach_ifnet_event, NULL, EVENTHANDLER_PRI_ANY); 173 pfi_detach_cookie = EVENTHANDLER_REGISTER(ifnet_departure_event, 174 pfi_detach_ifnet_event, NULL, EVENTHANDLER_PRI_ANY); 175 pfi_attach_group_cookie = EVENTHANDLER_REGISTER(group_attach_event, 176 pfi_attach_group_event, NULL, EVENTHANDLER_PRI_ANY); 177 pfi_change_group_cookie = EVENTHANDLER_REGISTER(group_change_event, 178 pfi_change_group_event, NULL, EVENTHANDLER_PRI_ANY); 179 pfi_detach_group_cookie = EVENTHANDLER_REGISTER(group_detach_event, 180 pfi_detach_group_event, NULL, EVENTHANDLER_PRI_ANY); 181 pfi_ifaddr_event_cookie = EVENTHANDLER_REGISTER(ifaddr_event, 182 pfi_ifaddr_event, NULL, EVENTHANDLER_PRI_ANY); 183 } 184 185 void 186 pfi_cleanup_vnet(void) 187 { 188 struct pfi_kkif *kif; 189 190 PF_RULES_WASSERT(); 191 192 V_pfi_all = NULL; 193 while ((kif = RB_MIN(pfi_ifhead, &V_pfi_ifs))) { 194 RB_REMOVE(pfi_ifhead, &V_pfi_ifs, kif); 195 if (kif->pfik_group) 196 kif->pfik_group->ifg_pf_kif = NULL; 197 if (kif->pfik_ifp) { 198 if_rele(kif->pfik_ifp); 199 kif->pfik_ifp->if_pf_kif = NULL; 200 } 201 pf_kkif_free(kif); 202 } 203 204 mtx_lock(&pfi_unlnkdkifs_mtx); 205 while ((kif = LIST_FIRST(&V_pfi_unlinked_kifs))) { 206 LIST_REMOVE(kif, pfik_list); 207 pf_kkif_free(kif); 208 } 209 mtx_unlock(&pfi_unlnkdkifs_mtx); 210 211 free(V_pfi_buffer, PFI_MTYPE); 212 } 213 214 void 215 pfi_cleanup(void) 216 { 217 218 EVENTHANDLER_DEREGISTER(ifnet_arrival_event, pfi_attach_cookie); 219 EVENTHANDLER_DEREGISTER(ifnet_departure_event, pfi_detach_cookie); 220 EVENTHANDLER_DEREGISTER(group_attach_event, pfi_attach_group_cookie); 221 EVENTHANDLER_DEREGISTER(group_change_event, pfi_change_group_cookie); 222 EVENTHANDLER_DEREGISTER(group_detach_event, pfi_detach_group_cookie); 223 EVENTHANDLER_DEREGISTER(ifaddr_event, pfi_ifaddr_event_cookie); 224 } 225 226 struct pfi_kkif* 227 pf_kkif_create(int flags) 228 { 229 struct pfi_kkif *kif; 230 #ifdef PF_WANT_32_TO_64_COUNTER 231 bool wowned; 232 #endif 233 234 kif = malloc(sizeof(*kif), PFI_MTYPE, flags | M_ZERO); 235 if (! kif) 236 return (kif); 237 238 for (int i = 0; i < 2; i++) { 239 for (int j = 0; j < 2; j++) { 240 for (int k = 0; k < 2; k++) { 241 if (pf_counter_u64_init(&kif->pfik_packets[i][j][k], flags) != 0) { 242 pf_kkif_free(kif); 243 return (NULL); 244 } 245 246 if (pf_counter_u64_init(&kif->pfik_bytes[i][j][k], flags) != 0) { 247 pf_kkif_free(kif); 248 return (NULL); 249 } 250 } 251 } 252 } 253 254 #ifdef PF_WANT_32_TO_64_COUNTER 255 wowned = PF_RULES_WOWNED(); 256 if (!wowned) 257 PF_RULES_WLOCK(); 258 LIST_INSERT_HEAD(&V_pf_allkiflist, kif, pfik_allkiflist); 259 V_pf_allkifcount++; 260 if (!wowned) 261 PF_RULES_WUNLOCK(); 262 #endif 263 264 return (kif); 265 } 266 267 void 268 pf_kkif_free(struct pfi_kkif *kif) 269 { 270 #ifdef PF_WANT_32_TO_64_COUNTER 271 bool wowned; 272 #endif 273 274 if (! kif) 275 return; 276 277 #ifdef INVARIANTS 278 if (kif->pfik_ifp) { 279 struct ifnet *ifp = kif->pfik_ifp; 280 MPASS(ifp->if_pf_kif == NULL || ifp->if_pf_kif == kif); 281 } 282 #endif 283 284 #ifdef PF_WANT_32_TO_64_COUNTER 285 wowned = PF_RULES_WOWNED(); 286 if (!wowned) 287 PF_RULES_WLOCK(); 288 LIST_REMOVE(kif, pfik_allkiflist); 289 V_pf_allkifcount--; 290 if (!wowned) 291 PF_RULES_WUNLOCK(); 292 #endif 293 294 for (int i = 0; i < 2; i++) { 295 for (int j = 0; j < 2; j++) { 296 for (int k = 0; k < 2; k++) { 297 pf_counter_u64_deinit(&kif->pfik_packets[i][j][k]); 298 pf_counter_u64_deinit(&kif->pfik_bytes[i][j][k]); 299 } 300 } 301 } 302 303 free(kif, PFI_MTYPE); 304 } 305 306 void 307 pf_kkif_zero(struct pfi_kkif *kif) 308 { 309 310 for (int i = 0; i < 2; i++) { 311 for (int j = 0; j < 2; j++) { 312 for (int k = 0; k < 2; k++) { 313 pf_counter_u64_zero(&kif->pfik_packets[i][j][k]); 314 pf_counter_u64_zero(&kif->pfik_bytes[i][j][k]); 315 } 316 } 317 } 318 kif->pfik_tzero = time_second; 319 } 320 321 struct pfi_kkif * 322 pfi_kkif_find(const char *kif_name) 323 { 324 struct pfi_kif_cmp s; 325 326 PF_RULES_ASSERT(); 327 328 memset(&s, 0, sizeof(s)); 329 strlcpy(s.pfik_name, kif_name, sizeof(s.pfik_name)); 330 331 return (RB_FIND(pfi_ifhead, &V_pfi_ifs, (struct pfi_kkif *)&s)); 332 } 333 334 struct pfi_kkif * 335 pfi_kkif_attach(struct pfi_kkif *kif, const char *kif_name) 336 { 337 struct pfi_kkif *kif1; 338 339 PF_RULES_WASSERT(); 340 KASSERT(kif != NULL, ("%s: null kif", __func__)); 341 342 kif1 = pfi_kkif_find(kif_name); 343 if (kif1 != NULL) { 344 pf_kkif_free(kif); 345 return (kif1); 346 } 347 348 pf_kkif_zero(kif); 349 strlcpy(kif->pfik_name, kif_name, sizeof(kif->pfik_name)); 350 /* 351 * It seems that the value of time_second is in unintialzied state 352 * when pf sets interface statistics clear time in boot phase if pf 353 * was statically linked to kernel. Instead of setting the bogus 354 * time value have pfi_get_ifaces handle this case. In 355 * pfi_get_ifaces it uses time_second if it sees the time is 0. 356 */ 357 kif->pfik_tzero = time_second > 1 ? time_second : 0; 358 TAILQ_INIT(&kif->pfik_dynaddrs); 359 360 RB_INSERT(pfi_ifhead, &V_pfi_ifs, kif); 361 362 return (kif); 363 } 364 365 void 366 pfi_kkif_ref(struct pfi_kkif *kif) 367 { 368 369 PF_RULES_WASSERT(); 370 kif->pfik_rulerefs++; 371 } 372 373 static void 374 pfi_kkif_remove_if_unref(struct pfi_kkif *kif) 375 { 376 377 PF_RULES_WASSERT(); 378 379 if (kif->pfik_rulerefs > 0) 380 return; 381 382 /* kif referencing an existing ifnet or group or holding flags should 383 * exist. */ 384 if (kif->pfik_ifp != NULL || kif->pfik_group != NULL || 385 kif == V_pfi_all || kif->pfik_flags != 0) 386 return; 387 388 /* 389 * We can get here in at least two distinct paths: 390 * - when the struct ifnet is removed, via pfi_detach_ifnet_event() 391 * - when a rule referencing us is removed, via pfi_kkif_unref(). 392 * These two events can race against each other, leading us to free this kif 393 * twice. That leads to a loop in V_pfi_unlinked_kifs, and an eventual 394 * deadlock. 395 * 396 * Avoid this by making sure we only ever insert the kif into 397 * V_pfi_unlinked_kifs once. 398 * If we don't find it in V_pfi_ifs it's already been removed. Check that it 399 * exists in V_pfi_unlinked_kifs. 400 */ 401 if (! RB_FIND(pfi_ifhead, &V_pfi_ifs, kif)) { 402 #ifdef INVARIANTS 403 struct pfi_kkif *tmp; 404 bool found = false; 405 mtx_lock(&pfi_unlnkdkifs_mtx); 406 LIST_FOREACH(tmp, &V_pfi_unlinked_kifs, pfik_list) { 407 if (tmp == kif) { 408 found = true; 409 break; 410 } 411 } 412 mtx_unlock(&pfi_unlnkdkifs_mtx); 413 MPASS(found); 414 #endif 415 return; 416 } 417 RB_REMOVE(pfi_ifhead, &V_pfi_ifs, kif); 418 419 kif->pfik_flags |= PFI_IFLAG_REFS; 420 421 mtx_lock(&pfi_unlnkdkifs_mtx); 422 LIST_INSERT_HEAD(&V_pfi_unlinked_kifs, kif, pfik_list); 423 mtx_unlock(&pfi_unlnkdkifs_mtx); 424 } 425 426 void 427 pfi_kkif_unref(struct pfi_kkif *kif) 428 { 429 430 PF_RULES_WASSERT(); 431 KASSERT(kif->pfik_rulerefs > 0, ("%s: %p has zero refs", __func__, kif)); 432 433 kif->pfik_rulerefs--; 434 435 pfi_kkif_remove_if_unref(kif); 436 } 437 438 void 439 pfi_kkif_purge(void) 440 { 441 struct pfi_kkif *kif, *kif1; 442 443 /* 444 * Do naive mark-and-sweep garbage collecting of old kifs. 445 * Reference flag is raised by pf_purge_expired_states(). 446 */ 447 mtx_lock(&pfi_unlnkdkifs_mtx); 448 LIST_FOREACH_SAFE(kif, &V_pfi_unlinked_kifs, pfik_list, kif1) { 449 if (!(kif->pfik_flags & PFI_IFLAG_REFS)) { 450 LIST_REMOVE(kif, pfik_list); 451 pf_kkif_free(kif); 452 } else 453 kif->pfik_flags &= ~PFI_IFLAG_REFS; 454 } 455 mtx_unlock(&pfi_unlnkdkifs_mtx); 456 } 457 458 int 459 pfi_kkif_match(struct pfi_kkif *rule_kif, struct pfi_kkif *packet_kif) 460 { 461 struct ifg_list *p; 462 463 NET_EPOCH_ASSERT(); 464 465 MPASS(packet_kif != NULL); 466 MPASS(packet_kif->pfik_ifp != NULL); 467 468 if (rule_kif == NULL || rule_kif == packet_kif) 469 return (1); 470 471 if (rule_kif->pfik_group != NULL) { 472 CK_STAILQ_FOREACH(p, &packet_kif->pfik_ifp->if_groups, ifgl_next) 473 if (p->ifgl_group == rule_kif->pfik_group) 474 return (1); 475 } 476 477 return (0); 478 } 479 480 static void 481 pfi_attach_ifnet(struct ifnet *ifp, struct pfi_kkif *kif) 482 { 483 484 PF_RULES_WASSERT(); 485 486 V_pfi_update++; 487 kif = pfi_kkif_attach(kif, ifp->if_xname); 488 if_ref(ifp); 489 kif->pfik_ifp = ifp; 490 ifp->if_pf_kif = kif; 491 pfi_kkif_update(kif); 492 } 493 494 static void 495 pfi_attach_ifgroup(struct ifg_group *ifg, struct pfi_kkif *kif) 496 { 497 498 PF_RULES_WASSERT(); 499 500 V_pfi_update++; 501 kif = pfi_kkif_attach(kif, ifg->ifg_group); 502 kif->pfik_group = ifg; 503 ifg->ifg_pf_kif = kif; 504 } 505 506 int 507 pfi_match_addr(struct pfi_dynaddr *dyn, struct pf_addr *a, sa_family_t af) 508 { 509 switch (af) { 510 #ifdef INET 511 case AF_INET: 512 switch (dyn->pfid_acnt4) { 513 case 0: 514 return (0); 515 case 1: 516 return (PF_MATCHA(0, &dyn->pfid_addr4, 517 &dyn->pfid_mask4, a, AF_INET)); 518 default: 519 return (pfr_match_addr(dyn->pfid_kt, a, AF_INET)); 520 } 521 break; 522 #endif /* INET */ 523 #ifdef INET6 524 case AF_INET6: 525 switch (dyn->pfid_acnt6) { 526 case 0: 527 return (0); 528 case 1: 529 return (PF_MATCHA(0, &dyn->pfid_addr6, 530 &dyn->pfid_mask6, a, AF_INET6)); 531 default: 532 return (pfr_match_addr(dyn->pfid_kt, a, AF_INET6)); 533 } 534 break; 535 #endif /* INET6 */ 536 default: 537 return (0); 538 } 539 } 540 541 int 542 pfi_dynaddr_setup(struct pf_addr_wrap *aw, sa_family_t af) 543 { 544 struct epoch_tracker et; 545 struct pfi_dynaddr *dyn; 546 char tblname[PF_TABLE_NAME_SIZE]; 547 struct pf_kruleset *ruleset = NULL; 548 struct pfi_kkif *kif; 549 int rv = 0; 550 551 PF_RULES_WASSERT(); 552 KASSERT(aw->type == PF_ADDR_DYNIFTL, ("%s: type %u", 553 __func__, aw->type)); 554 KASSERT(aw->p.dyn == NULL, ("%s: dyn is %p", __func__, aw->p.dyn)); 555 556 if ((dyn = malloc(sizeof(*dyn), PFI_MTYPE, M_NOWAIT | M_ZERO)) == NULL) 557 return (ENOMEM); 558 559 if ((kif = pf_kkif_create(M_NOWAIT)) == NULL) { 560 free(dyn, PFI_MTYPE); 561 return (ENOMEM); 562 } 563 564 if (!strcmp(aw->v.ifname, "self")) 565 dyn->pfid_kif = pfi_kkif_attach(kif, IFG_ALL); 566 else 567 dyn->pfid_kif = pfi_kkif_attach(kif, aw->v.ifname); 568 kif = NULL; 569 pfi_kkif_ref(dyn->pfid_kif); 570 571 dyn->pfid_net = pfi_unmask(&aw->v.a.mask); 572 if (af == AF_INET && dyn->pfid_net == 32) 573 dyn->pfid_net = 128; 574 strlcpy(tblname, aw->v.ifname, sizeof(tblname)); 575 if (aw->iflags & PFI_AFLAG_NETWORK) 576 strlcat(tblname, ":network", sizeof(tblname)); 577 if (aw->iflags & PFI_AFLAG_BROADCAST) 578 strlcat(tblname, ":broadcast", sizeof(tblname)); 579 if (aw->iflags & PFI_AFLAG_PEER) 580 strlcat(tblname, ":peer", sizeof(tblname)); 581 if (aw->iflags & PFI_AFLAG_NOALIAS) 582 strlcat(tblname, ":0", sizeof(tblname)); 583 if (dyn->pfid_net != 128) 584 snprintf(tblname + strlen(tblname), 585 sizeof(tblname) - strlen(tblname), "/%d", dyn->pfid_net); 586 if ((ruleset = pf_find_or_create_kruleset(PF_RESERVED_ANCHOR)) == NULL) { 587 rv = ENOMEM; 588 goto _bad; 589 } 590 591 if ((dyn->pfid_kt = pfr_attach_table(ruleset, tblname)) == NULL) { 592 rv = ENOMEM; 593 goto _bad; 594 } 595 596 dyn->pfid_kt->pfrkt_flags |= PFR_TFLAG_ACTIVE; 597 dyn->pfid_iflags = aw->iflags; 598 dyn->pfid_af = af; 599 600 TAILQ_INSERT_TAIL(&dyn->pfid_kif->pfik_dynaddrs, dyn, entry); 601 aw->p.dyn = dyn; 602 NET_EPOCH_ENTER(et); 603 pfi_kkif_update(dyn->pfid_kif); 604 NET_EPOCH_EXIT(et); 605 606 return (0); 607 608 _bad: 609 if (dyn->pfid_kt != NULL) 610 pfr_detach_table(dyn->pfid_kt); 611 if (ruleset != NULL) 612 pf_remove_if_empty_kruleset(ruleset); 613 pfi_kkif_unref(dyn->pfid_kif); 614 free(dyn, PFI_MTYPE); 615 616 return (rv); 617 } 618 619 static void 620 pfi_kkif_update(struct pfi_kkif *kif) 621 { 622 struct ifg_list *ifgl; 623 struct ifg_member *ifgm; 624 struct pfi_dynaddr *p; 625 struct pfi_kkif *tmpkif; 626 627 NET_EPOCH_ASSERT(); 628 PF_RULES_WASSERT(); 629 630 /* update all dynaddr */ 631 TAILQ_FOREACH(p, &kif->pfik_dynaddrs, entry) 632 pfi_dynaddr_update(p); 633 634 /* Apply group flags to new members. */ 635 if (kif->pfik_group != NULL) { 636 CK_STAILQ_FOREACH(ifgm, &kif->pfik_group->ifg_members, 637 ifgm_next) { 638 tmpkif = (struct pfi_kkif *)ifgm->ifgm_ifp->if_pf_kif; 639 if (tmpkif == NULL) 640 continue; 641 642 tmpkif->pfik_flags |= kif->pfik_flags; 643 } 644 } 645 646 /* again for all groups kif is member of */ 647 if (kif->pfik_ifp != NULL) { 648 CK_STAILQ_FOREACH(ifgl, &kif->pfik_ifp->if_groups, ifgl_next) 649 pfi_kkif_update((struct pfi_kkif *) 650 ifgl->ifgl_group->ifg_pf_kif); 651 } 652 } 653 654 static void 655 pfi_dynaddr_update(struct pfi_dynaddr *dyn) 656 { 657 struct pfi_kkif *kif; 658 struct pfr_ktable *kt; 659 660 PF_RULES_WASSERT(); 661 KASSERT(dyn && dyn->pfid_kif && dyn->pfid_kt, 662 ("%s: bad argument", __func__)); 663 664 kif = dyn->pfid_kif; 665 kt = dyn->pfid_kt; 666 667 if (kt->pfrkt_larg != V_pfi_update) { 668 /* this table needs to be brought up-to-date */ 669 pfi_table_update(kt, kif, dyn->pfid_net, dyn->pfid_iflags); 670 kt->pfrkt_larg = V_pfi_update; 671 } 672 pfr_dynaddr_update(kt, dyn); 673 } 674 675 static void 676 pfi_table_update(struct pfr_ktable *kt, struct pfi_kkif *kif, int net, int flags) 677 { 678 int e, size2 = 0; 679 struct ifg_member *ifgm; 680 681 NET_EPOCH_ASSERT(); 682 683 V_pfi_buffer_cnt = 0; 684 685 if (kif->pfik_ifp != NULL) 686 pfi_instance_add(kif->pfik_ifp, net, flags); 687 else if (kif->pfik_group != NULL) { 688 CK_STAILQ_FOREACH(ifgm, &kif->pfik_group->ifg_members, ifgm_next) 689 pfi_instance_add(ifgm->ifgm_ifp, net, flags); 690 } 691 692 if ((e = pfr_set_addrs(&kt->pfrkt_t, V_pfi_buffer, V_pfi_buffer_cnt, &size2, 693 NULL, NULL, NULL, 0, PFR_TFLAG_ALLMASK))) 694 printf("%s: cannot set %d new addresses into table %s: %d\n", 695 __func__, V_pfi_buffer_cnt, kt->pfrkt_name, e); 696 } 697 698 static void 699 pfi_instance_add(struct ifnet *ifp, int net, int flags) 700 { 701 struct ifaddr *ia; 702 int got4 = 0, got6 = 0; 703 int net2, af; 704 705 NET_EPOCH_ASSERT(); 706 707 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 708 if (ia->ifa_addr == NULL) 709 continue; 710 af = ia->ifa_addr->sa_family; 711 if (af != AF_INET && af != AF_INET6) 712 continue; 713 /* 714 * XXX: For point-to-point interfaces, (ifname:0) and IPv4, 715 * jump over addresses without a proper route to work 716 * around a problem with ppp not fully removing the 717 * address used during IPCP. 718 */ 719 if ((ifp->if_flags & IFF_POINTOPOINT) && 720 !(ia->ifa_flags & IFA_ROUTE) && 721 (flags & PFI_AFLAG_NOALIAS) && (af == AF_INET)) 722 continue; 723 if ((flags & PFI_AFLAG_BROADCAST) && af == AF_INET6) 724 continue; 725 if ((flags & PFI_AFLAG_BROADCAST) && 726 !(ifp->if_flags & IFF_BROADCAST)) 727 continue; 728 if ((flags & PFI_AFLAG_PEER) && 729 !(ifp->if_flags & IFF_POINTOPOINT)) 730 continue; 731 if ((flags & (PFI_AFLAG_NETWORK | PFI_AFLAG_NOALIAS)) && 732 af == AF_INET6 && 733 IN6_IS_ADDR_LINKLOCAL( 734 &((struct sockaddr_in6 *)ia->ifa_addr)->sin6_addr)) 735 continue; 736 if (flags & PFI_AFLAG_NOALIAS) { 737 if (af == AF_INET && got4) 738 continue; 739 if (af == AF_INET6 && got6) 740 continue; 741 } 742 if (af == AF_INET) 743 got4 = 1; 744 else if (af == AF_INET6) 745 got6 = 1; 746 net2 = net; 747 if (net2 == 128 && (flags & PFI_AFLAG_NETWORK)) { 748 if (af == AF_INET) 749 net2 = pfi_unmask(&((struct sockaddr_in *) 750 ia->ifa_netmask)->sin_addr); 751 else if (af == AF_INET6) 752 net2 = pfi_unmask(&((struct sockaddr_in6 *) 753 ia->ifa_netmask)->sin6_addr); 754 } 755 if (af == AF_INET && net2 > 32) 756 net2 = 32; 757 if (flags & PFI_AFLAG_BROADCAST) 758 pfi_address_add(ia->ifa_broadaddr, af, net2); 759 else if (flags & PFI_AFLAG_PEER) 760 pfi_address_add(ia->ifa_dstaddr, af, net2); 761 else 762 pfi_address_add(ia->ifa_addr, af, net2); 763 } 764 } 765 766 static void 767 pfi_address_add(struct sockaddr *sa, int af, int net) 768 { 769 struct pfr_addr *p; 770 int i; 771 772 if (V_pfi_buffer_cnt >= V_pfi_buffer_max) { 773 int new_max = V_pfi_buffer_max * 2; 774 775 if (new_max > PFI_BUFFER_MAX) { 776 printf("%s: address buffer full (%d/%d)\n", __func__, 777 V_pfi_buffer_cnt, PFI_BUFFER_MAX); 778 return; 779 } 780 p = malloc(new_max * sizeof(*V_pfi_buffer), PFI_MTYPE, 781 M_NOWAIT); 782 if (p == NULL) { 783 printf("%s: no memory to grow buffer (%d/%d)\n", 784 __func__, V_pfi_buffer_cnt, PFI_BUFFER_MAX); 785 return; 786 } 787 memcpy(p, V_pfi_buffer, V_pfi_buffer_max * sizeof(*V_pfi_buffer)); 788 /* no need to zero buffer */ 789 free(V_pfi_buffer, PFI_MTYPE); 790 V_pfi_buffer = p; 791 V_pfi_buffer_max = new_max; 792 } 793 if (af == AF_INET && net > 32) 794 net = 128; 795 p = V_pfi_buffer + V_pfi_buffer_cnt++; 796 memset(p, 0, sizeof(*p)); 797 p->pfra_af = af; 798 p->pfra_net = net; 799 if (af == AF_INET) 800 p->pfra_ip4addr = ((struct sockaddr_in *)sa)->sin_addr; 801 else if (af == AF_INET6) { 802 p->pfra_ip6addr = ((struct sockaddr_in6 *)sa)->sin6_addr; 803 if (IN6_IS_SCOPE_EMBED(&p->pfra_ip6addr)) 804 p->pfra_ip6addr.s6_addr16[1] = 0; 805 } 806 /* mask network address bits */ 807 if (net < 128) 808 ((caddr_t)p)[p->pfra_net/8] &= ~(0xFF >> (p->pfra_net%8)); 809 for (i = (p->pfra_net+7)/8; i < sizeof(p->pfra_u); i++) 810 ((caddr_t)p)[i] = 0; 811 } 812 813 void 814 pfi_dynaddr_remove(struct pfi_dynaddr *dyn) 815 { 816 817 KASSERT(dyn->pfid_kif != NULL, ("%s: null pfid_kif", __func__)); 818 KASSERT(dyn->pfid_kt != NULL, ("%s: null pfid_kt", __func__)); 819 820 TAILQ_REMOVE(&dyn->pfid_kif->pfik_dynaddrs, dyn, entry); 821 pfi_kkif_unref(dyn->pfid_kif); 822 pfr_detach_table(dyn->pfid_kt); 823 free(dyn, PFI_MTYPE); 824 } 825 826 void 827 pfi_dynaddr_copyout(struct pf_addr_wrap *aw) 828 { 829 830 KASSERT(aw->type == PF_ADDR_DYNIFTL, 831 ("%s: type %u", __func__, aw->type)); 832 833 if (aw->p.dyn == NULL || aw->p.dyn->pfid_kif == NULL) 834 return; 835 aw->p.dyncnt = aw->p.dyn->pfid_acnt4 + aw->p.dyn->pfid_acnt6; 836 } 837 838 static int 839 pfi_kkif_compare(struct pfi_kkif *p, struct pfi_kkif *q) 840 { 841 return (strncmp(p->pfik_name, q->pfik_name, IFNAMSIZ)); 842 } 843 844 void 845 pfi_update_status(const char *name, struct pf_status *pfs) 846 { 847 struct pfi_kkif *p; 848 struct pfi_kif_cmp key; 849 struct ifg_member p_member, *ifgm; 850 CK_STAILQ_HEAD(, ifg_member) ifg_members; 851 int i, j, k; 852 853 if (pfs) { 854 memset(pfs->pcounters, 0, sizeof(pfs->pcounters)); 855 memset(pfs->bcounters, 0, sizeof(pfs->bcounters)); 856 } 857 858 strlcpy(key.pfik_name, name, sizeof(key.pfik_name)); 859 p = RB_FIND(pfi_ifhead, &V_pfi_ifs, (struct pfi_kkif *)&key); 860 if (p == NULL) { 861 return; 862 } 863 864 if (p->pfik_group != NULL) { 865 memcpy(&ifg_members, &p->pfik_group->ifg_members, 866 sizeof(ifg_members)); 867 } else { 868 /* build a temporary list for p only */ 869 memset(&p_member, 0, sizeof(p_member)); 870 p_member.ifgm_ifp = p->pfik_ifp; 871 CK_STAILQ_INIT(&ifg_members); 872 CK_STAILQ_INSERT_TAIL(&ifg_members, &p_member, ifgm_next); 873 } 874 CK_STAILQ_FOREACH(ifgm, &ifg_members, ifgm_next) { 875 if (ifgm->ifgm_ifp == NULL || ifgm->ifgm_ifp->if_pf_kif == NULL) 876 continue; 877 p = (struct pfi_kkif *)ifgm->ifgm_ifp->if_pf_kif; 878 879 /* just clear statistics */ 880 if (pfs == NULL) { 881 pf_kkif_zero(p); 882 continue; 883 } 884 for (i = 0; i < 2; i++) 885 for (j = 0; j < 2; j++) 886 for (k = 0; k < 2; k++) { 887 pfs->pcounters[i][j][k] += 888 pf_counter_u64_fetch(&p->pfik_packets[i][j][k]); 889 pfs->bcounters[i][j] += 890 pf_counter_u64_fetch(&p->pfik_bytes[i][j][k]); 891 } 892 } 893 } 894 895 static void 896 pf_kkif_to_kif(struct pfi_kkif *kkif, struct pfi_kif *kif) 897 { 898 899 memset(kif, 0, sizeof(*kif)); 900 strlcpy(kif->pfik_name, kkif->pfik_name, sizeof(kif->pfik_name)); 901 for (int i = 0; i < 2; i++) { 902 for (int j = 0; j < 2; j++) { 903 for (int k = 0; k < 2; k++) { 904 kif->pfik_packets[i][j][k] = 905 pf_counter_u64_fetch(&kkif->pfik_packets[i][j][k]); 906 kif->pfik_bytes[i][j][k] = 907 pf_counter_u64_fetch(&kkif->pfik_bytes[i][j][k]); 908 } 909 } 910 } 911 kif->pfik_flags = kkif->pfik_flags; 912 kif->pfik_tzero = kkif->pfik_tzero; 913 kif->pfik_rulerefs = kkif->pfik_rulerefs; 914 /* 915 * Userspace relies on this pointer to decide if this is a group or 916 * not. We don't want to share the actual pointer, because it's 917 * useless to userspace and leaks kernel memory layout information. 918 * So instead we provide 0xfeedcode as 'true' and NULL as 'false'. 919 */ 920 kif->pfik_group = 921 kkif->pfik_group ? (struct ifg_group *)0xfeedc0de : NULL; 922 } 923 924 void 925 pfi_get_ifaces(const char *name, struct pfi_kif *buf, int *size) 926 { 927 struct epoch_tracker et; 928 struct pfi_kkif *p, *nextp; 929 int n = 0; 930 931 NET_EPOCH_ENTER(et); 932 for (p = RB_MIN(pfi_ifhead, &V_pfi_ifs); p; p = nextp) { 933 nextp = RB_NEXT(pfi_ifhead, &V_pfi_ifs, p); 934 if (pfi_skip_if(name, p)) 935 continue; 936 if (*size <= n++) 937 break; 938 if (!p->pfik_tzero) 939 p->pfik_tzero = time_second; 940 pf_kkif_to_kif(p, buf++); 941 nextp = RB_NEXT(pfi_ifhead, &V_pfi_ifs, p); 942 } 943 *size = n; 944 NET_EPOCH_EXIT(et); 945 } 946 947 static int 948 pfi_skip_if(const char *filter, struct pfi_kkif *p) 949 { 950 struct ifg_list *i; 951 int n; 952 953 NET_EPOCH_ASSERT(); 954 955 if (filter == NULL || !*filter) 956 return (0); 957 if (!strcmp(p->pfik_name, filter)) 958 return (0); /* exact match */ 959 n = strlen(filter); 960 if (n < 1 || n >= IFNAMSIZ) 961 return (1); /* sanity check */ 962 if (filter[n-1] >= '0' && filter[n-1] <= '9') 963 return (1); /* group names may not end in a digit */ 964 if (p->pfik_ifp == NULL) 965 return (1); 966 CK_STAILQ_FOREACH(i, &p->pfik_ifp->if_groups, ifgl_next) 967 if (!strncmp(i->ifgl_group->ifg_group, filter, IFNAMSIZ)) 968 return (0); /* iface is in group "filter" */ 969 return (1); 970 } 971 972 int 973 pfi_set_flags(const char *name, int flags) 974 { 975 struct epoch_tracker et; 976 struct pfi_kkif *p, *kif; 977 978 kif = pf_kkif_create(M_NOWAIT); 979 if (kif == NULL) 980 return (ENOMEM); 981 982 NET_EPOCH_ENTER(et); 983 984 kif = pfi_kkif_attach(kif, name); 985 986 RB_FOREACH(p, pfi_ifhead, &V_pfi_ifs) { 987 if (pfi_skip_if(name, p)) 988 continue; 989 p->pfik_flags |= flags; 990 } 991 NET_EPOCH_EXIT(et); 992 return (0); 993 } 994 995 int 996 pfi_clear_flags(const char *name, int flags) 997 { 998 struct epoch_tracker et; 999 struct pfi_kkif *p, *tmp; 1000 1001 NET_EPOCH_ENTER(et); 1002 RB_FOREACH_SAFE(p, pfi_ifhead, &V_pfi_ifs, tmp) { 1003 if (pfi_skip_if(name, p)) 1004 continue; 1005 p->pfik_flags &= ~flags; 1006 1007 if (p->pfik_ifp == NULL && p->pfik_group == NULL && 1008 p->pfik_flags == 0 && p->pfik_rulerefs == 0) { 1009 /* Delete this kif. */ 1010 RB_REMOVE(pfi_ifhead, &V_pfi_ifs, p); 1011 pf_kkif_free(p); 1012 } 1013 } 1014 NET_EPOCH_EXIT(et); 1015 return (0); 1016 } 1017 1018 /* from pf_print_state.c */ 1019 static int 1020 pfi_unmask(void *addr) 1021 { 1022 struct pf_addr *m = addr; 1023 int i = 31, j = 0, b = 0; 1024 u_int32_t tmp; 1025 1026 while (j < 4 && m->addr32[j] == 0xffffffff) { 1027 b += 32; 1028 j++; 1029 } 1030 if (j < 4) { 1031 tmp = ntohl(m->addr32[j]); 1032 for (i = 31; tmp & (1 << i); --i) 1033 b++; 1034 } 1035 return (b); 1036 } 1037 1038 static void 1039 pfi_attach_ifnet_event(void *arg __unused, struct ifnet *ifp) 1040 { 1041 struct epoch_tracker et; 1042 struct pfi_kkif *kif; 1043 1044 if (V_pf_vnet_active == 0) { 1045 /* Avoid teardown race in the least expensive way. */ 1046 return; 1047 } 1048 kif = pf_kkif_create(M_NOWAIT); 1049 NET_EPOCH_ENTER(et); 1050 PF_RULES_WLOCK(); 1051 pfi_attach_ifnet(ifp, kif); 1052 #ifdef ALTQ 1053 pf_altq_ifnet_event(ifp, 0); 1054 #endif 1055 PF_RULES_WUNLOCK(); 1056 NET_EPOCH_EXIT(et); 1057 } 1058 1059 static void 1060 pfi_detach_ifnet_event(void *arg __unused, struct ifnet *ifp) 1061 { 1062 struct epoch_tracker et; 1063 struct pfi_kkif *kif = (struct pfi_kkif *)ifp->if_pf_kif; 1064 1065 if (pfsync_detach_ifnet_ptr) 1066 pfsync_detach_ifnet_ptr(ifp); 1067 1068 if (kif == NULL) 1069 return; 1070 1071 if (V_pf_vnet_active == 0) { 1072 /* Avoid teardown race in the least expensive way. */ 1073 return; 1074 } 1075 1076 NET_EPOCH_ENTER(et); 1077 PF_RULES_WLOCK(); 1078 V_pfi_update++; 1079 pfi_kkif_update(kif); 1080 1081 if (kif->pfik_ifp) 1082 if_rele(kif->pfik_ifp); 1083 1084 kif->pfik_ifp = NULL; 1085 ifp->if_pf_kif = NULL; 1086 #ifdef ALTQ 1087 pf_altq_ifnet_event(ifp, 1); 1088 #endif 1089 pfi_kkif_remove_if_unref(kif); 1090 1091 PF_RULES_WUNLOCK(); 1092 NET_EPOCH_EXIT(et); 1093 } 1094 1095 static void 1096 pfi_attach_group_event(void *arg __unused, struct ifg_group *ifg) 1097 { 1098 struct epoch_tracker et; 1099 struct pfi_kkif *kif; 1100 1101 if (V_pf_vnet_active == 0) { 1102 /* Avoid teardown race in the least expensive way. */ 1103 return; 1104 } 1105 kif = pf_kkif_create(M_WAITOK); 1106 NET_EPOCH_ENTER(et); 1107 PF_RULES_WLOCK(); 1108 pfi_attach_ifgroup(ifg, kif); 1109 PF_RULES_WUNLOCK(); 1110 NET_EPOCH_EXIT(et); 1111 } 1112 1113 static void 1114 pfi_change_group_event(void *arg __unused, char *gname) 1115 { 1116 struct epoch_tracker et; 1117 struct pfi_kkif *kif; 1118 1119 if (V_pf_vnet_active == 0) { 1120 /* Avoid teardown race in the least expensive way. */ 1121 return; 1122 } 1123 1124 kif = pf_kkif_create(M_WAITOK); 1125 NET_EPOCH_ENTER(et); 1126 PF_RULES_WLOCK(); 1127 V_pfi_update++; 1128 kif = pfi_kkif_attach(kif, gname); 1129 pfi_kkif_update(kif); 1130 PF_RULES_WUNLOCK(); 1131 NET_EPOCH_EXIT(et); 1132 } 1133 1134 static void 1135 pfi_detach_group_event(void *arg __unused, struct ifg_group *ifg) 1136 { 1137 struct pfi_kkif *kif = (struct pfi_kkif *)ifg->ifg_pf_kif; 1138 1139 if (kif == NULL) 1140 return; 1141 1142 if (V_pf_vnet_active == 0) { 1143 /* Avoid teardown race in the least expensive way. */ 1144 return; 1145 } 1146 PF_RULES_WLOCK(); 1147 V_pfi_update++; 1148 1149 kif->pfik_group = NULL; 1150 ifg->ifg_pf_kif = NULL; 1151 1152 pfi_kkif_remove_if_unref(kif); 1153 PF_RULES_WUNLOCK(); 1154 } 1155 1156 static void 1157 pfi_ifaddr_event(void *arg __unused, struct ifnet *ifp) 1158 { 1159 1160 KASSERT(ifp, ("ifp == NULL")); 1161 1162 if (ifp->if_pf_kif == NULL) 1163 return; 1164 1165 if (V_pf_vnet_active == 0) { 1166 /* Avoid teardown race in the least expensive way. */ 1167 return; 1168 } 1169 PF_RULES_WLOCK(); 1170 if (ifp->if_pf_kif) { 1171 struct epoch_tracker et; 1172 1173 V_pfi_update++; 1174 NET_EPOCH_ENTER(et); 1175 pfi_kkif_update(ifp->if_pf_kif); 1176 NET_EPOCH_EXIT(et); 1177 } 1178 PF_RULES_WUNLOCK(); 1179 } 1180