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 *, uint8_t, 90 int); 91 static void pfi_instance_add(struct ifnet *, uint8_t, int); 92 static void pfi_address_add(struct sockaddr *, sa_family_t, uint8_t); 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 if (!strcmp(kif->pfik_name, "any")) { 361 /* both so it works in the ioctl and the regular case */ 362 kif->pfik_flags |= PFI_IFLAG_ANY; 363 } 364 365 RB_INSERT(pfi_ifhead, &V_pfi_ifs, kif); 366 367 return (kif); 368 } 369 370 void 371 pfi_kkif_ref(struct pfi_kkif *kif) 372 { 373 374 PF_RULES_WASSERT(); 375 kif->pfik_rulerefs++; 376 } 377 378 static void 379 pfi_kkif_remove_if_unref(struct pfi_kkif *kif) 380 { 381 382 PF_RULES_WASSERT(); 383 384 if (kif->pfik_rulerefs > 0) 385 return; 386 387 /* kif referencing an existing ifnet or group or holding flags should 388 * exist. */ 389 if (kif->pfik_ifp != NULL || kif->pfik_group != NULL || 390 kif == V_pfi_all || kif->pfik_flags != 0) 391 return; 392 393 /* 394 * We can get here in at least two distinct paths: 395 * - when the struct ifnet is removed, via pfi_detach_ifnet_event() 396 * - when a rule referencing us is removed, via pfi_kkif_unref(). 397 * These two events can race against each other, leading us to free this kif 398 * twice. That leads to a loop in V_pfi_unlinked_kifs, and an eventual 399 * deadlock. 400 * 401 * Avoid this by making sure we only ever insert the kif into 402 * V_pfi_unlinked_kifs once. 403 * If we don't find it in V_pfi_ifs it's already been removed. Check that it 404 * exists in V_pfi_unlinked_kifs. 405 */ 406 if (! RB_FIND(pfi_ifhead, &V_pfi_ifs, kif)) { 407 #ifdef INVARIANTS 408 struct pfi_kkif *tmp; 409 bool found = false; 410 mtx_lock(&pfi_unlnkdkifs_mtx); 411 LIST_FOREACH(tmp, &V_pfi_unlinked_kifs, pfik_list) { 412 if (tmp == kif) { 413 found = true; 414 break; 415 } 416 } 417 mtx_unlock(&pfi_unlnkdkifs_mtx); 418 MPASS(found); 419 #endif 420 return; 421 } 422 RB_REMOVE(pfi_ifhead, &V_pfi_ifs, kif); 423 424 kif->pfik_flags |= PFI_IFLAG_REFS; 425 426 mtx_lock(&pfi_unlnkdkifs_mtx); 427 LIST_INSERT_HEAD(&V_pfi_unlinked_kifs, kif, pfik_list); 428 mtx_unlock(&pfi_unlnkdkifs_mtx); 429 } 430 431 void 432 pfi_kkif_unref(struct pfi_kkif *kif) 433 { 434 435 PF_RULES_WASSERT(); 436 KASSERT(kif->pfik_rulerefs > 0, ("%s: %p has zero refs", __func__, kif)); 437 438 kif->pfik_rulerefs--; 439 440 pfi_kkif_remove_if_unref(kif); 441 } 442 443 void 444 pfi_kkif_purge(void) 445 { 446 struct pfi_kkif *kif, *kif1; 447 448 /* 449 * Do naive mark-and-sweep garbage collecting of old kifs. 450 * Reference flag is raised by pf_purge_expired_states(). 451 */ 452 mtx_lock(&pfi_unlnkdkifs_mtx); 453 LIST_FOREACH_SAFE(kif, &V_pfi_unlinked_kifs, pfik_list, kif1) { 454 if (!(kif->pfik_flags & PFI_IFLAG_REFS)) { 455 LIST_REMOVE(kif, pfik_list); 456 pf_kkif_free(kif); 457 } else 458 kif->pfik_flags &= ~PFI_IFLAG_REFS; 459 } 460 mtx_unlock(&pfi_unlnkdkifs_mtx); 461 } 462 463 int 464 pfi_kkif_match(struct pfi_kkif *rule_kif, struct pfi_kkif *packet_kif) 465 { 466 struct ifg_list *p; 467 468 NET_EPOCH_ASSERT(); 469 470 MPASS(packet_kif != NULL); 471 MPASS(packet_kif->pfik_ifp != NULL); 472 473 if (rule_kif == NULL || rule_kif == packet_kif) 474 return (1); 475 476 if (rule_kif->pfik_group != NULL) { 477 CK_STAILQ_FOREACH(p, &packet_kif->pfik_ifp->if_groups, ifgl_next) 478 if (p->ifgl_group == rule_kif->pfik_group) 479 return (1); 480 } 481 482 if (rule_kif->pfik_flags & PFI_IFLAG_ANY && packet_kif->pfik_ifp && 483 !(packet_kif->pfik_ifp->if_flags & IFF_LOOPBACK)) 484 return (1); 485 486 return (0); 487 } 488 489 static void 490 pfi_attach_ifnet(struct ifnet *ifp, struct pfi_kkif *kif) 491 { 492 493 PF_RULES_WASSERT(); 494 495 V_pfi_update++; 496 kif = pfi_kkif_attach(kif, ifp->if_xname); 497 if_ref(ifp); 498 kif->pfik_ifp = ifp; 499 ifp->if_pf_kif = kif; 500 pfi_kkif_update(kif); 501 } 502 503 static void 504 pfi_attach_ifgroup(struct ifg_group *ifg, struct pfi_kkif *kif) 505 { 506 507 PF_RULES_WASSERT(); 508 509 V_pfi_update++; 510 kif = pfi_kkif_attach(kif, ifg->ifg_group); 511 kif->pfik_group = ifg; 512 ifg->ifg_pf_kif = kif; 513 } 514 515 int 516 pfi_match_addr(struct pfi_dynaddr *dyn, struct pf_addr *a, sa_family_t af) 517 { 518 switch (af) { 519 #ifdef INET 520 case AF_INET: 521 switch (dyn->pfid_acnt4) { 522 case 0: 523 return (0); 524 case 1: 525 return (pf_match_addr(0, &dyn->pfid_addr4, 526 &dyn->pfid_mask4, a, AF_INET)); 527 default: 528 return (pfr_match_addr(dyn->pfid_kt, a, AF_INET)); 529 } 530 break; 531 #endif /* INET */ 532 #ifdef INET6 533 case AF_INET6: 534 switch (dyn->pfid_acnt6) { 535 case 0: 536 return (0); 537 case 1: 538 return (pf_match_addr(0, &dyn->pfid_addr6, 539 &dyn->pfid_mask6, a, AF_INET6)); 540 default: 541 return (pfr_match_addr(dyn->pfid_kt, a, AF_INET6)); 542 } 543 break; 544 #endif /* INET6 */ 545 default: 546 return (0); 547 } 548 } 549 550 int 551 pfi_dynaddr_setup(struct pf_addr_wrap *aw, sa_family_t af) 552 { 553 struct epoch_tracker et; 554 struct pfi_dynaddr *dyn; 555 char tblname[PF_TABLE_NAME_SIZE]; 556 struct pf_kruleset *ruleset = NULL; 557 struct pfi_kkif *kif; 558 int rv = 0; 559 560 PF_RULES_WASSERT(); 561 KASSERT(aw->type == PF_ADDR_DYNIFTL, ("%s: type %u", 562 __func__, aw->type)); 563 KASSERT(aw->p.dyn == NULL, ("%s: dyn is %p", __func__, aw->p.dyn)); 564 565 if ((dyn = malloc(sizeof(*dyn), PFI_MTYPE, M_NOWAIT | M_ZERO)) == NULL) 566 return (ENOMEM); 567 568 if ((kif = pf_kkif_create(M_NOWAIT)) == NULL) { 569 free(dyn, PFI_MTYPE); 570 return (ENOMEM); 571 } 572 573 if (!strcmp(aw->v.ifname, "self")) 574 dyn->pfid_kif = pfi_kkif_attach(kif, IFG_ALL); 575 else 576 dyn->pfid_kif = pfi_kkif_attach(kif, aw->v.ifname); 577 kif = NULL; 578 pfi_kkif_ref(dyn->pfid_kif); 579 580 dyn->pfid_net = pfi_unmask(&aw->v.a.mask); 581 if (af == AF_INET && dyn->pfid_net == 32) 582 dyn->pfid_net = 128; 583 strlcpy(tblname, aw->v.ifname, sizeof(tblname)); 584 if (aw->iflags & PFI_AFLAG_NETWORK) 585 strlcat(tblname, ":network", sizeof(tblname)); 586 if (aw->iflags & PFI_AFLAG_BROADCAST) 587 strlcat(tblname, ":broadcast", sizeof(tblname)); 588 if (aw->iflags & PFI_AFLAG_PEER) 589 strlcat(tblname, ":peer", sizeof(tblname)); 590 if (aw->iflags & PFI_AFLAG_NOALIAS) 591 strlcat(tblname, ":0", sizeof(tblname)); 592 if (dyn->pfid_net != 128) 593 snprintf(tblname + strlen(tblname), 594 sizeof(tblname) - strlen(tblname), "/%d", dyn->pfid_net); 595 if ((ruleset = pf_find_or_create_kruleset(PF_RESERVED_ANCHOR)) == NULL) { 596 rv = ENOMEM; 597 goto _bad; 598 } 599 600 if ((dyn->pfid_kt = pfr_attach_table(ruleset, tblname)) == NULL) { 601 rv = ENOMEM; 602 goto _bad; 603 } 604 605 dyn->pfid_kt->pfrkt_flags |= PFR_TFLAG_ACTIVE; 606 dyn->pfid_iflags = aw->iflags; 607 dyn->pfid_af = af; 608 609 TAILQ_INSERT_TAIL(&dyn->pfid_kif->pfik_dynaddrs, dyn, entry); 610 aw->p.dyn = dyn; 611 NET_EPOCH_ENTER(et); 612 pfi_kkif_update(dyn->pfid_kif); 613 NET_EPOCH_EXIT(et); 614 615 return (0); 616 617 _bad: 618 if (dyn->pfid_kt != NULL) 619 pfr_detach_table(dyn->pfid_kt); 620 if (ruleset != NULL) 621 pf_remove_if_empty_kruleset(ruleset); 622 pfi_kkif_unref(dyn->pfid_kif); 623 free(dyn, PFI_MTYPE); 624 625 return (rv); 626 } 627 628 static void 629 pfi_kkif_update(struct pfi_kkif *kif) 630 { 631 struct ifg_list *ifgl; 632 struct ifg_member *ifgm; 633 struct pfi_dynaddr *p; 634 struct pfi_kkif *tmpkif; 635 636 NET_EPOCH_ASSERT(); 637 PF_RULES_WASSERT(); 638 639 /* update all dynaddr */ 640 TAILQ_FOREACH(p, &kif->pfik_dynaddrs, entry) 641 pfi_dynaddr_update(p); 642 643 /* Apply group flags to new members. */ 644 if (kif->pfik_group != NULL) { 645 CK_STAILQ_FOREACH(ifgm, &kif->pfik_group->ifg_members, 646 ifgm_next) { 647 tmpkif = (struct pfi_kkif *)ifgm->ifgm_ifp->if_pf_kif; 648 if (tmpkif == NULL) 649 continue; 650 651 tmpkif->pfik_flags |= kif->pfik_flags; 652 } 653 } 654 655 /* again for all groups kif is member of */ 656 if (kif->pfik_ifp != NULL) { 657 CK_STAILQ_FOREACH(ifgl, &kif->pfik_ifp->if_groups, ifgl_next) 658 if (ifgl->ifgl_group->ifg_pf_kif) { 659 pfi_kkif_update((struct pfi_kkif *) 660 ifgl->ifgl_group->ifg_pf_kif); 661 } 662 } 663 } 664 665 static void 666 pfi_dynaddr_update(struct pfi_dynaddr *dyn) 667 { 668 struct pfi_kkif *kif; 669 struct pfr_ktable *kt; 670 671 PF_RULES_WASSERT(); 672 KASSERT(dyn && dyn->pfid_kif && dyn->pfid_kt, 673 ("%s: bad argument", __func__)); 674 675 kif = dyn->pfid_kif; 676 kt = dyn->pfid_kt; 677 678 if (kt->pfrkt_larg != V_pfi_update) { 679 /* this table needs to be brought up-to-date */ 680 pfi_table_update(kt, kif, dyn->pfid_net, dyn->pfid_iflags); 681 kt->pfrkt_larg = V_pfi_update; 682 } 683 pfr_dynaddr_update(kt, dyn); 684 } 685 686 static void 687 pfi_table_update(struct pfr_ktable *kt, struct pfi_kkif *kif, uint8_t net, 688 int flags) 689 { 690 int e, size2 = 0; 691 struct ifg_member *ifgm; 692 693 NET_EPOCH_ASSERT(); 694 695 V_pfi_buffer_cnt = 0; 696 697 if (kif->pfik_ifp != NULL) 698 pfi_instance_add(kif->pfik_ifp, net, flags); 699 else if (kif->pfik_group != NULL) { 700 CK_STAILQ_FOREACH(ifgm, &kif->pfik_group->ifg_members, ifgm_next) 701 pfi_instance_add(ifgm->ifgm_ifp, net, flags); 702 } 703 704 if ((e = pfr_set_addrs(&kt->pfrkt_t, V_pfi_buffer, V_pfi_buffer_cnt, &size2, 705 NULL, NULL, NULL, 0, PFR_TFLAG_ALLMASK))) 706 printf("%s: cannot set %d new addresses into table %s: %d\n", 707 __func__, V_pfi_buffer_cnt, kt->pfrkt_name, e); 708 } 709 710 static void 711 pfi_instance_add(struct ifnet *ifp, uint8_t net, int flags) 712 { 713 struct ifaddr *ia; 714 int got4 = 0, got6 = 0; 715 sa_family_t af; 716 uint8_t net2; 717 718 NET_EPOCH_ASSERT(); 719 720 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 721 if (ia->ifa_addr == NULL) 722 continue; 723 af = ia->ifa_addr->sa_family; 724 if (af != AF_INET && af != AF_INET6) 725 continue; 726 /* 727 * XXX: For point-to-point interfaces, (ifname:0) and IPv4, 728 * jump over addresses without a proper route to work 729 * around a problem with ppp not fully removing the 730 * address used during IPCP. 731 */ 732 if ((ifp->if_flags & IFF_POINTOPOINT) && 733 !(ia->ifa_flags & IFA_ROUTE) && 734 (flags & PFI_AFLAG_NOALIAS) && (af == AF_INET)) 735 continue; 736 if ((flags & PFI_AFLAG_BROADCAST) && af == AF_INET6) 737 continue; 738 if ((flags & PFI_AFLAG_BROADCAST) && 739 !(ifp->if_flags & IFF_BROADCAST)) 740 continue; 741 if ((flags & PFI_AFLAG_PEER) && 742 !(ifp->if_flags & IFF_POINTOPOINT)) 743 continue; 744 if ((flags & (PFI_AFLAG_NETWORK | PFI_AFLAG_NOALIAS)) && 745 af == AF_INET6 && 746 IN6_IS_ADDR_LINKLOCAL( 747 &((struct sockaddr_in6 *)ia->ifa_addr)->sin6_addr)) 748 continue; 749 if (flags & PFI_AFLAG_NOALIAS) { 750 if (af == AF_INET && got4) 751 continue; 752 if (af == AF_INET6 && got6) 753 continue; 754 } 755 if (af == AF_INET) 756 got4 = 1; 757 else if (af == AF_INET6) 758 got6 = 1; 759 net2 = net; 760 if (net2 == 128 && (flags & PFI_AFLAG_NETWORK)) { 761 if (af == AF_INET) 762 net2 = pfi_unmask(&((struct sockaddr_in *) 763 ia->ifa_netmask)->sin_addr); 764 else if (af == AF_INET6) 765 net2 = pfi_unmask(&((struct sockaddr_in6 *) 766 ia->ifa_netmask)->sin6_addr); 767 } 768 if (af == AF_INET && net2 > 32) 769 net2 = 32; 770 if (flags & PFI_AFLAG_BROADCAST) 771 pfi_address_add(ia->ifa_broadaddr, af, net2); 772 else if (flags & PFI_AFLAG_PEER) 773 pfi_address_add(ia->ifa_dstaddr, af, net2); 774 else 775 pfi_address_add(ia->ifa_addr, af, net2); 776 } 777 } 778 779 static void 780 pfi_address_add(struct sockaddr *sa, sa_family_t af, uint8_t net) 781 { 782 struct pfr_addr *p; 783 int i; 784 785 if (V_pfi_buffer_cnt >= V_pfi_buffer_max) { 786 int new_max = V_pfi_buffer_max * 2; 787 788 if (new_max > PFI_BUFFER_MAX) { 789 printf("%s: address buffer full (%d/%d)\n", __func__, 790 V_pfi_buffer_cnt, PFI_BUFFER_MAX); 791 return; 792 } 793 p = malloc(new_max * sizeof(*V_pfi_buffer), PFI_MTYPE, 794 M_NOWAIT); 795 if (p == NULL) { 796 printf("%s: no memory to grow buffer (%d/%d)\n", 797 __func__, V_pfi_buffer_cnt, PFI_BUFFER_MAX); 798 return; 799 } 800 memcpy(p, V_pfi_buffer, V_pfi_buffer_max * sizeof(*V_pfi_buffer)); 801 /* no need to zero buffer */ 802 free(V_pfi_buffer, PFI_MTYPE); 803 V_pfi_buffer = p; 804 V_pfi_buffer_max = new_max; 805 } 806 if (af == AF_INET && net > 32) 807 net = 128; 808 p = V_pfi_buffer + V_pfi_buffer_cnt++; 809 memset(p, 0, sizeof(*p)); 810 p->pfra_af = af; 811 p->pfra_net = net; 812 if (af == AF_INET) 813 p->pfra_ip4addr = ((struct sockaddr_in *)sa)->sin_addr; 814 else if (af == AF_INET6) { 815 p->pfra_ip6addr = ((struct sockaddr_in6 *)sa)->sin6_addr; 816 if (IN6_IS_SCOPE_EMBED(&p->pfra_ip6addr)) 817 p->pfra_ip6addr.s6_addr16[1] = 0; 818 } 819 /* mask network address bits */ 820 if (net < 128) 821 ((caddr_t)p)[p->pfra_net/8] &= ~(0xFF >> (p->pfra_net%8)); 822 for (i = (p->pfra_net+7)/8; i < sizeof(p->pfra_u); i++) 823 ((caddr_t)p)[i] = 0; 824 } 825 826 void 827 pfi_dynaddr_remove(struct pfi_dynaddr *dyn) 828 { 829 830 KASSERT(dyn->pfid_kif != NULL, ("%s: null pfid_kif", __func__)); 831 KASSERT(dyn->pfid_kt != NULL, ("%s: null pfid_kt", __func__)); 832 833 TAILQ_REMOVE(&dyn->pfid_kif->pfik_dynaddrs, dyn, entry); 834 pfi_kkif_unref(dyn->pfid_kif); 835 pfr_detach_table(dyn->pfid_kt); 836 free(dyn, PFI_MTYPE); 837 } 838 839 void 840 pfi_dynaddr_copyout(struct pf_addr_wrap *aw) 841 { 842 843 KASSERT(aw->type == PF_ADDR_DYNIFTL, 844 ("%s: type %u", __func__, aw->type)); 845 846 if (aw->p.dyn == NULL || aw->p.dyn->pfid_kif == NULL) 847 return; 848 aw->p.dyncnt = aw->p.dyn->pfid_acnt4 + aw->p.dyn->pfid_acnt6; 849 } 850 851 static int 852 pfi_kkif_compare(struct pfi_kkif *p, struct pfi_kkif *q) 853 { 854 return (strncmp(p->pfik_name, q->pfik_name, IFNAMSIZ)); 855 } 856 857 void 858 pfi_update_status(const char *name, struct pf_status *pfs) 859 { 860 struct pfi_kkif *p; 861 struct pfi_kif_cmp key; 862 struct ifg_member p_member, *ifgm; 863 CK_STAILQ_HEAD(, ifg_member) ifg_members; 864 int i, j, k; 865 866 if (pfs) { 867 memset(pfs->pcounters, 0, sizeof(pfs->pcounters)); 868 memset(pfs->bcounters, 0, sizeof(pfs->bcounters)); 869 } 870 871 strlcpy(key.pfik_name, name, sizeof(key.pfik_name)); 872 p = RB_FIND(pfi_ifhead, &V_pfi_ifs, (struct pfi_kkif *)&key); 873 if (p == NULL) { 874 return; 875 } 876 877 if (p->pfik_group != NULL) { 878 memcpy(&ifg_members, &p->pfik_group->ifg_members, 879 sizeof(ifg_members)); 880 } else { 881 /* build a temporary list for p only */ 882 memset(&p_member, 0, sizeof(p_member)); 883 p_member.ifgm_ifp = p->pfik_ifp; 884 CK_STAILQ_INIT(&ifg_members); 885 CK_STAILQ_INSERT_TAIL(&ifg_members, &p_member, ifgm_next); 886 } 887 CK_STAILQ_FOREACH(ifgm, &ifg_members, ifgm_next) { 888 if (ifgm->ifgm_ifp == NULL || ifgm->ifgm_ifp->if_pf_kif == NULL) 889 continue; 890 p = (struct pfi_kkif *)ifgm->ifgm_ifp->if_pf_kif; 891 892 /* just clear statistics */ 893 if (pfs == NULL) { 894 pf_kkif_zero(p); 895 continue; 896 } 897 for (i = 0; i < 2; i++) 898 for (j = 0; j < 2; j++) 899 for (k = 0; k < 2; k++) { 900 pfs->pcounters[i][j][k] += 901 pf_counter_u64_fetch(&p->pfik_packets[i][j][k]); 902 pfs->bcounters[i][j] += 903 pf_counter_u64_fetch(&p->pfik_bytes[i][j][k]); 904 } 905 } 906 } 907 908 static void 909 pf_kkif_to_kif(struct pfi_kkif *kkif, struct pfi_kif *kif) 910 { 911 912 memset(kif, 0, sizeof(*kif)); 913 strlcpy(kif->pfik_name, kkif->pfik_name, sizeof(kif->pfik_name)); 914 for (int i = 0; i < 2; i++) { 915 for (int j = 0; j < 2; j++) { 916 for (int k = 0; k < 2; k++) { 917 kif->pfik_packets[i][j][k] = 918 pf_counter_u64_fetch(&kkif->pfik_packets[i][j][k]); 919 kif->pfik_bytes[i][j][k] = 920 pf_counter_u64_fetch(&kkif->pfik_bytes[i][j][k]); 921 } 922 } 923 } 924 kif->pfik_flags = kkif->pfik_flags; 925 kif->pfik_tzero = kkif->pfik_tzero; 926 kif->pfik_rulerefs = kkif->pfik_rulerefs; 927 /* 928 * Userspace relies on this pointer to decide if this is a group or 929 * not. We don't want to share the actual pointer, because it's 930 * useless to userspace and leaks kernel memory layout information. 931 * So instead we provide 0xfeedcode as 'true' and NULL as 'false'. 932 */ 933 kif->pfik_group = 934 kkif->pfik_group ? (struct ifg_group *)0xfeedc0de : NULL; 935 } 936 937 void 938 pfi_get_ifaces(const char *name, struct pfi_kif *buf, int *size) 939 { 940 struct epoch_tracker et; 941 struct pfi_kkif *p, *nextp; 942 int n = 0; 943 944 NET_EPOCH_ENTER(et); 945 for (p = RB_MIN(pfi_ifhead, &V_pfi_ifs); p; p = nextp) { 946 nextp = RB_NEXT(pfi_ifhead, &V_pfi_ifs, p); 947 if (pfi_skip_if(name, p)) 948 continue; 949 if (*size <= n++) 950 break; 951 if (!p->pfik_tzero) 952 p->pfik_tzero = time_second; 953 pf_kkif_to_kif(p, buf++); 954 nextp = RB_NEXT(pfi_ifhead, &V_pfi_ifs, p); 955 } 956 *size = n; 957 NET_EPOCH_EXIT(et); 958 } 959 960 static int 961 pfi_skip_if(const char *filter, struct pfi_kkif *p) 962 { 963 struct ifg_list *i; 964 int n; 965 966 NET_EPOCH_ASSERT(); 967 968 if (filter == NULL || !*filter) 969 return (0); 970 if (!strcmp(p->pfik_name, filter)) 971 return (0); /* exact match */ 972 n = strlen(filter); 973 if (n < 1 || n >= IFNAMSIZ) 974 return (1); /* sanity check */ 975 if (filter[n-1] >= '0' && filter[n-1] <= '9') 976 return (1); /* group names may not end in a digit */ 977 if (p->pfik_ifp == NULL) 978 return (1); 979 CK_STAILQ_FOREACH(i, &p->pfik_ifp->if_groups, ifgl_next) 980 if (!strncmp(i->ifgl_group->ifg_group, filter, IFNAMSIZ)) 981 return (0); /* iface is in group "filter" */ 982 return (1); 983 } 984 985 int 986 pfi_set_flags(const char *name, int flags) 987 { 988 struct epoch_tracker et; 989 struct pfi_kkif *p, *kif; 990 991 kif = pf_kkif_create(M_NOWAIT); 992 if (kif == NULL) 993 return (ENOMEM); 994 995 NET_EPOCH_ENTER(et); 996 997 kif = pfi_kkif_attach(kif, name); 998 999 RB_FOREACH(p, pfi_ifhead, &V_pfi_ifs) { 1000 if (pfi_skip_if(name, p)) 1001 continue; 1002 p->pfik_flags |= flags; 1003 } 1004 NET_EPOCH_EXIT(et); 1005 return (0); 1006 } 1007 1008 int 1009 pfi_clear_flags(const char *name, int flags) 1010 { 1011 struct epoch_tracker et; 1012 struct pfi_kkif *p, *tmp; 1013 1014 NET_EPOCH_ENTER(et); 1015 RB_FOREACH_SAFE(p, pfi_ifhead, &V_pfi_ifs, tmp) { 1016 if (pfi_skip_if(name, p)) 1017 continue; 1018 p->pfik_flags &= ~flags; 1019 1020 if (p->pfik_ifp == NULL && p->pfik_group == NULL && 1021 p->pfik_flags == 0 && p->pfik_rulerefs == 0) { 1022 /* Delete this kif. */ 1023 RB_REMOVE(pfi_ifhead, &V_pfi_ifs, p); 1024 pf_kkif_free(p); 1025 } 1026 } 1027 NET_EPOCH_EXIT(et); 1028 return (0); 1029 } 1030 1031 /* from pf_print_state.c */ 1032 static int 1033 pfi_unmask(void *addr) 1034 { 1035 struct pf_addr *m = addr; 1036 int i = 31, j = 0, b = 0; 1037 u_int32_t tmp; 1038 1039 while (j < 4 && m->addr32[j] == 0xffffffff) { 1040 b += 32; 1041 j++; 1042 } 1043 if (j < 4) { 1044 tmp = ntohl(m->addr32[j]); 1045 for (i = 31; tmp & (1 << i); --i) 1046 b++; 1047 } 1048 return (b); 1049 } 1050 1051 static void 1052 pfi_attach_ifnet_event(void *arg __unused, struct ifnet *ifp) 1053 { 1054 struct epoch_tracker et; 1055 struct pfi_kkif *kif; 1056 1057 if (V_pf_vnet_active == 0) { 1058 /* Avoid teardown race in the least expensive way. */ 1059 return; 1060 } 1061 kif = pf_kkif_create(M_NOWAIT); 1062 NET_EPOCH_ENTER(et); 1063 PF_RULES_WLOCK(); 1064 pfi_attach_ifnet(ifp, kif); 1065 #ifdef ALTQ 1066 pf_altq_ifnet_event(ifp, 0); 1067 #endif 1068 PF_RULES_WUNLOCK(); 1069 NET_EPOCH_EXIT(et); 1070 } 1071 1072 static void 1073 pfi_detach_ifnet_event(void *arg __unused, struct ifnet *ifp) 1074 { 1075 struct epoch_tracker et; 1076 struct pfi_kkif *kif = (struct pfi_kkif *)ifp->if_pf_kif; 1077 1078 if (pfsync_detach_ifnet_ptr) 1079 pfsync_detach_ifnet_ptr(ifp); 1080 1081 if (kif == NULL) 1082 return; 1083 1084 if (V_pf_vnet_active == 0) { 1085 /* Avoid teardown race in the least expensive way. */ 1086 return; 1087 } 1088 1089 NET_EPOCH_ENTER(et); 1090 PF_RULES_WLOCK(); 1091 V_pfi_update++; 1092 pfi_kkif_update(kif); 1093 1094 if (kif->pfik_ifp) 1095 if_rele(kif->pfik_ifp); 1096 1097 kif->pfik_ifp = NULL; 1098 ifp->if_pf_kif = NULL; 1099 #ifdef ALTQ 1100 pf_altq_ifnet_event(ifp, 1); 1101 #endif 1102 pfi_kkif_remove_if_unref(kif); 1103 1104 PF_RULES_WUNLOCK(); 1105 NET_EPOCH_EXIT(et); 1106 } 1107 1108 static void 1109 pfi_attach_group_event(void *arg __unused, struct ifg_group *ifg) 1110 { 1111 struct epoch_tracker et; 1112 struct pfi_kkif *kif; 1113 1114 if (V_pf_vnet_active == 0) { 1115 /* Avoid teardown race in the least expensive way. */ 1116 return; 1117 } 1118 kif = pf_kkif_create(M_WAITOK); 1119 NET_EPOCH_ENTER(et); 1120 PF_RULES_WLOCK(); 1121 pfi_attach_ifgroup(ifg, kif); 1122 PF_RULES_WUNLOCK(); 1123 NET_EPOCH_EXIT(et); 1124 } 1125 1126 static void 1127 pfi_change_group_event(void *arg __unused, char *gname) 1128 { 1129 struct epoch_tracker et; 1130 struct pfi_kkif *kif; 1131 1132 if (V_pf_vnet_active == 0) { 1133 /* Avoid teardown race in the least expensive way. */ 1134 return; 1135 } 1136 1137 kif = pf_kkif_create(M_WAITOK); 1138 NET_EPOCH_ENTER(et); 1139 PF_RULES_WLOCK(); 1140 V_pfi_update++; 1141 kif = pfi_kkif_attach(kif, gname); 1142 pfi_kkif_update(kif); 1143 PF_RULES_WUNLOCK(); 1144 NET_EPOCH_EXIT(et); 1145 } 1146 1147 static void 1148 pfi_detach_group_event(void *arg __unused, struct ifg_group *ifg) 1149 { 1150 struct pfi_kkif *kif = (struct pfi_kkif *)ifg->ifg_pf_kif; 1151 1152 if (kif == NULL) 1153 return; 1154 1155 if (V_pf_vnet_active == 0) { 1156 /* Avoid teardown race in the least expensive way. */ 1157 return; 1158 } 1159 PF_RULES_WLOCK(); 1160 V_pfi_update++; 1161 1162 kif->pfik_group = NULL; 1163 ifg->ifg_pf_kif = NULL; 1164 1165 pfi_kkif_remove_if_unref(kif); 1166 PF_RULES_WUNLOCK(); 1167 } 1168 1169 static void 1170 pfi_ifaddr_event(void *arg __unused, struct ifnet *ifp) 1171 { 1172 1173 KASSERT(ifp, ("ifp == NULL")); 1174 1175 if (ifp->if_pf_kif == NULL) 1176 return; 1177 1178 if (V_pf_vnet_active == 0) { 1179 /* Avoid teardown race in the least expensive way. */ 1180 return; 1181 } 1182 PF_RULES_WLOCK(); 1183 if (ifp->if_pf_kif) { 1184 struct epoch_tracker et; 1185 1186 V_pfi_update++; 1187 NET_EPOCH_ENTER(et); 1188 pfi_kkif_update(ifp->if_pf_kif); 1189 NET_EPOCH_EXIT(et); 1190 } 1191 PF_RULES_WUNLOCK(); 1192 } 1193