1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * xfrm_policy.c 4 * 5 * Changes: 6 * Mitsuru KANDA @USAGI 7 * Kazunori MIYAZAWA @USAGI 8 * Kunihiro Ishiguro <kunihiro@ipinfusion.com> 9 * IPv6 support 10 * Kazunori MIYAZAWA @USAGI 11 * YOSHIFUJI Hideaki 12 * Split up af-specific portion 13 * Derek Atkins <derek@ihtfp.com> Add the post_input processor 14 * 15 */ 16 17 #include <linux/err.h> 18 #include <linux/slab.h> 19 #include <linux/kmod.h> 20 #include <linux/list.h> 21 #include <linux/spinlock.h> 22 #include <linux/workqueue.h> 23 #include <linux/notifier.h> 24 #include <linux/netdevice.h> 25 #include <linux/netfilter.h> 26 #include <linux/module.h> 27 #include <linux/cache.h> 28 #include <linux/cpu.h> 29 #include <linux/audit.h> 30 #include <linux/rhashtable.h> 31 #include <linux/if_tunnel.h> 32 #include <net/dst.h> 33 #include <net/flow.h> 34 #include <net/inet_ecn.h> 35 #include <net/xfrm.h> 36 #include <net/ip.h> 37 #include <net/gre.h> 38 #if IS_ENABLED(CONFIG_IPV6_MIP6) 39 #include <net/mip6.h> 40 #endif 41 #ifdef CONFIG_XFRM_STATISTICS 42 #include <net/snmp.h> 43 #endif 44 #ifdef CONFIG_XFRM_ESPINTCP 45 #include <net/espintcp.h> 46 #endif 47 48 #include "xfrm_hash.h" 49 50 #define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10)) 51 #define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ)) 52 #define XFRM_MAX_QUEUE_LEN 100 53 54 struct xfrm_flo { 55 struct dst_entry *dst_orig; 56 u8 flags; 57 }; 58 59 /* prefixes smaller than this are stored in lists, not trees. */ 60 #define INEXACT_PREFIXLEN_IPV4 16 61 #define INEXACT_PREFIXLEN_IPV6 48 62 63 struct xfrm_pol_inexact_node { 64 struct rb_node node; 65 union { 66 xfrm_address_t addr; 67 struct rcu_head rcu; 68 }; 69 u8 prefixlen; 70 71 struct rb_root root; 72 73 /* the policies matching this node, can be empty list */ 74 struct hlist_head hhead; 75 }; 76 77 /* xfrm inexact policy search tree: 78 * xfrm_pol_inexact_bin = hash(dir,type,family,if_id); 79 * | 80 * +---- root_d: sorted by daddr:prefix 81 * | | 82 * | xfrm_pol_inexact_node 83 * | | 84 * | +- root: sorted by saddr/prefix 85 * | | | 86 * | | xfrm_pol_inexact_node 87 * | | | 88 * | | + root: unused 89 * | | | 90 * | | + hhead: saddr:daddr policies 91 * | | 92 * | +- coarse policies and all any:daddr policies 93 * | 94 * +---- root_s: sorted by saddr:prefix 95 * | | 96 * | xfrm_pol_inexact_node 97 * | | 98 * | + root: unused 99 * | | 100 * | + hhead: saddr:any policies 101 * | 102 * +---- coarse policies and all any:any policies 103 * 104 * Lookups return four candidate lists: 105 * 1. any:any list from top-level xfrm_pol_inexact_bin 106 * 2. any:daddr list from daddr tree 107 * 3. saddr:daddr list from 2nd level daddr tree 108 * 4. saddr:any list from saddr tree 109 * 110 * This result set then needs to be searched for the policy with 111 * the lowest priority. If two results have same prio, youngest one wins. 112 */ 113 114 struct xfrm_pol_inexact_key { 115 possible_net_t net; 116 u32 if_id; 117 u16 family; 118 u8 dir, type; 119 }; 120 121 struct xfrm_pol_inexact_bin { 122 struct xfrm_pol_inexact_key k; 123 struct rhash_head head; 124 /* list containing '*:*' policies */ 125 struct hlist_head hhead; 126 127 seqcount_spinlock_t count; 128 /* tree sorted by daddr/prefix */ 129 struct rb_root root_d; 130 131 /* tree sorted by saddr/prefix */ 132 struct rb_root root_s; 133 134 /* slow path below */ 135 struct list_head inexact_bins; 136 struct rcu_head rcu; 137 }; 138 139 enum xfrm_pol_inexact_candidate_type { 140 XFRM_POL_CAND_BOTH, 141 XFRM_POL_CAND_SADDR, 142 XFRM_POL_CAND_DADDR, 143 XFRM_POL_CAND_ANY, 144 145 XFRM_POL_CAND_MAX, 146 }; 147 148 struct xfrm_pol_inexact_candidates { 149 struct hlist_head *res[XFRM_POL_CAND_MAX]; 150 }; 151 152 struct xfrm_flow_keys { 153 struct flow_dissector_key_basic basic; 154 struct flow_dissector_key_control control; 155 union { 156 struct flow_dissector_key_ipv4_addrs ipv4; 157 struct flow_dissector_key_ipv6_addrs ipv6; 158 } addrs; 159 struct flow_dissector_key_ip ip; 160 struct flow_dissector_key_icmp icmp; 161 struct flow_dissector_key_ports ports; 162 struct flow_dissector_key_keyid gre; 163 }; 164 165 static struct flow_dissector xfrm_session_dissector __ro_after_init; 166 167 static DEFINE_SPINLOCK(xfrm_if_cb_lock); 168 static struct xfrm_if_cb const __rcu *xfrm_if_cb __read_mostly; 169 170 static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock); 171 static struct xfrm_policy_afinfo const __rcu *xfrm_policy_afinfo[AF_INET6 + 1] 172 __read_mostly; 173 174 static struct kmem_cache *xfrm_dst_cache __ro_after_init; 175 176 static struct rhashtable xfrm_policy_inexact_table; 177 static const struct rhashtable_params xfrm_pol_inexact_params; 178 179 static void xfrm_init_pmtu(struct xfrm_dst **bundle, int nr); 180 static int stale_bundle(struct dst_entry *dst); 181 static int xfrm_bundle_ok(struct xfrm_dst *xdst); 182 static void xfrm_policy_queue_process(struct timer_list *t); 183 184 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir); 185 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol, 186 int dir); 187 188 static struct xfrm_pol_inexact_bin * 189 xfrm_policy_inexact_lookup(struct net *net, u8 type, u16 family, u8 dir, 190 u32 if_id); 191 192 static struct xfrm_pol_inexact_bin * 193 xfrm_policy_inexact_lookup_rcu(struct net *net, 194 u8 type, u16 family, u8 dir, u32 if_id); 195 static struct xfrm_policy * 196 xfrm_policy_insert_list(struct hlist_head *chain, struct xfrm_policy *policy, 197 bool excl); 198 static void xfrm_policy_insert_inexact_list(struct hlist_head *chain, 199 struct xfrm_policy *policy); 200 201 static bool 202 xfrm_policy_find_inexact_candidates(struct xfrm_pol_inexact_candidates *cand, 203 struct xfrm_pol_inexact_bin *b, 204 const xfrm_address_t *saddr, 205 const xfrm_address_t *daddr); 206 207 static inline bool xfrm_pol_hold_rcu(struct xfrm_policy *policy) 208 { 209 return refcount_inc_not_zero(&policy->refcnt); 210 } 211 212 static inline bool 213 __xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl) 214 { 215 const struct flowi4 *fl4 = &fl->u.ip4; 216 217 return addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) && 218 addr4_match(fl4->saddr, sel->saddr.a4, sel->prefixlen_s) && 219 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) && 220 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) && 221 (fl4->flowi4_proto == sel->proto || !sel->proto) && 222 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex); 223 } 224 225 static inline bool 226 __xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl) 227 { 228 const struct flowi6 *fl6 = &fl->u.ip6; 229 230 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) && 231 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) && 232 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) && 233 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) && 234 (fl6->flowi6_proto == sel->proto || !sel->proto) && 235 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex); 236 } 237 238 bool xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl, 239 unsigned short family) 240 { 241 switch (family) { 242 case AF_INET: 243 return __xfrm4_selector_match(sel, fl); 244 case AF_INET6: 245 return __xfrm6_selector_match(sel, fl); 246 } 247 return false; 248 } 249 250 static const struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family) 251 { 252 const struct xfrm_policy_afinfo *afinfo; 253 254 if (unlikely(family >= ARRAY_SIZE(xfrm_policy_afinfo))) 255 return NULL; 256 rcu_read_lock(); 257 afinfo = rcu_dereference(xfrm_policy_afinfo[family]); 258 if (unlikely(!afinfo)) 259 rcu_read_unlock(); 260 return afinfo; 261 } 262 263 /* Called with rcu_read_lock(). */ 264 static const struct xfrm_if_cb *xfrm_if_get_cb(void) 265 { 266 return rcu_dereference(xfrm_if_cb); 267 } 268 269 struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos, int oif, 270 const xfrm_address_t *saddr, 271 const xfrm_address_t *daddr, 272 int family, u32 mark) 273 { 274 const struct xfrm_policy_afinfo *afinfo; 275 struct dst_entry *dst; 276 277 afinfo = xfrm_policy_get_afinfo(family); 278 if (unlikely(afinfo == NULL)) 279 return ERR_PTR(-EAFNOSUPPORT); 280 281 dst = afinfo->dst_lookup(net, tos, oif, saddr, daddr, mark); 282 283 rcu_read_unlock(); 284 285 return dst; 286 } 287 EXPORT_SYMBOL(__xfrm_dst_lookup); 288 289 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, 290 int tos, int oif, 291 xfrm_address_t *prev_saddr, 292 xfrm_address_t *prev_daddr, 293 int family, u32 mark) 294 { 295 struct net *net = xs_net(x); 296 xfrm_address_t *saddr = &x->props.saddr; 297 xfrm_address_t *daddr = &x->id.daddr; 298 struct dst_entry *dst; 299 300 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) { 301 saddr = x->coaddr; 302 daddr = prev_daddr; 303 } 304 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) { 305 saddr = prev_saddr; 306 daddr = x->coaddr; 307 } 308 309 dst = __xfrm_dst_lookup(net, tos, oif, saddr, daddr, family, mark); 310 311 if (!IS_ERR(dst)) { 312 if (prev_saddr != saddr) 313 memcpy(prev_saddr, saddr, sizeof(*prev_saddr)); 314 if (prev_daddr != daddr) 315 memcpy(prev_daddr, daddr, sizeof(*prev_daddr)); 316 } 317 318 return dst; 319 } 320 321 static inline unsigned long make_jiffies(long secs) 322 { 323 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ) 324 return MAX_SCHEDULE_TIMEOUT-1; 325 else 326 return secs*HZ; 327 } 328 329 static void xfrm_policy_timer(struct timer_list *t) 330 { 331 struct xfrm_policy *xp = from_timer(xp, t, timer); 332 time64_t now = ktime_get_real_seconds(); 333 time64_t next = TIME64_MAX; 334 int warn = 0; 335 int dir; 336 337 read_lock(&xp->lock); 338 339 if (unlikely(xp->walk.dead)) 340 goto out; 341 342 dir = xfrm_policy_id2dir(xp->index); 343 344 if (xp->lft.hard_add_expires_seconds) { 345 time64_t tmo = xp->lft.hard_add_expires_seconds + 346 xp->curlft.add_time - now; 347 if (tmo <= 0) 348 goto expired; 349 if (tmo < next) 350 next = tmo; 351 } 352 if (xp->lft.hard_use_expires_seconds) { 353 time64_t tmo = xp->lft.hard_use_expires_seconds + 354 (READ_ONCE(xp->curlft.use_time) ? : xp->curlft.add_time) - now; 355 if (tmo <= 0) 356 goto expired; 357 if (tmo < next) 358 next = tmo; 359 } 360 if (xp->lft.soft_add_expires_seconds) { 361 time64_t tmo = xp->lft.soft_add_expires_seconds + 362 xp->curlft.add_time - now; 363 if (tmo <= 0) { 364 warn = 1; 365 tmo = XFRM_KM_TIMEOUT; 366 } 367 if (tmo < next) 368 next = tmo; 369 } 370 if (xp->lft.soft_use_expires_seconds) { 371 time64_t tmo = xp->lft.soft_use_expires_seconds + 372 (READ_ONCE(xp->curlft.use_time) ? : xp->curlft.add_time) - now; 373 if (tmo <= 0) { 374 warn = 1; 375 tmo = XFRM_KM_TIMEOUT; 376 } 377 if (tmo < next) 378 next = tmo; 379 } 380 381 if (warn) 382 km_policy_expired(xp, dir, 0, 0); 383 if (next != TIME64_MAX && 384 !mod_timer(&xp->timer, jiffies + make_jiffies(next))) 385 xfrm_pol_hold(xp); 386 387 out: 388 read_unlock(&xp->lock); 389 xfrm_pol_put(xp); 390 return; 391 392 expired: 393 read_unlock(&xp->lock); 394 if (!xfrm_policy_delete(xp, dir)) 395 km_policy_expired(xp, dir, 1, 0); 396 xfrm_pol_put(xp); 397 } 398 399 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2 400 * SPD calls. 401 */ 402 403 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp) 404 { 405 struct xfrm_policy *policy; 406 407 policy = kzalloc(sizeof(struct xfrm_policy), gfp); 408 409 if (policy) { 410 write_pnet(&policy->xp_net, net); 411 INIT_LIST_HEAD(&policy->walk.all); 412 INIT_HLIST_NODE(&policy->bydst_inexact_list); 413 INIT_HLIST_NODE(&policy->bydst); 414 INIT_HLIST_NODE(&policy->byidx); 415 rwlock_init(&policy->lock); 416 refcount_set(&policy->refcnt, 1); 417 skb_queue_head_init(&policy->polq.hold_queue); 418 timer_setup(&policy->timer, xfrm_policy_timer, 0); 419 timer_setup(&policy->polq.hold_timer, 420 xfrm_policy_queue_process, 0); 421 } 422 return policy; 423 } 424 EXPORT_SYMBOL(xfrm_policy_alloc); 425 426 static void xfrm_policy_destroy_rcu(struct rcu_head *head) 427 { 428 struct xfrm_policy *policy = container_of(head, struct xfrm_policy, rcu); 429 430 security_xfrm_policy_free(policy->security); 431 kfree(policy); 432 } 433 434 /* Destroy xfrm_policy: descendant resources must be released to this moment. */ 435 436 void xfrm_policy_destroy(struct xfrm_policy *policy) 437 { 438 BUG_ON(!policy->walk.dead); 439 440 if (del_timer(&policy->timer) || del_timer(&policy->polq.hold_timer)) 441 BUG(); 442 443 xfrm_dev_policy_free(policy); 444 call_rcu(&policy->rcu, xfrm_policy_destroy_rcu); 445 } 446 EXPORT_SYMBOL(xfrm_policy_destroy); 447 448 /* Rule must be locked. Release descendant resources, announce 449 * entry dead. The rule must be unlinked from lists to the moment. 450 */ 451 452 static void xfrm_policy_kill(struct xfrm_policy *policy) 453 { 454 write_lock_bh(&policy->lock); 455 policy->walk.dead = 1; 456 write_unlock_bh(&policy->lock); 457 458 atomic_inc(&policy->genid); 459 460 if (del_timer(&policy->polq.hold_timer)) 461 xfrm_pol_put(policy); 462 skb_queue_purge(&policy->polq.hold_queue); 463 464 if (del_timer(&policy->timer)) 465 xfrm_pol_put(policy); 466 467 xfrm_pol_put(policy); 468 } 469 470 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024; 471 472 static inline unsigned int idx_hash(struct net *net, u32 index) 473 { 474 return __idx_hash(index, net->xfrm.policy_idx_hmask); 475 } 476 477 /* calculate policy hash thresholds */ 478 static void __get_hash_thresh(struct net *net, 479 unsigned short family, int dir, 480 u8 *dbits, u8 *sbits) 481 { 482 switch (family) { 483 case AF_INET: 484 *dbits = net->xfrm.policy_bydst[dir].dbits4; 485 *sbits = net->xfrm.policy_bydst[dir].sbits4; 486 break; 487 488 case AF_INET6: 489 *dbits = net->xfrm.policy_bydst[dir].dbits6; 490 *sbits = net->xfrm.policy_bydst[dir].sbits6; 491 break; 492 493 default: 494 *dbits = 0; 495 *sbits = 0; 496 } 497 } 498 499 static struct hlist_head *policy_hash_bysel(struct net *net, 500 const struct xfrm_selector *sel, 501 unsigned short family, int dir) 502 { 503 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask; 504 unsigned int hash; 505 u8 dbits; 506 u8 sbits; 507 508 __get_hash_thresh(net, family, dir, &dbits, &sbits); 509 hash = __sel_hash(sel, family, hmask, dbits, sbits); 510 511 if (hash == hmask + 1) 512 return NULL; 513 514 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table, 515 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash; 516 } 517 518 static struct hlist_head *policy_hash_direct(struct net *net, 519 const xfrm_address_t *daddr, 520 const xfrm_address_t *saddr, 521 unsigned short family, int dir) 522 { 523 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask; 524 unsigned int hash; 525 u8 dbits; 526 u8 sbits; 527 528 __get_hash_thresh(net, family, dir, &dbits, &sbits); 529 hash = __addr_hash(daddr, saddr, family, hmask, dbits, sbits); 530 531 return rcu_dereference_check(net->xfrm.policy_bydst[dir].table, 532 lockdep_is_held(&net->xfrm.xfrm_policy_lock)) + hash; 533 } 534 535 static void xfrm_dst_hash_transfer(struct net *net, 536 struct hlist_head *list, 537 struct hlist_head *ndsttable, 538 unsigned int nhashmask, 539 int dir) 540 { 541 struct hlist_node *tmp, *entry0 = NULL; 542 struct xfrm_policy *pol; 543 unsigned int h0 = 0; 544 u8 dbits; 545 u8 sbits; 546 547 redo: 548 hlist_for_each_entry_safe(pol, tmp, list, bydst) { 549 unsigned int h; 550 551 __get_hash_thresh(net, pol->family, dir, &dbits, &sbits); 552 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr, 553 pol->family, nhashmask, dbits, sbits); 554 if (!entry0 || pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) { 555 hlist_del_rcu(&pol->bydst); 556 hlist_add_head_rcu(&pol->bydst, ndsttable + h); 557 h0 = h; 558 } else { 559 if (h != h0) 560 continue; 561 hlist_del_rcu(&pol->bydst); 562 hlist_add_behind_rcu(&pol->bydst, entry0); 563 } 564 entry0 = &pol->bydst; 565 } 566 if (!hlist_empty(list)) { 567 entry0 = NULL; 568 goto redo; 569 } 570 } 571 572 static void xfrm_idx_hash_transfer(struct hlist_head *list, 573 struct hlist_head *nidxtable, 574 unsigned int nhashmask) 575 { 576 struct hlist_node *tmp; 577 struct xfrm_policy *pol; 578 579 hlist_for_each_entry_safe(pol, tmp, list, byidx) { 580 unsigned int h; 581 582 h = __idx_hash(pol->index, nhashmask); 583 hlist_add_head(&pol->byidx, nidxtable+h); 584 } 585 } 586 587 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask) 588 { 589 return ((old_hmask + 1) << 1) - 1; 590 } 591 592 static void xfrm_bydst_resize(struct net *net, int dir) 593 { 594 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask; 595 unsigned int nhashmask = xfrm_new_hash_mask(hmask); 596 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head); 597 struct hlist_head *ndst = xfrm_hash_alloc(nsize); 598 struct hlist_head *odst; 599 int i; 600 601 if (!ndst) 602 return; 603 604 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 605 write_seqcount_begin(&net->xfrm.xfrm_policy_hash_generation); 606 607 odst = rcu_dereference_protected(net->xfrm.policy_bydst[dir].table, 608 lockdep_is_held(&net->xfrm.xfrm_policy_lock)); 609 610 for (i = hmask; i >= 0; i--) 611 xfrm_dst_hash_transfer(net, odst + i, ndst, nhashmask, dir); 612 613 rcu_assign_pointer(net->xfrm.policy_bydst[dir].table, ndst); 614 net->xfrm.policy_bydst[dir].hmask = nhashmask; 615 616 write_seqcount_end(&net->xfrm.xfrm_policy_hash_generation); 617 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 618 619 synchronize_rcu(); 620 621 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head)); 622 } 623 624 static void xfrm_byidx_resize(struct net *net) 625 { 626 unsigned int hmask = net->xfrm.policy_idx_hmask; 627 unsigned int nhashmask = xfrm_new_hash_mask(hmask); 628 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head); 629 struct hlist_head *oidx = net->xfrm.policy_byidx; 630 struct hlist_head *nidx = xfrm_hash_alloc(nsize); 631 int i; 632 633 if (!nidx) 634 return; 635 636 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 637 638 for (i = hmask; i >= 0; i--) 639 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask); 640 641 net->xfrm.policy_byidx = nidx; 642 net->xfrm.policy_idx_hmask = nhashmask; 643 644 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 645 646 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head)); 647 } 648 649 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total) 650 { 651 unsigned int cnt = net->xfrm.policy_count[dir]; 652 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask; 653 654 if (total) 655 *total += cnt; 656 657 if ((hmask + 1) < xfrm_policy_hashmax && 658 cnt > hmask) 659 return 1; 660 661 return 0; 662 } 663 664 static inline int xfrm_byidx_should_resize(struct net *net, int total) 665 { 666 unsigned int hmask = net->xfrm.policy_idx_hmask; 667 668 if ((hmask + 1) < xfrm_policy_hashmax && 669 total > hmask) 670 return 1; 671 672 return 0; 673 } 674 675 void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si) 676 { 677 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN]; 678 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT]; 679 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD]; 680 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX]; 681 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX]; 682 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX]; 683 si->spdhcnt = net->xfrm.policy_idx_hmask; 684 si->spdhmcnt = xfrm_policy_hashmax; 685 } 686 EXPORT_SYMBOL(xfrm_spd_getinfo); 687 688 static DEFINE_MUTEX(hash_resize_mutex); 689 static void xfrm_hash_resize(struct work_struct *work) 690 { 691 struct net *net = container_of(work, struct net, xfrm.policy_hash_work); 692 int dir, total; 693 694 mutex_lock(&hash_resize_mutex); 695 696 total = 0; 697 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) { 698 if (xfrm_bydst_should_resize(net, dir, &total)) 699 xfrm_bydst_resize(net, dir); 700 } 701 if (xfrm_byidx_should_resize(net, total)) 702 xfrm_byidx_resize(net); 703 704 mutex_unlock(&hash_resize_mutex); 705 } 706 707 /* Make sure *pol can be inserted into fastbin. 708 * Useful to check that later insert requests will be successful 709 * (provided xfrm_policy_lock is held throughout). 710 */ 711 static struct xfrm_pol_inexact_bin * 712 xfrm_policy_inexact_alloc_bin(const struct xfrm_policy *pol, u8 dir) 713 { 714 struct xfrm_pol_inexact_bin *bin, *prev; 715 struct xfrm_pol_inexact_key k = { 716 .family = pol->family, 717 .type = pol->type, 718 .dir = dir, 719 .if_id = pol->if_id, 720 }; 721 struct net *net = xp_net(pol); 722 723 lockdep_assert_held(&net->xfrm.xfrm_policy_lock); 724 725 write_pnet(&k.net, net); 726 bin = rhashtable_lookup_fast(&xfrm_policy_inexact_table, &k, 727 xfrm_pol_inexact_params); 728 if (bin) 729 return bin; 730 731 bin = kzalloc(sizeof(*bin), GFP_ATOMIC); 732 if (!bin) 733 return NULL; 734 735 bin->k = k; 736 INIT_HLIST_HEAD(&bin->hhead); 737 bin->root_d = RB_ROOT; 738 bin->root_s = RB_ROOT; 739 seqcount_spinlock_init(&bin->count, &net->xfrm.xfrm_policy_lock); 740 741 prev = rhashtable_lookup_get_insert_key(&xfrm_policy_inexact_table, 742 &bin->k, &bin->head, 743 xfrm_pol_inexact_params); 744 if (!prev) { 745 list_add(&bin->inexact_bins, &net->xfrm.inexact_bins); 746 return bin; 747 } 748 749 kfree(bin); 750 751 return IS_ERR(prev) ? NULL : prev; 752 } 753 754 static bool xfrm_pol_inexact_addr_use_any_list(const xfrm_address_t *addr, 755 int family, u8 prefixlen) 756 { 757 if (xfrm_addr_any(addr, family)) 758 return true; 759 760 if (family == AF_INET6 && prefixlen < INEXACT_PREFIXLEN_IPV6) 761 return true; 762 763 if (family == AF_INET && prefixlen < INEXACT_PREFIXLEN_IPV4) 764 return true; 765 766 return false; 767 } 768 769 static bool 770 xfrm_policy_inexact_insert_use_any_list(const struct xfrm_policy *policy) 771 { 772 const xfrm_address_t *addr; 773 bool saddr_any, daddr_any; 774 u8 prefixlen; 775 776 addr = &policy->selector.saddr; 777 prefixlen = policy->selector.prefixlen_s; 778 779 saddr_any = xfrm_pol_inexact_addr_use_any_list(addr, 780 policy->family, 781 prefixlen); 782 addr = &policy->selector.daddr; 783 prefixlen = policy->selector.prefixlen_d; 784 daddr_any = xfrm_pol_inexact_addr_use_any_list(addr, 785 policy->family, 786 prefixlen); 787 return saddr_any && daddr_any; 788 } 789 790 static void xfrm_pol_inexact_node_init(struct xfrm_pol_inexact_node *node, 791 const xfrm_address_t *addr, u8 prefixlen) 792 { 793 node->addr = *addr; 794 node->prefixlen = prefixlen; 795 } 796 797 static struct xfrm_pol_inexact_node * 798 xfrm_pol_inexact_node_alloc(const xfrm_address_t *addr, u8 prefixlen) 799 { 800 struct xfrm_pol_inexact_node *node; 801 802 node = kzalloc(sizeof(*node), GFP_ATOMIC); 803 if (node) 804 xfrm_pol_inexact_node_init(node, addr, prefixlen); 805 806 return node; 807 } 808 809 static int xfrm_policy_addr_delta(const xfrm_address_t *a, 810 const xfrm_address_t *b, 811 u8 prefixlen, u16 family) 812 { 813 u32 ma, mb, mask; 814 unsigned int pdw, pbi; 815 int delta = 0; 816 817 switch (family) { 818 case AF_INET: 819 if (prefixlen == 0) 820 return 0; 821 mask = ~0U << (32 - prefixlen); 822 ma = ntohl(a->a4) & mask; 823 mb = ntohl(b->a4) & mask; 824 if (ma < mb) 825 delta = -1; 826 else if (ma > mb) 827 delta = 1; 828 break; 829 case AF_INET6: 830 pdw = prefixlen >> 5; 831 pbi = prefixlen & 0x1f; 832 833 if (pdw) { 834 delta = memcmp(a->a6, b->a6, pdw << 2); 835 if (delta) 836 return delta; 837 } 838 if (pbi) { 839 mask = ~0U << (32 - pbi); 840 ma = ntohl(a->a6[pdw]) & mask; 841 mb = ntohl(b->a6[pdw]) & mask; 842 if (ma < mb) 843 delta = -1; 844 else if (ma > mb) 845 delta = 1; 846 } 847 break; 848 default: 849 break; 850 } 851 852 return delta; 853 } 854 855 static void xfrm_policy_inexact_list_reinsert(struct net *net, 856 struct xfrm_pol_inexact_node *n, 857 u16 family) 858 { 859 unsigned int matched_s, matched_d; 860 struct xfrm_policy *policy, *p; 861 862 matched_s = 0; 863 matched_d = 0; 864 865 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) { 866 struct hlist_node *newpos = NULL; 867 bool matches_s, matches_d; 868 869 if (policy->walk.dead || !policy->bydst_reinsert) 870 continue; 871 872 WARN_ON_ONCE(policy->family != family); 873 874 policy->bydst_reinsert = false; 875 hlist_for_each_entry(p, &n->hhead, bydst) { 876 if (policy->priority > p->priority) 877 newpos = &p->bydst; 878 else if (policy->priority == p->priority && 879 policy->pos > p->pos) 880 newpos = &p->bydst; 881 else 882 break; 883 } 884 885 if (newpos && policy->xdo.type != XFRM_DEV_OFFLOAD_PACKET) 886 hlist_add_behind_rcu(&policy->bydst, newpos); 887 else 888 hlist_add_head_rcu(&policy->bydst, &n->hhead); 889 890 /* paranoia checks follow. 891 * Check that the reinserted policy matches at least 892 * saddr or daddr for current node prefix. 893 * 894 * Matching both is fine, matching saddr in one policy 895 * (but not daddr) and then matching only daddr in another 896 * is a bug. 897 */ 898 matches_s = xfrm_policy_addr_delta(&policy->selector.saddr, 899 &n->addr, 900 n->prefixlen, 901 family) == 0; 902 matches_d = xfrm_policy_addr_delta(&policy->selector.daddr, 903 &n->addr, 904 n->prefixlen, 905 family) == 0; 906 if (matches_s && matches_d) 907 continue; 908 909 WARN_ON_ONCE(!matches_s && !matches_d); 910 if (matches_s) 911 matched_s++; 912 if (matches_d) 913 matched_d++; 914 WARN_ON_ONCE(matched_s && matched_d); 915 } 916 } 917 918 static void xfrm_policy_inexact_node_reinsert(struct net *net, 919 struct xfrm_pol_inexact_node *n, 920 struct rb_root *new, 921 u16 family) 922 { 923 struct xfrm_pol_inexact_node *node; 924 struct rb_node **p, *parent; 925 926 /* we should not have another subtree here */ 927 WARN_ON_ONCE(!RB_EMPTY_ROOT(&n->root)); 928 restart: 929 parent = NULL; 930 p = &new->rb_node; 931 while (*p) { 932 u8 prefixlen; 933 int delta; 934 935 parent = *p; 936 node = rb_entry(*p, struct xfrm_pol_inexact_node, node); 937 938 prefixlen = min(node->prefixlen, n->prefixlen); 939 940 delta = xfrm_policy_addr_delta(&n->addr, &node->addr, 941 prefixlen, family); 942 if (delta < 0) { 943 p = &parent->rb_left; 944 } else if (delta > 0) { 945 p = &parent->rb_right; 946 } else { 947 bool same_prefixlen = node->prefixlen == n->prefixlen; 948 struct xfrm_policy *tmp; 949 950 hlist_for_each_entry(tmp, &n->hhead, bydst) { 951 tmp->bydst_reinsert = true; 952 hlist_del_rcu(&tmp->bydst); 953 } 954 955 node->prefixlen = prefixlen; 956 957 xfrm_policy_inexact_list_reinsert(net, node, family); 958 959 if (same_prefixlen) { 960 kfree_rcu(n, rcu); 961 return; 962 } 963 964 rb_erase(*p, new); 965 kfree_rcu(n, rcu); 966 n = node; 967 goto restart; 968 } 969 } 970 971 rb_link_node_rcu(&n->node, parent, p); 972 rb_insert_color(&n->node, new); 973 } 974 975 /* merge nodes v and n */ 976 static void xfrm_policy_inexact_node_merge(struct net *net, 977 struct xfrm_pol_inexact_node *v, 978 struct xfrm_pol_inexact_node *n, 979 u16 family) 980 { 981 struct xfrm_pol_inexact_node *node; 982 struct xfrm_policy *tmp; 983 struct rb_node *rnode; 984 985 /* To-be-merged node v has a subtree. 986 * 987 * Dismantle it and insert its nodes to n->root. 988 */ 989 while ((rnode = rb_first(&v->root)) != NULL) { 990 node = rb_entry(rnode, struct xfrm_pol_inexact_node, node); 991 rb_erase(&node->node, &v->root); 992 xfrm_policy_inexact_node_reinsert(net, node, &n->root, 993 family); 994 } 995 996 hlist_for_each_entry(tmp, &v->hhead, bydst) { 997 tmp->bydst_reinsert = true; 998 hlist_del_rcu(&tmp->bydst); 999 } 1000 1001 xfrm_policy_inexact_list_reinsert(net, n, family); 1002 } 1003 1004 static struct xfrm_pol_inexact_node * 1005 xfrm_policy_inexact_insert_node(struct net *net, 1006 struct rb_root *root, 1007 xfrm_address_t *addr, 1008 u16 family, u8 prefixlen, u8 dir) 1009 { 1010 struct xfrm_pol_inexact_node *cached = NULL; 1011 struct rb_node **p, *parent = NULL; 1012 struct xfrm_pol_inexact_node *node; 1013 1014 p = &root->rb_node; 1015 while (*p) { 1016 int delta; 1017 1018 parent = *p; 1019 node = rb_entry(*p, struct xfrm_pol_inexact_node, node); 1020 1021 delta = xfrm_policy_addr_delta(addr, &node->addr, 1022 node->prefixlen, 1023 family); 1024 if (delta == 0 && prefixlen >= node->prefixlen) { 1025 WARN_ON_ONCE(cached); /* ipsec policies got lost */ 1026 return node; 1027 } 1028 1029 if (delta < 0) 1030 p = &parent->rb_left; 1031 else 1032 p = &parent->rb_right; 1033 1034 if (prefixlen < node->prefixlen) { 1035 delta = xfrm_policy_addr_delta(addr, &node->addr, 1036 prefixlen, 1037 family); 1038 if (delta) 1039 continue; 1040 1041 /* This node is a subnet of the new prefix. It needs 1042 * to be removed and re-inserted with the smaller 1043 * prefix and all nodes that are now also covered 1044 * by the reduced prefixlen. 1045 */ 1046 rb_erase(&node->node, root); 1047 1048 if (!cached) { 1049 xfrm_pol_inexact_node_init(node, addr, 1050 prefixlen); 1051 cached = node; 1052 } else { 1053 /* This node also falls within the new 1054 * prefixlen. Merge the to-be-reinserted 1055 * node and this one. 1056 */ 1057 xfrm_policy_inexact_node_merge(net, node, 1058 cached, family); 1059 kfree_rcu(node, rcu); 1060 } 1061 1062 /* restart */ 1063 p = &root->rb_node; 1064 parent = NULL; 1065 } 1066 } 1067 1068 node = cached; 1069 if (!node) { 1070 node = xfrm_pol_inexact_node_alloc(addr, prefixlen); 1071 if (!node) 1072 return NULL; 1073 } 1074 1075 rb_link_node_rcu(&node->node, parent, p); 1076 rb_insert_color(&node->node, root); 1077 1078 return node; 1079 } 1080 1081 static void xfrm_policy_inexact_gc_tree(struct rb_root *r, bool rm) 1082 { 1083 struct xfrm_pol_inexact_node *node; 1084 struct rb_node *rn = rb_first(r); 1085 1086 while (rn) { 1087 node = rb_entry(rn, struct xfrm_pol_inexact_node, node); 1088 1089 xfrm_policy_inexact_gc_tree(&node->root, rm); 1090 rn = rb_next(rn); 1091 1092 if (!hlist_empty(&node->hhead) || !RB_EMPTY_ROOT(&node->root)) { 1093 WARN_ON_ONCE(rm); 1094 continue; 1095 } 1096 1097 rb_erase(&node->node, r); 1098 kfree_rcu(node, rcu); 1099 } 1100 } 1101 1102 static void __xfrm_policy_inexact_prune_bin(struct xfrm_pol_inexact_bin *b, bool net_exit) 1103 { 1104 write_seqcount_begin(&b->count); 1105 xfrm_policy_inexact_gc_tree(&b->root_d, net_exit); 1106 xfrm_policy_inexact_gc_tree(&b->root_s, net_exit); 1107 write_seqcount_end(&b->count); 1108 1109 if (!RB_EMPTY_ROOT(&b->root_d) || !RB_EMPTY_ROOT(&b->root_s) || 1110 !hlist_empty(&b->hhead)) { 1111 WARN_ON_ONCE(net_exit); 1112 return; 1113 } 1114 1115 if (rhashtable_remove_fast(&xfrm_policy_inexact_table, &b->head, 1116 xfrm_pol_inexact_params) == 0) { 1117 list_del(&b->inexact_bins); 1118 kfree_rcu(b, rcu); 1119 } 1120 } 1121 1122 static void xfrm_policy_inexact_prune_bin(struct xfrm_pol_inexact_bin *b) 1123 { 1124 struct net *net = read_pnet(&b->k.net); 1125 1126 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1127 __xfrm_policy_inexact_prune_bin(b, false); 1128 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1129 } 1130 1131 static void __xfrm_policy_inexact_flush(struct net *net) 1132 { 1133 struct xfrm_pol_inexact_bin *bin, *t; 1134 1135 lockdep_assert_held(&net->xfrm.xfrm_policy_lock); 1136 1137 list_for_each_entry_safe(bin, t, &net->xfrm.inexact_bins, inexact_bins) 1138 __xfrm_policy_inexact_prune_bin(bin, false); 1139 } 1140 1141 static struct hlist_head * 1142 xfrm_policy_inexact_alloc_chain(struct xfrm_pol_inexact_bin *bin, 1143 struct xfrm_policy *policy, u8 dir) 1144 { 1145 struct xfrm_pol_inexact_node *n; 1146 struct net *net; 1147 1148 net = xp_net(policy); 1149 lockdep_assert_held(&net->xfrm.xfrm_policy_lock); 1150 1151 if (xfrm_policy_inexact_insert_use_any_list(policy)) 1152 return &bin->hhead; 1153 1154 if (xfrm_pol_inexact_addr_use_any_list(&policy->selector.daddr, 1155 policy->family, 1156 policy->selector.prefixlen_d)) { 1157 write_seqcount_begin(&bin->count); 1158 n = xfrm_policy_inexact_insert_node(net, 1159 &bin->root_s, 1160 &policy->selector.saddr, 1161 policy->family, 1162 policy->selector.prefixlen_s, 1163 dir); 1164 write_seqcount_end(&bin->count); 1165 if (!n) 1166 return NULL; 1167 1168 return &n->hhead; 1169 } 1170 1171 /* daddr is fixed */ 1172 write_seqcount_begin(&bin->count); 1173 n = xfrm_policy_inexact_insert_node(net, 1174 &bin->root_d, 1175 &policy->selector.daddr, 1176 policy->family, 1177 policy->selector.prefixlen_d, dir); 1178 write_seqcount_end(&bin->count); 1179 if (!n) 1180 return NULL; 1181 1182 /* saddr is wildcard */ 1183 if (xfrm_pol_inexact_addr_use_any_list(&policy->selector.saddr, 1184 policy->family, 1185 policy->selector.prefixlen_s)) 1186 return &n->hhead; 1187 1188 write_seqcount_begin(&bin->count); 1189 n = xfrm_policy_inexact_insert_node(net, 1190 &n->root, 1191 &policy->selector.saddr, 1192 policy->family, 1193 policy->selector.prefixlen_s, dir); 1194 write_seqcount_end(&bin->count); 1195 if (!n) 1196 return NULL; 1197 1198 return &n->hhead; 1199 } 1200 1201 static struct xfrm_policy * 1202 xfrm_policy_inexact_insert(struct xfrm_policy *policy, u8 dir, int excl) 1203 { 1204 struct xfrm_pol_inexact_bin *bin; 1205 struct xfrm_policy *delpol; 1206 struct hlist_head *chain; 1207 struct net *net; 1208 1209 bin = xfrm_policy_inexact_alloc_bin(policy, dir); 1210 if (!bin) 1211 return ERR_PTR(-ENOMEM); 1212 1213 net = xp_net(policy); 1214 lockdep_assert_held(&net->xfrm.xfrm_policy_lock); 1215 1216 chain = xfrm_policy_inexact_alloc_chain(bin, policy, dir); 1217 if (!chain) { 1218 __xfrm_policy_inexact_prune_bin(bin, false); 1219 return ERR_PTR(-ENOMEM); 1220 } 1221 1222 delpol = xfrm_policy_insert_list(chain, policy, excl); 1223 if (delpol && excl) { 1224 __xfrm_policy_inexact_prune_bin(bin, false); 1225 return ERR_PTR(-EEXIST); 1226 } 1227 1228 chain = &net->xfrm.policy_inexact[dir]; 1229 xfrm_policy_insert_inexact_list(chain, policy); 1230 1231 if (delpol) 1232 __xfrm_policy_inexact_prune_bin(bin, false); 1233 1234 return delpol; 1235 } 1236 1237 static void xfrm_hash_rebuild(struct work_struct *work) 1238 { 1239 struct net *net = container_of(work, struct net, 1240 xfrm.policy_hthresh.work); 1241 unsigned int hmask; 1242 struct xfrm_policy *pol; 1243 struct xfrm_policy *policy; 1244 struct hlist_head *chain; 1245 struct hlist_head *odst; 1246 struct hlist_node *newpos; 1247 int i; 1248 int dir; 1249 unsigned seq; 1250 u8 lbits4, rbits4, lbits6, rbits6; 1251 1252 mutex_lock(&hash_resize_mutex); 1253 1254 /* read selector prefixlen thresholds */ 1255 do { 1256 seq = read_seqbegin(&net->xfrm.policy_hthresh.lock); 1257 1258 lbits4 = net->xfrm.policy_hthresh.lbits4; 1259 rbits4 = net->xfrm.policy_hthresh.rbits4; 1260 lbits6 = net->xfrm.policy_hthresh.lbits6; 1261 rbits6 = net->xfrm.policy_hthresh.rbits6; 1262 } while (read_seqretry(&net->xfrm.policy_hthresh.lock, seq)); 1263 1264 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1265 write_seqcount_begin(&net->xfrm.xfrm_policy_hash_generation); 1266 1267 /* make sure that we can insert the indirect policies again before 1268 * we start with destructive action. 1269 */ 1270 list_for_each_entry(policy, &net->xfrm.policy_all, walk.all) { 1271 struct xfrm_pol_inexact_bin *bin; 1272 u8 dbits, sbits; 1273 1274 if (policy->walk.dead) 1275 continue; 1276 1277 dir = xfrm_policy_id2dir(policy->index); 1278 if (dir >= XFRM_POLICY_MAX) 1279 continue; 1280 1281 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) { 1282 if (policy->family == AF_INET) { 1283 dbits = rbits4; 1284 sbits = lbits4; 1285 } else { 1286 dbits = rbits6; 1287 sbits = lbits6; 1288 } 1289 } else { 1290 if (policy->family == AF_INET) { 1291 dbits = lbits4; 1292 sbits = rbits4; 1293 } else { 1294 dbits = lbits6; 1295 sbits = rbits6; 1296 } 1297 } 1298 1299 if (policy->selector.prefixlen_d < dbits || 1300 policy->selector.prefixlen_s < sbits) 1301 continue; 1302 1303 bin = xfrm_policy_inexact_alloc_bin(policy, dir); 1304 if (!bin) 1305 goto out_unlock; 1306 1307 if (!xfrm_policy_inexact_alloc_chain(bin, policy, dir)) 1308 goto out_unlock; 1309 } 1310 1311 /* reset the bydst and inexact table in all directions */ 1312 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) { 1313 struct hlist_node *n; 1314 1315 hlist_for_each_entry_safe(policy, n, 1316 &net->xfrm.policy_inexact[dir], 1317 bydst_inexact_list) { 1318 hlist_del_rcu(&policy->bydst); 1319 hlist_del_init(&policy->bydst_inexact_list); 1320 } 1321 1322 hmask = net->xfrm.policy_bydst[dir].hmask; 1323 odst = net->xfrm.policy_bydst[dir].table; 1324 for (i = hmask; i >= 0; i--) { 1325 hlist_for_each_entry_safe(policy, n, odst + i, bydst) 1326 hlist_del_rcu(&policy->bydst); 1327 } 1328 if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) { 1329 /* dir out => dst = remote, src = local */ 1330 net->xfrm.policy_bydst[dir].dbits4 = rbits4; 1331 net->xfrm.policy_bydst[dir].sbits4 = lbits4; 1332 net->xfrm.policy_bydst[dir].dbits6 = rbits6; 1333 net->xfrm.policy_bydst[dir].sbits6 = lbits6; 1334 } else { 1335 /* dir in/fwd => dst = local, src = remote */ 1336 net->xfrm.policy_bydst[dir].dbits4 = lbits4; 1337 net->xfrm.policy_bydst[dir].sbits4 = rbits4; 1338 net->xfrm.policy_bydst[dir].dbits6 = lbits6; 1339 net->xfrm.policy_bydst[dir].sbits6 = rbits6; 1340 } 1341 } 1342 1343 /* re-insert all policies by order of creation */ 1344 list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) { 1345 if (policy->walk.dead) 1346 continue; 1347 dir = xfrm_policy_id2dir(policy->index); 1348 if (dir >= XFRM_POLICY_MAX) { 1349 /* skip socket policies */ 1350 continue; 1351 } 1352 newpos = NULL; 1353 chain = policy_hash_bysel(net, &policy->selector, 1354 policy->family, dir); 1355 1356 if (!chain) { 1357 void *p = xfrm_policy_inexact_insert(policy, dir, 0); 1358 1359 WARN_ONCE(IS_ERR(p), "reinsert: %ld\n", PTR_ERR(p)); 1360 continue; 1361 } 1362 1363 hlist_for_each_entry(pol, chain, bydst) { 1364 if (policy->priority >= pol->priority) 1365 newpos = &pol->bydst; 1366 else 1367 break; 1368 } 1369 if (newpos && policy->xdo.type != XFRM_DEV_OFFLOAD_PACKET) 1370 hlist_add_behind_rcu(&policy->bydst, newpos); 1371 else 1372 hlist_add_head_rcu(&policy->bydst, chain); 1373 } 1374 1375 out_unlock: 1376 __xfrm_policy_inexact_flush(net); 1377 write_seqcount_end(&net->xfrm.xfrm_policy_hash_generation); 1378 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1379 1380 mutex_unlock(&hash_resize_mutex); 1381 } 1382 1383 void xfrm_policy_hash_rebuild(struct net *net) 1384 { 1385 schedule_work(&net->xfrm.policy_hthresh.work); 1386 } 1387 EXPORT_SYMBOL(xfrm_policy_hash_rebuild); 1388 1389 /* Generate new index... KAME seems to generate them ordered by cost 1390 * of an absolute inpredictability of ordering of rules. This will not pass. */ 1391 static u32 xfrm_gen_index(struct net *net, int dir, u32 index) 1392 { 1393 for (;;) { 1394 struct hlist_head *list; 1395 struct xfrm_policy *p; 1396 u32 idx; 1397 int found; 1398 1399 if (!index) { 1400 idx = (net->xfrm.idx_generator | dir); 1401 net->xfrm.idx_generator += 8; 1402 } else { 1403 idx = index; 1404 index = 0; 1405 } 1406 1407 if (idx == 0) 1408 idx = 8; 1409 list = net->xfrm.policy_byidx + idx_hash(net, idx); 1410 found = 0; 1411 hlist_for_each_entry(p, list, byidx) { 1412 if (p->index == idx) { 1413 found = 1; 1414 break; 1415 } 1416 } 1417 if (!found) 1418 return idx; 1419 } 1420 } 1421 1422 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2) 1423 { 1424 u32 *p1 = (u32 *) s1; 1425 u32 *p2 = (u32 *) s2; 1426 int len = sizeof(struct xfrm_selector) / sizeof(u32); 1427 int i; 1428 1429 for (i = 0; i < len; i++) { 1430 if (p1[i] != p2[i]) 1431 return 1; 1432 } 1433 1434 return 0; 1435 } 1436 1437 static void xfrm_policy_requeue(struct xfrm_policy *old, 1438 struct xfrm_policy *new) 1439 { 1440 struct xfrm_policy_queue *pq = &old->polq; 1441 struct sk_buff_head list; 1442 1443 if (skb_queue_empty(&pq->hold_queue)) 1444 return; 1445 1446 __skb_queue_head_init(&list); 1447 1448 spin_lock_bh(&pq->hold_queue.lock); 1449 skb_queue_splice_init(&pq->hold_queue, &list); 1450 if (del_timer(&pq->hold_timer)) 1451 xfrm_pol_put(old); 1452 spin_unlock_bh(&pq->hold_queue.lock); 1453 1454 pq = &new->polq; 1455 1456 spin_lock_bh(&pq->hold_queue.lock); 1457 skb_queue_splice(&list, &pq->hold_queue); 1458 pq->timeout = XFRM_QUEUE_TMO_MIN; 1459 if (!mod_timer(&pq->hold_timer, jiffies)) 1460 xfrm_pol_hold(new); 1461 spin_unlock_bh(&pq->hold_queue.lock); 1462 } 1463 1464 static inline bool xfrm_policy_mark_match(const struct xfrm_mark *mark, 1465 struct xfrm_policy *pol) 1466 { 1467 return mark->v == pol->mark.v && mark->m == pol->mark.m; 1468 } 1469 1470 static u32 xfrm_pol_bin_key(const void *data, u32 len, u32 seed) 1471 { 1472 const struct xfrm_pol_inexact_key *k = data; 1473 u32 a = k->type << 24 | k->dir << 16 | k->family; 1474 1475 return jhash_3words(a, k->if_id, net_hash_mix(read_pnet(&k->net)), 1476 seed); 1477 } 1478 1479 static u32 xfrm_pol_bin_obj(const void *data, u32 len, u32 seed) 1480 { 1481 const struct xfrm_pol_inexact_bin *b = data; 1482 1483 return xfrm_pol_bin_key(&b->k, 0, seed); 1484 } 1485 1486 static int xfrm_pol_bin_cmp(struct rhashtable_compare_arg *arg, 1487 const void *ptr) 1488 { 1489 const struct xfrm_pol_inexact_key *key = arg->key; 1490 const struct xfrm_pol_inexact_bin *b = ptr; 1491 int ret; 1492 1493 if (!net_eq(read_pnet(&b->k.net), read_pnet(&key->net))) 1494 return -1; 1495 1496 ret = b->k.dir ^ key->dir; 1497 if (ret) 1498 return ret; 1499 1500 ret = b->k.type ^ key->type; 1501 if (ret) 1502 return ret; 1503 1504 ret = b->k.family ^ key->family; 1505 if (ret) 1506 return ret; 1507 1508 return b->k.if_id ^ key->if_id; 1509 } 1510 1511 static const struct rhashtable_params xfrm_pol_inexact_params = { 1512 .head_offset = offsetof(struct xfrm_pol_inexact_bin, head), 1513 .hashfn = xfrm_pol_bin_key, 1514 .obj_hashfn = xfrm_pol_bin_obj, 1515 .obj_cmpfn = xfrm_pol_bin_cmp, 1516 .automatic_shrinking = true, 1517 }; 1518 1519 static void xfrm_policy_insert_inexact_list(struct hlist_head *chain, 1520 struct xfrm_policy *policy) 1521 { 1522 struct xfrm_policy *pol, *delpol = NULL; 1523 struct hlist_node *newpos = NULL; 1524 int i = 0; 1525 1526 hlist_for_each_entry(pol, chain, bydst_inexact_list) { 1527 if (pol->type == policy->type && 1528 pol->if_id == policy->if_id && 1529 !selector_cmp(&pol->selector, &policy->selector) && 1530 xfrm_policy_mark_match(&policy->mark, pol) && 1531 xfrm_sec_ctx_match(pol->security, policy->security) && 1532 !WARN_ON(delpol)) { 1533 delpol = pol; 1534 if (policy->priority > pol->priority) 1535 continue; 1536 } else if (policy->priority >= pol->priority) { 1537 newpos = &pol->bydst_inexact_list; 1538 continue; 1539 } 1540 if (delpol) 1541 break; 1542 } 1543 1544 if (newpos && policy->xdo.type != XFRM_DEV_OFFLOAD_PACKET) 1545 hlist_add_behind_rcu(&policy->bydst_inexact_list, newpos); 1546 else 1547 hlist_add_head_rcu(&policy->bydst_inexact_list, chain); 1548 1549 hlist_for_each_entry(pol, chain, bydst_inexact_list) { 1550 pol->pos = i; 1551 i++; 1552 } 1553 } 1554 1555 static struct xfrm_policy *xfrm_policy_insert_list(struct hlist_head *chain, 1556 struct xfrm_policy *policy, 1557 bool excl) 1558 { 1559 struct xfrm_policy *pol, *newpos = NULL, *delpol = NULL; 1560 1561 hlist_for_each_entry(pol, chain, bydst) { 1562 if (pol->type == policy->type && 1563 pol->if_id == policy->if_id && 1564 !selector_cmp(&pol->selector, &policy->selector) && 1565 xfrm_policy_mark_match(&policy->mark, pol) && 1566 xfrm_sec_ctx_match(pol->security, policy->security) && 1567 !WARN_ON(delpol)) { 1568 if (excl) 1569 return ERR_PTR(-EEXIST); 1570 delpol = pol; 1571 if (policy->priority > pol->priority) 1572 continue; 1573 } else if (policy->priority >= pol->priority) { 1574 newpos = pol; 1575 continue; 1576 } 1577 if (delpol) 1578 break; 1579 } 1580 1581 if (newpos && policy->xdo.type != XFRM_DEV_OFFLOAD_PACKET) 1582 hlist_add_behind_rcu(&policy->bydst, &newpos->bydst); 1583 else 1584 /* Packet offload policies enter to the head 1585 * to speed-up lookups. 1586 */ 1587 hlist_add_head_rcu(&policy->bydst, chain); 1588 1589 return delpol; 1590 } 1591 1592 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl) 1593 { 1594 struct net *net = xp_net(policy); 1595 struct xfrm_policy *delpol; 1596 struct hlist_head *chain; 1597 1598 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1599 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir); 1600 if (chain) 1601 delpol = xfrm_policy_insert_list(chain, policy, excl); 1602 else 1603 delpol = xfrm_policy_inexact_insert(policy, dir, excl); 1604 1605 if (IS_ERR(delpol)) { 1606 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1607 return PTR_ERR(delpol); 1608 } 1609 1610 __xfrm_policy_link(policy, dir); 1611 1612 /* After previous checking, family can either be AF_INET or AF_INET6 */ 1613 if (policy->family == AF_INET) 1614 rt_genid_bump_ipv4(net); 1615 else 1616 rt_genid_bump_ipv6(net); 1617 1618 if (delpol) { 1619 xfrm_policy_requeue(delpol, policy); 1620 __xfrm_policy_unlink(delpol, dir); 1621 } 1622 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir, policy->index); 1623 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index)); 1624 policy->curlft.add_time = ktime_get_real_seconds(); 1625 policy->curlft.use_time = 0; 1626 if (!mod_timer(&policy->timer, jiffies + HZ)) 1627 xfrm_pol_hold(policy); 1628 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1629 1630 if (delpol) 1631 xfrm_policy_kill(delpol); 1632 else if (xfrm_bydst_should_resize(net, dir, NULL)) 1633 schedule_work(&net->xfrm.policy_hash_work); 1634 1635 return 0; 1636 } 1637 EXPORT_SYMBOL(xfrm_policy_insert); 1638 1639 static struct xfrm_policy * 1640 __xfrm_policy_bysel_ctx(struct hlist_head *chain, const struct xfrm_mark *mark, 1641 u32 if_id, u8 type, int dir, struct xfrm_selector *sel, 1642 struct xfrm_sec_ctx *ctx) 1643 { 1644 struct xfrm_policy *pol; 1645 1646 if (!chain) 1647 return NULL; 1648 1649 hlist_for_each_entry(pol, chain, bydst) { 1650 if (pol->type == type && 1651 pol->if_id == if_id && 1652 xfrm_policy_mark_match(mark, pol) && 1653 !selector_cmp(sel, &pol->selector) && 1654 xfrm_sec_ctx_match(ctx, pol->security)) 1655 return pol; 1656 } 1657 1658 return NULL; 1659 } 1660 1661 struct xfrm_policy * 1662 xfrm_policy_bysel_ctx(struct net *net, const struct xfrm_mark *mark, u32 if_id, 1663 u8 type, int dir, struct xfrm_selector *sel, 1664 struct xfrm_sec_ctx *ctx, int delete, int *err) 1665 { 1666 struct xfrm_pol_inexact_bin *bin = NULL; 1667 struct xfrm_policy *pol, *ret = NULL; 1668 struct hlist_head *chain; 1669 1670 *err = 0; 1671 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1672 chain = policy_hash_bysel(net, sel, sel->family, dir); 1673 if (!chain) { 1674 struct xfrm_pol_inexact_candidates cand; 1675 int i; 1676 1677 bin = xfrm_policy_inexact_lookup(net, type, 1678 sel->family, dir, if_id); 1679 if (!bin) { 1680 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1681 return NULL; 1682 } 1683 1684 if (!xfrm_policy_find_inexact_candidates(&cand, bin, 1685 &sel->saddr, 1686 &sel->daddr)) { 1687 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1688 return NULL; 1689 } 1690 1691 pol = NULL; 1692 for (i = 0; i < ARRAY_SIZE(cand.res); i++) { 1693 struct xfrm_policy *tmp; 1694 1695 tmp = __xfrm_policy_bysel_ctx(cand.res[i], mark, 1696 if_id, type, dir, 1697 sel, ctx); 1698 if (!tmp) 1699 continue; 1700 1701 if (!pol || tmp->pos < pol->pos) 1702 pol = tmp; 1703 } 1704 } else { 1705 pol = __xfrm_policy_bysel_ctx(chain, mark, if_id, type, dir, 1706 sel, ctx); 1707 } 1708 1709 if (pol) { 1710 xfrm_pol_hold(pol); 1711 if (delete) { 1712 *err = security_xfrm_policy_delete(pol->security); 1713 if (*err) { 1714 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1715 return pol; 1716 } 1717 __xfrm_policy_unlink(pol, dir); 1718 } 1719 ret = pol; 1720 } 1721 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1722 1723 if (ret && delete) 1724 xfrm_policy_kill(ret); 1725 if (bin && delete) 1726 xfrm_policy_inexact_prune_bin(bin); 1727 return ret; 1728 } 1729 EXPORT_SYMBOL(xfrm_policy_bysel_ctx); 1730 1731 struct xfrm_policy * 1732 xfrm_policy_byid(struct net *net, const struct xfrm_mark *mark, u32 if_id, 1733 u8 type, int dir, u32 id, int delete, int *err) 1734 { 1735 struct xfrm_policy *pol, *ret; 1736 struct hlist_head *chain; 1737 1738 *err = -ENOENT; 1739 if (xfrm_policy_id2dir(id) != dir) 1740 return NULL; 1741 1742 *err = 0; 1743 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1744 chain = net->xfrm.policy_byidx + idx_hash(net, id); 1745 ret = NULL; 1746 hlist_for_each_entry(pol, chain, byidx) { 1747 if (pol->type == type && pol->index == id && 1748 pol->if_id == if_id && xfrm_policy_mark_match(mark, pol)) { 1749 xfrm_pol_hold(pol); 1750 if (delete) { 1751 *err = security_xfrm_policy_delete( 1752 pol->security); 1753 if (*err) { 1754 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1755 return pol; 1756 } 1757 __xfrm_policy_unlink(pol, dir); 1758 } 1759 ret = pol; 1760 break; 1761 } 1762 } 1763 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1764 1765 if (ret && delete) 1766 xfrm_policy_kill(ret); 1767 return ret; 1768 } 1769 EXPORT_SYMBOL(xfrm_policy_byid); 1770 1771 #ifdef CONFIG_SECURITY_NETWORK_XFRM 1772 static inline int 1773 xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid) 1774 { 1775 struct xfrm_policy *pol; 1776 int err = 0; 1777 1778 list_for_each_entry(pol, &net->xfrm.policy_all, walk.all) { 1779 if (pol->walk.dead || 1780 xfrm_policy_id2dir(pol->index) >= XFRM_POLICY_MAX || 1781 pol->type != type) 1782 continue; 1783 1784 err = security_xfrm_policy_delete(pol->security); 1785 if (err) { 1786 xfrm_audit_policy_delete(pol, 0, task_valid); 1787 return err; 1788 } 1789 } 1790 return err; 1791 } 1792 1793 static inline int xfrm_dev_policy_flush_secctx_check(struct net *net, 1794 struct net_device *dev, 1795 bool task_valid) 1796 { 1797 struct xfrm_policy *pol; 1798 int err = 0; 1799 1800 list_for_each_entry(pol, &net->xfrm.policy_all, walk.all) { 1801 if (pol->walk.dead || 1802 xfrm_policy_id2dir(pol->index) >= XFRM_POLICY_MAX || 1803 pol->xdo.dev != dev) 1804 continue; 1805 1806 err = security_xfrm_policy_delete(pol->security); 1807 if (err) { 1808 xfrm_audit_policy_delete(pol, 0, task_valid); 1809 return err; 1810 } 1811 } 1812 return err; 1813 } 1814 #else 1815 static inline int 1816 xfrm_policy_flush_secctx_check(struct net *net, u8 type, bool task_valid) 1817 { 1818 return 0; 1819 } 1820 1821 static inline int xfrm_dev_policy_flush_secctx_check(struct net *net, 1822 struct net_device *dev, 1823 bool task_valid) 1824 { 1825 return 0; 1826 } 1827 #endif 1828 1829 int xfrm_policy_flush(struct net *net, u8 type, bool task_valid) 1830 { 1831 int dir, err = 0, cnt = 0; 1832 struct xfrm_policy *pol; 1833 1834 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1835 1836 err = xfrm_policy_flush_secctx_check(net, type, task_valid); 1837 if (err) 1838 goto out; 1839 1840 again: 1841 list_for_each_entry(pol, &net->xfrm.policy_all, walk.all) { 1842 if (pol->walk.dead) 1843 continue; 1844 1845 dir = xfrm_policy_id2dir(pol->index); 1846 if (dir >= XFRM_POLICY_MAX || 1847 pol->type != type) 1848 continue; 1849 1850 __xfrm_policy_unlink(pol, dir); 1851 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1852 xfrm_dev_policy_delete(pol); 1853 cnt++; 1854 xfrm_audit_policy_delete(pol, 1, task_valid); 1855 xfrm_policy_kill(pol); 1856 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1857 goto again; 1858 } 1859 if (cnt) 1860 __xfrm_policy_inexact_flush(net); 1861 else 1862 err = -ESRCH; 1863 out: 1864 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1865 return err; 1866 } 1867 EXPORT_SYMBOL(xfrm_policy_flush); 1868 1869 int xfrm_dev_policy_flush(struct net *net, struct net_device *dev, 1870 bool task_valid) 1871 { 1872 int dir, err = 0, cnt = 0; 1873 struct xfrm_policy *pol; 1874 1875 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1876 1877 err = xfrm_dev_policy_flush_secctx_check(net, dev, task_valid); 1878 if (err) 1879 goto out; 1880 1881 again: 1882 list_for_each_entry(pol, &net->xfrm.policy_all, walk.all) { 1883 if (pol->walk.dead) 1884 continue; 1885 1886 dir = xfrm_policy_id2dir(pol->index); 1887 if (dir >= XFRM_POLICY_MAX || 1888 pol->xdo.dev != dev) 1889 continue; 1890 1891 __xfrm_policy_unlink(pol, dir); 1892 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1893 xfrm_dev_policy_delete(pol); 1894 cnt++; 1895 xfrm_audit_policy_delete(pol, 1, task_valid); 1896 xfrm_policy_kill(pol); 1897 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1898 goto again; 1899 } 1900 if (cnt) 1901 __xfrm_policy_inexact_flush(net); 1902 else 1903 err = -ESRCH; 1904 out: 1905 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1906 return err; 1907 } 1908 EXPORT_SYMBOL(xfrm_dev_policy_flush); 1909 1910 int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk, 1911 int (*func)(struct xfrm_policy *, int, int, void*), 1912 void *data) 1913 { 1914 struct xfrm_policy *pol; 1915 struct xfrm_policy_walk_entry *x; 1916 int error = 0; 1917 1918 if (walk->type >= XFRM_POLICY_TYPE_MAX && 1919 walk->type != XFRM_POLICY_TYPE_ANY) 1920 return -EINVAL; 1921 1922 if (list_empty(&walk->walk.all) && walk->seq != 0) 1923 return 0; 1924 1925 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 1926 if (list_empty(&walk->walk.all)) 1927 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all); 1928 else 1929 x = list_first_entry(&walk->walk.all, 1930 struct xfrm_policy_walk_entry, all); 1931 1932 list_for_each_entry_from(x, &net->xfrm.policy_all, all) { 1933 if (x->dead) 1934 continue; 1935 pol = container_of(x, struct xfrm_policy, walk); 1936 if (walk->type != XFRM_POLICY_TYPE_ANY && 1937 walk->type != pol->type) 1938 continue; 1939 error = func(pol, xfrm_policy_id2dir(pol->index), 1940 walk->seq, data); 1941 if (error) { 1942 list_move_tail(&walk->walk.all, &x->all); 1943 goto out; 1944 } 1945 walk->seq++; 1946 } 1947 if (walk->seq == 0) { 1948 error = -ENOENT; 1949 goto out; 1950 } 1951 list_del_init(&walk->walk.all); 1952 out: 1953 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1954 return error; 1955 } 1956 EXPORT_SYMBOL(xfrm_policy_walk); 1957 1958 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type) 1959 { 1960 INIT_LIST_HEAD(&walk->walk.all); 1961 walk->walk.dead = 1; 1962 walk->type = type; 1963 walk->seq = 0; 1964 } 1965 EXPORT_SYMBOL(xfrm_policy_walk_init); 1966 1967 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk, struct net *net) 1968 { 1969 if (list_empty(&walk->walk.all)) 1970 return; 1971 1972 spin_lock_bh(&net->xfrm.xfrm_policy_lock); /*FIXME where is net? */ 1973 list_del(&walk->walk.all); 1974 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 1975 } 1976 EXPORT_SYMBOL(xfrm_policy_walk_done); 1977 1978 /* 1979 * Find policy to apply to this flow. 1980 * 1981 * Returns 0 if policy found, else an -errno. 1982 */ 1983 static int xfrm_policy_match(const struct xfrm_policy *pol, 1984 const struct flowi *fl, 1985 u8 type, u16 family, u32 if_id) 1986 { 1987 const struct xfrm_selector *sel = &pol->selector; 1988 int ret = -ESRCH; 1989 bool match; 1990 1991 if (pol->family != family || 1992 pol->if_id != if_id || 1993 (fl->flowi_mark & pol->mark.m) != pol->mark.v || 1994 pol->type != type) 1995 return ret; 1996 1997 match = xfrm_selector_match(sel, fl, family); 1998 if (match) 1999 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid); 2000 return ret; 2001 } 2002 2003 static struct xfrm_pol_inexact_node * 2004 xfrm_policy_lookup_inexact_addr(const struct rb_root *r, 2005 seqcount_spinlock_t *count, 2006 const xfrm_address_t *addr, u16 family) 2007 { 2008 const struct rb_node *parent; 2009 int seq; 2010 2011 again: 2012 seq = read_seqcount_begin(count); 2013 2014 parent = rcu_dereference_raw(r->rb_node); 2015 while (parent) { 2016 struct xfrm_pol_inexact_node *node; 2017 int delta; 2018 2019 node = rb_entry(parent, struct xfrm_pol_inexact_node, node); 2020 2021 delta = xfrm_policy_addr_delta(addr, &node->addr, 2022 node->prefixlen, family); 2023 if (delta < 0) { 2024 parent = rcu_dereference_raw(parent->rb_left); 2025 continue; 2026 } else if (delta > 0) { 2027 parent = rcu_dereference_raw(parent->rb_right); 2028 continue; 2029 } 2030 2031 return node; 2032 } 2033 2034 if (read_seqcount_retry(count, seq)) 2035 goto again; 2036 2037 return NULL; 2038 } 2039 2040 static bool 2041 xfrm_policy_find_inexact_candidates(struct xfrm_pol_inexact_candidates *cand, 2042 struct xfrm_pol_inexact_bin *b, 2043 const xfrm_address_t *saddr, 2044 const xfrm_address_t *daddr) 2045 { 2046 struct xfrm_pol_inexact_node *n; 2047 u16 family; 2048 2049 if (!b) 2050 return false; 2051 2052 family = b->k.family; 2053 memset(cand, 0, sizeof(*cand)); 2054 cand->res[XFRM_POL_CAND_ANY] = &b->hhead; 2055 2056 n = xfrm_policy_lookup_inexact_addr(&b->root_d, &b->count, daddr, 2057 family); 2058 if (n) { 2059 cand->res[XFRM_POL_CAND_DADDR] = &n->hhead; 2060 n = xfrm_policy_lookup_inexact_addr(&n->root, &b->count, saddr, 2061 family); 2062 if (n) 2063 cand->res[XFRM_POL_CAND_BOTH] = &n->hhead; 2064 } 2065 2066 n = xfrm_policy_lookup_inexact_addr(&b->root_s, &b->count, saddr, 2067 family); 2068 if (n) 2069 cand->res[XFRM_POL_CAND_SADDR] = &n->hhead; 2070 2071 return true; 2072 } 2073 2074 static struct xfrm_pol_inexact_bin * 2075 xfrm_policy_inexact_lookup_rcu(struct net *net, u8 type, u16 family, 2076 u8 dir, u32 if_id) 2077 { 2078 struct xfrm_pol_inexact_key k = { 2079 .family = family, 2080 .type = type, 2081 .dir = dir, 2082 .if_id = if_id, 2083 }; 2084 2085 write_pnet(&k.net, net); 2086 2087 return rhashtable_lookup(&xfrm_policy_inexact_table, &k, 2088 xfrm_pol_inexact_params); 2089 } 2090 2091 static struct xfrm_pol_inexact_bin * 2092 xfrm_policy_inexact_lookup(struct net *net, u8 type, u16 family, 2093 u8 dir, u32 if_id) 2094 { 2095 struct xfrm_pol_inexact_bin *bin; 2096 2097 lockdep_assert_held(&net->xfrm.xfrm_policy_lock); 2098 2099 rcu_read_lock(); 2100 bin = xfrm_policy_inexact_lookup_rcu(net, type, family, dir, if_id); 2101 rcu_read_unlock(); 2102 2103 return bin; 2104 } 2105 2106 static struct xfrm_policy * 2107 __xfrm_policy_eval_candidates(struct hlist_head *chain, 2108 struct xfrm_policy *prefer, 2109 const struct flowi *fl, 2110 u8 type, u16 family, u32 if_id) 2111 { 2112 u32 priority = prefer ? prefer->priority : ~0u; 2113 struct xfrm_policy *pol; 2114 2115 if (!chain) 2116 return NULL; 2117 2118 hlist_for_each_entry_rcu(pol, chain, bydst) { 2119 int err; 2120 2121 if (pol->priority > priority) 2122 break; 2123 2124 err = xfrm_policy_match(pol, fl, type, family, if_id); 2125 if (err) { 2126 if (err != -ESRCH) 2127 return ERR_PTR(err); 2128 2129 continue; 2130 } 2131 2132 if (prefer) { 2133 /* matches. Is it older than *prefer? */ 2134 if (pol->priority == priority && 2135 prefer->pos < pol->pos) 2136 return prefer; 2137 } 2138 2139 return pol; 2140 } 2141 2142 return NULL; 2143 } 2144 2145 static struct xfrm_policy * 2146 xfrm_policy_eval_candidates(struct xfrm_pol_inexact_candidates *cand, 2147 struct xfrm_policy *prefer, 2148 const struct flowi *fl, 2149 u8 type, u16 family, u32 if_id) 2150 { 2151 struct xfrm_policy *tmp; 2152 int i; 2153 2154 for (i = 0; i < ARRAY_SIZE(cand->res); i++) { 2155 tmp = __xfrm_policy_eval_candidates(cand->res[i], 2156 prefer, 2157 fl, type, family, if_id); 2158 if (!tmp) 2159 continue; 2160 2161 if (IS_ERR(tmp)) 2162 return tmp; 2163 prefer = tmp; 2164 } 2165 2166 return prefer; 2167 } 2168 2169 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type, 2170 const struct flowi *fl, 2171 u16 family, u8 dir, 2172 u32 if_id) 2173 { 2174 struct xfrm_pol_inexact_candidates cand; 2175 const xfrm_address_t *daddr, *saddr; 2176 struct xfrm_pol_inexact_bin *bin; 2177 struct xfrm_policy *pol, *ret; 2178 struct hlist_head *chain; 2179 unsigned int sequence; 2180 int err; 2181 2182 daddr = xfrm_flowi_daddr(fl, family); 2183 saddr = xfrm_flowi_saddr(fl, family); 2184 if (unlikely(!daddr || !saddr)) 2185 return NULL; 2186 2187 rcu_read_lock(); 2188 retry: 2189 do { 2190 sequence = read_seqcount_begin(&net->xfrm.xfrm_policy_hash_generation); 2191 chain = policy_hash_direct(net, daddr, saddr, family, dir); 2192 } while (read_seqcount_retry(&net->xfrm.xfrm_policy_hash_generation, sequence)); 2193 2194 ret = NULL; 2195 hlist_for_each_entry_rcu(pol, chain, bydst) { 2196 err = xfrm_policy_match(pol, fl, type, family, if_id); 2197 if (err) { 2198 if (err == -ESRCH) 2199 continue; 2200 else { 2201 ret = ERR_PTR(err); 2202 goto fail; 2203 } 2204 } else { 2205 ret = pol; 2206 break; 2207 } 2208 } 2209 if (ret && ret->xdo.type == XFRM_DEV_OFFLOAD_PACKET) 2210 goto skip_inexact; 2211 2212 bin = xfrm_policy_inexact_lookup_rcu(net, type, family, dir, if_id); 2213 if (!bin || !xfrm_policy_find_inexact_candidates(&cand, bin, saddr, 2214 daddr)) 2215 goto skip_inexact; 2216 2217 pol = xfrm_policy_eval_candidates(&cand, ret, fl, type, 2218 family, if_id); 2219 if (pol) { 2220 ret = pol; 2221 if (IS_ERR(pol)) 2222 goto fail; 2223 } 2224 2225 skip_inexact: 2226 if (read_seqcount_retry(&net->xfrm.xfrm_policy_hash_generation, sequence)) 2227 goto retry; 2228 2229 if (ret && !xfrm_pol_hold_rcu(ret)) 2230 goto retry; 2231 fail: 2232 rcu_read_unlock(); 2233 2234 return ret; 2235 } 2236 2237 static struct xfrm_policy *xfrm_policy_lookup(struct net *net, 2238 const struct flowi *fl, 2239 u16 family, u8 dir, u32 if_id) 2240 { 2241 #ifdef CONFIG_XFRM_SUB_POLICY 2242 struct xfrm_policy *pol; 2243 2244 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, 2245 dir, if_id); 2246 if (pol != NULL) 2247 return pol; 2248 #endif 2249 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, 2250 dir, if_id); 2251 } 2252 2253 static struct xfrm_policy *xfrm_sk_policy_lookup(const struct sock *sk, int dir, 2254 const struct flowi *fl, 2255 u16 family, u32 if_id) 2256 { 2257 struct xfrm_policy *pol; 2258 2259 rcu_read_lock(); 2260 again: 2261 pol = rcu_dereference(sk->sk_policy[dir]); 2262 if (pol != NULL) { 2263 bool match; 2264 int err = 0; 2265 2266 if (pol->family != family) { 2267 pol = NULL; 2268 goto out; 2269 } 2270 2271 match = xfrm_selector_match(&pol->selector, fl, family); 2272 if (match) { 2273 if ((READ_ONCE(sk->sk_mark) & pol->mark.m) != pol->mark.v || 2274 pol->if_id != if_id) { 2275 pol = NULL; 2276 goto out; 2277 } 2278 err = security_xfrm_policy_lookup(pol->security, 2279 fl->flowi_secid); 2280 if (!err) { 2281 if (!xfrm_pol_hold_rcu(pol)) 2282 goto again; 2283 } else if (err == -ESRCH) { 2284 pol = NULL; 2285 } else { 2286 pol = ERR_PTR(err); 2287 } 2288 } else 2289 pol = NULL; 2290 } 2291 out: 2292 rcu_read_unlock(); 2293 return pol; 2294 } 2295 2296 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir) 2297 { 2298 struct net *net = xp_net(pol); 2299 2300 list_add(&pol->walk.all, &net->xfrm.policy_all); 2301 net->xfrm.policy_count[dir]++; 2302 xfrm_pol_hold(pol); 2303 } 2304 2305 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol, 2306 int dir) 2307 { 2308 struct net *net = xp_net(pol); 2309 2310 if (list_empty(&pol->walk.all)) 2311 return NULL; 2312 2313 /* Socket policies are not hashed. */ 2314 if (!hlist_unhashed(&pol->bydst)) { 2315 hlist_del_rcu(&pol->bydst); 2316 hlist_del_init(&pol->bydst_inexact_list); 2317 hlist_del(&pol->byidx); 2318 } 2319 2320 list_del_init(&pol->walk.all); 2321 net->xfrm.policy_count[dir]--; 2322 2323 return pol; 2324 } 2325 2326 static void xfrm_sk_policy_link(struct xfrm_policy *pol, int dir) 2327 { 2328 __xfrm_policy_link(pol, XFRM_POLICY_MAX + dir); 2329 } 2330 2331 static void xfrm_sk_policy_unlink(struct xfrm_policy *pol, int dir) 2332 { 2333 __xfrm_policy_unlink(pol, XFRM_POLICY_MAX + dir); 2334 } 2335 2336 int xfrm_policy_delete(struct xfrm_policy *pol, int dir) 2337 { 2338 struct net *net = xp_net(pol); 2339 2340 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 2341 pol = __xfrm_policy_unlink(pol, dir); 2342 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 2343 if (pol) { 2344 xfrm_dev_policy_delete(pol); 2345 xfrm_policy_kill(pol); 2346 return 0; 2347 } 2348 return -ENOENT; 2349 } 2350 EXPORT_SYMBOL(xfrm_policy_delete); 2351 2352 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol) 2353 { 2354 struct net *net = sock_net(sk); 2355 struct xfrm_policy *old_pol; 2356 2357 #ifdef CONFIG_XFRM_SUB_POLICY 2358 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN) 2359 return -EINVAL; 2360 #endif 2361 2362 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 2363 old_pol = rcu_dereference_protected(sk->sk_policy[dir], 2364 lockdep_is_held(&net->xfrm.xfrm_policy_lock)); 2365 if (pol) { 2366 pol->curlft.add_time = ktime_get_real_seconds(); 2367 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir, 0); 2368 xfrm_sk_policy_link(pol, dir); 2369 } 2370 rcu_assign_pointer(sk->sk_policy[dir], pol); 2371 if (old_pol) { 2372 if (pol) 2373 xfrm_policy_requeue(old_pol, pol); 2374 2375 /* Unlinking succeeds always. This is the only function 2376 * allowed to delete or replace socket policy. 2377 */ 2378 xfrm_sk_policy_unlink(old_pol, dir); 2379 } 2380 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 2381 2382 if (old_pol) { 2383 xfrm_policy_kill(old_pol); 2384 } 2385 return 0; 2386 } 2387 2388 static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir) 2389 { 2390 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC); 2391 struct net *net = xp_net(old); 2392 2393 if (newp) { 2394 newp->selector = old->selector; 2395 if (security_xfrm_policy_clone(old->security, 2396 &newp->security)) { 2397 kfree(newp); 2398 return NULL; /* ENOMEM */ 2399 } 2400 newp->lft = old->lft; 2401 newp->curlft = old->curlft; 2402 newp->mark = old->mark; 2403 newp->if_id = old->if_id; 2404 newp->action = old->action; 2405 newp->flags = old->flags; 2406 newp->xfrm_nr = old->xfrm_nr; 2407 newp->index = old->index; 2408 newp->type = old->type; 2409 newp->family = old->family; 2410 memcpy(newp->xfrm_vec, old->xfrm_vec, 2411 newp->xfrm_nr*sizeof(struct xfrm_tmpl)); 2412 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 2413 xfrm_sk_policy_link(newp, dir); 2414 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 2415 xfrm_pol_put(newp); 2416 } 2417 return newp; 2418 } 2419 2420 int __xfrm_sk_clone_policy(struct sock *sk, const struct sock *osk) 2421 { 2422 const struct xfrm_policy *p; 2423 struct xfrm_policy *np; 2424 int i, ret = 0; 2425 2426 rcu_read_lock(); 2427 for (i = 0; i < 2; i++) { 2428 p = rcu_dereference(osk->sk_policy[i]); 2429 if (p) { 2430 np = clone_policy(p, i); 2431 if (unlikely(!np)) { 2432 ret = -ENOMEM; 2433 break; 2434 } 2435 rcu_assign_pointer(sk->sk_policy[i], np); 2436 } 2437 } 2438 rcu_read_unlock(); 2439 return ret; 2440 } 2441 2442 static int 2443 xfrm_get_saddr(struct net *net, int oif, xfrm_address_t *local, 2444 xfrm_address_t *remote, unsigned short family, u32 mark) 2445 { 2446 int err; 2447 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family); 2448 2449 if (unlikely(afinfo == NULL)) 2450 return -EINVAL; 2451 err = afinfo->get_saddr(net, oif, local, remote, mark); 2452 rcu_read_unlock(); 2453 return err; 2454 } 2455 2456 /* Resolve list of templates for the flow, given policy. */ 2457 2458 static int 2459 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl, 2460 struct xfrm_state **xfrm, unsigned short family) 2461 { 2462 struct net *net = xp_net(policy); 2463 int nx; 2464 int i, error; 2465 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family); 2466 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family); 2467 xfrm_address_t tmp; 2468 2469 for (nx = 0, i = 0; i < policy->xfrm_nr; i++) { 2470 struct xfrm_state *x; 2471 xfrm_address_t *remote = daddr; 2472 xfrm_address_t *local = saddr; 2473 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i]; 2474 2475 if (tmpl->mode == XFRM_MODE_TUNNEL || 2476 tmpl->mode == XFRM_MODE_BEET) { 2477 remote = &tmpl->id.daddr; 2478 local = &tmpl->saddr; 2479 if (xfrm_addr_any(local, tmpl->encap_family)) { 2480 error = xfrm_get_saddr(net, fl->flowi_oif, 2481 &tmp, remote, 2482 tmpl->encap_family, 0); 2483 if (error) 2484 goto fail; 2485 local = &tmp; 2486 } 2487 } 2488 2489 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, 2490 family, policy->if_id); 2491 2492 if (x && x->km.state == XFRM_STATE_VALID) { 2493 xfrm[nx++] = x; 2494 daddr = remote; 2495 saddr = local; 2496 continue; 2497 } 2498 if (x) { 2499 error = (x->km.state == XFRM_STATE_ERROR ? 2500 -EINVAL : -EAGAIN); 2501 xfrm_state_put(x); 2502 } else if (error == -ESRCH) { 2503 error = -EAGAIN; 2504 } 2505 2506 if (!tmpl->optional) 2507 goto fail; 2508 } 2509 return nx; 2510 2511 fail: 2512 for (nx--; nx >= 0; nx--) 2513 xfrm_state_put(xfrm[nx]); 2514 return error; 2515 } 2516 2517 static int 2518 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl, 2519 struct xfrm_state **xfrm, unsigned short family) 2520 { 2521 struct xfrm_state *tp[XFRM_MAX_DEPTH]; 2522 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm; 2523 int cnx = 0; 2524 int error; 2525 int ret; 2526 int i; 2527 2528 for (i = 0; i < npols; i++) { 2529 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) { 2530 error = -ENOBUFS; 2531 goto fail; 2532 } 2533 2534 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family); 2535 if (ret < 0) { 2536 error = ret; 2537 goto fail; 2538 } else 2539 cnx += ret; 2540 } 2541 2542 /* found states are sorted for outbound processing */ 2543 if (npols > 1) 2544 xfrm_state_sort(xfrm, tpp, cnx, family); 2545 2546 return cnx; 2547 2548 fail: 2549 for (cnx--; cnx >= 0; cnx--) 2550 xfrm_state_put(tpp[cnx]); 2551 return error; 2552 2553 } 2554 2555 static int xfrm_get_tos(const struct flowi *fl, int family) 2556 { 2557 if (family == AF_INET) 2558 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; 2559 2560 return 0; 2561 } 2562 2563 static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family) 2564 { 2565 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family); 2566 struct dst_ops *dst_ops; 2567 struct xfrm_dst *xdst; 2568 2569 if (!afinfo) 2570 return ERR_PTR(-EINVAL); 2571 2572 switch (family) { 2573 case AF_INET: 2574 dst_ops = &net->xfrm.xfrm4_dst_ops; 2575 break; 2576 #if IS_ENABLED(CONFIG_IPV6) 2577 case AF_INET6: 2578 dst_ops = &net->xfrm.xfrm6_dst_ops; 2579 break; 2580 #endif 2581 default: 2582 BUG(); 2583 } 2584 xdst = dst_alloc(dst_ops, NULL, DST_OBSOLETE_NONE, 0); 2585 2586 if (likely(xdst)) { 2587 memset_after(xdst, 0, u.dst); 2588 } else 2589 xdst = ERR_PTR(-ENOBUFS); 2590 2591 rcu_read_unlock(); 2592 2593 return xdst; 2594 } 2595 2596 static void xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst, 2597 int nfheader_len) 2598 { 2599 if (dst->ops->family == AF_INET6) { 2600 struct rt6_info *rt = (struct rt6_info *)dst; 2601 path->path_cookie = rt6_get_cookie(rt); 2602 path->u.rt6.rt6i_nfheader_len = nfheader_len; 2603 } 2604 } 2605 2606 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 2607 const struct flowi *fl) 2608 { 2609 const struct xfrm_policy_afinfo *afinfo = 2610 xfrm_policy_get_afinfo(xdst->u.dst.ops->family); 2611 int err; 2612 2613 if (!afinfo) 2614 return -EINVAL; 2615 2616 err = afinfo->fill_dst(xdst, dev, fl); 2617 2618 rcu_read_unlock(); 2619 2620 return err; 2621 } 2622 2623 2624 /* Allocate chain of dst_entry's, attach known xfrm's, calculate 2625 * all the metrics... Shortly, bundle a bundle. 2626 */ 2627 2628 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy, 2629 struct xfrm_state **xfrm, 2630 struct xfrm_dst **bundle, 2631 int nx, 2632 const struct flowi *fl, 2633 struct dst_entry *dst) 2634 { 2635 const struct xfrm_state_afinfo *afinfo; 2636 const struct xfrm_mode *inner_mode; 2637 struct net *net = xp_net(policy); 2638 unsigned long now = jiffies; 2639 struct net_device *dev; 2640 struct xfrm_dst *xdst_prev = NULL; 2641 struct xfrm_dst *xdst0 = NULL; 2642 int i = 0; 2643 int err; 2644 int header_len = 0; 2645 int nfheader_len = 0; 2646 int trailer_len = 0; 2647 int tos; 2648 int family = policy->selector.family; 2649 xfrm_address_t saddr, daddr; 2650 2651 xfrm_flowi_addr_get(fl, &saddr, &daddr, family); 2652 2653 tos = xfrm_get_tos(fl, family); 2654 2655 dst_hold(dst); 2656 2657 for (; i < nx; i++) { 2658 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family); 2659 struct dst_entry *dst1 = &xdst->u.dst; 2660 2661 err = PTR_ERR(xdst); 2662 if (IS_ERR(xdst)) { 2663 dst_release(dst); 2664 goto put_states; 2665 } 2666 2667 bundle[i] = xdst; 2668 if (!xdst_prev) 2669 xdst0 = xdst; 2670 else 2671 /* Ref count is taken during xfrm_alloc_dst() 2672 * No need to do dst_clone() on dst1 2673 */ 2674 xfrm_dst_set_child(xdst_prev, &xdst->u.dst); 2675 2676 if (xfrm[i]->sel.family == AF_UNSPEC) { 2677 inner_mode = xfrm_ip2inner_mode(xfrm[i], 2678 xfrm_af2proto(family)); 2679 if (!inner_mode) { 2680 err = -EAFNOSUPPORT; 2681 dst_release(dst); 2682 goto put_states; 2683 } 2684 } else 2685 inner_mode = &xfrm[i]->inner_mode; 2686 2687 xdst->route = dst; 2688 dst_copy_metrics(dst1, dst); 2689 2690 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) { 2691 __u32 mark = 0; 2692 int oif; 2693 2694 if (xfrm[i]->props.smark.v || xfrm[i]->props.smark.m) 2695 mark = xfrm_smark_get(fl->flowi_mark, xfrm[i]); 2696 2697 family = xfrm[i]->props.family; 2698 oif = fl->flowi_oif ? : fl->flowi_l3mdev; 2699 dst = xfrm_dst_lookup(xfrm[i], tos, oif, 2700 &saddr, &daddr, family, mark); 2701 err = PTR_ERR(dst); 2702 if (IS_ERR(dst)) 2703 goto put_states; 2704 } else 2705 dst_hold(dst); 2706 2707 dst1->xfrm = xfrm[i]; 2708 xdst->xfrm_genid = xfrm[i]->genid; 2709 2710 dst1->obsolete = DST_OBSOLETE_FORCE_CHK; 2711 dst1->lastuse = now; 2712 2713 dst1->input = dst_discard; 2714 2715 rcu_read_lock(); 2716 afinfo = xfrm_state_afinfo_get_rcu(inner_mode->family); 2717 if (likely(afinfo)) 2718 dst1->output = afinfo->output; 2719 else 2720 dst1->output = dst_discard_out; 2721 rcu_read_unlock(); 2722 2723 xdst_prev = xdst; 2724 2725 header_len += xfrm[i]->props.header_len; 2726 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT) 2727 nfheader_len += xfrm[i]->props.header_len; 2728 trailer_len += xfrm[i]->props.trailer_len; 2729 } 2730 2731 xfrm_dst_set_child(xdst_prev, dst); 2732 xdst0->path = dst; 2733 2734 err = -ENODEV; 2735 dev = dst->dev; 2736 if (!dev) 2737 goto free_dst; 2738 2739 xfrm_init_path(xdst0, dst, nfheader_len); 2740 xfrm_init_pmtu(bundle, nx); 2741 2742 for (xdst_prev = xdst0; xdst_prev != (struct xfrm_dst *)dst; 2743 xdst_prev = (struct xfrm_dst *) xfrm_dst_child(&xdst_prev->u.dst)) { 2744 err = xfrm_fill_dst(xdst_prev, dev, fl); 2745 if (err) 2746 goto free_dst; 2747 2748 xdst_prev->u.dst.header_len = header_len; 2749 xdst_prev->u.dst.trailer_len = trailer_len; 2750 header_len -= xdst_prev->u.dst.xfrm->props.header_len; 2751 trailer_len -= xdst_prev->u.dst.xfrm->props.trailer_len; 2752 } 2753 2754 return &xdst0->u.dst; 2755 2756 put_states: 2757 for (; i < nx; i++) 2758 xfrm_state_put(xfrm[i]); 2759 free_dst: 2760 if (xdst0) 2761 dst_release_immediate(&xdst0->u.dst); 2762 2763 return ERR_PTR(err); 2764 } 2765 2766 static int xfrm_expand_policies(const struct flowi *fl, u16 family, 2767 struct xfrm_policy **pols, 2768 int *num_pols, int *num_xfrms) 2769 { 2770 int i; 2771 2772 if (*num_pols == 0 || !pols[0]) { 2773 *num_pols = 0; 2774 *num_xfrms = 0; 2775 return 0; 2776 } 2777 if (IS_ERR(pols[0])) { 2778 *num_pols = 0; 2779 return PTR_ERR(pols[0]); 2780 } 2781 2782 *num_xfrms = pols[0]->xfrm_nr; 2783 2784 #ifdef CONFIG_XFRM_SUB_POLICY 2785 if (pols[0]->action == XFRM_POLICY_ALLOW && 2786 pols[0]->type != XFRM_POLICY_TYPE_MAIN) { 2787 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]), 2788 XFRM_POLICY_TYPE_MAIN, 2789 fl, family, 2790 XFRM_POLICY_OUT, 2791 pols[0]->if_id); 2792 if (pols[1]) { 2793 if (IS_ERR(pols[1])) { 2794 xfrm_pols_put(pols, *num_pols); 2795 *num_pols = 0; 2796 return PTR_ERR(pols[1]); 2797 } 2798 (*num_pols)++; 2799 (*num_xfrms) += pols[1]->xfrm_nr; 2800 } 2801 } 2802 #endif 2803 for (i = 0; i < *num_pols; i++) { 2804 if (pols[i]->action != XFRM_POLICY_ALLOW) { 2805 *num_xfrms = -1; 2806 break; 2807 } 2808 } 2809 2810 return 0; 2811 2812 } 2813 2814 static struct xfrm_dst * 2815 xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols, 2816 const struct flowi *fl, u16 family, 2817 struct dst_entry *dst_orig) 2818 { 2819 struct net *net = xp_net(pols[0]); 2820 struct xfrm_state *xfrm[XFRM_MAX_DEPTH]; 2821 struct xfrm_dst *bundle[XFRM_MAX_DEPTH]; 2822 struct xfrm_dst *xdst; 2823 struct dst_entry *dst; 2824 int err; 2825 2826 /* Try to instantiate a bundle */ 2827 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family); 2828 if (err <= 0) { 2829 if (err == 0) 2830 return NULL; 2831 2832 if (err != -EAGAIN) 2833 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR); 2834 return ERR_PTR(err); 2835 } 2836 2837 dst = xfrm_bundle_create(pols[0], xfrm, bundle, err, fl, dst_orig); 2838 if (IS_ERR(dst)) { 2839 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR); 2840 return ERR_CAST(dst); 2841 } 2842 2843 xdst = (struct xfrm_dst *)dst; 2844 xdst->num_xfrms = err; 2845 xdst->num_pols = num_pols; 2846 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols); 2847 xdst->policy_genid = atomic_read(&pols[0]->genid); 2848 2849 return xdst; 2850 } 2851 2852 static void xfrm_policy_queue_process(struct timer_list *t) 2853 { 2854 struct sk_buff *skb; 2855 struct sock *sk; 2856 struct dst_entry *dst; 2857 struct xfrm_policy *pol = from_timer(pol, t, polq.hold_timer); 2858 struct net *net = xp_net(pol); 2859 struct xfrm_policy_queue *pq = &pol->polq; 2860 struct flowi fl; 2861 struct sk_buff_head list; 2862 __u32 skb_mark; 2863 2864 spin_lock(&pq->hold_queue.lock); 2865 skb = skb_peek(&pq->hold_queue); 2866 if (!skb) { 2867 spin_unlock(&pq->hold_queue.lock); 2868 goto out; 2869 } 2870 dst = skb_dst(skb); 2871 sk = skb->sk; 2872 2873 /* Fixup the mark to support VTI. */ 2874 skb_mark = skb->mark; 2875 skb->mark = pol->mark.v; 2876 xfrm_decode_session(net, skb, &fl, dst->ops->family); 2877 skb->mark = skb_mark; 2878 spin_unlock(&pq->hold_queue.lock); 2879 2880 dst_hold(xfrm_dst_path(dst)); 2881 dst = xfrm_lookup(net, xfrm_dst_path(dst), &fl, sk, XFRM_LOOKUP_QUEUE); 2882 if (IS_ERR(dst)) 2883 goto purge_queue; 2884 2885 if (dst->flags & DST_XFRM_QUEUE) { 2886 dst_release(dst); 2887 2888 if (pq->timeout >= XFRM_QUEUE_TMO_MAX) 2889 goto purge_queue; 2890 2891 pq->timeout = pq->timeout << 1; 2892 if (!mod_timer(&pq->hold_timer, jiffies + pq->timeout)) 2893 xfrm_pol_hold(pol); 2894 goto out; 2895 } 2896 2897 dst_release(dst); 2898 2899 __skb_queue_head_init(&list); 2900 2901 spin_lock(&pq->hold_queue.lock); 2902 pq->timeout = 0; 2903 skb_queue_splice_init(&pq->hold_queue, &list); 2904 spin_unlock(&pq->hold_queue.lock); 2905 2906 while (!skb_queue_empty(&list)) { 2907 skb = __skb_dequeue(&list); 2908 2909 /* Fixup the mark to support VTI. */ 2910 skb_mark = skb->mark; 2911 skb->mark = pol->mark.v; 2912 xfrm_decode_session(net, skb, &fl, skb_dst(skb)->ops->family); 2913 skb->mark = skb_mark; 2914 2915 dst_hold(xfrm_dst_path(skb_dst(skb))); 2916 dst = xfrm_lookup(net, xfrm_dst_path(skb_dst(skb)), &fl, skb->sk, 0); 2917 if (IS_ERR(dst)) { 2918 kfree_skb(skb); 2919 continue; 2920 } 2921 2922 nf_reset_ct(skb); 2923 skb_dst_drop(skb); 2924 skb_dst_set(skb, dst); 2925 2926 dst_output(net, skb->sk, skb); 2927 } 2928 2929 out: 2930 xfrm_pol_put(pol); 2931 return; 2932 2933 purge_queue: 2934 pq->timeout = 0; 2935 skb_queue_purge(&pq->hold_queue); 2936 xfrm_pol_put(pol); 2937 } 2938 2939 static int xdst_queue_output(struct net *net, struct sock *sk, struct sk_buff *skb) 2940 { 2941 unsigned long sched_next; 2942 struct dst_entry *dst = skb_dst(skb); 2943 struct xfrm_dst *xdst = (struct xfrm_dst *) dst; 2944 struct xfrm_policy *pol = xdst->pols[0]; 2945 struct xfrm_policy_queue *pq = &pol->polq; 2946 2947 if (unlikely(skb_fclone_busy(sk, skb))) { 2948 kfree_skb(skb); 2949 return 0; 2950 } 2951 2952 if (pq->hold_queue.qlen > XFRM_MAX_QUEUE_LEN) { 2953 kfree_skb(skb); 2954 return -EAGAIN; 2955 } 2956 2957 skb_dst_force(skb); 2958 2959 spin_lock_bh(&pq->hold_queue.lock); 2960 2961 if (!pq->timeout) 2962 pq->timeout = XFRM_QUEUE_TMO_MIN; 2963 2964 sched_next = jiffies + pq->timeout; 2965 2966 if (del_timer(&pq->hold_timer)) { 2967 if (time_before(pq->hold_timer.expires, sched_next)) 2968 sched_next = pq->hold_timer.expires; 2969 xfrm_pol_put(pol); 2970 } 2971 2972 __skb_queue_tail(&pq->hold_queue, skb); 2973 if (!mod_timer(&pq->hold_timer, sched_next)) 2974 xfrm_pol_hold(pol); 2975 2976 spin_unlock_bh(&pq->hold_queue.lock); 2977 2978 return 0; 2979 } 2980 2981 static struct xfrm_dst *xfrm_create_dummy_bundle(struct net *net, 2982 struct xfrm_flo *xflo, 2983 const struct flowi *fl, 2984 int num_xfrms, 2985 u16 family) 2986 { 2987 int err; 2988 struct net_device *dev; 2989 struct dst_entry *dst; 2990 struct dst_entry *dst1; 2991 struct xfrm_dst *xdst; 2992 2993 xdst = xfrm_alloc_dst(net, family); 2994 if (IS_ERR(xdst)) 2995 return xdst; 2996 2997 if (!(xflo->flags & XFRM_LOOKUP_QUEUE) || 2998 net->xfrm.sysctl_larval_drop || 2999 num_xfrms <= 0) 3000 return xdst; 3001 3002 dst = xflo->dst_orig; 3003 dst1 = &xdst->u.dst; 3004 dst_hold(dst); 3005 xdst->route = dst; 3006 3007 dst_copy_metrics(dst1, dst); 3008 3009 dst1->obsolete = DST_OBSOLETE_FORCE_CHK; 3010 dst1->flags |= DST_XFRM_QUEUE; 3011 dst1->lastuse = jiffies; 3012 3013 dst1->input = dst_discard; 3014 dst1->output = xdst_queue_output; 3015 3016 dst_hold(dst); 3017 xfrm_dst_set_child(xdst, dst); 3018 xdst->path = dst; 3019 3020 xfrm_init_path((struct xfrm_dst *)dst1, dst, 0); 3021 3022 err = -ENODEV; 3023 dev = dst->dev; 3024 if (!dev) 3025 goto free_dst; 3026 3027 err = xfrm_fill_dst(xdst, dev, fl); 3028 if (err) 3029 goto free_dst; 3030 3031 out: 3032 return xdst; 3033 3034 free_dst: 3035 dst_release(dst1); 3036 xdst = ERR_PTR(err); 3037 goto out; 3038 } 3039 3040 static struct xfrm_dst *xfrm_bundle_lookup(struct net *net, 3041 const struct flowi *fl, 3042 u16 family, u8 dir, 3043 struct xfrm_flo *xflo, u32 if_id) 3044 { 3045 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX]; 3046 int num_pols = 0, num_xfrms = 0, err; 3047 struct xfrm_dst *xdst; 3048 3049 /* Resolve policies to use if we couldn't get them from 3050 * previous cache entry */ 3051 num_pols = 1; 3052 pols[0] = xfrm_policy_lookup(net, fl, family, dir, if_id); 3053 err = xfrm_expand_policies(fl, family, pols, 3054 &num_pols, &num_xfrms); 3055 if (err < 0) 3056 goto inc_error; 3057 if (num_pols == 0) 3058 return NULL; 3059 if (num_xfrms <= 0) 3060 goto make_dummy_bundle; 3061 3062 xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, 3063 xflo->dst_orig); 3064 if (IS_ERR(xdst)) { 3065 err = PTR_ERR(xdst); 3066 if (err == -EREMOTE) { 3067 xfrm_pols_put(pols, num_pols); 3068 return NULL; 3069 } 3070 3071 if (err != -EAGAIN) 3072 goto error; 3073 goto make_dummy_bundle; 3074 } else if (xdst == NULL) { 3075 num_xfrms = 0; 3076 goto make_dummy_bundle; 3077 } 3078 3079 return xdst; 3080 3081 make_dummy_bundle: 3082 /* We found policies, but there's no bundles to instantiate: 3083 * either because the policy blocks, has no transformations or 3084 * we could not build template (no xfrm_states).*/ 3085 xdst = xfrm_create_dummy_bundle(net, xflo, fl, num_xfrms, family); 3086 if (IS_ERR(xdst)) { 3087 xfrm_pols_put(pols, num_pols); 3088 return ERR_CAST(xdst); 3089 } 3090 xdst->num_pols = num_pols; 3091 xdst->num_xfrms = num_xfrms; 3092 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy *) * num_pols); 3093 3094 return xdst; 3095 3096 inc_error: 3097 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR); 3098 error: 3099 xfrm_pols_put(pols, num_pols); 3100 return ERR_PTR(err); 3101 } 3102 3103 static struct dst_entry *make_blackhole(struct net *net, u16 family, 3104 struct dst_entry *dst_orig) 3105 { 3106 const struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family); 3107 struct dst_entry *ret; 3108 3109 if (!afinfo) { 3110 dst_release(dst_orig); 3111 return ERR_PTR(-EINVAL); 3112 } else { 3113 ret = afinfo->blackhole_route(net, dst_orig); 3114 } 3115 rcu_read_unlock(); 3116 3117 return ret; 3118 } 3119 3120 /* Finds/creates a bundle for given flow and if_id 3121 * 3122 * At the moment we eat a raw IP route. Mostly to speed up lookups 3123 * on interfaces with disabled IPsec. 3124 * 3125 * xfrm_lookup uses an if_id of 0 by default, and is provided for 3126 * compatibility 3127 */ 3128 struct dst_entry *xfrm_lookup_with_ifid(struct net *net, 3129 struct dst_entry *dst_orig, 3130 const struct flowi *fl, 3131 const struct sock *sk, 3132 int flags, u32 if_id) 3133 { 3134 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX]; 3135 struct xfrm_dst *xdst; 3136 struct dst_entry *dst, *route; 3137 u16 family = dst_orig->ops->family; 3138 u8 dir = XFRM_POLICY_OUT; 3139 int i, err, num_pols, num_xfrms = 0, drop_pols = 0; 3140 3141 dst = NULL; 3142 xdst = NULL; 3143 route = NULL; 3144 3145 sk = sk_const_to_full_sk(sk); 3146 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) { 3147 num_pols = 1; 3148 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl, family, 3149 if_id); 3150 err = xfrm_expand_policies(fl, family, pols, 3151 &num_pols, &num_xfrms); 3152 if (err < 0) 3153 goto dropdst; 3154 3155 if (num_pols) { 3156 if (num_xfrms <= 0) { 3157 drop_pols = num_pols; 3158 goto no_transform; 3159 } 3160 3161 xdst = xfrm_resolve_and_create_bundle( 3162 pols, num_pols, fl, 3163 family, dst_orig); 3164 3165 if (IS_ERR(xdst)) { 3166 xfrm_pols_put(pols, num_pols); 3167 err = PTR_ERR(xdst); 3168 if (err == -EREMOTE) 3169 goto nopol; 3170 3171 goto dropdst; 3172 } else if (xdst == NULL) { 3173 num_xfrms = 0; 3174 drop_pols = num_pols; 3175 goto no_transform; 3176 } 3177 3178 route = xdst->route; 3179 } 3180 } 3181 3182 if (xdst == NULL) { 3183 struct xfrm_flo xflo; 3184 3185 xflo.dst_orig = dst_orig; 3186 xflo.flags = flags; 3187 3188 /* To accelerate a bit... */ 3189 if (!if_id && ((dst_orig->flags & DST_NOXFRM) || 3190 !net->xfrm.policy_count[XFRM_POLICY_OUT])) 3191 goto nopol; 3192 3193 xdst = xfrm_bundle_lookup(net, fl, family, dir, &xflo, if_id); 3194 if (xdst == NULL) 3195 goto nopol; 3196 if (IS_ERR(xdst)) { 3197 err = PTR_ERR(xdst); 3198 goto dropdst; 3199 } 3200 3201 num_pols = xdst->num_pols; 3202 num_xfrms = xdst->num_xfrms; 3203 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy *) * num_pols); 3204 route = xdst->route; 3205 } 3206 3207 dst = &xdst->u.dst; 3208 if (route == NULL && num_xfrms > 0) { 3209 /* The only case when xfrm_bundle_lookup() returns a 3210 * bundle with null route, is when the template could 3211 * not be resolved. It means policies are there, but 3212 * bundle could not be created, since we don't yet 3213 * have the xfrm_state's. We need to wait for KM to 3214 * negotiate new SA's or bail out with error.*/ 3215 if (net->xfrm.sysctl_larval_drop) { 3216 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES); 3217 err = -EREMOTE; 3218 goto error; 3219 } 3220 3221 err = -EAGAIN; 3222 3223 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES); 3224 goto error; 3225 } 3226 3227 no_transform: 3228 if (num_pols == 0) 3229 goto nopol; 3230 3231 if ((flags & XFRM_LOOKUP_ICMP) && 3232 !(pols[0]->flags & XFRM_POLICY_ICMP)) { 3233 err = -ENOENT; 3234 goto error; 3235 } 3236 3237 for (i = 0; i < num_pols; i++) 3238 WRITE_ONCE(pols[i]->curlft.use_time, ktime_get_real_seconds()); 3239 3240 if (num_xfrms < 0) { 3241 /* Prohibit the flow */ 3242 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK); 3243 err = -EPERM; 3244 goto error; 3245 } else if (num_xfrms > 0) { 3246 /* Flow transformed */ 3247 dst_release(dst_orig); 3248 } else { 3249 /* Flow passes untransformed */ 3250 dst_release(dst); 3251 dst = dst_orig; 3252 } 3253 ok: 3254 xfrm_pols_put(pols, drop_pols); 3255 if (dst && dst->xfrm && 3256 dst->xfrm->props.mode == XFRM_MODE_TUNNEL) 3257 dst->flags |= DST_XFRM_TUNNEL; 3258 return dst; 3259 3260 nopol: 3261 if ((!dst_orig->dev || !(dst_orig->dev->flags & IFF_LOOPBACK)) && 3262 net->xfrm.policy_default[dir] == XFRM_USERPOLICY_BLOCK) { 3263 err = -EPERM; 3264 goto error; 3265 } 3266 if (!(flags & XFRM_LOOKUP_ICMP)) { 3267 dst = dst_orig; 3268 goto ok; 3269 } 3270 err = -ENOENT; 3271 error: 3272 dst_release(dst); 3273 dropdst: 3274 if (!(flags & XFRM_LOOKUP_KEEP_DST_REF)) 3275 dst_release(dst_orig); 3276 xfrm_pols_put(pols, drop_pols); 3277 return ERR_PTR(err); 3278 } 3279 EXPORT_SYMBOL(xfrm_lookup_with_ifid); 3280 3281 /* Main function: finds/creates a bundle for given flow. 3282 * 3283 * At the moment we eat a raw IP route. Mostly to speed up lookups 3284 * on interfaces with disabled IPsec. 3285 */ 3286 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig, 3287 const struct flowi *fl, const struct sock *sk, 3288 int flags) 3289 { 3290 return xfrm_lookup_with_ifid(net, dst_orig, fl, sk, flags, 0); 3291 } 3292 EXPORT_SYMBOL(xfrm_lookup); 3293 3294 /* Callers of xfrm_lookup_route() must ensure a call to dst_output(). 3295 * Otherwise we may send out blackholed packets. 3296 */ 3297 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig, 3298 const struct flowi *fl, 3299 const struct sock *sk, int flags) 3300 { 3301 struct dst_entry *dst = xfrm_lookup(net, dst_orig, fl, sk, 3302 flags | XFRM_LOOKUP_QUEUE | 3303 XFRM_LOOKUP_KEEP_DST_REF); 3304 3305 if (PTR_ERR(dst) == -EREMOTE) 3306 return make_blackhole(net, dst_orig->ops->family, dst_orig); 3307 3308 if (IS_ERR(dst)) 3309 dst_release(dst_orig); 3310 3311 return dst; 3312 } 3313 EXPORT_SYMBOL(xfrm_lookup_route); 3314 3315 static inline int 3316 xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl) 3317 { 3318 struct sec_path *sp = skb_sec_path(skb); 3319 struct xfrm_state *x; 3320 3321 if (!sp || idx < 0 || idx >= sp->len) 3322 return 0; 3323 x = sp->xvec[idx]; 3324 if (!x->type->reject) 3325 return 0; 3326 return x->type->reject(x, skb, fl); 3327 } 3328 3329 /* When skb is transformed back to its "native" form, we have to 3330 * check policy restrictions. At the moment we make this in maximally 3331 * stupid way. Shame on me. :-) Of course, connected sockets must 3332 * have policy cached at them. 3333 */ 3334 3335 static inline int 3336 xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x, 3337 unsigned short family, u32 if_id) 3338 { 3339 if (xfrm_state_kern(x)) 3340 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family); 3341 return x->id.proto == tmpl->id.proto && 3342 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) && 3343 (x->props.reqid == tmpl->reqid || !tmpl->reqid) && 3344 x->props.mode == tmpl->mode && 3345 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) || 3346 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) && 3347 !(x->props.mode != XFRM_MODE_TRANSPORT && 3348 xfrm_state_addr_cmp(tmpl, x, family)) && 3349 (if_id == 0 || if_id == x->if_id); 3350 } 3351 3352 /* 3353 * 0 or more than 0 is returned when validation is succeeded (either bypass 3354 * because of optional transport mode, or next index of the matched secpath 3355 * state with the template. 3356 * -1 is returned when no matching template is found. 3357 * Otherwise "-2 - errored_index" is returned. 3358 */ 3359 static inline int 3360 xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start, 3361 unsigned short family, u32 if_id) 3362 { 3363 int idx = start; 3364 3365 if (tmpl->optional) { 3366 if (tmpl->mode == XFRM_MODE_TRANSPORT) 3367 return start; 3368 } else 3369 start = -1; 3370 for (; idx < sp->len; idx++) { 3371 if (xfrm_state_ok(tmpl, sp->xvec[idx], family, if_id)) 3372 return ++idx; 3373 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) { 3374 if (idx < sp->verified_cnt) { 3375 /* Secpath entry previously verified, consider optional and 3376 * continue searching 3377 */ 3378 continue; 3379 } 3380 3381 if (start == -1) 3382 start = -2-idx; 3383 break; 3384 } 3385 } 3386 return start; 3387 } 3388 3389 static void 3390 decode_session4(const struct xfrm_flow_keys *flkeys, struct flowi *fl, bool reverse) 3391 { 3392 struct flowi4 *fl4 = &fl->u.ip4; 3393 3394 memset(fl4, 0, sizeof(struct flowi4)); 3395 3396 if (reverse) { 3397 fl4->saddr = flkeys->addrs.ipv4.dst; 3398 fl4->daddr = flkeys->addrs.ipv4.src; 3399 fl4->fl4_sport = flkeys->ports.dst; 3400 fl4->fl4_dport = flkeys->ports.src; 3401 } else { 3402 fl4->saddr = flkeys->addrs.ipv4.src; 3403 fl4->daddr = flkeys->addrs.ipv4.dst; 3404 fl4->fl4_sport = flkeys->ports.src; 3405 fl4->fl4_dport = flkeys->ports.dst; 3406 } 3407 3408 switch (flkeys->basic.ip_proto) { 3409 case IPPROTO_GRE: 3410 fl4->fl4_gre_key = flkeys->gre.keyid; 3411 break; 3412 case IPPROTO_ICMP: 3413 fl4->fl4_icmp_type = flkeys->icmp.type; 3414 fl4->fl4_icmp_code = flkeys->icmp.code; 3415 break; 3416 } 3417 3418 fl4->flowi4_proto = flkeys->basic.ip_proto; 3419 fl4->flowi4_tos = flkeys->ip.tos; 3420 } 3421 3422 #if IS_ENABLED(CONFIG_IPV6) 3423 static void 3424 decode_session6(const struct xfrm_flow_keys *flkeys, struct flowi *fl, bool reverse) 3425 { 3426 struct flowi6 *fl6 = &fl->u.ip6; 3427 3428 memset(fl6, 0, sizeof(struct flowi6)); 3429 3430 if (reverse) { 3431 fl6->saddr = flkeys->addrs.ipv6.dst; 3432 fl6->daddr = flkeys->addrs.ipv6.src; 3433 fl6->fl6_sport = flkeys->ports.dst; 3434 fl6->fl6_dport = flkeys->ports.src; 3435 } else { 3436 fl6->saddr = flkeys->addrs.ipv6.src; 3437 fl6->daddr = flkeys->addrs.ipv6.dst; 3438 fl6->fl6_sport = flkeys->ports.src; 3439 fl6->fl6_dport = flkeys->ports.dst; 3440 } 3441 3442 switch (flkeys->basic.ip_proto) { 3443 case IPPROTO_GRE: 3444 fl6->fl6_gre_key = flkeys->gre.keyid; 3445 break; 3446 case IPPROTO_ICMPV6: 3447 fl6->fl6_icmp_type = flkeys->icmp.type; 3448 fl6->fl6_icmp_code = flkeys->icmp.code; 3449 break; 3450 } 3451 3452 fl6->flowi6_proto = flkeys->basic.ip_proto; 3453 } 3454 #endif 3455 3456 int __xfrm_decode_session(struct net *net, struct sk_buff *skb, struct flowi *fl, 3457 unsigned int family, int reverse) 3458 { 3459 struct xfrm_flow_keys flkeys; 3460 3461 memset(&flkeys, 0, sizeof(flkeys)); 3462 __skb_flow_dissect(net, skb, &xfrm_session_dissector, &flkeys, 3463 NULL, 0, 0, 0, FLOW_DISSECTOR_F_STOP_AT_ENCAP); 3464 3465 switch (family) { 3466 case AF_INET: 3467 decode_session4(&flkeys, fl, reverse); 3468 break; 3469 #if IS_ENABLED(CONFIG_IPV6) 3470 case AF_INET6: 3471 decode_session6(&flkeys, fl, reverse); 3472 break; 3473 #endif 3474 default: 3475 return -EAFNOSUPPORT; 3476 } 3477 3478 fl->flowi_mark = skb->mark; 3479 if (reverse) { 3480 fl->flowi_oif = skb->skb_iif; 3481 } else { 3482 int oif = 0; 3483 3484 if (skb_dst(skb) && skb_dst(skb)->dev) 3485 oif = skb_dst(skb)->dev->ifindex; 3486 3487 fl->flowi_oif = oif; 3488 } 3489 3490 return security_xfrm_decode_session(skb, &fl->flowi_secid); 3491 } 3492 EXPORT_SYMBOL(__xfrm_decode_session); 3493 3494 static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp) 3495 { 3496 for (; k < sp->len; k++) { 3497 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) { 3498 *idxp = k; 3499 return 1; 3500 } 3501 } 3502 3503 return 0; 3504 } 3505 3506 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb, 3507 unsigned short family) 3508 { 3509 struct net *net = dev_net(skb->dev); 3510 struct xfrm_policy *pol; 3511 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX]; 3512 int npols = 0; 3513 int xfrm_nr; 3514 int pi; 3515 int reverse; 3516 struct flowi fl; 3517 int xerr_idx = -1; 3518 const struct xfrm_if_cb *ifcb; 3519 struct sec_path *sp; 3520 u32 if_id = 0; 3521 3522 rcu_read_lock(); 3523 ifcb = xfrm_if_get_cb(); 3524 3525 if (ifcb) { 3526 struct xfrm_if_decode_session_result r; 3527 3528 if (ifcb->decode_session(skb, family, &r)) { 3529 if_id = r.if_id; 3530 net = r.net; 3531 } 3532 } 3533 rcu_read_unlock(); 3534 3535 reverse = dir & ~XFRM_POLICY_MASK; 3536 dir &= XFRM_POLICY_MASK; 3537 3538 if (__xfrm_decode_session(net, skb, &fl, family, reverse) < 0) { 3539 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR); 3540 return 0; 3541 } 3542 3543 nf_nat_decode_session(skb, &fl, family); 3544 3545 /* First, check used SA against their selectors. */ 3546 sp = skb_sec_path(skb); 3547 if (sp) { 3548 int i; 3549 3550 for (i = sp->len - 1; i >= 0; i--) { 3551 struct xfrm_state *x = sp->xvec[i]; 3552 if (!xfrm_selector_match(&x->sel, &fl, family)) { 3553 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH); 3554 return 0; 3555 } 3556 } 3557 } 3558 3559 pol = NULL; 3560 sk = sk_to_full_sk(sk); 3561 if (sk && sk->sk_policy[dir]) { 3562 pol = xfrm_sk_policy_lookup(sk, dir, &fl, family, if_id); 3563 if (IS_ERR(pol)) { 3564 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR); 3565 return 0; 3566 } 3567 } 3568 3569 if (!pol) 3570 pol = xfrm_policy_lookup(net, &fl, family, dir, if_id); 3571 3572 if (IS_ERR(pol)) { 3573 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR); 3574 return 0; 3575 } 3576 3577 if (!pol) { 3578 if (net->xfrm.policy_default[dir] == XFRM_USERPOLICY_BLOCK) { 3579 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS); 3580 return 0; 3581 } 3582 3583 if (sp && secpath_has_nontransport(sp, 0, &xerr_idx)) { 3584 xfrm_secpath_reject(xerr_idx, skb, &fl); 3585 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS); 3586 return 0; 3587 } 3588 return 1; 3589 } 3590 3591 /* This lockless write can happen from different cpus. */ 3592 WRITE_ONCE(pol->curlft.use_time, ktime_get_real_seconds()); 3593 3594 pols[0] = pol; 3595 npols++; 3596 #ifdef CONFIG_XFRM_SUB_POLICY 3597 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) { 3598 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, 3599 &fl, family, 3600 XFRM_POLICY_IN, if_id); 3601 if (pols[1]) { 3602 if (IS_ERR(pols[1])) { 3603 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR); 3604 xfrm_pol_put(pols[0]); 3605 return 0; 3606 } 3607 /* This write can happen from different cpus. */ 3608 WRITE_ONCE(pols[1]->curlft.use_time, 3609 ktime_get_real_seconds()); 3610 npols++; 3611 } 3612 } 3613 #endif 3614 3615 if (pol->action == XFRM_POLICY_ALLOW) { 3616 static struct sec_path dummy; 3617 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH]; 3618 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH]; 3619 struct xfrm_tmpl **tpp = tp; 3620 int ti = 0; 3621 int i, k; 3622 3623 sp = skb_sec_path(skb); 3624 if (!sp) 3625 sp = &dummy; 3626 3627 for (pi = 0; pi < npols; pi++) { 3628 if (pols[pi] != pol && 3629 pols[pi]->action != XFRM_POLICY_ALLOW) { 3630 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK); 3631 goto reject; 3632 } 3633 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) { 3634 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR); 3635 goto reject_error; 3636 } 3637 for (i = 0; i < pols[pi]->xfrm_nr; i++) 3638 tpp[ti++] = &pols[pi]->xfrm_vec[i]; 3639 } 3640 xfrm_nr = ti; 3641 3642 if (npols > 1) { 3643 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family); 3644 tpp = stp; 3645 } 3646 3647 /* For each tunnel xfrm, find the first matching tmpl. 3648 * For each tmpl before that, find corresponding xfrm. 3649 * Order is _important_. Later we will implement 3650 * some barriers, but at the moment barriers 3651 * are implied between each two transformations. 3652 * Upon success, marks secpath entries as having been 3653 * verified to allow them to be skipped in future policy 3654 * checks (e.g. nested tunnels). 3655 */ 3656 for (i = xfrm_nr-1, k = 0; i >= 0; i--) { 3657 k = xfrm_policy_ok(tpp[i], sp, k, family, if_id); 3658 if (k < 0) { 3659 if (k < -1) 3660 /* "-2 - errored_index" returned */ 3661 xerr_idx = -(2+k); 3662 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH); 3663 goto reject; 3664 } 3665 } 3666 3667 if (secpath_has_nontransport(sp, k, &xerr_idx)) { 3668 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH); 3669 goto reject; 3670 } 3671 3672 xfrm_pols_put(pols, npols); 3673 sp->verified_cnt = k; 3674 3675 return 1; 3676 } 3677 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK); 3678 3679 reject: 3680 xfrm_secpath_reject(xerr_idx, skb, &fl); 3681 reject_error: 3682 xfrm_pols_put(pols, npols); 3683 return 0; 3684 } 3685 EXPORT_SYMBOL(__xfrm_policy_check); 3686 3687 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family) 3688 { 3689 struct net *net = dev_net(skb->dev); 3690 struct flowi fl; 3691 struct dst_entry *dst; 3692 int res = 1; 3693 3694 if (xfrm_decode_session(net, skb, &fl, family) < 0) { 3695 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR); 3696 return 0; 3697 } 3698 3699 skb_dst_force(skb); 3700 if (!skb_dst(skb)) { 3701 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR); 3702 return 0; 3703 } 3704 3705 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, XFRM_LOOKUP_QUEUE); 3706 if (IS_ERR(dst)) { 3707 res = 0; 3708 dst = NULL; 3709 } 3710 skb_dst_set(skb, dst); 3711 return res; 3712 } 3713 EXPORT_SYMBOL(__xfrm_route_forward); 3714 3715 /* Optimize later using cookies and generation ids. */ 3716 3717 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie) 3718 { 3719 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete 3720 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to 3721 * get validated by dst_ops->check on every use. We do this 3722 * because when a normal route referenced by an XFRM dst is 3723 * obsoleted we do not go looking around for all parent 3724 * referencing XFRM dsts so that we can invalidate them. It 3725 * is just too much work. Instead we make the checks here on 3726 * every use. For example: 3727 * 3728 * XFRM dst A --> IPv4 dst X 3729 * 3730 * X is the "xdst->route" of A (X is also the "dst->path" of A 3731 * in this example). If X is marked obsolete, "A" will not 3732 * notice. That's what we are validating here via the 3733 * stale_bundle() check. 3734 * 3735 * When a dst is removed from the fib tree, DST_OBSOLETE_DEAD will 3736 * be marked on it. 3737 * This will force stale_bundle() to fail on any xdst bundle with 3738 * this dst linked in it. 3739 */ 3740 if (dst->obsolete < 0 && !stale_bundle(dst)) 3741 return dst; 3742 3743 return NULL; 3744 } 3745 3746 static int stale_bundle(struct dst_entry *dst) 3747 { 3748 return !xfrm_bundle_ok((struct xfrm_dst *)dst); 3749 } 3750 3751 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev) 3752 { 3753 while ((dst = xfrm_dst_child(dst)) && dst->xfrm && dst->dev == dev) { 3754 dst->dev = blackhole_netdev; 3755 dev_hold(dst->dev); 3756 dev_put(dev); 3757 } 3758 } 3759 EXPORT_SYMBOL(xfrm_dst_ifdown); 3760 3761 static void xfrm_link_failure(struct sk_buff *skb) 3762 { 3763 /* Impossible. Such dst must be popped before reaches point of failure. */ 3764 } 3765 3766 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst) 3767 { 3768 if (dst) { 3769 if (dst->obsolete) { 3770 dst_release(dst); 3771 dst = NULL; 3772 } 3773 } 3774 return dst; 3775 } 3776 3777 static void xfrm_init_pmtu(struct xfrm_dst **bundle, int nr) 3778 { 3779 while (nr--) { 3780 struct xfrm_dst *xdst = bundle[nr]; 3781 u32 pmtu, route_mtu_cached; 3782 struct dst_entry *dst; 3783 3784 dst = &xdst->u.dst; 3785 pmtu = dst_mtu(xfrm_dst_child(dst)); 3786 xdst->child_mtu_cached = pmtu; 3787 3788 pmtu = xfrm_state_mtu(dst->xfrm, pmtu); 3789 3790 route_mtu_cached = dst_mtu(xdst->route); 3791 xdst->route_mtu_cached = route_mtu_cached; 3792 3793 if (pmtu > route_mtu_cached) 3794 pmtu = route_mtu_cached; 3795 3796 dst_metric_set(dst, RTAX_MTU, pmtu); 3797 } 3798 } 3799 3800 /* Check that the bundle accepts the flow and its components are 3801 * still valid. 3802 */ 3803 3804 static int xfrm_bundle_ok(struct xfrm_dst *first) 3805 { 3806 struct xfrm_dst *bundle[XFRM_MAX_DEPTH]; 3807 struct dst_entry *dst = &first->u.dst; 3808 struct xfrm_dst *xdst; 3809 int start_from, nr; 3810 u32 mtu; 3811 3812 if (!dst_check(xfrm_dst_path(dst), ((struct xfrm_dst *)dst)->path_cookie) || 3813 (dst->dev && !netif_running(dst->dev))) 3814 return 0; 3815 3816 if (dst->flags & DST_XFRM_QUEUE) 3817 return 1; 3818 3819 start_from = nr = 0; 3820 do { 3821 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 3822 3823 if (dst->xfrm->km.state != XFRM_STATE_VALID) 3824 return 0; 3825 if (xdst->xfrm_genid != dst->xfrm->genid) 3826 return 0; 3827 if (xdst->num_pols > 0 && 3828 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid)) 3829 return 0; 3830 3831 bundle[nr++] = xdst; 3832 3833 mtu = dst_mtu(xfrm_dst_child(dst)); 3834 if (xdst->child_mtu_cached != mtu) { 3835 start_from = nr; 3836 xdst->child_mtu_cached = mtu; 3837 } 3838 3839 if (!dst_check(xdst->route, xdst->route_cookie)) 3840 return 0; 3841 mtu = dst_mtu(xdst->route); 3842 if (xdst->route_mtu_cached != mtu) { 3843 start_from = nr; 3844 xdst->route_mtu_cached = mtu; 3845 } 3846 3847 dst = xfrm_dst_child(dst); 3848 } while (dst->xfrm); 3849 3850 if (likely(!start_from)) 3851 return 1; 3852 3853 xdst = bundle[start_from - 1]; 3854 mtu = xdst->child_mtu_cached; 3855 while (start_from--) { 3856 dst = &xdst->u.dst; 3857 3858 mtu = xfrm_state_mtu(dst->xfrm, mtu); 3859 if (mtu > xdst->route_mtu_cached) 3860 mtu = xdst->route_mtu_cached; 3861 dst_metric_set(dst, RTAX_MTU, mtu); 3862 if (!start_from) 3863 break; 3864 3865 xdst = bundle[start_from - 1]; 3866 xdst->child_mtu_cached = mtu; 3867 } 3868 3869 return 1; 3870 } 3871 3872 static unsigned int xfrm_default_advmss(const struct dst_entry *dst) 3873 { 3874 return dst_metric_advmss(xfrm_dst_path(dst)); 3875 } 3876 3877 static unsigned int xfrm_mtu(const struct dst_entry *dst) 3878 { 3879 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU); 3880 3881 return mtu ? : dst_mtu(xfrm_dst_path(dst)); 3882 } 3883 3884 static const void *xfrm_get_dst_nexthop(const struct dst_entry *dst, 3885 const void *daddr) 3886 { 3887 while (dst->xfrm) { 3888 const struct xfrm_state *xfrm = dst->xfrm; 3889 3890 dst = xfrm_dst_child(dst); 3891 3892 if (xfrm->props.mode == XFRM_MODE_TRANSPORT) 3893 continue; 3894 if (xfrm->type->flags & XFRM_TYPE_REMOTE_COADDR) 3895 daddr = xfrm->coaddr; 3896 else if (!(xfrm->type->flags & XFRM_TYPE_LOCAL_COADDR)) 3897 daddr = &xfrm->id.daddr; 3898 } 3899 return daddr; 3900 } 3901 3902 static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst, 3903 struct sk_buff *skb, 3904 const void *daddr) 3905 { 3906 const struct dst_entry *path = xfrm_dst_path(dst); 3907 3908 if (!skb) 3909 daddr = xfrm_get_dst_nexthop(dst, daddr); 3910 return path->ops->neigh_lookup(path, skb, daddr); 3911 } 3912 3913 static void xfrm_confirm_neigh(const struct dst_entry *dst, const void *daddr) 3914 { 3915 const struct dst_entry *path = xfrm_dst_path(dst); 3916 3917 daddr = xfrm_get_dst_nexthop(dst, daddr); 3918 path->ops->confirm_neigh(path, daddr); 3919 } 3920 3921 int xfrm_policy_register_afinfo(const struct xfrm_policy_afinfo *afinfo, int family) 3922 { 3923 int err = 0; 3924 3925 if (WARN_ON(family >= ARRAY_SIZE(xfrm_policy_afinfo))) 3926 return -EAFNOSUPPORT; 3927 3928 spin_lock(&xfrm_policy_afinfo_lock); 3929 if (unlikely(xfrm_policy_afinfo[family] != NULL)) 3930 err = -EEXIST; 3931 else { 3932 struct dst_ops *dst_ops = afinfo->dst_ops; 3933 if (likely(dst_ops->kmem_cachep == NULL)) 3934 dst_ops->kmem_cachep = xfrm_dst_cache; 3935 if (likely(dst_ops->check == NULL)) 3936 dst_ops->check = xfrm_dst_check; 3937 if (likely(dst_ops->default_advmss == NULL)) 3938 dst_ops->default_advmss = xfrm_default_advmss; 3939 if (likely(dst_ops->mtu == NULL)) 3940 dst_ops->mtu = xfrm_mtu; 3941 if (likely(dst_ops->negative_advice == NULL)) 3942 dst_ops->negative_advice = xfrm_negative_advice; 3943 if (likely(dst_ops->link_failure == NULL)) 3944 dst_ops->link_failure = xfrm_link_failure; 3945 if (likely(dst_ops->neigh_lookup == NULL)) 3946 dst_ops->neigh_lookup = xfrm_neigh_lookup; 3947 if (likely(!dst_ops->confirm_neigh)) 3948 dst_ops->confirm_neigh = xfrm_confirm_neigh; 3949 rcu_assign_pointer(xfrm_policy_afinfo[family], afinfo); 3950 } 3951 spin_unlock(&xfrm_policy_afinfo_lock); 3952 3953 return err; 3954 } 3955 EXPORT_SYMBOL(xfrm_policy_register_afinfo); 3956 3957 void xfrm_policy_unregister_afinfo(const struct xfrm_policy_afinfo *afinfo) 3958 { 3959 struct dst_ops *dst_ops = afinfo->dst_ops; 3960 int i; 3961 3962 for (i = 0; i < ARRAY_SIZE(xfrm_policy_afinfo); i++) { 3963 if (xfrm_policy_afinfo[i] != afinfo) 3964 continue; 3965 RCU_INIT_POINTER(xfrm_policy_afinfo[i], NULL); 3966 break; 3967 } 3968 3969 synchronize_rcu(); 3970 3971 dst_ops->kmem_cachep = NULL; 3972 dst_ops->check = NULL; 3973 dst_ops->negative_advice = NULL; 3974 dst_ops->link_failure = NULL; 3975 } 3976 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo); 3977 3978 void xfrm_if_register_cb(const struct xfrm_if_cb *ifcb) 3979 { 3980 spin_lock(&xfrm_if_cb_lock); 3981 rcu_assign_pointer(xfrm_if_cb, ifcb); 3982 spin_unlock(&xfrm_if_cb_lock); 3983 } 3984 EXPORT_SYMBOL(xfrm_if_register_cb); 3985 3986 void xfrm_if_unregister_cb(void) 3987 { 3988 RCU_INIT_POINTER(xfrm_if_cb, NULL); 3989 synchronize_rcu(); 3990 } 3991 EXPORT_SYMBOL(xfrm_if_unregister_cb); 3992 3993 #ifdef CONFIG_XFRM_STATISTICS 3994 static int __net_init xfrm_statistics_init(struct net *net) 3995 { 3996 int rv; 3997 net->mib.xfrm_statistics = alloc_percpu(struct linux_xfrm_mib); 3998 if (!net->mib.xfrm_statistics) 3999 return -ENOMEM; 4000 rv = xfrm_proc_init(net); 4001 if (rv < 0) 4002 free_percpu(net->mib.xfrm_statistics); 4003 return rv; 4004 } 4005 4006 static void xfrm_statistics_fini(struct net *net) 4007 { 4008 xfrm_proc_fini(net); 4009 free_percpu(net->mib.xfrm_statistics); 4010 } 4011 #else 4012 static int __net_init xfrm_statistics_init(struct net *net) 4013 { 4014 return 0; 4015 } 4016 4017 static void xfrm_statistics_fini(struct net *net) 4018 { 4019 } 4020 #endif 4021 4022 static int __net_init xfrm_policy_init(struct net *net) 4023 { 4024 unsigned int hmask, sz; 4025 int dir, err; 4026 4027 if (net_eq(net, &init_net)) { 4028 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache", 4029 sizeof(struct xfrm_dst), 4030 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, 4031 NULL); 4032 err = rhashtable_init(&xfrm_policy_inexact_table, 4033 &xfrm_pol_inexact_params); 4034 BUG_ON(err); 4035 } 4036 4037 hmask = 8 - 1; 4038 sz = (hmask+1) * sizeof(struct hlist_head); 4039 4040 net->xfrm.policy_byidx = xfrm_hash_alloc(sz); 4041 if (!net->xfrm.policy_byidx) 4042 goto out_byidx; 4043 net->xfrm.policy_idx_hmask = hmask; 4044 4045 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) { 4046 struct xfrm_policy_hash *htab; 4047 4048 net->xfrm.policy_count[dir] = 0; 4049 net->xfrm.policy_count[XFRM_POLICY_MAX + dir] = 0; 4050 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]); 4051 4052 htab = &net->xfrm.policy_bydst[dir]; 4053 htab->table = xfrm_hash_alloc(sz); 4054 if (!htab->table) 4055 goto out_bydst; 4056 htab->hmask = hmask; 4057 htab->dbits4 = 32; 4058 htab->sbits4 = 32; 4059 htab->dbits6 = 128; 4060 htab->sbits6 = 128; 4061 } 4062 net->xfrm.policy_hthresh.lbits4 = 32; 4063 net->xfrm.policy_hthresh.rbits4 = 32; 4064 net->xfrm.policy_hthresh.lbits6 = 128; 4065 net->xfrm.policy_hthresh.rbits6 = 128; 4066 4067 seqlock_init(&net->xfrm.policy_hthresh.lock); 4068 4069 INIT_LIST_HEAD(&net->xfrm.policy_all); 4070 INIT_LIST_HEAD(&net->xfrm.inexact_bins); 4071 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize); 4072 INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild); 4073 return 0; 4074 4075 out_bydst: 4076 for (dir--; dir >= 0; dir--) { 4077 struct xfrm_policy_hash *htab; 4078 4079 htab = &net->xfrm.policy_bydst[dir]; 4080 xfrm_hash_free(htab->table, sz); 4081 } 4082 xfrm_hash_free(net->xfrm.policy_byidx, sz); 4083 out_byidx: 4084 return -ENOMEM; 4085 } 4086 4087 static void xfrm_policy_fini(struct net *net) 4088 { 4089 struct xfrm_pol_inexact_bin *b, *t; 4090 unsigned int sz; 4091 int dir; 4092 4093 flush_work(&net->xfrm.policy_hash_work); 4094 #ifdef CONFIG_XFRM_SUB_POLICY 4095 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, false); 4096 #endif 4097 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, false); 4098 4099 WARN_ON(!list_empty(&net->xfrm.policy_all)); 4100 4101 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) { 4102 struct xfrm_policy_hash *htab; 4103 4104 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir])); 4105 4106 htab = &net->xfrm.policy_bydst[dir]; 4107 sz = (htab->hmask + 1) * sizeof(struct hlist_head); 4108 WARN_ON(!hlist_empty(htab->table)); 4109 xfrm_hash_free(htab->table, sz); 4110 } 4111 4112 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head); 4113 WARN_ON(!hlist_empty(net->xfrm.policy_byidx)); 4114 xfrm_hash_free(net->xfrm.policy_byidx, sz); 4115 4116 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 4117 list_for_each_entry_safe(b, t, &net->xfrm.inexact_bins, inexact_bins) 4118 __xfrm_policy_inexact_prune_bin(b, true); 4119 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 4120 } 4121 4122 static int __net_init xfrm_net_init(struct net *net) 4123 { 4124 int rv; 4125 4126 /* Initialize the per-net locks here */ 4127 spin_lock_init(&net->xfrm.xfrm_state_lock); 4128 spin_lock_init(&net->xfrm.xfrm_policy_lock); 4129 seqcount_spinlock_init(&net->xfrm.xfrm_policy_hash_generation, &net->xfrm.xfrm_policy_lock); 4130 mutex_init(&net->xfrm.xfrm_cfg_mutex); 4131 net->xfrm.policy_default[XFRM_POLICY_IN] = XFRM_USERPOLICY_ACCEPT; 4132 net->xfrm.policy_default[XFRM_POLICY_FWD] = XFRM_USERPOLICY_ACCEPT; 4133 net->xfrm.policy_default[XFRM_POLICY_OUT] = XFRM_USERPOLICY_ACCEPT; 4134 4135 rv = xfrm_statistics_init(net); 4136 if (rv < 0) 4137 goto out_statistics; 4138 rv = xfrm_state_init(net); 4139 if (rv < 0) 4140 goto out_state; 4141 rv = xfrm_policy_init(net); 4142 if (rv < 0) 4143 goto out_policy; 4144 rv = xfrm_sysctl_init(net); 4145 if (rv < 0) 4146 goto out_sysctl; 4147 4148 return 0; 4149 4150 out_sysctl: 4151 xfrm_policy_fini(net); 4152 out_policy: 4153 xfrm_state_fini(net); 4154 out_state: 4155 xfrm_statistics_fini(net); 4156 out_statistics: 4157 return rv; 4158 } 4159 4160 static void __net_exit xfrm_net_exit(struct net *net) 4161 { 4162 xfrm_sysctl_fini(net); 4163 xfrm_policy_fini(net); 4164 xfrm_state_fini(net); 4165 xfrm_statistics_fini(net); 4166 } 4167 4168 static struct pernet_operations __net_initdata xfrm_net_ops = { 4169 .init = xfrm_net_init, 4170 .exit = xfrm_net_exit, 4171 }; 4172 4173 static const struct flow_dissector_key xfrm_flow_dissector_keys[] = { 4174 { 4175 .key_id = FLOW_DISSECTOR_KEY_CONTROL, 4176 .offset = offsetof(struct xfrm_flow_keys, control), 4177 }, 4178 { 4179 .key_id = FLOW_DISSECTOR_KEY_BASIC, 4180 .offset = offsetof(struct xfrm_flow_keys, basic), 4181 }, 4182 { 4183 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, 4184 .offset = offsetof(struct xfrm_flow_keys, addrs.ipv4), 4185 }, 4186 { 4187 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, 4188 .offset = offsetof(struct xfrm_flow_keys, addrs.ipv6), 4189 }, 4190 { 4191 .key_id = FLOW_DISSECTOR_KEY_PORTS, 4192 .offset = offsetof(struct xfrm_flow_keys, ports), 4193 }, 4194 { 4195 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, 4196 .offset = offsetof(struct xfrm_flow_keys, gre), 4197 }, 4198 { 4199 .key_id = FLOW_DISSECTOR_KEY_IP, 4200 .offset = offsetof(struct xfrm_flow_keys, ip), 4201 }, 4202 { 4203 .key_id = FLOW_DISSECTOR_KEY_ICMP, 4204 .offset = offsetof(struct xfrm_flow_keys, icmp), 4205 }, 4206 }; 4207 4208 void __init xfrm_init(void) 4209 { 4210 skb_flow_dissector_init(&xfrm_session_dissector, 4211 xfrm_flow_dissector_keys, 4212 ARRAY_SIZE(xfrm_flow_dissector_keys)); 4213 4214 register_pernet_subsys(&xfrm_net_ops); 4215 xfrm_dev_init(); 4216 xfrm_input_init(); 4217 4218 #ifdef CONFIG_XFRM_ESPINTCP 4219 espintcp_init(); 4220 #endif 4221 } 4222 4223 #ifdef CONFIG_AUDITSYSCALL 4224 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp, 4225 struct audit_buffer *audit_buf) 4226 { 4227 struct xfrm_sec_ctx *ctx = xp->security; 4228 struct xfrm_selector *sel = &xp->selector; 4229 4230 if (ctx) 4231 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s", 4232 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str); 4233 4234 switch (sel->family) { 4235 case AF_INET: 4236 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4); 4237 if (sel->prefixlen_s != 32) 4238 audit_log_format(audit_buf, " src_prefixlen=%d", 4239 sel->prefixlen_s); 4240 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4); 4241 if (sel->prefixlen_d != 32) 4242 audit_log_format(audit_buf, " dst_prefixlen=%d", 4243 sel->prefixlen_d); 4244 break; 4245 case AF_INET6: 4246 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6); 4247 if (sel->prefixlen_s != 128) 4248 audit_log_format(audit_buf, " src_prefixlen=%d", 4249 sel->prefixlen_s); 4250 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6); 4251 if (sel->prefixlen_d != 128) 4252 audit_log_format(audit_buf, " dst_prefixlen=%d", 4253 sel->prefixlen_d); 4254 break; 4255 } 4256 } 4257 4258 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result, bool task_valid) 4259 { 4260 struct audit_buffer *audit_buf; 4261 4262 audit_buf = xfrm_audit_start("SPD-add"); 4263 if (audit_buf == NULL) 4264 return; 4265 xfrm_audit_helper_usrinfo(task_valid, audit_buf); 4266 audit_log_format(audit_buf, " res=%u", result); 4267 xfrm_audit_common_policyinfo(xp, audit_buf); 4268 audit_log_end(audit_buf); 4269 } 4270 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add); 4271 4272 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result, 4273 bool task_valid) 4274 { 4275 struct audit_buffer *audit_buf; 4276 4277 audit_buf = xfrm_audit_start("SPD-delete"); 4278 if (audit_buf == NULL) 4279 return; 4280 xfrm_audit_helper_usrinfo(task_valid, audit_buf); 4281 audit_log_format(audit_buf, " res=%u", result); 4282 xfrm_audit_common_policyinfo(xp, audit_buf); 4283 audit_log_end(audit_buf); 4284 } 4285 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete); 4286 #endif 4287 4288 #ifdef CONFIG_XFRM_MIGRATE 4289 static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp, 4290 const struct xfrm_selector *sel_tgt) 4291 { 4292 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) { 4293 if (sel_tgt->family == sel_cmp->family && 4294 xfrm_addr_equal(&sel_tgt->daddr, &sel_cmp->daddr, 4295 sel_cmp->family) && 4296 xfrm_addr_equal(&sel_tgt->saddr, &sel_cmp->saddr, 4297 sel_cmp->family) && 4298 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d && 4299 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) { 4300 return true; 4301 } 4302 } else { 4303 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) { 4304 return true; 4305 } 4306 } 4307 return false; 4308 } 4309 4310 static struct xfrm_policy *xfrm_migrate_policy_find(const struct xfrm_selector *sel, 4311 u8 dir, u8 type, struct net *net, u32 if_id) 4312 { 4313 struct xfrm_policy *pol, *ret = NULL; 4314 struct hlist_head *chain; 4315 u32 priority = ~0U; 4316 4317 spin_lock_bh(&net->xfrm.xfrm_policy_lock); 4318 chain = policy_hash_direct(net, &sel->daddr, &sel->saddr, sel->family, dir); 4319 hlist_for_each_entry(pol, chain, bydst) { 4320 if ((if_id == 0 || pol->if_id == if_id) && 4321 xfrm_migrate_selector_match(sel, &pol->selector) && 4322 pol->type == type) { 4323 ret = pol; 4324 priority = ret->priority; 4325 break; 4326 } 4327 } 4328 chain = &net->xfrm.policy_inexact[dir]; 4329 hlist_for_each_entry(pol, chain, bydst_inexact_list) { 4330 if ((pol->priority >= priority) && ret) 4331 break; 4332 4333 if ((if_id == 0 || pol->if_id == if_id) && 4334 xfrm_migrate_selector_match(sel, &pol->selector) && 4335 pol->type == type) { 4336 ret = pol; 4337 break; 4338 } 4339 } 4340 4341 xfrm_pol_hold(ret); 4342 4343 spin_unlock_bh(&net->xfrm.xfrm_policy_lock); 4344 4345 return ret; 4346 } 4347 4348 static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t) 4349 { 4350 int match = 0; 4351 4352 if (t->mode == m->mode && t->id.proto == m->proto && 4353 (m->reqid == 0 || t->reqid == m->reqid)) { 4354 switch (t->mode) { 4355 case XFRM_MODE_TUNNEL: 4356 case XFRM_MODE_BEET: 4357 if (xfrm_addr_equal(&t->id.daddr, &m->old_daddr, 4358 m->old_family) && 4359 xfrm_addr_equal(&t->saddr, &m->old_saddr, 4360 m->old_family)) { 4361 match = 1; 4362 } 4363 break; 4364 case XFRM_MODE_TRANSPORT: 4365 /* in case of transport mode, template does not store 4366 any IP addresses, hence we just compare mode and 4367 protocol */ 4368 match = 1; 4369 break; 4370 default: 4371 break; 4372 } 4373 } 4374 return match; 4375 } 4376 4377 /* update endpoint address(es) of template(s) */ 4378 static int xfrm_policy_migrate(struct xfrm_policy *pol, 4379 struct xfrm_migrate *m, int num_migrate, 4380 struct netlink_ext_ack *extack) 4381 { 4382 struct xfrm_migrate *mp; 4383 int i, j, n = 0; 4384 4385 write_lock_bh(&pol->lock); 4386 if (unlikely(pol->walk.dead)) { 4387 /* target policy has been deleted */ 4388 NL_SET_ERR_MSG(extack, "Target policy not found"); 4389 write_unlock_bh(&pol->lock); 4390 return -ENOENT; 4391 } 4392 4393 for (i = 0; i < pol->xfrm_nr; i++) { 4394 for (j = 0, mp = m; j < num_migrate; j++, mp++) { 4395 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i])) 4396 continue; 4397 n++; 4398 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL && 4399 pol->xfrm_vec[i].mode != XFRM_MODE_BEET) 4400 continue; 4401 /* update endpoints */ 4402 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr, 4403 sizeof(pol->xfrm_vec[i].id.daddr)); 4404 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr, 4405 sizeof(pol->xfrm_vec[i].saddr)); 4406 pol->xfrm_vec[i].encap_family = mp->new_family; 4407 /* flush bundles */ 4408 atomic_inc(&pol->genid); 4409 } 4410 } 4411 4412 write_unlock_bh(&pol->lock); 4413 4414 if (!n) 4415 return -ENODATA; 4416 4417 return 0; 4418 } 4419 4420 static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate, 4421 struct netlink_ext_ack *extack) 4422 { 4423 int i, j; 4424 4425 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH) { 4426 NL_SET_ERR_MSG(extack, "Invalid number of SAs to migrate, must be 0 < num <= XFRM_MAX_DEPTH (6)"); 4427 return -EINVAL; 4428 } 4429 4430 for (i = 0; i < num_migrate; i++) { 4431 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) || 4432 xfrm_addr_any(&m[i].new_saddr, m[i].new_family)) { 4433 NL_SET_ERR_MSG(extack, "Addresses in the MIGRATE attribute's list cannot be null"); 4434 return -EINVAL; 4435 } 4436 4437 /* check if there is any duplicated entry */ 4438 for (j = i + 1; j < num_migrate; j++) { 4439 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr, 4440 sizeof(m[i].old_daddr)) && 4441 !memcmp(&m[i].old_saddr, &m[j].old_saddr, 4442 sizeof(m[i].old_saddr)) && 4443 m[i].proto == m[j].proto && 4444 m[i].mode == m[j].mode && 4445 m[i].reqid == m[j].reqid && 4446 m[i].old_family == m[j].old_family) { 4447 NL_SET_ERR_MSG(extack, "Entries in the MIGRATE attribute's list must be unique"); 4448 return -EINVAL; 4449 } 4450 } 4451 } 4452 4453 return 0; 4454 } 4455 4456 int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type, 4457 struct xfrm_migrate *m, int num_migrate, 4458 struct xfrm_kmaddress *k, struct net *net, 4459 struct xfrm_encap_tmpl *encap, u32 if_id, 4460 struct netlink_ext_ack *extack) 4461 { 4462 int i, err, nx_cur = 0, nx_new = 0; 4463 struct xfrm_policy *pol = NULL; 4464 struct xfrm_state *x, *xc; 4465 struct xfrm_state *x_cur[XFRM_MAX_DEPTH]; 4466 struct xfrm_state *x_new[XFRM_MAX_DEPTH]; 4467 struct xfrm_migrate *mp; 4468 4469 /* Stage 0 - sanity checks */ 4470 err = xfrm_migrate_check(m, num_migrate, extack); 4471 if (err < 0) 4472 goto out; 4473 4474 if (dir >= XFRM_POLICY_MAX) { 4475 NL_SET_ERR_MSG(extack, "Invalid policy direction"); 4476 err = -EINVAL; 4477 goto out; 4478 } 4479 4480 /* Stage 1 - find policy */ 4481 pol = xfrm_migrate_policy_find(sel, dir, type, net, if_id); 4482 if (!pol) { 4483 NL_SET_ERR_MSG(extack, "Target policy not found"); 4484 err = -ENOENT; 4485 goto out; 4486 } 4487 4488 /* Stage 2 - find and update state(s) */ 4489 for (i = 0, mp = m; i < num_migrate; i++, mp++) { 4490 if ((x = xfrm_migrate_state_find(mp, net, if_id))) { 4491 x_cur[nx_cur] = x; 4492 nx_cur++; 4493 xc = xfrm_state_migrate(x, mp, encap); 4494 if (xc) { 4495 x_new[nx_new] = xc; 4496 nx_new++; 4497 } else { 4498 err = -ENODATA; 4499 goto restore_state; 4500 } 4501 } 4502 } 4503 4504 /* Stage 3 - update policy */ 4505 err = xfrm_policy_migrate(pol, m, num_migrate, extack); 4506 if (err < 0) 4507 goto restore_state; 4508 4509 /* Stage 4 - delete old state(s) */ 4510 if (nx_cur) { 4511 xfrm_states_put(x_cur, nx_cur); 4512 xfrm_states_delete(x_cur, nx_cur); 4513 } 4514 4515 /* Stage 5 - announce */ 4516 km_migrate(sel, dir, type, m, num_migrate, k, encap); 4517 4518 xfrm_pol_put(pol); 4519 4520 return 0; 4521 out: 4522 return err; 4523 4524 restore_state: 4525 if (pol) 4526 xfrm_pol_put(pol); 4527 if (nx_cur) 4528 xfrm_states_put(x_cur, nx_cur); 4529 if (nx_new) 4530 xfrm_states_delete(x_new, nx_new); 4531 4532 return err; 4533 } 4534 EXPORT_SYMBOL(xfrm_migrate); 4535 #endif 4536