1// SPDX-License-Identifier: GPL-2.0-or-later 2/* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Generic INET transport hashtables 8 * 9 * Authors: Lotsa people, from code originally in tcp 10 */ 11 12#include <linux/module.h> 13#include <linux/random.h> 14#include <linux/sched.h> 15#include <linux/slab.h> 16#include <linux/wait.h> 17#include <linux/vmalloc.h> 18#include <linux/memblock.h> 19 20#include <net/addrconf.h> 21#include <net/inet_connection_sock.h> 22#include <net/inet_hashtables.h> 23#if IS_ENABLED(CONFIG_IPV6) 24#include <net/inet6_hashtables.h> 25#endif 26#include <net/secure_seq.h> 27#include <net/ip.h> 28#include <net/tcp.h> 29#include <net/sock_reuseport.h> 30 31u32 inet_ehashfn(const struct net *net, const __be32 laddr, 32 const __u16 lport, const __be32 faddr, 33 const __be16 fport) 34{ 35 static u32 inet_ehash_secret __read_mostly; 36 37 net_get_random_once(&inet_ehash_secret, sizeof(inet_ehash_secret)); 38 39 return __inet_ehashfn(laddr, lport, faddr, fport, 40 inet_ehash_secret + net_hash_mix(net)); 41} 42EXPORT_SYMBOL_GPL(inet_ehashfn); 43 44/* This function handles inet_sock, but also timewait and request sockets 45 * for IPv4/IPv6. 46 */ 47static u32 sk_ehashfn(const struct sock *sk) 48{ 49#if IS_ENABLED(CONFIG_IPV6) 50 if (sk->sk_family == AF_INET6 && 51 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) 52 return inet6_ehashfn(sock_net(sk), 53 &sk->sk_v6_rcv_saddr, sk->sk_num, 54 &sk->sk_v6_daddr, sk->sk_dport); 55#endif 56 return inet_ehashfn(sock_net(sk), 57 sk->sk_rcv_saddr, sk->sk_num, 58 sk->sk_daddr, sk->sk_dport); 59} 60 61/* 62 * Allocate and initialize a new local port bind bucket. 63 * The bindhash mutex for snum's hash chain must be held here. 64 */ 65struct inet_bind_bucket *inet_bind_bucket_create(struct kmem_cache *cachep, 66 struct net *net, 67 struct inet_bind_hashbucket *head, 68 const unsigned short snum, 69 int l3mdev) 70{ 71 struct inet_bind_bucket *tb = kmem_cache_alloc(cachep, GFP_ATOMIC); 72 73 if (tb) { 74 write_pnet(&tb->ib_net, net); 75 tb->l3mdev = l3mdev; 76 tb->port = snum; 77 tb->fastreuse = 0; 78 tb->fastreuseport = 0; 79 INIT_HLIST_HEAD(&tb->owners); 80 hlist_add_head(&tb->node, &head->chain); 81 } 82 return tb; 83} 84 85/* 86 * Caller must hold hashbucket lock for this tb with local BH disabled 87 */ 88void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb) 89{ 90 if (hlist_empty(&tb->owners)) { 91 __hlist_del(&tb->node); 92 kmem_cache_free(cachep, tb); 93 } 94} 95 96bool inet_bind_bucket_match(const struct inet_bind_bucket *tb, const struct net *net, 97 unsigned short port, int l3mdev) 98{ 99 return net_eq(ib_net(tb), net) && tb->port == port && 100 tb->l3mdev == l3mdev; 101} 102 103static void inet_bind2_bucket_init(struct inet_bind2_bucket *tb, 104 struct net *net, 105 struct inet_bind_hashbucket *head, 106 unsigned short port, int l3mdev, 107 const struct sock *sk) 108{ 109 write_pnet(&tb->ib_net, net); 110 tb->l3mdev = l3mdev; 111 tb->port = port; 112#if IS_ENABLED(CONFIG_IPV6) 113 tb->family = sk->sk_family; 114 if (sk->sk_family == AF_INET6) 115 tb->v6_rcv_saddr = sk->sk_v6_rcv_saddr; 116 else 117#endif 118 tb->rcv_saddr = sk->sk_rcv_saddr; 119 INIT_HLIST_HEAD(&tb->owners); 120 INIT_HLIST_HEAD(&tb->deathrow); 121 hlist_add_head(&tb->node, &head->chain); 122} 123 124struct inet_bind2_bucket *inet_bind2_bucket_create(struct kmem_cache *cachep, 125 struct net *net, 126 struct inet_bind_hashbucket *head, 127 unsigned short port, 128 int l3mdev, 129 const struct sock *sk) 130{ 131 struct inet_bind2_bucket *tb = kmem_cache_alloc(cachep, GFP_ATOMIC); 132 133 if (tb) 134 inet_bind2_bucket_init(tb, net, head, port, l3mdev, sk); 135 136 return tb; 137} 138 139/* Caller must hold hashbucket lock for this tb with local BH disabled */ 140void inet_bind2_bucket_destroy(struct kmem_cache *cachep, struct inet_bind2_bucket *tb) 141{ 142 if (hlist_empty(&tb->owners) && hlist_empty(&tb->deathrow)) { 143 __hlist_del(&tb->node); 144 kmem_cache_free(cachep, tb); 145 } 146} 147 148static bool inet_bind2_bucket_addr_match(const struct inet_bind2_bucket *tb2, 149 const struct sock *sk) 150{ 151#if IS_ENABLED(CONFIG_IPV6) 152 if (sk->sk_family != tb2->family) { 153 if (sk->sk_family == AF_INET) 154 return ipv6_addr_v4mapped(&tb2->v6_rcv_saddr) && 155 tb2->v6_rcv_saddr.s6_addr32[3] == sk->sk_rcv_saddr; 156 157 return ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr) && 158 sk->sk_v6_rcv_saddr.s6_addr32[3] == tb2->rcv_saddr; 159 } 160 161 if (sk->sk_family == AF_INET6) 162 return ipv6_addr_equal(&tb2->v6_rcv_saddr, 163 &sk->sk_v6_rcv_saddr); 164#endif 165 return tb2->rcv_saddr == sk->sk_rcv_saddr; 166} 167 168void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb, 169 struct inet_bind2_bucket *tb2, unsigned short port) 170{ 171 inet_sk(sk)->inet_num = port; 172 sk_add_bind_node(sk, &tb->owners); 173 inet_csk(sk)->icsk_bind_hash = tb; 174 sk_add_bind2_node(sk, &tb2->owners); 175 inet_csk(sk)->icsk_bind2_hash = tb2; 176} 177 178/* 179 * Get rid of any references to a local port held by the given sock. 180 */ 181static void __inet_put_port(struct sock *sk) 182{ 183 struct inet_hashinfo *hashinfo = tcp_or_dccp_get_hashinfo(sk); 184 struct inet_bind_hashbucket *head, *head2; 185 struct net *net = sock_net(sk); 186 struct inet_bind_bucket *tb; 187 int bhash; 188 189 bhash = inet_bhashfn(net, inet_sk(sk)->inet_num, hashinfo->bhash_size); 190 head = &hashinfo->bhash[bhash]; 191 head2 = inet_bhashfn_portaddr(hashinfo, sk, net, inet_sk(sk)->inet_num); 192 193 spin_lock(&head->lock); 194 tb = inet_csk(sk)->icsk_bind_hash; 195 __sk_del_bind_node(sk); 196 inet_csk(sk)->icsk_bind_hash = NULL; 197 inet_sk(sk)->inet_num = 0; 198 inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb); 199 200 spin_lock(&head2->lock); 201 if (inet_csk(sk)->icsk_bind2_hash) { 202 struct inet_bind2_bucket *tb2 = inet_csk(sk)->icsk_bind2_hash; 203 204 __sk_del_bind2_node(sk); 205 inet_csk(sk)->icsk_bind2_hash = NULL; 206 inet_bind2_bucket_destroy(hashinfo->bind2_bucket_cachep, tb2); 207 } 208 spin_unlock(&head2->lock); 209 210 spin_unlock(&head->lock); 211} 212 213void inet_put_port(struct sock *sk) 214{ 215 local_bh_disable(); 216 __inet_put_port(sk); 217 local_bh_enable(); 218} 219EXPORT_SYMBOL(inet_put_port); 220 221int __inet_inherit_port(const struct sock *sk, struct sock *child) 222{ 223 struct inet_hashinfo *table = tcp_or_dccp_get_hashinfo(sk); 224 unsigned short port = inet_sk(child)->inet_num; 225 struct inet_bind_hashbucket *head, *head2; 226 bool created_inet_bind_bucket = false; 227 struct net *net = sock_net(sk); 228 bool update_fastreuse = false; 229 struct inet_bind2_bucket *tb2; 230 struct inet_bind_bucket *tb; 231 int bhash, l3mdev; 232 233 bhash = inet_bhashfn(net, port, table->bhash_size); 234 head = &table->bhash[bhash]; 235 head2 = inet_bhashfn_portaddr(table, child, net, port); 236 237 spin_lock(&head->lock); 238 spin_lock(&head2->lock); 239 tb = inet_csk(sk)->icsk_bind_hash; 240 tb2 = inet_csk(sk)->icsk_bind2_hash; 241 if (unlikely(!tb || !tb2)) { 242 spin_unlock(&head2->lock); 243 spin_unlock(&head->lock); 244 return -ENOENT; 245 } 246 if (tb->port != port) { 247 l3mdev = inet_sk_bound_l3mdev(sk); 248 249 /* NOTE: using tproxy and redirecting skbs to a proxy 250 * on a different listener port breaks the assumption 251 * that the listener socket's icsk_bind_hash is the same 252 * as that of the child socket. We have to look up or 253 * create a new bind bucket for the child here. */ 254 inet_bind_bucket_for_each(tb, &head->chain) { 255 if (inet_bind_bucket_match(tb, net, port, l3mdev)) 256 break; 257 } 258 if (!tb) { 259 tb = inet_bind_bucket_create(table->bind_bucket_cachep, 260 net, head, port, l3mdev); 261 if (!tb) { 262 spin_unlock(&head2->lock); 263 spin_unlock(&head->lock); 264 return -ENOMEM; 265 } 266 created_inet_bind_bucket = true; 267 } 268 update_fastreuse = true; 269 270 goto bhash2_find; 271 } else if (!inet_bind2_bucket_addr_match(tb2, child)) { 272 l3mdev = inet_sk_bound_l3mdev(sk); 273 274bhash2_find: 275 tb2 = inet_bind2_bucket_find(head2, net, port, l3mdev, child); 276 if (!tb2) { 277 tb2 = inet_bind2_bucket_create(table->bind2_bucket_cachep, 278 net, head2, port, 279 l3mdev, child); 280 if (!tb2) 281 goto error; 282 } 283 } 284 if (update_fastreuse) 285 inet_csk_update_fastreuse(tb, child); 286 inet_bind_hash(child, tb, tb2, port); 287 spin_unlock(&head2->lock); 288 spin_unlock(&head->lock); 289 290 return 0; 291 292error: 293 if (created_inet_bind_bucket) 294 inet_bind_bucket_destroy(table->bind_bucket_cachep, tb); 295 spin_unlock(&head2->lock); 296 spin_unlock(&head->lock); 297 return -ENOMEM; 298} 299EXPORT_SYMBOL_GPL(__inet_inherit_port); 300 301static struct inet_listen_hashbucket * 302inet_lhash2_bucket_sk(struct inet_hashinfo *h, struct sock *sk) 303{ 304 u32 hash; 305 306#if IS_ENABLED(CONFIG_IPV6) 307 if (sk->sk_family == AF_INET6) 308 hash = ipv6_portaddr_hash(sock_net(sk), 309 &sk->sk_v6_rcv_saddr, 310 inet_sk(sk)->inet_num); 311 else 312#endif 313 hash = ipv4_portaddr_hash(sock_net(sk), 314 inet_sk(sk)->inet_rcv_saddr, 315 inet_sk(sk)->inet_num); 316 return inet_lhash2_bucket(h, hash); 317} 318 319static inline int compute_score(struct sock *sk, struct net *net, 320 const unsigned short hnum, const __be32 daddr, 321 const int dif, const int sdif) 322{ 323 int score = -1; 324 325 if (net_eq(sock_net(sk), net) && sk->sk_num == hnum && 326 !ipv6_only_sock(sk)) { 327 if (sk->sk_rcv_saddr != daddr) 328 return -1; 329 330 if (!inet_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif)) 331 return -1; 332 score = sk->sk_bound_dev_if ? 2 : 1; 333 334 if (sk->sk_family == PF_INET) 335 score++; 336 if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id()) 337 score++; 338 } 339 return score; 340} 341 342/** 343 * inet_lookup_reuseport() - execute reuseport logic on AF_INET socket if necessary. 344 * @net: network namespace. 345 * @sk: AF_INET socket, must be in TCP_LISTEN state for TCP or TCP_CLOSE for UDP. 346 * @skb: context for a potential SK_REUSEPORT program. 347 * @doff: header offset. 348 * @saddr: source address. 349 * @sport: source port. 350 * @daddr: destination address. 351 * @hnum: destination port in host byte order. 352 * @ehashfn: hash function used to generate the fallback hash. 353 * 354 * Return: NULL if sk doesn't have SO_REUSEPORT set, otherwise a pointer to 355 * the selected sock or an error. 356 */ 357struct sock *inet_lookup_reuseport(struct net *net, struct sock *sk, 358 struct sk_buff *skb, int doff, 359 __be32 saddr, __be16 sport, 360 __be32 daddr, unsigned short hnum, 361 inet_ehashfn_t *ehashfn) 362{ 363 struct sock *reuse_sk = NULL; 364 u32 phash; 365 366 if (sk->sk_reuseport) { 367 phash = INDIRECT_CALL_2(ehashfn, udp_ehashfn, inet_ehashfn, 368 net, daddr, hnum, saddr, sport); 369 reuse_sk = reuseport_select_sock(sk, phash, skb, doff); 370 } 371 return reuse_sk; 372} 373EXPORT_SYMBOL_GPL(inet_lookup_reuseport); 374 375/* 376 * Here are some nice properties to exploit here. The BSD API 377 * does not allow a listening sock to specify the remote port nor the 378 * remote address for the connection. So always assume those are both 379 * wildcarded during the search since they can never be otherwise. 380 */ 381 382/* called with rcu_read_lock() : No refcount taken on the socket */ 383static struct sock *inet_lhash2_lookup(struct net *net, 384 struct inet_listen_hashbucket *ilb2, 385 struct sk_buff *skb, int doff, 386 const __be32 saddr, __be16 sport, 387 const __be32 daddr, const unsigned short hnum, 388 const int dif, const int sdif) 389{ 390 struct sock *sk, *result = NULL; 391 struct hlist_nulls_node *node; 392 int score, hiscore = 0; 393 394 sk_nulls_for_each_rcu(sk, node, &ilb2->nulls_head) { 395 score = compute_score(sk, net, hnum, daddr, dif, sdif); 396 if (score > hiscore) { 397 result = inet_lookup_reuseport(net, sk, skb, doff, 398 saddr, sport, daddr, hnum, inet_ehashfn); 399 if (result) 400 return result; 401 402 result = sk; 403 hiscore = score; 404 } 405 } 406 407 return result; 408} 409 410struct sock *inet_lookup_run_sk_lookup(struct net *net, 411 int protocol, 412 struct sk_buff *skb, int doff, 413 __be32 saddr, __be16 sport, 414 __be32 daddr, u16 hnum, const int dif, 415 inet_ehashfn_t *ehashfn) 416{ 417 struct sock *sk, *reuse_sk; 418 bool no_reuseport; 419 420 no_reuseport = bpf_sk_lookup_run_v4(net, protocol, saddr, sport, 421 daddr, hnum, dif, &sk); 422 if (no_reuseport || IS_ERR_OR_NULL(sk)) 423 return sk; 424 425 reuse_sk = inet_lookup_reuseport(net, sk, skb, doff, saddr, sport, daddr, hnum, 426 ehashfn); 427 if (reuse_sk) 428 sk = reuse_sk; 429 return sk; 430} 431 432struct sock *__inet_lookup_listener(struct net *net, 433 struct inet_hashinfo *hashinfo, 434 struct sk_buff *skb, int doff, 435 const __be32 saddr, __be16 sport, 436 const __be32 daddr, const unsigned short hnum, 437 const int dif, const int sdif) 438{ 439 struct inet_listen_hashbucket *ilb2; 440 struct sock *result = NULL; 441 unsigned int hash2; 442 443 /* Lookup redirect from BPF */ 444 if (static_branch_unlikely(&bpf_sk_lookup_enabled) && 445 hashinfo == net->ipv4.tcp_death_row.hashinfo) { 446 result = inet_lookup_run_sk_lookup(net, IPPROTO_TCP, skb, doff, 447 saddr, sport, daddr, hnum, dif, 448 inet_ehashfn); 449 if (result) 450 goto done; 451 } 452 453 hash2 = ipv4_portaddr_hash(net, daddr, hnum); 454 ilb2 = inet_lhash2_bucket(hashinfo, hash2); 455 456 result = inet_lhash2_lookup(net, ilb2, skb, doff, 457 saddr, sport, daddr, hnum, 458 dif, sdif); 459 if (result) 460 goto done; 461 462 /* Lookup lhash2 with INADDR_ANY */ 463 hash2 = ipv4_portaddr_hash(net, htonl(INADDR_ANY), hnum); 464 ilb2 = inet_lhash2_bucket(hashinfo, hash2); 465 466 result = inet_lhash2_lookup(net, ilb2, skb, doff, 467 saddr, sport, htonl(INADDR_ANY), hnum, 468 dif, sdif); 469done: 470 if (IS_ERR(result)) 471 return NULL; 472 return result; 473} 474EXPORT_SYMBOL_GPL(__inet_lookup_listener); 475 476/* All sockets share common refcount, but have different destructors */ 477void sock_gen_put(struct sock *sk) 478{ 479 if (!refcount_dec_and_test(&sk->sk_refcnt)) 480 return; 481 482 if (sk->sk_state == TCP_TIME_WAIT) 483 inet_twsk_free(inet_twsk(sk)); 484 else if (sk->sk_state == TCP_NEW_SYN_RECV) 485 reqsk_free(inet_reqsk(sk)); 486 else 487 sk_free(sk); 488} 489EXPORT_SYMBOL_GPL(sock_gen_put); 490 491void sock_edemux(struct sk_buff *skb) 492{ 493 sock_gen_put(skb->sk); 494} 495EXPORT_SYMBOL(sock_edemux); 496 497struct sock *__inet_lookup_established(struct net *net, 498 struct inet_hashinfo *hashinfo, 499 const __be32 saddr, const __be16 sport, 500 const __be32 daddr, const u16 hnum, 501 const int dif, const int sdif) 502{ 503 INET_ADDR_COOKIE(acookie, saddr, daddr); 504 const __portpair ports = INET_COMBINED_PORTS(sport, hnum); 505 struct sock *sk; 506 const struct hlist_nulls_node *node; 507 /* Optimize here for direct hit, only listening connections can 508 * have wildcards anyways. 509 */ 510 unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport); 511 unsigned int slot = hash & hashinfo->ehash_mask; 512 struct inet_ehash_bucket *head = &hashinfo->ehash[slot]; 513 514begin: 515 sk_nulls_for_each_rcu(sk, node, &head->chain) { 516 if (sk->sk_hash != hash) 517 continue; 518 if (likely(inet_match(net, sk, acookie, ports, dif, sdif))) { 519 if (unlikely(!refcount_inc_not_zero(&sk->sk_refcnt))) 520 goto out; 521 if (unlikely(!inet_match(net, sk, acookie, 522 ports, dif, sdif))) { 523 sock_gen_put(sk); 524 goto begin; 525 } 526 goto found; 527 } 528 } 529 /* 530 * if the nulls value we got at the end of this lookup is 531 * not the expected one, we must restart lookup. 532 * We probably met an item that was moved to another chain. 533 */ 534 if (get_nulls_value(node) != slot) 535 goto begin; 536out: 537 sk = NULL; 538found: 539 return sk; 540} 541EXPORT_SYMBOL_GPL(__inet_lookup_established); 542 543/* called with local bh disabled */ 544static int __inet_check_established(struct inet_timewait_death_row *death_row, 545 struct sock *sk, __u16 lport, 546 struct inet_timewait_sock **twp) 547{ 548 struct inet_hashinfo *hinfo = death_row->hashinfo; 549 struct inet_sock *inet = inet_sk(sk); 550 __be32 daddr = inet->inet_rcv_saddr; 551 __be32 saddr = inet->inet_daddr; 552 int dif = sk->sk_bound_dev_if; 553 struct net *net = sock_net(sk); 554 int sdif = l3mdev_master_ifindex_by_index(net, dif); 555 INET_ADDR_COOKIE(acookie, saddr, daddr); 556 const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport); 557 unsigned int hash = inet_ehashfn(net, daddr, lport, 558 saddr, inet->inet_dport); 559 struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash); 560 spinlock_t *lock = inet_ehash_lockp(hinfo, hash); 561 struct sock *sk2; 562 const struct hlist_nulls_node *node; 563 struct inet_timewait_sock *tw = NULL; 564 565 spin_lock(lock); 566 567 sk_nulls_for_each(sk2, node, &head->chain) { 568 if (sk2->sk_hash != hash) 569 continue; 570 571 if (likely(inet_match(net, sk2, acookie, ports, dif, sdif))) { 572 if (sk2->sk_state == TCP_TIME_WAIT) { 573 tw = inet_twsk(sk2); 574 if (twsk_unique(sk, sk2, twp)) 575 break; 576 } 577 goto not_unique; 578 } 579 } 580 581 /* Must record num and sport now. Otherwise we will see 582 * in hash table socket with a funny identity. 583 */ 584 inet->inet_num = lport; 585 inet->inet_sport = htons(lport); 586 sk->sk_hash = hash; 587 WARN_ON(!sk_unhashed(sk)); 588 __sk_nulls_add_node_rcu(sk, &head->chain); 589 if (tw) { 590 sk_nulls_del_node_init_rcu((struct sock *)tw); 591 __NET_INC_STATS(net, LINUX_MIB_TIMEWAITRECYCLED); 592 } 593 spin_unlock(lock); 594 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 595 596 if (twp) { 597 *twp = tw; 598 } else if (tw) { 599 /* Silly. Should hash-dance instead... */ 600 inet_twsk_deschedule_put(tw); 601 } 602 return 0; 603 604not_unique: 605 spin_unlock(lock); 606 return -EADDRNOTAVAIL; 607} 608 609static u64 inet_sk_port_offset(const struct sock *sk) 610{ 611 const struct inet_sock *inet = inet_sk(sk); 612 613 return secure_ipv4_port_ephemeral(inet->inet_rcv_saddr, 614 inet->inet_daddr, 615 inet->inet_dport); 616} 617 618/* Searches for an exsiting socket in the ehash bucket list. 619 * Returns true if found, false otherwise. 620 */ 621static bool inet_ehash_lookup_by_sk(struct sock *sk, 622 struct hlist_nulls_head *list) 623{ 624 const __portpair ports = INET_COMBINED_PORTS(sk->sk_dport, sk->sk_num); 625 const int sdif = sk->sk_bound_dev_if; 626 const int dif = sk->sk_bound_dev_if; 627 const struct hlist_nulls_node *node; 628 struct net *net = sock_net(sk); 629 struct sock *esk; 630 631 INET_ADDR_COOKIE(acookie, sk->sk_daddr, sk->sk_rcv_saddr); 632 633 sk_nulls_for_each_rcu(esk, node, list) { 634 if (esk->sk_hash != sk->sk_hash) 635 continue; 636 if (sk->sk_family == AF_INET) { 637 if (unlikely(inet_match(net, esk, acookie, 638 ports, dif, sdif))) { 639 return true; 640 } 641 } 642#if IS_ENABLED(CONFIG_IPV6) 643 else if (sk->sk_family == AF_INET6) { 644 if (unlikely(inet6_match(net, esk, 645 &sk->sk_v6_daddr, 646 &sk->sk_v6_rcv_saddr, 647 ports, dif, sdif))) { 648 return true; 649 } 650 } 651#endif 652 } 653 return false; 654} 655 656/* Insert a socket into ehash, and eventually remove another one 657 * (The another one can be a SYN_RECV or TIMEWAIT) 658 * If an existing socket already exists, socket sk is not inserted, 659 * and sets found_dup_sk parameter to true. 660 */ 661bool inet_ehash_insert(struct sock *sk, struct sock *osk, bool *found_dup_sk) 662{ 663 struct inet_hashinfo *hashinfo = tcp_or_dccp_get_hashinfo(sk); 664 struct inet_ehash_bucket *head; 665 struct hlist_nulls_head *list; 666 spinlock_t *lock; 667 bool ret = true; 668 669 WARN_ON_ONCE(!sk_unhashed(sk)); 670 671 sk->sk_hash = sk_ehashfn(sk); 672 head = inet_ehash_bucket(hashinfo, sk->sk_hash); 673 list = &head->chain; 674 lock = inet_ehash_lockp(hashinfo, sk->sk_hash); 675 676 spin_lock(lock); 677 if (osk) { 678 WARN_ON_ONCE(sk->sk_hash != osk->sk_hash); 679 ret = sk_nulls_del_node_init_rcu(osk); 680 } else if (found_dup_sk) { 681 *found_dup_sk = inet_ehash_lookup_by_sk(sk, list); 682 if (*found_dup_sk) 683 ret = false; 684 } 685 686 if (ret) 687 __sk_nulls_add_node_rcu(sk, list); 688 689 spin_unlock(lock); 690 691 return ret; 692} 693 694bool inet_ehash_nolisten(struct sock *sk, struct sock *osk, bool *found_dup_sk) 695{ 696 bool ok = inet_ehash_insert(sk, osk, found_dup_sk); 697 698 if (ok) { 699 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 700 } else { 701 this_cpu_inc(*sk->sk_prot->orphan_count); 702 inet_sk_set_state(sk, TCP_CLOSE); 703 sock_set_flag(sk, SOCK_DEAD); 704 inet_csk_destroy_sock(sk); 705 } 706 return ok; 707} 708EXPORT_SYMBOL_GPL(inet_ehash_nolisten); 709 710static int inet_reuseport_add_sock(struct sock *sk, 711 struct inet_listen_hashbucket *ilb) 712{ 713 struct inet_bind_bucket *tb = inet_csk(sk)->icsk_bind_hash; 714 const struct hlist_nulls_node *node; 715 struct sock *sk2; 716 kuid_t uid = sock_i_uid(sk); 717 718 sk_nulls_for_each_rcu(sk2, node, &ilb->nulls_head) { 719 if (sk2 != sk && 720 sk2->sk_family == sk->sk_family && 721 ipv6_only_sock(sk2) == ipv6_only_sock(sk) && 722 sk2->sk_bound_dev_if == sk->sk_bound_dev_if && 723 inet_csk(sk2)->icsk_bind_hash == tb && 724 sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) && 725 inet_rcv_saddr_equal(sk, sk2, false)) 726 return reuseport_add_sock(sk, sk2, 727 inet_rcv_saddr_any(sk)); 728 } 729 730 return reuseport_alloc(sk, inet_rcv_saddr_any(sk)); 731} 732 733int __inet_hash(struct sock *sk, struct sock *osk) 734{ 735 struct inet_hashinfo *hashinfo = tcp_or_dccp_get_hashinfo(sk); 736 struct inet_listen_hashbucket *ilb2; 737 int err = 0; 738 739 if (sk->sk_state != TCP_LISTEN) { 740 local_bh_disable(); 741 inet_ehash_nolisten(sk, osk, NULL); 742 local_bh_enable(); 743 return 0; 744 } 745 WARN_ON(!sk_unhashed(sk)); 746 ilb2 = inet_lhash2_bucket_sk(hashinfo, sk); 747 748 spin_lock(&ilb2->lock); 749 if (sk->sk_reuseport) { 750 err = inet_reuseport_add_sock(sk, ilb2); 751 if (err) 752 goto unlock; 753 } 754 sock_set_flag(sk, SOCK_RCU_FREE); 755 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport && 756 sk->sk_family == AF_INET6) 757 __sk_nulls_add_node_tail_rcu(sk, &ilb2->nulls_head); 758 else 759 __sk_nulls_add_node_rcu(sk, &ilb2->nulls_head); 760 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 761unlock: 762 spin_unlock(&ilb2->lock); 763 764 return err; 765} 766EXPORT_SYMBOL(__inet_hash); 767 768int inet_hash(struct sock *sk) 769{ 770 int err = 0; 771 772 if (sk->sk_state != TCP_CLOSE) 773 err = __inet_hash(sk, NULL); 774 775 return err; 776} 777EXPORT_SYMBOL_GPL(inet_hash); 778 779void inet_unhash(struct sock *sk) 780{ 781 struct inet_hashinfo *hashinfo = tcp_or_dccp_get_hashinfo(sk); 782 783 if (sk_unhashed(sk)) 784 return; 785 786 if (sk->sk_state == TCP_LISTEN) { 787 struct inet_listen_hashbucket *ilb2; 788 789 ilb2 = inet_lhash2_bucket_sk(hashinfo, sk); 790 /* Don't disable bottom halves while acquiring the lock to 791 * avoid circular locking dependency on PREEMPT_RT. 792 */ 793 spin_lock(&ilb2->lock); 794 if (sk_unhashed(sk)) { 795 spin_unlock(&ilb2->lock); 796 return; 797 } 798 799 if (rcu_access_pointer(sk->sk_reuseport_cb)) 800 reuseport_stop_listen_sock(sk); 801 802 __sk_nulls_del_node_init_rcu(sk); 803 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 804 spin_unlock(&ilb2->lock); 805 } else { 806 spinlock_t *lock = inet_ehash_lockp(hashinfo, sk->sk_hash); 807 808 spin_lock_bh(lock); 809 if (sk_unhashed(sk)) { 810 spin_unlock_bh(lock); 811 return; 812 } 813 __sk_nulls_del_node_init_rcu(sk); 814 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 815 spin_unlock_bh(lock); 816 } 817} 818EXPORT_SYMBOL_GPL(inet_unhash); 819 820static bool inet_bind2_bucket_match(const struct inet_bind2_bucket *tb, 821 const struct net *net, unsigned short port, 822 int l3mdev, const struct sock *sk) 823{ 824 if (!net_eq(ib2_net(tb), net) || tb->port != port || 825 tb->l3mdev != l3mdev) 826 return false; 827 828 return inet_bind2_bucket_addr_match(tb, sk); 829} 830 831bool inet_bind2_bucket_match_addr_any(const struct inet_bind2_bucket *tb, const struct net *net, 832 unsigned short port, int l3mdev, const struct sock *sk) 833{ 834 if (!net_eq(ib2_net(tb), net) || tb->port != port || 835 tb->l3mdev != l3mdev) 836 return false; 837 838#if IS_ENABLED(CONFIG_IPV6) 839 if (sk->sk_family != tb->family) { 840 if (sk->sk_family == AF_INET) 841 return ipv6_addr_any(&tb->v6_rcv_saddr) || 842 ipv6_addr_v4mapped_any(&tb->v6_rcv_saddr); 843 844 return false; 845 } 846 847 if (sk->sk_family == AF_INET6) 848 return ipv6_addr_any(&tb->v6_rcv_saddr); 849#endif 850 return tb->rcv_saddr == 0; 851} 852 853/* The socket's bhash2 hashbucket spinlock must be held when this is called */ 854struct inet_bind2_bucket * 855inet_bind2_bucket_find(const struct inet_bind_hashbucket *head, const struct net *net, 856 unsigned short port, int l3mdev, const struct sock *sk) 857{ 858 struct inet_bind2_bucket *bhash2 = NULL; 859 860 inet_bind_bucket_for_each(bhash2, &head->chain) 861 if (inet_bind2_bucket_match(bhash2, net, port, l3mdev, sk)) 862 break; 863 864 return bhash2; 865} 866 867struct inet_bind_hashbucket * 868inet_bhash2_addr_any_hashbucket(const struct sock *sk, const struct net *net, int port) 869{ 870 struct inet_hashinfo *hinfo = tcp_or_dccp_get_hashinfo(sk); 871 u32 hash; 872 873#if IS_ENABLED(CONFIG_IPV6) 874 if (sk->sk_family == AF_INET6) 875 hash = ipv6_portaddr_hash(net, &in6addr_any, port); 876 else 877#endif 878 hash = ipv4_portaddr_hash(net, 0, port); 879 880 return &hinfo->bhash2[hash & (hinfo->bhash_size - 1)]; 881} 882 883static void inet_update_saddr(struct sock *sk, void *saddr, int family) 884{ 885 if (family == AF_INET) { 886 inet_sk(sk)->inet_saddr = *(__be32 *)saddr; 887 sk_rcv_saddr_set(sk, inet_sk(sk)->inet_saddr); 888 } 889#if IS_ENABLED(CONFIG_IPV6) 890 else { 891 sk->sk_v6_rcv_saddr = *(struct in6_addr *)saddr; 892 } 893#endif 894} 895 896static int __inet_bhash2_update_saddr(struct sock *sk, void *saddr, int family, bool reset) 897{ 898 struct inet_hashinfo *hinfo = tcp_or_dccp_get_hashinfo(sk); 899 struct inet_bind_hashbucket *head, *head2; 900 struct inet_bind2_bucket *tb2, *new_tb2; 901 int l3mdev = inet_sk_bound_l3mdev(sk); 902 int port = inet_sk(sk)->inet_num; 903 struct net *net = sock_net(sk); 904 int bhash; 905 906 if (!inet_csk(sk)->icsk_bind2_hash) { 907 /* Not bind()ed before. */ 908 if (reset) 909 inet_reset_saddr(sk); 910 else 911 inet_update_saddr(sk, saddr, family); 912 913 return 0; 914 } 915 916 /* Allocate a bind2 bucket ahead of time to avoid permanently putting 917 * the bhash2 table in an inconsistent state if a new tb2 bucket 918 * allocation fails. 919 */ 920 new_tb2 = kmem_cache_alloc(hinfo->bind2_bucket_cachep, GFP_ATOMIC); 921 if (!new_tb2) { 922 if (reset) { 923 /* The (INADDR_ANY, port) bucket might have already 924 * been freed, then we cannot fixup icsk_bind2_hash, 925 * so we give up and unlink sk from bhash/bhash2 not 926 * to leave inconsistency in bhash2. 927 */ 928 inet_put_port(sk); 929 inet_reset_saddr(sk); 930 } 931 932 return -ENOMEM; 933 } 934 935 bhash = inet_bhashfn(net, port, hinfo->bhash_size); 936 head = &hinfo->bhash[bhash]; 937 head2 = inet_bhashfn_portaddr(hinfo, sk, net, port); 938 939 /* If we change saddr locklessly, another thread 940 * iterating over bhash might see corrupted address. 941 */ 942 spin_lock_bh(&head->lock); 943 944 spin_lock(&head2->lock); 945 __sk_del_bind2_node(sk); 946 inet_bind2_bucket_destroy(hinfo->bind2_bucket_cachep, inet_csk(sk)->icsk_bind2_hash); 947 spin_unlock(&head2->lock); 948 949 if (reset) 950 inet_reset_saddr(sk); 951 else 952 inet_update_saddr(sk, saddr, family); 953 954 head2 = inet_bhashfn_portaddr(hinfo, sk, net, port); 955 956 spin_lock(&head2->lock); 957 tb2 = inet_bind2_bucket_find(head2, net, port, l3mdev, sk); 958 if (!tb2) { 959 tb2 = new_tb2; 960 inet_bind2_bucket_init(tb2, net, head2, port, l3mdev, sk); 961 } 962 sk_add_bind2_node(sk, &tb2->owners); 963 inet_csk(sk)->icsk_bind2_hash = tb2; 964 spin_unlock(&head2->lock); 965 966 spin_unlock_bh(&head->lock); 967 968 if (tb2 != new_tb2) 969 kmem_cache_free(hinfo->bind2_bucket_cachep, new_tb2); 970 971 return 0; 972} 973 974int inet_bhash2_update_saddr(struct sock *sk, void *saddr, int family) 975{ 976 return __inet_bhash2_update_saddr(sk, saddr, family, false); 977} 978EXPORT_SYMBOL_GPL(inet_bhash2_update_saddr); 979 980void inet_bhash2_reset_saddr(struct sock *sk) 981{ 982 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) 983 __inet_bhash2_update_saddr(sk, NULL, 0, true); 984} 985EXPORT_SYMBOL_GPL(inet_bhash2_reset_saddr); 986 987/* RFC 6056 3.3.4. Algorithm 4: Double-Hash Port Selection Algorithm 988 * Note that we use 32bit integers (vs RFC 'short integers') 989 * because 2^16 is not a multiple of num_ephemeral and this 990 * property might be used by clever attacker. 991 * 992 * RFC claims using TABLE_LENGTH=10 buckets gives an improvement, though 993 * attacks were since demonstrated, thus we use 65536 by default instead 994 * to really give more isolation and privacy, at the expense of 256kB 995 * of kernel memory. 996 */ 997#define INET_TABLE_PERTURB_SIZE (1 << CONFIG_INET_TABLE_PERTURB_ORDER) 998static u32 *table_perturb; 999 1000int __inet_hash_connect(struct inet_timewait_death_row *death_row, 1001 struct sock *sk, u64 port_offset, 1002 int (*check_established)(struct inet_timewait_death_row *, 1003 struct sock *, __u16, struct inet_timewait_sock **)) 1004{ 1005 struct inet_hashinfo *hinfo = death_row->hashinfo; 1006 struct inet_bind_hashbucket *head, *head2; 1007 struct inet_timewait_sock *tw = NULL; 1008 int port = inet_sk(sk)->inet_num; 1009 struct net *net = sock_net(sk); 1010 struct inet_bind2_bucket *tb2; 1011 struct inet_bind_bucket *tb; 1012 bool tb_created = false; 1013 u32 remaining, offset; 1014 int ret, i, low, high; 1015 int l3mdev; 1016 u32 index; 1017 1018 if (port) { 1019 local_bh_disable(); 1020 ret = check_established(death_row, sk, port, NULL); 1021 local_bh_enable(); 1022 return ret; 1023 } 1024 1025 l3mdev = inet_sk_bound_l3mdev(sk); 1026 1027 inet_sk_get_local_port_range(sk, &low, &high); 1028 high++; /* [32768, 60999] -> [32768, 61000[ */ 1029 remaining = high - low; 1030 if (likely(remaining > 1)) 1031 remaining &= ~1U; 1032 1033 get_random_sleepable_once(table_perturb, 1034 INET_TABLE_PERTURB_SIZE * sizeof(*table_perturb)); 1035 index = port_offset & (INET_TABLE_PERTURB_SIZE - 1); 1036 1037 offset = READ_ONCE(table_perturb[index]) + (port_offset >> 32); 1038 offset %= remaining; 1039 1040 /* In first pass we try ports of @low parity. 1041 * inet_csk_get_port() does the opposite choice. 1042 */ 1043 offset &= ~1U; 1044other_parity_scan: 1045 port = low + offset; 1046 for (i = 0; i < remaining; i += 2, port += 2) { 1047 if (unlikely(port >= high)) 1048 port -= remaining; 1049 if (inet_is_local_reserved_port(net, port)) 1050 continue; 1051 head = &hinfo->bhash[inet_bhashfn(net, port, 1052 hinfo->bhash_size)]; 1053 spin_lock_bh(&head->lock); 1054 1055 /* Does not bother with rcv_saddr checks, because 1056 * the established check is already unique enough. 1057 */ 1058 inet_bind_bucket_for_each(tb, &head->chain) { 1059 if (inet_bind_bucket_match(tb, net, port, l3mdev)) { 1060 if (tb->fastreuse >= 0 || 1061 tb->fastreuseport >= 0) 1062 goto next_port; 1063 WARN_ON(hlist_empty(&tb->owners)); 1064 if (!check_established(death_row, sk, 1065 port, &tw)) 1066 goto ok; 1067 goto next_port; 1068 } 1069 } 1070 1071 tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep, 1072 net, head, port, l3mdev); 1073 if (!tb) { 1074 spin_unlock_bh(&head->lock); 1075 return -ENOMEM; 1076 } 1077 tb_created = true; 1078 tb->fastreuse = -1; 1079 tb->fastreuseport = -1; 1080 goto ok; 1081next_port: 1082 spin_unlock_bh(&head->lock); 1083 cond_resched(); 1084 } 1085 1086 offset++; 1087 if ((offset & 1) && remaining > 1) 1088 goto other_parity_scan; 1089 1090 return -EADDRNOTAVAIL; 1091 1092ok: 1093 /* Find the corresponding tb2 bucket since we need to 1094 * add the socket to the bhash2 table as well 1095 */ 1096 head2 = inet_bhashfn_portaddr(hinfo, sk, net, port); 1097 spin_lock(&head2->lock); 1098 1099 tb2 = inet_bind2_bucket_find(head2, net, port, l3mdev, sk); 1100 if (!tb2) { 1101 tb2 = inet_bind2_bucket_create(hinfo->bind2_bucket_cachep, net, 1102 head2, port, l3mdev, sk); 1103 if (!tb2) 1104 goto error; 1105 } 1106 1107 /* Here we want to add a little bit of randomness to the next source 1108 * port that will be chosen. We use a max() with a random here so that 1109 * on low contention the randomness is maximal and on high contention 1110 * it may be inexistent. 1111 */ 1112 i = max_t(int, i, get_random_u32_below(8) * 2); 1113 WRITE_ONCE(table_perturb[index], READ_ONCE(table_perturb[index]) + i + 2); 1114 1115 /* Head lock still held and bh's disabled */ 1116 inet_bind_hash(sk, tb, tb2, port); 1117 1118 if (sk_unhashed(sk)) { 1119 inet_sk(sk)->inet_sport = htons(port); 1120 inet_ehash_nolisten(sk, (struct sock *)tw, NULL); 1121 } 1122 if (tw) 1123 inet_twsk_bind_unhash(tw, hinfo); 1124 1125 spin_unlock(&head2->lock); 1126 spin_unlock(&head->lock); 1127 1128 if (tw) 1129 inet_twsk_deschedule_put(tw); 1130 local_bh_enable(); 1131 return 0; 1132 1133error: 1134 if (sk_hashed(sk)) { 1135 spinlock_t *lock = inet_ehash_lockp(hinfo, sk->sk_hash); 1136 1137 sock_prot_inuse_add(net, sk->sk_prot, -1); 1138 1139 spin_lock(lock); 1140 __sk_nulls_del_node_init_rcu(sk); 1141 spin_unlock(lock); 1142 1143 sk->sk_hash = 0; 1144 inet_sk(sk)->inet_sport = 0; 1145 inet_sk(sk)->inet_num = 0; 1146 1147 if (tw) 1148 inet_twsk_bind_unhash(tw, hinfo); 1149 } 1150 1151 spin_unlock(&head2->lock); 1152 if (tb_created) 1153 inet_bind_bucket_destroy(hinfo->bind_bucket_cachep, tb); 1154 spin_unlock(&head->lock); 1155 1156 if (tw) 1157 inet_twsk_deschedule_put(tw); 1158 1159 local_bh_enable(); 1160 1161 return -ENOMEM; 1162} 1163 1164/* 1165 * Bind a port for a connect operation and hash it. 1166 */ 1167int inet_hash_connect(struct inet_timewait_death_row *death_row, 1168 struct sock *sk) 1169{ 1170 u64 port_offset = 0; 1171 1172 if (!inet_sk(sk)->inet_num) 1173 port_offset = inet_sk_port_offset(sk); 1174 return __inet_hash_connect(death_row, sk, port_offset, 1175 __inet_check_established); 1176} 1177EXPORT_SYMBOL_GPL(inet_hash_connect); 1178 1179static void init_hashinfo_lhash2(struct inet_hashinfo *h) 1180{ 1181 int i; 1182 1183 for (i = 0; i <= h->lhash2_mask; i++) { 1184 spin_lock_init(&h->lhash2[i].lock); 1185 INIT_HLIST_NULLS_HEAD(&h->lhash2[i].nulls_head, 1186 i + LISTENING_NULLS_BASE); 1187 } 1188} 1189 1190void __init inet_hashinfo2_init(struct inet_hashinfo *h, const char *name, 1191 unsigned long numentries, int scale, 1192 unsigned long low_limit, 1193 unsigned long high_limit) 1194{ 1195 h->lhash2 = alloc_large_system_hash(name, 1196 sizeof(*h->lhash2), 1197 numentries, 1198 scale, 1199 0, 1200 NULL, 1201 &h->lhash2_mask, 1202 low_limit, 1203 high_limit); 1204 init_hashinfo_lhash2(h); 1205 1206 /* this one is used for source ports of outgoing connections */ 1207 table_perturb = alloc_large_system_hash("Table-perturb", 1208 sizeof(*table_perturb), 1209 INET_TABLE_PERTURB_SIZE, 1210 0, 0, NULL, NULL, 1211 INET_TABLE_PERTURB_SIZE, 1212 INET_TABLE_PERTURB_SIZE); 1213} 1214 1215int inet_hashinfo2_init_mod(struct inet_hashinfo *h) 1216{ 1217 h->lhash2 = kmalloc_array(INET_LHTABLE_SIZE, sizeof(*h->lhash2), GFP_KERNEL); 1218 if (!h->lhash2) 1219 return -ENOMEM; 1220 1221 h->lhash2_mask = INET_LHTABLE_SIZE - 1; 1222 /* INET_LHTABLE_SIZE must be a power of 2 */ 1223 BUG_ON(INET_LHTABLE_SIZE & h->lhash2_mask); 1224 1225 init_hashinfo_lhash2(h); 1226 return 0; 1227} 1228EXPORT_SYMBOL_GPL(inet_hashinfo2_init_mod); 1229 1230int inet_ehash_locks_alloc(struct inet_hashinfo *hashinfo) 1231{ 1232 unsigned int locksz = sizeof(spinlock_t); 1233 unsigned int i, nblocks = 1; 1234 1235 if (locksz != 0) { 1236 /* allocate 2 cache lines or at least one spinlock per cpu */ 1237 nblocks = max(2U * L1_CACHE_BYTES / locksz, 1U); 1238 nblocks = roundup_pow_of_two(nblocks * num_possible_cpus()); 1239 1240 /* no more locks than number of hash buckets */ 1241 nblocks = min(nblocks, hashinfo->ehash_mask + 1); 1242 1243 hashinfo->ehash_locks = kvmalloc_array(nblocks, locksz, GFP_KERNEL); 1244 if (!hashinfo->ehash_locks) 1245 return -ENOMEM; 1246 1247 for (i = 0; i < nblocks; i++) 1248 spin_lock_init(&hashinfo->ehash_locks[i]); 1249 } 1250 hashinfo->ehash_locks_mask = nblocks - 1; 1251 return 0; 1252} 1253EXPORT_SYMBOL_GPL(inet_ehash_locks_alloc); 1254 1255struct inet_hashinfo *inet_pernet_hashinfo_alloc(struct inet_hashinfo *hashinfo, 1256 unsigned int ehash_entries) 1257{ 1258 struct inet_hashinfo *new_hashinfo; 1259 int i; 1260 1261 new_hashinfo = kmemdup(hashinfo, sizeof(*hashinfo), GFP_KERNEL); 1262 if (!new_hashinfo) 1263 goto err; 1264 1265 new_hashinfo->ehash = vmalloc_huge(ehash_entries * sizeof(struct inet_ehash_bucket), 1266 GFP_KERNEL_ACCOUNT); 1267 if (!new_hashinfo->ehash) 1268 goto free_hashinfo; 1269 1270 new_hashinfo->ehash_mask = ehash_entries - 1; 1271 1272 if (inet_ehash_locks_alloc(new_hashinfo)) 1273 goto free_ehash; 1274 1275 for (i = 0; i < ehash_entries; i++) 1276 INIT_HLIST_NULLS_HEAD(&new_hashinfo->ehash[i].chain, i); 1277 1278 new_hashinfo->pernet = true; 1279 1280 return new_hashinfo; 1281 1282free_ehash: 1283 vfree(new_hashinfo->ehash); 1284free_hashinfo: 1285 kfree(new_hashinfo); 1286err: 1287 return NULL; 1288} 1289EXPORT_SYMBOL_GPL(inet_pernet_hashinfo_alloc); 1290 1291void inet_pernet_hashinfo_free(struct inet_hashinfo *hashinfo) 1292{ 1293 if (!hashinfo->pernet) 1294 return; 1295 1296 inet_ehash_locks_free(hashinfo); 1297 vfree(hashinfo->ehash); 1298 kfree(hashinfo); 1299} 1300EXPORT_SYMBOL_GPL(inet_pernet_hashinfo_free); 1301