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 INET6 transport hashtables 8 * 9 * Authors: Lotsa people, from code originally in tcp, generalised here 10 * by Arnaldo Carvalho de Melo <acme@mandriva.com> 11 */ 12 13 #include <linux/module.h> 14 #include <linux/random.h> 15 16 #include <net/addrconf.h> 17 #include <net/hotdata.h> 18 #include <net/inet_connection_sock.h> 19 #include <net/inet_hashtables.h> 20 #include <net/inet6_hashtables.h> 21 #include <net/secure_seq.h> 22 #include <net/ip.h> 23 #include <net/sock_reuseport.h> 24 25 u32 inet6_ehashfn(const struct net *net, 26 const struct in6_addr *laddr, const u16 lport, 27 const struct in6_addr *faddr, const __be16 fport) 28 { 29 u32 lhash, fhash; 30 31 net_get_random_once(&inet6_ehash_secret, sizeof(inet6_ehash_secret)); 32 net_get_random_once(&tcp_ipv6_hash_secret, sizeof(tcp_ipv6_hash_secret)); 33 34 lhash = (__force u32)laddr->s6_addr32[3]; 35 fhash = __ipv6_addr_jhash(faddr, tcp_ipv6_hash_secret); 36 37 return __inet6_ehashfn(lhash, lport, fhash, fport, 38 inet6_ehash_secret + net_hash_mix(net)); 39 } 40 EXPORT_SYMBOL_GPL(inet6_ehashfn); 41 42 /* 43 * Sockets in TCP_CLOSE state are _always_ taken out of the hash, so 44 * we need not check it for TCP lookups anymore, thanks Alexey. -DaveM 45 * 46 * The sockhash lock must be held as a reader here. 47 */ 48 struct sock *__inet6_lookup_established(struct net *net, 49 struct inet_hashinfo *hashinfo, 50 const struct in6_addr *saddr, 51 const __be16 sport, 52 const struct in6_addr *daddr, 53 const u16 hnum, 54 const int dif, const int sdif) 55 { 56 struct sock *sk; 57 const struct hlist_nulls_node *node; 58 const __portpair ports = INET_COMBINED_PORTS(sport, hnum); 59 /* Optimize here for direct hit, only listening connections can 60 * have wildcards anyways. 61 */ 62 unsigned int hash = inet6_ehashfn(net, daddr, hnum, saddr, sport); 63 unsigned int slot = hash & hashinfo->ehash_mask; 64 struct inet_ehash_bucket *head = &hashinfo->ehash[slot]; 65 66 67 begin: 68 sk_nulls_for_each_rcu(sk, node, &head->chain) { 69 if (sk->sk_hash != hash) 70 continue; 71 if (!inet6_match(net, sk, saddr, daddr, ports, dif, sdif)) 72 continue; 73 if (unlikely(!refcount_inc_not_zero(&sk->sk_refcnt))) 74 goto out; 75 76 if (unlikely(!inet6_match(net, sk, saddr, daddr, ports, dif, sdif))) { 77 sock_gen_put(sk); 78 goto begin; 79 } 80 goto found; 81 } 82 if (get_nulls_value(node) != slot) 83 goto begin; 84 out: 85 sk = NULL; 86 found: 87 return sk; 88 } 89 EXPORT_SYMBOL(__inet6_lookup_established); 90 91 static inline int compute_score(struct sock *sk, struct net *net, 92 const unsigned short hnum, 93 const struct in6_addr *daddr, 94 const int dif, const int sdif) 95 { 96 int score = -1; 97 98 if (net_eq(sock_net(sk), net) && inet_sk(sk)->inet_num == hnum && 99 sk->sk_family == PF_INET6) { 100 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr)) 101 return -1; 102 103 if (!inet_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif)) 104 return -1; 105 106 score = sk->sk_bound_dev_if ? 2 : 1; 107 if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id()) 108 score++; 109 } 110 return score; 111 } 112 113 /** 114 * inet6_lookup_reuseport() - execute reuseport logic on AF_INET6 socket if necessary. 115 * @net: network namespace. 116 * @sk: AF_INET6 socket, must be in TCP_LISTEN state for TCP or TCP_CLOSE for UDP. 117 * @skb: context for a potential SK_REUSEPORT program. 118 * @doff: header offset. 119 * @saddr: source address. 120 * @sport: source port. 121 * @daddr: destination address. 122 * @hnum: destination port in host byte order. 123 * @ehashfn: hash function used to generate the fallback hash. 124 * 125 * Return: NULL if sk doesn't have SO_REUSEPORT set, otherwise a pointer to 126 * the selected sock or an error. 127 */ 128 struct sock *inet6_lookup_reuseport(struct net *net, struct sock *sk, 129 struct sk_buff *skb, int doff, 130 const struct in6_addr *saddr, 131 __be16 sport, 132 const struct in6_addr *daddr, 133 unsigned short hnum, 134 inet6_ehashfn_t *ehashfn) 135 { 136 struct sock *reuse_sk = NULL; 137 u32 phash; 138 139 if (sk->sk_reuseport) { 140 phash = INDIRECT_CALL_INET(ehashfn, udp6_ehashfn, inet6_ehashfn, 141 net, daddr, hnum, saddr, sport); 142 reuse_sk = reuseport_select_sock(sk, phash, skb, doff); 143 } 144 return reuse_sk; 145 } 146 EXPORT_SYMBOL_GPL(inet6_lookup_reuseport); 147 148 /* called with rcu_read_lock() */ 149 static struct sock *inet6_lhash2_lookup(struct net *net, 150 struct inet_listen_hashbucket *ilb2, 151 struct sk_buff *skb, int doff, 152 const struct in6_addr *saddr, 153 const __be16 sport, const struct in6_addr *daddr, 154 const unsigned short hnum, const int dif, const int sdif) 155 { 156 struct sock *sk, *result = NULL; 157 struct hlist_nulls_node *node; 158 int score, hiscore = 0; 159 160 sk_nulls_for_each_rcu(sk, node, &ilb2->nulls_head) { 161 score = compute_score(sk, net, hnum, daddr, dif, sdif); 162 if (score > hiscore) { 163 result = inet6_lookup_reuseport(net, sk, skb, doff, 164 saddr, sport, daddr, hnum, inet6_ehashfn); 165 if (result) 166 return result; 167 168 result = sk; 169 hiscore = score; 170 } 171 } 172 173 return result; 174 } 175 176 struct sock *inet6_lookup_run_sk_lookup(struct net *net, 177 int protocol, 178 struct sk_buff *skb, int doff, 179 const struct in6_addr *saddr, 180 const __be16 sport, 181 const struct in6_addr *daddr, 182 const u16 hnum, const int dif, 183 inet6_ehashfn_t *ehashfn) 184 { 185 struct sock *sk, *reuse_sk; 186 bool no_reuseport; 187 188 no_reuseport = bpf_sk_lookup_run_v6(net, protocol, saddr, sport, 189 daddr, hnum, dif, &sk); 190 if (no_reuseport || IS_ERR_OR_NULL(sk)) 191 return sk; 192 193 reuse_sk = inet6_lookup_reuseport(net, sk, skb, doff, 194 saddr, sport, daddr, hnum, ehashfn); 195 if (reuse_sk) 196 sk = reuse_sk; 197 return sk; 198 } 199 EXPORT_SYMBOL_GPL(inet6_lookup_run_sk_lookup); 200 201 struct sock *inet6_lookup_listener(struct net *net, 202 struct inet_hashinfo *hashinfo, 203 struct sk_buff *skb, int doff, 204 const struct in6_addr *saddr, 205 const __be16 sport, const struct in6_addr *daddr, 206 const unsigned short hnum, const int dif, const int sdif) 207 { 208 struct inet_listen_hashbucket *ilb2; 209 struct sock *result = NULL; 210 unsigned int hash2; 211 212 /* Lookup redirect from BPF */ 213 if (static_branch_unlikely(&bpf_sk_lookup_enabled) && 214 hashinfo == net->ipv4.tcp_death_row.hashinfo) { 215 result = inet6_lookup_run_sk_lookup(net, IPPROTO_TCP, skb, doff, 216 saddr, sport, daddr, hnum, dif, 217 inet6_ehashfn); 218 if (result) 219 goto done; 220 } 221 222 hash2 = ipv6_portaddr_hash(net, daddr, hnum); 223 ilb2 = inet_lhash2_bucket(hashinfo, hash2); 224 225 result = inet6_lhash2_lookup(net, ilb2, skb, doff, 226 saddr, sport, daddr, hnum, 227 dif, sdif); 228 if (result) 229 goto done; 230 231 /* Lookup lhash2 with in6addr_any */ 232 hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum); 233 ilb2 = inet_lhash2_bucket(hashinfo, hash2); 234 235 result = inet6_lhash2_lookup(net, ilb2, skb, doff, 236 saddr, sport, &in6addr_any, hnum, 237 dif, sdif); 238 done: 239 if (IS_ERR(result)) 240 return NULL; 241 return result; 242 } 243 EXPORT_SYMBOL_GPL(inet6_lookup_listener); 244 245 struct sock *inet6_lookup(struct net *net, struct inet_hashinfo *hashinfo, 246 struct sk_buff *skb, int doff, 247 const struct in6_addr *saddr, const __be16 sport, 248 const struct in6_addr *daddr, const __be16 dport, 249 const int dif) 250 { 251 struct sock *sk; 252 bool refcounted; 253 254 sk = __inet6_lookup(net, hashinfo, skb, doff, saddr, sport, daddr, 255 ntohs(dport), dif, 0, &refcounted); 256 if (sk && !refcounted && !refcount_inc_not_zero(&sk->sk_refcnt)) 257 sk = NULL; 258 return sk; 259 } 260 EXPORT_SYMBOL_GPL(inet6_lookup); 261 262 static int __inet6_check_established(struct inet_timewait_death_row *death_row, 263 struct sock *sk, const __u16 lport, 264 struct inet_timewait_sock **twp) 265 { 266 struct inet_hashinfo *hinfo = death_row->hashinfo; 267 struct inet_sock *inet = inet_sk(sk); 268 const struct in6_addr *daddr = &sk->sk_v6_rcv_saddr; 269 const struct in6_addr *saddr = &sk->sk_v6_daddr; 270 const int dif = sk->sk_bound_dev_if; 271 struct net *net = sock_net(sk); 272 const int sdif = l3mdev_master_ifindex_by_index(net, dif); 273 const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport); 274 const unsigned int hash = inet6_ehashfn(net, daddr, lport, saddr, 275 inet->inet_dport); 276 struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash); 277 spinlock_t *lock = inet_ehash_lockp(hinfo, hash); 278 struct sock *sk2; 279 const struct hlist_nulls_node *node; 280 struct inet_timewait_sock *tw = NULL; 281 282 spin_lock(lock); 283 284 sk_nulls_for_each(sk2, node, &head->chain) { 285 if (sk2->sk_hash != hash) 286 continue; 287 288 if (likely(inet6_match(net, sk2, saddr, daddr, ports, 289 dif, sdif))) { 290 if (sk2->sk_state == TCP_TIME_WAIT) { 291 tw = inet_twsk(sk2); 292 if (twsk_unique(sk, sk2, twp)) 293 break; 294 } 295 goto not_unique; 296 } 297 } 298 299 /* Must record num and sport now. Otherwise we will see 300 * in hash table socket with a funny identity. 301 */ 302 inet->inet_num = lport; 303 inet->inet_sport = htons(lport); 304 sk->sk_hash = hash; 305 WARN_ON(!sk_unhashed(sk)); 306 __sk_nulls_add_node_rcu(sk, &head->chain); 307 if (tw) { 308 sk_nulls_del_node_init_rcu((struct sock *)tw); 309 __NET_INC_STATS(net, LINUX_MIB_TIMEWAITRECYCLED); 310 } 311 spin_unlock(lock); 312 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); 313 314 if (twp) { 315 *twp = tw; 316 } else if (tw) { 317 /* Silly. Should hash-dance instead... */ 318 inet_twsk_deschedule_put(tw); 319 } 320 return 0; 321 322 not_unique: 323 spin_unlock(lock); 324 return -EADDRNOTAVAIL; 325 } 326 327 static u64 inet6_sk_port_offset(const struct sock *sk) 328 { 329 const struct inet_sock *inet = inet_sk(sk); 330 331 return secure_ipv6_port_ephemeral(sk->sk_v6_rcv_saddr.s6_addr32, 332 sk->sk_v6_daddr.s6_addr32, 333 inet->inet_dport); 334 } 335 336 int inet6_hash_connect(struct inet_timewait_death_row *death_row, 337 struct sock *sk) 338 { 339 u64 port_offset = 0; 340 341 if (!inet_sk(sk)->inet_num) 342 port_offset = inet6_sk_port_offset(sk); 343 return __inet_hash_connect(death_row, sk, port_offset, 344 __inet6_check_established); 345 } 346 EXPORT_SYMBOL_GPL(inet6_hash_connect); 347 348 int inet6_hash(struct sock *sk) 349 { 350 int err = 0; 351 352 if (sk->sk_state != TCP_CLOSE) 353 err = __inet_hash(sk, NULL); 354 355 return err; 356 } 357 EXPORT_SYMBOL_GPL(inet6_hash); 358