Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 1 | /* |
| 2 | * INET An implementation of the TCP/IP protocol suite for the LINUX |
| 3 | * operating system. INET is implemented using the BSD Socket |
| 4 | * interface as the means of communication with the user level. |
| 5 | * |
| 6 | * Authors: Lotsa people, from code originally in tcp |
| 7 | * |
| 8 | * This program is free software; you can redistribute it and/or |
| 9 | * modify it under the terms of the GNU General Public License |
| 10 | * as published by the Free Software Foundation; either version |
| 11 | * 2 of the License, or (at your option) any later version. |
| 12 | */ |
| 13 | |
| 14 | #ifndef _INET_HASHTABLES_H |
| 15 | #define _INET_HASHTABLES_H |
| 16 | |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 17 | |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 18 | #include <linux/interrupt.h> |
Arnaldo Carvalho de Melo | 9a1f27c | 2008-10-07 11:41:57 -0700 | [diff] [blame] | 19 | #include <linux/ip.h> |
Arnaldo Carvalho de Melo | 33b6223 | 2005-08-09 20:09:06 -0700 | [diff] [blame] | 20 | #include <linux/ipv6.h> |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 21 | #include <linux/list.h> |
| 22 | #include <linux/slab.h> |
Arnaldo Carvalho de Melo | 33b6223 | 2005-08-09 20:09:06 -0700 | [diff] [blame] | 23 | #include <linux/socket.h> |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 24 | #include <linux/spinlock.h> |
Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 25 | #include <linux/types.h> |
Arnaldo Carvalho de Melo | f3f05f7 | 2005-08-09 20:08:09 -0700 | [diff] [blame] | 26 | #include <linux/wait.h> |
Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 27 | |
Arnaldo Carvalho de Melo | 463c84b | 2005-08-09 20:10:42 -0700 | [diff] [blame] | 28 | #include <net/inet_connection_sock.h> |
Arnaldo Carvalho de Melo | 14c8502 | 2005-12-27 02:43:12 -0200 | [diff] [blame] | 29 | #include <net/inet_sock.h> |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 30 | #include <net/sock.h> |
Arnaldo Carvalho de Melo | 9a1f27c | 2008-10-07 11:41:57 -0700 | [diff] [blame] | 31 | #include <net/route.h> |
Arnaldo Carvalho de Melo | c752f07 | 2005-08-09 20:08:28 -0700 | [diff] [blame] | 32 | #include <net/tcp_states.h> |
Pavel Emelyanov | 0b44191 | 2008-06-16 17:14:11 -0700 | [diff] [blame] | 33 | #include <net/netns/hash.h> |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 34 | |
Arun Sharma | 60063497 | 2011-07-26 16:09:06 -0700 | [diff] [blame] | 35 | #include <linux/atomic.h> |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 36 | #include <asm/byteorder.h> |
Arnaldo Carvalho de Melo | f3f05f7 | 2005-08-09 20:08:09 -0700 | [diff] [blame] | 37 | |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 38 | /* This is for all connections with a full identity, no wildcards. |
Eric Dumazet | 05dbc7b | 2013-10-03 00:22:02 -0700 | [diff] [blame] | 39 | * The 'e' prefix stands for Establish, but we really put all sockets |
| 40 | * but LISTEN ones. |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 41 | */ |
| 42 | struct inet_ehash_bucket { |
Eric Dumazet | 3ab5aee | 2008-11-16 19:40:17 -0800 | [diff] [blame] | 43 | struct hlist_nulls_head chain; |
Eric Dumazet | 6d25536 | 2005-10-04 15:55:51 -0700 | [diff] [blame] | 44 | }; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 45 | |
| 46 | /* There are a few simple rules, which allow for local port reuse by |
| 47 | * an application. In essence: |
| 48 | * |
| 49 | * 1) Sockets bound to different interfaces may share a local port. |
| 50 | * Failing that, goto test 2. |
| 51 | * 2) If all sockets have sk->sk_reuse set, and none of them are in |
| 52 | * TCP_LISTEN state, the port may be shared. |
| 53 | * Failing that, goto test 3. |
| 54 | * 3) If all sockets are bound to a specific inet_sk(sk)->rcv_saddr local |
| 55 | * address, and none of them are the same, the port may be |
| 56 | * shared. |
| 57 | * Failing this, the port cannot be shared. |
| 58 | * |
| 59 | * The interesting point, is test #2. This is what an FTP server does |
| 60 | * all day. To optimize this case we use a specific flag bit defined |
| 61 | * below. As we add sockets to a bind bucket list, we perform a |
| 62 | * check of: (newsk->sk_reuse && (newsk->sk_state != TCP_LISTEN)) |
| 63 | * As long as all sockets added to a bind bucket pass this test, |
| 64 | * the flag bit will be set. |
| 65 | * The resulting situation is that tcp_v[46]_verify_bind() can just check |
| 66 | * for this flag bit, if it is set and the socket trying to bind has |
| 67 | * sk->sk_reuse set, we don't even have to walk the owners list at all, |
| 68 | * we return that it is ok to bind this socket to the requested local port. |
| 69 | * |
| 70 | * Sounds like a lot of work, but it is worth it. In a more naive |
| 71 | * implementation (ie. current FreeBSD etc.) the entire list of ports |
| 72 | * must be walked for each data port opened by an ftp server. Needless |
| 73 | * to say, this does not scale at all. With a couple thousand FTP |
| 74 | * users logged onto your box, isn't it nice to know that new data |
| 75 | * ports are created in O(1) time? I thought so. ;-) -DaveM |
| 76 | */ |
Josef Bacik | 637bc8b | 2017-01-17 07:51:06 -0800 | [diff] [blame] | 77 | #define FASTREUSEPORT_ANY 1 |
| 78 | #define FASTREUSEPORT_STRICT 2 |
| 79 | |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 80 | struct inet_bind_bucket { |
Eric W. Biederman | 0c5c9fb | 2015-03-11 23:06:44 -0500 | [diff] [blame] | 81 | possible_net_t ib_net; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 82 | unsigned short port; |
Tom Herbert | da5e363 | 2013-01-22 09:50:24 +0000 | [diff] [blame] | 83 | signed char fastreuse; |
| 84 | signed char fastreuseport; |
| 85 | kuid_t fastuid; |
Josef Bacik | 637bc8b | 2017-01-17 07:51:06 -0800 | [diff] [blame] | 86 | #if IS_ENABLED(CONFIG_IPV6) |
| 87 | struct in6_addr fast_v6_rcv_saddr; |
| 88 | #endif |
| 89 | __be32 fast_rcv_saddr; |
| 90 | unsigned short fast_sk_family; |
| 91 | bool fast_ipv6_only; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 92 | struct hlist_node node; |
| 93 | struct hlist_head owners; |
| 94 | }; |
| 95 | |
Eric Dumazet | 7a9546e | 2008-11-12 00:54:20 -0800 | [diff] [blame] | 96 | static inline struct net *ib_net(struct inet_bind_bucket *ib) |
| 97 | { |
| 98 | return read_pnet(&ib->ib_net); |
| 99 | } |
| 100 | |
Sasha Levin | b67bfe0 | 2013-02-27 17:06:00 -0800 | [diff] [blame] | 101 | #define inet_bind_bucket_for_each(tb, head) \ |
| 102 | hlist_for_each_entry(tb, head, node) |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 103 | |
| 104 | struct inet_bind_hashbucket { |
| 105 | spinlock_t lock; |
| 106 | struct hlist_head chain; |
| 107 | }; |
| 108 | |
Eric Dumazet | c25eb3b | 2008-11-23 17:22:55 -0800 | [diff] [blame] | 109 | /* |
| 110 | * Sockets can be hashed in established or listening table |
Eric Dumazet | c25eb3b | 2008-11-23 17:22:55 -0800 | [diff] [blame] | 111 | */ |
Eric Dumazet | 5caea4e | 2008-11-20 00:40:07 -0800 | [diff] [blame] | 112 | struct inet_listen_hashbucket { |
| 113 | spinlock_t lock; |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 114 | struct hlist_head head; |
Eric Dumazet | 5caea4e | 2008-11-20 00:40:07 -0800 | [diff] [blame] | 115 | }; |
| 116 | |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 117 | /* This is for listening sockets, thus all sockets which possess wildcards. */ |
| 118 | #define INET_LHTABLE_SIZE 32 /* Yes, really, this is all you need. */ |
| 119 | |
| 120 | struct inet_hashinfo { |
| 121 | /* This is for sockets with full identity only. Sockets here will |
| 122 | * always be without wildcards and will have the following invariant: |
| 123 | * |
| 124 | * TCP_ESTABLISHED <= sk->sk_state < TCP_CLOSE |
| 125 | * |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 126 | */ |
| 127 | struct inet_ehash_bucket *ehash; |
Eric Dumazet | 9db66bd | 2008-11-20 20:39:09 -0800 | [diff] [blame] | 128 | spinlock_t *ehash_locks; |
Eric Dumazet | f373b53 | 2009-10-09 00:16:19 +0000 | [diff] [blame] | 129 | unsigned int ehash_mask; |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 130 | unsigned int ehash_locks_mask; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 131 | |
| 132 | /* Ok, let's try this, I give up, we do need a local binding |
| 133 | * TCP hash as well as the others for fast bind/connect. |
| 134 | */ |
| 135 | struct inet_bind_hashbucket *bhash; |
| 136 | |
Stephen Hemminger | cfcabdc | 2007-10-09 01:59:42 -0700 | [diff] [blame] | 137 | unsigned int bhash_size; |
Eric Dumazet | 24dd1fa | 2009-02-01 12:31:33 -0800 | [diff] [blame] | 138 | /* 4 bytes hole on 64 bit */ |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 139 | |
Eric Dumazet | 5caea4e | 2008-11-20 00:40:07 -0800 | [diff] [blame] | 140 | struct kmem_cache *bind_bucket_cachep; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 141 | |
| 142 | /* All the above members are written once at bootup and |
| 143 | * never written again _or_ are predominantly read-access. |
| 144 | * |
| 145 | * Now align to a new cache line as all the following members |
Eric Dumazet | 5caea4e | 2008-11-20 00:40:07 -0800 | [diff] [blame] | 146 | * might be often dirty. |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 147 | */ |
Eric Dumazet | 5caea4e | 2008-11-20 00:40:07 -0800 | [diff] [blame] | 148 | /* All sockets in TCP_LISTEN state will be in here. This is the only |
| 149 | * table where wildcard'd TCP sockets can exist. Hash function here |
| 150 | * is just local port number. |
| 151 | */ |
| 152 | struct inet_listen_hashbucket listening_hash[INET_LHTABLE_SIZE] |
| 153 | ____cacheline_aligned_in_smp; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 154 | }; |
| 155 | |
Eric Dumazet | 81c3d54 | 2005-10-03 14:13:38 -0700 | [diff] [blame] | 156 | static inline struct inet_ehash_bucket *inet_ehash_bucket( |
| 157 | struct inet_hashinfo *hashinfo, |
| 158 | unsigned int hash) |
| 159 | { |
Eric Dumazet | f373b53 | 2009-10-09 00:16:19 +0000 | [diff] [blame] | 160 | return &hashinfo->ehash[hash & hashinfo->ehash_mask]; |
Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 161 | } |
| 162 | |
Eric Dumazet | 9db66bd | 2008-11-20 20:39:09 -0800 | [diff] [blame] | 163 | static inline spinlock_t *inet_ehash_lockp( |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 164 | struct inet_hashinfo *hashinfo, |
| 165 | unsigned int hash) |
| 166 | { |
| 167 | return &hashinfo->ehash_locks[hash & hashinfo->ehash_locks_mask]; |
| 168 | } |
| 169 | |
Eric Dumazet | 095dc8e | 2015-05-26 07:55:34 -0700 | [diff] [blame] | 170 | int inet_ehash_locks_alloc(struct inet_hashinfo *hashinfo); |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 171 | |
| 172 | static inline void inet_ehash_locks_free(struct inet_hashinfo *hashinfo) |
| 173 | { |
Eric Dumazet | 095dc8e | 2015-05-26 07:55:34 -0700 | [diff] [blame] | 174 | kvfree(hashinfo->ehash_locks); |
| 175 | hashinfo->ehash_locks = NULL; |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 176 | } |
| 177 | |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 178 | struct inet_bind_bucket * |
| 179 | inet_bind_bucket_create(struct kmem_cache *cachep, struct net *net, |
| 180 | struct inet_bind_hashbucket *head, |
| 181 | const unsigned short snum); |
| 182 | void inet_bind_bucket_destroy(struct kmem_cache *cachep, |
| 183 | struct inet_bind_bucket *tb); |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 184 | |
Eric Dumazet | 6eada01 | 2015-03-18 14:05:33 -0700 | [diff] [blame] | 185 | static inline u32 inet_bhashfn(const struct net *net, const __u16 lport, |
| 186 | const u32 bhash_size) |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 187 | { |
Pavel Emelyanov | 0b44191 | 2008-06-16 17:14:11 -0700 | [diff] [blame] | 188 | return (lport + net_hash_mix(net)) & (bhash_size - 1); |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 189 | } |
| 190 | |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 191 | void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb, |
| 192 | const unsigned short snum); |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 193 | |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 194 | /* These can have wildcards, don't try too hard. */ |
Eric Dumazet | 6eada01 | 2015-03-18 14:05:33 -0700 | [diff] [blame] | 195 | static inline u32 inet_lhashfn(const struct net *net, const unsigned short num) |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 196 | { |
Pavel Emelyanov | 0b44191 | 2008-06-16 17:14:11 -0700 | [diff] [blame] | 197 | return (num + net_hash_mix(net)) & (INET_LHTABLE_SIZE - 1); |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 198 | } |
| 199 | |
| 200 | static inline int inet_sk_listen_hashfn(const struct sock *sk) |
| 201 | { |
Eric Dumazet | c720c7e8 | 2009-10-15 06:30:45 +0000 | [diff] [blame] | 202 | return inet_lhashfn(sock_net(sk), inet_sk(sk)->inet_num); |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 203 | } |
| 204 | |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 205 | /* Caller must disable local BH processing. */ |
Eric Dumazet | 1ce31c9 | 2015-09-29 07:42:44 -0700 | [diff] [blame] | 206 | int __inet_inherit_port(const struct sock *sk, struct sock *child); |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 207 | |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 208 | void inet_put_port(struct sock *sk); |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 209 | |
Eric Dumazet | 5caea4e | 2008-11-20 00:40:07 -0800 | [diff] [blame] | 210 | void inet_hashinfo_init(struct inet_hashinfo *h); |
Arnaldo Carvalho de Melo | f3f05f7 | 2005-08-09 20:08:09 -0700 | [diff] [blame] | 211 | |
Eric Dumazet | 5e0724d | 2015-10-22 08:20:46 -0700 | [diff] [blame] | 212 | bool inet_ehash_insert(struct sock *sk, struct sock *osk); |
| 213 | bool inet_ehash_nolisten(struct sock *sk, struct sock *osk); |
Josef Bacik | fe38d2a | 2017-01-17 07:51:01 -0800 | [diff] [blame] | 214 | int __inet_hash(struct sock *sk, struct sock *osk); |
Craig Gallek | 086c653 | 2016-02-10 11:50:35 -0500 | [diff] [blame] | 215 | int inet_hash(struct sock *sk); |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 216 | void inet_unhash(struct sock *sk); |
Arnaldo Carvalho de Melo | 33b6223 | 2005-08-09 20:09:06 -0700 | [diff] [blame] | 217 | |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 218 | struct sock *__inet_lookup_listener(struct net *net, |
| 219 | struct inet_hashinfo *hashinfo, |
Craig Gallek | a583636 | 2016-02-10 11:50:38 -0500 | [diff] [blame] | 220 | struct sk_buff *skb, int doff, |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 221 | const __be32 saddr, const __be16 sport, |
| 222 | const __be32 daddr, |
| 223 | const unsigned short hnum, |
| 224 | const int dif); |
Herbert Xu | 8f491069 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 225 | |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 226 | static inline struct sock *inet_lookup_listener(struct net *net, |
| 227 | struct inet_hashinfo *hashinfo, |
Craig Gallek | a583636 | 2016-02-10 11:50:38 -0500 | [diff] [blame] | 228 | struct sk_buff *skb, int doff, |
Tom Herbert | da5e363 | 2013-01-22 09:50:24 +0000 | [diff] [blame] | 229 | __be32 saddr, __be16 sport, |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 230 | __be32 daddr, __be16 dport, int dif) |
Herbert Xu | 8f491069 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 231 | { |
Craig Gallek | a583636 | 2016-02-10 11:50:38 -0500 | [diff] [blame] | 232 | return __inet_lookup_listener(net, hashinfo, skb, doff, saddr, sport, |
Tom Herbert | da5e363 | 2013-01-22 09:50:24 +0000 | [diff] [blame] | 233 | daddr, ntohs(dport), dif); |
Herbert Xu | 8f491069 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 234 | } |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 235 | |
| 236 | /* Socket demux engine toys. */ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 237 | /* What happens here is ugly; there's a pair of adjacent fields in |
| 238 | struct inet_sock; __be16 dport followed by __u16 num. We want to |
| 239 | search by pair, so we combine the keys into a single 32bit value |
| 240 | and compare with 32bit value read from &...->dport. Let's at least |
| 241 | make sure that it's not mixed with anything else... |
| 242 | On 64bit targets we combine comparisons with pair of adjacent __be32 |
| 243 | fields in the same way. |
| 244 | */ |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 245 | #ifdef __BIG_ENDIAN |
| 246 | #define INET_COMBINED_PORTS(__sport, __dport) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 247 | ((__force __portpair)(((__force __u32)(__be16)(__sport) << 16) | (__u32)(__dport))) |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 248 | #else /* __LITTLE_ENDIAN */ |
| 249 | #define INET_COMBINED_PORTS(__sport, __dport) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 250 | ((__force __portpair)(((__u32)(__dport) << 16) | (__force __u32)(__be16)(__sport))) |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 251 | #endif |
| 252 | |
| 253 | #if (BITS_PER_LONG == 64) |
| 254 | #ifdef __BIG_ENDIAN |
| 255 | #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 256 | const __addrpair __name = (__force __addrpair) ( \ |
| 257 | (((__force __u64)(__be32)(__saddr)) << 32) | \ |
Joe Perches | c722831 | 2014-05-13 20:30:07 -0700 | [diff] [blame] | 258 | ((__force __u64)(__be32)(__daddr))) |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 259 | #else /* __LITTLE_ENDIAN */ |
| 260 | #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 261 | const __addrpair __name = (__force __addrpair) ( \ |
| 262 | (((__force __u64)(__be32)(__daddr)) << 32) | \ |
Joe Perches | c722831 | 2014-05-13 20:30:07 -0700 | [diff] [blame] | 263 | ((__force __u64)(__be32)(__saddr))) |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 264 | #endif /* __BIG_ENDIAN */ |
Eric Dumazet | ce43b03 | 2012-11-30 09:49:27 +0000 | [diff] [blame] | 265 | #define INET_MATCH(__sk, __net, __cookie, __saddr, __daddr, __ports, __dif) \ |
Eric Dumazet | 5080546 | 2013-10-02 04:29:50 -0700 | [diff] [blame] | 266 | (((__sk)->sk_portpair == (__ports)) && \ |
| 267 | ((__sk)->sk_addrpair == (__cookie)) && \ |
Eric Dumazet | ce43b03 | 2012-11-30 09:49:27 +0000 | [diff] [blame] | 268 | (!(__sk)->sk_bound_dev_if || \ |
| 269 | ((__sk)->sk_bound_dev_if == (__dif))) && \ |
| 270 | net_eq(sock_net(__sk), (__net))) |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 271 | #else /* 32-bit arch */ |
Joe Perches | c722831 | 2014-05-13 20:30:07 -0700 | [diff] [blame] | 272 | #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \ |
| 273 | const int __name __deprecated __attribute__((unused)) |
| 274 | |
Eric Dumazet | ce43b03 | 2012-11-30 09:49:27 +0000 | [diff] [blame] | 275 | #define INET_MATCH(__sk, __net, __cookie, __saddr, __daddr, __ports, __dif) \ |
Eric Dumazet | 5080546 | 2013-10-02 04:29:50 -0700 | [diff] [blame] | 276 | (((__sk)->sk_portpair == (__ports)) && \ |
| 277 | ((__sk)->sk_daddr == (__saddr)) && \ |
| 278 | ((__sk)->sk_rcv_saddr == (__daddr)) && \ |
Eric Dumazet | ce43b03 | 2012-11-30 09:49:27 +0000 | [diff] [blame] | 279 | (!(__sk)->sk_bound_dev_if || \ |
| 280 | ((__sk)->sk_bound_dev_if == (__dif))) && \ |
| 281 | net_eq(sock_net(__sk), (__net))) |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 282 | #endif /* 64-bit arch */ |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 283 | |
Eric Dumazet | ee3cf32 | 2016-04-01 08:52:14 -0700 | [diff] [blame] | 284 | /* Sockets in TCP_CLOSE state are _always_ taken out of the hash, so we need |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 285 | * not check it for lookups anymore, thanks Alexey. -DaveM |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 286 | */ |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 287 | struct sock *__inet_lookup_established(struct net *net, |
| 288 | struct inet_hashinfo *hashinfo, |
| 289 | const __be32 saddr, const __be16 sport, |
| 290 | const __be32 daddr, const u16 hnum, |
| 291 | const int dif); |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 292 | |
Herbert Xu | 8f491069 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 293 | static inline struct sock * |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 294 | inet_lookup_established(struct net *net, struct inet_hashinfo *hashinfo, |
Al Viro | fb99c84 | 2006-09-27 18:43:33 -0700 | [diff] [blame] | 295 | const __be32 saddr, const __be16 sport, |
| 296 | const __be32 daddr, const __be16 dport, |
Herbert Xu | 8f491069 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 297 | const int dif) |
| 298 | { |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 299 | return __inet_lookup_established(net, hashinfo, saddr, sport, daddr, |
Herbert Xu | 8f491069 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 300 | ntohs(dport), dif); |
| 301 | } |
| 302 | |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 303 | static inline struct sock *__inet_lookup(struct net *net, |
| 304 | struct inet_hashinfo *hashinfo, |
Craig Gallek | a583636 | 2016-02-10 11:50:38 -0500 | [diff] [blame] | 305 | struct sk_buff *skb, int doff, |
Al Viro | fb99c84 | 2006-09-27 18:43:33 -0700 | [diff] [blame] | 306 | const __be32 saddr, const __be16 sport, |
| 307 | const __be32 daddr, const __be16 dport, |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 308 | const int dif, |
| 309 | bool *refcounted) |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 310 | { |
Herbert Xu | 8f491069 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 311 | u16 hnum = ntohs(dport); |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 312 | struct sock *sk; |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 313 | |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 314 | sk = __inet_lookup_established(net, hashinfo, saddr, sport, |
| 315 | daddr, hnum, dif); |
| 316 | *refcounted = true; |
| 317 | if (sk) |
| 318 | return sk; |
| 319 | *refcounted = false; |
| 320 | return __inet_lookup_listener(net, hashinfo, skb, doff, saddr, |
| 321 | sport, daddr, hnum, dif); |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 322 | } |
| 323 | |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 324 | static inline struct sock *inet_lookup(struct net *net, |
| 325 | struct inet_hashinfo *hashinfo, |
Craig Gallek | a583636 | 2016-02-10 11:50:38 -0500 | [diff] [blame] | 326 | struct sk_buff *skb, int doff, |
Al Viro | fb99c84 | 2006-09-27 18:43:33 -0700 | [diff] [blame] | 327 | const __be32 saddr, const __be16 sport, |
| 328 | const __be32 daddr, const __be16 dport, |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 329 | const int dif) |
| 330 | { |
| 331 | struct sock *sk; |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 332 | bool refcounted; |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 333 | |
Craig Gallek | a583636 | 2016-02-10 11:50:38 -0500 | [diff] [blame] | 334 | sk = __inet_lookup(net, hashinfo, skb, doff, saddr, sport, daddr, |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 335 | dport, dif, &refcounted); |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 336 | |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 337 | if (sk && !refcounted && !atomic_inc_not_zero(&sk->sk_refcnt)) |
| 338 | sk = NULL; |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 339 | return sk; |
| 340 | } |
Arnaldo Carvalho de Melo | a7f5e7f | 2005-12-13 23:25:31 -0800 | [diff] [blame] | 341 | |
Arnaldo Carvalho de Melo | 9a1f27c | 2008-10-07 11:41:57 -0700 | [diff] [blame] | 342 | static inline struct sock *__inet_lookup_skb(struct inet_hashinfo *hashinfo, |
| 343 | struct sk_buff *skb, |
Craig Gallek | a583636 | 2016-02-10 11:50:38 -0500 | [diff] [blame] | 344 | int doff, |
Arnaldo Carvalho de Melo | 9a1f27c | 2008-10-07 11:41:57 -0700 | [diff] [blame] | 345 | const __be16 sport, |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 346 | const __be16 dport, |
| 347 | bool *refcounted) |
Arnaldo Carvalho de Melo | 9a1f27c | 2008-10-07 11:41:57 -0700 | [diff] [blame] | 348 | { |
David S. Miller | 41063e9 | 2012-06-19 21:22:05 -0700 | [diff] [blame] | 349 | struct sock *sk = skb_steal_sock(skb); |
Arnaldo Carvalho de Melo | 9a1f27c | 2008-10-07 11:41:57 -0700 | [diff] [blame] | 350 | const struct iphdr *iph = ip_hdr(skb); |
| 351 | |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 352 | *refcounted = true; |
David S. Miller | 41063e9 | 2012-06-19 21:22:05 -0700 | [diff] [blame] | 353 | if (sk) |
KOVACS Krisztian | 2354261 | 2008-10-07 12:41:01 -0700 | [diff] [blame] | 354 | return sk; |
Eric Dumazet | 3b24d85 | 2016-04-01 08:52:17 -0700 | [diff] [blame] | 355 | |
| 356 | return __inet_lookup(dev_net(skb_dst(skb)->dev), hashinfo, skb, |
| 357 | doff, iph->saddr, sport, |
| 358 | iph->daddr, dport, inet_iif(skb), |
| 359 | refcounted); |
Arnaldo Carvalho de Melo | 9a1f27c | 2008-10-07 11:41:57 -0700 | [diff] [blame] | 360 | } |
| 361 | |
Eric Dumazet | 5b441f7 | 2015-03-18 14:05:34 -0700 | [diff] [blame] | 362 | u32 sk_ehashfn(const struct sock *sk); |
Eric Dumazet | d1e559d | 2015-03-18 14:05:35 -0700 | [diff] [blame] | 363 | u32 inet6_ehashfn(const struct net *net, |
| 364 | const struct in6_addr *laddr, const u16 lport, |
| 365 | const struct in6_addr *faddr, const __be16 fport); |
| 366 | |
| 367 | static inline void sk_daddr_set(struct sock *sk, __be32 addr) |
| 368 | { |
| 369 | sk->sk_daddr = addr; /* alias of inet_daddr */ |
| 370 | #if IS_ENABLED(CONFIG_IPV6) |
| 371 | ipv6_addr_set_v4mapped(addr, &sk->sk_v6_daddr); |
| 372 | #endif |
| 373 | } |
| 374 | |
| 375 | static inline void sk_rcv_saddr_set(struct sock *sk, __be32 addr) |
| 376 | { |
| 377 | sk->sk_rcv_saddr = addr; /* alias of inet_rcv_saddr */ |
| 378 | #if IS_ENABLED(CONFIG_IPV6) |
| 379 | ipv6_addr_set_v4mapped(addr, &sk->sk_v6_rcv_saddr); |
| 380 | #endif |
| 381 | } |
Eric Dumazet | 5b441f7 | 2015-03-18 14:05:34 -0700 | [diff] [blame] | 382 | |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 383 | int __inet_hash_connect(struct inet_timewait_death_row *death_row, |
| 384 | struct sock *sk, u32 port_offset, |
| 385 | int (*check_established)(struct inet_timewait_death_row *, |
| 386 | struct sock *, __u16, |
Eric Dumazet | b4d6444 | 2015-03-18 14:05:37 -0700 | [diff] [blame] | 387 | struct inet_timewait_sock **)); |
Eric Dumazet | 9327f70 | 2009-12-04 03:46:54 +0000 | [diff] [blame] | 388 | |
Joe Perches | 1fd5115 | 2013-09-21 10:22:41 -0700 | [diff] [blame] | 389 | int inet_hash_connect(struct inet_timewait_death_row *death_row, |
| 390 | struct sock *sk); |
Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 391 | #endif /* _INET_HASHTABLES_H */ |