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 | 8feaf0c | 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 | 33b6223 | 2005-08-09 20:09:06 -0700 | [diff] [blame] | 19 | #include <linux/ipv6.h> |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 20 | #include <linux/list.h> |
| 21 | #include <linux/slab.h> |
Arnaldo Carvalho de Melo | 33b6223 | 2005-08-09 20:09:06 -0700 | [diff] [blame] | 22 | #include <linux/socket.h> |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 23 | #include <linux/spinlock.h> |
Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 24 | #include <linux/types.h> |
Arnaldo Carvalho de Melo | f3f05f7 | 2005-08-09 20:08:09 -0700 | [diff] [blame] | 25 | #include <linux/wait.h> |
Eric Dumazet | 9e4505c | 2007-11-10 21:18:39 -0800 | [diff] [blame] | 26 | #include <linux/vmalloc.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 | c752f07 | 2005-08-09 20:08:28 -0700 | [diff] [blame] | 31 | #include <net/tcp_states.h> |
Pavel Emelyanov | 0b44191 | 2008-06-16 17:14:11 -0700 | [diff] [blame] | 32 | #include <net/netns/hash.h> |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 33 | |
Arnaldo Carvalho de Melo | f3f05f7 | 2005-08-09 20:08:09 -0700 | [diff] [blame] | 34 | #include <asm/atomic.h> |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 35 | #include <asm/byteorder.h> |
Arnaldo Carvalho de Melo | f3f05f7 | 2005-08-09 20:08:09 -0700 | [diff] [blame] | 36 | |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 37 | /* This is for all connections with a full identity, no wildcards. |
Eric Dumazet | dbca9b275 | 2007-02-08 14:16:46 -0800 | [diff] [blame] | 38 | * One chain is dedicated to TIME_WAIT sockets. |
| 39 | * I'll experiment with dynamic table growth later. |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 40 | */ |
| 41 | struct inet_ehash_bucket { |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 42 | struct hlist_head chain; |
Eric Dumazet | dbca9b275 | 2007-02-08 14:16:46 -0800 | [diff] [blame] | 43 | struct hlist_head twchain; |
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 | */ |
| 77 | struct inet_bind_bucket { |
Pavel Emelyanov | 941b1d2 | 2008-01-31 05:05:50 -0800 | [diff] [blame] | 78 | struct net *ib_net; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 79 | unsigned short port; |
| 80 | signed short fastreuse; |
| 81 | struct hlist_node node; |
| 82 | struct hlist_head owners; |
| 83 | }; |
| 84 | |
| 85 | #define inet_bind_bucket_for_each(tb, node, head) \ |
| 86 | hlist_for_each_entry(tb, node, head, node) |
| 87 | |
| 88 | struct inet_bind_hashbucket { |
| 89 | spinlock_t lock; |
| 90 | struct hlist_head chain; |
| 91 | }; |
| 92 | |
| 93 | /* This is for listening sockets, thus all sockets which possess wildcards. */ |
| 94 | #define INET_LHTABLE_SIZE 32 /* Yes, really, this is all you need. */ |
| 95 | |
| 96 | struct inet_hashinfo { |
| 97 | /* This is for sockets with full identity only. Sockets here will |
| 98 | * always be without wildcards and will have the following invariant: |
| 99 | * |
| 100 | * TCP_ESTABLISHED <= sk->sk_state < TCP_CLOSE |
| 101 | * |
Eric Dumazet | dbca9b275 | 2007-02-08 14:16:46 -0800 | [diff] [blame] | 102 | * TIME_WAIT sockets use a separate chain (twchain). |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 103 | */ |
| 104 | struct inet_ehash_bucket *ehash; |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 105 | rwlock_t *ehash_locks; |
| 106 | unsigned int ehash_size; |
| 107 | unsigned int ehash_locks_mask; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 108 | |
| 109 | /* Ok, let's try this, I give up, we do need a local binding |
| 110 | * TCP hash as well as the others for fast bind/connect. |
| 111 | */ |
| 112 | struct inet_bind_hashbucket *bhash; |
| 113 | |
Stephen Hemminger | cfcabdc | 2007-10-09 01:59:42 -0700 | [diff] [blame] | 114 | unsigned int bhash_size; |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 115 | /* Note : 4 bytes padding on 64 bit arches */ |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 116 | |
| 117 | /* All sockets in TCP_LISTEN state will be in here. This is the only |
| 118 | * table where wildcard'd TCP sockets can exist. Hash function here |
| 119 | * is just local port number. |
| 120 | */ |
| 121 | struct hlist_head listening_hash[INET_LHTABLE_SIZE]; |
| 122 | |
| 123 | /* All the above members are written once at bootup and |
| 124 | * never written again _or_ are predominantly read-access. |
| 125 | * |
| 126 | * Now align to a new cache line as all the following members |
| 127 | * are often dirty. |
| 128 | */ |
| 129 | rwlock_t lhash_lock ____cacheline_aligned; |
| 130 | atomic_t lhash_users; |
| 131 | wait_queue_head_t lhash_wait; |
Christoph Lameter | e18b890 | 2006-12-06 20:33:20 -0800 | [diff] [blame] | 132 | struct kmem_cache *bind_bucket_cachep; |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 133 | }; |
| 134 | |
Eric Dumazet | 81c3d54 | 2005-10-03 14:13:38 -0700 | [diff] [blame] | 135 | static inline struct inet_ehash_bucket *inet_ehash_bucket( |
| 136 | struct inet_hashinfo *hashinfo, |
| 137 | unsigned int hash) |
| 138 | { |
| 139 | return &hashinfo->ehash[hash & (hashinfo->ehash_size - 1)]; |
Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 140 | } |
| 141 | |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 142 | static inline rwlock_t *inet_ehash_lockp( |
| 143 | struct inet_hashinfo *hashinfo, |
| 144 | unsigned int hash) |
| 145 | { |
| 146 | return &hashinfo->ehash_locks[hash & hashinfo->ehash_locks_mask]; |
| 147 | } |
| 148 | |
| 149 | static inline int inet_ehash_locks_alloc(struct inet_hashinfo *hashinfo) |
| 150 | { |
| 151 | unsigned int i, size = 256; |
| 152 | #if defined(CONFIG_PROVE_LOCKING) |
| 153 | unsigned int nr_pcpus = 2; |
| 154 | #else |
| 155 | unsigned int nr_pcpus = num_possible_cpus(); |
| 156 | #endif |
| 157 | if (nr_pcpus >= 4) |
| 158 | size = 512; |
| 159 | if (nr_pcpus >= 8) |
| 160 | size = 1024; |
| 161 | if (nr_pcpus >= 16) |
| 162 | size = 2048; |
| 163 | if (nr_pcpus >= 32) |
| 164 | size = 4096; |
| 165 | if (sizeof(rwlock_t) != 0) { |
| 166 | #ifdef CONFIG_NUMA |
| 167 | if (size * sizeof(rwlock_t) > PAGE_SIZE) |
| 168 | hashinfo->ehash_locks = vmalloc(size * sizeof(rwlock_t)); |
| 169 | else |
| 170 | #endif |
| 171 | hashinfo->ehash_locks = kmalloc(size * sizeof(rwlock_t), |
| 172 | GFP_KERNEL); |
| 173 | if (!hashinfo->ehash_locks) |
| 174 | return ENOMEM; |
| 175 | for (i = 0; i < size; i++) |
| 176 | rwlock_init(&hashinfo->ehash_locks[i]); |
| 177 | } |
| 178 | hashinfo->ehash_locks_mask = size - 1; |
| 179 | return 0; |
| 180 | } |
| 181 | |
| 182 | static inline void inet_ehash_locks_free(struct inet_hashinfo *hashinfo) |
| 183 | { |
| 184 | if (hashinfo->ehash_locks) { |
| 185 | #ifdef CONFIG_NUMA |
| 186 | unsigned int size = (hashinfo->ehash_locks_mask + 1) * |
| 187 | sizeof(rwlock_t); |
| 188 | if (size > PAGE_SIZE) |
| 189 | vfree(hashinfo->ehash_locks); |
| 190 | else |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 191 | #endif |
Pavel Emelyanov | 218ad12 | 2007-11-26 20:23:31 +0800 | [diff] [blame] | 192 | kfree(hashinfo->ehash_locks); |
Eric Dumazet | 230140c | 2007-11-07 02:40:20 -0800 | [diff] [blame] | 193 | hashinfo->ehash_locks = NULL; |
| 194 | } |
| 195 | } |
| 196 | |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 197 | extern struct inet_bind_bucket * |
Christoph Lameter | e18b890 | 2006-12-06 20:33:20 -0800 | [diff] [blame] | 198 | inet_bind_bucket_create(struct kmem_cache *cachep, |
Pavel Emelyanov | 941b1d2 | 2008-01-31 05:05:50 -0800 | [diff] [blame] | 199 | struct net *net, |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 200 | struct inet_bind_hashbucket *head, |
| 201 | const unsigned short snum); |
Christoph Lameter | e18b890 | 2006-12-06 20:33:20 -0800 | [diff] [blame] | 202 | extern void inet_bind_bucket_destroy(struct kmem_cache *cachep, |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 203 | struct inet_bind_bucket *tb); |
| 204 | |
Pavel Emelyanov | 7f635ab | 2008-06-16 17:12:49 -0700 | [diff] [blame] | 205 | static inline int inet_bhashfn(struct net *net, |
| 206 | const __u16 lport, const int bhash_size) |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 207 | { |
Pavel Emelyanov | 0b44191 | 2008-06-16 17:14:11 -0700 | [diff] [blame] | 208 | return (lport + net_hash_mix(net)) & (bhash_size - 1); |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 209 | } |
| 210 | |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 211 | extern void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb, |
| 212 | const unsigned short snum); |
| 213 | |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 214 | /* These can have wildcards, don't try too hard. */ |
Pavel Emelyanov | 2086a65 | 2008-06-16 17:13:08 -0700 | [diff] [blame] | 215 | static inline int inet_lhashfn(struct net *net, const unsigned short num) |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 216 | { |
Pavel Emelyanov | 0b44191 | 2008-06-16 17:14:11 -0700 | [diff] [blame] | 217 | return (num + net_hash_mix(net)) & (INET_LHTABLE_SIZE - 1); |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 218 | } |
| 219 | |
| 220 | static inline int inet_sk_listen_hashfn(const struct sock *sk) |
| 221 | { |
Pavel Emelyanov | 2086a65 | 2008-06-16 17:13:08 -0700 | [diff] [blame] | 222 | return inet_lhashfn(sock_net(sk), inet_sk(sk)->num); |
Arnaldo Carvalho de Melo | 77d8bf9 | 2005-08-09 20:00:51 -0700 | [diff] [blame] | 223 | } |
| 224 | |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 225 | /* Caller must disable local BH processing. */ |
Pavel Emelyanov | 5308377 | 2008-04-17 23:18:15 -0700 | [diff] [blame] | 226 | extern void __inet_inherit_port(struct sock *sk, struct sock *child); |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 227 | |
Arnaldo Carvalho de Melo | ab1e0a1 | 2008-02-03 04:06:04 -0800 | [diff] [blame] | 228 | extern void inet_put_port(struct sock *sk); |
Arnaldo Carvalho de Melo | 2d8c4ce | 2005-08-09 20:07:13 -0700 | [diff] [blame] | 229 | |
Arnaldo Carvalho de Melo | f3f05f7 | 2005-08-09 20:08:09 -0700 | [diff] [blame] | 230 | extern void inet_listen_wlock(struct inet_hashinfo *hashinfo); |
| 231 | |
| 232 | /* |
| 233 | * - We may sleep inside this lock. |
| 234 | * - If sleeping is not required (or called from BH), |
| 235 | * use plain read_(un)lock(&inet_hashinfo.lhash_lock). |
| 236 | */ |
| 237 | static inline void inet_listen_lock(struct inet_hashinfo *hashinfo) |
| 238 | { |
| 239 | /* read_lock synchronizes to candidates to writers */ |
| 240 | read_lock(&hashinfo->lhash_lock); |
| 241 | atomic_inc(&hashinfo->lhash_users); |
| 242 | read_unlock(&hashinfo->lhash_lock); |
| 243 | } |
| 244 | |
| 245 | static inline void inet_listen_unlock(struct inet_hashinfo *hashinfo) |
| 246 | { |
| 247 | if (atomic_dec_and_test(&hashinfo->lhash_users)) |
| 248 | wake_up(&hashinfo->lhash_wait); |
| 249 | } |
| 250 | |
Arnaldo Carvalho de Melo | ab1e0a1 | 2008-02-03 04:06:04 -0800 | [diff] [blame] | 251 | extern void __inet_hash_nolisten(struct sock *sk); |
| 252 | extern void inet_hash(struct sock *sk); |
| 253 | extern void inet_unhash(struct sock *sk); |
Arnaldo Carvalho de Melo | 33b6223 | 2005-08-09 20:09:06 -0700 | [diff] [blame] | 254 | |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 255 | extern struct sock *__inet_lookup_listener(struct net *net, |
| 256 | struct inet_hashinfo *hashinfo, |
Al Viro | fb99c84 | 2006-09-27 18:43:33 -0700 | [diff] [blame] | 257 | const __be32 daddr, |
Herbert Xu | 8f49106 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 258 | const unsigned short hnum, |
| 259 | const int dif); |
| 260 | |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 261 | static inline struct sock *inet_lookup_listener(struct net *net, |
| 262 | struct inet_hashinfo *hashinfo, |
| 263 | __be32 daddr, __be16 dport, int dif) |
Herbert Xu | 8f49106 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 264 | { |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 265 | return __inet_lookup_listener(net, hashinfo, daddr, ntohs(dport), dif); |
Herbert Xu | 8f49106 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 266 | } |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 267 | |
| 268 | /* Socket demux engine toys. */ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 269 | /* What happens here is ugly; there's a pair of adjacent fields in |
| 270 | struct inet_sock; __be16 dport followed by __u16 num. We want to |
| 271 | search by pair, so we combine the keys into a single 32bit value |
| 272 | and compare with 32bit value read from &...->dport. Let's at least |
| 273 | make sure that it's not mixed with anything else... |
| 274 | On 64bit targets we combine comparisons with pair of adjacent __be32 |
| 275 | fields in the same way. |
| 276 | */ |
| 277 | typedef __u32 __bitwise __portpair; |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 278 | #ifdef __BIG_ENDIAN |
| 279 | #define INET_COMBINED_PORTS(__sport, __dport) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 280 | ((__force __portpair)(((__force __u32)(__be16)(__sport) << 16) | (__u32)(__dport))) |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 281 | #else /* __LITTLE_ENDIAN */ |
| 282 | #define INET_COMBINED_PORTS(__sport, __dport) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 283 | ((__force __portpair)(((__u32)(__dport) << 16) | (__force __u32)(__be16)(__sport))) |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 284 | #endif |
| 285 | |
| 286 | #if (BITS_PER_LONG == 64) |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 287 | typedef __u64 __bitwise __addrpair; |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 288 | #ifdef __BIG_ENDIAN |
| 289 | #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 290 | const __addrpair __name = (__force __addrpair) ( \ |
| 291 | (((__force __u64)(__be32)(__saddr)) << 32) | \ |
| 292 | ((__force __u64)(__be32)(__daddr))); |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 293 | #else /* __LITTLE_ENDIAN */ |
| 294 | #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 295 | const __addrpair __name = (__force __addrpair) ( \ |
| 296 | (((__force __u64)(__be32)(__daddr)) << 32) | \ |
| 297 | ((__force __u64)(__be32)(__saddr))); |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 298 | #endif /* __BIG_ENDIAN */ |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 299 | #define INET_MATCH(__sk, __net, __hash, __cookie, __saddr, __daddr, __ports, __dif)\ |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 300 | (((__sk)->sk_hash == (__hash)) && sock_net((__sk)) == (__net) && \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 301 | ((*((__addrpair *)&(inet_sk(__sk)->daddr))) == (__cookie)) && \ |
| 302 | ((*((__portpair *)&(inet_sk(__sk)->dport))) == (__ports)) && \ |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 303 | (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif)))) |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 304 | #define INET_TW_MATCH(__sk, __net, __hash, __cookie, __saddr, __daddr, __ports, __dif)\ |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 305 | (((__sk)->sk_hash == (__hash)) && sock_net((__sk)) == (__net) && \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 306 | ((*((__addrpair *)&(inet_twsk(__sk)->tw_daddr))) == (__cookie)) && \ |
| 307 | ((*((__portpair *)&(inet_twsk(__sk)->tw_dport))) == (__ports)) && \ |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 308 | (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif)))) |
| 309 | #else /* 32-bit arch */ |
| 310 | #define INET_ADDR_COOKIE(__name, __saddr, __daddr) |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 311 | #define INET_MATCH(__sk, __net, __hash, __cookie, __saddr, __daddr, __ports, __dif) \ |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 312 | (((__sk)->sk_hash == (__hash)) && sock_net((__sk)) == (__net) && \ |
Eric Dumazet | 81c3d54 | 2005-10-03 14:13:38 -0700 | [diff] [blame] | 313 | (inet_sk(__sk)->daddr == (__saddr)) && \ |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 314 | (inet_sk(__sk)->rcv_saddr == (__daddr)) && \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 315 | ((*((__portpair *)&(inet_sk(__sk)->dport))) == (__ports)) && \ |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 316 | (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif)))) |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 317 | #define INET_TW_MATCH(__sk, __net, __hash,__cookie, __saddr, __daddr, __ports, __dif) \ |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 318 | (((__sk)->sk_hash == (__hash)) && sock_net((__sk)) == (__net) && \ |
Eric Dumazet | 81c3d54 | 2005-10-03 14:13:38 -0700 | [diff] [blame] | 319 | (inet_twsk(__sk)->tw_daddr == (__saddr)) && \ |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 320 | (inet_twsk(__sk)->tw_rcv_saddr == (__daddr)) && \ |
Al Viro | 4f765d8 | 2006-09-27 18:43:07 -0700 | [diff] [blame] | 321 | ((*((__portpair *)&(inet_twsk(__sk)->tw_dport))) == (__ports)) && \ |
Arnaldo Carvalho de Melo | 8feaf0c | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 322 | (!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif)))) |
| 323 | #endif /* 64-bit arch */ |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 324 | |
| 325 | /* |
| 326 | * Sockets in TCP_CLOSE state are _always_ taken out of the hash, so we need |
| 327 | * not check it for lookups anymore, thanks Alexey. -DaveM |
| 328 | * |
| 329 | * Local BH must be disabled here. |
| 330 | */ |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 331 | extern struct sock * __inet_lookup_established(struct net *net, |
| 332 | struct inet_hashinfo *hashinfo, |
Pavel Emelyanov | 77a5ba5 | 2007-12-20 15:32:17 -0800 | [diff] [blame] | 333 | const __be32 saddr, const __be16 sport, |
| 334 | const __be32 daddr, const u16 hnum, const int dif); |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 335 | |
Herbert Xu | 8f49106 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 336 | static inline struct sock * |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 337 | inet_lookup_established(struct net *net, struct inet_hashinfo *hashinfo, |
Al Viro | fb99c84 | 2006-09-27 18:43:33 -0700 | [diff] [blame] | 338 | const __be32 saddr, const __be16 sport, |
| 339 | const __be32 daddr, const __be16 dport, |
Herbert Xu | 8f49106 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 340 | const int dif) |
| 341 | { |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 342 | return __inet_lookup_established(net, hashinfo, saddr, sport, daddr, |
Herbert Xu | 8f49106 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 343 | ntohs(dport), dif); |
| 344 | } |
| 345 | |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 346 | static inline struct sock *__inet_lookup(struct net *net, |
| 347 | struct inet_hashinfo *hashinfo, |
Al Viro | fb99c84 | 2006-09-27 18:43:33 -0700 | [diff] [blame] | 348 | const __be32 saddr, const __be16 sport, |
| 349 | const __be32 daddr, const __be16 dport, |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 350 | const int dif) |
| 351 | { |
Herbert Xu | 8f49106 | 2006-08-09 15:47:12 -0700 | [diff] [blame] | 352 | u16 hnum = ntohs(dport); |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 353 | struct sock *sk = __inet_lookup_established(net, hashinfo, |
| 354 | saddr, sport, daddr, hnum, dif); |
| 355 | |
| 356 | return sk ? : __inet_lookup_listener(net, hashinfo, daddr, hnum, dif); |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 357 | } |
| 358 | |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 359 | static inline struct sock *inet_lookup(struct net *net, |
| 360 | struct inet_hashinfo *hashinfo, |
Al Viro | fb99c84 | 2006-09-27 18:43:33 -0700 | [diff] [blame] | 361 | const __be32 saddr, const __be16 sport, |
| 362 | const __be32 daddr, const __be16 dport, |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 363 | const int dif) |
| 364 | { |
| 365 | struct sock *sk; |
| 366 | |
| 367 | local_bh_disable(); |
Pavel Emelyanov | c67499c | 2008-01-31 05:06:40 -0800 | [diff] [blame] | 368 | sk = __inet_lookup(net, hashinfo, saddr, sport, daddr, dport, dif); |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 369 | local_bh_enable(); |
| 370 | |
| 371 | return sk; |
| 372 | } |
Arnaldo Carvalho de Melo | a7f5e7f | 2005-12-13 23:25:31 -0800 | [diff] [blame] | 373 | |
Pavel Emelyanov | 5ee31fc | 2008-01-31 05:04:45 -0800 | [diff] [blame] | 374 | extern int __inet_hash_connect(struct inet_timewait_death_row *death_row, |
Pavel Emelyanov | 5d8c0aa | 2008-02-05 03:14:44 -0800 | [diff] [blame] | 375 | struct sock *sk, u32 port_offset, |
Pavel Emelyanov | 5ee31fc | 2008-01-31 05:04:45 -0800 | [diff] [blame] | 376 | int (*check_established)(struct inet_timewait_death_row *, |
| 377 | struct sock *, __u16, struct inet_timewait_sock **), |
Arnaldo Carvalho de Melo | ab1e0a1 | 2008-02-03 04:06:04 -0800 | [diff] [blame] | 378 | void (*hash)(struct sock *sk)); |
Arnaldo Carvalho de Melo | a7f5e7f | 2005-12-13 23:25:31 -0800 | [diff] [blame] | 379 | extern int inet_hash_connect(struct inet_timewait_death_row *death_row, |
| 380 | struct sock *sk); |
Arnaldo Carvalho de Melo | 304a161 | 2005-08-09 19:59:20 -0700 | [diff] [blame] | 381 | #endif /* _INET_HASHTABLES_H */ |