~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/net/inet_hashtables.h

Version: ~ [ linux-5.7-rc7 ] ~ [ linux-5.6.14 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.42 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.124 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.181 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.224 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.224 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.84 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 
 17 
 18 #include <linux/interrupt.h>
 19 #include <linux/ip.h>
 20 #include <linux/ipv6.h>
 21 #include <linux/list.h>
 22 #include <linux/slab.h>
 23 #include <linux/socket.h>
 24 #include <linux/spinlock.h>
 25 #include <linux/types.h>
 26 #include <linux/wait.h>
 27 #include <linux/vmalloc.h>
 28 
 29 #include <net/inet_connection_sock.h>
 30 #include <net/inet_sock.h>
 31 #include <net/sock.h>
 32 #include <net/route.h>
 33 #include <net/tcp_states.h>
 34 #include <net/netns/hash.h>
 35 
 36 #include <linux/atomic.h>
 37 #include <asm/byteorder.h>
 38 
 39 /* This is for all connections with a full identity, no wildcards.
 40  * The 'e' prefix stands for Establish, but we really put all sockets
 41  * but LISTEN ones.
 42  */
 43 struct inet_ehash_bucket {
 44         struct hlist_nulls_head chain;
 45 };
 46 
 47 /* There are a few simple rules, which allow for local port reuse by
 48  * an application.  In essence:
 49  *
 50  *      1) Sockets bound to different interfaces may share a local port.
 51  *         Failing that, goto test 2.
 52  *      2) If all sockets have sk->sk_reuse set, and none of them are in
 53  *         TCP_LISTEN state, the port may be shared.
 54  *         Failing that, goto test 3.
 55  *      3) If all sockets are bound to a specific inet_sk(sk)->rcv_saddr local
 56  *         address, and none of them are the same, the port may be
 57  *         shared.
 58  *         Failing this, the port cannot be shared.
 59  *
 60  * The interesting point, is test #2.  This is what an FTP server does
 61  * all day.  To optimize this case we use a specific flag bit defined
 62  * below.  As we add sockets to a bind bucket list, we perform a
 63  * check of: (newsk->sk_reuse && (newsk->sk_state != TCP_LISTEN))
 64  * As long as all sockets added to a bind bucket pass this test,
 65  * the flag bit will be set.
 66  * The resulting situation is that tcp_v[46]_verify_bind() can just check
 67  * for this flag bit, if it is set and the socket trying to bind has
 68  * sk->sk_reuse set, we don't even have to walk the owners list at all,
 69  * we return that it is ok to bind this socket to the requested local port.
 70  *
 71  * Sounds like a lot of work, but it is worth it.  In a more naive
 72  * implementation (ie. current FreeBSD etc.) the entire list of ports
 73  * must be walked for each data port opened by an ftp server.  Needless
 74  * to say, this does not scale at all.  With a couple thousand FTP
 75  * users logged onto your box, isn't it nice to know that new data
 76  * ports are created in O(1) time?  I thought so. ;-)   -DaveM
 77  */
 78 struct inet_bind_bucket {
 79 #ifdef CONFIG_NET_NS
 80         struct net              *ib_net;
 81 #endif
 82         unsigned short          port;
 83         signed char             fastreuse;
 84         signed char             fastreuseport;
 85         kuid_t                  fastuid;
 86         int                     num_owners;
 87         struct hlist_node       node;
 88         struct hlist_head       owners;
 89 };
 90 
 91 static inline struct net *ib_net(struct inet_bind_bucket *ib)
 92 {
 93         return read_pnet(&ib->ib_net);
 94 }
 95 
 96 #define inet_bind_bucket_for_each(tb, head) \
 97         hlist_for_each_entry(tb, head, node)
 98 
 99 struct inet_bind_hashbucket {
100         spinlock_t              lock;
101         struct hlist_head       chain;
102 };
103 
104 /*
105  * Sockets can be hashed in established or listening table
106  * We must use different 'nulls' end-of-chain value for listening
107  * hash table, or we might find a socket that was closed and
108  * reallocated/inserted into established hash table
109  */
110 #define LISTENING_NULLS_BASE (1U << 29)
111 struct inet_listen_hashbucket {
112         spinlock_t              lock;
113         struct hlist_nulls_head head;
114 };
115 
116 /* This is for listening sockets, thus all sockets which possess wildcards. */
117 #define INET_LHTABLE_SIZE       32      /* Yes, really, this is all you need. */
118 
119 struct inet_hashinfo {
120         /* This is for sockets with full identity only.  Sockets here will
121          * always be without wildcards and will have the following invariant:
122          *
123          *          TCP_ESTABLISHED <= sk->sk_state < TCP_CLOSE
124          *
125          */
126         struct inet_ehash_bucket        *ehash;
127         spinlock_t                      *ehash_locks;
128         unsigned int                    ehash_mask;
129         unsigned int                    ehash_locks_mask;
130 
131         /* Ok, let's try this, I give up, we do need a local binding
132          * TCP hash as well as the others for fast bind/connect.
133          */
134         struct inet_bind_hashbucket     *bhash;
135 
136         unsigned int                    bhash_size;
137         /* 4 bytes hole on 64 bit */
138 
139         struct kmem_cache               *bind_bucket_cachep;
140 
141         /* All the above members are written once at bootup and
142          * never written again _or_ are predominantly read-access.
143          *
144          * Now align to a new cache line as all the following members
145          * might be often dirty.
146          */
147         /* All sockets in TCP_LISTEN state will be in here.  This is the only
148          * table where wildcard'd TCP sockets can exist.  Hash function here
149          * is just local port number.
150          */
151         struct inet_listen_hashbucket   listening_hash[INET_LHTABLE_SIZE]
152                                         ____cacheline_aligned_in_smp;
153 
154         atomic_t                        bsockets;
155 };
156 
157 static inline struct inet_ehash_bucket *inet_ehash_bucket(
158         struct inet_hashinfo *hashinfo,
159         unsigned int hash)
160 {
161         return &hashinfo->ehash[hash & hashinfo->ehash_mask];
162 }
163 
164 static inline spinlock_t *inet_ehash_lockp(
165         struct inet_hashinfo *hashinfo,
166         unsigned int hash)
167 {
168         return &hashinfo->ehash_locks[hash & hashinfo->ehash_locks_mask];
169 }
170 
171 static inline int inet_ehash_locks_alloc(struct inet_hashinfo *hashinfo)
172 {
173         unsigned int i, size = 256;
174 #if defined(CONFIG_PROVE_LOCKING)
175         unsigned int nr_pcpus = 2;
176 #else
177         unsigned int nr_pcpus = num_possible_cpus();
178 #endif
179         if (nr_pcpus >= 4)
180                 size = 512;
181         if (nr_pcpus >= 8)
182                 size = 1024;
183         if (nr_pcpus >= 16)
184                 size = 2048;
185         if (nr_pcpus >= 32)
186                 size = 4096;
187         if (sizeof(spinlock_t) != 0) {
188 #ifdef CONFIG_NUMA
189                 if (size * sizeof(spinlock_t) > PAGE_SIZE)
190                         hashinfo->ehash_locks = vmalloc(size * sizeof(spinlock_t));
191                 else
192 #endif
193                 hashinfo->ehash_locks = kmalloc(size * sizeof(spinlock_t),
194                                                 GFP_KERNEL);
195                 if (!hashinfo->ehash_locks)
196                         return ENOMEM;
197                 for (i = 0; i < size; i++)
198                         spin_lock_init(&hashinfo->ehash_locks[i]);
199         }
200         hashinfo->ehash_locks_mask = size - 1;
201         return 0;
202 }
203 
204 static inline void inet_ehash_locks_free(struct inet_hashinfo *hashinfo)
205 {
206         if (hashinfo->ehash_locks) {
207 #ifdef CONFIG_NUMA
208                 unsigned int size = (hashinfo->ehash_locks_mask + 1) *
209                                                         sizeof(spinlock_t);
210                 if (size > PAGE_SIZE)
211                         vfree(hashinfo->ehash_locks);
212                 else
213 #endif
214                 kfree(hashinfo->ehash_locks);
215                 hashinfo->ehash_locks = NULL;
216         }
217 }
218 
219 struct inet_bind_bucket *
220 inet_bind_bucket_create(struct kmem_cache *cachep, struct net *net,
221                         struct inet_bind_hashbucket *head,
222                         const unsigned short snum);
223 void inet_bind_bucket_destroy(struct kmem_cache *cachep,
224                               struct inet_bind_bucket *tb);
225 
226 static inline int inet_bhashfn(struct net *net, const __u16 lport,
227                                const int bhash_size)
228 {
229         return (lport + net_hash_mix(net)) & (bhash_size - 1);
230 }
231 
232 void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
233                     const unsigned short snum);
234 
235 /* These can have wildcards, don't try too hard. */
236 static inline int inet_lhashfn(struct net *net, const unsigned short num)
237 {
238         return (num + net_hash_mix(net)) & (INET_LHTABLE_SIZE - 1);
239 }
240 
241 static inline int inet_sk_listen_hashfn(const struct sock *sk)
242 {
243         return inet_lhashfn(sock_net(sk), inet_sk(sk)->inet_num);
244 }
245 
246 /* Caller must disable local BH processing. */
247 int __inet_inherit_port(struct sock *sk, struct sock *child);
248 
249 void inet_put_port(struct sock *sk);
250 
251 void inet_hashinfo_init(struct inet_hashinfo *h);
252 
253 int __inet_hash_nolisten(struct sock *sk, struct inet_timewait_sock *tw);
254 void inet_hash(struct sock *sk);
255 void inet_unhash(struct sock *sk);
256 
257 struct sock *__inet_lookup_listener(struct net *net,
258                                     struct inet_hashinfo *hashinfo,
259                                     const __be32 saddr, const __be16 sport,
260                                     const __be32 daddr,
261                                     const unsigned short hnum,
262                                     const int dif);
263 
264 static inline struct sock *inet_lookup_listener(struct net *net,
265                 struct inet_hashinfo *hashinfo,
266                 __be32 saddr, __be16 sport,
267                 __be32 daddr, __be16 dport, int dif)
268 {
269         return __inet_lookup_listener(net, hashinfo, saddr, sport,
270                                       daddr, ntohs(dport), dif);
271 }
272 
273 /* Socket demux engine toys. */
274 /* What happens here is ugly; there's a pair of adjacent fields in
275    struct inet_sock; __be16 dport followed by __u16 num.  We want to
276    search by pair, so we combine the keys into a single 32bit value
277    and compare with 32bit value read from &...->dport.  Let's at least
278    make sure that it's not mixed with anything else...
279    On 64bit targets we combine comparisons with pair of adjacent __be32
280    fields in the same way.
281 */
282 #ifdef __BIG_ENDIAN
283 #define INET_COMBINED_PORTS(__sport, __dport) \
284         ((__force __portpair)(((__force __u32)(__be16)(__sport) << 16) | (__u32)(__dport)))
285 #else /* __LITTLE_ENDIAN */
286 #define INET_COMBINED_PORTS(__sport, __dport) \
287         ((__force __portpair)(((__u32)(__dport) << 16) | (__force __u32)(__be16)(__sport)))
288 #endif
289 
290 #if (BITS_PER_LONG == 64)
291 #ifdef __BIG_ENDIAN
292 #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \
293         const __addrpair __name = (__force __addrpair) ( \
294                                    (((__force __u64)(__be32)(__saddr)) << 32) | \
295                                    ((__force __u64)(__be32)(__daddr)));
296 #else /* __LITTLE_ENDIAN */
297 #define INET_ADDR_COOKIE(__name, __saddr, __daddr) \
298         const __addrpair __name = (__force __addrpair) ( \
299                                    (((__force __u64)(__be32)(__daddr)) << 32) | \
300                                    ((__force __u64)(__be32)(__saddr)));
301 #endif /* __BIG_ENDIAN */
302 #define INET_MATCH(__sk, __net, __cookie, __saddr, __daddr, __ports, __dif)     \
303         (((__sk)->sk_portpair == (__ports))                     &&      \
304          ((__sk)->sk_addrpair == (__cookie))                    &&      \
305          (!(__sk)->sk_bound_dev_if      ||                              \
306            ((__sk)->sk_bound_dev_if == (__dif)))                &&      \
307          net_eq(sock_net(__sk), (__net)))
308 #else /* 32-bit arch */
309 #define INET_ADDR_COOKIE(__name, __saddr, __daddr)
310 #define INET_MATCH(__sk, __net, __cookie, __saddr, __daddr, __ports, __dif) \
311         (((__sk)->sk_portpair == (__ports))             &&              \
312          ((__sk)->sk_daddr      == (__saddr))           &&              \
313          ((__sk)->sk_rcv_saddr  == (__daddr))           &&              \
314          (!(__sk)->sk_bound_dev_if      ||                              \
315            ((__sk)->sk_bound_dev_if == (__dif)))        &&              \
316          net_eq(sock_net(__sk), (__net)))
317 #endif /* 64-bit arch */
318 
319 /*
320  * Sockets in TCP_CLOSE state are _always_ taken out of the hash, so we need
321  * not check it for lookups anymore, thanks Alexey. -DaveM
322  *
323  * Local BH must be disabled here.
324  */
325 struct sock *__inet_lookup_established(struct net *net,
326                                        struct inet_hashinfo *hashinfo,
327                                        const __be32 saddr, const __be16 sport,
328                                        const __be32 daddr, const u16 hnum,
329                                        const int dif);
330 
331 static inline struct sock *
332         inet_lookup_established(struct net *net, struct inet_hashinfo *hashinfo,
333                                 const __be32 saddr, const __be16 sport,
334                                 const __be32 daddr, const __be16 dport,
335                                 const int dif)
336 {
337         return __inet_lookup_established(net, hashinfo, saddr, sport, daddr,
338                                          ntohs(dport), dif);
339 }
340 
341 static inline struct sock *__inet_lookup(struct net *net,
342                                          struct inet_hashinfo *hashinfo,
343                                          const __be32 saddr, const __be16 sport,
344                                          const __be32 daddr, const __be16 dport,
345                                          const int dif)
346 {
347         u16 hnum = ntohs(dport);
348         struct sock *sk = __inet_lookup_established(net, hashinfo,
349                                 saddr, sport, daddr, hnum, dif);
350 
351         return sk ? : __inet_lookup_listener(net, hashinfo, saddr, sport,
352                                              daddr, hnum, dif);
353 }
354 
355 static inline struct sock *inet_lookup(struct net *net,
356                                        struct inet_hashinfo *hashinfo,
357                                        const __be32 saddr, const __be16 sport,
358                                        const __be32 daddr, const __be16 dport,
359                                        const int dif)
360 {
361         struct sock *sk;
362 
363         local_bh_disable();
364         sk = __inet_lookup(net, hashinfo, saddr, sport, daddr, dport, dif);
365         local_bh_enable();
366 
367         return sk;
368 }
369 
370 static inline struct sock *__inet_lookup_skb(struct inet_hashinfo *hashinfo,
371                                              struct sk_buff *skb,
372                                              const __be16 sport,
373                                              const __be16 dport)
374 {
375         struct sock *sk = skb_steal_sock(skb);
376         const struct iphdr *iph = ip_hdr(skb);
377 
378         if (sk)
379                 return sk;
380         else
381                 return __inet_lookup(dev_net(skb_dst(skb)->dev), hashinfo,
382                                      iph->saddr, sport,
383                                      iph->daddr, dport, inet_iif(skb));
384 }
385 
386 int __inet_hash_connect(struct inet_timewait_death_row *death_row,
387                         struct sock *sk, u32 port_offset,
388                         int (*check_established)(struct inet_timewait_death_row *,
389                                                  struct sock *, __u16,
390                                                  struct inet_timewait_sock **),
391                         int (*hash)(struct sock *sk,
392                                     struct inet_timewait_sock *twp));
393 
394 int inet_hash_connect(struct inet_timewait_death_row *death_row,
395                       struct sock *sk);
396 #endif /* _INET_HASHTABLES_H */
397 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp