[linux-block.git] / include / net / inet_hashtables.h

/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the BSD Socket
 *		interface as the means of communication with the user level.
 *
 * Authors:	Lotsa people, from code originally in tcp
 *
 *	This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

#ifndef _INET_HASHTABLES_H
#define _INET_HASHTABLES_H

#include <linux/interrupt.h>
#include <linux/ip.h>
#include <linux/ipv6.h>
#include <linux/list.h>
#include <linux/slab.h>
#include <linux/socket.h>
#include <linux/spinlock.h>
#include <linux/types.h>
#include <linux/wait.h>

#include <net/sock.h>
#include <net/tcp_states.h>

#include <asm/atomic.h>

/* This is for all connections with a full identity, no wildcards.
 * New scheme, half the table is for TIME_WAIT, the other half is
 * for the rest.  I'll experiment with dynamic table growth later.
 */
struct inet_ehash_bucket {
	rwlock_t	  lock;
	struct hlist_head chain;
} __attribute__((__aligned__(8)));

/* There are a few simple rules, which allow for local port reuse by
 * an application.  In essence:
 *
 *	1) Sockets bound to different interfaces may share a local port.
 *	   Failing that, goto test 2.
 *	2) If all sockets have sk->sk_reuse set, and none of them are in
 *	   TCP_LISTEN state, the port may be shared.
 *	   Failing that, goto test 3.
 *	3) If all sockets are bound to a specific inet_sk(sk)->rcv_saddr local
 *	   address, and none of them are the same, the port may be
 *	   shared.
 *	   Failing this, the port cannot be shared.
 *
 * The interesting point, is test #2.  This is what an FTP server does
 * all day.  To optimize this case we use a specific flag bit defined
 * below.  As we add sockets to a bind bucket list, we perform a
 * check of: (newsk->sk_reuse && (newsk->sk_state != TCP_LISTEN))
 * As long as all sockets added to a bind bucket pass this test,
 * the flag bit will be set.
 * The resulting situation is that tcp_v[46]_verify_bind() can just check
 * for this flag bit, if it is set and the socket trying to bind has
 * sk->sk_reuse set, we don't even have to walk the owners list at all,
 * we return that it is ok to bind this socket to the requested local port.
 *
 * Sounds like a lot of work, but it is worth it.  In a more naive
 * implementation (ie. current FreeBSD etc.) the entire list of ports
 * must be walked for each data port opened by an ftp server.  Needless
 * to say, this does not scale at all.  With a couple thousand FTP
 * users logged onto your box, isn't it nice to know that new data
 * ports are created in O(1) time?  I thought so. ;-)	-DaveM
 */
struct inet_bind_bucket {
	unsigned short		port;
	signed short		fastreuse;
	struct hlist_node	node;
	struct hlist_head	owners;
};

#define inet_bind_bucket_for_each(tb, node, head) \
	hlist_for_each_entry(tb, node, head, node)

struct inet_bind_hashbucket {
	spinlock_t		lock;
	struct hlist_head	chain;
};

/* This is for listening sockets, thus all sockets which possess wildcards. */
#define INET_LHTABLE_SIZE	32	/* Yes, really, this is all you need. */

struct inet_hashinfo {
	/* This is for sockets with full identity only.  Sockets here will
	 * always be without wildcards and will have the following invariant:
	 *
	 *          TCP_ESTABLISHED <= sk->sk_state < TCP_CLOSE
	 *
	 * First half of the table is for sockets not in TIME_WAIT, second half
	 * is for TIME_WAIT sockets only.
	 */
	struct inet_ehash_bucket	*ehash;

	/* Ok, let's try this, I give up, we do need a local binding
	 * TCP hash as well as the others for fast bind/connect.
	 */
	struct inet_bind_hashbucket	*bhash;

	int				bhash_size;
	int				ehash_size;

	/* All sockets in TCP_LISTEN state will be in here.  This is the only
	 * table where wildcard'd TCP sockets can exist.  Hash function here
	 * is just local port number.
	 */
	struct hlist_head		listening_hash[INET_LHTABLE_SIZE];

	/* All the above members are written once at bootup and
	 * never written again _or_ are predominantly read-access.
	 *
	 * Now align to a new cache line as all the following members
	 * are often dirty.
	 */
	rwlock_t			lhash_lock ____cacheline_aligned;
	atomic_t			lhash_users;
	wait_queue_head_t		lhash_wait;
	spinlock_t			portalloc_lock;
	kmem_cache_t			*bind_bucket_cachep;
	int				port_rover;
};

static inline int inet_ehashfn(const __u32 laddr, const __u16 lport,
			       const __u32 faddr, const __u16 fport,
			       const int ehash_size)
{
	int h = (laddr ^ lport) ^ (faddr ^ fport);
	h ^= h >> 16;
	h ^= h >> 8;
	return h & (ehash_size - 1);
}

static inline int inet_sk_ehashfn(const struct sock *sk, const int ehash_size)
{
	const struct inet_sock *inet = inet_sk(sk);
	const __u32 laddr = inet->rcv_saddr;
	const __u16 lport = inet->num;
	const __u32 faddr = inet->daddr;
	const __u16 fport = inet->dport;

	return inet_ehashfn(laddr, lport, faddr, fport, ehash_size);
}

extern struct inet_bind_bucket *
		    inet_bind_bucket_create(kmem_cache_t *cachep,
					    struct inet_bind_hashbucket *head,
					    const unsigned short snum);
extern void inet_bind_bucket_destroy(kmem_cache_t *cachep,
				     struct inet_bind_bucket *tb);

static inline int inet_bhashfn(const __u16 lport, const int bhash_size)
{
	return lport & (bhash_size - 1);
}

extern void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
			   const unsigned short snum);

/* These can have wildcards, don't try too hard. */
static inline int inet_lhashfn(const unsigned short num)
{
	return num & (INET_LHTABLE_SIZE - 1);
}

static inline int inet_sk_listen_hashfn(const struct sock *sk)
{
	return inet_lhashfn(inet_sk(sk)->num);
}

/* Caller must disable local BH processing. */
static inline void __inet_inherit_port(struct inet_hashinfo *table,
				       struct sock *sk, struct sock *child)
{
	const int bhash = inet_bhashfn(inet_sk(child)->num, table->bhash_size);
	struct inet_bind_hashbucket *head = &table->bhash[bhash];
	struct inet_bind_bucket *tb;

	spin_lock(&head->lock);
	tb = inet_sk(sk)->bind_hash;
	sk_add_bind_node(child, &tb->owners);
	inet_sk(child)->bind_hash = tb;
	spin_unlock(&head->lock);
}

static inline void inet_inherit_port(struct inet_hashinfo *table,
				     struct sock *sk, struct sock *child)
{
	local_bh_disable();
	__inet_inherit_port(table, sk, child);
	local_bh_enable();
}

extern void inet_put_port(struct inet_hashinfo *table, struct sock *sk);

extern void inet_listen_wlock(struct inet_hashinfo *hashinfo);

/*
 * - We may sleep inside this lock.
 * - If sleeping is not required (or called from BH),
 *   use plain read_(un)lock(&inet_hashinfo.lhash_lock).
 */
static inline void inet_listen_lock(struct inet_hashinfo *hashinfo)
{
	/* read_lock synchronizes to candidates to writers */
	read_lock(&hashinfo->lhash_lock);
	atomic_inc(&hashinfo->lhash_users);
	read_unlock(&hashinfo->lhash_lock);
}

static inline void inet_listen_unlock(struct inet_hashinfo *hashinfo)
{
	if (atomic_dec_and_test(&hashinfo->lhash_users))
		wake_up(&hashinfo->lhash_wait);
}

static inline void __inet_hash(struct inet_hashinfo *hashinfo,
			       struct sock *sk, const int listen_possible)
{
	struct hlist_head *list;
	rwlock_t *lock;

	BUG_TRAP(sk_unhashed(sk));
	if (listen_possible && sk->sk_state == TCP_LISTEN) {
		list = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
		lock = &hashinfo->lhash_lock;
		inet_listen_wlock(hashinfo);
	} else {
		sk->sk_hashent = inet_sk_ehashfn(sk, hashinfo->ehash_size);
		list = &hashinfo->ehash[sk->sk_hashent].chain;
		lock = &hashinfo->ehash[sk->sk_hashent].lock;
		write_lock(lock);
	}
	__sk_add_node(sk, list);
	sock_prot_inc_use(sk->sk_prot);
	write_unlock(lock);
	if (listen_possible && sk->sk_state == TCP_LISTEN)
		wake_up(&hashinfo->lhash_wait);
}

static inline void inet_hash(struct inet_hashinfo *hashinfo, struct sock *sk)
{
	if (sk->sk_state != TCP_CLOSE) {
		local_bh_disable();
		__inet_hash(hashinfo, sk, 1);
		local_bh_enable();
	}
}

static inline void inet_unhash(struct inet_hashinfo *hashinfo, struct sock *sk)
{
	rwlock_t *lock;

	if (sk_unhashed(sk))
		goto out;

	if (sk->sk_state == TCP_LISTEN) {
		local_bh_disable();
		inet_listen_wlock(hashinfo);
		lock = &hashinfo->lhash_lock;
	} else {
		struct inet_ehash_bucket *head = &hashinfo->ehash[sk->sk_hashent];
		lock = &head->lock;
		write_lock_bh(&head->lock);
	}

	if (__sk_del_node_init(sk))
		sock_prot_dec_use(sk->sk_prot);
	write_unlock_bh(lock);
out:
	if (sk->sk_state == TCP_LISTEN)
		wake_up(&hashinfo->lhash_wait);
}

extern struct sock *__inet_lookup_listener(const struct hlist_head *head,
					   const u32 daddr,
					   const unsigned short hnum,
					   const int dif);

/* Optimize the common listener case. */
static inline struct sock *inet_lookup_listener(struct inet_hashinfo *hashinfo,
						const u32 daddr,
						const unsigned short hnum,
						const int dif)
{
	struct sock *sk = NULL;
	struct hlist_head *head;

	read_lock(&hashinfo->lhash_lock);
	head = &hashinfo->listening_hash[inet_lhashfn(hnum)];
	if (!hlist_empty(head)) {
		const struct inet_sock *inet = inet_sk((sk = __sk_head(head)));

		if (inet->num == hnum && !sk->sk_node.next &&
		    (!inet->rcv_saddr || inet->rcv_saddr == daddr) &&
		    (sk->sk_family == PF_INET || !ipv6_only_sock(sk)) &&
		    !sk->sk_bound_dev_if)
			goto sherry_cache;
		sk = __inet_lookup_listener(head, daddr, hnum, dif);
	}
	if (sk) {
sherry_cache:
		sock_hold(sk);
	}
	read_unlock(&hashinfo->lhash_lock);
	return sk;
}
#endif /* _INET_HASHTABLES_H */
Commit	Line	Data
304a1618 ACM	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
2d8c4ce5	17	#include <linux/interrupt.h>
77d8bf9c	18	#include <linux/ip.h>
33b62231	19	#include <linux/ipv6.h>
77d8bf9c ACM	20	#include <linux/list.h>
77d8bf9c ACM	21	#include <linux/slab.h>
33b62231	22	#include <linux/socket.h>
77d8bf9c	23	#include <linux/spinlock.h>
304a1618	24	#include <linux/types.h>
f3f05f70	25	#include <linux/wait.h>
304a1618	26
2d8c4ce5	27	#include <net/sock.h>
c752f073	28	#include <net/tcp_states.h>
2d8c4ce5	29
f3f05f70 ACM	30	#include <asm/atomic.h>
f3f05f70 ACM	31
77d8bf9c ACM	32	/* This is for all connections with a full identity, no wildcards.
	33	* New scheme, half the table is for TIME_WAIT, the other half is
	34	* for the rest. I'll experiment with dynamic table growth later.
	35	*/
	36	struct inet_ehash_bucket {
	37	rwlock_t lock;
	38	struct hlist_head chain;
	39	} __attribute__((__aligned__(8)));
	40
	41	/* There are a few simple rules, which allow for local port reuse by
	42	* an application. In essence:
	43	*
	44	* 1) Sockets bound to different interfaces may share a local port.
	45	* Failing that, goto test 2.
	46	* 2) If all sockets have sk->sk_reuse set, and none of them are in
	47	* TCP_LISTEN state, the port may be shared.
	48	* Failing that, goto test 3.
	49	* 3) If all sockets are bound to a specific inet_sk(sk)->rcv_saddr local
	50	* address, and none of them are the same, the port may be
	51	* shared.
	52	* Failing this, the port cannot be shared.
	53	*
	54	* The interesting point, is test #2. This is what an FTP server does
	55	* all day. To optimize this case we use a specific flag bit defined
	56	* below. As we add sockets to a bind bucket list, we perform a
	57	* check of: (newsk->sk_reuse && (newsk->sk_state != TCP_LISTEN))
	58	* As long as all sockets added to a bind bucket pass this test,
	59	* the flag bit will be set.
	60	* The resulting situation is that tcp_v[46]_verify_bind() can just check
	61	* for this flag bit, if it is set and the socket trying to bind has
	62	* sk->sk_reuse set, we don't even have to walk the owners list at all,
	63	* we return that it is ok to bind this socket to the requested local port.
	64	*
	65	* Sounds like a lot of work, but it is worth it. In a more naive
	66	* implementation (ie. current FreeBSD etc.) the entire list of ports
	67	* must be walked for each data port opened by an ftp server. Needless
	68	* to say, this does not scale at all. With a couple thousand FTP
	69	* users logged onto your box, isn't it nice to know that new data
	70	* ports are created in O(1) time? I thought so. ;-) -DaveM
	71	*/
	72	struct inet_bind_bucket {
	73	unsigned short port;
	74	signed short fastreuse;
	75	struct hlist_node node;
	76	struct hlist_head owners;
	77	};
	78
	79	#define inet_bind_bucket_for_each(tb, node, head) \
	80	hlist_for_each_entry(tb, node, head, node)
	81
	82	struct inet_bind_hashbucket {
	83	spinlock_t lock;
	84	struct hlist_head chain;
	85	};
	86
	87	/* This is for listening sockets, thus all sockets which possess wildcards. */
	88	#define INET_LHTABLE_SIZE 32 /* Yes, really, this is all you need. */
	89
	90	struct inet_hashinfo {
	91	/* This is for sockets with full identity only. Sockets here will
	92	* always be without wildcards and will have the following invariant:
	93	*
	94	* TCP_ESTABLISHED <= sk->sk_state < TCP_CLOSE
	95	*
96	* First half of the table is for sockets not in TIME_WAIT, second half
97	* is for TIME_WAIT sockets only.
98	*/
99	struct inet_ehash_bucket *ehash;
100
101	/* Ok, let's try this, I give up, we do need a local binding
102	* TCP hash as well as the others for fast bind/connect.
103	*/
104	struct inet_bind_hashbucket *bhash;
105
106	int bhash_size;
107	int ehash_size;
108
109	/* All sockets in TCP_LISTEN state will be in here. This is the only
110	* table where wildcard'd TCP sockets can exist. Hash function here
111	* is just local port number.
112	*/
113	struct hlist_head listening_hash[INET_LHTABLE_SIZE];
114
115	/* All the above members are written once at bootup and
116	* never written again _or_ are predominantly read-access.
117	*
118	* Now align to a new cache line as all the following members
119	* are often dirty.
120	*/
121	rwlock_t lhash_lock ____cacheline_aligned;
122	atomic_t lhash_users;
123	wait_queue_head_t lhash_wait;
124	spinlock_t portalloc_lock;
2d8c4ce5	125	kmem_cache_t *bind_bucket_cachep;
6e04e021	126	int port_rover;
77d8bf9c ACM	127	};
77d8bf9c ACM	128
304a1618 ACM	129	static inline int inet_ehashfn(const __u32 laddr, const __u16 lport,
	130	const __u32 faddr, const __u16 fport,
	131	const int ehash_size)
	132	{
	133	int h = (laddr ^ lport) ^ (faddr ^ fport);
	134	h ^= h >> 16;
	135	h ^= h >> 8;
	136	return h & (ehash_size - 1);
	137	}
	138
	139	static inline int inet_sk_ehashfn(const struct sock *sk, const int ehash_size)
	140	{
	141	const struct inet_sock *inet = inet_sk(sk);
	142	const __u32 laddr = inet->rcv_saddr;
	143	const __u16 lport = inet->num;
	144	const __u32 faddr = inet->daddr;
	145	const __u16 fport = inet->dport;
	146
	147	return inet_ehashfn(laddr, lport, faddr, fport, ehash_size);
	148	}
	149
77d8bf9c ACM	150	extern struct inet_bind_bucket *
	151	inet_bind_bucket_create(kmem_cache_t *cachep,
	152	struct inet_bind_hashbucket *head,
	153	const unsigned short snum);
	154	extern void inet_bind_bucket_destroy(kmem_cache_t *cachep,
	155	struct inet_bind_bucket *tb);
	156
	157	static inline int inet_bhashfn(const __u16 lport, const int bhash_size)
	158	{
	159	return lport & (bhash_size - 1);
	160	}
	161
2d8c4ce5 ACM	162	extern void inet_bind_hash(struct sock sk, struct inet_bind_bucket tb,
	163	const unsigned short snum);
	164
77d8bf9c ACM	165	/* These can have wildcards, don't try too hard. */
	166	static inline int inet_lhashfn(const unsigned short num)
	167	{
	168	return num & (INET_LHTABLE_SIZE - 1);
	169	}
	170
	171	static inline int inet_sk_listen_hashfn(const struct sock *sk)
	172	{
	173	return inet_lhashfn(inet_sk(sk)->num);
	174	}
	175
2d8c4ce5 ACM	176	/* Caller must disable local BH processing. */
	177	static inline void __inet_inherit_port(struct inet_hashinfo *table,
	178	struct sock sk, struct sock child)
	179	{
	180	const int bhash = inet_bhashfn(inet_sk(child)->num, table->bhash_size);
	181	struct inet_bind_hashbucket *head = &table->bhash[bhash];
	182	struct inet_bind_bucket *tb;
	183
	184	spin_lock(&head->lock);
	185	tb = inet_sk(sk)->bind_hash;
	186	sk_add_bind_node(child, &tb->owners);
	187	inet_sk(child)->bind_hash = tb;
	188	spin_unlock(&head->lock);
	189	}
	190
	191	static inline void inet_inherit_port(struct inet_hashinfo *table,
	192	struct sock sk, struct sock child)
	193	{
	194	local_bh_disable();
	195	__inet_inherit_port(table, sk, child);
	196	local_bh_enable();
	197	}
	198
	199	extern void inet_put_port(struct inet_hashinfo table, struct sock sk);
	200
f3f05f70 ACM	201	extern void inet_listen_wlock(struct inet_hashinfo *hashinfo);
	202
	203	/*
	204	* - We may sleep inside this lock.
	205	* - If sleeping is not required (or called from BH),
	206	* use plain read_(un)lock(&inet_hashinfo.lhash_lock).
	207	*/
	208	static inline void inet_listen_lock(struct inet_hashinfo *hashinfo)
	209	{
	210	/* read_lock synchronizes to candidates to writers */
	211	read_lock(&hashinfo->lhash_lock);
	212	atomic_inc(&hashinfo->lhash_users);
	213	read_unlock(&hashinfo->lhash_lock);
	214	}
	215
	216	static inline void inet_listen_unlock(struct inet_hashinfo *hashinfo)
	217	{
	218	if (atomic_dec_and_test(&hashinfo->lhash_users))
	219	wake_up(&hashinfo->lhash_wait);
	220	}
	221
	222	static inline void __inet_hash(struct inet_hashinfo *hashinfo,
	223	struct sock *sk, const int listen_possible)
	224	{
	225	struct hlist_head *list;
	226	rwlock_t *lock;
	227
	228	BUG_TRAP(sk_unhashed(sk));
	229	if (listen_possible && sk->sk_state == TCP_LISTEN) {
	230	list = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
	231	lock = &hashinfo->lhash_lock;
	232	inet_listen_wlock(hashinfo);
	233	} else {
	234	sk->sk_hashent = inet_sk_ehashfn(sk, hashinfo->ehash_size);
	235	list = &hashinfo->ehash[sk->sk_hashent].chain;
	236	lock = &hashinfo->ehash[sk->sk_hashent].lock;
	237	write_lock(lock);
	238	}
	239	__sk_add_node(sk, list);
	240	sock_prot_inc_use(sk->sk_prot);
	241	write_unlock(lock);
	242	if (listen_possible && sk->sk_state == TCP_LISTEN)
	243	wake_up(&hashinfo->lhash_wait);
	244	}
81849d10 ACM	245
	246	static inline void inet_hash(struct inet_hashinfo hashinfo, struct sock sk)
	247	{
	248	if (sk->sk_state != TCP_CLOSE) {
	249	local_bh_disable();
	250	__inet_hash(hashinfo, sk, 1);
	251	local_bh_enable();
	252	}
	253	}
	254
	255	static inline void inet_unhash(struct inet_hashinfo hashinfo, struct sock sk)
	256	{
	257	rwlock_t *lock;
	258
	259	if (sk_unhashed(sk))
	260	goto out;
	261
	262	if (sk->sk_state == TCP_LISTEN) {
	263	local_bh_disable();
	264	inet_listen_wlock(hashinfo);
	265	lock = &hashinfo->lhash_lock;
	266	} else {
	267	struct inet_ehash_bucket *head = &hashinfo->ehash[sk->sk_hashent];
	268	lock = &head->lock;
	269	write_lock_bh(&head->lock);
	270	}
	271
	272	if (__sk_del_node_init(sk))
	273	sock_prot_dec_use(sk->sk_prot);
	274	write_unlock_bh(lock);
	275	out:
	276	if (sk->sk_state == TCP_LISTEN)
	277	wake_up(&hashinfo->lhash_wait);
	278	}
33b62231 ACM	279
	280	extern struct sock __inet_lookup_listener(const struct hlist_head head,
	281	const u32 daddr,
	282	const unsigned short hnum,
	283	const int dif);
	284
	285	/* Optimize the common listener case. */
	286	static inline struct sock inet_lookup_listener(struct inet_hashinfo hashinfo,
	287	const u32 daddr,
	288	const unsigned short hnum,
	289	const int dif)
	290	{
	291	struct sock *sk = NULL;
	292	struct hlist_head *head;
	293
	294	read_lock(&hashinfo->lhash_lock);
	295	head = &hashinfo->listening_hash[inet_lhashfn(hnum)];
	296	if (!hlist_empty(head)) {
	297	const struct inet_sock *inet = inet_sk((sk = __sk_head(head)));
	298
	299	if (inet->num == hnum && !sk->sk_node.next &&
	300	(!inet->rcv_saddr \|\| inet->rcv_saddr == daddr) &&
	301	(sk->sk_family == PF_INET \|\| !ipv6_only_sock(sk)) &&
	302	!sk->sk_bound_dev_if)
	303	goto sherry_cache;
	304	sk = __inet_lookup_listener(head, daddr, hnum, dif);
	305	}
	306	if (sk) {
	307	sherry_cache:
	308	sock_hold(sk);
	309	}
	310	read_unlock(&hashinfo->lhash_lock);
	311	return sk;
	312	}
304a1618	313	#endif /* _INET_HASHTABLES_H */