[net-next-2.6.git] / include / net / inet_timewait_sock.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.
 *
 *		Definitions for a generic INET TIMEWAIT sock
 *
 *		From code originally in net/tcp.h
 *
 *		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_TIMEWAIT_SOCK_
#define _INET_TIMEWAIT_SOCK_


#include <linux/list.h>
#include <linux/module.h>
#include <linux/timer.h>
#include <linux/types.h>
#include <linux/workqueue.h>

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

#include <asm/atomic.h>

struct inet_hashinfo;

#define INET_TWDR_RECYCLE_SLOTS_LOG	5
#define INET_TWDR_RECYCLE_SLOTS		(1 << INET_TWDR_RECYCLE_SLOTS_LOG)

/*
 * If time > 4sec, it is "slow" path, no recycling is required,
 * so that we select tick to get range about 4 seconds.
 */
#if HZ <= 16 || HZ > 4096
# error Unsupported: HZ <= 16 or HZ > 4096
#elif HZ <= 32
# define INET_TWDR_RECYCLE_TICK (5 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#elif HZ <= 64
# define INET_TWDR_RECYCLE_TICK (6 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#elif HZ <= 128
# define INET_TWDR_RECYCLE_TICK (7 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#elif HZ <= 256
# define INET_TWDR_RECYCLE_TICK (8 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#elif HZ <= 512
# define INET_TWDR_RECYCLE_TICK (9 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#elif HZ <= 1024
# define INET_TWDR_RECYCLE_TICK (10 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#elif HZ <= 2048
# define INET_TWDR_RECYCLE_TICK (11 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#else
# define INET_TWDR_RECYCLE_TICK (12 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
#endif

/* TIME_WAIT reaping mechanism. */
#define INET_TWDR_TWKILL_SLOTS	8 /* Please keep this a power of 2. */

#define INET_TWDR_TWKILL_QUOTA 100

struct inet_timewait_death_row {
	/* Short-time timewait calendar */
	int			twcal_hand;
	unsigned long		twcal_jiffie;
	struct timer_list	twcal_timer;
	struct hlist_head	twcal_row[INET_TWDR_RECYCLE_SLOTS];

	spinlock_t		death_lock;
	int			tw_count;
	int			period;
	u32			thread_slots;
	struct work_struct	twkill_work;
	struct timer_list	tw_timer;
	int			slot;
	struct hlist_head	cells[INET_TWDR_TWKILL_SLOTS];
	struct inet_hashinfo 	*hashinfo;
	int			sysctl_tw_recycle;
	int			sysctl_max_tw_buckets;
};

extern void inet_twdr_hangman(unsigned long data);
extern void inet_twdr_twkill_work(struct work_struct *work);
extern void inet_twdr_twcal_tick(unsigned long data);

#if (BITS_PER_LONG == 64)
#define INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES 8
#else
#define INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES 4
#endif

struct inet_bind_bucket;

/*
 * This is a TIME_WAIT sock. It works around the memory consumption
 * problems of sockets in such a state on heavily loaded servers, but
 * without violating the protocol specification.
 */
struct inet_timewait_sock {
	/*
	 * Now struct sock also uses sock_common, so please just
	 * don't add nothing before this first member (__tw_common) --acme
	 */
	struct sock_common	__tw_common;
#define tw_family		__tw_common.skc_family
#define tw_state		__tw_common.skc_state
#define tw_reuse		__tw_common.skc_reuse
#define tw_bound_dev_if		__tw_common.skc_bound_dev_if
#define tw_node			__tw_common.skc_node
#define tw_bind_node		__tw_common.skc_bind_node
#define tw_refcnt		__tw_common.skc_refcnt
#define tw_hash			__tw_common.skc_hash
#define tw_prot			__tw_common.skc_prot
#define tw_net			__tw_common.skc_net
	volatile unsigned char	tw_substate;
	/* 3 bits hole, try to pack */
	unsigned char		tw_rcv_wscale;
	/* Socket demultiplex comparisons on incoming packets. */
	/* these five are in inet_sock */
	__be16			tw_sport;
	__be32			tw_daddr __attribute__((aligned(INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES)));
	__be32			tw_rcv_saddr;
	__be16			tw_dport;
	__u16			tw_num;
	/* And these are ours. */
	__u8			tw_ipv6only:1;
	/* 15 bits hole, try to pack */
	__u16			tw_ipv6_offset;
	int			tw_timeout;
	unsigned long		tw_ttd;
	struct inet_bind_bucket	*tw_tb;
	struct hlist_node	tw_death_node;
};

static inline void inet_twsk_add_node(struct inet_timewait_sock *tw,
				      struct hlist_head *list)
{
	hlist_add_head(&tw->tw_node, list);
}

static inline void inet_twsk_add_bind_node(struct inet_timewait_sock *tw,
					   struct hlist_head *list)
{
	hlist_add_head(&tw->tw_bind_node, list);
}

static inline int inet_twsk_dead_hashed(const struct inet_timewait_sock *tw)
{
	return !hlist_unhashed(&tw->tw_death_node);
}

static inline void inet_twsk_dead_node_init(struct inet_timewait_sock *tw)
{
	tw->tw_death_node.pprev = NULL;
}

static inline void __inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
{
	__hlist_del(&tw->tw_death_node);
	inet_twsk_dead_node_init(tw);
}

static inline int inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
{
	if (inet_twsk_dead_hashed(tw)) {
		__inet_twsk_del_dead_node(tw);
		return 1;
	}
	return 0;
}

#define inet_twsk_for_each(tw, node, head) \
	hlist_for_each_entry(tw, node, head, tw_node)

#define inet_twsk_for_each_inmate(tw, node, jail) \
	hlist_for_each_entry(tw, node, jail, tw_death_node)

#define inet_twsk_for_each_inmate_safe(tw, node, safe, jail) \
	hlist_for_each_entry_safe(tw, node, safe, jail, tw_death_node)

static inline struct inet_timewait_sock *inet_twsk(const struct sock *sk)
{
	return (struct inet_timewait_sock *)sk;
}

static inline __be32 inet_rcv_saddr(const struct sock *sk)
{
	return likely(sk->sk_state != TCP_TIME_WAIT) ?
		inet_sk(sk)->rcv_saddr : inet_twsk(sk)->tw_rcv_saddr;
}

static inline void inet_twsk_put(struct inet_timewait_sock *tw)
{
	if (atomic_dec_and_test(&tw->tw_refcnt)) {
		struct module *owner = tw->tw_prot->owner;
		twsk_destructor((struct sock *)tw);
#ifdef SOCK_REFCNT_DEBUG
		printk(KERN_DEBUG "%s timewait_sock %p released\n",
		       tw->tw_prot->name, tw);
#endif
		kmem_cache_free(tw->tw_prot->twsk_prot->twsk_slab, tw);
		module_put(owner);
	}
}

extern struct inet_timewait_sock *inet_twsk_alloc(const struct sock *sk,
						  const int state);

extern void __inet_twsk_hashdance(struct inet_timewait_sock *tw,
				  struct sock *sk,
				  struct inet_hashinfo *hashinfo);

extern void inet_twsk_schedule(struct inet_timewait_sock *tw,
			       struct inet_timewait_death_row *twdr,
			       const int timeo, const int timewait_len);
extern void inet_twsk_deschedule(struct inet_timewait_sock *tw,
				 struct inet_timewait_death_row *twdr);
#endif	/* _INET_TIMEWAIT_SOCK_ */
Commit	Line	Data
8feaf0c0 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	* Definitions for a generic INET TIMEWAIT sock
	7	*
	8	* From code originally in net/tcp.h
	9	*
	10	* This program is free software; you can redistribute it and/or
	11	* modify it under the terms of the GNU General Public License
	12	* as published by the Free Software Foundation; either version
	13	* 2 of the License, or (at your option) any later version.
	14	*/
	15	#ifndef _INET_TIMEWAIT_SOCK_
	16	#define _INET_TIMEWAIT_SOCK_
	17
8feaf0c0 ACM	18
8feaf0c0 ACM	19	#include <linux/list.h>
eeb2b856	20	#include <linux/module.h>
295ff7ed	21	#include <linux/timer.h>
8feaf0c0	22	#include <linux/types.h>
295ff7ed	23	#include <linux/workqueue.h>
8feaf0c0	24
14c85021	25	#include <net/inet_sock.h>
8feaf0c0 ACM	26	#include <net/sock.h>
8feaf0c0 ACM	27	#include <net/tcp_states.h>
6d6ee43e	28	#include <net/timewait_sock.h>
8feaf0c0 ACM	29
	30	#include <asm/atomic.h>
	31
295ff7ed ACM	32	struct inet_hashinfo;
	33
	34	#define INET_TWDR_RECYCLE_SLOTS_LOG 5
	35	#define INET_TWDR_RECYCLE_SLOTS (1 << INET_TWDR_RECYCLE_SLOTS_LOG)
	36
	37	/*
	38	* If time > 4sec, it is "slow" path, no recycling is required,
	39	* so that we select tick to get range about 4 seconds.
	40	*/
	41	#if HZ <= 16 \|\| HZ > 4096
	42	# error Unsupported: HZ <= 16 or HZ > 4096
	43	#elif HZ <= 32
	44	# define INET_TWDR_RECYCLE_TICK (5 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	45	#elif HZ <= 64
	46	# define INET_TWDR_RECYCLE_TICK (6 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	47	#elif HZ <= 128
	48	# define INET_TWDR_RECYCLE_TICK (7 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	49	#elif HZ <= 256
	50	# define INET_TWDR_RECYCLE_TICK (8 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	51	#elif HZ <= 512
	52	# define INET_TWDR_RECYCLE_TICK (9 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	53	#elif HZ <= 1024
	54	# define INET_TWDR_RECYCLE_TICK (10 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	55	#elif HZ <= 2048
	56	# define INET_TWDR_RECYCLE_TICK (11 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	57	#else
	58	# define INET_TWDR_RECYCLE_TICK (12 + 2 - INET_TWDR_RECYCLE_SLOTS_LOG)
	59	#endif
	60
	61	/* TIME_WAIT reaping mechanism. */
	62	#define INET_TWDR_TWKILL_SLOTS 8 /* Please keep this a power of 2. */
	63
	64	#define INET_TWDR_TWKILL_QUOTA 100
	65
	66	struct inet_timewait_death_row {
	67	/* Short-time timewait calendar */
	68	int twcal_hand;
187f5f84	69	unsigned long twcal_jiffie;
295ff7ed ACM	70	struct timer_list twcal_timer;
	71	struct hlist_head twcal_row[INET_TWDR_RECYCLE_SLOTS];
	72
	73	spinlock_t death_lock;
	74	int tw_count;
	75	int period;
	76	u32 thread_slots;
	77	struct work_struct twkill_work;
	78	struct timer_list tw_timer;
	79	int slot;
	80	struct hlist_head cells[INET_TWDR_TWKILL_SLOTS];
	81	struct inet_hashinfo *hashinfo;
	82	int sysctl_tw_recycle;
	83	int sysctl_max_tw_buckets;
	84	};
	85
696ab2d3	86	extern void inet_twdr_hangman(unsigned long data);
65f27f38	87	extern void inet_twdr_twkill_work(struct work_struct *work);
696ab2d3 ACM	88	extern void inet_twdr_twcal_tick(unsigned long data);
696ab2d3 ACM	89
8feaf0c0 ACM	90	#if (BITS_PER_LONG == 64)
	91	#define INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES 8
	92	#else
	93	#define INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES 4
	94	#endif
	95
	96	struct inet_bind_bucket;
	97
	98	/*
	99	* This is a TIME_WAIT sock. It works around the memory consumption
	100	* problems of sockets in such a state on heavily loaded servers, but
	101	* without violating the protocol specification.
	102	*/
	103	struct inet_timewait_sock {
	104	/*
	105	* Now struct sock also uses sock_common, so please just
	106	* don't add nothing before this first member (__tw_common) --acme
	107	*/
	108	struct sock_common __tw_common;
	109	#define tw_family __tw_common.skc_family
	110	#define tw_state __tw_common.skc_state
	111	#define tw_reuse __tw_common.skc_reuse
	112	#define tw_bound_dev_if __tw_common.skc_bound_dev_if
	113	#define tw_node __tw_common.skc_node
	114	#define tw_bind_node __tw_common.skc_bind_node
	115	#define tw_refcnt __tw_common.skc_refcnt
81c3d547	116	#define tw_hash __tw_common.skc_hash
8feaf0c0	117	#define tw_prot __tw_common.skc_prot
07feaebf	118	#define tw_net __tw_common.skc_net
8feaf0c0 ACM	119	volatile unsigned char tw_substate;
	120	/* 3 bits hole, try to pack */
	121	unsigned char tw_rcv_wscale;
	122	/* Socket demultiplex comparisons on incoming packets. */
	123	/* these five are in inet_sock */
23f33c2d AV	124	__be16 tw_sport;
	125	__be32 tw_daddr __attribute__((aligned(INET_TIMEWAIT_ADDRCMP_ALIGN_BYTES)));
	126	__be32 tw_rcv_saddr;
	127	__be16 tw_dport;
8feaf0c0 ACM	128	__u16 tw_num;
	129	/* And these are ours. */
	130	__u8 tw_ipv6only:1;
0fa1a53e ACM	131	/* 15 bits hole, try to pack */
0fa1a53e ACM	132	__u16 tw_ipv6_offset;
8feaf0c0 ACM	133	int tw_timeout;
	134	unsigned long tw_ttd;
	135	struct inet_bind_bucket *tw_tb;
	136	struct hlist_node tw_death_node;
	137	};
	138
	139	static inline void inet_twsk_add_node(struct inet_timewait_sock *tw,
	140	struct hlist_head *list)
	141	{
	142	hlist_add_head(&tw->tw_node, list);
	143	}
	144
	145	static inline void inet_twsk_add_bind_node(struct inet_timewait_sock *tw,
	146	struct hlist_head *list)
	147	{
	148	hlist_add_head(&tw->tw_bind_node, list);
	149	}
	150
	151	static inline int inet_twsk_dead_hashed(const struct inet_timewait_sock *tw)
	152	{
da753bea	153	return !hlist_unhashed(&tw->tw_death_node);
8feaf0c0 ACM	154	}
	155
	156	static inline void inet_twsk_dead_node_init(struct inet_timewait_sock *tw)
	157	{
	158	tw->tw_death_node.pprev = NULL;
	159	}
	160
	161	static inline void __inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
	162	{
	163	__hlist_del(&tw->tw_death_node);
	164	inet_twsk_dead_node_init(tw);
	165	}
	166
	167	static inline int inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
	168	{
	169	if (inet_twsk_dead_hashed(tw)) {
	170	__inet_twsk_del_dead_node(tw);
	171	return 1;
	172	}
	173	return 0;
	174	}
	175
	176	#define inet_twsk_for_each(tw, node, head) \
	177	hlist_for_each_entry(tw, node, head, tw_node)
	178
	179	#define inet_twsk_for_each_inmate(tw, node, jail) \
	180	hlist_for_each_entry(tw, node, jail, tw_death_node)
	181
	182	#define inet_twsk_for_each_inmate_safe(tw, node, safe, jail) \
	183	hlist_for_each_entry_safe(tw, node, safe, jail, tw_death_node)
	184
	185	static inline struct inet_timewait_sock inet_twsk(const struct sock sk)
	186	{
	187	return (struct inet_timewait_sock *)sk;
	188	}
	189
82103232	190	static inline __be32 inet_rcv_saddr(const struct sock *sk)
8feaf0c0 ACM	191	{
	192	return likely(sk->sk_state != TCP_TIME_WAIT) ?
	193	inet_sk(sk)->rcv_saddr : inet_twsk(sk)->tw_rcv_saddr;
	194	}
	195
	196	static inline void inet_twsk_put(struct inet_timewait_sock *tw)
	197	{
	198	if (atomic_dec_and_test(&tw->tw_refcnt)) {
eeb2b856	199	struct module *owner = tw->tw_prot->owner;
42b6785e	200	twsk_destructor((struct sock *)tw);
8feaf0c0 ACM	201	#ifdef SOCK_REFCNT_DEBUG
	202	printk(KERN_DEBUG "%s timewait_sock %p released\n",
	203	tw->tw_prot->name, tw);
	204	#endif
6d6ee43e	205	kmem_cache_free(tw->tw_prot->twsk_prot->twsk_slab, tw);
eeb2b856	206	module_put(owner);
8feaf0c0 ACM	207	}
8feaf0c0 ACM	208	}
e48c414e	209
c676270b ACM	210	extern struct inet_timewait_sock inet_twsk_alloc(const struct sock sk,
	211	const int state);
	212
e48c414e ACM	213	extern void __inet_twsk_hashdance(struct inet_timewait_sock *tw,
	214	struct sock *sk,
	215	struct inet_hashinfo *hashinfo);
696ab2d3 ACM	216
	217	extern void inet_twsk_schedule(struct inet_timewait_sock *tw,
	218	struct inet_timewait_death_row *twdr,
	219	const int timeo, const int timewait_len);
	220	extern void inet_twsk_deschedule(struct inet_timewait_sock *tw,
	221	struct inet_timewait_death_row *twdr);
8feaf0c0	222	#endif /* _INET_TIMEWAIT_SOCK_ */