[linux.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/kmemcheck.h>
#include <linux/list.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 <linux/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);

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_nulls_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
#define tw_daddr        	__tw_common.skc_daddr
#define tw_rcv_saddr    	__tw_common.skc_rcv_saddr
	int			tw_timeout;
	volatile unsigned char	tw_substate;
	unsigned char		tw_rcv_wscale;

	/* Socket demultiplex comparisons on incoming packets. */
	/* these three are in inet_sock */
	__be16			tw_sport;
	__be16			tw_dport;
	__u16			tw_num;
	kmemcheck_bitfield_begin(flags);
	/* And these are ours. */
	unsigned int		tw_ipv6only     : 1,
				tw_transparent  : 1,
				tw_pad		: 6,	/* 6 bits hole */
				tw_tos		: 8,
				tw_ipv6_offset  : 16;
	kmemcheck_bitfield_end(flags);
	unsigned long		tw_ttd;
	struct inet_bind_bucket	*tw_tb;
	struct hlist_node	tw_death_node;
};
#define tw_tclass tw_tos

static inline void inet_twsk_add_node_rcu(struct inet_timewait_sock *tw,
				      struct hlist_nulls_head *list)
{
	hlist_nulls_add_head_rcu(&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_nulls_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 sk_rcv_saddr(const struct sock *sk)
{
/* both inet_sk() and inet_twsk() store rcv_saddr in skc_rcv_saddr */
	return sk->__sk_common.skc_rcv_saddr;
}

extern void inet_twsk_put(struct inet_timewait_sock *tw);

extern int inet_twsk_unhash(struct inet_timewait_sock *tw);

extern int inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
				 struct inet_hashinfo *hashinfo);

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);

extern void inet_twsk_purge(struct inet_hashinfo *hashinfo,
			    struct inet_timewait_death_row *twdr, int family);

static inline
struct net *twsk_net(const struct inet_timewait_sock *twsk)
{
#ifdef CONFIG_NET_NS
	return rcu_dereference_raw(twsk->tw_net); /* protected by locking, */
						  /* reference counting, */
						  /* initialization, or RCU. */
#else
	return &init_net;
#endif
}

static inline
void twsk_net_set(struct inet_timewait_sock *twsk, struct net *net)
{
#ifdef CONFIG_NET_NS
	rcu_assign_pointer(twsk->tw_net, net);
#endif
}
#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	18
9e337b0f	19	#include <linux/kmemcheck.h>
8feaf0c0	20	#include <linux/list.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	29
60063497	30	#include <linux/atomic.h>
8feaf0c0	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	struct inet_bind_bucket;
	91
	92	/*
	93	* This is a TIME_WAIT sock. It works around the memory consumption
	94	* problems of sockets in such a state on heavily loaded servers, but
	95	* without violating the protocol specification.
	96	*/
	97	struct inet_timewait_sock {
	98	/*
	99	* Now struct sock also uses sock_common, so please just
	100	* don't add nothing before this first member (__tw_common) --acme
	101	*/
	102	struct sock_common __tw_common;
	103	#define tw_family __tw_common.skc_family
	104	#define tw_state __tw_common.skc_state
	105	#define tw_reuse __tw_common.skc_reuse
	106	#define tw_bound_dev_if __tw_common.skc_bound_dev_if
3ab5aee7	107	#define tw_node __tw_common.skc_nulls_node
8feaf0c0 ACM	108	#define tw_bind_node __tw_common.skc_bind_node
8feaf0c0 ACM	109	#define tw_refcnt __tw_common.skc_refcnt
81c3d547	110	#define tw_hash __tw_common.skc_hash
8feaf0c0	111	#define tw_prot __tw_common.skc_prot
07feaebf	112	#define tw_net __tw_common.skc_net
68835aba ED	113	#define tw_daddr __tw_common.skc_daddr
68835aba ED	114	#define tw_rcv_saddr __tw_common.skc_rcv_saddr
ad8bb780	115	int tw_timeout;
8feaf0c0	116	volatile unsigned char tw_substate;
8feaf0c0	117	unsigned char tw_rcv_wscale;
68835aba	118
8feaf0c0	119	/* Socket demultiplex comparisons on incoming packets. */
68835aba	120	/* these three are in inet_sock */
23f33c2d	121	__be16 tw_sport;
23f33c2d	122	__be16 tw_dport;
8feaf0c0	123	__u16 tw_num;
9e337b0f	124	kmemcheck_bitfield_begin(flags);
8feaf0c0	125	/* And these are ours. */
abf90cca ED	126	unsigned int tw_ipv6only : 1,
abf90cca ED	127	tw_transparent : 1,
66b13d99 ED	128	tw_pad : 6, /* 6 bits hole */
66b13d99 ED	129	tw_tos : 8,
abf90cca	130	tw_ipv6_offset : 16;
9e337b0f	131	kmemcheck_bitfield_end(flags);
8feaf0c0 ACM	132	unsigned long tw_ttd;
	133	struct inet_bind_bucket *tw_tb;
	134	struct hlist_node tw_death_node;
	135	};
b903d324	136	#define tw_tclass tw_tos
8feaf0c0	137
3ab5aee7 ED	138	static inline void inet_twsk_add_node_rcu(struct inet_timewait_sock *tw,
3ab5aee7 ED	139	struct hlist_nulls_head *list)
8feaf0c0	140	{
3ab5aee7	141	hlist_nulls_add_head_rcu(&tw->tw_node, list);
8feaf0c0 ACM	142	}
	143
	144	static inline void inet_twsk_add_bind_node(struct inet_timewait_sock *tw,
	145	struct hlist_head *list)
	146	{
	147	hlist_add_head(&tw->tw_bind_node, list);
	148	}
	149
	150	static inline int inet_twsk_dead_hashed(const struct inet_timewait_sock *tw)
	151	{
da753bea	152	return !hlist_unhashed(&tw->tw_death_node);
8feaf0c0 ACM	153	}
	154
	155	static inline void inet_twsk_dead_node_init(struct inet_timewait_sock *tw)
	156	{
	157	tw->tw_death_node.pprev = NULL;
	158	}
	159
	160	static inline void __inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
	161	{
	162	__hlist_del(&tw->tw_death_node);
	163	inet_twsk_dead_node_init(tw);
	164	}
	165
	166	static inline int inet_twsk_del_dead_node(struct inet_timewait_sock *tw)
	167	{
	168	if (inet_twsk_dead_hashed(tw)) {
	169	__inet_twsk_del_dead_node(tw);
	170	return 1;
	171	}
	172	return 0;
	173	}
	174
	175	#define inet_twsk_for_each(tw, node, head) \
3ab5aee7	176	hlist_nulls_for_each_entry(tw, node, head, tw_node)
8feaf0c0 ACM	177
	178	#define inet_twsk_for_each_inmate(tw, node, jail) \
	179	hlist_for_each_entry(tw, node, jail, tw_death_node)
	180
	181	#define inet_twsk_for_each_inmate_safe(tw, node, safe, jail) \
	182	hlist_for_each_entry_safe(tw, node, safe, jail, tw_death_node)
	183
	184	static inline struct inet_timewait_sock inet_twsk(const struct sock sk)
	185	{
	186	return (struct inet_timewait_sock *)sk;
	187	}
	188
68835aba	189	static inline __be32 sk_rcv_saddr(const struct sock *sk)
8feaf0c0	190	{
68835aba ED	191	/* both inet_sk() and inet_twsk() store rcv_saddr in skc_rcv_saddr */
68835aba ED	192	return sk->__sk_common.skc_rcv_saddr;
8feaf0c0 ACM	193	}
8feaf0c0 ACM	194
7054fb93	195	extern void inet_twsk_put(struct inet_timewait_sock *tw);
e48c414e	196
13475a30 ED	197	extern int inet_twsk_unhash(struct inet_timewait_sock *tw);
13475a30 ED	198
3cdaedae ED	199	extern int inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
	200	struct inet_hashinfo *hashinfo);
	201
c676270b ACM	202	extern struct inet_timewait_sock inet_twsk_alloc(const struct sock sk,
	203	const int state);
	204
e48c414e ACM	205	extern void __inet_twsk_hashdance(struct inet_timewait_sock *tw,
	206	struct sock *sk,
	207	struct inet_hashinfo *hashinfo);
696ab2d3 ACM	208
	209	extern void inet_twsk_schedule(struct inet_timewait_sock *tw,
	210	struct inet_timewait_death_row *twdr,
	211	const int timeo, const int timewait_len);
	212	extern void inet_twsk_deschedule(struct inet_timewait_sock *tw,
	213	struct inet_timewait_death_row *twdr);
3b1e0a65	214
b099ce26	215	extern void inet_twsk_purge(struct inet_hashinfo *hashinfo,
d315492b DL	216	struct inet_timewait_death_row *twdr, int family);
d315492b DL	217
3b1e0a65 YH	218	static inline
	219	struct net twsk_net(const struct inet_timewait_sock twsk)
	220	{
	221	#ifdef CONFIG_NET_NS
7ec75c58 PM	222	return rcu_dereference_raw(twsk->tw_net); /* protected by locking, */
	223	/* reference counting, */
	224	/* initialization, or RCU. */
3b1e0a65 YH	225	#else
	226	return &init_net;
	227	#endif
	228	}
	229
	230	static inline
f5aa23fd	231	void twsk_net_set(struct inet_timewait_sock twsk, struct net net)
3b1e0a65 YH	232	{
3b1e0a65 YH	233	#ifdef CONFIG_NET_NS
b099ce26	234	rcu_assign_pointer(twsk->tw_net, net);
3b1e0a65 YH	235	#endif
3b1e0a65 YH	236	}
8feaf0c0	237	#endif /* _INET_TIMEWAIT_SOCK_ */