[deliverable/linux.git] / net / ipv4 / inet_timewait_sock.c

/*
 * 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.
 *
 *		Generic TIME_WAIT sockets functions
 *
 *		From code orinally in TCP
 */

#include <linux/kernel.h>
#include <net/inet_hashtables.h>
#include <net/inet_timewait_sock.h>
#include <net/ip.h>

/* Must be called with locally disabled BHs. */
static void __inet_twsk_kill(struct inet_timewait_sock *tw,
			     struct inet_hashinfo *hashinfo)
{
	struct inet_bind_hashbucket *bhead;
	struct inet_bind_bucket *tb;
	/* Unlink from established hashes. */
	rwlock_t *lock = inet_ehash_lockp(hashinfo, tw->tw_hash);

	write_lock(lock);
	if (hlist_unhashed(&tw->tw_node)) {
		write_unlock(lock);
		return;
	}
	__hlist_del(&tw->tw_node);
	sk_node_init(&tw->tw_node);
	write_unlock(lock);

	/* Disassociate with bind bucket. */
	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), tw->tw_num,
			hashinfo->bhash_size)];
	spin_lock(&bhead->lock);
	tb = tw->tw_tb;
	__hlist_del(&tw->tw_bind_node);
	tw->tw_tb = NULL;
	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	spin_unlock(&bhead->lock);
#ifdef SOCK_REFCNT_DEBUG
	if (atomic_read(&tw->tw_refcnt) != 1) {
		printk(KERN_DEBUG "%s timewait_sock %p refcnt=%d\n",
		       tw->tw_prot->name, tw, atomic_read(&tw->tw_refcnt));
	}
#endif
	inet_twsk_put(tw);
}

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
		release_net(twsk_net(tw));
		kmem_cache_free(tw->tw_prot->twsk_prot->twsk_slab, tw);
		module_put(owner);
	}
}
EXPORT_SYMBOL_GPL(inet_twsk_put);

/*
 * Enter the time wait state. This is called with locally disabled BH.
 * Essentially we whip up a timewait bucket, copy the relevant info into it
 * from the SK, and mess with hash chains and list linkage.
 */
void __inet_twsk_hashdance(struct inet_timewait_sock *tw, struct sock *sk,
			   struct inet_hashinfo *hashinfo)
{
	const struct inet_sock *inet = inet_sk(sk);
	const struct inet_connection_sock *icsk = inet_csk(sk);
	struct inet_ehash_bucket *ehead = inet_ehash_bucket(hashinfo, sk->sk_hash);
	rwlock_t *lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
	struct inet_bind_hashbucket *bhead;
	/* Step 1: Put TW into bind hash. Original socket stays there too.
	   Note, that any socket with inet->num != 0 MUST be bound in
	   binding cache, even if it is closed.
	 */
	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), inet->num,
			hashinfo->bhash_size)];
	spin_lock(&bhead->lock);
	tw->tw_tb = icsk->icsk_bind_hash;
	BUG_TRAP(icsk->icsk_bind_hash);
	inet_twsk_add_bind_node(tw, &tw->tw_tb->owners);
	spin_unlock(&bhead->lock);

	write_lock(lock);

	/* Step 2: Remove SK from established hash. */
	if (__sk_del_node_init(sk))
		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);

	/* Step 3: Hash TW into TIMEWAIT chain. */
	inet_twsk_add_node(tw, &ehead->twchain);
	atomic_inc(&tw->tw_refcnt);

	write_unlock(lock);
}

EXPORT_SYMBOL_GPL(__inet_twsk_hashdance);

struct inet_timewait_sock *inet_twsk_alloc(const struct sock *sk, const int state)
{
	struct inet_timewait_sock *tw =
		kmem_cache_alloc(sk->sk_prot_creator->twsk_prot->twsk_slab,
				 GFP_ATOMIC);
	if (tw != NULL) {
		const struct inet_sock *inet = inet_sk(sk);

		/* Give us an identity. */
		tw->tw_daddr	    = inet->daddr;
		tw->tw_rcv_saddr    = inet->rcv_saddr;
		tw->tw_bound_dev_if = sk->sk_bound_dev_if;
		tw->tw_num	    = inet->num;
		tw->tw_state	    = TCP_TIME_WAIT;
		tw->tw_substate	    = state;
		tw->tw_sport	    = inet->sport;
		tw->tw_dport	    = inet->dport;
		tw->tw_family	    = sk->sk_family;
		tw->tw_reuse	    = sk->sk_reuse;
		tw->tw_hash	    = sk->sk_hash;
		tw->tw_ipv6only	    = 0;
		tw->tw_prot	    = sk->sk_prot_creator;
		twsk_net_set(tw, hold_net(sock_net(sk)));
		atomic_set(&tw->tw_refcnt, 1);
		inet_twsk_dead_node_init(tw);
		__module_get(tw->tw_prot->owner);
	}

	return tw;
}

EXPORT_SYMBOL_GPL(inet_twsk_alloc);

/* Returns non-zero if quota exceeded.  */
static int inet_twdr_do_twkill_work(struct inet_timewait_death_row *twdr,
				    const int slot)
{
	struct inet_timewait_sock *tw;
	struct hlist_node *node;
	unsigned int killed;
	int ret;

	/* NOTE: compare this to previous version where lock
	 * was released after detaching chain. It was racy,
	 * because tw buckets are scheduled in not serialized context
	 * in 2.3 (with netfilter), and with softnet it is common, because
	 * soft irqs are not sequenced.
	 */
	killed = 0;
	ret = 0;
rescan:
	inet_twsk_for_each_inmate(tw, node, &twdr->cells[slot]) {
		__inet_twsk_del_dead_node(tw);
		spin_unlock(&twdr->death_lock);
		__inet_twsk_kill(tw, twdr->hashinfo);
		inet_twsk_put(tw);
		killed++;
		spin_lock(&twdr->death_lock);
		if (killed > INET_TWDR_TWKILL_QUOTA) {
			ret = 1;
			break;
		}

		/* While we dropped twdr->death_lock, another cpu may have
		 * killed off the next TW bucket in the list, therefore
		 * do a fresh re-read of the hlist head node with the
		 * lock reacquired.  We still use the hlist traversal
		 * macro in order to get the prefetches.
		 */
		goto rescan;
	}

	twdr->tw_count -= killed;
	NET_ADD_STATS_BH(LINUX_MIB_TIMEWAITED, killed);

	return ret;
}

void inet_twdr_hangman(unsigned long data)
{
	struct inet_timewait_death_row *twdr;
	int unsigned need_timer;

	twdr = (struct inet_timewait_death_row *)data;
	spin_lock(&twdr->death_lock);

	if (twdr->tw_count == 0)
		goto out;

	need_timer = 0;
	if (inet_twdr_do_twkill_work(twdr, twdr->slot)) {
		twdr->thread_slots |= (1 << twdr->slot);
		schedule_work(&twdr->twkill_work);
		need_timer = 1;
	} else {
		/* We purged the entire slot, anything left?  */
		if (twdr->tw_count)
			need_timer = 1;
	}
	twdr->slot = ((twdr->slot + 1) & (INET_TWDR_TWKILL_SLOTS - 1));
	if (need_timer)
		mod_timer(&twdr->tw_timer, jiffies + twdr->period);
out:
	spin_unlock(&twdr->death_lock);
}

EXPORT_SYMBOL_GPL(inet_twdr_hangman);

void inet_twdr_twkill_work(struct work_struct *work)
{
	struct inet_timewait_death_row *twdr =
		container_of(work, struct inet_timewait_death_row, twkill_work);
	int i;

	BUILD_BUG_ON((INET_TWDR_TWKILL_SLOTS - 1) >
			(sizeof(twdr->thread_slots) * 8));

	while (twdr->thread_slots) {
		spin_lock_bh(&twdr->death_lock);
		for (i = 0; i < INET_TWDR_TWKILL_SLOTS; i++) {
			if (!(twdr->thread_slots & (1 << i)))
				continue;

			while (inet_twdr_do_twkill_work(twdr, i) != 0) {
				if (need_resched()) {
					spin_unlock_bh(&twdr->death_lock);
					schedule();
					spin_lock_bh(&twdr->death_lock);
				}
			}

			twdr->thread_slots &= ~(1 << i);
		}
		spin_unlock_bh(&twdr->death_lock);
	}
}

EXPORT_SYMBOL_GPL(inet_twdr_twkill_work);

/* These are always called from BH context.  See callers in
 * tcp_input.c to verify this.
 */

/* This is for handling early-kills of TIME_WAIT sockets. */
void inet_twsk_deschedule(struct inet_timewait_sock *tw,
			  struct inet_timewait_death_row *twdr)
{
	spin_lock(&twdr->death_lock);
	if (inet_twsk_del_dead_node(tw)) {
		inet_twsk_put(tw);
		if (--twdr->tw_count == 0)
			del_timer(&twdr->tw_timer);
	}
	spin_unlock(&twdr->death_lock);
	__inet_twsk_kill(tw, twdr->hashinfo);
}

EXPORT_SYMBOL(inet_twsk_deschedule);

void inet_twsk_schedule(struct inet_timewait_sock *tw,
		       struct inet_timewait_death_row *twdr,
		       const int timeo, const int timewait_len)
{
	struct hlist_head *list;
	int slot;

	/* timeout := RTO * 3.5
	 *
	 * 3.5 = 1+2+0.5 to wait for two retransmits.
	 *
	 * RATIONALE: if FIN arrived and we entered TIME-WAIT state,
	 * our ACK acking that FIN can be lost. If N subsequent retransmitted
	 * FINs (or previous seqments) are lost (probability of such event
	 * is p^(N+1), where p is probability to lose single packet and
	 * time to detect the loss is about RTO*(2^N - 1) with exponential
	 * backoff). Normal timewait length is calculated so, that we
	 * waited at least for one retransmitted FIN (maximal RTO is 120sec).
	 * [ BTW Linux. following BSD, violates this requirement waiting
	 *   only for 60sec, we should wait at least for 240 secs.
	 *   Well, 240 consumes too much of resources 8)
	 * ]
	 * This interval is not reduced to catch old duplicate and
	 * responces to our wandering segments living for two MSLs.
	 * However, if we use PAWS to detect
	 * old duplicates, we can reduce the interval to bounds required
	 * by RTO, rather than MSL. So, if peer understands PAWS, we
	 * kill tw bucket after 3.5*RTO (it is important that this number
	 * is greater than TS tick!) and detect old duplicates with help
	 * of PAWS.
	 */
	slot = (timeo + (1 << INET_TWDR_RECYCLE_TICK) - 1) >> INET_TWDR_RECYCLE_TICK;

	spin_lock(&twdr->death_lock);

	/* Unlink it, if it was scheduled */
	if (inet_twsk_del_dead_node(tw))
		twdr->tw_count--;
	else
		atomic_inc(&tw->tw_refcnt);

	if (slot >= INET_TWDR_RECYCLE_SLOTS) {
		/* Schedule to slow timer */
		if (timeo >= timewait_len) {
			slot = INET_TWDR_TWKILL_SLOTS - 1;
		} else {
			slot = DIV_ROUND_UP(timeo, twdr->period);
			if (slot >= INET_TWDR_TWKILL_SLOTS)
				slot = INET_TWDR_TWKILL_SLOTS - 1;
		}
		tw->tw_ttd = jiffies + timeo;
		slot = (twdr->slot + slot) & (INET_TWDR_TWKILL_SLOTS - 1);
		list = &twdr->cells[slot];
	} else {
		tw->tw_ttd = jiffies + (slot << INET_TWDR_RECYCLE_TICK);

		if (twdr->twcal_hand < 0) {
			twdr->twcal_hand = 0;
			twdr->twcal_jiffie = jiffies;
			twdr->twcal_timer.expires = twdr->twcal_jiffie +
					      (slot << INET_TWDR_RECYCLE_TICK);
			add_timer(&twdr->twcal_timer);
		} else {
			if (time_after(twdr->twcal_timer.expires,
				       jiffies + (slot << INET_TWDR_RECYCLE_TICK)))
				mod_timer(&twdr->twcal_timer,
					  jiffies + (slot << INET_TWDR_RECYCLE_TICK));
			slot = (twdr->twcal_hand + slot) & (INET_TWDR_RECYCLE_SLOTS - 1);
		}
		list = &twdr->twcal_row[slot];
	}

	hlist_add_head(&tw->tw_death_node, list);

	if (twdr->tw_count++ == 0)
		mod_timer(&twdr->tw_timer, jiffies + twdr->period);
	spin_unlock(&twdr->death_lock);
}

EXPORT_SYMBOL_GPL(inet_twsk_schedule);

void inet_twdr_twcal_tick(unsigned long data)
{
	struct inet_timewait_death_row *twdr;
	int n, slot;
	unsigned long j;
	unsigned long now = jiffies;
	int killed = 0;
	int adv = 0;

	twdr = (struct inet_timewait_death_row *)data;

	spin_lock(&twdr->death_lock);
	if (twdr->twcal_hand < 0)
		goto out;

	slot = twdr->twcal_hand;
	j = twdr->twcal_jiffie;

	for (n = 0; n < INET_TWDR_RECYCLE_SLOTS; n++) {
		if (time_before_eq(j, now)) {
			struct hlist_node *node, *safe;
			struct inet_timewait_sock *tw;

			inet_twsk_for_each_inmate_safe(tw, node, safe,
						       &twdr->twcal_row[slot]) {
				__inet_twsk_del_dead_node(tw);
				__inet_twsk_kill(tw, twdr->hashinfo);
				inet_twsk_put(tw);
				killed++;
			}
		} else {
			if (!adv) {
				adv = 1;
				twdr->twcal_jiffie = j;
				twdr->twcal_hand = slot;
			}

			if (!hlist_empty(&twdr->twcal_row[slot])) {
				mod_timer(&twdr->twcal_timer, j);
				goto out;
			}
		}
		j += 1 << INET_TWDR_RECYCLE_TICK;
		slot = (slot + 1) & (INET_TWDR_RECYCLE_SLOTS - 1);
	}
	twdr->twcal_hand = -1;

out:
	if ((twdr->tw_count -= killed) == 0)
		del_timer(&twdr->tw_timer);
	NET_ADD_STATS_BH(LINUX_MIB_TIMEWAITKILLED, killed);
	spin_unlock(&twdr->death_lock);
}

EXPORT_SYMBOL_GPL(inet_twdr_twcal_tick);
Commit	Line	Data
e48c414e 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	* Generic TIME_WAIT sockets functions
	7	*
	8	* From code orinally in TCP
	9	*/
	10
172589cc	11	#include <linux/kernel.h>
e48c414e ACM	12	#include <net/inet_hashtables.h>
e48c414e ACM	13	#include <net/inet_timewait_sock.h>
696ab2d3	14	#include <net/ip.h>
e48c414e ACM	15
e48c414e ACM	16	/* Must be called with locally disabled BHs. */
acd159b6 AB	17	static void __inet_twsk_kill(struct inet_timewait_sock *tw,
acd159b6 AB	18	struct inet_hashinfo *hashinfo)
e48c414e ACM	19	{
	20	struct inet_bind_hashbucket *bhead;
	21	struct inet_bind_bucket *tb;
	22	/* Unlink from established hashes. */
230140cf	23	rwlock_t *lock = inet_ehash_lockp(hashinfo, tw->tw_hash);
e48c414e	24
230140cf	25	write_lock(lock);
e48c414e	26	if (hlist_unhashed(&tw->tw_node)) {
230140cf	27	write_unlock(lock);
e48c414e ACM	28	return;
	29	}
	30	__hlist_del(&tw->tw_node);
	31	sk_node_init(&tw->tw_node);
230140cf	32	write_unlock(lock);
e48c414e ACM	33
e48c414e ACM	34	/* Disassociate with bind bucket. */
7f635ab7 PE	35	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), tw->tw_num,
7f635ab7 PE	36	hashinfo->bhash_size)];
e48c414e ACM	37	spin_lock(&bhead->lock);
	38	tb = tw->tw_tb;
	39	__hlist_del(&tw->tw_bind_node);
	40	tw->tw_tb = NULL;
	41	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	42	spin_unlock(&bhead->lock);
	43	#ifdef SOCK_REFCNT_DEBUG
	44	if (atomic_read(&tw->tw_refcnt) != 1) {
	45	printk(KERN_DEBUG "%s timewait_sock %p refcnt=%d\n",
	46	tw->tw_prot->name, tw, atomic_read(&tw->tw_refcnt));
	47	}
	48	#endif
	49	inet_twsk_put(tw);
	50	}
	51
7054fb93 PE	52	void inet_twsk_put(struct inet_timewait_sock *tw)
	53	{
	54	if (atomic_dec_and_test(&tw->tw_refcnt)) {
	55	struct module *owner = tw->tw_prot->owner;
	56	twsk_destructor((struct sock *)tw);
	57	#ifdef SOCK_REFCNT_DEBUG
	58	printk(KERN_DEBUG "%s timewait_sock %p released\n",
	59	tw->tw_prot->name, tw);
	60	#endif
cd5342d9	61	release_net(twsk_net(tw));
7054fb93 PE	62	kmem_cache_free(tw->tw_prot->twsk_prot->twsk_slab, tw);
	63	module_put(owner);
	64	}
	65	}
	66	EXPORT_SYMBOL_GPL(inet_twsk_put);
	67
e48c414e ACM	68	/*
	69	* Enter the time wait state. This is called with locally disabled BH.
	70	* Essentially we whip up a timewait bucket, copy the relevant info into it
	71	* from the SK, and mess with hash chains and list linkage.
	72	*/
	73	void __inet_twsk_hashdance(struct inet_timewait_sock tw, struct sock sk,
	74	struct inet_hashinfo *hashinfo)
	75	{
	76	const struct inet_sock *inet = inet_sk(sk);
463c84b9	77	const struct inet_connection_sock *icsk = inet_csk(sk);
81c3d547	78	struct inet_ehash_bucket *ehead = inet_ehash_bucket(hashinfo, sk->sk_hash);
230140cf	79	rwlock_t *lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
e48c414e ACM	80	struct inet_bind_hashbucket *bhead;
	81	/* Step 1: Put TW into bind hash. Original socket stays there too.
	82	Note, that any socket with inet->num != 0 MUST be bound in
	83	binding cache, even if it is closed.
	84	*/
7f635ab7 PE	85	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), inet->num,
7f635ab7 PE	86	hashinfo->bhash_size)];
e48c414e	87	spin_lock(&bhead->lock);
463c84b9 ACM	88	tw->tw_tb = icsk->icsk_bind_hash;
463c84b9 ACM	89	BUG_TRAP(icsk->icsk_bind_hash);
e48c414e ACM	90	inet_twsk_add_bind_node(tw, &tw->tw_tb->owners);
	91	spin_unlock(&bhead->lock);
	92
230140cf	93	write_lock(lock);
e48c414e ACM	94
	95	/* Step 2: Remove SK from established hash. */
	96	if (__sk_del_node_init(sk))
c29a0bc4	97	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
e48c414e	98
dbca9b27 ED	99	/* Step 3: Hash TW into TIMEWAIT chain. */
dbca9b27 ED	100	inet_twsk_add_node(tw, &ehead->twchain);
e48c414e ACM	101	atomic_inc(&tw->tw_refcnt);
e48c414e ACM	102
230140cf	103	write_unlock(lock);
e48c414e	104	}
c676270b	105
696ab2d3 ACM	106	EXPORT_SYMBOL_GPL(__inet_twsk_hashdance);
696ab2d3 ACM	107
c676270b ACM	108	struct inet_timewait_sock inet_twsk_alloc(const struct sock sk, const int state)
c676270b ACM	109	{
6d6ee43e ACM	110	struct inet_timewait_sock *tw =
6d6ee43e ACM	111	kmem_cache_alloc(sk->sk_prot_creator->twsk_prot->twsk_slab,
54e6ecb2	112	GFP_ATOMIC);
c676270b ACM	113	if (tw != NULL) {
	114	const struct inet_sock *inet = inet_sk(sk);
	115
	116	/* Give us an identity. */
	117	tw->tw_daddr = inet->daddr;
	118	tw->tw_rcv_saddr = inet->rcv_saddr;
	119	tw->tw_bound_dev_if = sk->sk_bound_dev_if;
	120	tw->tw_num = inet->num;
	121	tw->tw_state = TCP_TIME_WAIT;
	122	tw->tw_substate = state;
	123	tw->tw_sport = inet->sport;
	124	tw->tw_dport = inet->dport;
	125	tw->tw_family = sk->sk_family;
	126	tw->tw_reuse = sk->sk_reuse;
81c3d547	127	tw->tw_hash = sk->sk_hash;
c676270b ACM	128	tw->tw_ipv6only = 0;
c676270b ACM	129	tw->tw_prot = sk->sk_prot_creator;
cd5342d9	130	twsk_net_set(tw, hold_net(sock_net(sk)));
c676270b ACM	131	atomic_set(&tw->tw_refcnt, 1);
c676270b ACM	132	inet_twsk_dead_node_init(tw);
eeb2b856	133	__module_get(tw->tw_prot->owner);
c676270b ACM	134	}
	135
	136	return tw;
	137	}
696ab2d3 ACM	138
	139	EXPORT_SYMBOL_GPL(inet_twsk_alloc);
	140
	141	/* Returns non-zero if quota exceeded. */
	142	static int inet_twdr_do_twkill_work(struct inet_timewait_death_row *twdr,
	143	const int slot)
	144	{
	145	struct inet_timewait_sock *tw;
	146	struct hlist_node *node;
	147	unsigned int killed;
	148	int ret;
	149
	150	/* NOTE: compare this to previous version where lock
	151	* was released after detaching chain. It was racy,
	152	* because tw buckets are scheduled in not serialized context
	153	* in 2.3 (with netfilter), and with softnet it is common, because
	154	* soft irqs are not sequenced.
	155	*/
	156	killed = 0;
	157	ret = 0;
	158	rescan:
	159	inet_twsk_for_each_inmate(tw, node, &twdr->cells[slot]) {
	160	__inet_twsk_del_dead_node(tw);
	161	spin_unlock(&twdr->death_lock);
	162	__inet_twsk_kill(tw, twdr->hashinfo);
	163	inet_twsk_put(tw);
	164	killed++;
	165	spin_lock(&twdr->death_lock);
	166	if (killed > INET_TWDR_TWKILL_QUOTA) {
	167	ret = 1;
	168	break;
	169	}
	170
	171	/* While we dropped twdr->death_lock, another cpu may have
	172	* killed off the next TW bucket in the list, therefore
	173	* do a fresh re-read of the hlist head node with the
	174	* lock reacquired. We still use the hlist traversal
	175	* macro in order to get the prefetches.
	176	*/
	177	goto rescan;
	178	}
	179
	180	twdr->tw_count -= killed;
	181	NET_ADD_STATS_BH(LINUX_MIB_TIMEWAITED, killed);
	182
	183	return ret;
	184	}
	185
	186	void inet_twdr_hangman(unsigned long data)
	187	{
	188	struct inet_timewait_death_row *twdr;
	189	int unsigned need_timer;
	190
	191	twdr = (struct inet_timewait_death_row *)data;
	192	spin_lock(&twdr->death_lock);
	193
	194	if (twdr->tw_count == 0)
	195	goto out;
	196
	197	need_timer = 0;
	198	if (inet_twdr_do_twkill_work(twdr, twdr->slot)) {
	199	twdr->thread_slots \|= (1 << twdr->slot);
696ab2d3 ACM	200	schedule_work(&twdr->twkill_work);
	201	need_timer = 1;
	202	} else {
	203	/* We purged the entire slot, anything left? */
	204	if (twdr->tw_count)
	205	need_timer = 1;
	206	}
	207	twdr->slot = ((twdr->slot + 1) & (INET_TWDR_TWKILL_SLOTS - 1));
	208	if (need_timer)
	209	mod_timer(&twdr->tw_timer, jiffies + twdr->period);
	210	out:
	211	spin_unlock(&twdr->death_lock);
	212	}
	213
	214	EXPORT_SYMBOL_GPL(inet_twdr_hangman);
	215
65f27f38	216	void inet_twdr_twkill_work(struct work_struct *work)
696ab2d3	217	{
65f27f38 DH	218	struct inet_timewait_death_row *twdr =
65f27f38 DH	219	container_of(work, struct inet_timewait_death_row, twkill_work);
696ab2d3 ACM	220	int i;
696ab2d3 ACM	221
95c9382a PE	222	BUILD_BUG_ON((INET_TWDR_TWKILL_SLOTS - 1) >
95c9382a PE	223	(sizeof(twdr->thread_slots) * 8));
696ab2d3 ACM	224
	225	while (twdr->thread_slots) {
	226	spin_lock_bh(&twdr->death_lock);
	227	for (i = 0; i < INET_TWDR_TWKILL_SLOTS; i++) {
	228	if (!(twdr->thread_slots & (1 << i)))
	229	continue;
	230
	231	while (inet_twdr_do_twkill_work(twdr, i) != 0) {
	232	if (need_resched()) {
	233	spin_unlock_bh(&twdr->death_lock);
	234	schedule();
	235	spin_lock_bh(&twdr->death_lock);
	236	}
	237	}
	238
	239	twdr->thread_slots &= ~(1 << i);
	240	}
	241	spin_unlock_bh(&twdr->death_lock);
	242	}
	243	}
	244
	245	EXPORT_SYMBOL_GPL(inet_twdr_twkill_work);
	246
	247	/* These are always called from BH context. See callers in
	248	* tcp_input.c to verify this.
	249	*/
	250
	251	/* This is for handling early-kills of TIME_WAIT sockets. */
	252	void inet_twsk_deschedule(struct inet_timewait_sock *tw,
	253	struct inet_timewait_death_row *twdr)
	254	{
	255	spin_lock(&twdr->death_lock);
	256	if (inet_twsk_del_dead_node(tw)) {
	257	inet_twsk_put(tw);
	258	if (--twdr->tw_count == 0)
	259	del_timer(&twdr->tw_timer);
	260	}
	261	spin_unlock(&twdr->death_lock);
	262	__inet_twsk_kill(tw, twdr->hashinfo);
	263	}
	264
	265	EXPORT_SYMBOL(inet_twsk_deschedule);
	266
	267	void inet_twsk_schedule(struct inet_timewait_sock *tw,
	268	struct inet_timewait_death_row *twdr,
	269	const int timeo, const int timewait_len)
	270	{
	271	struct hlist_head *list;
	272	int slot;
	273
	274	/* timeout := RTO * 3.5
	275	*
	276	* 3.5 = 1+2+0.5 to wait for two retransmits.
	277	*
	278	* RATIONALE: if FIN arrived and we entered TIME-WAIT state,
	279	* our ACK acking that FIN can be lost. If N subsequent retransmitted
	280	* FINs (or previous seqments) are lost (probability of such event
	281	* is p^(N+1), where p is probability to lose single packet and
	282	* time to detect the loss is about RTO*(2^N - 1) with exponential
	283	* backoff). Normal timewait length is calculated so, that we
	284	* waited at least for one retransmitted FIN (maximal RTO is 120sec).
	285	* [ BTW Linux. following BSD, violates this requirement waiting
	286	* only for 60sec, we should wait at least for 240 secs.
	287	* Well, 240 consumes too much of resources 8)
288	* ]
289	* This interval is not reduced to catch old duplicate and
290	* responces to our wandering segments living for two MSLs.
291	* However, if we use PAWS to detect
292	* old duplicates, we can reduce the interval to bounds required
293	* by RTO, rather than MSL. So, if peer understands PAWS, we
294	* kill tw bucket after 3.5*RTO (it is important that this number
295	* is greater than TS tick!) and detect old duplicates with help
296	* of PAWS.
297	*/
298	slot = (timeo + (1 << INET_TWDR_RECYCLE_TICK) - 1) >> INET_TWDR_RECYCLE_TICK;
299
300	spin_lock(&twdr->death_lock);
301
302	/* Unlink it, if it was scheduled */
303	if (inet_twsk_del_dead_node(tw))
304	twdr->tw_count--;
305	else
306	atomic_inc(&tw->tw_refcnt);
307
308	if (slot >= INET_TWDR_RECYCLE_SLOTS) {
309	/* Schedule to slow timer */
310	if (timeo >= timewait_len) {
311	slot = INET_TWDR_TWKILL_SLOTS - 1;
312	} else {
172589cc	313	slot = DIV_ROUND_UP(timeo, twdr->period);
696ab2d3 ACM	314	if (slot >= INET_TWDR_TWKILL_SLOTS)
	315	slot = INET_TWDR_TWKILL_SLOTS - 1;
	316	}
	317	tw->tw_ttd = jiffies + timeo;
	318	slot = (twdr->slot + slot) & (INET_TWDR_TWKILL_SLOTS - 1);
	319	list = &twdr->cells[slot];
	320	} else {
	321	tw->tw_ttd = jiffies + (slot << INET_TWDR_RECYCLE_TICK);
	322
	323	if (twdr->twcal_hand < 0) {
	324	twdr->twcal_hand = 0;
	325	twdr->twcal_jiffie = jiffies;
	326	twdr->twcal_timer.expires = twdr->twcal_jiffie +
	327	(slot << INET_TWDR_RECYCLE_TICK);
	328	add_timer(&twdr->twcal_timer);
	329	} else {
	330	if (time_after(twdr->twcal_timer.expires,
	331	jiffies + (slot << INET_TWDR_RECYCLE_TICK)))
	332	mod_timer(&twdr->twcal_timer,
	333	jiffies + (slot << INET_TWDR_RECYCLE_TICK));
	334	slot = (twdr->twcal_hand + slot) & (INET_TWDR_RECYCLE_SLOTS - 1);
	335	}
	336	list = &twdr->twcal_row[slot];
	337	}
	338
	339	hlist_add_head(&tw->tw_death_node, list);
	340
	341	if (twdr->tw_count++ == 0)
	342	mod_timer(&twdr->tw_timer, jiffies + twdr->period);
	343	spin_unlock(&twdr->death_lock);
	344	}
	345
	346	EXPORT_SYMBOL_GPL(inet_twsk_schedule);
	347
	348	void inet_twdr_twcal_tick(unsigned long data)
	349	{
	350	struct inet_timewait_death_row *twdr;
	351	int n, slot;
	352	unsigned long j;
	353	unsigned long now = jiffies;
	354	int killed = 0;
	355	int adv = 0;
	356
	357	twdr = (struct inet_timewait_death_row *)data;
	358
	359	spin_lock(&twdr->death_lock);
	360	if (twdr->twcal_hand < 0)
	361	goto out;
	362
	363	slot = twdr->twcal_hand;
	364	j = twdr->twcal_jiffie;
	365
	366	for (n = 0; n < INET_TWDR_RECYCLE_SLOTS; n++) {
	367	if (time_before_eq(j, now)) {
	368	struct hlist_node node, safe;
	369	struct inet_timewait_sock *tw;
	370
	371	inet_twsk_for_each_inmate_safe(tw, node, safe,
	372	&twdr->twcal_row[slot]) {
	373	__inet_twsk_del_dead_node(tw);
	374	__inet_twsk_kill(tw, twdr->hashinfo);
	375	inet_twsk_put(tw);
	376	killed++;
	377	}
378	} else {
379	if (!adv) {
380	adv = 1;
381	twdr->twcal_jiffie = j;
382	twdr->twcal_hand = slot;
383	}
384
385	if (!hlist_empty(&twdr->twcal_row[slot])) {
386	mod_timer(&twdr->twcal_timer, j);
387	goto out;
388	}
389	}
390	j += 1 << INET_TWDR_RECYCLE_TICK;
391	slot = (slot + 1) & (INET_TWDR_RECYCLE_SLOTS - 1);
392	}
393	twdr->twcal_hand = -1;
394
395	out:
396	if ((twdr->tw_count -= killed) == 0)
397	del_timer(&twdr->tw_timer);
398	NET_ADD_STATS_BH(LINUX_MIB_TIMEWAITKILLED, killed);
399	spin_unlock(&twdr->death_lock);
400	}
401
402	EXPORT_SYMBOL_GPL(inet_twdr_twcal_tick);