[GitHub/mt8127/android_kernel_alcatel_ttab.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_nulls_unhashed(&tw->tw_node)) {
		write_unlock(lock);
		return;
	}
	hlist_nulls_del_rcu(&tw->tw_node);
	sk_nulls_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;
	WARN_ON(!icsk->icsk_bind_hash);
	inet_twsk_add_bind_node(tw, &tw->tw_tb->owners);
	spin_unlock(&bhead->lock);

	write_lock(lock);

	/*
	 * Step 2: Hash TW into TIMEWAIT chain.
	 * Should be done before removing sk from established chain
	 * because readers are lockless and search established first.
	 */
	atomic_inc(&tw->tw_refcnt);
	inet_twsk_add_node_rcu(tw, &ehead->twchain);

	/* Step 3: Remove SK from established hash. */
	if (__sk_nulls_del_node_init_rcu(sk))
		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);

	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_transparent  = inet->transparent;
		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);
#ifdef CONFIG_NET_NS
		NET_INC_STATS_BH(twsk_net(tw), LINUX_MIB_TIMEWAITED);
#endif
		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;
#ifndef CONFIG_NET_NS
	NET_ADD_STATS_BH(&init_net, LINUX_MIB_TIMEWAITED, killed);
#endif
	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);
#ifdef CONFIG_NET_NS
				NET_INC_STATS_BH(twsk_net(tw), LINUX_MIB_TIMEWAITKILLED);
#endif
				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);
#ifndef CONFIG_NET_NS
	NET_ADD_STATS_BH(&init_net, LINUX_MIB_TIMEWAITKILLED, killed);
#endif
	spin_unlock(&twdr->death_lock);
}

EXPORT_SYMBOL_GPL(inet_twdr_twcal_tick);

void inet_twsk_purge(struct net *net, struct inet_hashinfo *hashinfo,
		     struct inet_timewait_death_row *twdr, int family)
{
	struct inet_timewait_sock *tw;
	struct sock *sk;
	struct hlist_nulls_node *node;
	int h;

	local_bh_disable();
	for (h = 0; h < (hashinfo->ehash_size); h++) {
		struct inet_ehash_bucket *head =
			inet_ehash_bucket(hashinfo, h);
		rwlock_t *lock = inet_ehash_lockp(hashinfo, h);
restart:
		write_lock(lock);
		sk_nulls_for_each(sk, node, &head->twchain) {

			tw = inet_twsk(sk);
			if (!net_eq(twsk_net(tw), net) ||
			    tw->tw_family != family)
				continue;

			atomic_inc(&tw->tw_refcnt);
			write_unlock(lock);
			inet_twsk_deschedule(tw, twdr);
			inet_twsk_put(tw);

			goto restart;
		}
		write_unlock(lock);
	}
	local_bh_enable();
}
EXPORT_SYMBOL_GPL(inet_twsk_purge);
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);
3ab5aee7	26	if (hlist_nulls_unhashed(&tw->tw_node)) {
230140cf	27	write_unlock(lock);
e48c414e ACM	28	return;
e48c414e ACM	29	}
3ab5aee7 ED	30	hlist_nulls_del_rcu(&tw->tw_node);
3ab5aee7 ED	31	sk_nulls_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	88	tw->tw_tb = icsk->icsk_bind_hash;
547b792c	89	WARN_ON(!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	94
3ab5aee7 ED	95	/*
	96	* Step 2: Hash TW into TIMEWAIT chain.
	97	* Should be done before removing sk from established chain
	98	* because readers are lockless and search established first.
	99	*/
e48c414e	100	atomic_inc(&tw->tw_refcnt);
3ab5aee7 ED	101	inet_twsk_add_node_rcu(tw, &ehead->twchain);
	102
	103	/* Step 3: Remove SK from established hash. */
	104	if (__sk_nulls_del_node_init_rcu(sk))
	105	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
e48c414e	106
230140cf	107	write_unlock(lock);
e48c414e	108	}
c676270b	109
696ab2d3 ACM	110	EXPORT_SYMBOL_GPL(__inet_twsk_hashdance);
696ab2d3 ACM	111
c676270b ACM	112	struct inet_timewait_sock inet_twsk_alloc(const struct sock sk, const int state)
c676270b ACM	113	{
6d6ee43e ACM	114	struct inet_timewait_sock *tw =
6d6ee43e ACM	115	kmem_cache_alloc(sk->sk_prot_creator->twsk_prot->twsk_slab,
54e6ecb2	116	GFP_ATOMIC);
c676270b ACM	117	if (tw != NULL) {
	118	const struct inet_sock *inet = inet_sk(sk);
	119
	120	/* Give us an identity. */
	121	tw->tw_daddr = inet->daddr;
	122	tw->tw_rcv_saddr = inet->rcv_saddr;
	123	tw->tw_bound_dev_if = sk->sk_bound_dev_if;
	124	tw->tw_num = inet->num;
	125	tw->tw_state = TCP_TIME_WAIT;
	126	tw->tw_substate = state;
	127	tw->tw_sport = inet->sport;
	128	tw->tw_dport = inet->dport;
	129	tw->tw_family = sk->sk_family;
	130	tw->tw_reuse = sk->sk_reuse;
81c3d547	131	tw->tw_hash = sk->sk_hash;
c676270b	132	tw->tw_ipv6only = 0;
f5715aea	133	tw->tw_transparent = inet->transparent;
c676270b	134	tw->tw_prot = sk->sk_prot_creator;
cd5342d9	135	twsk_net_set(tw, hold_net(sock_net(sk)));
c676270b ACM	136	atomic_set(&tw->tw_refcnt, 1);
c676270b ACM	137	inet_twsk_dead_node_init(tw);
eeb2b856	138	__module_get(tw->tw_prot->owner);
c676270b ACM	139	}
	140
	141	return tw;
	142	}
696ab2d3 ACM	143
	144	EXPORT_SYMBOL_GPL(inet_twsk_alloc);
	145
	146	/* Returns non-zero if quota exceeded. */
	147	static int inet_twdr_do_twkill_work(struct inet_timewait_death_row *twdr,
	148	const int slot)
	149	{
	150	struct inet_timewait_sock *tw;
	151	struct hlist_node *node;
	152	unsigned int killed;
	153	int ret;
	154
	155	/* NOTE: compare this to previous version where lock
	156	* was released after detaching chain. It was racy,
	157	* because tw buckets are scheduled in not serialized context
	158	* in 2.3 (with netfilter), and with softnet it is common, because
	159	* soft irqs are not sequenced.
	160	*/
	161	killed = 0;
	162	ret = 0;
	163	rescan:
	164	inet_twsk_for_each_inmate(tw, node, &twdr->cells[slot]) {
	165	__inet_twsk_del_dead_node(tw);
	166	spin_unlock(&twdr->death_lock);
	167	__inet_twsk_kill(tw, twdr->hashinfo);
f2bf415c PE	168	#ifdef CONFIG_NET_NS
	169	NET_INC_STATS_BH(twsk_net(tw), LINUX_MIB_TIMEWAITED);
	170	#endif
696ab2d3 ACM	171	inet_twsk_put(tw);
	172	killed++;
	173	spin_lock(&twdr->death_lock);
	174	if (killed > INET_TWDR_TWKILL_QUOTA) {
	175	ret = 1;
	176	break;
	177	}
	178
	179	/* While we dropped twdr->death_lock, another cpu may have
	180	* killed off the next TW bucket in the list, therefore
	181	* do a fresh re-read of the hlist head node with the
	182	* lock reacquired. We still use the hlist traversal
	183	* macro in order to get the prefetches.
	184	*/
	185	goto rescan;
	186	}
	187
	188	twdr->tw_count -= killed;
f2bf415c PE	189	#ifndef CONFIG_NET_NS
	190	NET_ADD_STATS_BH(&init_net, LINUX_MIB_TIMEWAITED, killed);
	191	#endif
696ab2d3 ACM	192	return ret;
	193	}
	194
	195	void inet_twdr_hangman(unsigned long data)
	196	{
	197	struct inet_timewait_death_row *twdr;
	198	int unsigned need_timer;
	199
	200	twdr = (struct inet_timewait_death_row *)data;
	201	spin_lock(&twdr->death_lock);
	202
	203	if (twdr->tw_count == 0)
	204	goto out;
	205
	206	need_timer = 0;
	207	if (inet_twdr_do_twkill_work(twdr, twdr->slot)) {
	208	twdr->thread_slots \|= (1 << twdr->slot);
696ab2d3 ACM	209	schedule_work(&twdr->twkill_work);
	210	need_timer = 1;
	211	} else {
	212	/* We purged the entire slot, anything left? */
	213	if (twdr->tw_count)
	214	need_timer = 1;
	215	}
	216	twdr->slot = ((twdr->slot + 1) & (INET_TWDR_TWKILL_SLOTS - 1));
	217	if (need_timer)
	218	mod_timer(&twdr->tw_timer, jiffies + twdr->period);
	219	out:
	220	spin_unlock(&twdr->death_lock);
	221	}
	222
	223	EXPORT_SYMBOL_GPL(inet_twdr_hangman);
	224
65f27f38	225	void inet_twdr_twkill_work(struct work_struct *work)
696ab2d3	226	{
65f27f38 DH	227	struct inet_timewait_death_row *twdr =
65f27f38 DH	228	container_of(work, struct inet_timewait_death_row, twkill_work);
696ab2d3 ACM	229	int i;
696ab2d3 ACM	230
95c9382a PE	231	BUILD_BUG_ON((INET_TWDR_TWKILL_SLOTS - 1) >
95c9382a PE	232	(sizeof(twdr->thread_slots) * 8));
696ab2d3 ACM	233
	234	while (twdr->thread_slots) {
	235	spin_lock_bh(&twdr->death_lock);
	236	for (i = 0; i < INET_TWDR_TWKILL_SLOTS; i++) {
	237	if (!(twdr->thread_slots & (1 << i)))
	238	continue;
	239
	240	while (inet_twdr_do_twkill_work(twdr, i) != 0) {
	241	if (need_resched()) {
	242	spin_unlock_bh(&twdr->death_lock);
	243	schedule();
	244	spin_lock_bh(&twdr->death_lock);
	245	}
	246	}
	247
	248	twdr->thread_slots &= ~(1 << i);
	249	}
	250	spin_unlock_bh(&twdr->death_lock);
	251	}
	252	}
	253
	254	EXPORT_SYMBOL_GPL(inet_twdr_twkill_work);
	255
	256	/* These are always called from BH context. See callers in
	257	* tcp_input.c to verify this.
	258	*/
	259
	260	/* This is for handling early-kills of TIME_WAIT sockets. */
	261	void inet_twsk_deschedule(struct inet_timewait_sock *tw,
	262	struct inet_timewait_death_row *twdr)
	263	{
	264	spin_lock(&twdr->death_lock);
	265	if (inet_twsk_del_dead_node(tw)) {
	266	inet_twsk_put(tw);
	267	if (--twdr->tw_count == 0)
	268	del_timer(&twdr->tw_timer);
	269	}
	270	spin_unlock(&twdr->death_lock);
	271	__inet_twsk_kill(tw, twdr->hashinfo);
	272	}
	273
	274	EXPORT_SYMBOL(inet_twsk_deschedule);
	275
	276	void inet_twsk_schedule(struct inet_timewait_sock *tw,
	277	struct inet_timewait_death_row *twdr,
	278	const int timeo, const int timewait_len)
	279	{
	280	struct hlist_head *list;
	281	int slot;
	282
	283	/* timeout := RTO * 3.5
	284	*
	285	* 3.5 = 1+2+0.5 to wait for two retransmits.
	286	*
	287	* RATIONALE: if FIN arrived and we entered TIME-WAIT state,
	288	* our ACK acking that FIN can be lost. If N subsequent retransmitted
	289	* FINs (or previous seqments) are lost (probability of such event
	290	* is p^(N+1), where p is probability to lose single packet and
	291	* time to detect the loss is about RTO*(2^N - 1) with exponential
	292	* backoff). Normal timewait length is calculated so, that we
	293	* waited at least for one retransmitted FIN (maximal RTO is 120sec).
	294	* [ BTW Linux. following BSD, violates this requirement waiting
	295	* only for 60sec, we should wait at least for 240 secs.
	296	* Well, 240 consumes too much of resources 8)
297	* ]
298	* This interval is not reduced to catch old duplicate and
299	* responces to our wandering segments living for two MSLs.
300	* However, if we use PAWS to detect
301	* old duplicates, we can reduce the interval to bounds required
302	* by RTO, rather than MSL. So, if peer understands PAWS, we
303	* kill tw bucket after 3.5*RTO (it is important that this number
304	* is greater than TS tick!) and detect old duplicates with help
305	* of PAWS.
306	*/
307	slot = (timeo + (1 << INET_TWDR_RECYCLE_TICK) - 1) >> INET_TWDR_RECYCLE_TICK;
308
309	spin_lock(&twdr->death_lock);
310
311	/* Unlink it, if it was scheduled */
312	if (inet_twsk_del_dead_node(tw))
313	twdr->tw_count--;
314	else
315	atomic_inc(&tw->tw_refcnt);
316
317	if (slot >= INET_TWDR_RECYCLE_SLOTS) {
318	/* Schedule to slow timer */
319	if (timeo >= timewait_len) {
320	slot = INET_TWDR_TWKILL_SLOTS - 1;
321	} else {
172589cc	322	slot = DIV_ROUND_UP(timeo, twdr->period);
696ab2d3 ACM	323	if (slot >= INET_TWDR_TWKILL_SLOTS)
	324	slot = INET_TWDR_TWKILL_SLOTS - 1;
	325	}
	326	tw->tw_ttd = jiffies + timeo;
	327	slot = (twdr->slot + slot) & (INET_TWDR_TWKILL_SLOTS - 1);
	328	list = &twdr->cells[slot];
	329	} else {
	330	tw->tw_ttd = jiffies + (slot << INET_TWDR_RECYCLE_TICK);
	331
	332	if (twdr->twcal_hand < 0) {
	333	twdr->twcal_hand = 0;
	334	twdr->twcal_jiffie = jiffies;
	335	twdr->twcal_timer.expires = twdr->twcal_jiffie +
	336	(slot << INET_TWDR_RECYCLE_TICK);
	337	add_timer(&twdr->twcal_timer);
	338	} else {
	339	if (time_after(twdr->twcal_timer.expires,
	340	jiffies + (slot << INET_TWDR_RECYCLE_TICK)))
	341	mod_timer(&twdr->twcal_timer,
	342	jiffies + (slot << INET_TWDR_RECYCLE_TICK));
	343	slot = (twdr->twcal_hand + slot) & (INET_TWDR_RECYCLE_SLOTS - 1);
	344	}
	345	list = &twdr->twcal_row[slot];
	346	}
	347
	348	hlist_add_head(&tw->tw_death_node, list);
	349
	350	if (twdr->tw_count++ == 0)
	351	mod_timer(&twdr->tw_timer, jiffies + twdr->period);
	352	spin_unlock(&twdr->death_lock);
	353	}
	354
	355	EXPORT_SYMBOL_GPL(inet_twsk_schedule);
	356
	357	void inet_twdr_twcal_tick(unsigned long data)
	358	{
	359	struct inet_timewait_death_row *twdr;
	360	int n, slot;
	361	unsigned long j;
	362	unsigned long now = jiffies;
	363	int killed = 0;
	364	int adv = 0;
	365
	366	twdr = (struct inet_timewait_death_row *)data;
	367
	368	spin_lock(&twdr->death_lock);
	369	if (twdr->twcal_hand < 0)
	370	goto out;
	371
	372	slot = twdr->twcal_hand;
	373	j = twdr->twcal_jiffie;
	374
	375	for (n = 0; n < INET_TWDR_RECYCLE_SLOTS; n++) {
	376	if (time_before_eq(j, now)) {
	377	struct hlist_node node, safe;
	378	struct inet_timewait_sock *tw;
	379
	380	inet_twsk_for_each_inmate_safe(tw, node, safe,
	381	&twdr->twcal_row[slot]) {
	382	__inet_twsk_del_dead_node(tw);
	383	__inet_twsk_kill(tw, twdr->hashinfo);
f2bf415c PE	384	#ifdef CONFIG_NET_NS
	385	NET_INC_STATS_BH(twsk_net(tw), LINUX_MIB_TIMEWAITKILLED);
	386	#endif
696ab2d3 ACM	387	inet_twsk_put(tw);
	388	killed++;
	389	}
	390	} else {
	391	if (!adv) {
	392	adv = 1;
	393	twdr->twcal_jiffie = j;
	394	twdr->twcal_hand = slot;
	395	}
	396
	397	if (!hlist_empty(&twdr->twcal_row[slot])) {
	398	mod_timer(&twdr->twcal_timer, j);
	399	goto out;
	400	}
	401	}
	402	j += 1 << INET_TWDR_RECYCLE_TICK;
	403	slot = (slot + 1) & (INET_TWDR_RECYCLE_SLOTS - 1);
	404	}
	405	twdr->twcal_hand = -1;
	406
	407	out:
	408	if ((twdr->tw_count -= killed) == 0)
	409	del_timer(&twdr->tw_timer);
f2bf415c PE	410	#ifndef CONFIG_NET_NS
	411	NET_ADD_STATS_BH(&init_net, LINUX_MIB_TIMEWAITKILLED, killed);
	412	#endif
696ab2d3 ACM	413	spin_unlock(&twdr->death_lock);
	414	}
	415
	416	EXPORT_SYMBOL_GPL(inet_twdr_twcal_tick);
d315492b DL	417
	418	void inet_twsk_purge(struct net net, struct inet_hashinfo hashinfo,
	419	struct inet_timewait_death_row *twdr, int family)
	420	{
	421	struct inet_timewait_sock *tw;
	422	struct sock *sk;
3ab5aee7	423	struct hlist_nulls_node *node;
d315492b DL	424	int h;
	425
	426	local_bh_disable();
	427	for (h = 0; h < (hashinfo->ehash_size); h++) {
	428	struct inet_ehash_bucket *head =
	429	inet_ehash_bucket(hashinfo, h);
	430	rwlock_t *lock = inet_ehash_lockp(hashinfo, h);
	431	restart:
	432	write_lock(lock);
3ab5aee7	433	sk_nulls_for_each(sk, node, &head->twchain) {
d315492b DL	434
	435	tw = inet_twsk(sk);
	436	if (!net_eq(twsk_net(tw), net) \|\|
	437	tw->tw_family != family)
	438	continue;
	439
	440	atomic_inc(&tw->tw_refcnt);
	441	write_unlock(lock);
	442	inet_twsk_deschedule(tw, twdr);
	443	inet_twsk_put(tw);
	444
	445	goto restart;
	446	}
	447	write_unlock(lock);
	448	}
	449	local_bh_enable();
	450	}
	451	EXPORT_SYMBOL_GPL(inet_twsk_purge);