[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 <linux/kmemcheck.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <net/inet_hashtables.h>
#include <net/inet_timewait_sock.h>
#include <net/ip.h>


/**
 *	inet_twsk_unhash - unhash a timewait socket from established hash
 *	@tw: timewait socket
 *
 *	unhash a timewait socket from established hash, if hashed.
 *	ehash lock must be held by caller.
 *	Returns 1 if caller should call inet_twsk_put() after lock release.
 */
int inet_twsk_unhash(struct inet_timewait_sock *tw)
{
	if (hlist_nulls_unhashed(&tw->tw_node))
		return 0;

	hlist_nulls_del_rcu(&tw->tw_node);
	sk_nulls_node_init(&tw->tw_node);
	/*
	 * We cannot call inet_twsk_put() ourself under lock,
	 * caller must call it for us.
	 */
	return 1;
}

/**
 *	inet_twsk_bind_unhash - unhash a timewait socket from bind hash
 *	@tw: timewait socket
 *	@hashinfo: hashinfo pointer
 *
 *	unhash a timewait socket from bind hash, if hashed.
 *	bind hash lock must be held by caller.
 *	Returns 1 if caller should call inet_twsk_put() after lock release.
 */
int inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
			  struct inet_hashinfo *hashinfo)
{
	struct inet_bind_bucket *tb = tw->tw_tb;

	if (!tb)
		return 0;

	__hlist_del(&tw->tw_bind_node);
	tw->tw_tb = NULL;
	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	/*
	 * We cannot call inet_twsk_put() ourself under lock,
	 * caller must call it for us.
	 */
	return 1;
}

/* 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;
	int refcnt;
	/* Unlink from established hashes. */
	spinlock_t *lock = inet_ehash_lockp(hashinfo, tw->tw_hash);

	spin_lock(lock);
	refcnt = inet_twsk_unhash(tw);
	spin_unlock(lock);

	/* Disassociate with bind bucket. */
	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), tw->tw_num,
			hashinfo->bhash_size)];

	spin_lock(&bhead->lock);
	refcnt += inet_twsk_bind_unhash(tw, hashinfo);
	spin_unlock(&bhead->lock);

#ifdef SOCK_REFCNT_DEBUG
	if (atomic_read(&tw->tw_refcnt) != 1) {
		pr_debug("%s timewait_sock %p refcnt=%d\n",
			 tw->tw_prot->name, tw, atomic_read(&tw->tw_refcnt));
	}
#endif
	while (refcnt) {
		inet_twsk_put(tw);
		refcnt--;
	}
}

static noinline void inet_twsk_free(struct inet_timewait_sock *tw)
{
	struct module *owner = tw->tw_prot->owner;
	twsk_destructor((struct sock *)tw);
#ifdef SOCK_REFCNT_DEBUG
	pr_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);
}

void inet_twsk_put(struct inet_timewait_sock *tw)
{
	if (atomic_dec_and_test(&tw->tw_refcnt))
		inet_twsk_free(tw);
}
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);
	spinlock_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->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);

	spin_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.
	 */
	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);

	/*
	 * Notes :
	 * - We initially set tw_refcnt to 0 in inet_twsk_alloc()
	 * - We add one reference for the bhash link
	 * - We add one reference for the ehash link
	 * - We want this refcnt update done before allowing other
	 *   threads to find this tw in ehash chain.
	 */
	atomic_add(1 + 1 + 1, &tw->tw_refcnt);

	spin_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);

		kmemcheck_annotate_bitfield(tw, flags);

		/* Give us an identity. */
		tw->tw_daddr	    = inet->inet_daddr;
		tw->tw_rcv_saddr    = inet->inet_rcv_saddr;
		tw->tw_bound_dev_if = sk->sk_bound_dev_if;
		tw->tw_tos	    = inet->tos;
		tw->tw_num	    = inet->inet_num;
		tw->tw_state	    = TCP_TIME_WAIT;
		tw->tw_substate	    = state;
		tw->tw_sport	    = inet->inet_sport;
		tw->tw_dport	    = inet->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)));
		/*
		 * Because we use RCU lookups, we should not set tw_refcnt
		 * to a non null value before everything is setup for this
		 * timewait socket.
		 */
		atomic_set(&tw->tw_refcnt, 0);
		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;
	unsigned int 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 inet_hashinfo *hashinfo,
		     struct inet_timewait_death_row *twdr, int family)
{
	struct inet_timewait_sock *tw;
	struct sock *sk;
	struct hlist_nulls_node *node;
	unsigned int slot;

	for (slot = 0; slot <= hashinfo->ehash_mask; slot++) {
		struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
restart_rcu:
		rcu_read_lock();
restart:
		sk_nulls_for_each_rcu(sk, node, &head->twchain) {
			tw = inet_twsk(sk);
			if ((tw->tw_family != family) ||
				atomic_read(&twsk_net(tw)->count))
				continue;

			if (unlikely(!atomic_inc_not_zero(&tw->tw_refcnt)))
				continue;

			if (unlikely((tw->tw_family != family) ||
				     atomic_read(&twsk_net(tw)->count))) {
				inet_twsk_put(tw);
				goto restart;
			}

			rcu_read_unlock();
			local_bh_disable();
			inet_twsk_deschedule(tw, twdr);
			local_bh_enable();
			inet_twsk_put(tw);
			goto restart_rcu;
		}
		/* If the nulls value we got at the end of this lookup is
		 * not the expected one, we must restart lookup.
		 * We probably met an item that was moved to another chain.
		 */
		if (get_nulls_value(node) != slot)
			goto restart;
		rcu_read_unlock();
	}
}
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>
9e337b0f	12	#include <linux/kmemcheck.h>
5a0e3ad6	13	#include <linux/slab.h>
3a9a231d	14	#include <linux/module.h>
e48c414e ACM	15	#include <net/inet_hashtables.h>
e48c414e ACM	16	#include <net/inet_timewait_sock.h>
696ab2d3	17	#include <net/ip.h>
e48c414e	18
13475a30	19
2a8875e7 ED	20	/**
	21	* inet_twsk_unhash - unhash a timewait socket from established hash
	22	* @tw: timewait socket
	23	*
	24	* unhash a timewait socket from established hash, if hashed.
	25	* ehash lock must be held by caller.
	26	* Returns 1 if caller should call inet_twsk_put() after lock release.
13475a30 ED	27	*/
	28	int inet_twsk_unhash(struct inet_timewait_sock *tw)
	29	{
	30	if (hlist_nulls_unhashed(&tw->tw_node))
	31	return 0;
	32
	33	hlist_nulls_del_rcu(&tw->tw_node);
	34	sk_nulls_node_init(&tw->tw_node);
2a8875e7 ED	35	/*
	36	* We cannot call inet_twsk_put() ourself under lock,
	37	* caller must call it for us.
	38	*/
13475a30 ED	39	return 1;
	40	}
	41
2a8875e7 ED	42	/**
	43	* inet_twsk_bind_unhash - unhash a timewait socket from bind hash
	44	* @tw: timewait socket
	45	* @hashinfo: hashinfo pointer
	46	*
	47	* unhash a timewait socket from bind hash, if hashed.
	48	* bind hash lock must be held by caller.
	49	* Returns 1 if caller should call inet_twsk_put() after lock release.
3cdaedae ED	50	*/
	51	int inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
	52	struct inet_hashinfo *hashinfo)
	53	{
	54	struct inet_bind_bucket *tb = tw->tw_tb;
	55
	56	if (!tb)
	57	return 0;
	58
	59	__hlist_del(&tw->tw_bind_node);
	60	tw->tw_tb = NULL;
	61	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
2a8875e7 ED	62	/*
	63	* We cannot call inet_twsk_put() ourself under lock,
	64	* caller must call it for us.
	65	*/
3cdaedae ED	66	return 1;
	67	}
	68
e48c414e	69	/* Must be called with locally disabled BHs. */
acd159b6 AB	70	static void __inet_twsk_kill(struct inet_timewait_sock *tw,
acd159b6 AB	71	struct inet_hashinfo *hashinfo)
e48c414e ACM	72	{
e48c414e ACM	73	struct inet_bind_hashbucket *bhead;
13475a30	74	int refcnt;
e48c414e	75	/* Unlink from established hashes. */
9db66bdc	76	spinlock_t *lock = inet_ehash_lockp(hashinfo, tw->tw_hash);
e48c414e	77
9db66bdc	78	spin_lock(lock);
13475a30	79	refcnt = inet_twsk_unhash(tw);
9db66bdc	80	spin_unlock(lock);
e48c414e ACM	81
e48c414e ACM	82	/* Disassociate with bind bucket. */
7f635ab7 PE	83	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), tw->tw_num,
7f635ab7 PE	84	hashinfo->bhash_size)];
3cdaedae	85
e48c414e	86	spin_lock(&bhead->lock);
3cdaedae	87	refcnt += inet_twsk_bind_unhash(tw, hashinfo);
e48c414e	88	spin_unlock(&bhead->lock);
3cdaedae	89
e48c414e ACM	90	#ifdef SOCK_REFCNT_DEBUG
e48c414e ACM	91	if (atomic_read(&tw->tw_refcnt) != 1) {
91df42be JP	92	pr_debug("%s timewait_sock %p refcnt=%d\n",
91df42be JP	93	tw->tw_prot->name, tw, atomic_read(&tw->tw_refcnt));
e48c414e ACM	94	}
e48c414e ACM	95	#endif
13475a30 ED	96	while (refcnt) {
	97	inet_twsk_put(tw);
	98	refcnt--;
	99	}
e48c414e ACM	100	}
e48c414e ACM	101
4dbc8ef7	102	static noinline void inet_twsk_free(struct inet_timewait_sock *tw)
7054fb93	103	{
4dbc8ef7 ACM	104	struct module *owner = tw->tw_prot->owner;
4dbc8ef7 ACM	105	twsk_destructor((struct sock *)tw);
7054fb93	106	#ifdef SOCK_REFCNT_DEBUG
4dbc8ef7	107	pr_debug("%s timewait_sock %p released\n", tw->tw_prot->name, tw);
7054fb93	108	#endif
4dbc8ef7 ACM	109	release_net(twsk_net(tw));
	110	kmem_cache_free(tw->tw_prot->twsk_prot->twsk_slab, tw);
	111	module_put(owner);
	112	}
	113
	114	void inet_twsk_put(struct inet_timewait_sock *tw)
	115	{
	116	if (atomic_dec_and_test(&tw->tw_refcnt))
	117	inet_twsk_free(tw);
7054fb93 PE	118	}
	119	EXPORT_SYMBOL_GPL(inet_twsk_put);
	120
e48c414e ACM	121	/*
	122	* Enter the time wait state. This is called with locally disabled BH.
	123	* Essentially we whip up a timewait bucket, copy the relevant info into it
	124	* from the SK, and mess with hash chains and list linkage.
	125	*/
	126	void __inet_twsk_hashdance(struct inet_timewait_sock tw, struct sock sk,
	127	struct inet_hashinfo *hashinfo)
	128	{
	129	const struct inet_sock *inet = inet_sk(sk);
463c84b9	130	const struct inet_connection_sock *icsk = inet_csk(sk);
81c3d547	131	struct inet_ehash_bucket *ehead = inet_ehash_bucket(hashinfo, sk->sk_hash);
9db66bdc	132	spinlock_t *lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
e48c414e ACM	133	struct inet_bind_hashbucket *bhead;
	134	/* Step 1: Put TW into bind hash. Original socket stays there too.
	135	Note, that any socket with inet->num != 0 MUST be bound in
	136	binding cache, even if it is closed.
	137	*/
c720c7e8	138	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), inet->inet_num,
7f635ab7	139	hashinfo->bhash_size)];
e48c414e	140	spin_lock(&bhead->lock);
463c84b9	141	tw->tw_tb = icsk->icsk_bind_hash;
547b792c	142	WARN_ON(!icsk->icsk_bind_hash);
e48c414e ACM	143	inet_twsk_add_bind_node(tw, &tw->tw_tb->owners);
	144	spin_unlock(&bhead->lock);
	145
9db66bdc	146	spin_lock(lock);
e48c414e	147
3ab5aee7 ED	148	/*
	149	* Step 2: Hash TW into TIMEWAIT chain.
	150	* Should be done before removing sk from established chain
	151	* because readers are lockless and search established first.
	152	*/
3ab5aee7 ED	153	inet_twsk_add_node_rcu(tw, &ehead->twchain);
	154
	155	/* Step 3: Remove SK from established hash. */
	156	if (__sk_nulls_del_node_init_rcu(sk))
	157	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
e48c414e	158
47e1c323 ED	159	/*
47e1c323 ED	160	* Notes :
2a8875e7	161	* - We initially set tw_refcnt to 0 in inet_twsk_alloc()
47e1c323 ED	162	* - We add one reference for the bhash link
	163	* - We add one reference for the ehash link
	164	* - We want this refcnt update done before allowing other
	165	* threads to find this tw in ehash chain.
	166	*/
	167	atomic_add(1 + 1 + 1, &tw->tw_refcnt);
	168
9db66bdc	169	spin_unlock(lock);
e48c414e	170	}
696ab2d3 ACM	171	EXPORT_SYMBOL_GPL(__inet_twsk_hashdance);
696ab2d3 ACM	172
c676270b ACM	173	struct inet_timewait_sock inet_twsk_alloc(const struct sock sk, const int state)
c676270b ACM	174	{
6d6ee43e ACM	175	struct inet_timewait_sock *tw =
6d6ee43e ACM	176	kmem_cache_alloc(sk->sk_prot_creator->twsk_prot->twsk_slab,
54e6ecb2	177	GFP_ATOMIC);
c676270b ACM	178	if (tw != NULL) {
	179	const struct inet_sock *inet = inet_sk(sk);
	180
9e337b0f VN	181	kmemcheck_annotate_bitfield(tw, flags);
9e337b0f VN	182
c676270b	183	/* Give us an identity. */
c720c7e8 ED	184	tw->tw_daddr = inet->inet_daddr;
c720c7e8 ED	185	tw->tw_rcv_saddr = inet->inet_rcv_saddr;
c676270b	186	tw->tw_bound_dev_if = sk->sk_bound_dev_if;
66b13d99	187	tw->tw_tos = inet->tos;
c720c7e8	188	tw->tw_num = inet->inet_num;
c676270b ACM	189	tw->tw_state = TCP_TIME_WAIT;
c676270b ACM	190	tw->tw_substate = state;
c720c7e8 ED	191	tw->tw_sport = inet->inet_sport;
c720c7e8 ED	192	tw->tw_dport = inet->inet_dport;
c676270b ACM	193	tw->tw_family = sk->sk_family;
c676270b ACM	194	tw->tw_reuse = sk->sk_reuse;
81c3d547	195	tw->tw_hash = sk->sk_hash;
c676270b	196	tw->tw_ipv6only = 0;
f5715aea	197	tw->tw_transparent = inet->transparent;
c676270b	198	tw->tw_prot = sk->sk_prot_creator;
cd5342d9	199	twsk_net_set(tw, hold_net(sock_net(sk)));
47e1c323 ED	200	/*
	201	* Because we use RCU lookups, we should not set tw_refcnt
	202	* to a non null value before everything is setup for this
	203	* timewait socket.
	204	*/
	205	atomic_set(&tw->tw_refcnt, 0);
c676270b	206	inet_twsk_dead_node_init(tw);
eeb2b856	207	__module_get(tw->tw_prot->owner);
c676270b ACM	208	}
	209
	210	return tw;
	211	}
696ab2d3 ACM	212	EXPORT_SYMBOL_GPL(inet_twsk_alloc);
	213
	214	/* Returns non-zero if quota exceeded. */
	215	static int inet_twdr_do_twkill_work(struct inet_timewait_death_row *twdr,
	216	const int slot)
	217	{
	218	struct inet_timewait_sock *tw;
	219	struct hlist_node *node;
	220	unsigned int killed;
	221	int ret;
	222
	223	/* NOTE: compare this to previous version where lock
	224	* was released after detaching chain. It was racy,
	225	* because tw buckets are scheduled in not serialized context
	226	* in 2.3 (with netfilter), and with softnet it is common, because
	227	* soft irqs are not sequenced.
	228	*/
	229	killed = 0;
	230	ret = 0;
	231	rescan:
	232	inet_twsk_for_each_inmate(tw, node, &twdr->cells[slot]) {
	233	__inet_twsk_del_dead_node(tw);
	234	spin_unlock(&twdr->death_lock);
	235	__inet_twsk_kill(tw, twdr->hashinfo);
f2bf415c PE	236	#ifdef CONFIG_NET_NS
	237	NET_INC_STATS_BH(twsk_net(tw), LINUX_MIB_TIMEWAITED);
	238	#endif
696ab2d3 ACM	239	inet_twsk_put(tw);
	240	killed++;
	241	spin_lock(&twdr->death_lock);
	242	if (killed > INET_TWDR_TWKILL_QUOTA) {
	243	ret = 1;
	244	break;
	245	}
	246
	247	/* While we dropped twdr->death_lock, another cpu may have
	248	* killed off the next TW bucket in the list, therefore
	249	* do a fresh re-read of the hlist head node with the
	250	* lock reacquired. We still use the hlist traversal
	251	* macro in order to get the prefetches.
	252	*/
	253	goto rescan;
	254	}
	255
	256	twdr->tw_count -= killed;
f2bf415c PE	257	#ifndef CONFIG_NET_NS
	258	NET_ADD_STATS_BH(&init_net, LINUX_MIB_TIMEWAITED, killed);
	259	#endif
696ab2d3 ACM	260	return ret;
	261	}
	262
	263	void inet_twdr_hangman(unsigned long data)
	264	{
	265	struct inet_timewait_death_row *twdr;
95c96174	266	unsigned int need_timer;
696ab2d3 ACM	267
	268	twdr = (struct inet_timewait_death_row *)data;
	269	spin_lock(&twdr->death_lock);
	270
	271	if (twdr->tw_count == 0)
	272	goto out;
	273
	274	need_timer = 0;
	275	if (inet_twdr_do_twkill_work(twdr, twdr->slot)) {
	276	twdr->thread_slots \|= (1 << twdr->slot);
696ab2d3 ACM	277	schedule_work(&twdr->twkill_work);
	278	need_timer = 1;
	279	} else {
	280	/* We purged the entire slot, anything left? */
	281	if (twdr->tw_count)
	282	need_timer = 1;
80a1096b	283	twdr->slot = ((twdr->slot + 1) & (INET_TWDR_TWKILL_SLOTS - 1));
696ab2d3	284	}
696ab2d3 ACM	285	if (need_timer)
	286	mod_timer(&twdr->tw_timer, jiffies + twdr->period);
	287	out:
	288	spin_unlock(&twdr->death_lock);
	289	}
696ab2d3 ACM	290	EXPORT_SYMBOL_GPL(inet_twdr_hangman);
696ab2d3 ACM	291
65f27f38	292	void inet_twdr_twkill_work(struct work_struct *work)
696ab2d3	293	{
65f27f38 DH	294	struct inet_timewait_death_row *twdr =
65f27f38 DH	295	container_of(work, struct inet_timewait_death_row, twkill_work);
696ab2d3 ACM	296	int i;
696ab2d3 ACM	297
95c9382a PE	298	BUILD_BUG_ON((INET_TWDR_TWKILL_SLOTS - 1) >
95c9382a PE	299	(sizeof(twdr->thread_slots) * 8));
696ab2d3 ACM	300
	301	while (twdr->thread_slots) {
	302	spin_lock_bh(&twdr->death_lock);
	303	for (i = 0; i < INET_TWDR_TWKILL_SLOTS; i++) {
	304	if (!(twdr->thread_slots & (1 << i)))
	305	continue;
	306
	307	while (inet_twdr_do_twkill_work(twdr, i) != 0) {
	308	if (need_resched()) {
	309	spin_unlock_bh(&twdr->death_lock);
	310	schedule();
	311	spin_lock_bh(&twdr->death_lock);
	312	}
	313	}
	314
	315	twdr->thread_slots &= ~(1 << i);
	316	}
	317	spin_unlock_bh(&twdr->death_lock);
	318	}
	319	}
696ab2d3 ACM	320	EXPORT_SYMBOL_GPL(inet_twdr_twkill_work);
	321
	322	/* These are always called from BH context. See callers in
	323	* tcp_input.c to verify this.
	324	*/
	325
	326	/* This is for handling early-kills of TIME_WAIT sockets. */
	327	void inet_twsk_deschedule(struct inet_timewait_sock *tw,
	328	struct inet_timewait_death_row *twdr)
	329	{
	330	spin_lock(&twdr->death_lock);
	331	if (inet_twsk_del_dead_node(tw)) {
	332	inet_twsk_put(tw);
	333	if (--twdr->tw_count == 0)
	334	del_timer(&twdr->tw_timer);
	335	}
	336	spin_unlock(&twdr->death_lock);
	337	__inet_twsk_kill(tw, twdr->hashinfo);
	338	}
696ab2d3 ACM	339	EXPORT_SYMBOL(inet_twsk_deschedule);
	340
	341	void inet_twsk_schedule(struct inet_timewait_sock *tw,
	342	struct inet_timewait_death_row *twdr,
	343	const int timeo, const int timewait_len)
	344	{
	345	struct hlist_head *list;
	346	int slot;
	347
	348	/* timeout := RTO * 3.5
	349	*
	350	* 3.5 = 1+2+0.5 to wait for two retransmits.
	351	*
	352	* RATIONALE: if FIN arrived and we entered TIME-WAIT state,
	353	* our ACK acking that FIN can be lost. If N subsequent retransmitted
	354	* FINs (or previous seqments) are lost (probability of such event
	355	* is p^(N+1), where p is probability to lose single packet and
	356	* time to detect the loss is about RTO*(2^N - 1) with exponential
	357	* backoff). Normal timewait length is calculated so, that we
	358	* waited at least for one retransmitted FIN (maximal RTO is 120sec).
	359	* [ BTW Linux. following BSD, violates this requirement waiting
	360	* only for 60sec, we should wait at least for 240 secs.
	361	* Well, 240 consumes too much of resources 8)
	362	* ]
	363	* This interval is not reduced to catch old duplicate and
	364	* responces to our wandering segments living for two MSLs.
	365	* However, if we use PAWS to detect
	366	* old duplicates, we can reduce the interval to bounds required
	367	* by RTO, rather than MSL. So, if peer understands PAWS, we
	368	* kill tw bucket after 3.5*RTO (it is important that this number
	369	* is greater than TS tick!) and detect old duplicates with help
	370	* of PAWS.
	371	*/
	372	slot = (timeo + (1 << INET_TWDR_RECYCLE_TICK) - 1) >> INET_TWDR_RECYCLE_TICK;
	373
	374	spin_lock(&twdr->death_lock);
	375
	376	/* Unlink it, if it was scheduled */
	377	if (inet_twsk_del_dead_node(tw))
	378	twdr->tw_count--;
	379	else
	380	atomic_inc(&tw->tw_refcnt);
	381
	382	if (slot >= INET_TWDR_RECYCLE_SLOTS) {
	383	/* Schedule to slow timer */
	384	if (timeo >= timewait_len) {
	385	slot = INET_TWDR_TWKILL_SLOTS - 1;
	386	} else {
172589cc	387	slot = DIV_ROUND_UP(timeo, twdr->period);
696ab2d3 ACM	388	if (slot >= INET_TWDR_TWKILL_SLOTS)
	389	slot = INET_TWDR_TWKILL_SLOTS - 1;
	390	}
	391	tw->tw_ttd = jiffies + timeo;
	392	slot = (twdr->slot + slot) & (INET_TWDR_TWKILL_SLOTS - 1);
	393	list = &twdr->cells[slot];
	394	} else {
	395	tw->tw_ttd = jiffies + (slot << INET_TWDR_RECYCLE_TICK);
	396
	397	if (twdr->twcal_hand < 0) {
	398	twdr->twcal_hand = 0;
	399	twdr->twcal_jiffie = jiffies;
	400	twdr->twcal_timer.expires = twdr->twcal_jiffie +
	401	(slot << INET_TWDR_RECYCLE_TICK);
	402	add_timer(&twdr->twcal_timer);
	403	} else {
	404	if (time_after(twdr->twcal_timer.expires,
	405	jiffies + (slot << INET_TWDR_RECYCLE_TICK)))
	406	mod_timer(&twdr->twcal_timer,
	407	jiffies + (slot << INET_TWDR_RECYCLE_TICK));
	408	slot = (twdr->twcal_hand + slot) & (INET_TWDR_RECYCLE_SLOTS - 1);
	409	}
	410	list = &twdr->twcal_row[slot];
	411	}
	412
	413	hlist_add_head(&tw->tw_death_node, list);
	414
	415	if (twdr->tw_count++ == 0)
	416	mod_timer(&twdr->tw_timer, jiffies + twdr->period);
	417	spin_unlock(&twdr->death_lock);
	418	}
696ab2d3 ACM	419	EXPORT_SYMBOL_GPL(inet_twsk_schedule);
	420
	421	void inet_twdr_twcal_tick(unsigned long data)
	422	{
	423	struct inet_timewait_death_row *twdr;
	424	int n, slot;
	425	unsigned long j;
	426	unsigned long now = jiffies;
	427	int killed = 0;
	428	int adv = 0;
	429
	430	twdr = (struct inet_timewait_death_row *)data;
	431
	432	spin_lock(&twdr->death_lock);
	433	if (twdr->twcal_hand < 0)
	434	goto out;
	435
	436	slot = twdr->twcal_hand;
	437	j = twdr->twcal_jiffie;
	438
	439	for (n = 0; n < INET_TWDR_RECYCLE_SLOTS; n++) {
	440	if (time_before_eq(j, now)) {
	441	struct hlist_node node, safe;
	442	struct inet_timewait_sock *tw;
	443
	444	inet_twsk_for_each_inmate_safe(tw, node, safe,
	445	&twdr->twcal_row[slot]) {
	446	__inet_twsk_del_dead_node(tw);
	447	__inet_twsk_kill(tw, twdr->hashinfo);
f2bf415c PE	448	#ifdef CONFIG_NET_NS
	449	NET_INC_STATS_BH(twsk_net(tw), LINUX_MIB_TIMEWAITKILLED);
	450	#endif
696ab2d3 ACM	451	inet_twsk_put(tw);
	452	killed++;
	453	}
	454	} else {
	455	if (!adv) {
	456	adv = 1;
	457	twdr->twcal_jiffie = j;
	458	twdr->twcal_hand = slot;
	459	}
	460
	461	if (!hlist_empty(&twdr->twcal_row[slot])) {
	462	mod_timer(&twdr->twcal_timer, j);
	463	goto out;
	464	}
	465	}
	466	j += 1 << INET_TWDR_RECYCLE_TICK;
	467	slot = (slot + 1) & (INET_TWDR_RECYCLE_SLOTS - 1);
	468	}
	469	twdr->twcal_hand = -1;
	470
	471	out:
	472	if ((twdr->tw_count -= killed) == 0)
	473	del_timer(&twdr->tw_timer);
f2bf415c PE	474	#ifndef CONFIG_NET_NS
	475	NET_ADD_STATS_BH(&init_net, LINUX_MIB_TIMEWAITKILLED, killed);
	476	#endif
696ab2d3 ACM	477	spin_unlock(&twdr->death_lock);
696ab2d3 ACM	478	}
696ab2d3	479	EXPORT_SYMBOL_GPL(inet_twdr_twcal_tick);
d315492b	480
b099ce26	481	void inet_twsk_purge(struct inet_hashinfo *hashinfo,
d315492b DL	482	struct inet_timewait_death_row *twdr, int family)
	483	{
	484	struct inet_timewait_sock *tw;
	485	struct sock *sk;
3ab5aee7	486	struct hlist_nulls_node *node;
575f4cd5	487	unsigned int slot;
d315492b	488
575f4cd5 EB	489	for (slot = 0; slot <= hashinfo->ehash_mask; slot++) {
	490	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
	491	restart_rcu:
	492	rcu_read_lock();
d315492b	493	restart:
575f4cd5	494	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
d315492b	495	tw = inet_twsk(sk);
b099ce26 EB	496	if ((tw->tw_family != family) \|\|
b099ce26 EB	497	atomic_read(&twsk_net(tw)->count))
d315492b DL	498	continue;
d315492b DL	499
575f4cd5 EB	500	if (unlikely(!atomic_inc_not_zero(&tw->tw_refcnt)))
	501	continue;
	502
b099ce26 EB	503	if (unlikely((tw->tw_family != family) \|\|
b099ce26 EB	504	atomic_read(&twsk_net(tw)->count))) {
575f4cd5 EB	505	inet_twsk_put(tw);
	506	goto restart;
	507	}
	508
	509	rcu_read_unlock();
91035f0b	510	local_bh_disable();
d315492b	511	inet_twsk_deschedule(tw, twdr);
91035f0b	512	local_bh_enable();
d315492b	513	inet_twsk_put(tw);
575f4cd5	514	goto restart_rcu;
d315492b	515	}
575f4cd5 EB	516	/* If the nulls value we got at the end of this lookup is
	517	* not the expected one, we must restart lookup.
	518	* We probably met an item that was moved to another chain.
	519	*/
	520	if (get_nulls_value(node) != slot)
	521	goto restart;
	522	rcu_read_unlock();
d315492b	523	}
d315492b DL	524	}
d315492b DL	525	EXPORT_SYMBOL_GPL(inet_twsk_purge);