tcp_minisocks.c

Linux内核源代码 为压缩文件 是<<Linux内核>>一书中的源代码

/*
 * 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.
 *
 *		Implementation of the Transmission Control Protocol(TCP).
 *
 * Version:	$Id: tcp_minisocks.c,v 1.5 2000/11/28 17:04:10 davem Exp $
 *
 * Authors:	Ross Biro, <[e-mail]>
 *		Fred N. van Kempen, <[e-mail]>
 *		Mark Evans, <[e-mail]>
 *		Corey Minyard <wf-rch![e-mail]>
 *		Florian La Roche, <[e-mail]>
 *		Charles Hedrick, <[e-mail]>
 *		Linus Torvalds, <[e-mail]>
 *		Alan Cox, <[e-mail]>
 *		Matthew Dillon, <[e-mail]>
 *		Arnt Gulbrandsen, <[e-mail]>
 *		Jorge Cwik, <[e-mail]>
 */

#include <linux/config.h>
#include <linux/mm.h>
#include <linux/sysctl.h>
#include <net/tcp.h>
#include <net/inet_common.h>

#ifdef CONFIG_SYSCTL
#define SYNC_INIT 0 /* let the user enable it */
#else
#define SYNC_INIT 1
#endif

int sysctl_tcp_tw_recycle = 0;
int sysctl_tcp_max_tw_buckets = NR_FILE*2;

int sysctl_tcp_syncookies = SYNC_INIT; 
int sysctl_tcp_abort_on_overflow = 0;

static __inline__ int tcp_in_window(u32 seq, u32 end_seq, u32 s_win, u32 e_win)
{
	if (seq == s_win)
		return 1;
	if (after(end_seq, s_win) && before(seq, e_win))
		return 1;
	return (seq == e_win && seq == end_seq);
}

/* New-style handling of TIME_WAIT sockets. */

int tcp_tw_count = 0;


/* Must be called with locally disabled BHs. */
void tcp_timewait_kill(struct tcp_tw_bucket *tw)
{
	struct tcp_ehash_bucket *ehead;
	struct tcp_bind_hashbucket *bhead;
	struct tcp_bind_bucket *tb;

	/* Unlink from established hashes. */
	ehead = &tcp_ehash[tw->hashent];
	write_lock(&ehead->lock);
	if (!tw->pprev) {
		write_unlock(&ehead->lock);
		return;
	}
	if(tw->next)
		tw->next->pprev = tw->pprev;
	*(tw->pprev) = tw->next;
	tw->pprev = NULL;
	write_unlock(&ehead->lock);

	/* Disassociate with bind bucket. */
	bhead = &tcp_bhash[tcp_bhashfn(tw->num)];
	spin_lock(&bhead->lock);
	if ((tb = tw->tb) != NULL) {
		if(tw->bind_next)
			tw->bind_next->bind_pprev = tw->bind_pprev;
		*(tw->bind_pprev) = tw->bind_next;
		tw->tb = NULL;
		if (tb->owners == NULL) {
			if (tb->next)
				tb->next->pprev = tb->pprev;
			*(tb->pprev) = tb->next;
			kmem_cache_free(tcp_bucket_cachep, tb);
		}
	}
	spin_unlock(&bhead->lock);

#ifdef INET_REFCNT_DEBUG
	if (atomic_read(&tw->refcnt) != 1) {
		printk(KERN_DEBUG "tw_bucket %p refcnt=%d\n", tw, atomic_read(&tw->refcnt));
	}
#endif
	tcp_tw_put(tw);
}

/* 
 * * Main purpose of TIME-WAIT state is to close connection gracefully,
 *   when one of ends sits in LAST-ACK or CLOSING retransmitting FIN
 *   (and, probably, tail of data) and one or more our ACKs are lost.
 * * What is TIME-WAIT timeout? It is associated with maximal packet
 *   lifetime in the internet, which results in wrong conclusion, that
 *   it is set to catch "old duplicate segments" wandering out of their path.
 *   It is not quite correct. This timeout is calculated so that it exceeds
 *   maximal retransmision timeout enough to allow to lose one (or more)
 *   segments sent by peer and our ACKs. This time may be calculated from RTO.
 * * When TIME-WAIT socket receives RST, it means that another end
 *   finally closed and we are allowed to kill TIME-WAIT too.
 * * Second purpose of TIME-WAIT is catching old duplicate segments.
 *   Well, certainly it is pure paranoia, but if we load TIME-WAIT
 *   with this semantics, we MUST NOT kill TIME-WAIT state with RSTs.
 * * If we invented some more clever way to catch duplicates
 *   (f.e. based on PAWS), we could truncate TIME-WAIT to several RTOs.
 *
 * The algorithm below is based on FORMAL INTERPRETATION of RFCs.
 * When you compare it to RFCs, please, read section SEGMENT ARRIVES
 * from the very beginning.
 *
 * NOTE. With recycling (and later with fin-wait-2) TW bucket
 * is _not_ stateless. It means, that strictly speaking we must
 * spinlock it. I do not want! Well, probability of misbehaviour
 * is ridiculously low and, seems, we could use some mb() tricks
 * to avoid misread sequence numbers, states etc.  --ANK
 */
enum tcp_tw_status
tcp_timewait_state_process(struct tcp_tw_bucket *tw, struct sk_buff *skb,
			   struct tcphdr *th, unsigned len)
{
	struct tcp_opt tp;
	int paws_reject = 0;

	tp.saw_tstamp = 0;
	if (th->doff > (sizeof(struct tcphdr)>>2) && tw->ts_recent_stamp) {
		tcp_parse_options(skb, &tp, 0);

		if (tp.saw_tstamp) {
			tp.ts_recent = tw->ts_recent;
			tp.ts_recent_stamp = tw->ts_recent_stamp;
			paws_reject = tcp_paws_check(&tp, th->rst);
		}
	}

	if (tw->substate == TCP_FIN_WAIT2) {
		/* Just repeat all the checks of tcp_rcv_state_process() */

		/* Out of window, send ACK */
		if (paws_reject ||
		    !tcp_in_window(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq,
				   tw->rcv_nxt, tw->rcv_nxt + tw->rcv_wnd))
			return TCP_TW_ACK;

		if (th->rst)
			goto kill;

		if (th->syn && TCP_SKB_CB(skb)->seq != tw->syn_seq)
			goto kill_with_rst;

		/* Dup ACK? */
		if (!after(TCP_SKB_CB(skb)->end_seq, tw->rcv_nxt) ||
		    TCP_SKB_CB(skb)->end_seq == TCP_SKB_CB(skb)->seq) {
			tcp_tw_put(tw);
			return TCP_TW_SUCCESS;
		}

		/* New data or FIN. If new data arrive after half-duplex close,
		 * reset.
		 */
		if (!th->fin || TCP_SKB_CB(skb)->end_seq != tw->rcv_nxt+1) {
kill_with_rst:
			tcp_tw_deschedule(tw);
			tcp_timewait_kill(tw);
			tcp_tw_put(tw);
			return TCP_TW_RST;
		}

		/* FIN arrived, enter true time-wait state. */
		tw->substate = TCP_TIME_WAIT;
		tw->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
		if (tp.saw_tstamp) {
			tw->ts_recent_stamp = xtime.tv_sec;
			tw->ts_recent = tp.rcv_tsval;
		}

		/* I am shamed, but failed to make it more elegant.
		 * Yes, it is direct reference to IP, which is impossible
		 * to generalize to IPv6. Taking into account that IPv6
		 * do not undertsnad recycling in any case, it not
		 * a big problem in practice. --ANK */
		if (tw->family == AF_INET &&
		    sysctl_tcp_tw_recycle && tw->ts_recent_stamp &&
		    tcp_v4_tw_remember_stamp(tw))
			tcp_tw_schedule(tw, tw->timeout);
		else
			tcp_tw_schedule(tw, TCP_TIMEWAIT_LEN);
		return TCP_TW_ACK;
	}

	/*
	 *	Now real TIME-WAIT state.
	 *
	 *	RFC 1122:
	 *	"When a connection is [...] on TIME-WAIT state [...]
	 *	[a TCP] MAY accept a new SYN from the remote TCP to
	 *	reopen the connection directly, if it:
	 *	
	 *	(1)  assigns its initial sequence number for the new
	 *	connection to be larger than the largest sequence
	 *	number it used on the previous connection incarnation,
	 *	and
	 *
	 *	(2)  returns to TIME-WAIT state if the SYN turns out 
	 *	to be an old duplicate".
	 */

	if (!paws_reject &&
	    (TCP_SKB_CB(skb)->seq == tw->rcv_nxt &&
	     (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq || th->rst))) {
		/* In window segment, it may be only reset or bare ack. */

		if (th->rst) {
			/* This is TIME_WAIT assasination, in two flavors.
			 * Oh well... nobody has a sufficient solution to this
			 * protocol bug yet.
			 */
			if (sysctl_tcp_rfc1337 == 0) {
kill:
				tcp_tw_deschedule(tw);
				tcp_timewait_kill(tw);
				tcp_tw_put(tw);
				return TCP_TW_SUCCESS;
			}
		}
		tcp_tw_schedule(tw, TCP_TIMEWAIT_LEN);

		if (tp.saw_tstamp) {
			tw->ts_recent = tp.rcv_tsval;
			tw->ts_recent_stamp = xtime.tv_sec;
		}

		tcp_tw_put(tw);
		return TCP_TW_SUCCESS;
	}

	/* Out of window segment.

	   All the segments are ACKed immediately.

	   The only exception is new SYN. We accept it, if it is
	   not old duplicate and we are not in danger to be killed
	   by delayed old duplicates. RFC check is that it has
	   newer sequence number works at rates <40Mbit/sec.
	   However, if paws works, it is reliable AND even more,
	   we even may relax silly seq space cutoff.

	   RED-PEN: we violate main RFC requirement, if this SYN will appear
	   old duplicate (i.e. we receive RST in reply to SYN-ACK),
	   we must return socket to time-wait state. It is not good,
	   but not fatal yet.
	 */

	if (th->syn && !th->rst && !th->ack && !paws_reject &&
	    (after(TCP_SKB_CB(skb)->seq, tw->rcv_nxt) ||
	     (tp.saw_tstamp && (s32)(tw->ts_recent - tp.rcv_tsval) < 0))) {
		u32 isn = tw->snd_nxt+65535+2;
		if (isn == 0)
			isn++;
		TCP_SKB_CB(skb)->when = isn;
		return TCP_TW_SYN;
	}

	if (paws_reject)
		NET_INC_STATS_BH(PAWSEstabRejected);

	if(!th->rst) {
		/* In this case we must reset the TIMEWAIT timer.
		 *
		 * If it is ACKless SYN it may be both old duplicate
		 * and new good SYN with random sequence number <rcv_nxt.
		 * Do not reschedule in the last case.
		 */
		if (paws_reject || th->ack)
			tcp_tw_schedule(tw, TCP_TIMEWAIT_LEN);

		/* Send ACK. Note, we do not put the bucket,
		 * it will be released by caller.
		 */
		return TCP_TW_ACK;
	}
	tcp_tw_put(tw);
	return TCP_TW_SUCCESS;
}

/* 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.
 */
static void __tcp_tw_hashdance(struct sock *sk, struct tcp_tw_bucket *tw)
{
	struct tcp_ehash_bucket *ehead = &tcp_ehash[sk->hashent];
	struct tcp_bind_hashbucket *bhead;
	struct sock **head, *sktw;

	write_lock(&ehead->lock);

	/* Step 1: Remove SK from established hash. */
	if (sk->pprev) {
		if(sk->next)
			sk->next->pprev = sk->pprev;
		*sk->pprev = sk->next;
		sk->pprev = NULL;
		sock_prot_dec_use(sk->prot);
	}

	/* Step 2: Hash TW into TIMEWAIT half of established hash table. */
	head = &(ehead + tcp_ehash_size)->chain;
	sktw = (struct sock *)tw;
	if((sktw->next = *head) != NULL)
		(*head)->pprev = &sktw->next;
	*head = sktw;
	sktw->pprev = head;
	atomic_inc(&tw->refcnt);

	write_unlock(&ehead->lock);

	/* Step 3: Put TW into bind hash. Original socket stays there too.
	   Note, that any socket with sk->num!=0 MUST be bound in binding
	   cache, even if it is closed.
	 */
	bhead = &tcp_bhash[tcp_bhashfn(sk->num)];
	spin_lock(&bhead->lock);
	tw->tb = (struct tcp_bind_bucket *)sk->prev;
	BUG_TRAP(sk->prev!=NULL);
	if ((tw->bind_next = tw->tb->owners) != NULL)
		tw->tb->owners->bind_pprev = &tw->bind_next;
	tw->tb->owners = (struct sock*)tw;
	tw->bind_pprev = &tw->tb->owners;
	spin_unlock(&bhead->lock);
}

/* 
 * Move a socket to time-wait or dead fin-wait-2 state.
 */ 
void tcp_time_wait(struct sock *sk, int state, int timeo)
{
	struct tcp_tw_bucket *tw = NULL;
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	int recycle_ok = 0;

	if (sysctl_tcp_tw_recycle && tp->ts_recent_stamp)
		recycle_ok = tp->af_specific->remember_stamp(sk);

	if (tcp_tw_count < sysctl_tcp_max_tw_buckets)
		tw = kmem_cache_alloc(tcp_timewait_cachep, SLAB_ATOMIC);

	if(tw != NULL) {
		int rto = (tp->rto<<2) - (tp->rto>>1);

		/* Give us an identity. */
		tw->daddr	= sk->daddr;
		tw->rcv_saddr	= sk->rcv_saddr;
		tw->bound_dev_if= sk->bound_dev_if;
		tw->num		= sk->num;
		tw->state	= TCP_TIME_WAIT;
		tw->substate	= state;
		tw->sport	= sk->sport;
		tw->dport	= sk->dport;
		tw->family	= sk->family;
		tw->reuse	= sk->reuse;
		tw->rcv_wscale	= tp->rcv_wscale;
		atomic_set(&tw->refcnt, 0);

		tw->hashent	= sk->hashent;
		tw->rcv_nxt	= tp->rcv_nxt;
		tw->snd_nxt	= tp->snd_nxt;
		tw->rcv_wnd	= tcp_receive_window(tp);
		tw->syn_seq	= tp->syn_seq;
		tw->ts_recent	= tp->ts_recent;
		tw->ts_recent_stamp= tp->ts_recent_stamp;
		tw->pprev_death = NULL;

#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
		if(tw->family == PF_INET6) {
			memcpy(&tw->v6_daddr,
			       &sk->net_pinfo.af_inet6.daddr,
			       sizeof(struct in6_addr));
			memcpy(&tw->v6_rcv_saddr,
			       &sk->net_pinfo.af_inet6.rcv_saddr,
			       sizeof(struct in6_addr));
		}
#endif
		/* Linkage updates. */
		__tcp_tw_hashdance(sk, tw);

		/* Get the TIME_WAIT timeout firing. */
		if (timeo < rto)
			timeo = rto;

		if (recycle_ok) {
			tw->timeout = rto;
		} else {
			tw->timeout = TCP_TIMEWAIT_LEN;
			if (state == TCP_TIME_WAIT)
				timeo = TCP_TIMEWAIT_LEN;
		}

		tcp_tw_schedule(tw, timeo);
	} else {
		/* Sorry, if we're out of memory, just CLOSE this
		 * socket up.  We've got bigger problems than
		 * non-graceful socket closings.
		 */
		if (net_ratelimit())
			printk(KERN_INFO "TCP: time wait bucket table overflow\n");
	}

	tcp_update_metrics(sk);
	tcp_done(sk);
}

/* Kill off TIME_WAIT sockets once their lifetime has expired. */
static int tcp_tw_death_row_slot = 0;

static void tcp_twkill(unsigned long);

static struct tcp_tw_bucket *tcp_tw_death_row[TCP_TWKILL_SLOTS];
static spinlock_t tw_death_lock = SPIN_LOCK_UNLOCKED;
static struct timer_list tcp_tw_timer = { function: tcp_twkill };

static void SMP_TIMER_NAME(tcp_twkill)(unsigned long dummy)
{
	struct tcp_tw_bucket *tw;
	int killed = 0;

	/* 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.
	 */
	spin_lock(&tw_death_lock);

	if (tcp_tw_count == 0)
		goto out;

	while((tw = tcp_tw_death_row[tcp_tw_death_row_slot]) != NULL) {
		tcp_tw_death_row[tcp_tw_death_row_slot] = tw->next_death;
		tw->pprev_death = NULL;
		spin_unlock(&tw_death_lock);

		tcp_timewait_kill(tw);
		tcp_tw_put(tw);

		killed++;

		spin_lock(&tw_death_lock);
	}
	tcp_tw_death_row_slot =
		((tcp_tw_death_row_slot + 1) & (TCP_TWKILL_SLOTS - 1));

	if ((tcp_tw_count -= killed) != 0)
		mod_timer(&tcp_tw_timer, jiffies+TCP_TWKILL_PERIOD);
	net_statistics[smp_processor_id()*2].TimeWaited += killed;
out:
	spin_unlock(&tw_death_lock);
}

SMP_TIMER_DEFINE(tcp_twkill, tcp_twkill_task);

/* 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 tcp_tw_deschedule(struct tcp_tw_bucket *tw)
{
	spin_lock(&tw_death_lock);
	if (tw->pprev_death) {
		if(tw->next_death)
			tw->next_death->pprev_death = tw->pprev_death;
		*tw->pprev_death = tw->next_death;
		tw->pprev_death = NULL;
		tcp_tw_put(tw);
		if (--tcp_tw_count == 0)
			del_timer(&tcp_tw_timer);