tcp.c

GNU Hurd 源代码

static unsigned char new_state[16] = {
  /* current state:        new state:      action:	*/
  /* (Invalid)		*/ TCP_CLOSE,
  /* TCP_ESTABLISHED	*/ TCP_FIN_WAIT1 | TCP_ACTION_FIN,
  /* TCP_SYN_SENT	*/ TCP_CLOSE,
  /* TCP_SYN_RECV	*/ TCP_FIN_WAIT1 | TCP_ACTION_FIN,
  /* TCP_FIN_WAIT1	*/ TCP_FIN_WAIT1,
  /* TCP_FIN_WAIT2	*/ TCP_FIN_WAIT2,
  /* TCP_TIME_WAIT	*/ TCP_CLOSE,
  /* TCP_CLOSE		*/ TCP_CLOSE,
  /* TCP_CLOSE_WAIT	*/ TCP_LAST_ACK  | TCP_ACTION_FIN,
  /* TCP_LAST_ACK	*/ TCP_LAST_ACK,
  /* TCP_LISTEN		*/ TCP_CLOSE,
  /* TCP_CLOSING	*/ TCP_CLOSING,
};

static int tcp_close_state(struct sock *sk, int dead)
{
	int next = (int) new_state[sk->state];
	int ns = (next & TCP_STATE_MASK);

	tcp_set_state(sk, ns);

	/*	This is a (useful) BSD violating of the RFC. There is a
	 *	problem with TCP as specified in that the other end could
	 *	keep a socket open forever with no application left this end.
	 *	We use a 3 minute timeout (about the same as BSD) then kill
	 *	our end. If they send after that then tough - BUT: long enough
	 *	that we won't make the old 4*rto = almost no time - whoops
	 *	reset mistake.
	 */
	if (dead)
		tcp_check_fin_timer(sk);

	return (next & TCP_ACTION_FIN);
}

/*
 *	Shutdown the sending side of a connection. Much like close except
 *	that we don't receive shut down or set sk->dead.
 */

void tcp_shutdown(struct sock *sk, int how)
{
	/*	We need to grab some memory, and put together a FIN,
	 *	and then put it into the queue to be sent.
	 *		Tim MacKenzie([e-mail]) 4 Dec '92.
	 */
	if (!(how & SEND_SHUTDOWN))
		return;

	/* If we've already sent a FIN, or it's a closed state, skip this. */
	if ((1 << sk->state) &
	    (TCPF_ESTABLISHED|TCPF_SYN_SENT|TCPF_SYN_RECV|TCPF_CLOSE_WAIT)) {
		lock_sock(sk);

		/* Clear out any half completed packets.  FIN if needed. */
		if (tcp_close_state(sk,0))
			tcp_send_fin(sk);

		release_sock(sk);
	}
}


/*
 *	Return 1 if we still have things to send in our buffers.
 */

static inline int closing(struct sock * sk)
{
	return ((1 << sk->state) & (TCPF_FIN_WAIT1|TCPF_CLOSING|TCPF_LAST_ACK));
}

/*
 *	This routine closes sockets which have been at least partially
 *	opened, but not yet accepted. Currently it is only called by
 *	tcp_close, and timeout mirrors the value there.
 */

static void tcp_close_pending (struct sock *sk)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	struct open_request *req = tp->syn_wait_queue;

	while(req) {
		struct open_request *iter;

		if (req->sk)
			tcp_close(req->sk, 0);

		iter = req;
		req = req->dl_next;

		(*iter->class->destructor)(iter);
		tcp_dec_slow_timer(TCP_SLT_SYNACK);
		sk->ack_backlog--;
		tcp_openreq_free(iter);
	}

	tcp_synq_init(tp);
}

void tcp_close(struct sock *sk, long timeout)
{
	struct sk_buff *skb;
	int data_was_unread = 0;

	/*
	 * Check whether the socket is locked ... supposedly
	 * it's impossible to tcp_close() a locked socket.
	 */
	if (atomic_read(&sk->sock_readers))
		printk("tcp_close: socket already locked!\n");

	/* We need to grab some memory, and put together a FIN,
	 * and then put it into the queue to be sent.
	 */
	lock_sock(sk);
	if(sk->state == TCP_LISTEN) {
		/* Special case. */
		tcp_set_state(sk, TCP_CLOSE);
		tcp_close_pending(sk);
		release_sock(sk);
		sk->dead = 1;
		return;
	}

	/* It is questionable, what the role of this is now.
	 * In any event either it should be removed, or
	 * increment of SLT_KEEPALIVE be done, this is causing
	 * big problems.  For now I comment it out.  -DaveM
	 */
	/* sk->keepopen = 1; */
	sk->shutdown = SHUTDOWN_MASK;

	if (!sk->dead)
	  	sk->state_change(sk);

	/*  We need to flush the recv. buffs.  We do this only on the
	 *  descriptor close, not protocol-sourced closes, because the
	 *  reader process may not have drained the data yet!
	 */
	while((skb=__skb_dequeue(&sk->receive_queue))!=NULL) {
		u32 len = TCP_SKB_CB(skb)->end_seq - TCP_SKB_CB(skb)->seq - skb->h.th->fin;
		data_was_unread += len;
		kfree_skb(skb);
	}

	/* As outlined in draft-ietf-tcpimpl-prob-03.txt, section
	 * 3.10, we send a RST here because data was lost.  To
	 * witness the awful effects of the old behavior of always
	 * doing a FIN, run an older 2.1.x kernel or 2.0.x, start
	 * a bulk GET in an FTP client, suspend the process, wait
	 * for the client to advertise a zero window, then kill -9
	 * the FTP client, wheee...  Note: timeout is always zero
	 * in such a case.
	 */
	if(data_was_unread != 0) {
		/* Unread data was tossed, zap the connection. */
		tcp_set_state(sk, TCP_CLOSE);
		tcp_send_active_reset(sk);
	} else if (tcp_close_state(sk,1)) {
		/* We FIN if the application ate all the data before
		 * zapping the connection.
		 */
		tcp_send_fin(sk);
	}

	if (timeout) {
		struct task_struct *tsk = current;
		struct wait_queue wait = { tsk, NULL };

		add_wait_queue(sk->sleep, &wait);
		release_sock(sk);

		while (1) {
			tsk->state = TASK_INTERRUPTIBLE;
			if (!closing(sk))
				break;
			timeout = schedule_timeout(timeout);
			if (signal_pending(tsk) || !timeout)
				break;
		}

		tsk->state = TASK_RUNNING;
		remove_wait_queue(sk->sleep, &wait);

		lock_sock(sk);
	}

	/* Now that the socket is dead, if we are in the FIN_WAIT2 state
	 * we may need to set up a timer.
         */
	tcp_check_fin_timer(sk);

	release_sock(sk);
	sk->dead = 1;
}

/*
 *	Wait for an incoming connection, avoid race
 *	conditions. This must be called with the socket locked.
 */
static struct open_request * wait_for_connect(struct sock * sk,
					      struct open_request **pprev)
{
	struct wait_queue wait = { current, NULL };
	struct open_request *req;

	add_wait_queue(sk->sleep, &wait);
	for (;;) {
		current->state = TASK_INTERRUPTIBLE;
		release_sock(sk);
		schedule();
		lock_sock(sk);
		req = tcp_find_established(&(sk->tp_pinfo.af_tcp), pprev);
		if (req)
			break;
		if (signal_pending(current))
			break;
	}
	current->state = TASK_RUNNING;
	remove_wait_queue(sk->sleep, &wait);
	return req;
}

/*
 *	This will accept the next outstanding connection.
 *
 *	Be careful about race conditions here - this is subtle.
 */

struct sock *tcp_accept(struct sock *sk, int flags)
{
	struct tcp_opt *tp = &sk->tp_pinfo.af_tcp;
	struct open_request *req, *prev;
	struct sock *newsk = NULL;
	int error;

	lock_sock(sk);

	/* We need to make sure that this socket is listening,
	 * and that it has something pending.
	 */
	error = EINVAL;
	if (sk->state != TCP_LISTEN)
		goto out;

	/* Find already established connection */
	req = tcp_find_established(tp, &prev);
	if (!req) {
		/* If this is a non blocking socket don't sleep */
		error = EAGAIN;
		if (flags & O_NONBLOCK)
			goto out;

		error = ERESTARTSYS;
		req = wait_for_connect(sk, &prev);
		if (!req)
			goto out;
	}

	tcp_synq_unlink(tp, req, prev);
	newsk = req->sk;
	req->class->destructor(req);
	tcp_openreq_free(req);
	sk->ack_backlog--;
	if(sk->keepopen)
		tcp_inc_slow_timer(TCP_SLT_KEEPALIVE);

	release_sock(sk);
	return newsk;

out:
	/* sk should be in LISTEN state, thus accept can use sk->err for
	 * internal purposes without stomping one anyone's feed.
	 */
	sk->err = error;
	release_sock(sk);
	return newsk;
}

/*
 *	Socket option code for TCP.
 */

int tcp_setsockopt(struct sock *sk, int level, int optname, char *optval,
		   int optlen)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	int val;

	if (level != SOL_TCP)
		return tp->af_specific->setsockopt(sk, level, optname,
						   optval, optlen);

	if(optlen<sizeof(int))
		return -EINVAL;

  	if (get_user(val, (int *)optval))
		return -EFAULT;

	switch(optname) {
	case TCP_MAXSEG:
		/* values greater than interface MTU won't take effect.  however at
		 * the point when this call is done we typically don't yet know
		 * which interface is going to be used
		 */
		if(val < 1 || val > MAX_WINDOW)
			return -EINVAL;
		tp->user_mss = val;
		return 0;

	case TCP_NODELAY:
		/* You cannot try to use this and TCP_CORK in
		 * tandem, so let the user know.
		 */
		if (sk->nonagle == 2)
			return -EINVAL;
		sk->nonagle = (val == 0) ? 0 : 1;
		return 0;

	case TCP_CORK:
		/* When set indicates to always queue non-full frames.
		 * Later the user clears this option and we transmit
		 * any pending partial frames in the queue.  This is
		 * meant to be used alongside sendfile() to get properly
		 * filled frames when the user (for example) must write
		 * out headers with a write() call first and then use
		 * sendfile to send out the data parts.
		 *
		 * You cannot try to use TCP_NODELAY and this mechanism
		 * at the same time, so let the user know.
		 */
		if (sk->nonagle == 1)
			return -EINVAL;
		if (val != 0) {
			sk->nonagle = 2;
		} else {
			sk->nonagle = 0;

			lock_sock(sk);
			tcp_push_pending_frames(sk, tp);
			release_sock(sk);
		}
		return 0;

	default:
		return -ENOPROTOOPT;
	};
}

int tcp_getsockopt(struct sock *sk, int level, int optname, char *optval,
		   int *optlen)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	int val, len;

	if(level != SOL_TCP)
		return tp->af_specific->getsockopt(sk, level, optname,
						   optval, optlen);

	if(get_user(len,optlen))
		return -EFAULT;

	len = min(len, sizeof(int));

	switch(optname) {
	case TCP_MAXSEG:
		val = tp->user_mss;
		break;
	case TCP_NODELAY:
		val = (sk->nonagle == 1);
		break;
	case TCP_CORK:
		val = (sk->nonagle == 2);
		break;
	default:
		return -ENOPROTOOPT;
	};

  	if(put_user(len, optlen))
  		return -EFAULT;
	if(copy_to_user(optval, &val,len))
		return -EFAULT;
  	return 0;
}

void tcp_set_keepalive(struct sock *sk, int val)
{
	if (!sk->keepopen && val)
		tcp_inc_slow_timer(TCP_SLT_KEEPALIVE);
	else if (sk->keepopen && !val)
		tcp_dec_slow_timer(TCP_SLT_KEEPALIVE);
}

extern void __skb_cb_too_small_for_tcp(int, int);

void __init tcp_init(void)
{
	struct sk_buff *skb = NULL;
	unsigned long goal;
	int order;

	if(sizeof(struct tcp_skb_cb) > sizeof(skb->cb))
		__skb_cb_too_small_for_tcp(sizeof(struct tcp_skb_cb),
					   sizeof(skb->cb));

	tcp_openreq_cachep = kmem_cache_create("tcp_open_request",
						   sizeof(struct open_request),
					       0, SLAB_HWCACHE_ALIGN,
					       NULL, NULL);
	if(!tcp_openreq_cachep)
		panic("tcp_init: Cannot alloc open_request cache.");

	tcp_bucket_cachep = kmem_cache_create("tcp_bind_bucket",
					      sizeof(struct tcp_bind_bucket),
					      0, SLAB_HWCACHE_ALIGN,
					      NULL, NULL);
	if(!tcp_bucket_cachep)
		panic("tcp_init: Cannot alloc tcp_bind_bucket cache.");

	tcp_timewait_cachep = kmem_cache_create("tcp_tw_bucket",
						sizeof(struct tcp_tw_bucket),
						0, SLAB_HWCACHE_ALIGN,
						NULL, NULL);
	if(!tcp_timewait_cachep)
		panic("tcp_init: Cannot alloc tcp_tw_bucket cache.");

	/* Size and allocate TCP hash tables. */
	goal = num_physpages >> (20 - PAGE_SHIFT);
	for (order = 0; (1UL << order) < goal; order++)
		;
	do {
		tcp_ehash_size = (1UL << order) * PAGE_SIZE /
			sizeof(struct sock *);
		tcp_ehash = (struct sock **)
			__get_free_pages(GFP_ATOMIC, order);
	} while (tcp_ehash == NULL && --order >= 0);

	if (!tcp_ehash)
		panic("Failed to allocate TCP established hash table\n");
	memset(tcp_ehash, 0, tcp_ehash_size * sizeof(struct sock *));

	goal = (((1UL << order) * PAGE_SIZE) / sizeof(struct tcp_bind_bucket *));
	if (goal > (64 * 1024)) {
		/* Don't size the bind-hash larger than the port
		 * space, that is just silly.
		 */
		goal = (((64 * 1024) * sizeof(struct tcp_bind_bucket *)) / PAGE_SIZE);
		for (order = 0; (1UL << order) < goal; order++)
			;
	}

	do {
		tcp_bhash_size = (1UL << order) * PAGE_SIZE /
			sizeof(struct tcp_bind_bucket *);
		tcp_bhash = (struct tcp_bind_bucket **)
			__get_free_pages(GFP_ATOMIC, order);
	} while (tcp_bhash == NULL && --order >= 0);

	if (!tcp_bhash)
		panic("Failed to allocate TCP bind hash table\n");
	memset(tcp_bhash, 0, tcp_bhash_size * sizeof(struct tcp_bind_bucket *));

	printk("TCP: Hash tables configured (ehash %d bhash %d)\n",
	       tcp_ehash_size, tcp_bhash_size);
}