tcp.c

讲述linux的初始化过程

	if (atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]) {
		tcp_enter_memory_pressure();
		goto suppress_allocation;
	}

	/* Under pressure. */
	if (atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[1])
		tcp_enter_memory_pressure();

	if (kind) {
		if (atomic_read(&sk->rmem_alloc) < sysctl_tcp_rmem[0])
			return 1;
	} else {
		if (sk->wmem_queued < sysctl_tcp_wmem[0])
			return 1;
	}

	if (!tcp_memory_pressure ||
	    sysctl_tcp_mem[2] > atomic_read(&tcp_sockets_allocated)
	    * TCP_PAGES(sk->wmem_queued+atomic_read(&sk->rmem_alloc)+
			sk->forward_alloc))
		return 1;

suppress_allocation:

	if (kind == 0) {
		tcp_moderate_sndbuf(sk);

		/* Fail only if socket is _under_ its sndbuf.
		 * In this case we cannot block, so that we have to fail.
		 */
		if (sk->wmem_queued+size >= sk->sndbuf)
			return 1;
	}

	/* Alas. Undo changes. */
	sk->forward_alloc -= amt*TCP_MEM_QUANTUM;
	atomic_sub(amt, &tcp_memory_allocated);
	return 0;
}

void __tcp_mem_reclaim(struct sock *sk)
{
	if (sk->forward_alloc >= TCP_MEM_QUANTUM) {
		atomic_sub(sk->forward_alloc/TCP_MEM_QUANTUM, &tcp_memory_allocated);
		sk->forward_alloc &= (TCP_MEM_QUANTUM-1);
		if (tcp_memory_pressure &&
		    atomic_read(&tcp_memory_allocated) < sysctl_tcp_mem[0])
			tcp_memory_pressure = 0;
	}
}

void tcp_rfree(struct sk_buff *skb)
{
	struct sock *sk = skb->sk;

	atomic_sub(skb->truesize, &sk->rmem_alloc);
	sk->forward_alloc += skb->truesize;
}

/*
 * LISTEN is a special case for poll..
 */
static __inline__ unsigned int tcp_listen_poll(struct sock *sk, poll_table *wait)
{
	return sk->tp_pinfo.af_tcp.accept_queue ? (POLLIN | POLLRDNORM) : 0;
}

/*
 *	Wait for a TCP event.
 *
 *	Note that we don't need to lock the socket, as the upper poll layers
 *	take care of normal races (between the test and the event) and we don't
 *	go look at any of the socket buffers directly.
 */
unsigned int tcp_poll(struct file * file, struct socket *sock, poll_table *wait)
{
	unsigned int mask;
	struct sock *sk = sock->sk;
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);

	poll_wait(file, sk->sleep, wait);
	if (sk->state == TCP_LISTEN)
		return tcp_listen_poll(sk, wait);

	/* Socket is not locked. We are protected from async events
	   by poll logic and correct handling of state changes
	   made by another threads is impossible in any case.
	 */

	mask = 0;
	if (sk->err)
		mask = POLLERR;

	/*
	 * POLLHUP is certainly not done right. But poll() doesn't
	 * have a notion of HUP in just one direction, and for a
	 * socket the read side is more interesting.
	 *
	 * Some poll() documentation says that POLLHUP is incompatible
	 * with the POLLOUT/POLLWR flags, so somebody should check this
	 * all. But careful, it tends to be safer to return too many
	 * bits than too few, and you can easily break real applications
	 * if you don't tell them that something has hung up!
	 *
	 * Check-me.
	 *
	 * Check number 1. POLLHUP is _UNMASKABLE_ event (see UNIX98 and
	 * our fs/select.c). It means that after we received EOF,
	 * poll always returns immediately, making impossible poll() on write()
	 * in state CLOSE_WAIT. One solution is evident --- to set POLLHUP
	 * if and only if shutdown has been made in both directions.
	 * Actually, it is interesting to look how Solaris and DUX
	 * solve this dilemma. I would prefer, if PULLHUP were maskable,
	 * then we could set it on SND_SHUTDOWN. BTW examples given
	 * in Stevens' books assume exactly this behaviour, it explains
	 * why PULLHUP is incompatible with POLLOUT.	--ANK
	 *
	 * NOTE. Check for TCP_CLOSE is added. The goal is to prevent
	 * blocking on fresh not-connected or disconnected socket. --ANK
	 */
	if (sk->shutdown == SHUTDOWN_MASK || sk->state == TCP_CLOSE)
		mask |= POLLHUP;
	if (sk->shutdown & RCV_SHUTDOWN)
		mask |= POLLIN | POLLRDNORM;

	/* Connected? */
	if ((1 << sk->state) & ~(TCPF_SYN_SENT|TCPF_SYN_RECV)) {
		/* Potential race condition. If read of tp below will
		 * escape above sk->state, we can be illegally awaken
		 * in SYN_* states. */
		if ((tp->rcv_nxt != tp->copied_seq) &&
		    (tp->urg_seq != tp->copied_seq ||
		     tp->rcv_nxt != tp->copied_seq+1 ||
		     sk->urginline || !tp->urg_data))
			mask |= POLLIN | POLLRDNORM;

		if (!(sk->shutdown & SEND_SHUTDOWN)) {
			if (tcp_wspace(sk) >= tcp_min_write_space(sk)) {
				mask |= POLLOUT | POLLWRNORM;
			} else {  /* send SIGIO later */
				set_bit(SOCK_ASYNC_NOSPACE, &sk->socket->flags);
				set_bit(SOCK_NOSPACE, &sk->socket->flags);

				/* Race breaker. If space is freed after
				 * wspace test but before the flags are set,
				 * IO signal will be lost.
				 */
				if (tcp_wspace(sk) >= tcp_min_write_space(sk))
					mask |= POLLOUT | POLLWRNORM;
			}
		}

		if (tp->urg_data & TCP_URG_VALID)
			mask |= POLLPRI;
	}
	return mask;
}

/*
 *	TCP socket write_space callback. Not used.
 */
void tcp_write_space(struct sock *sk)
{
}

int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	int answ;

	switch(cmd) {
	case SIOCINQ:
		if (sk->state == TCP_LISTEN)
			return(-EINVAL);

		lock_sock(sk);
		if ((1<<sk->state) & (TCPF_SYN_SENT|TCPF_SYN_RECV))
			answ = 0;
		else if (sk->urginline || !tp->urg_data ||
			 before(tp->urg_seq,tp->copied_seq) ||
			 !before(tp->urg_seq,tp->rcv_nxt)) {
			answ = tp->rcv_nxt - tp->copied_seq;

			/* Subtract 1, if FIN is in queue. */
			if (answ && !skb_queue_empty(&sk->receive_queue))
				answ -= ((struct sk_buff*)sk->receive_queue.prev)->h.th->fin;
		} else
			answ = tp->urg_seq - tp->copied_seq;
		release_sock(sk);
		break;
	case SIOCATMARK:
		{
			answ = tp->urg_data && tp->urg_seq == tp->copied_seq;
			break;
		}
	case SIOCOUTQ:
		if (sk->state == TCP_LISTEN)
			return(-EINVAL);

		if ((1<<sk->state) & (TCPF_SYN_SENT|TCPF_SYN_RECV))
			answ = 0;
		else
			answ = tp->write_seq - tp->snd_una;
		break;
	default:
		return(-ENOIOCTLCMD);
	};

	return put_user(answ, (int *)arg);
}


int tcp_listen_start(struct sock *sk)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	struct tcp_listen_opt *lopt;

	sk->max_ack_backlog = 0;
	sk->ack_backlog = 0;
	tp->accept_queue = tp->accept_queue_tail = NULL;
	tp->syn_wait_lock = RW_LOCK_UNLOCKED;

	lopt = kmalloc(sizeof(struct tcp_listen_opt), GFP_KERNEL);
	if (!lopt)
		return -ENOMEM;

	memset(lopt, 0, sizeof(struct tcp_listen_opt));
	for (lopt->max_qlen_log = 6; ; lopt->max_qlen_log++)
		if ((1<<lopt->max_qlen_log) >= sysctl_max_syn_backlog)
			break;

	write_lock_bh(&tp->syn_wait_lock);
	tp->listen_opt = lopt;
	write_unlock_bh(&tp->syn_wait_lock);

	/* There is race window here: we announce ourselves listening,
	 * but this transition is still not validated by get_port().
	 * It is OK, because this socket enters to hash table only
	 * after validation is complete.
	 */
	sk->state = TCP_LISTEN;
	if (sk->prot->get_port(sk, sk->num) == 0) {
		sk->sport = htons(sk->num);

		sk_dst_reset(sk);
		sk->prot->hash(sk);

		return 0;
	}

	sk->state = TCP_CLOSE;
	write_lock_bh(&tp->syn_wait_lock);
	tp->listen_opt = NULL;
	write_unlock_bh(&tp->syn_wait_lock);
	kfree(lopt);
	return -EADDRINUSE;
}

/*
 *	This routine closes sockets which have been at least partially
 *	opened, but not yet accepted.
 */

static void tcp_listen_stop (struct sock *sk)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	struct tcp_listen_opt *lopt = tp->listen_opt;
	struct open_request *acc_req = tp->accept_queue;
	struct open_request *req;
	int i;

	tcp_delete_keepalive_timer(sk);

	/* make all the listen_opt local to us */
	write_lock_bh(&tp->syn_wait_lock);
	tp->listen_opt =NULL;
	write_unlock_bh(&tp->syn_wait_lock);
	tp->accept_queue = tp->accept_queue_tail = NULL;

	if (lopt->qlen) {
		for (i=0; i<TCP_SYNQ_HSIZE; i++) {
			while ((req = lopt->syn_table[i]) != NULL) {
				lopt->syn_table[i] = req->dl_next;
				lopt->qlen--;
				tcp_openreq_free(req);

		/* Following specs, it would be better either to send FIN
		 * (and enter FIN-WAIT-1, it is normal close)
		 * or to send active reset (abort). 
		 * Certainly, it is pretty dangerous while synflood, but it is
		 * bad justification for our negligence 8)
		 * To be honest, we are not able to make either
		 * of the variants now.			--ANK
		 */
			}
		}
	}
	BUG_TRAP(lopt->qlen == 0);

	kfree(lopt);

	while ((req=acc_req) != NULL) {
		struct sock *child = req->sk;

		acc_req = req->dl_next;

		local_bh_disable();
		bh_lock_sock(child);
		BUG_TRAP(child->lock.users==0);
		sock_hold(child);

		tcp_disconnect(child, O_NONBLOCK);

		sock_orphan(child);

		atomic_inc(&tcp_orphan_count);

		tcp_destroy_sock(child);

		bh_unlock_sock(child);
		local_bh_enable();
		sock_put(child);

		tcp_acceptq_removed(sk);
		tcp_openreq_fastfree(req);
	}
	BUG_TRAP(sk->ack_backlog == 0);
}

/*
 *	Wait for a socket to get into the connected state
 *
 *	Note: Must be called with the socket locked.
 */
static int wait_for_tcp_connect(struct sock * sk, int flags, long *timeo_p)
{
	struct task_struct *tsk = current;
	DECLARE_WAITQUEUE(wait, tsk);

	while((1 << sk->state) & ~(TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)) {
		if(sk->err)
			return sock_error(sk);
		if((1 << sk->state) &
		   ~(TCPF_SYN_SENT | TCPF_SYN_RECV)) {
			if(sk->keepopen && !(flags&MSG_NOSIGNAL))
				send_sig(SIGPIPE, tsk, 0);
			return -EPIPE;
		}
		if(!*timeo_p)
			return -EAGAIN;
		if(signal_pending(tsk))
			return sock_intr_errno(*timeo_p);

		__set_task_state(tsk, TASK_INTERRUPTIBLE);
		add_wait_queue(sk->sleep, &wait);
		sk->tp_pinfo.af_tcp.write_pending++;

		release_sock(sk);
		*timeo_p = schedule_timeout(*timeo_p);
		lock_sock(sk);

		__set_task_state(tsk, TASK_RUNNING);
		remove_wait_queue(sk->sleep, &wait);
		sk->tp_pinfo.af_tcp.write_pending--;
	}
	return 0;
}

static inline int tcp_memory_free(struct sock *sk)
{
	return sk->wmem_queued < sk->sndbuf;
}

/*
 *	Wait for more memory for a socket
 */
static long wait_for_tcp_memory(struct sock * sk, long timeo)
{
	long vm_wait = 0;
	long current_timeo = timeo;
	DECLARE_WAITQUEUE(wait, current);

	if (tcp_memory_free(sk))
		current_timeo = vm_wait = (net_random()%(HZ/5))+2;

	clear_bit(SOCK_ASYNC_NOSPACE, &sk->socket->flags);

	add_wait_queue(sk->sleep, &wait);
	for (;;) {
		set_bit(SOCK_NOSPACE, &sk->socket->flags);

		set_current_state(TASK_INTERRUPTIBLE);

		if (signal_pending(current))
			break;
		if (tcp_memory_free(sk) && !vm_wait)
			break;
		if (sk->shutdown & SEND_SHUTDOWN)
			break;
		if (sk->err)
			break;
		release_sock(sk);
		if (!tcp_memory_free(sk) || vm_wait)
			current_timeo = schedule_timeout(current_timeo);
		lock_sock(sk);
		if (vm_wait) {
			if (timeo != MAX_SCHEDULE_TIMEOUT &&
			    (timeo -= vm_wait-current_timeo) < 0)
				timeo = 0;
			break;
		} else {
			timeo = current_timeo;
		}
	}
	current->state = TASK_RUNNING;
	remove_wait_queue(sk->sleep, &wait);
	return timeo;
}

/* When all user supplied data has been queued set the PSH bit */
#define PSH_NEEDED (seglen == 0 && iovlen == 0)

/*
 *	This routine copies from a user buffer into a socket,
 *	and starts the transmit system.
 */

int tcp_sendmsg(struct sock *sk, struct msghdr *msg, int size)
{
	struct iovec *iov;
	struct tcp_opt *tp;
	struct sk_buff *skb;
	int iovlen, flags;
	int mss_now;
	int err, copied;
	long timeo;

	err = 0;
	tp = &(sk->tp_pinfo.af_tcp);

	lock_sock(sk);
	TCP_CHECK_TIMER(sk);

	flags = msg->msg_flags;

	timeo = sock_sndtimeo(sk, flags&MSG_DONTWAIT);

	/* Wait for a connection to finish. */
	if ((1 << sk->state) & ~(TCPF_ESTABLISHED | TCPF_CLOSE_WAIT))
		if((err = wait_for_tcp_connect(sk, flags, &timeo)) != 0)
			goto out_unlock;

	/* This should be in poll */
	clear_bit(SOCK_ASYNC_NOSPACE, &sk->socket->flags);

	mss_now = tcp_current_mss(sk);

	/* Ok commence sending. */
	iovlen = msg->msg_iovlen;
	iov = msg->msg_iov;
	copied = 0;

	while (--iovlen >= 0) {
		int seglen=iov->iov_len;
		unsigned char * from=iov->iov_base;

		iov++;