tcp.c

GNU Hurd 源代码

	  	return(0);
	}

	counted = sk->tp_pinfo.af_tcp.copied_seq;	/* Where we are at the moment */
	amount = 0;

	/* Do until a push or until we are out of data. */
	do {
		/* Found a hole so stops here. */
		if (before(counted, TCP_SKB_CB(skb)->seq))	/* should not happen */
			break;

		/* Length - header but start from where we are up to
		 * avoid overlaps.
		 */
		sum = skb->len - (counted - TCP_SKB_CB(skb)->seq);
		if (sum >= 0) {
			/* Add it up, move on. */
			amount += sum;
			counted += sum;
			if (skb->h.th->syn)
				counted++;
		}

		/* Don't count urg data ... but do it in the right place!
		 * Consider: "old_data (ptr is here) URG PUSH data"
		 * The old code would stop at the first push because
		 * it counted the urg (amount==1) and then does amount--
		 * *after* the loop.  This means tcp_readable() always
		 * returned zero if any URG PUSH was in the queue, even
		 * though there was normal data available. If we subtract
		 * the urg data right here, we even get it to work for more
		 * than one URG PUSH skb without normal data.
		 * This means that poll() finally works now with urg data
		 * in the queue.  Note that rlogin was never affected
		 * because it doesn't use poll(); it uses two processes
		 * and a blocking read().  And the queue scan in tcp_read()
		 * was correct.  Mike <[e-mail]>
		 */

		/* Don't count urg data. */
		if (skb->h.th->urg)
			amount--;
#if 0
		if (amount && skb->h.th->psh) break;
#endif
		skb = skb->next;
	} while(skb != (struct sk_buff *)&sk->receive_queue);

	SOCK_DEBUG(sk, "got %lu bytes.\n",amount);
	return(amount);
}

/*
 * LISTEN is a special case for poll..
 */
static unsigned int tcp_listen_poll(struct sock *sk, poll_table *wait)
{
	struct open_request *req, *dummy;

	lock_sock(sk);
	req = tcp_find_established(&sk->tp_pinfo.af_tcp, &dummy);
	release_sock(sk);
	if (req)
		return POLLIN | POLLRDNORM;
	return 0;
}

/*
 *	Compute minimal free write space needed to queue new packets.
 */
#define tcp_min_write_space(__sk) \
	(atomic_read(&(__sk)->wmem_alloc) / 2)

/*
 *	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);

	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.
	 */
	if (sk->shutdown & RCV_SHUTDOWN)
		mask |= POLLHUP;

	/* Connected? */
	if ((1 << sk->state) & ~(TCPF_SYN_SENT|TCPF_SYN_RECV)) {
		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 (sock_wspace(sk) >= tcp_min_write_space(sk)) {
				mask |= POLLOUT | POLLWRNORM;
			} else {  /* send SIGIO later */
				sk->socket->flags |= SO_NOSPACE;
			}
		}

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

/*
 *	Socket write_space callback.
 *	This (or rather the sock_wake_async) should agree with poll.
 */
void tcp_write_space(struct sock *sk)
{
	if (sk->dead)
		return;

	wake_up_interruptible(sk->sleep);
	if (sock_wspace(sk) >=
	    tcp_min_write_space(sk))
		sock_wake_async(sk->socket, 2);
}


#ifdef _HURD_

#define tcp_ioctl 0

error_t
tcp_tiocinq(struct sock *sk, mach_msg_type_number_t *amount)
{
  if (sk->state == TCP_LISTEN)
    return EINVAL;
  lock_sock(sk);
  *amount = tcp_readable(sk);
  release_sock(sk);
  return 0;
}

#else

int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg)
{
	int answ;

	switch(cmd) {
	case TIOCINQ:
#ifdef FIXME	/* FIXME: */
	case FIONREAD:
#endif
		if (sk->state == TCP_LISTEN)
			return(-EINVAL);
		lock_sock(sk);
		answ = tcp_readable(sk);
		release_sock(sk);
		break;
	case SIOCATMARK:
		{
			struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
			answ = tp->urg_data && tp->urg_seq == tp->copied_seq;
			break;
		}
	case TIOCOUTQ:
		if (sk->state == TCP_LISTEN)
			return(-EINVAL);
		answ = sock_wspace(sk);
		break;
	default:
		return(-ENOIOCTLCMD);
	};

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

#endif

/*
 *	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)
{
	struct task_struct *tsk = current;
	struct wait_queue wait = { tsk, NULL };

	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(flags & MSG_DONTWAIT)
			return -EAGAIN;
		if(signal_pending(tsk))
			return -ERESTARTSYS;

		tsk->state = TASK_INTERRUPTIBLE;
		add_wait_queue(sk->sleep, &wait);
		release_sock(sk);

		if (((1 << sk->state) & ~(TCPF_ESTABLISHED|TCPF_CLOSE_WAIT)) &&
		    sk->err == 0)
			schedule();

		tsk->state = TASK_RUNNING;
		remove_wait_queue(sk->sleep, &wait);
		lock_sock(sk);
	}
	return 0;
}

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

/*
 *	Wait for more memory for a socket
 */
static void wait_for_tcp_memory(struct sock * sk)
{
	release_sock(sk);
	if (!tcp_memory_free(sk)) {
		struct wait_queue wait = { current, NULL };

		sk->socket->flags &= ~SO_NOSPACE;
		add_wait_queue(sk->sleep, &wait);
		for (;;) {
			if (signal_pending(current))
				break;
			current->state = TASK_INTERRUPTIBLE;
			if (tcp_memory_free(sk))
				break;
			if (sk->shutdown & SEND_SHUTDOWN)
				break;
			if (sk->err)
				break;
			schedule();
		}
		current->state = TASK_RUNNING;
		remove_wait_queue(sk->sleep, &wait);
	}
	lock_sock(sk);
}

/*
 * Wait for a buffer.
 */
static int wait_for_buffer(struct sock *sk)
{
	struct wait_queue wait = { current, NULL };

	release_sock(sk);
	add_wait_queue(sk->sleep, &wait);
	current->state = TASK_INTERRUPTIBLE;
	schedule();
	current->state = TASK_RUNNING;
	remove_wait_queue(sk->sleep, &wait);
	lock_sock(sk);
	return 0;
}

/* 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.
 *
 *	Note: must be called with the socket locked.
 */

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

	lock_sock(sk);

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

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

	/* This should be in poll */
	sk->socket->flags &= ~SO_NOSPACE; /* clear SIGIO XXX */

	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++;

		while(seglen > 0) {
			int copy, tmp, queue_it, psh;

			if (err)
				goto do_fault2;

			/* Stop on errors. */
			if (sk->err)
				goto do_sock_err;

			/* Make sure that we are established. */
			if (sk->shutdown & SEND_SHUTDOWN)
				goto do_shutdown;

			/* Now we need to check if we have a half
			 * built packet we can tack some data onto.
			 */
			if (tp->send_head && !(flags & MSG_OOB)) {
				skb = sk->write_queue.prev;
				copy = skb->len;
				/* If the remote does SWS avoidance we should
				 * queue the best we can if not we should in
				 * fact send multiple packets...
				 * A method for detecting this would be most
				 * welcome.
				 */
				if (skb_tailroom(skb) > 0 &&
				    (mss_now - copy) > 0 &&
				    tp->snd_nxt < TCP_SKB_CB(skb)->end_seq) {
					int last_byte_was_odd = (copy % 4);

					/*
					 * Check for parallel writers sleeping in user access.
					 */
					if (tp->partial_writers++ > 0) {
						wait_for_buffer(sk);
						tp->partial_writers--;
						continue;
					}

					copy = mss_now - copy;
					if(copy > skb_tailroom(skb))
						copy = skb_tailroom(skb);
					if(copy > seglen)
						copy = seglen;

					if(last_byte_was_odd) {
						if(copy_from_user(skb_put(skb, copy),
								  from, copy))
							err = -EFAULT;
						skb->csum = csum_partial(skb->data,
									 skb->len, 0);
					} else {
						skb->csum =
							csum_and_copy_from_user(
							from, skb_put(skb, copy),
							copy, skb->csum, &err);
					}

					/*
					 * FIXME: the *_user functions should
					 *	  return how much data was
					 *	  copied before the fault
					 *	  occurred and then a partial
					 *	  packet with this data should
					 *	  be sent.  Unfortunately
					 *	  csum_and_copy_from_user doesn't
					 *	  return this information.
					 *	  ATM it might send partly zeroed
					 *	  data in this case.
					 */
					tp->write_seq += copy;
					TCP_SKB_CB(skb)->end_seq += copy;
					from += copy;
					copied += copy;
					seglen -= copy;
					if (PSH_NEEDED)
						TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;

					if (--tp->partial_writers > 0)
						wake_up_interruptible(sk->sleep);

					continue;
				}
			}

			/* We also need to worry about the window.  If
			 * window < 1/2 the maximum window we've seen
			 * from this host, don't use it.  This is
			 * sender side silly window prevention, as
			 * specified in RFC1122.  (Note that this is
			 * different than earlier versions of SWS
			 * prevention, e.g. RFC813.).  What we
			 * actually do is use the whole MSS.  Since
			 * the results in the right edge of the packet
			 * being outside the window, it will be queued
			 * for later rather than sent.
			 */
			psh = 0;
			copy = tp->snd_wnd - (tp->snd_nxt - tp->snd_una);
			if(copy > (tp->max_window >> 1)) {
				copy = min(copy, mss_now);
				psh = 1;
			} else {
				copy = mss_now;
			}
			if(copy > seglen)
				copy = seglen;

			/* Determine how large of a buffer to allocate.  */
			tmp = MAX_HEADER + sk->prot->max_header;
			if (copy < min(mss_now, tp->max_window >> 1) &&
			    !(flags & MSG_OOB)) {
				tmp += min(mss_now, tp->max_window);

				/* What is happening here is that we want to
				 * tack on later members of the users iovec
				 * if possible into a single frame.  When we
				 * leave this loop our caller checks to see if
				 * we can send queued frames onto the wire.
				 * See tcp_v[46]_sendmsg() for this.
				 */
				queue_it = 1;
			} else {
				tmp += copy;
				queue_it = 0;
			}
			skb = sock_wmalloc(sk, tmp, 0, GFP_KERNEL);

			/* If we didn't get any memory, we need to sleep. */
			if (skb == NULL) {
				sk->socket->flags |= SO_NOSPACE;
				if (flags&MSG_DONTWAIT) {
					err = -EAGAIN;
					goto do_interrupted;