ulpqueue.c

在linux环境下的流控制传输协议（sctp）的源代码

/* SCTP kernel implementation
 * (C) Copyright IBM Corp. 2001, 2004
 * Copyright (c) 1999-2000 Cisco, Inc.
 * Copyright (c) 1999-2001 Motorola, Inc.
 * Copyright (c) 2001 Intel Corp.
 * Copyright (c) 2001 Nokia, Inc.
 * Copyright (c) 2001 La Monte H.P. Yarroll
 *
 * This abstraction carries sctp events to the ULP (sockets).
 *
 * This SCTP implementation is free software;
 * you can redistribute it and/or modify it under the terms of
 * the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * This SCTP implementation is distributed in the hope that it
 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
 *                 ************************
 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 * See the GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with GNU CC; see the file COPYING.  If not, write to
 * the Free Software Foundation, 59 Temple Place - Suite 330,
 * Boston, MA 02111-1307, USA.
 *
 * Please send any bug reports or fixes you make to the
 * email address(es):
 *    lksctp developers <[e-mail]>
 *
 * Or submit a bug report through the following website:
 *    http://www.sf.net/projects/lksctp
 *
 * Written or modified by:
 *    Jon Grimm             <[e-mail]>
 *    La Monte H.P. Yarroll <[e-mail]>
 *    Sridhar Samudrala     <[e-mail]>
 *
 * Any bugs reported given to us we will try to fix... any fixes shared will
 * be incorporated into the next SCTP release.
 */

#include <linux/types.h>
#include <linux/skbuff.h>
#include <net/sock.h>
#include <net/sctp/structs.h>
#include <net/sctp/sctp.h>
#include <net/sctp/sm.h>

/* Forward declarations for internal helpers.  */
static struct sctp_ulpevent * sctp_ulpq_reasm(struct sctp_ulpq *ulpq,
					      struct sctp_ulpevent *);
static struct sctp_ulpevent * sctp_ulpq_order(struct sctp_ulpq *,
					      struct sctp_ulpevent *);
static void sctp_ulpq_reasm_drain(struct sctp_ulpq *ulpq);

/* 1st Level Abstractions */

/* Initialize a ULP queue from a block of memory.  */
struct sctp_ulpq *sctp_ulpq_init(struct sctp_ulpq *ulpq,
				 struct sctp_association *asoc)
{
	memset(ulpq, 0, sizeof(struct sctp_ulpq));

	ulpq->asoc = asoc;
	skb_queue_head_init(&ulpq->reasm);
	skb_queue_head_init(&ulpq->lobby);
	ulpq->pd_mode  = 0;
	ulpq->malloced = 0;

	return ulpq;
}


/* Flush the reassembly and ordering queues.  */
void sctp_ulpq_flush(struct sctp_ulpq *ulpq)
{
	struct sk_buff *skb;
	struct sctp_ulpevent *event;

	while ((skb = __skb_dequeue(&ulpq->lobby)) != NULL) {
		event = sctp_skb2event(skb);
		sctp_ulpevent_free(event);
	}

	while ((skb = __skb_dequeue(&ulpq->reasm)) != NULL) {
		event = sctp_skb2event(skb);
		sctp_ulpevent_free(event);
	}

}

/* Dispose of a ulpqueue.  */
void sctp_ulpq_free(struct sctp_ulpq *ulpq)
{
	sctp_ulpq_flush(ulpq);
	if (ulpq->malloced)
		kfree(ulpq);
}

/* Process an incoming DATA chunk.  */
int sctp_ulpq_tail_data(struct sctp_ulpq *ulpq, struct sctp_chunk *chunk,
			gfp_t gfp)
{
	struct sk_buff_head temp;
	sctp_data_chunk_t *hdr;
	struct sctp_ulpevent *event;

	hdr = (sctp_data_chunk_t *) chunk->chunk_hdr;

	/* Create an event from the incoming chunk. */
	event = sctp_ulpevent_make_rcvmsg(chunk->asoc, chunk, gfp);
	if (!event)
		return -ENOMEM;

	/* Do reassembly if needed.  */
	event = sctp_ulpq_reasm(ulpq, event);

	/* Do ordering if needed.  */
	if ((event) && (event->msg_flags & MSG_EOR)){
		/* Create a temporary list to collect chunks on.  */
		skb_queue_head_init(&temp);
		__skb_queue_tail(&temp, sctp_event2skb(event));

		event = sctp_ulpq_order(ulpq, event);
	}

	/* Send event to the ULP.  'event' is the sctp_ulpevent for
	 * very first SKB on the 'temp' list.
	 */
	if (event)
		sctp_ulpq_tail_event(ulpq, event);

	return 0;
}

/* Add a new event for propagation to the ULP.  */
/* Clear the partial delivery mode for this socket.   Note: This
 * assumes that no association is currently in partial delivery mode.
 */
int sctp_clear_pd(struct sock *sk, struct sctp_association *asoc)
{
	struct sctp_sock *sp = sctp_sk(sk);

	if (atomic_dec_and_test(&sp->pd_mode)) {
		/* This means there are no other associations in PD, so
		 * we can go ahead and clear out the lobby in one shot
		 */
		if (!skb_queue_empty(&sp->pd_lobby)) {
			struct list_head *list;
			sctp_skb_list_tail(&sp->pd_lobby, &sk->sk_receive_queue);
			list = (struct list_head *)&sctp_sk(sk)->pd_lobby;
			INIT_LIST_HEAD(list);
			return 1;
		}
	} else {
		/* There are other associations in PD, so we only need to
		 * pull stuff out of the lobby that belongs to the
		 * associations that is exiting PD (all of its notifications
		 * are posted here).
		 */
		if (!skb_queue_empty(&sp->pd_lobby) && asoc) {
			struct sk_buff *skb, *tmp;
			struct sctp_ulpevent *event;

			sctp_skb_for_each(skb, &sp->pd_lobby, tmp) {
				event = sctp_skb2event(skb);
				if (event->asoc == asoc) {
					__skb_unlink(skb, &sp->pd_lobby);
					__skb_queue_tail(&sk->sk_receive_queue,
							 skb);
				}
			}
		}
	}

	return 0;
}

/* Set the pd_mode on the socket and ulpq */
static void sctp_ulpq_set_pd(struct sctp_ulpq *ulpq)
{
	struct sctp_sock *sp = sctp_sk(ulpq->asoc->base.sk);

	atomic_inc(&sp->pd_mode);
	ulpq->pd_mode = 1;
}

/* Clear the pd_mode and restart any pending messages waiting for delivery. */
static int sctp_ulpq_clear_pd(struct sctp_ulpq *ulpq)
{
	ulpq->pd_mode = 0;
	sctp_ulpq_reasm_drain(ulpq);
	return sctp_clear_pd(ulpq->asoc->base.sk, ulpq->asoc);
}

/* If the SKB of 'event' is on a list, it is the first such member
 * of that list.
 */
int sctp_ulpq_tail_event(struct sctp_ulpq *ulpq, struct sctp_ulpevent *event)
{
	struct sock *sk = ulpq->asoc->base.sk;
	struct sk_buff_head *queue, *skb_list;
	struct sk_buff *skb = sctp_event2skb(event);
	int clear_pd = 0;

	skb_list = (struct sk_buff_head *) skb->prev;

	/* If the socket is just going to throw this away, do not
	 * even try to deliver it.
	 */
	if (sock_flag(sk, SOCK_DEAD) || (sk->sk_shutdown & RCV_SHUTDOWN))
		goto out_free;

	/* Check if the user wishes to receive this event.  */
	if (!sctp_ulpevent_is_enabled(event, &sctp_sk(sk)->subscribe))
		goto out_free;

	/* If we are in partial delivery mode, post to the lobby until
	 * partial delivery is cleared, unless, of course _this_ is
	 * the association the cause of the partial delivery.
	 */

	if (atomic_read(&sctp_sk(sk)->pd_mode) == 0) {
		queue = &sk->sk_receive_queue;
	} else {
		if (ulpq->pd_mode) {
			/* If the association is in partial delivery, we
			 * need to finish delivering the partially processed
			 * packet before passing any other data.  This is
			 * because we don't truly support stream interleaving.
			 */
			if ((event->msg_flags & MSG_NOTIFICATION) ||
			    (SCTP_DATA_NOT_FRAG ==
				    (event->msg_flags & SCTP_DATA_FRAG_MASK)))
				queue = &sctp_sk(sk)->pd_lobby;
			else {
				clear_pd = event->msg_flags & MSG_EOR;
				queue = &sk->sk_receive_queue;
			}
		} else {
			/*
			 * If fragment interleave is enabled, we
			 * can queue this to the recieve queue instead
			 * of the lobby.
			 */
			if (sctp_sk(sk)->frag_interleave)
				queue = &sk->sk_receive_queue;
			else
				queue = &sctp_sk(sk)->pd_lobby;
		}
	}

	/* If we are harvesting multiple skbs they will be
	 * collected on a list.
	 */
	if (skb_list)
		sctp_skb_list_tail(skb_list, queue);
	else
		__skb_queue_tail(queue, skb);

	/* Did we just complete partial delivery and need to get
	 * rolling again?  Move pending data to the receive
	 * queue.
	 */
	if (clear_pd)
		sctp_ulpq_clear_pd(ulpq);

	if (queue == &sk->sk_receive_queue)
		sk->sk_data_ready(sk, 0);
	return 1;

out_free:
	if (skb_list)
		sctp_queue_purge_ulpevents(skb_list);
	else
		sctp_ulpevent_free(event);

	return 0;
}

/* 2nd Level Abstractions */

/* Helper function to store chunks that need to be reassembled.  */
static void sctp_ulpq_store_reasm(struct sctp_ulpq *ulpq,
					 struct sctp_ulpevent *event)
{
	struct sk_buff *pos;
	struct sctp_ulpevent *cevent;
	__u32 tsn, ctsn;

	tsn = event->tsn;

	/* See if it belongs at the end. */
	pos = skb_peek_tail(&ulpq->reasm);
	if (!pos) {
		__skb_queue_tail(&ulpq->reasm, sctp_event2skb(event));
		return;
	}

	/* Short circuit just dropping it at the end. */
	cevent = sctp_skb2event(pos);
	ctsn = cevent->tsn;
	if (TSN_lt(ctsn, tsn)) {
		__skb_queue_tail(&ulpq->reasm, sctp_event2skb(event));
		return;
	}

	/* Find the right place in this list. We store them by TSN.  */
	skb_queue_walk(&ulpq->reasm, pos) {
		cevent = sctp_skb2event(pos);
		ctsn = cevent->tsn;

		if (TSN_lt(tsn, ctsn))
			break;
	}

	/* Insert before pos. */
	__skb_insert(sctp_event2skb(event), pos->prev, pos, &ulpq->reasm);

}

/* Helper function to return an event corresponding to the reassembled
 * datagram.
 * This routine creates a re-assembled skb given the first and last skb's
 * as stored in the reassembly queue. The skb's may be non-linear if the sctp
 * payload was fragmented on the way and ip had to reassemble them.
 * We add the rest of skb's to the first skb's fraglist.
 */
static struct sctp_ulpevent *sctp_make_reassembled_event(struct sk_buff_head *queue, struct sk_buff *f_frag, struct sk_buff *l_frag)
{
	struct sk_buff *pos;
	struct sk_buff *new = NULL;
	struct sctp_ulpevent *event;
	struct sk_buff *pnext, *last;
	struct sk_buff *list = skb_shinfo(f_frag)->frag_list;

	/* Store the pointer to the 2nd skb */
	if (f_frag == l_frag)
		pos = NULL;
	else
		pos = f_frag->next;

	/* Get the last skb in the f_frag's frag_list if present. */
	for (last = list; list; last = list, list = list->next);

	/* Add the list of remaining fragments to the first fragments
	 * frag_list.
	 */
	if (last)
		last->next = pos;
	else {
		if (skb_cloned(f_frag)) {
			/* This is a cloned skb, we can't just modify
			 * the frag_list.  We need a new skb to do that.
			 * Instead of calling skb_unshare(), we'll do it
			 * ourselves since we need to delay the free.
			 */
			new = skb_copy(f_frag, GFP_ATOMIC);
			if (!new)
				return NULL;	/* try again later */

			sctp_skb_set_owner_r(new, f_frag->sk);

			skb_shinfo(new)->frag_list = pos;
		} else
			skb_shinfo(f_frag)->frag_list = pos;
	}

	/* Remove the first fragment from the reassembly queue.  */
	__skb_unlink(f_frag, queue);

	/* if we did unshare, then free the old skb and re-assign */
	if (new) {
		kfree_skb(f_frag);
		f_frag = new;
	}

	while (pos) {

		pnext = pos->next;

		/* Update the len and data_len fields of the first fragment. */
		f_frag->len += pos->len;
		f_frag->data_len += pos->len;

		/* Remove the fragment from the reassembly queue.  */
		__skb_unlink(pos, queue);

		/* Break if we have reached the last fragment.  */
		if (pos == l_frag)
			break;
		pos->next = pnext;
		pos = pnext;
	}

	event = sctp_skb2event(f_frag);
	SCTP_INC_STATS(SCTP_MIB_REASMUSRMSGS);

	return event;
}


/* Helper function to check if an incoming chunk has filled up the last
 * missing fragment in a SCTP datagram and return the corresponding event.
 */
static struct sctp_ulpevent *sctp_ulpq_retrieve_reassembled(struct sctp_ulpq *ulpq)
{
	struct sk_buff *pos;
	struct sctp_ulpevent *cevent;
	struct sk_buff *first_frag = NULL;
	__u32 ctsn, next_tsn;
	struct sctp_ulpevent *retval = NULL;
	struct sk_buff *pd_first = NULL;
	struct sk_buff *pd_last = NULL;
	size_t pd_len = 0;
	struct sctp_association *asoc;
	u32 pd_point;

	/* Initialized to 0 just to avoid compiler warning message.  Will
	 * never be used with this value. It is referenced only after it
	 * is set when we find the first fragment of a message.
	 */
	next_tsn = 0;

	/* The chunks are held in the reasm queue sorted by TSN.
	 * Walk through the queue sequentially and look for a sequence of
	 * fragmented chunks that complete a datagram.
	 * 'first_frag' and next_tsn are reset when we find a chunk which
	 * is the first fragment of a datagram. Once these 2 fields are set
	 * we expect to find the remaining middle fragments and the last
	 * fragment in order. If not, first_frag is reset to NULL and we
	 * start the next pass when we find another first fragment.
	 *
	 * There is a potential to do partial delivery if user sets
	 * SCTP_PARTIAL_DELIVERY_POINT option. Lets count some things here
	 * to see if can do PD.
	 */
	skb_queue_walk(&ulpq->reasm, pos) {
		cevent = sctp_skb2event(pos);
		ctsn = cevent->tsn;

		switch (cevent->msg_flags & SCTP_DATA_FRAG_MASK) {
		case SCTP_DATA_FIRST_FRAG:
			/* If this "FIRST_FRAG" is the first
			 * element in the queue, then count it towards
			 * possible PD.
			 */
			if (pos == ulpq->reasm.next) {
			    pd_first = pos;
			    pd_last = pos;
			    pd_len = pos->len;
			} else {
			    pd_first = NULL;
			    pd_last = NULL;
			    pd_len = 0;
			}

			first_frag = pos;
			next_tsn = ctsn + 1;
			break;

		case SCTP_DATA_MIDDLE_FRAG:
			if ((first_frag) && (ctsn == next_tsn)) {
				next_tsn++;
				if (pd_first) {
				    pd_last = pos;
				    pd_len += pos->len;
				}
			} else
				first_frag = NULL;
			break;

		case SCTP_DATA_LAST_FRAG:
			if (first_frag && (ctsn == next_tsn))
				goto found;
			else
				first_frag = NULL;
			break;
		}
	}

	asoc = ulpq->asoc;
	if (pd_first) {
		/* Make sure we can enter partial deliver.
		 * We can trigger partial delivery only if framgent
		 * interleave is set, or the socket is not already
		 * in  partial delivery.
		 */
		if (!sctp_sk(asoc->base.sk)->frag_interleave &&
		    atomic_read(&sctp_sk(asoc->base.sk)->pd_mode))
			goto done;

		cevent = sctp_skb2event(pd_first);
		pd_point = sctp_sk(asoc->base.sk)->pd_point;
		if (pd_point && pd_point <= pd_len) {
			retval = sctp_make_reassembled_event(&ulpq->reasm,
							     pd_first,
							     pd_last);
			if (retval)
				sctp_ulpq_set_pd(ulpq);
		}
	}
done:
	return retval;
found:
	retval = sctp_make_reassembled_event(&ulpq->reasm, first_frag, pos);
	if (retval)
		retval->msg_flags |= MSG_EOR;
	goto done;
}

/* Retrieve the next set of fragments of a partial message. */
static struct sctp_ulpevent *sctp_ulpq_retrieve_partial(struct sctp_ulpq *ulpq)
{
	struct sk_buff *pos, *last_frag, *first_frag;
	struct sctp_ulpevent *cevent;
	__u32 ctsn, next_tsn;
	int is_last;
	struct sctp_ulpevent *retval;

	/* The chunks are held in the reasm queue sorted by TSN.
	 * Walk through the queue sequentially and look for the first
	 * sequence of fragmented chunks.
	 */

	if (skb_queue_empty(&ulpq->reasm))
		return NULL;

	last_frag = first_frag = NULL;
	retval = NULL;
	next_tsn = 0;
	is_last = 0;

	skb_queue_walk(&ulpq->reasm, pos) {
		cevent = sctp_skb2event(pos);
		ctsn = cevent->tsn;

		switch (cevent->msg_flags & SCTP_DATA_FRAG_MASK) {
		case SCTP_DATA_MIDDLE_FRAG:
			if (!first_frag) {
				first_frag = pos;
				next_tsn = ctsn + 1;
				last_frag = pos;
			} else if (next_tsn == ctsn)
				next_tsn++;
			else