[deliverable/linux.git] / net / dccp / output.c

/*
 *  net/dccp/output.c
 * 
 *  An implementation of the DCCP protocol
 *  Arnaldo Carvalho de Melo <acme@conectiva.com.br>
 *
 *	This program 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 of the License, or (at your option) any later version.
 */

#include <linux/dccp.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>

#include <net/inet_sock.h>
#include <net/sock.h>

#include "ackvec.h"
#include "ccid.h"
#include "dccp.h"

static inline void dccp_event_ack_sent(struct sock *sk)
{
	inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
}

static void dccp_skb_entail(struct sock *sk, struct sk_buff *skb)
{
	skb_set_owner_w(skb, sk);
	WARN_ON(sk->sk_send_head);
	sk->sk_send_head = skb;
}

/*
 * All SKB's seen here are completely headerless. It is our
 * job to build the DCCP header, and pass the packet down to
 * IP so it can do the same plus pass the packet off to the
 * device.
 */
static int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb)
{
	if (likely(skb != NULL)) {
		const struct inet_sock *inet = inet_sk(sk);
		const struct inet_connection_sock *icsk = inet_csk(sk);
		struct dccp_sock *dp = dccp_sk(sk);
		struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
		struct dccp_hdr *dh;
		/* XXX For now we're using only 48 bits sequence numbers */
		const u32 dccp_header_size = sizeof(*dh) +
					     sizeof(struct dccp_hdr_ext) +
					  dccp_packet_hdr_len(dcb->dccpd_type);
		int err, set_ack = 1;
		u64 ackno = dp->dccps_gsr;

		dccp_inc_seqno(&dp->dccps_gss);

		switch (dcb->dccpd_type) {
		case DCCP_PKT_DATA:
			set_ack = 0;
			/* fall through */
		case DCCP_PKT_DATAACK:
			break;

		case DCCP_PKT_REQUEST:
			set_ack = 0;
			/* fall through */

		case DCCP_PKT_SYNC:
		case DCCP_PKT_SYNCACK:
			ackno = dcb->dccpd_seq;
			/* fall through */
		default:
			/*
			 * Only data packets should come through with skb->sk
			 * set.
			 */
			WARN_ON(skb->sk);
			skb_set_owner_w(skb, sk);
			break;
		}

		dcb->dccpd_seq = dp->dccps_gss;

		if (dccp_insert_options(sk, skb)) {
			kfree_skb(skb);
			return -EPROTO;
		}
		

		/* Build DCCP header and checksum it. */
		dh = dccp_zeroed_hdr(skb, dccp_header_size);
		dh->dccph_type	= dcb->dccpd_type;
		dh->dccph_sport	= inet->sport;
		dh->dccph_dport	= inet->dport;
		dh->dccph_doff	= (dccp_header_size + dcb->dccpd_opt_len) / 4;
		dh->dccph_ccval	= dcb->dccpd_ccval;
		dh->dccph_cscov = dp->dccps_pcslen;
		/* XXX For now we're using only 48 bits sequence numbers */
		dh->dccph_x	= 1;

		dp->dccps_awh = dp->dccps_gss;
		dccp_hdr_set_seq(dh, dp->dccps_gss);
		if (set_ack)
			dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), ackno);

		switch (dcb->dccpd_type) {
		case DCCP_PKT_REQUEST:
			dccp_hdr_request(skb)->dccph_req_service =
							dp->dccps_service;
			break;
		case DCCP_PKT_RESET:
			dccp_hdr_reset(skb)->dccph_reset_code =
							dcb->dccpd_reset_code;
			break;
		}

		icsk->icsk_af_ops->send_check(sk, 0, skb);

		if (set_ack)
			dccp_event_ack_sent(sk);

		DCCP_INC_STATS(DCCP_MIB_OUTSEGS);

		memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
		err = icsk->icsk_af_ops->queue_xmit(skb, sk, 0);
		return net_xmit_eval(err);
	}
	return -ENOBUFS;
}

unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu)
{
	struct inet_connection_sock *icsk = inet_csk(sk);
	struct dccp_sock *dp = dccp_sk(sk);
	int mss_now = (pmtu - icsk->icsk_af_ops->net_header_len -
		       sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext));

	/* Now subtract optional transport overhead */
	mss_now -= icsk->icsk_ext_hdr_len;

	/*
	 * FIXME: this should come from the CCID infrastructure, where, say,
	 * TFRC will say it wants TIMESTAMPS, ELAPSED time, etc, for now lets
	 * put a rough estimate for NDP + TIMESTAMP + TIMESTAMP_ECHO + ELAPSED
	 * TIME + TFRC_OPT_LOSS_EVENT_RATE + TFRC_OPT_RECEIVE_RATE + padding to
	 * make it a multiple of 4
	 */

	mss_now -= ((5 + 6 + 10 + 6 + 6 + 6 + 3) / 4) * 4;

	/* And store cached results */
	icsk->icsk_pmtu_cookie = pmtu;
	dp->dccps_mss_cache = mss_now;

	return mss_now;
}

EXPORT_SYMBOL_GPL(dccp_sync_mss);

void dccp_write_space(struct sock *sk)
{
	read_lock(&sk->sk_callback_lock);

	if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
		wake_up_interruptible(sk->sk_sleep);
	/* Should agree with poll, otherwise some programs break */
	if (sock_writeable(sk))
		sk_wake_async(sk, 2, POLL_OUT);

	read_unlock(&sk->sk_callback_lock);
}

/**
 * dccp_wait_for_ccid - Wait for ccid to tell us we can send a packet
 * @sk: socket to wait for
 */
static int dccp_wait_for_ccid(struct sock *sk, struct sk_buff *skb)
{
	struct dccp_sock *dp = dccp_sk(sk);
	DEFINE_WAIT(wait);
	unsigned long delay;
	int rc;

	while (1) {
		prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);

		if (sk->sk_err)
			goto do_error;
		if (signal_pending(current))
			goto do_interrupted;

		rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb);
		if (rc <= 0)
			break;
		delay = msecs_to_jiffies(rc);
		sk->sk_write_pending++;
		release_sock(sk);
		schedule_timeout(delay);
		lock_sock(sk);
		sk->sk_write_pending--;
	}
out:
	finish_wait(sk->sk_sleep, &wait);
	return rc;

do_error:
	rc = -EPIPE;
	goto out;
do_interrupted:
	rc = -EINTR;
	goto out;
}

static void dccp_write_xmit_timer(unsigned long data) {
	struct sock *sk = (struct sock *)data;
	struct dccp_sock *dp = dccp_sk(sk);

	bh_lock_sock(sk);
	if (sock_owned_by_user(sk))
		sk_reset_timer(sk, &dp->dccps_xmit_timer, jiffies+1);
	else
		dccp_write_xmit(sk, 0);
	bh_unlock_sock(sk);
	sock_put(sk);
}

void dccp_write_xmit(struct sock *sk, int block)
{
	struct dccp_sock *dp = dccp_sk(sk);
	struct sk_buff *skb;

	while ((skb = skb_peek(&sk->sk_write_queue))) {
		int err = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb);

		if (err > 0) {
			if (!block) {
				sk_reset_timer(sk, &dp->dccps_xmit_timer,
						msecs_to_jiffies(err)+jiffies);
				break;
			} else
				err = dccp_wait_for_ccid(sk, skb);
			if (err)
				DCCP_BUG("err=%d after dccp_wait_for_ccid", err);
		}

		skb_dequeue(&sk->sk_write_queue);
		if (err == 0) {
			struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
			const int len = skb->len;

			if (sk->sk_state == DCCP_PARTOPEN) {
				/* See 8.1.5.  Handshake Completion */
				inet_csk_schedule_ack(sk);
				inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  inet_csk(sk)->icsk_rto,
						  DCCP_RTO_MAX);
				dcb->dccpd_type = DCCP_PKT_DATAACK;
			} else if (dccp_ack_pending(sk))
				dcb->dccpd_type = DCCP_PKT_DATAACK;
			else
				dcb->dccpd_type = DCCP_PKT_DATA;

			err = dccp_transmit_skb(sk, skb);
			ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, 0, len);
			if (err)
				DCCP_BUG("err=%d after ccid_hc_tx_packet_sent",
					 err);
		} else
			kfree(skb);
	}
}

int dccp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
{
	if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk) != 0)
		return -EHOSTUNREACH; /* Routing failure or similar. */

	return dccp_transmit_skb(sk, (skb_cloned(skb) ?
				      pskb_copy(skb, GFP_ATOMIC):
				      skb_clone(skb, GFP_ATOMIC)));
}

struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst,
				   struct request_sock *req)
{
	struct dccp_hdr *dh;
	struct dccp_request_sock *dreq;
	const u32 dccp_header_size = sizeof(struct dccp_hdr) +
				     sizeof(struct dccp_hdr_ext) +
				     sizeof(struct dccp_hdr_response);
	struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1,
					   GFP_ATOMIC);
	if (skb == NULL)
		return NULL;

	/* Reserve space for headers. */
	skb_reserve(skb, sk->sk_prot->max_header);

	skb->dst = dst_clone(dst);

	dreq = dccp_rsk(req);
	if (inet_rsk(req)->acked)	/* increase ISS upon retransmission */
		dccp_inc_seqno(&dreq->dreq_iss);
	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESPONSE;
	DCCP_SKB_CB(skb)->dccpd_seq  = dreq->dreq_iss;

	if (dccp_insert_options(sk, skb)) {
		kfree_skb(skb);
		return NULL;
	}

	/* Build and checksum header */
	dh = dccp_zeroed_hdr(skb, dccp_header_size);

	dh->dccph_sport	= inet_sk(sk)->sport;
	dh->dccph_dport	= inet_rsk(req)->rmt_port;
	dh->dccph_doff	= (dccp_header_size +
			   DCCP_SKB_CB(skb)->dccpd_opt_len) / 4;
	dh->dccph_type	= DCCP_PKT_RESPONSE;
	dh->dccph_x	= 1;
	dccp_hdr_set_seq(dh, dreq->dreq_iss);
	dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dreq->dreq_isr);
	dccp_hdr_response(skb)->dccph_resp_service = dreq->dreq_service;

	dccp_csum_outgoing(skb);

	/* We use `acked' to remember that a Response was already sent. */
	inet_rsk(req)->acked = 1;
	DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
	return skb;
}

EXPORT_SYMBOL_GPL(dccp_make_response);

static struct sk_buff *dccp_make_reset(struct sock *sk, struct dst_entry *dst,
				       const enum dccp_reset_codes code)
				   
{
	struct dccp_hdr *dh;
	struct dccp_sock *dp = dccp_sk(sk);
	const u32 dccp_header_size = sizeof(struct dccp_hdr) +
				     sizeof(struct dccp_hdr_ext) +
				     sizeof(struct dccp_hdr_reset);
	struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1,
					   GFP_ATOMIC);
	if (skb == NULL)
		return NULL;

	/* Reserve space for headers. */
	skb_reserve(skb, sk->sk_prot->max_header);

	skb->dst = dst_clone(dst);

	dccp_inc_seqno(&dp->dccps_gss);

	DCCP_SKB_CB(skb)->dccpd_reset_code = code;
	DCCP_SKB_CB(skb)->dccpd_type	   = DCCP_PKT_RESET;
	DCCP_SKB_CB(skb)->dccpd_seq	   = dp->dccps_gss;

	if (dccp_insert_options(sk, skb)) {
		kfree_skb(skb);
		return NULL;
	}

	dh = dccp_zeroed_hdr(skb, dccp_header_size);

	dh->dccph_sport	= inet_sk(sk)->sport;
	dh->dccph_dport	= inet_sk(sk)->dport;
	dh->dccph_doff	= (dccp_header_size +
			   DCCP_SKB_CB(skb)->dccpd_opt_len) / 4;
	dh->dccph_type	= DCCP_PKT_RESET;
	dh->dccph_x	= 1;
	dccp_hdr_set_seq(dh, dp->dccps_gss);
	dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dp->dccps_gsr);

	dccp_hdr_reset(skb)->dccph_reset_code = code;
	inet_csk(sk)->icsk_af_ops->send_check(sk, 0, skb);

	DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
	return skb;
}

int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code)
{
	/*
	 * FIXME: what if rebuild_header fails?
	 * Should we be doing a rebuild_header here?
	 */
	int err = inet_sk_rebuild_header(sk);

	if (err == 0) {
		struct sk_buff *skb = dccp_make_reset(sk, sk->sk_dst_cache,
						      code);
		if (skb != NULL) {
			memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
			err = inet_csk(sk)->icsk_af_ops->queue_xmit(skb, sk, 0);
			return net_xmit_eval(err);
		}
	}

	return err;
}

/*
 * Do all connect socket setups that can be done AF independent.
 */
static inline void dccp_connect_init(struct sock *sk)
{
	struct dccp_sock *dp = dccp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);

	sk->sk_err = 0;
	sock_reset_flag(sk, SOCK_DONE);
	
	dccp_sync_mss(sk, dst_mtu(dst));

 	/*
	 * SWL and AWL are initially adjusted so that they are not less than
	 * the initial Sequence Numbers received and sent, respectively:
	 *	SWL := max(GSR + 1 - floor(W/4), ISR),
	 *	AWL := max(GSS - W' + 1, ISS).
	 * These adjustments MUST be applied only at the beginning of the
	 * connection.
 	 */
	dccp_update_gss(sk, dp->dccps_iss);
	dccp_set_seqno(&dp->dccps_awl, max48(dp->dccps_awl, dp->dccps_iss));

	/* S.GAR - greatest valid acknowledgement number received on a non-Sync;
	 *         initialized to S.ISS (sec. 8.5)                            */
	dp->dccps_gar = dp->dccps_iss;

	icsk->icsk_retransmits = 0;
	init_timer(&dp->dccps_xmit_timer);
	dp->dccps_xmit_timer.data = (unsigned long)sk;
	dp->dccps_xmit_timer.function = dccp_write_xmit_timer;
}

int dccp_connect(struct sock *sk)
{
	struct sk_buff *skb;
	struct inet_connection_sock *icsk = inet_csk(sk);

	dccp_connect_init(sk);

	skb = alloc_skb(sk->sk_prot->max_header, sk->sk_allocation);
	if (unlikely(skb == NULL))
		return -ENOBUFS;

	/* Reserve space for headers. */
	skb_reserve(skb, sk->sk_prot->max_header);

	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_REQUEST;

	dccp_skb_entail(sk, skb);
	dccp_transmit_skb(sk, skb_clone(skb, GFP_KERNEL));
	DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS);

	/* Timer for repeating the REQUEST until an answer. */
	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
				  icsk->icsk_rto, DCCP_RTO_MAX);
	return 0;
}

EXPORT_SYMBOL_GPL(dccp_connect);

void dccp_send_ack(struct sock *sk)
{
	/* If we have been reset, we may not send again. */
	if (sk->sk_state != DCCP_CLOSED) {
		struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header,
						GFP_ATOMIC);

		if (skb == NULL) {
			inet_csk_schedule_ack(sk);
			inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX,
						  DCCP_RTO_MAX);
			return;
		}

		/* Reserve space for headers */
		skb_reserve(skb, sk->sk_prot->max_header);
		DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_ACK;
		dccp_transmit_skb(sk, skb);
	}
}

EXPORT_SYMBOL_GPL(dccp_send_ack);

void dccp_send_delayed_ack(struct sock *sk)
{
	struct inet_connection_sock *icsk = inet_csk(sk);
	/*
	 * FIXME: tune this timer. elapsed time fixes the skew, so no problem
	 * with using 2s, and active senders also piggyback the ACK into a
	 * DATAACK packet, so this is really for quiescent senders.
	 */
	unsigned long timeout = jiffies + 2 * HZ;

	/* Use new timeout only if there wasn't a older one earlier. */
	if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
		/* If delack timer was blocked or is about to expire,
		 * send ACK now.
		 *
		 * FIXME: check the "about to expire" part
		 */
		if (icsk->icsk_ack.blocked) {
			dccp_send_ack(sk);
			return;
		}

		if (!time_before(timeout, icsk->icsk_ack.timeout))
			timeout = icsk->icsk_ack.timeout;
	}
	icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER;
	icsk->icsk_ack.timeout = timeout;
	sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout);
}

void dccp_send_sync(struct sock *sk, const u64 seq,
		    const enum dccp_pkt_type pkt_type)
{
	/*
	 * We are not putting this on the write queue, so
	 * dccp_transmit_skb() will set the ownership to this
	 * sock.
	 */
	struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header, GFP_ATOMIC);

	if (skb == NULL)
		/* FIXME: how to make sure the sync is sent? */
		return;

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(skb, sk->sk_prot->max_header);
	DCCP_SKB_CB(skb)->dccpd_type = pkt_type;
	DCCP_SKB_CB(skb)->dccpd_seq = seq;

	dccp_transmit_skb(sk, skb);
}

EXPORT_SYMBOL_GPL(dccp_send_sync);

/*
 * Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This
 * cannot be allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under
 * any circumstances.
 */
void dccp_send_close(struct sock *sk, const int active)
{
	struct dccp_sock *dp = dccp_sk(sk);
	struct sk_buff *skb;
	const gfp_t prio = active ? GFP_KERNEL : GFP_ATOMIC;

	skb = alloc_skb(sk->sk_prot->max_header, prio);
	if (skb == NULL)
		return;

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(skb, sk->sk_prot->max_header);
	DCCP_SKB_CB(skb)->dccpd_type = dp->dccps_role == DCCP_ROLE_CLIENT ?
					DCCP_PKT_CLOSE : DCCP_PKT_CLOSEREQ;

	if (active) {
		dccp_write_xmit(sk, 1);
		dccp_skb_entail(sk, skb);
		dccp_transmit_skb(sk, skb_clone(skb, prio));
		/* FIXME do we need a retransmit timer here? */
	} else
		dccp_transmit_skb(sk, skb);
}
Commit	Line	Data
7c657876 ACM	1	/*
	2	* net/dccp/output.c
	3	*
	4	* An implementation of the DCCP protocol
	5	* Arnaldo Carvalho de Melo <acme@conectiva.com.br>
	6	*
	7	* This program is free software; you can redistribute it and/or
	8	* modify it under the terms of the GNU General Public License
	9	* as published by the Free Software Foundation; either version
	10	* 2 of the License, or (at your option) any later version.
	11	*/
	12
7c657876	13	#include <linux/dccp.h>
48918a4d	14	#include <linux/kernel.h>
7c657876 ACM	15	#include <linux/skbuff.h>
7c657876 ACM	16
14c85021	17	#include <net/inet_sock.h>
7c657876 ACM	18	#include <net/sock.h>
7c657876 ACM	19
ae31c339	20	#include "ackvec.h"
7c657876 ACM	21	#include "ccid.h"
	22	#include "dccp.h"
	23
	24	static inline void dccp_event_ack_sent(struct sock *sk)
	25	{
	26	inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
	27	}
	28
c25a18ba	29	static void dccp_skb_entail(struct sock sk, struct sk_buff skb)
48918a4d HX	30	{
	31	skb_set_owner_w(skb, sk);
	32	WARN_ON(sk->sk_send_head);
	33	sk->sk_send_head = skb;
	34	}
	35
7c657876 ACM	36	/*
	37	* All SKB's seen here are completely headerless. It is our
	38	* job to build the DCCP header, and pass the packet down to
	39	* IP so it can do the same plus pass the packet off to the
	40	* device.
	41	*/
48918a4d	42	static int dccp_transmit_skb(struct sock sk, struct sk_buff skb)
7c657876 ACM	43	{
	44	if (likely(skb != NULL)) {
	45	const struct inet_sock *inet = inet_sk(sk);
57cca05a	46	const struct inet_connection_sock *icsk = inet_csk(sk);
7c657876 ACM	47	struct dccp_sock *dp = dccp_sk(sk);
	48	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
	49	struct dccp_hdr *dh;
	50	/* XXX For now we're using only 48 bits sequence numbers */
118b2c95	51	const u32 dccp_header_size = sizeof(*dh) +
7c657876	52	sizeof(struct dccp_hdr_ext) +
7690af3f	53	dccp_packet_hdr_len(dcb->dccpd_type);
7c657876 ACM	54	int err, set_ack = 1;
	55	u64 ackno = dp->dccps_gsr;
	56
7c657876 ACM	57	dccp_inc_seqno(&dp->dccps_gss);
7c657876 ACM	58
7c657876 ACM	59	switch (dcb->dccpd_type) {
	60	case DCCP_PKT_DATA:
	61	set_ack = 0;
edc9e819 HX	62	/* fall through */
edc9e819 HX	63	case DCCP_PKT_DATAACK:
7c657876	64	break;
edc9e819	65
afe00251 AB	66	case DCCP_PKT_REQUEST:
	67	set_ack = 0;
	68	/* fall through */
	69
7c657876 ACM	70	case DCCP_PKT_SYNC:
	71	case DCCP_PKT_SYNCACK:
	72	ackno = dcb->dccpd_seq;
edc9e819 HX	73	/* fall through */
	74	default:
	75	/*
	76	* Only data packets should come through with skb->sk
	77	* set.
	78	*/
	79	WARN_ON(skb->sk);
	80	skb_set_owner_w(skb, sk);
7c657876 ACM	81	break;
7c657876 ACM	82	}
24117727 ACM	83
24117727 ACM	84	dcb->dccpd_seq = dp->dccps_gss;
2d0817d1 ACM	85
	86	if (dccp_insert_options(sk, skb)) {
	87	kfree_skb(skb);
	88	return -EPROTO;
	89	}
7c657876	90
fda0fd6c	91
7c657876	92	/* Build DCCP header and checksum it. */
9b42078e	93	dh = dccp_zeroed_hdr(skb, dccp_header_size);
7c657876 ACM	94	dh->dccph_type = dcb->dccpd_type;
	95	dh->dccph_sport = inet->sport;
	96	dh->dccph_dport = inet->dport;
	97	dh->dccph_doff = (dccp_header_size + dcb->dccpd_opt_len) / 4;
	98	dh->dccph_ccval = dcb->dccpd_ccval;
6f4e5fff	99	dh->dccph_cscov = dp->dccps_pcslen;
7c657876 ACM	100	/* XXX For now we're using only 48 bits sequence numbers */
	101	dh->dccph_x = 1;
	102
	103	dp->dccps_awh = dp->dccps_gss;
	104	dccp_hdr_set_seq(dh, dp->dccps_gss);
	105	if (set_ack)
	106	dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), ackno);
	107
	108	switch (dcb->dccpd_type) {
	109	case DCCP_PKT_REQUEST:
7690af3f	110	dccp_hdr_request(skb)->dccph_req_service =
67e6b629	111	dp->dccps_service;
7c657876 ACM	112	break;
7c657876 ACM	113	case DCCP_PKT_RESET:
7690af3f ACM	114	dccp_hdr_reset(skb)->dccph_reset_code =
7690af3f ACM	115	dcb->dccpd_reset_code;
7c657876 ACM	116	break;
	117	}
	118
6f4e5fff	119	icsk->icsk_af_ops->send_check(sk, 0, skb);
7c657876	120
7ad07e7c	121	if (set_ack)
7c657876 ACM	122	dccp_event_ack_sent(sk);
	123
	124	DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
	125
49c5bfaf	126	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
93173112	127	err = icsk->icsk_af_ops->queue_xmit(skb, sk, 0);
b9df3cb8	128	return net_xmit_eval(err);
7c657876 ACM	129	}
	130	return -ENOBUFS;
	131	}
	132
	133	unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu)
	134	{
d83d8461	135	struct inet_connection_sock *icsk = inet_csk(sk);
7c657876	136	struct dccp_sock *dp = dccp_sk(sk);
d83d8461	137	int mss_now = (pmtu - icsk->icsk_af_ops->net_header_len -
57cca05a	138	sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext));
7c657876 ACM	139
7c657876 ACM	140	/* Now subtract optional transport overhead */
d83d8461	141	mss_now -= icsk->icsk_ext_hdr_len;
7c657876 ACM	142
	143	/*
	144	* FIXME: this should come from the CCID infrastructure, where, say,
	145	* TFRC will say it wants TIMESTAMPS, ELAPSED time, etc, for now lets
	146	* put a rough estimate for NDP + TIMESTAMP + TIMESTAMP_ECHO + ELAPSED
	147	* TIME + TFRC_OPT_LOSS_EVENT_RATE + TFRC_OPT_RECEIVE_RATE + padding to
	148	* make it a multiple of 4
	149	*/
	150
	151	mss_now -= ((5 + 6 + 10 + 6 + 6 + 6 + 3) / 4) * 4;
	152
	153	/* And store cached results */
d83d8461	154	icsk->icsk_pmtu_cookie = pmtu;
7c657876 ACM	155	dp->dccps_mss_cache = mss_now;
	156
	157	return mss_now;
	158	}
	159
f21e68ca ACM	160	EXPORT_SYMBOL_GPL(dccp_sync_mss);
f21e68ca ACM	161
c530cfb1 ACM	162	void dccp_write_space(struct sock *sk)
	163	{
	164	read_lock(&sk->sk_callback_lock);
	165
	166	if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
	167	wake_up_interruptible(sk->sk_sleep);
	168	/* Should agree with poll, otherwise some programs break */
	169	if (sock_writeable(sk))
	170	sk_wake_async(sk, 2, POLL_OUT);
	171
	172	read_unlock(&sk->sk_callback_lock);
	173	}
	174
d6809c12 ACM	175	/**
	176	* dccp_wait_for_ccid - Wait for ccid to tell us we can send a packet
	177	* @sk: socket to wait for
d6809c12	178	*/
5cc3741d	179	static int dccp_wait_for_ccid(struct sock sk, struct sk_buff skb)
d6809c12 ACM	180	{
	181	struct dccp_sock *dp = dccp_sk(sk);
	182	DEFINE_WAIT(wait);
5cc3741d	183	unsigned long delay;
d6809c12 ACM	184	int rc;
	185
	186	while (1) {
	187	prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
	188
97e5848d	189	if (sk->sk_err)
d6809c12	190	goto do_error;
d6809c12 ACM	191	if (signal_pending(current))
	192	goto do_interrupted;
	193
6b57c93d	194	rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb);
d6809c12 ACM	195	if (rc <= 0)
	196	break;
	197	delay = msecs_to_jiffies(rc);
d6809c12 ACM	198	sk->sk_write_pending++;
d6809c12 ACM	199	release_sock(sk);
5cc3741d	200	schedule_timeout(delay);
d6809c12 ACM	201	lock_sock(sk);
	202	sk->sk_write_pending--;
	203	}
	204	out:
	205	finish_wait(sk->sk_sleep, &wait);
	206	return rc;
	207
	208	do_error:
	209	rc = -EPIPE;
	210	goto out;
d6809c12	211	do_interrupted:
5cc3741d	212	rc = -EINTR;
d6809c12 ACM	213	goto out;
	214	}
	215
97e5848d IM	216	static void dccp_write_xmit_timer(unsigned long data) {
	217	struct sock sk = (struct sock )data;
	218	struct dccp_sock *dp = dccp_sk(sk);
	219
	220	bh_lock_sock(sk);
	221	if (sock_owned_by_user(sk))
	222	sk_reset_timer(sk, &dp->dccps_xmit_timer, jiffies+1);
	223	else
	224	dccp_write_xmit(sk, 0);
	225	bh_unlock_sock(sk);
	226	sock_put(sk);
	227	}
	228
	229	void dccp_write_xmit(struct sock *sk, int block)
27258ee5	230	{
97e5848d IM	231	struct dccp_sock *dp = dccp_sk(sk);
97e5848d IM	232	struct sk_buff *skb;
97e5848d IM	233
97e5848d IM	234	while ((skb = skb_peek(&sk->sk_write_queue))) {
6b57c93d	235	int err = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb);
d6809c12	236
97e5848d IM	237	if (err > 0) {
	238	if (!block) {
	239	sk_reset_timer(sk, &dp->dccps_xmit_timer,
	240	msecs_to_jiffies(err)+jiffies);
	241	break;
5cc3741d IM	242	} else
5cc3741d IM	243	err = dccp_wait_for_ccid(sk, skb);
59348b19 GR	244	if (err)
59348b19 GR	245	DCCP_BUG("err=%d after dccp_wait_for_ccid", err);
97e5848d	246	}
27258ee5	247
97e5848d IM	248	skb_dequeue(&sk->sk_write_queue);
	249	if (err == 0) {
	250	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
	251	const int len = skb->len;
27258ee5	252
97e5848d IM	253	if (sk->sk_state == DCCP_PARTOPEN) {
	254	/* See 8.1.5. Handshake Completion */
	255	inet_csk_schedule_ack(sk);
	256	inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
27258ee5 ACM	257	inet_csk(sk)->icsk_rto,
27258ee5 ACM	258	DCCP_RTO_MAX);
97e5848d IM	259	dcb->dccpd_type = DCCP_PKT_DATAACK;
	260	} else if (dccp_ack_pending(sk))
	261	dcb->dccpd_type = DCCP_PKT_DATAACK;
	262	else
	263	dcb->dccpd_type = DCCP_PKT_DATA;
	264
	265	err = dccp_transmit_skb(sk, skb);
	266	ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, 0, len);
59348b19 GR	267	if (err)
	268	DCCP_BUG("err=%d after ccid_hc_tx_packet_sent",
	269	err);
97e5848d IM	270	} else
	271	kfree(skb);
	272	}
27258ee5 ACM	273	}
27258ee5 ACM	274
7c657876 ACM	275	int dccp_retransmit_skb(struct sock sk, struct sk_buff skb)
7c657876 ACM	276	{
57cca05a	277	if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk) != 0)
7c657876 ACM	278	return -EHOSTUNREACH; /* Routing failure or similar. */
	279
	280	return dccp_transmit_skb(sk, (skb_cloned(skb) ?
	281	pskb_copy(skb, GFP_ATOMIC):
	282	skb_clone(skb, GFP_ATOMIC)));
	283	}
	284
	285	struct sk_buff dccp_make_response(struct sock sk, struct dst_entry *dst,
	286	struct request_sock *req)
	287	{
	288	struct dccp_hdr *dh;
67e6b629	289	struct dccp_request_sock *dreq;
118b2c95	290	const u32 dccp_header_size = sizeof(struct dccp_hdr) +
7c657876 ACM	291	sizeof(struct dccp_hdr_ext) +
7c657876 ACM	292	sizeof(struct dccp_hdr_response);
118b2c95	293	struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1,
7c657876 ACM	294	GFP_ATOMIC);
	295	if (skb == NULL)
	296	return NULL;
	297
	298	/* Reserve space for headers. */
118b2c95	299	skb_reserve(skb, sk->sk_prot->max_header);
7c657876 ACM	300
7c657876 ACM	301	skb->dst = dst_clone(dst);
7c657876	302
67e6b629	303	dreq = dccp_rsk(req);
e11d9d30 GR	304	if (inet_rsk(req)->acked) /* increase ISS upon retransmission */
e11d9d30 GR	305	dccp_inc_seqno(&dreq->dreq_iss);
7c657876	306	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESPONSE;
67e6b629	307	DCCP_SKB_CB(skb)->dccpd_seq = dreq->dreq_iss;
2d0817d1 ACM	308
	309	if (dccp_insert_options(sk, skb)) {
	310	kfree_skb(skb);
	311	return NULL;
	312	}
7c657876	313
09dbc389	314	/* Build and checksum header */
9b42078e	315	dh = dccp_zeroed_hdr(skb, dccp_header_size);
7c657876 ACM	316
	317	dh->dccph_sport = inet_sk(sk)->sport;
	318	dh->dccph_dport = inet_rsk(req)->rmt_port;
7690af3f ACM	319	dh->dccph_doff = (dccp_header_size +
7690af3f ACM	320	DCCP_SKB_CB(skb)->dccpd_opt_len) / 4;
7c657876 ACM	321	dh->dccph_type = DCCP_PKT_RESPONSE;
7c657876 ACM	322	dh->dccph_x = 1;
67e6b629 ACM	323	dccp_hdr_set_seq(dh, dreq->dreq_iss);
	324	dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dreq->dreq_isr);
	325	dccp_hdr_response(skb)->dccph_resp_service = dreq->dreq_service;
7c657876	326
6f4e5fff GR	327	dccp_csum_outgoing(skb);
6f4e5fff GR	328
e11d9d30 GR	329	/* We use `acked' to remember that a Response was already sent. */
e11d9d30 GR	330	inet_rsk(req)->acked = 1;
7c657876 ACM	331	DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
	332	return skb;
	333	}
	334
f21e68ca ACM	335	EXPORT_SYMBOL_GPL(dccp_make_response);
f21e68ca ACM	336
017487d7 ACM	337	static struct sk_buff dccp_make_reset(struct sock sk, struct dst_entry *dst,
017487d7 ACM	338	const enum dccp_reset_codes code)
7c657876 ACM	339
	340	{
	341	struct dccp_hdr *dh;
	342	struct dccp_sock *dp = dccp_sk(sk);
118b2c95	343	const u32 dccp_header_size = sizeof(struct dccp_hdr) +
7c657876 ACM	344	sizeof(struct dccp_hdr_ext) +
7c657876 ACM	345	sizeof(struct dccp_hdr_reset);
118b2c95	346	struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1,
7c657876 ACM	347	GFP_ATOMIC);
	348	if (skb == NULL)
	349	return NULL;
	350
	351	/* Reserve space for headers. */
118b2c95	352	skb_reserve(skb, sk->sk_prot->max_header);
7c657876 ACM	353
7c657876 ACM	354	skb->dst = dst_clone(dst);
7c657876 ACM	355
	356	dccp_inc_seqno(&dp->dccps_gss);
	357
	358	DCCP_SKB_CB(skb)->dccpd_reset_code = code;
	359	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESET;
	360	DCCP_SKB_CB(skb)->dccpd_seq = dp->dccps_gss;
2d0817d1 ACM	361
	362	if (dccp_insert_options(sk, skb)) {
	363	kfree_skb(skb);
	364	return NULL;
	365	}
7c657876	366
9b42078e	367	dh = dccp_zeroed_hdr(skb, dccp_header_size);
7c657876 ACM	368
	369	dh->dccph_sport = inet_sk(sk)->sport;
	370	dh->dccph_dport = inet_sk(sk)->dport;
7690af3f ACM	371	dh->dccph_doff = (dccp_header_size +
7690af3f ACM	372	DCCP_SKB_CB(skb)->dccpd_opt_len) / 4;
7c657876 ACM	373	dh->dccph_type = DCCP_PKT_RESET;
	374	dh->dccph_x = 1;
	375	dccp_hdr_set_seq(dh, dp->dccps_gss);
	376	dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dp->dccps_gsr);
	377
	378	dccp_hdr_reset(skb)->dccph_reset_code = code;
6f4e5fff	379	inet_csk(sk)->icsk_af_ops->send_check(sk, 0, skb);
7c657876 ACM	380
	381	DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
	382	return skb;
	383	}
	384
017487d7 ACM	385	int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code)
	386	{
	387	/*
	388	* FIXME: what if rebuild_header fails?
	389	* Should we be doing a rebuild_header here?
	390	*/
	391	int err = inet_sk_rebuild_header(sk);
	392
	393	if (err == 0) {
	394	struct sk_buff *skb = dccp_make_reset(sk, sk->sk_dst_cache,
	395	code);
	396	if (skb != NULL) {
	397	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
93173112	398	err = inet_csk(sk)->icsk_af_ops->queue_xmit(skb, sk, 0);
b9df3cb8	399	return net_xmit_eval(err);
017487d7 ACM	400	}
	401	}
	402
	403	return err;
	404	}
	405
7c657876 ACM	406	/*
	407	* Do all connect socket setups that can be done AF independent.
	408	*/
	409	static inline void dccp_connect_init(struct sock *sk)
	410	{
f21e68ca	411	struct dccp_sock *dp = dccp_sk(sk);
7c657876 ACM	412	struct dst_entry *dst = __sk_dst_get(sk);
	413	struct inet_connection_sock *icsk = inet_csk(sk);
	414
	415	sk->sk_err = 0;
	416	sock_reset_flag(sk, SOCK_DONE);
	417
	418	dccp_sync_mss(sk, dst_mtu(dst));
	419
f21e68ca ACM	420	/*
	421	* SWL and AWL are initially adjusted so that they are not less than
	422	* the initial Sequence Numbers received and sent, respectively:
	423	* SWL := max(GSR + 1 - floor(W/4), ISR),
	424	* AWL := max(GSS - W' + 1, ISS).
	425	* These adjustments MUST be applied only at the beginning of the
	426	* connection.
	427	*/
d7f7365f	428	dccp_update_gss(sk, dp->dccps_iss);
f21e68ca	429	dccp_set_seqno(&dp->dccps_awl, max48(dp->dccps_awl, dp->dccps_iss));
7c657876	430
d7f7365f GR	431	/* S.GAR - greatest valid acknowledgement number received on a non-Sync;
	432	* initialized to S.ISS (sec. 8.5) */
	433	dp->dccps_gar = dp->dccps_iss;
	434
7c657876	435	icsk->icsk_retransmits = 0;
97e5848d IM	436	init_timer(&dp->dccps_xmit_timer);
	437	dp->dccps_xmit_timer.data = (unsigned long)sk;
	438	dp->dccps_xmit_timer.function = dccp_write_xmit_timer;
7c657876 ACM	439	}
	440
	441	int dccp_connect(struct sock *sk)
	442	{
	443	struct sk_buff *skb;
	444	struct inet_connection_sock *icsk = inet_csk(sk);
	445
	446	dccp_connect_init(sk);
	447
118b2c95	448	skb = alloc_skb(sk->sk_prot->max_header, sk->sk_allocation);
7c657876 ACM	449	if (unlikely(skb == NULL))
	450	return -ENOBUFS;
	451
	452	/* Reserve space for headers. */
118b2c95	453	skb_reserve(skb, sk->sk_prot->max_header);
7c657876 ACM	454
7c657876 ACM	455	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_REQUEST;
7c657876	456
48918a4d	457	dccp_skb_entail(sk, skb);
7c657876 ACM	458	dccp_transmit_skb(sk, skb_clone(skb, GFP_KERNEL));
	459	DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS);
	460
	461	/* Timer for repeating the REQUEST until an answer. */
27258ee5 ACM	462	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
27258ee5 ACM	463	icsk->icsk_rto, DCCP_RTO_MAX);
7c657876 ACM	464	return 0;
	465	}
	466
f21e68ca ACM	467	EXPORT_SYMBOL_GPL(dccp_connect);
f21e68ca ACM	468
7c657876 ACM	469	void dccp_send_ack(struct sock *sk)
	470	{
	471	/* If we have been reset, we may not send again. */
	472	if (sk->sk_state != DCCP_CLOSED) {
118b2c95 ACM	473	struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header,
118b2c95 ACM	474	GFP_ATOMIC);
7c657876 ACM	475
	476	if (skb == NULL) {
	477	inet_csk_schedule_ack(sk);
	478	inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
7690af3f ACM	479	inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
	480	TCP_DELACK_MAX,
	481	DCCP_RTO_MAX);
7c657876 ACM	482	return;
	483	}
	484
	485	/* Reserve space for headers */
118b2c95	486	skb_reserve(skb, sk->sk_prot->max_header);
7c657876	487	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_ACK;
7c657876 ACM	488	dccp_transmit_skb(sk, skb);
	489	}
	490	}
	491
	492	EXPORT_SYMBOL_GPL(dccp_send_ack);
	493
	494	void dccp_send_delayed_ack(struct sock *sk)
	495	{
	496	struct inet_connection_sock *icsk = inet_csk(sk);
	497	/*
	498	* FIXME: tune this timer. elapsed time fixes the skew, so no problem
	499	* with using 2s, and active senders also piggyback the ACK into a
	500	* DATAACK packet, so this is really for quiescent senders.
	501	*/
	502	unsigned long timeout = jiffies + 2 * HZ;
	503
	504	/* Use new timeout only if there wasn't a older one earlier. */
	505	if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
	506	/* If delack timer was blocked or is about to expire,
	507	* send ACK now.
	508	*
	509	* FIXME: check the "about to expire" part
	510	*/
	511	if (icsk->icsk_ack.blocked) {
	512	dccp_send_ack(sk);
	513	return;
	514	}
	515
	516	if (!time_before(timeout, icsk->icsk_ack.timeout))
	517	timeout = icsk->icsk_ack.timeout;
	518	}
	519	icsk->icsk_ack.pending \|= ICSK_ACK_SCHED \| ICSK_ACK_TIMER;
	520	icsk->icsk_ack.timeout = timeout;
	521	sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout);
	522	}
	523
e92ae93a ACM	524	void dccp_send_sync(struct sock *sk, const u64 seq,
e92ae93a ACM	525	const enum dccp_pkt_type pkt_type)
7c657876 ACM	526	{
	527	/*
	528	* We are not putting this on the write queue, so
	529	* dccp_transmit_skb() will set the ownership to this
	530	* sock.
	531	*/
118b2c95	532	struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header, GFP_ATOMIC);
7c657876 ACM	533
	534	if (skb == NULL)
	535	/* FIXME: how to make sure the sync is sent? */
	536	return;
	537
	538	/* Reserve space for headers and prepare control bits. */
118b2c95	539	skb_reserve(skb, sk->sk_prot->max_header);
e92ae93a	540	DCCP_SKB_CB(skb)->dccpd_type = pkt_type;
7c657876 ACM	541	DCCP_SKB_CB(skb)->dccpd_seq = seq;
7c657876 ACM	542
7c657876 ACM	543	dccp_transmit_skb(sk, skb);
	544	}
	545
b61fafc4 ACM	546	EXPORT_SYMBOL_GPL(dccp_send_sync);
b61fafc4 ACM	547
7690af3f ACM	548	/*
	549	* Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This
	550	* cannot be allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under
	551	* any circumstances.
7c657876	552	*/
7ad07e7c	553	void dccp_send_close(struct sock *sk, const int active)
7c657876 ACM	554	{
	555	struct dccp_sock *dp = dccp_sk(sk);
	556	struct sk_buff *skb;
7d877f3b	557	const gfp_t prio = active ? GFP_KERNEL : GFP_ATOMIC;
7c657876	558
7ad07e7c ACM	559	skb = alloc_skb(sk->sk_prot->max_header, prio);
	560	if (skb == NULL)
	561	return;
7c657876 ACM	562
	563	/* Reserve space for headers and prepare control bits. */
	564	skb_reserve(skb, sk->sk_prot->max_header);
7690af3f ACM	565	DCCP_SKB_CB(skb)->dccpd_type = dp->dccps_role == DCCP_ROLE_CLIENT ?
7690af3f ACM	566	DCCP_PKT_CLOSE : DCCP_PKT_CLOSEREQ;
7c657876	567
7ad07e7c	568	if (active) {
97e5848d	569	dccp_write_xmit(sk, 1);
48918a4d	570	dccp_skb_entail(sk, skb);
7ad07e7c	571	dccp_transmit_skb(sk, skb_clone(skb, prio));
97e5848d	572	/* FIXME do we need a retransmit timer here? */
7ad07e7c ACM	573	} else
7ad07e7c ACM	574	dccp_transmit_skb(sk, skb);
7c657876	575	}