[net-next-2.6.git] / net / dccp / ackvec.c

/*
 *  net/dccp/ackvec.c
 *
 *  An implementation of the DCCP protocol
 *  Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
 *
 *      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; version 2 of the License;
 */

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

#include <linux/dccp.h>
#include <linux/init.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/slab.h>

#include <net/sock.h>

static kmem_cache_t *dccp_ackvec_slab;
static kmem_cache_t *dccp_ackvec_record_slab;

static struct dccp_ackvec_record *dccp_ackvec_record_new(void)
{
	struct dccp_ackvec_record *avr =
			kmem_cache_alloc(dccp_ackvec_record_slab, GFP_ATOMIC);

	if (avr != NULL)
		INIT_LIST_HEAD(&avr->dccpavr_node);

	return avr;
}

static void dccp_ackvec_record_delete(struct dccp_ackvec_record *avr)
{
	if (unlikely(avr == NULL))
		return;
	/* Check if deleting a linked record */
	WARN_ON(!list_empty(&avr->dccpavr_node));
	kmem_cache_free(dccp_ackvec_record_slab, avr);
}

static void dccp_ackvec_insert_avr(struct dccp_ackvec *av,
				   struct dccp_ackvec_record *avr)
{
	/*
	 * AVRs are sorted by seqno. Since we are sending them in order, we
	 * just add the AVR at the head of the list.
	 * -sorbo.
	 */
	if (!list_empty(&av->dccpav_records)) {
		const struct dccp_ackvec_record *head =
					list_entry(av->dccpav_records.next,
						   struct dccp_ackvec_record,
						   dccpavr_node);
		BUG_ON(before48(avr->dccpavr_ack_seqno,
				head->dccpavr_ack_seqno));
	}

	list_add(&avr->dccpavr_node, &av->dccpav_records);
}

int dccp_insert_option_ackvec(struct sock *sk, struct sk_buff *skb)
{
	struct dccp_sock *dp = dccp_sk(sk);
#ifdef CONFIG_IP_DCCP_DEBUG
	const char *debug_prefix = dp->dccps_role == DCCP_ROLE_CLIENT ?
				"CLIENT tx: " : "server tx: ";
#endif
	struct dccp_ackvec *av = dp->dccps_hc_rx_ackvec;
	int len = av->dccpav_vec_len + 2;
	struct timeval now;
	u32 elapsed_time;
	unsigned char *to, *from;
	struct dccp_ackvec_record *avr;

	if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
		return -1;

	dccp_timestamp(sk, &now);
	elapsed_time = timeval_delta(&now, &av->dccpav_time) / 10;

	if (elapsed_time != 0 &&
	    dccp_insert_option_elapsed_time(sk, skb, elapsed_time))
		return -1;

	avr = dccp_ackvec_record_new();
	if (avr == NULL)
		return -1;

	DCCP_SKB_CB(skb)->dccpd_opt_len += len;

	to    = skb_push(skb, len);
	*to++ = DCCPO_ACK_VECTOR_0;
	*to++ = len;

	len  = av->dccpav_vec_len;
	from = av->dccpav_buf + av->dccpav_buf_head;

	/* Check if buf_head wraps */
	if ((int)av->dccpav_buf_head + len > DCCP_MAX_ACKVEC_LEN) {
		const u32 tailsize = DCCP_MAX_ACKVEC_LEN - av->dccpav_buf_head;

		memcpy(to, from, tailsize);
		to   += tailsize;
		len  -= tailsize;
		from = av->dccpav_buf;
	}

	memcpy(to, from, len);
	/*
	 *	From RFC 4340, A.2:
	 *
	 *	For each acknowledgement it sends, the HC-Receiver will add an
	 *	acknowledgement record.  ack_seqno will equal the HC-Receiver
	 *	sequence number it used for the ack packet; ack_ptr will equal
	 *	buf_head; ack_ackno will equal buf_ackno; and ack_nonce will
	 *	equal buf_nonce.
	 */
	avr->dccpavr_ack_seqno = DCCP_SKB_CB(skb)->dccpd_seq;
	avr->dccpavr_ack_ptr   = av->dccpav_buf_head;
	avr->dccpavr_ack_ackno = av->dccpav_buf_ackno;
	avr->dccpavr_ack_nonce = av->dccpav_buf_nonce;
	avr->dccpavr_sent_len  = av->dccpav_vec_len;

	dccp_ackvec_insert_avr(av, avr);

	dccp_pr_debug("%sACK Vector 0, len=%d, ack_seqno=%llu, "
		      "ack_ackno=%llu\n",
		      debug_prefix, avr->dccpavr_sent_len,
		      (unsigned long long)avr->dccpavr_ack_seqno,
		      (unsigned long long)avr->dccpavr_ack_ackno);
	return 0;
}

struct dccp_ackvec *dccp_ackvec_alloc(const gfp_t priority)
{
	struct dccp_ackvec *av = kmem_cache_alloc(dccp_ackvec_slab, priority);

	if (av != NULL) {
		av->dccpav_buf_head	= DCCP_MAX_ACKVEC_LEN - 1;
		av->dccpav_buf_ackno	= DCCP_MAX_SEQNO + 1;
		av->dccpav_buf_nonce = av->dccpav_buf_nonce = 0;
		av->dccpav_ack_ptr	= 0;
		av->dccpav_time.tv_sec	= 0;
		av->dccpav_time.tv_usec	= 0;
		av->dccpav_vec_len	= 0;
		INIT_LIST_HEAD(&av->dccpav_records);
	}

	return av;
}

void dccp_ackvec_free(struct dccp_ackvec *av)
{
	if (unlikely(av == NULL))
		return;

	if (!list_empty(&av->dccpav_records)) {
		struct dccp_ackvec_record *avr, *next;

		list_for_each_entry_safe(avr, next, &av->dccpav_records,
					 dccpavr_node) {
			list_del_init(&avr->dccpavr_node);
			dccp_ackvec_record_delete(avr);
		}
	}

	kmem_cache_free(dccp_ackvec_slab, av);
}

static inline u8 dccp_ackvec_state(const struct dccp_ackvec *av,
				   const u8 index)
{
	return av->dccpav_buf[index] & DCCP_ACKVEC_STATE_MASK;
}

static inline u8 dccp_ackvec_len(const struct dccp_ackvec *av,
				 const u8 index)
{
	return av->dccpav_buf[index] & DCCP_ACKVEC_LEN_MASK;
}

/*
 * If several packets are missing, the HC-Receiver may prefer to enter multiple
 * bytes with run length 0, rather than a single byte with a larger run length;
 * this simplifies table updates if one of the missing packets arrives.
 */
static inline int dccp_ackvec_set_buf_head_state(struct dccp_ackvec *av,
						 const unsigned int packets,
						 const unsigned char state)
{
	unsigned int gap;
	long new_head;

	if (av->dccpav_vec_len + packets > DCCP_MAX_ACKVEC_LEN)
		return -ENOBUFS;

	gap	 = packets - 1;
	new_head = av->dccpav_buf_head - packets;

	if (new_head < 0) {
		if (gap > 0) {
			memset(av->dccpav_buf, DCCP_ACKVEC_STATE_NOT_RECEIVED,
			       gap + new_head + 1);
			gap = -new_head;
		}
		new_head += DCCP_MAX_ACKVEC_LEN;
	} 

	av->dccpav_buf_head = new_head;

	if (gap > 0)
		memset(av->dccpav_buf + av->dccpav_buf_head + 1,
		       DCCP_ACKVEC_STATE_NOT_RECEIVED, gap);

	av->dccpav_buf[av->dccpav_buf_head] = state;
	av->dccpav_vec_len += packets;
	return 0;
}

/*
 * Implements the RFC 4340, Appendix A
 */
int dccp_ackvec_add(struct dccp_ackvec *av, const struct sock *sk,
		    const u64 ackno, const u8 state)
{
	/*
	 * Check at the right places if the buffer is full, if it is, tell the
	 * caller to start dropping packets till the HC-Sender acks our ACK
	 * vectors, when we will free up space in dccpav_buf.
	 *
	 * We may well decide to do buffer compression, etc, but for now lets
	 * just drop.
	 *
	 * From Appendix A.1.1 (`New Packets'):
	 *
	 *	Of course, the circular buffer may overflow, either when the
	 *	HC-Sender is sending data at a very high rate, when the
	 *	HC-Receiver's acknowledgements are not reaching the HC-Sender,
	 *	or when the HC-Sender is forgetting to acknowledge those acks
	 *	(so the HC-Receiver is unable to clean up old state). In this
	 *	case, the HC-Receiver should either compress the buffer (by
	 *	increasing run lengths when possible), transfer its state to
	 *	a larger buffer, or, as a last resort, drop all received
	 *	packets, without processing them whatsoever, until its buffer
	 *	shrinks again.
	 */

	/* See if this is the first ackno being inserted */
	if (av->dccpav_vec_len == 0) {
		av->dccpav_buf[av->dccpav_buf_head] = state;
		av->dccpav_vec_len = 1;
	} else if (after48(ackno, av->dccpav_buf_ackno)) {
		const u64 delta = dccp_delta_seqno(av->dccpav_buf_ackno,
						   ackno);

		/*
		 * Look if the state of this packet is the same as the
		 * previous ackno and if so if we can bump the head len.
		 */
		if (delta == 1 &&
		    dccp_ackvec_state(av, av->dccpav_buf_head) == state &&
		    (dccp_ackvec_len(av, av->dccpav_buf_head) <
		     DCCP_ACKVEC_LEN_MASK))
			av->dccpav_buf[av->dccpav_buf_head]++;
		else if (dccp_ackvec_set_buf_head_state(av, delta, state))
			return -ENOBUFS;
	} else {
		/*
		 * A.1.2.  Old Packets
		 *
		 *	When a packet with Sequence Number S <= buf_ackno
		 *	arrives, the HC-Receiver will scan the table for
		 *	the byte corresponding to S. (Indexing structures
		 *	could reduce the complexity of this scan.)
		 */
		u64 delta = dccp_delta_seqno(ackno, av->dccpav_buf_ackno);
		u8 index = av->dccpav_buf_head;

		while (1) {
			const u8 len = dccp_ackvec_len(av, index);
			const u8 state = dccp_ackvec_state(av, index);
			/*
			 * valid packets not yet in dccpav_buf have a reserved
			 * entry, with a len equal to 0.
			 */
			if (state == DCCP_ACKVEC_STATE_NOT_RECEIVED &&
			    len == 0 && delta == 0) { /* Found our
							 reserved seat! */
				dccp_pr_debug("Found %llu reserved seat!\n",
					      (unsigned long long)ackno);
				av->dccpav_buf[index] = state;
				goto out;
			}
			/* len == 0 means one packet */
			if (delta < len + 1)
				goto out_duplicate;

			delta -= len + 1;
			if (++index == DCCP_MAX_ACKVEC_LEN)
				index = 0;
		}
	}

	av->dccpav_buf_ackno = ackno;
	dccp_timestamp(sk, &av->dccpav_time);
out:
	return 0;

out_duplicate:
	/* Duplicate packet */
	dccp_pr_debug("Received a dup or already considered lost "
		      "packet: %llu\n", (unsigned long long)ackno);
	return -EILSEQ;
}

#ifdef CONFIG_IP_DCCP_DEBUG
void dccp_ackvector_print(const u64 ackno, const unsigned char *vector, int len)
{
	if (!dccp_debug)
		return;

	printk("ACK vector len=%d, ackno=%llu |", len,
	       (unsigned long long)ackno);

	while (len--) {
		const u8 state = (*vector & DCCP_ACKVEC_STATE_MASK) >> 6;
		const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;

		printk("%d,%d|", state, rl);
		++vector;
	}

	printk("\n");
}

void dccp_ackvec_print(const struct dccp_ackvec *av)
{
	dccp_ackvector_print(av->dccpav_buf_ackno,
			     av->dccpav_buf + av->dccpav_buf_head,
			     av->dccpav_vec_len);
}
#endif

static void dccp_ackvec_throw_record(struct dccp_ackvec *av,
				     struct dccp_ackvec_record *avr)
{
	struct dccp_ackvec_record *next;

	/* sort out vector length */
	if (av->dccpav_buf_head <= avr->dccpavr_ack_ptr)
		av->dccpav_vec_len = avr->dccpavr_ack_ptr - av->dccpav_buf_head;
	else
		av->dccpav_vec_len = DCCP_MAX_ACKVEC_LEN - 1
				     - av->dccpav_buf_head
				     + avr->dccpavr_ack_ptr;

	/* free records */
	list_for_each_entry_safe_from(avr, next, &av->dccpav_records,
				      dccpavr_node) {
		list_del_init(&avr->dccpavr_node);
		dccp_ackvec_record_delete(avr);
	}
}

void dccp_ackvec_check_rcv_ackno(struct dccp_ackvec *av, struct sock *sk,
				 const u64 ackno)
{
	struct dccp_ackvec_record *avr;

	/*
	 * If we traverse backwards, it should be faster when we have large
	 * windows. We will be receiving ACKs for stuff we sent a while back
	 * -sorbo.
	 */
	list_for_each_entry_reverse(avr, &av->dccpav_records, dccpavr_node) {
		if (ackno == avr->dccpavr_ack_seqno) {
#ifdef CONFIG_IP_DCCP_DEBUG
			struct dccp_sock *dp = dccp_sk(sk);
			const char *debug_prefix = dp->dccps_role == DCCP_ROLE_CLIENT ?
						"CLIENT rx ack: " : "server rx ack: ";
#endif
			dccp_pr_debug("%sACK packet 0, len=%d, ack_seqno=%llu, "
				      "ack_ackno=%llu, ACKED!\n",
				      debug_prefix, 1,
				      (unsigned long long)avr->dccpavr_ack_seqno,
				      (unsigned long long)avr->dccpavr_ack_ackno);
			dccp_ackvec_throw_record(av, avr);
			break;
		}
	}
}

static void dccp_ackvec_check_rcv_ackvector(struct dccp_ackvec *av,
					    struct sock *sk, u64 ackno,
					    const unsigned char len,
					    const unsigned char *vector)
{
	unsigned char i;
	struct dccp_ackvec_record *avr;

	/* Check if we actually sent an ACK vector */
	if (list_empty(&av->dccpav_records))
		return;

	i = len;
	/*
	 * XXX
	 * I think it might be more efficient to work backwards. See comment on
	 * rcv_ackno. -sorbo.
	 */
	avr = list_entry(av->dccpav_records.next, struct dccp_ackvec_record,
			 dccpavr_node);
	while (i--) {
		const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
		u64 ackno_end_rl;

		dccp_set_seqno(&ackno_end_rl, ackno - rl);

		/*
		 * If our AVR sequence number is greater than the ack, go
		 * forward in the AVR list until it is not so.
		 */
		list_for_each_entry_from(avr, &av->dccpav_records,
					 dccpavr_node) {
			if (!after48(avr->dccpavr_ack_seqno, ackno))
				goto found;
		}
		/* End of the dccpav_records list, not found, exit */
		break;
found:
		if (between48(avr->dccpavr_ack_seqno, ackno_end_rl, ackno)) {
			const u8 state = *vector & DCCP_ACKVEC_STATE_MASK;
			if (state != DCCP_ACKVEC_STATE_NOT_RECEIVED) {
#ifdef CONFIG_IP_DCCP_DEBUG
				struct dccp_sock *dp = dccp_sk(sk);
				const char *debug_prefix =
					dp->dccps_role == DCCP_ROLE_CLIENT ?
					"CLIENT rx ack: " : "server rx ack: ";
#endif
				dccp_pr_debug("%sACK vector 0, len=%d, "
					      "ack_seqno=%llu, ack_ackno=%llu, "
					      "ACKED!\n",
					      debug_prefix, len,
					      (unsigned long long)
					      avr->dccpavr_ack_seqno,
					      (unsigned long long)
					      avr->dccpavr_ack_ackno);
				dccp_ackvec_throw_record(av, avr);
				break;
			}
			/*
			 * If it wasn't received, continue scanning... we might
			 * find another one.
			 */
		}

		dccp_set_seqno(&ackno, ackno_end_rl - 1);
		++vector;
	}
}

int dccp_ackvec_parse(struct sock *sk, const struct sk_buff *skb,
		      const u8 opt, const u8 *value, const u8 len)
{
	if (len > DCCP_MAX_ACKVEC_LEN)
		return -1;

	/* dccp_ackvector_print(DCCP_SKB_CB(skb)->dccpd_ack_seq, value, len); */
	dccp_ackvec_check_rcv_ackvector(dccp_sk(sk)->dccps_hc_rx_ackvec, sk,
					DCCP_SKB_CB(skb)->dccpd_ack_seq,
				        len, value);
	return 0;
}

static char dccp_ackvec_slab_msg[] __initdata =
	KERN_CRIT "DCCP: Unable to create ack vectors slab caches\n";

int __init dccp_ackvec_init(void)
{
	dccp_ackvec_slab = kmem_cache_create("dccp_ackvec",
					     sizeof(struct dccp_ackvec), 0,
					     SLAB_HWCACHE_ALIGN, NULL, NULL);
	if (dccp_ackvec_slab == NULL)
		goto out_err;

	dccp_ackvec_record_slab =
			kmem_cache_create("dccp_ackvec_record",
					  sizeof(struct dccp_ackvec_record),
					  0, SLAB_HWCACHE_ALIGN, NULL, NULL);
	if (dccp_ackvec_record_slab == NULL)
		goto out_destroy_slab;

	return 0;

out_destroy_slab:
	kmem_cache_destroy(dccp_ackvec_slab);
	dccp_ackvec_slab = NULL;
out_err:
	printk(dccp_ackvec_slab_msg);
	return -ENOBUFS;
}

void dccp_ackvec_exit(void)
{
	if (dccp_ackvec_slab != NULL) {
		kmem_cache_destroy(dccp_ackvec_slab);
		dccp_ackvec_slab = NULL;
	}
	if (dccp_ackvec_record_slab != NULL) {
		kmem_cache_destroy(dccp_ackvec_record_slab);
		dccp_ackvec_record_slab = NULL;
	}
}
Commit	Line	Data
ae31c339 ACM	1	/*
	2	* net/dccp/ackvec.c
	3	*
	4	* An implementation of the DCCP protocol
	5	* Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
	6	*
	7	* This program is free software; you can redistribute it and/or modify it
	8	* under the terms of the GNU General Public License as published by the
	9	* Free Software Foundation; version 2 of the License;
	10	*/
	11
	12	#include "ackvec.h"
	13	#include "dccp.h"
	14
	15	#include <linux/dccp.h>
9b07ef5d ACM	16	#include <linux/init.h>
	17	#include <linux/errno.h>
	18	#include <linux/kernel.h>
ae31c339	19	#include <linux/skbuff.h>
9b07ef5d	20	#include <linux/slab.h>
ae31c339 ACM	21
	22	#include <net/sock.h>
	23
9b07ef5d	24	static kmem_cache_t *dccp_ackvec_slab;
02bcf28c AB	25	static kmem_cache_t *dccp_ackvec_record_slab;
	26
	27	static struct dccp_ackvec_record *dccp_ackvec_record_new(void)
	28	{
	29	struct dccp_ackvec_record *avr =
	30	kmem_cache_alloc(dccp_ackvec_record_slab, GFP_ATOMIC);
	31
	32	if (avr != NULL)
	33	INIT_LIST_HEAD(&avr->dccpavr_node);
	34
	35	return avr;
	36	}
	37
	38	static void dccp_ackvec_record_delete(struct dccp_ackvec_record *avr)
	39	{
	40	if (unlikely(avr == NULL))
	41	return;
	42	/* Check if deleting a linked record */
	43	WARN_ON(!list_empty(&avr->dccpavr_node));
	44	kmem_cache_free(dccp_ackvec_record_slab, avr);
	45	}
	46
	47	static void dccp_ackvec_insert_avr(struct dccp_ackvec *av,
	48	struct dccp_ackvec_record *avr)
	49	{
	50	/*
	51	* AVRs are sorted by seqno. Since we are sending them in order, we
	52	* just add the AVR at the head of the list.
	53	* -sorbo.
	54	*/
	55	if (!list_empty(&av->dccpav_records)) {
	56	const struct dccp_ackvec_record *head =
	57	list_entry(av->dccpav_records.next,
	58	struct dccp_ackvec_record,
	59	dccpavr_node);
	60	BUG_ON(before48(avr->dccpavr_ack_seqno,
	61	head->dccpavr_ack_seqno));
	62	}
	63
	64	list_add(&avr->dccpavr_node, &av->dccpav_records);
	65	}
9b07ef5d	66
ae31c339 ACM	67	int dccp_insert_option_ackvec(struct sock sk, struct sk_buff skb)
	68	{
	69	struct dccp_sock *dp = dccp_sk(sk);
7400d781 ACM	70	#ifdef CONFIG_IP_DCCP_DEBUG
	71	const char *debug_prefix = dp->dccps_role == DCCP_ROLE_CLIENT ?
	72	"CLIENT tx: " : "server tx: ";
	73	#endif
ae31c339 ACM	74	struct dccp_ackvec *av = dp->dccps_hc_rx_ackvec;
	75	int len = av->dccpav_vec_len + 2;
	76	struct timeval now;
	77	u32 elapsed_time;
	78	unsigned char to, from;
02bcf28c AB	79	struct dccp_ackvec_record *avr;
	80
	81	if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
	82	return -1;
	83
ae31c339 ACM	84	dccp_timestamp(sk, &now);
	85	elapsed_time = timeval_delta(&now, &av->dccpav_time) / 10;
	86
2d0817d1 ACM	87	if (elapsed_time != 0 &&
	88	dccp_insert_option_elapsed_time(sk, skb, elapsed_time))
	89	return -1;
	90
	91	avr = dccp_ackvec_record_new();
	92	if (avr == NULL)
	93	return -1;
ae31c339	94
ae31c339 ACM	95	DCCP_SKB_CB(skb)->dccpd_opt_len += len;
	96
	97	to = skb_push(skb, len);
	98	*to++ = DCCPO_ACK_VECTOR_0;
	99	*to++ = len;
	100
	101	len = av->dccpav_vec_len;
	102	from = av->dccpav_buf + av->dccpav_buf_head;
	103
	104	/* Check if buf_head wraps */
02bcf28c AB	105	if ((int)av->dccpav_buf_head + len > DCCP_MAX_ACKVEC_LEN) {
02bcf28c AB	106	const u32 tailsize = DCCP_MAX_ACKVEC_LEN - av->dccpav_buf_head;
ae31c339 ACM	107
	108	memcpy(to, from, tailsize);
	109	to += tailsize;
	110	len -= tailsize;
	111	from = av->dccpav_buf;
	112	}
	113
	114	memcpy(to, from, len);
	115	/*
0e64e94e	116	* From RFC 4340, A.2:
ae31c339 ACM	117	*
	118	* For each acknowledgement it sends, the HC-Receiver will add an
	119	* acknowledgement record. ack_seqno will equal the HC-Receiver
	120	* sequence number it used for the ack packet; ack_ptr will equal
	121	* buf_head; ack_ackno will equal buf_ackno; and ack_nonce will
	122	* equal buf_nonce.
ae31c339	123	*/
02bcf28c AB	124	avr->dccpavr_ack_seqno = DCCP_SKB_CB(skb)->dccpd_seq;
	125	avr->dccpavr_ack_ptr = av->dccpav_buf_head;
	126	avr->dccpavr_ack_ackno = av->dccpav_buf_ackno;
	127	avr->dccpavr_ack_nonce = av->dccpav_buf_nonce;
	128	avr->dccpavr_sent_len = av->dccpav_vec_len;
	129
	130	dccp_ackvec_insert_avr(av, avr);
ae31c339 ACM	131
	132	dccp_pr_debug("%sACK Vector 0, len=%d, ack_seqno=%llu, "
	133	"ack_ackno=%llu\n",
02bcf28c AB	134	debug_prefix, avr->dccpavr_sent_len,
	135	(unsigned long long)avr->dccpavr_ack_seqno,
	136	(unsigned long long)avr->dccpavr_ack_ackno);
	137	return 0;
ae31c339 ACM	138	}
ae31c339 ACM	139
7400d781	140	struct dccp_ackvec *dccp_ackvec_alloc(const gfp_t priority)
ae31c339	141	{
9b07ef5d	142	struct dccp_ackvec *av = kmem_cache_alloc(dccp_ackvec_slab, priority);
e4dfd449	143
ae31c339	144	if (av != NULL) {
4a0a50fb	145	av->dccpav_buf_head = DCCP_MAX_ACKVEC_LEN - 1;
02bcf28c	146	av->dccpav_buf_ackno = DCCP_MAX_SEQNO + 1;
ae31c339 ACM	147	av->dccpav_buf_nonce = av->dccpav_buf_nonce = 0;
	148	av->dccpav_ack_ptr = 0;
	149	av->dccpav_time.tv_sec = 0;
	150	av->dccpav_time.tv_usec = 0;
4a0a50fb	151	av->dccpav_vec_len = 0;
02bcf28c	152	INIT_LIST_HEAD(&av->dccpav_records);
ae31c339 ACM	153	}
	154
	155	return av;
	156	}
	157
	158	void dccp_ackvec_free(struct dccp_ackvec *av)
	159	{
02bcf28c AB	160	if (unlikely(av == NULL))
02bcf28c AB	161	return;
d5e9b2c7 ACM	162
	163	if (!list_empty(&av->dccpav_records)) {
	164	struct dccp_ackvec_record avr, next;
	165
	166	list_for_each_entry_safe(avr, next, &av->dccpav_records,
	167	dccpavr_node) {
	168	list_del_init(&avr->dccpavr_node);
	169	dccp_ackvec_record_delete(avr);
	170	}
	171	}
	172
9b07ef5d	173	kmem_cache_free(dccp_ackvec_slab, av);
ae31c339 ACM	174	}
	175
	176	static inline u8 dccp_ackvec_state(const struct dccp_ackvec *av,
e4dfd449	177	const u8 index)
ae31c339 ACM	178	{
	179	return av->dccpav_buf[index] & DCCP_ACKVEC_STATE_MASK;
	180	}
	181
	182	static inline u8 dccp_ackvec_len(const struct dccp_ackvec *av,
e4dfd449	183	const u8 index)
ae31c339 ACM	184	{
	185	return av->dccpav_buf[index] & DCCP_ACKVEC_LEN_MASK;
	186	}
	187
	188	/*
	189	* If several packets are missing, the HC-Receiver may prefer to enter multiple
	190	* bytes with run length 0, rather than a single byte with a larger run length;
	191	* this simplifies table updates if one of the missing packets arrives.
	192	*/
	193	static inline int dccp_ackvec_set_buf_head_state(struct dccp_ackvec *av,
	194	const unsigned int packets,
e4dfd449	195	const unsigned char state)
ae31c339 ACM	196	{
ae31c339 ACM	197	unsigned int gap;
a8fc3d8d	198	long new_head;
ae31c339	199
7400d781	200	if (av->dccpav_vec_len + packets > DCCP_MAX_ACKVEC_LEN)
ae31c339 ACM	201	return -ENOBUFS;
	202
	203	gap = packets - 1;
	204	new_head = av->dccpav_buf_head - packets;
	205
	206	if (new_head < 0) {
	207	if (gap > 0) {
	208	memset(av->dccpav_buf, DCCP_ACKVEC_STATE_NOT_RECEIVED,
	209	gap + new_head + 1);
	210	gap = -new_head;
	211	}
7400d781	212	new_head += DCCP_MAX_ACKVEC_LEN;
ae31c339 ACM	213	}
	214
	215	av->dccpav_buf_head = new_head;
	216
	217	if (gap > 0)
	218	memset(av->dccpav_buf + av->dccpav_buf_head + 1,
	219	DCCP_ACKVEC_STATE_NOT_RECEIVED, gap);
	220
	221	av->dccpav_buf[av->dccpav_buf_head] = state;
	222	av->dccpav_vec_len += packets;
	223	return 0;
	224	}
	225
	226	/*
0e64e94e	227	* Implements the RFC 4340, Appendix A
ae31c339 ACM	228	*/
	229	int dccp_ackvec_add(struct dccp_ackvec av, const struct sock sk,
	230	const u64 ackno, const u8 state)
	231	{
	232	/*
	233	* Check at the right places if the buffer is full, if it is, tell the
	234	* caller to start dropping packets till the HC-Sender acks our ACK
	235	* vectors, when we will free up space in dccpav_buf.
	236	*
	237	* We may well decide to do buffer compression, etc, but for now lets
	238	* just drop.
	239	*
0e64e94e	240	* From Appendix A.1.1 (`New Packets'):
ae31c339 ACM	241	*
	242	* Of course, the circular buffer may overflow, either when the
	243	* HC-Sender is sending data at a very high rate, when the
	244	* HC-Receiver's acknowledgements are not reaching the HC-Sender,
	245	* or when the HC-Sender is forgetting to acknowledge those acks
	246	* (so the HC-Receiver is unable to clean up old state). In this
	247	* case, the HC-Receiver should either compress the buffer (by
	248	* increasing run lengths when possible), transfer its state to
	249	* a larger buffer, or, as a last resort, drop all received
	250	* packets, without processing them whatsoever, until its buffer
	251	* shrinks again.
	252	*/
	253
	254	/* See if this is the first ackno being inserted */
	255	if (av->dccpav_vec_len == 0) {
	256	av->dccpav_buf[av->dccpav_buf_head] = state;
	257	av->dccpav_vec_len = 1;
	258	} else if (after48(ackno, av->dccpav_buf_ackno)) {
	259	const u64 delta = dccp_delta_seqno(av->dccpav_buf_ackno,
	260	ackno);
	261
	262	/*
	263	* Look if the state of this packet is the same as the
	264	* previous ackno and if so if we can bump the head len.
	265	*/
	266	if (delta == 1 &&
	267	dccp_ackvec_state(av, av->dccpav_buf_head) == state &&
	268	(dccp_ackvec_len(av, av->dccpav_buf_head) <
	269	DCCP_ACKVEC_LEN_MASK))
	270	av->dccpav_buf[av->dccpav_buf_head]++;
	271	else if (dccp_ackvec_set_buf_head_state(av, delta, state))
	272	return -ENOBUFS;
	273	} else {
	274	/*
	275	* A.1.2. Old Packets
	276	*
0e64e94e GR	277	* When a packet with Sequence Number S <= buf_ackno
	278	* arrives, the HC-Receiver will scan the table for
	279	* the byte corresponding to S. (Indexing structures
ae31c339 ACM	280	* could reduce the complexity of this scan.)
	281	*/
	282	u64 delta = dccp_delta_seqno(ackno, av->dccpav_buf_ackno);
e4dfd449	283	u8 index = av->dccpav_buf_head;
ae31c339 ACM	284
	285	while (1) {
	286	const u8 len = dccp_ackvec_len(av, index);
	287	const u8 state = dccp_ackvec_state(av, index);
	288	/*
	289	* valid packets not yet in dccpav_buf have a reserved
	290	* entry, with a len equal to 0.
	291	*/
	292	if (state == DCCP_ACKVEC_STATE_NOT_RECEIVED &&
	293	len == 0 && delta == 0) { /* Found our
	294	reserved seat! */
	295	dccp_pr_debug("Found %llu reserved seat!\n",
	296	(unsigned long long)ackno);
	297	av->dccpav_buf[index] = state;
	298	goto out;
	299	}
	300	/* len == 0 means one packet */
	301	if (delta < len + 1)
	302	goto out_duplicate;
	303
	304	delta -= len + 1;
7400d781	305	if (++index == DCCP_MAX_ACKVEC_LEN)
ae31c339 ACM	306	index = 0;
	307	}
	308	}
	309
	310	av->dccpav_buf_ackno = ackno;
	311	dccp_timestamp(sk, &av->dccpav_time);
	312	out:
ae31c339 ACM	313	return 0;
	314
	315	out_duplicate:
	316	/* Duplicate packet */
	317	dccp_pr_debug("Received a dup or already considered lost "
	318	"packet: %llu\n", (unsigned long long)ackno);
	319	return -EILSEQ;
	320	}
	321
	322	#ifdef CONFIG_IP_DCCP_DEBUG
	323	void dccp_ackvector_print(const u64 ackno, const unsigned char *vector, int len)
	324	{
	325	if (!dccp_debug)
	326	return;
	327
	328	printk("ACK vector len=%d, ackno=%llu \|", len,
	329	(unsigned long long)ackno);
	330
	331	while (len--) {
	332	const u8 state = (*vector & DCCP_ACKVEC_STATE_MASK) >> 6;
	333	const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
	334
	335	printk("%d,%d\|", state, rl);
	336	++vector;
	337	}
	338
	339	printk("\n");
	340	}
	341
	342	void dccp_ackvec_print(const struct dccp_ackvec *av)
	343	{
	344	dccp_ackvector_print(av->dccpav_buf_ackno,
	345	av->dccpav_buf + av->dccpav_buf_head,
	346	av->dccpav_vec_len);
	347	}
	348	#endif
	349
02bcf28c AB	350	static void dccp_ackvec_throw_record(struct dccp_ackvec *av,
02bcf28c AB	351	struct dccp_ackvec_record *avr)
ae31c339	352	{
02bcf28c AB	353	struct dccp_ackvec_record *next;
02bcf28c AB	354
23d06e3b AB	355	/* sort out vector length */
	356	if (av->dccpav_buf_head <= avr->dccpavr_ack_ptr)
	357	av->dccpav_vec_len = avr->dccpavr_ack_ptr - av->dccpav_buf_head;
	358	else
	359	av->dccpav_vec_len = DCCP_MAX_ACKVEC_LEN - 1
	360	- av->dccpav_buf_head
	361	+ avr->dccpavr_ack_ptr;
02bcf28c AB	362
	363	/* free records */
	364	list_for_each_entry_safe_from(avr, next, &av->dccpav_records,
	365	dccpavr_node) {
	366	list_del_init(&avr->dccpavr_node);
	367	dccp_ackvec_record_delete(avr);
	368	}
ae31c339 ACM	369	}
	370
	371	void dccp_ackvec_check_rcv_ackno(struct dccp_ackvec av, struct sock sk,
	372	const u64 ackno)
	373	{
02bcf28c	374	struct dccp_ackvec_record *avr;
ae31c339	375
02bcf28c AB	376	/*
	377	* If we traverse backwards, it should be faster when we have large
	378	* windows. We will be receiving ACKs for stuff we sent a while back
	379	* -sorbo.
	380	*/
	381	list_for_each_entry_reverse(avr, &av->dccpav_records, dccpavr_node) {
	382	if (ackno == avr->dccpavr_ack_seqno) {
ae31c339	383	#ifdef CONFIG_IP_DCCP_DEBUG
02bcf28c AB	384	struct dccp_sock *dp = dccp_sk(sk);
	385	const char *debug_prefix = dp->dccps_role == DCCP_ROLE_CLIENT ?
	386	"CLIENT rx ack: " : "server rx ack: ";
ae31c339	387	#endif
02bcf28c AB	388	dccp_pr_debug("%sACK packet 0, len=%d, ack_seqno=%llu, "
	389	"ack_ackno=%llu, ACKED!\n",
	390	debug_prefix, 1,
	391	(unsigned long long)avr->dccpavr_ack_seqno,
	392	(unsigned long long)avr->dccpavr_ack_ackno);
	393	dccp_ackvec_throw_record(av, avr);
	394	break;
	395	}
ae31c339 ACM	396	}
	397	}
	398
	399	static void dccp_ackvec_check_rcv_ackvector(struct dccp_ackvec *av,
	400	struct sock *sk, u64 ackno,
	401	const unsigned char len,
	402	const unsigned char *vector)
	403	{
	404	unsigned char i;
02bcf28c	405	struct dccp_ackvec_record *avr;
ae31c339 ACM	406
ae31c339 ACM	407	/* Check if we actually sent an ACK vector */
02bcf28c	408	if (list_empty(&av->dccpav_records))
ae31c339	409	return;
ae31c339 ACM	410
ae31c339 ACM	411	i = len;
02bcf28c AB	412	/*
	413	* XXX
	414	* I think it might be more efficient to work backwards. See comment on
	415	* rcv_ackno. -sorbo.
	416	*/
	417	avr = list_entry(av->dccpav_records.next, struct dccp_ackvec_record,
	418	dccpavr_node);
ae31c339 ACM	419	while (i--) {
	420	const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
	421	u64 ackno_end_rl;
	422
	423	dccp_set_seqno(&ackno_end_rl, ackno - rl);
	424
	425	/*
02bcf28c AB	426	* If our AVR sequence number is greater than the ack, go
02bcf28c AB	427	* forward in the AVR list until it is not so.
ae31c339	428	*/
02bcf28c AB	429	list_for_each_entry_from(avr, &av->dccpav_records,
	430	dccpavr_node) {
	431	if (!after48(avr->dccpavr_ack_seqno, ackno))
	432	goto found;
	433	}
	434	/* End of the dccpav_records list, not found, exit */
	435	break;
	436	found:
	437	if (between48(avr->dccpavr_ack_seqno, ackno_end_rl, ackno)) {
8e27e465	438	const u8 state = *vector & DCCP_ACKVEC_STATE_MASK;
ae31c339 ACM	439	if (state != DCCP_ACKVEC_STATE_NOT_RECEIVED) {
	440	#ifdef CONFIG_IP_DCCP_DEBUG
	441	struct dccp_sock *dp = dccp_sk(sk);
	442	const char *debug_prefix =
	443	dp->dccps_role == DCCP_ROLE_CLIENT ?
	444	"CLIENT rx ack: " : "server rx ack: ";
	445	#endif
	446	dccp_pr_debug("%sACK vector 0, len=%d, "
	447	"ack_seqno=%llu, ack_ackno=%llu, "
	448	"ACKED!\n",
	449	debug_prefix, len,
	450	(unsigned long long)
02bcf28c	451	avr->dccpavr_ack_seqno,
ae31c339	452	(unsigned long long)
02bcf28c AB	453	avr->dccpavr_ack_ackno);
02bcf28c AB	454	dccp_ackvec_throw_record(av, avr);
afec35e3	455	break;
ae31c339 ACM	456	}
ae31c339 ACM	457	/*
02bcf28c AB	458	* If it wasn't received, continue scanning... we might
02bcf28c AB	459	* find another one.
ae31c339	460	*/
ae31c339	461	}
ae31c339 ACM	462
	463	dccp_set_seqno(&ackno, ackno_end_rl - 1);
	464	++vector;
	465	}
	466	}
	467
	468	int dccp_ackvec_parse(struct sock sk, const struct sk_buff skb,
	469	const u8 opt, const u8 *value, const u8 len)
	470	{
	471	if (len > DCCP_MAX_ACKVEC_LEN)
	472	return -1;
	473
	474	/* dccp_ackvector_print(DCCP_SKB_CB(skb)->dccpd_ack_seq, value, len); */
	475	dccp_ackvec_check_rcv_ackvector(dccp_sk(sk)->dccps_hc_rx_ackvec, sk,
	476	DCCP_SKB_CB(skb)->dccpd_ack_seq,
	477	len, value);
	478	return 0;
	479	}
9b07ef5d ACM	480
9b07ef5d ACM	481	static char dccp_ackvec_slab_msg[] __initdata =
02bcf28c	482	KERN_CRIT "DCCP: Unable to create ack vectors slab caches\n";
9b07ef5d ACM	483
	484	int __init dccp_ackvec_init(void)
	485	{
	486	dccp_ackvec_slab = kmem_cache_create("dccp_ackvec",
	487	sizeof(struct dccp_ackvec), 0,
	488	SLAB_HWCACHE_ALIGN, NULL, NULL);
02bcf28c AB	489	if (dccp_ackvec_slab == NULL)
	490	goto out_err;
	491
	492	dccp_ackvec_record_slab =
	493	kmem_cache_create("dccp_ackvec_record",
	494	sizeof(struct dccp_ackvec_record),
	495	0, SLAB_HWCACHE_ALIGN, NULL, NULL);
	496	if (dccp_ackvec_record_slab == NULL)
	497	goto out_destroy_slab;
9b07ef5d ACM	498
9b07ef5d ACM	499	return 0;
02bcf28c AB	500
	501	out_destroy_slab:
	502	kmem_cache_destroy(dccp_ackvec_slab);
	503	dccp_ackvec_slab = NULL;
	504	out_err:
	505	printk(dccp_ackvec_slab_msg);
	506	return -ENOBUFS;
9b07ef5d ACM	507	}
	508
	509	void dccp_ackvec_exit(void)
	510	{
	511	if (dccp_ackvec_slab != NULL) {
	512	kmem_cache_destroy(dccp_ackvec_slab);
	513	dccp_ackvec_slab = NULL;
	514	}
02bcf28c AB	515	if (dccp_ackvec_record_slab != NULL) {
	516	kmem_cache_destroy(dccp_ackvec_record_slab);
	517	dccp_ackvec_record_slab = NULL;
	518	}
9b07ef5d	519	}