[linux.git] / net / dccp / ackvec.c

/*
 *  net/dccp/ackvec.c
 *
 *  An implementation of the DCCP protocol
 *  Copyright (c) 2005 Arnaldo Carvalho de Melo <[email protected]>
 *
 *      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 struct kmem_cache *dccp_ackvec_slab;
static struct kmem_cache *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);
	struct dccp_ackvec *av = dp->dccps_hc_rx_ackvec;
	/* Figure out how many options do we need to represent the ackvec */
	const u16 nr_opts = (av->dccpav_vec_len +
			     DCCP_MAX_ACKVEC_OPT_LEN - 1) /
			    DCCP_MAX_ACKVEC_OPT_LEN;
	u16 len = av->dccpav_vec_len + 2 * nr_opts, i;
	struct timeval now;
	u32 elapsed_time;
	const unsigned char *tail, *from;
	unsigned char *to;
	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);
	len  = av->dccpav_vec_len;
	from = av->dccpav_buf + av->dccpav_buf_head;
	tail = av->dccpav_buf + DCCP_MAX_ACKVEC_LEN;

	for (i = 0; i < nr_opts; ++i) {
		int copylen = len;

		if (len > DCCP_MAX_ACKVEC_OPT_LEN)
			copylen = DCCP_MAX_ACKVEC_OPT_LEN;

		*to++ = DCCPO_ACK_VECTOR_0;
		*to++ = copylen + 2;

		/* Check if buf_head wraps */
		if (from + copylen > tail) {
			const u16 tailsize = tail - from;

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

		memcpy(to, from, copylen);
		from += copylen;
		to   += copylen;
		len  -= copylen;
	}

	/*
	 *	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("%s ACK Vector 0, len=%d, ack_seqno=%llu, "
		      "ack_ackno=%llu\n",
		      dccp_role(sk), 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	= UINT48_MAX + 1;
		av->dccpav_buf_nonce = av->dccpav_buf_nonce = 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 u32 index)
{
	return av->dccpav_buf[index] & DCCP_ACKVEC_STATE_MASK;
}

static inline u8 dccp_ackvec_len(const struct dccp_ackvec *av,
				 const u32 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);
		u32 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)
{
	dccp_pr_debug_cat("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;

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

	dccp_pr_debug_cat("\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) {
			dccp_pr_debug("%s ACK packet 0, len=%d, ack_seqno=%llu, "
				      "ack_ackno=%llu, ACKED!\n",
				      dccp_role(sk), 1,
				      (unsigned long long)avr->dccpavr_ack_seqno,
				      (unsigned long long)avr->dccpavr_ack_ackno);
			dccp_ackvec_throw_record(av, avr);
			break;
		} else if (avr->dccpavr_ack_seqno > ackno)
			break; /* old news */
	}
}

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) {
				dccp_pr_debug("%s ACK vector 0, len=%d, "
					      "ack_seqno=%llu, ack_ackno=%llu, "
					      "ACKED!\n",
					      dccp_role(sk), 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,
		      u64 *ackno, const u8 opt, const u8 *value, const u8 len)
{
	if (len > DCCP_MAX_ACKVEC_OPT_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,
					ackno, len, value);
	return 0;
}

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:
	DCCP_CRIT("Unable to create Ack Vector slab cache");
	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 <[email protected]>
	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
e18b890b CL	24	static struct kmem_cache *dccp_ackvec_slab;
e18b890b CL	25	static struct kmem_cache *dccp_ackvec_record_slab;
02bcf28c AB	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);
	70	struct dccp_ackvec *av = dp->dccps_hc_rx_ackvec;
522f1d09 AB	71	/* Figure out how many options do we need to represent the ackvec */
	72	const u16 nr_opts = (av->dccpav_vec_len +
	73	DCCP_MAX_ACKVEC_OPT_LEN - 1) /
	74	DCCP_MAX_ACKVEC_OPT_LEN;
	75	u16 len = av->dccpav_vec_len + 2 * nr_opts, i;
ae31c339 ACM	76	struct timeval now;
ae31c339 ACM	77	u32 elapsed_time;
522f1d09 AB	78	const unsigned char tail, from;
522f1d09 AB	79	unsigned char *to;
02bcf28c AB	80	struct dccp_ackvec_record *avr;
	81
	82	if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
	83	return -1;
	84
ae31c339 ACM	85	dccp_timestamp(sk, &now);
	86	elapsed_time = timeval_delta(&now, &av->dccpav_time) / 10;
	87
2d0817d1 ACM	88	if (elapsed_time != 0 &&
	89	dccp_insert_option_elapsed_time(sk, skb, elapsed_time))
	90	return -1;
	91
	92	avr = dccp_ackvec_record_new();
	93	if (avr == NULL)
	94	return -1;
ae31c339	95
ae31c339 ACM	96	DCCP_SKB_CB(skb)->dccpd_opt_len += len;
ae31c339 ACM	97
522f1d09	98	to = skb_push(skb, len);
ae31c339 ACM	99	len = av->dccpav_vec_len;
ae31c339 ACM	100	from = av->dccpav_buf + av->dccpav_buf_head;
522f1d09 AB	101	tail = av->dccpav_buf + DCCP_MAX_ACKVEC_LEN;
	102
	103	for (i = 0; i < nr_opts; ++i) {
	104	int copylen = len;
	105
	106	if (len > DCCP_MAX_ACKVEC_OPT_LEN)
	107	copylen = DCCP_MAX_ACKVEC_OPT_LEN;
	108
	109	*to++ = DCCPO_ACK_VECTOR_0;
	110	*to++ = copylen + 2;
	111
	112	/* Check if buf_head wraps */
	113	if (from + copylen > tail) {
	114	const u16 tailsize = tail - from;
ae31c339	115
522f1d09 AB	116	memcpy(to, from, tailsize);
	117	to += tailsize;
	118	len -= tailsize;
	119	copylen -= tailsize;
	120	from = av->dccpav_buf;
	121	}
ae31c339	122
522f1d09 AB	123	memcpy(to, from, copylen);
	124	from += copylen;
	125	to += copylen;
	126	len -= copylen;
ae31c339 ACM	127	}
ae31c339 ACM	128
ae31c339	129	/*
0e64e94e	130	* From RFC 4340, A.2:
ae31c339 ACM	131	*
	132	* For each acknowledgement it sends, the HC-Receiver will add an
	133	* acknowledgement record. ack_seqno will equal the HC-Receiver
	134	* sequence number it used for the ack packet; ack_ptr will equal
	135	* buf_head; ack_ackno will equal buf_ackno; and ack_nonce will
	136	* equal buf_nonce.
ae31c339	137	*/
02bcf28c AB	138	avr->dccpavr_ack_seqno = DCCP_SKB_CB(skb)->dccpd_seq;
	139	avr->dccpavr_ack_ptr = av->dccpav_buf_head;
	140	avr->dccpavr_ack_ackno = av->dccpav_buf_ackno;
	141	avr->dccpavr_ack_nonce = av->dccpav_buf_nonce;
	142	avr->dccpavr_sent_len = av->dccpav_vec_len;
	143
	144	dccp_ackvec_insert_avr(av, avr);
ae31c339	145
09dbc389	146	dccp_pr_debug("%s ACK Vector 0, len=%d, ack_seqno=%llu, "
ae31c339	147	"ack_ackno=%llu\n",
09dbc389	148	dccp_role(sk), avr->dccpavr_sent_len,
02bcf28c AB	149	(unsigned long long)avr->dccpavr_ack_seqno,
	150	(unsigned long long)avr->dccpavr_ack_ackno);
	151	return 0;
ae31c339 ACM	152	}
ae31c339 ACM	153
7400d781	154	struct dccp_ackvec *dccp_ackvec_alloc(const gfp_t priority)
ae31c339	155	{
9b07ef5d	156	struct dccp_ackvec *av = kmem_cache_alloc(dccp_ackvec_slab, priority);
e4dfd449	157
ae31c339	158	if (av != NULL) {
4a0a50fb	159	av->dccpav_buf_head = DCCP_MAX_ACKVEC_LEN - 1;
6b811d43	160	av->dccpav_buf_ackno = UINT48_MAX + 1;
ae31c339	161	av->dccpav_buf_nonce = av->dccpav_buf_nonce = 0;
ae31c339 ACM	162	av->dccpav_time.tv_sec = 0;
ae31c339 ACM	163	av->dccpav_time.tv_usec = 0;
4a0a50fb	164	av->dccpav_vec_len = 0;
02bcf28c	165	INIT_LIST_HEAD(&av->dccpav_records);
ae31c339 ACM	166	}
	167
	168	return av;
	169	}
	170
	171	void dccp_ackvec_free(struct dccp_ackvec *av)
	172	{
02bcf28c AB	173	if (unlikely(av == NULL))
02bcf28c AB	174	return;
d5e9b2c7 ACM	175
	176	if (!list_empty(&av->dccpav_records)) {
	177	struct dccp_ackvec_record avr, next;
	178
	179	list_for_each_entry_safe(avr, next, &av->dccpav_records,
	180	dccpavr_node) {
	181	list_del_init(&avr->dccpavr_node);
	182	dccp_ackvec_record_delete(avr);
	183	}
	184	}
	185
9b07ef5d	186	kmem_cache_free(dccp_ackvec_slab, av);
ae31c339 ACM	187	}
	188
	189	static inline u8 dccp_ackvec_state(const struct dccp_ackvec *av,
bdf13d20	190	const u32 index)
ae31c339 ACM	191	{
	192	return av->dccpav_buf[index] & DCCP_ACKVEC_STATE_MASK;
	193	}
	194
	195	static inline u8 dccp_ackvec_len(const struct dccp_ackvec *av,
bdf13d20	196	const u32 index)
ae31c339 ACM	197	{
	198	return av->dccpav_buf[index] & DCCP_ACKVEC_LEN_MASK;
	199	}
	200
	201	/*
	202	* If several packets are missing, the HC-Receiver may prefer to enter multiple
	203	* bytes with run length 0, rather than a single byte with a larger run length;
	204	* this simplifies table updates if one of the missing packets arrives.
	205	*/
	206	static inline int dccp_ackvec_set_buf_head_state(struct dccp_ackvec *av,
	207	const unsigned int packets,
e4dfd449	208	const unsigned char state)
ae31c339 ACM	209	{
ae31c339 ACM	210	unsigned int gap;
a8fc3d8d	211	long new_head;
ae31c339	212
7400d781	213	if (av->dccpav_vec_len + packets > DCCP_MAX_ACKVEC_LEN)
ae31c339 ACM	214	return -ENOBUFS;
	215
	216	gap = packets - 1;
	217	new_head = av->dccpav_buf_head - packets;
	218
	219	if (new_head < 0) {
	220	if (gap > 0) {
	221	memset(av->dccpav_buf, DCCP_ACKVEC_STATE_NOT_RECEIVED,
	222	gap + new_head + 1);
	223	gap = -new_head;
	224	}
7400d781	225	new_head += DCCP_MAX_ACKVEC_LEN;
8109b02b	226	}
ae31c339 ACM	227
	228	av->dccpav_buf_head = new_head;
	229
	230	if (gap > 0)
	231	memset(av->dccpav_buf + av->dccpav_buf_head + 1,
	232	DCCP_ACKVEC_STATE_NOT_RECEIVED, gap);
	233
	234	av->dccpav_buf[av->dccpav_buf_head] = state;
	235	av->dccpav_vec_len += packets;
	236	return 0;
	237	}
	238
	239	/*
0e64e94e	240	* Implements the RFC 4340, Appendix A
ae31c339 ACM	241	*/
	242	int dccp_ackvec_add(struct dccp_ackvec av, const struct sock sk,
	243	const u64 ackno, const u8 state)
	244	{
	245	/*
	246	* Check at the right places if the buffer is full, if it is, tell the
	247	* caller to start dropping packets till the HC-Sender acks our ACK
	248	* vectors, when we will free up space in dccpav_buf.
	249	*
	250	* We may well decide to do buffer compression, etc, but for now lets
	251	* just drop.
	252	*
0e64e94e	253	* From Appendix A.1.1 (`New Packets'):
ae31c339 ACM	254	*
	255	* Of course, the circular buffer may overflow, either when the
	256	* HC-Sender is sending data at a very high rate, when the
	257	* HC-Receiver's acknowledgements are not reaching the HC-Sender,
	258	* or when the HC-Sender is forgetting to acknowledge those acks
	259	* (so the HC-Receiver is unable to clean up old state). In this
	260	* case, the HC-Receiver should either compress the buffer (by
	261	* increasing run lengths when possible), transfer its state to
	262	* a larger buffer, or, as a last resort, drop all received
	263	* packets, without processing them whatsoever, until its buffer
	264	* shrinks again.
	265	*/
	266
	267	/* See if this is the first ackno being inserted */
	268	if (av->dccpav_vec_len == 0) {
	269	av->dccpav_buf[av->dccpav_buf_head] = state;
	270	av->dccpav_vec_len = 1;
	271	} else if (after48(ackno, av->dccpav_buf_ackno)) {
	272	const u64 delta = dccp_delta_seqno(av->dccpav_buf_ackno,
	273	ackno);
	274
	275	/*
	276	* Look if the state of this packet is the same as the
	277	* previous ackno and if so if we can bump the head len.
	278	*/
	279	if (delta == 1 &&
	280	dccp_ackvec_state(av, av->dccpav_buf_head) == state &&
	281	(dccp_ackvec_len(av, av->dccpav_buf_head) <
	282	DCCP_ACKVEC_LEN_MASK))
	283	av->dccpav_buf[av->dccpav_buf_head]++;
	284	else if (dccp_ackvec_set_buf_head_state(av, delta, state))
	285	return -ENOBUFS;
	286	} else {
	287	/*
	288	* A.1.2. Old Packets
	289	*
0e64e94e GR	290	* When a packet with Sequence Number S <= buf_ackno
	291	* arrives, the HC-Receiver will scan the table for
	292	* the byte corresponding to S. (Indexing structures
ae31c339 ACM	293	* could reduce the complexity of this scan.)
	294	*/
	295	u64 delta = dccp_delta_seqno(ackno, av->dccpav_buf_ackno);
bdf13d20	296	u32 index = av->dccpav_buf_head;
ae31c339 ACM	297
	298	while (1) {
	299	const u8 len = dccp_ackvec_len(av, index);
	300	const u8 state = dccp_ackvec_state(av, index);
	301	/*
	302	* valid packets not yet in dccpav_buf have a reserved
	303	* entry, with a len equal to 0.
	304	*/
	305	if (state == DCCP_ACKVEC_STATE_NOT_RECEIVED &&
	306	len == 0 && delta == 0) { /* Found our
	307	reserved seat! */
	308	dccp_pr_debug("Found %llu reserved seat!\n",
	309	(unsigned long long)ackno);
	310	av->dccpav_buf[index] = state;
	311	goto out;
	312	}
	313	/* len == 0 means one packet */
	314	if (delta < len + 1)
	315	goto out_duplicate;
	316
	317	delta -= len + 1;
7400d781	318	if (++index == DCCP_MAX_ACKVEC_LEN)
ae31c339 ACM	319	index = 0;
	320	}
	321	}
	322
	323	av->dccpav_buf_ackno = ackno;
	324	dccp_timestamp(sk, &av->dccpav_time);
	325	out:
ae31c339 ACM	326	return 0;
	327
	328	out_duplicate:
	329	/* Duplicate packet */
	330	dccp_pr_debug("Received a dup or already considered lost "
	331	"packet: %llu\n", (unsigned long long)ackno);
	332	return -EILSEQ;
	333	}
	334
	335	#ifdef CONFIG_IP_DCCP_DEBUG
	336	void dccp_ackvector_print(const u64 ackno, const unsigned char *vector, int len)
	337	{
84116716	338	dccp_pr_debug_cat("ACK vector len=%d, ackno=%llu \|", len,
8109b02b	339	(unsigned long long)ackno);
ae31c339 ACM	340
	341	while (len--) {
	342	const u8 state = (*vector & DCCP_ACKVEC_STATE_MASK) >> 6;
	343	const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
	344
84116716	345	dccp_pr_debug_cat("%d,%d\|", state, rl);
ae31c339 ACM	346	++vector;
	347	}
	348
84116716	349	dccp_pr_debug_cat("\n");
ae31c339 ACM	350	}
	351
	352	void dccp_ackvec_print(const struct dccp_ackvec *av)
	353	{
	354	dccp_ackvector_print(av->dccpav_buf_ackno,
	355	av->dccpav_buf + av->dccpav_buf_head,
	356	av->dccpav_vec_len);
	357	}
	358	#endif
	359
02bcf28c AB	360	static void dccp_ackvec_throw_record(struct dccp_ackvec *av,
02bcf28c AB	361	struct dccp_ackvec_record *avr)
ae31c339	362	{
02bcf28c AB	363	struct dccp_ackvec_record *next;
02bcf28c AB	364
23d06e3b AB	365	/* sort out vector length */
	366	if (av->dccpav_buf_head <= avr->dccpavr_ack_ptr)
	367	av->dccpav_vec_len = avr->dccpavr_ack_ptr - av->dccpav_buf_head;
	368	else
	369	av->dccpav_vec_len = DCCP_MAX_ACKVEC_LEN - 1
	370	- av->dccpav_buf_head
	371	+ avr->dccpavr_ack_ptr;
02bcf28c AB	372
	373	/* free records */
	374	list_for_each_entry_safe_from(avr, next, &av->dccpav_records,
	375	dccpavr_node) {
	376	list_del_init(&avr->dccpavr_node);
	377	dccp_ackvec_record_delete(avr);
	378	}
ae31c339 ACM	379	}
	380
	381	void dccp_ackvec_check_rcv_ackno(struct dccp_ackvec av, struct sock sk,
	382	const u64 ackno)
	383	{
02bcf28c	384	struct dccp_ackvec_record *avr;
ae31c339	385
02bcf28c AB	386	/*
	387	* If we traverse backwards, it should be faster when we have large
	388	* windows. We will be receiving ACKs for stuff we sent a while back
	389	* -sorbo.
	390	*/
	391	list_for_each_entry_reverse(avr, &av->dccpav_records, dccpavr_node) {
	392	if (ackno == avr->dccpavr_ack_seqno) {
09dbc389	393	dccp_pr_debug("%s ACK packet 0, len=%d, ack_seqno=%llu, "
02bcf28c	394	"ack_ackno=%llu, ACKED!\n",
09dbc389	395	dccp_role(sk), 1,
02bcf28c AB	396	(unsigned long long)avr->dccpavr_ack_seqno,
	397	(unsigned long long)avr->dccpavr_ack_ackno);
	398	dccp_ackvec_throw_record(av, avr);
	399	break;
d23ca15a AB	400	} else if (avr->dccpavr_ack_seqno > ackno)
d23ca15a AB	401	break; /* old news */
ae31c339 ACM	402	}
	403	}
	404
	405	static void dccp_ackvec_check_rcv_ackvector(struct dccp_ackvec *av,
bdf13d20	406	struct sock sk, u64 ackno,
ae31c339 ACM	407	const unsigned char len,
	408	const unsigned char *vector)
	409	{
	410	unsigned char i;
02bcf28c	411	struct dccp_ackvec_record *avr;
ae31c339 ACM	412
ae31c339 ACM	413	/* Check if we actually sent an ACK vector */
02bcf28c	414	if (list_empty(&av->dccpav_records))
ae31c339	415	return;
ae31c339 ACM	416
ae31c339 ACM	417	i = len;
02bcf28c AB	418	/*
	419	* XXX
	420	* I think it might be more efficient to work backwards. See comment on
	421	* rcv_ackno. -sorbo.
	422	*/
	423	avr = list_entry(av->dccpav_records.next, struct dccp_ackvec_record,
	424	dccpavr_node);
ae31c339 ACM	425	while (i--) {
	426	const u8 rl = *vector & DCCP_ACKVEC_LEN_MASK;
	427	u64 ackno_end_rl;
	428
bdf13d20	429	dccp_set_seqno(&ackno_end_rl, *ackno - rl);
ae31c339 ACM	430
ae31c339 ACM	431	/*
02bcf28c AB	432	* If our AVR sequence number is greater than the ack, go
02bcf28c AB	433	* forward in the AVR list until it is not so.
ae31c339	434	*/
02bcf28c AB	435	list_for_each_entry_from(avr, &av->dccpav_records,
02bcf28c AB	436	dccpavr_node) {
bdf13d20	437	if (!after48(avr->dccpavr_ack_seqno, *ackno))
02bcf28c AB	438	goto found;
	439	}
	440	/* End of the dccpav_records list, not found, exit */
	441	break;
	442	found:
bdf13d20	443	if (between48(avr->dccpavr_ack_seqno, ackno_end_rl, *ackno)) {
8e27e465	444	const u8 state = *vector & DCCP_ACKVEC_STATE_MASK;
ae31c339	445	if (state != DCCP_ACKVEC_STATE_NOT_RECEIVED) {
09dbc389	446	dccp_pr_debug("%s ACK vector 0, len=%d, "
ae31c339 ACM	447	"ack_seqno=%llu, ack_ackno=%llu, "
ae31c339 ACM	448	"ACKED!\n",
09dbc389	449	dccp_role(sk), len,
ae31c339	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	462
bdf13d20	463	dccp_set_seqno(ackno, ackno_end_rl - 1);
ae31c339 ACM	464	++vector;
	465	}
	466	}
	467
	468	int dccp_ackvec_parse(struct sock sk, const struct sk_buff skb,
bdf13d20	469	u64 ackno, const u8 opt, const u8 value, const u8 len)
ae31c339	470	{
bdf13d20	471	if (len > DCCP_MAX_ACKVEC_OPT_LEN)
ae31c339 ACM	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,
bdf13d20	476	ackno, len, value);
ae31c339 ACM	477	return 0;
ae31c339 ACM	478	}
9b07ef5d	479
9b07ef5d ACM	480	int __init dccp_ackvec_init(void)
	481	{
	482	dccp_ackvec_slab = kmem_cache_create("dccp_ackvec",
	483	sizeof(struct dccp_ackvec), 0,
	484	SLAB_HWCACHE_ALIGN, NULL, NULL);
02bcf28c AB	485	if (dccp_ackvec_slab == NULL)
	486	goto out_err;
	487
	488	dccp_ackvec_record_slab =
	489	kmem_cache_create("dccp_ackvec_record",
	490	sizeof(struct dccp_ackvec_record),
	491	0, SLAB_HWCACHE_ALIGN, NULL, NULL);
	492	if (dccp_ackvec_record_slab == NULL)
	493	goto out_destroy_slab;
9b07ef5d ACM	494
9b07ef5d ACM	495	return 0;
02bcf28c AB	496
	497	out_destroy_slab:
	498	kmem_cache_destroy(dccp_ackvec_slab);
	499	dccp_ackvec_slab = NULL;
	500	out_err:
59348b19	501	DCCP_CRIT("Unable to create Ack Vector slab cache");
02bcf28c	502	return -ENOBUFS;
9b07ef5d ACM	503	}
	504
	505	void dccp_ackvec_exit(void)
	506	{
	507	if (dccp_ackvec_slab != NULL) {
	508	kmem_cache_destroy(dccp_ackvec_slab);
	509	dccp_ackvec_slab = NULL;
	510	}
02bcf28c AB	511	if (dccp_ackvec_record_slab != NULL) {
	512	kmem_cache_destroy(dccp_ackvec_record_slab);
	513	dccp_ackvec_record_slab = NULL;
	514	}
9b07ef5d	515	}