Merge http://linux-lksctp.bkbits.net/lksctp-2.5.work

into touki.austin.ibm.com:/home/jgrimm/bk/lksctp-2.5.work

include/net/sctp/command.h

@@ -110,13 +110,13 @@ typedef union {
 	sctp_event_timeout_t to;
 	sctp_counter_t counter;
 	void *ptr;
-	sctp_chunk_t *chunk;
-	sctp_association_t *asoc;
+	struct sctp_chunk *chunk;
+	struct sctp_association *asoc;
 	struct sctp_transport *transport;
-	sctp_bind_addr_t  *bp;
+	struct sctp_bind_addr *bp;
 	sctp_init_chunk_t *init;
 	struct sctp_ulpevent *ulpevent;
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 	sctp_sackhdr_t *sackh;
 } sctp_arg_t;
 
@@ -158,13 +158,13 @@ SCTP_ARG_CONSTRUCTOR(STATE,	sctp_state_t, state)
 SCTP_ARG_CONSTRUCTOR(COUNTER,	sctp_counter_t, counter)
 SCTP_ARG_CONSTRUCTOR(TO,	sctp_event_timeout_t, to)
 SCTP_ARG_CONSTRUCTOR(PTR,	void *, ptr)
-SCTP_ARG_CONSTRUCTOR(CHUNK,	sctp_chunk_t *, chunk)
-SCTP_ARG_CONSTRUCTOR(ASOC,	sctp_association_t *, asoc)
+SCTP_ARG_CONSTRUCTOR(CHUNK,	struct sctp_chunk *, chunk)
+SCTP_ARG_CONSTRUCTOR(ASOC,	struct sctp_association *, asoc)
 SCTP_ARG_CONSTRUCTOR(TRANSPORT,	struct sctp_transport *, transport)
-SCTP_ARG_CONSTRUCTOR(BA,	sctp_bind_addr_t *, bp)
+SCTP_ARG_CONSTRUCTOR(BA,	struct sctp_bind_addr *, bp)
 SCTP_ARG_CONSTRUCTOR(PEER_INIT,	sctp_init_chunk_t *, init)
 SCTP_ARG_CONSTRUCTOR(ULPEVENT,  struct sctp_ulpevent *, ulpevent)
-SCTP_ARG_CONSTRUCTOR(PACKET,	sctp_packet_t *, packet)
+SCTP_ARG_CONSTRUCTOR(PACKET,	struct sctp_packet *, packet)
 SCTP_ARG_CONSTRUCTOR(SACKH,	sctp_sackhdr_t *, sackh)
 
 typedef struct {

include/net/sctp/constants.h

@@ -345,6 +345,7 @@ typedef enum {
 	SCTP_XMIT_PMTU_FULL,
 	SCTP_XMIT_RWND_FULL,
 	SCTP_XMIT_MUST_FRAG,
+	SCTP_XMIT_NAGLE_DELAY,
 } sctp_xmit_t;
 
 /* These are the commands for manipulating transports.  */

include/net/sctp/sctp.h

@@ -313,31 +313,21 @@ static inline void sctp_sysctl_unregister(void) { return; }
 #endif
 
 
+/* Size of Supported Address Parameter for 'x' address types. */
+#define SCTP_SAT_LEN(x) (sizeof(struct sctp_paramhdr) + (x) * sizeof(__u16))
+
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 
 extern int sctp_v6_init(void);
 extern void sctp_v6_exit(void);
-
 static inline int sctp_ipv6_addr_type(const struct in6_addr *addr)
 {
 	return ipv6_addr_type((struct in6_addr*) addr);
 }
 
-/* Size of Supported Address Parameter for 'x' address types. */
-#define SCTP_SAT_LEN(x) (sizeof(struct sctp_paramhdr) + (x) * sizeof(__u16))
-
-/* Note: These V6 macros are obsolescent.  */
-/* Use this macro to enclose code fragments which are V6-dependent. */
-#define SCTP_V6(m...)	m
-#define SCTP_V6_SUPPORT 1
-
 #else /* #ifdef defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
 
 #define sctp_ipv6_addr_type(a) 0
-#define SCTP_SAT_LEN (sizeof(sctp_paramhdr_t) + 1 * sizeof(__u16))
-#define SCTP_V6(m...) /* Do nothing. */
-#undef SCTP_V6_SUPPORT
-
 static inline int sctp_v6_init(void) { return 0; }
 static inline void sctp_v6_exit(void) { return; }
 

include/net/sctp/sm.h

@@ -313,18 +313,18 @@ void sctp_generate_t3_rtx_event(unsigned long peer);
 void sctp_generate_heartbeat_event(unsigned long peer);
 
 sctp_sackhdr_t *sctp_sm_pull_sack(sctp_chunk_t *);
-sctp_packet_t *sctp_abort_pkt_new(const sctp_endpoint_t *ep,
-				  const sctp_association_t *asoc,
-				  sctp_chunk_t *chunk,
+struct sctp_packet *sctp_abort_pkt_new(const struct sctp_endpoint *,
+				       const struct sctp_association *,
+				       struct sctp_chunk *chunk,
 				       const void *payload,
 				       size_t paylen);
-sctp_packet_t *sctp_ootb_pkt_new(const sctp_association_t *asoc,
-				 const sctp_chunk_t *chunk);
-void sctp_ootb_pkt_free(sctp_packet_t *packet);
+struct sctp_packet *sctp_ootb_pkt_new(const struct sctp_association *,
+				      const struct sctp_chunk *);
+void sctp_ootb_pkt_free(struct sctp_packet *);
 
 sctp_cookie_param_t *
-sctp_pack_cookie(const sctp_endpoint_t *, const sctp_association_t *,
-		 const sctp_chunk_t *, int *cookie_len,
+sctp_pack_cookie(const struct sctp_endpoint *, const struct sctp_association *,
+		 const struct sctp_chunk *, int *cookie_len,
 		 const __u8 *, int addrs_len);
 sctp_association_t *sctp_unpack_cookie(const sctp_endpoint_t *,
 				       const sctp_association_t *,

include/net/sctp/structs.h

@@ -88,7 +88,6 @@ struct sctp_ssnmap;
 
 typedef struct sctp_endpoint sctp_endpoint_t;
 typedef struct sctp_association sctp_association_t;
-typedef struct sctp_packet sctp_packet_t;
 typedef struct sctp_chunk sctp_chunk_t;
 typedef struct sctp_bind_addr sctp_bind_addr_t;
 typedef struct sctp_endpoint_common sctp_endpoint_common_t;
@@ -604,26 +603,26 @@ struct sctp_packet {
 
 typedef int (sctp_outq_thandler_t)(struct sctp_outq *, void *);
 typedef int (sctp_outq_ehandler_t)(struct sctp_outq *);
-typedef sctp_packet_t *(sctp_outq_ohandler_init_t)
-	(sctp_packet_t *,
+typedef struct sctp_packet *(sctp_outq_ohandler_init_t)
+	(struct sctp_packet *,
          struct sctp_transport *,
          __u16 sport,
          __u16 dport);
-typedef sctp_packet_t *(sctp_outq_ohandler_config_t)
-        (sctp_packet_t *,
+typedef struct sctp_packet *(sctp_outq_ohandler_config_t)
+        (struct sctp_packet *,
 	 __u32 vtag,
 	 int ecn_capable,
 	 sctp_packet_phandler_t *get_prepend_chunk);
-typedef sctp_xmit_t (sctp_outq_ohandler_t)(sctp_packet_t *,
+typedef sctp_xmit_t (sctp_outq_ohandler_t)(struct sctp_packet *,
                                                sctp_chunk_t *);
-typedef int (sctp_outq_ohandler_force_t)(sctp_packet_t *);
+typedef int (sctp_outq_ohandler_force_t)(struct sctp_packet *);
 
 sctp_outq_ohandler_init_t    sctp_packet_init;
 sctp_outq_ohandler_config_t  sctp_packet_config;
 sctp_outq_ohandler_t         sctp_packet_append_chunk;
 sctp_outq_ohandler_t         sctp_packet_transmit_chunk;
 sctp_outq_ohandler_force_t   sctp_packet_transmit;
-void sctp_packet_free(sctp_packet_t *);
+void sctp_packet_free(struct sctp_packet *);
 
 
 /* This represents a remote transport address.
@@ -789,7 +788,7 @@ struct sctp_transport {
 	struct list_head transmitted;
 
 	/* We build bundle-able packets for this transport here.  */
-	sctp_packet_t packet;
+	struct sctp_packet packet;
 
 	/* This is the list of transports that have chunks to send.  */
 	struct list_head send_ready;
@@ -865,12 +864,11 @@ void sctp_inq_set_th_handler(struct sctp_inq *, void (*)(void *), void *);
 struct sctp_outq {
 	sctp_association_t *asoc;
 
-	/* BUG: This really should be an array of streams.
-	 * This really holds a list of chunks (one stream).
-	 * FIXME: If true, why so?
-	 */
+	/* Data pending that has never been transmitted.  */
 	struct sk_buff_head out;
 
+	unsigned out_qlen;	/* Total length of queued data chunks. */
+
 	/* These are control chunks we want to send.  */
 	struct sk_buff_head control;
 
@@ -885,7 +883,7 @@ struct sctp_outq {
 	struct list_head retransmit;
 
 	/* Call these functions to send chunks down to the next lower
-	 * layer.  This is always SCTP_packet, but we separate the two
+	 * layer.  This is always sctp_packet, but we separate the two
 	 * structures to make testing simpler.
 	 */
 	sctp_outq_ohandler_init_t	*init_output;

include/net/sctp/user.h

@@ -108,6 +108,8 @@ enum sctp_optname {
 #define SCTP_GET_LOCAL_ADDRS_NUM	SCTP_GET_LOCAL_ADDRS_NUM
 	SCTP_GET_LOCAL_ADDRS, 	/* Get all local addresss. */
 #define SCTP_GET_LOCAL_ADDRS	SCTP_GET_LOCAL_ADDRS
+	SCTP_NODELAY, 	/* Get/set nodelay option. */
+#define SCTP_NODELAY	SCTP_NODELAY
 };
 
 

net/sctp/output.c

@@ -62,16 +62,15 @@
 #include <net/sctp/sm.h>
 
 /* Forward declarations for private helpers. */
-static void sctp_packet_reset(sctp_packet_t *packet);
-static sctp_xmit_t sctp_packet_append_data(sctp_packet_t *packet,
-					   sctp_chunk_t *chunk);
+static void sctp_packet_reset(struct sctp_packet *packet);
+static sctp_xmit_t sctp_packet_append_data(struct sctp_packet *packet,
+					   struct sctp_chunk *chunk);
 
 /* Config a packet.
  * This appears to be a followup set of initializations.)
  */
-sctp_packet_t *sctp_packet_config(sctp_packet_t *packet,
-				  __u32 vtag,
-				  int ecn_capable,
+struct sctp_packet *sctp_packet_config(struct sctp_packet *packet,
+				       __u32 vtag, int ecn_capable,
 				       sctp_packet_phandler_t *prepend_handler)
 {
 	int packet_empty = (packet->size == SCTP_IP_OVERHEAD);
@@ -89,10 +88,9 @@ sctp_packet_t *sctp_packet_config(sctp_packet_t *packet,
 }
 
 /* Initialize the packet structure. */
-sctp_packet_t *sctp_packet_init(sctp_packet_t *packet,
+struct sctp_packet *sctp_packet_init(struct sctp_packet *packet,
 				     struct sctp_transport *transport,
-				__u16 sport,
-				__u16 dport)
+				     __u16 sport, __u16 dport)
 {
 	packet->transport = transport;
 	packet->source_port = sport;
@@ -109,14 +107,12 @@ sctp_packet_t *sctp_packet_init(sctp_packet_t *packet,
 }
 
 /* Free a packet.  */
-void sctp_packet_free(sctp_packet_t *packet)
+void sctp_packet_free(struct sctp_packet *packet)
 {
-	sctp_chunk_t *chunk;
+	struct sctp_chunk *chunk;
 
-        while (NULL != 
-	       (chunk = (sctp_chunk_t *)skb_dequeue(&packet->chunks))) {
+        while ((chunk = (struct sctp_chunk *)__skb_dequeue(&packet->chunks)))
 		sctp_free_chunk(chunk);
-	}
 
 	if (packet->malloced)
 		kfree(packet);
@@ -129,8 +125,8 @@ void sctp_packet_free(sctp_packet_t *packet)
  * as it can fit in the packet, but any more data that does not fit in this
  * packet can be sent only after receiving the COOKIE_ACK.
  */
-sctp_xmit_t sctp_packet_transmit_chunk(sctp_packet_t *packet,
-				       sctp_chunk_t *chunk)
+sctp_xmit_t sctp_packet_transmit_chunk(struct sctp_packet *packet,
+				       struct sctp_chunk *chunk)
 {
 	sctp_xmit_t retval;
 	int error = 0;
@@ -152,6 +148,7 @@ sctp_xmit_t sctp_packet_transmit_chunk(sctp_packet_t *packet,
 	case SCTP_XMIT_MUST_FRAG:
 	case SCTP_XMIT_RWND_FULL:
 	case SCTP_XMIT_OK:
+	case SCTP_XMIT_NAGLE_DELAY:
 		break;
 	};
 
@@ -161,7 +158,8 @@ sctp_xmit_t sctp_packet_transmit_chunk(sctp_packet_t *packet,
 /* Append a chunk to the offered packet reporting back any inability to do
  * so.
  */
-sctp_xmit_t sctp_packet_append_chunk(sctp_packet_t *packet, sctp_chunk_t *chunk)
+sctp_xmit_t sctp_packet_append_chunk(struct sctp_packet *packet, 
+				     struct sctp_chunk *chunk)
 {
 	sctp_xmit_t retval = SCTP_XMIT_OK;
 	__u16 chunk_len = WORD_ROUND(ntohs(chunk->chunk_hdr->length));
@@ -223,7 +221,7 @@ sctp_xmit_t sctp_packet_append_chunk(sctp_packet_t *packet, sctp_chunk_t *chunk)
 	}
 
 	/* It is OK to send this chunk.  */
-	skb_queue_tail(&packet->chunks, (struct sk_buff *)chunk);
+	__skb_queue_tail(&packet->chunks, (struct sk_buff *)chunk);
 	packet->size += chunk_len;
 finish:
 	return retval;
@@ -234,18 +232,18 @@ sctp_xmit_t sctp_packet_append_chunk(sctp_packet_t *packet, sctp_chunk_t *chunk)
  *
  * The return value is a normal kernel error return value.
  */
-int sctp_packet_transmit(sctp_packet_t *packet)
+int sctp_packet_transmit(struct sctp_packet *packet)
 {
 	struct sctp_transport *transport = packet->transport;
-	sctp_association_t *asoc = transport->asoc;
+	struct sctp_association *asoc = transport->asoc;
 	struct sctphdr *sh;
 	__u32 crc32;
 	struct sk_buff *nskb;
-	sctp_chunk_t *chunk;
+	struct sctp_chunk *chunk;
 	struct sock *sk;
 	int err = 0;
 	int padding;		/* How much padding do we need?  */
-	__u8 packet_has_data = 0;
+	__u8 has_data = 0;
 	struct dst_entry *dst;
 
 	/* Do NOT generate a chunkless packet... */
@@ -253,7 +251,7 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 		return err;
 
 	/* Set up convenience variables... */
-	chunk = (sctp_chunk_t *) (packet->chunks.next);
+	chunk = (struct sctp_chunk *) (packet->chunks.next);
 	sk = chunk->skb->sk;
 
 	/* Allocate the new skb.  */
@@ -291,8 +289,7 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 	 * [This whole comment explains WORD_ROUND() below.]
 	 */
 	SCTP_DEBUG_PRINTK("***sctp_transmit_packet***\n");
-	while (NULL != (chunk = (sctp_chunk_t *)
-			skb_dequeue(&packet->chunks))) {
+	while ((chunk = (struct sctp_chunk *)__skb_dequeue(&packet->chunks))) {
 		chunk->num_times_sent++;
 		chunk->sent_at = jiffies;
 		if (sctp_chunk_is_data(chunk)) {
@@ -309,7 +306,7 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 				chunk->rtt_in_progress = 1;
 				transport->rto_pending = 1;
 			}
-			packet_has_data = 1;
+			has_data = 1;
 		}
 		memcpy(skb_put(nskb, chunk->skb->len),
 		       chunk->skb->data, chunk->skb->len);
@@ -399,7 +396,7 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 		asoc->peer.last_sent_to = transport;
 	}
 
-	if (packet_has_data) {
+	if (has_data) {
 		struct timer_list *timer;
 		unsigned long timeout;
 
@@ -456,9 +453,9 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 /*
  * This private function resets the packet to a fresh state.
  */
-static void sctp_packet_reset(sctp_packet_t *packet)
+static void sctp_packet_reset(struct sctp_packet *packet)
 {
-	sctp_chunk_t *chunk = NULL;
+	struct sctp_chunk *chunk = NULL;
 
 	packet->size = SCTP_IP_OVERHEAD;
 
@@ -473,13 +470,15 @@ static void sctp_packet_reset(sctp_packet_t *packet)
 }
 
 /* This private function handles the specifics of appending DATA chunks.  */
-static sctp_xmit_t sctp_packet_append_data(sctp_packet_t *packet, 
-					   sctp_chunk_t *chunk)
+static sctp_xmit_t sctp_packet_append_data(struct sctp_packet *packet, 
+					   struct sctp_chunk *chunk)
 {
 	sctp_xmit_t retval = SCTP_XMIT_OK;
 	size_t datasize, rwnd, inflight;
 	struct sctp_transport *transport = packet->transport;
 	__u32 max_burst_bytes;
+	struct sctp_opt *sp = sctp_sk(transport->asoc->base.sk);
+	struct sctp_outq *q = &transport->asoc->outqueue;
 
 	/* RFC 2960 6.1  Transmission of DATA Chunks
 	 *
@@ -543,11 +542,34 @@ static sctp_xmit_t sctp_packet_append_data(sctp_packet_t *packet,
 	 *    When a Fast Retransmit is being performed the sender SHOULD
 	 *    ignore the value of cwnd and SHOULD NOT delay retransmission.
 	 */
-	if (!chunk->fast_retransmit) {
+	if (!chunk->fast_retransmit)
 		if (transport->flight_size >= transport->cwnd) {
 			retval = SCTP_XMIT_RWND_FULL;
 			goto finish;
 		}
+
+	/* Nagle's algorithm to solve small-packet problem:
+	 * inhibit the sending of new chunks when new outgoing data arrives 
+	 * if any proeviously transmitted data on the connection remains
+	 * unacknowledged. Unless the connection was previously idle. Check 
+	 * whether the connection is idle. No outstanding means idle, flush
+	 * it. If outstanding bytes are less than half cwnd, the connection
+	 * is not in the state of congestion, so also flush it. 
+	 */
+	if (!sp->nodelay && q->outstanding_bytes >= transport->cwnd >> 1) {
+		/* Check whether this chunk and all the rest of pending 
+		 * data will fit or whether we'll choose to delay in
+		 * hopes of bundling a full sized packet. 
+		 */
+		if ((datasize + q->out_qlen) < transport->asoc->frag_point) {
+						    
+			/* If the the chunk should be delay
+			 * for future sending, we could not
+			 * append it.
+			 */
+			retval = SCTP_XMIT_NAGLE_DELAY;
+			goto finish;		
+		}
 	}
 
 	/* Keep track of how many bytes are in flight over this transport. */

net/sctp/outqueue.c

 /* SCTP kernel reference Implementation
  * Copyright (c) 1999-2000 Cisco, Inc.
  * Copyright (c) 1999-2001 Motorola, Inc.
- * Copyright (c) 2001 Intel Corp.
+ * Copyright (c) 2001-2003 Intel Corp.
  * Copyright (c) 2001-2003 International Business Machines Corp.
  *
  * This file is part of the SCTP kernel reference Implementation
@@ -62,6 +62,33 @@ static void sctp_check_transmitted(struct sctp_outq *q,
 				   sctp_sackhdr_t *sack,
 				   __u32 highest_new_tsn);
 
+/* Add data to the front of the queue. */
+static inline void sctp_outq_head_data(struct sctp_outq *q,
+					struct sctp_chunk *ch)
+{
+	__skb_queue_head(&q->out, (struct sk_buff *)ch);
+	q->out_qlen += ch->skb->len;
+	return;
+}
+
+/* Take data from the front of the queue. */
+static inline struct sctp_chunk *sctp_outq_dequeue_data(struct sctp_outq *q)
+{
+	struct sctp_chunk *ch;
+	ch = (struct sctp_chunk *)__skb_dequeue(&q->out);
+	if (ch)
+		q->out_qlen -= ch->skb->len;
+	return ch;
+}
+/* Add data chunk to the end of the queue. */
+static inline void sctp_outq_tail_data(struct sctp_outq *q,
+				       struct sctp_chunk *ch)
+{
+	__skb_queue_tail(&q->out, (struct sk_buff *)ch);
+	q->out_qlen += ch->skb->len;
+	return;
+}
+
 /* Generate a new outqueue.  */
 struct sctp_outq *sctp_outq_new(sctp_association_t *asoc)
 {
@@ -97,6 +124,7 @@ void sctp_outq_init(sctp_association_t *asoc, struct sctp_outq *q)
 	q->empty = 1;
 
 	q->malloced = 0;
+	q->out_qlen = 0;
 }
 
 /* Free the outqueue structure and any related pending chunks.
@@ -133,7 +161,7 @@ void sctp_outq_teardown(struct sctp_outq *q)
 	}
 
 	/* Throw away any leftover data chunks. */
-	while ((chunk = (sctp_chunk_t *) skb_dequeue(&q->out)))
+	while ((chunk = sctp_outq_dequeue_data(q)))
 		sctp_free_chunk(chunk);
 
 	/* Throw away any leftover control chunks. */
@@ -192,7 +220,7 @@ int sctp_outq_tail(struct sctp_outq *q, sctp_chunk_t *chunk)
 			  sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type))
 			  : "Illegal Chunk");
 
-			skb_queue_tail(&q->out, (struct sk_buff *) chunk);
+			sctp_outq_tail_data(q, chunk);
 			if (chunk->chunk_hdr->flags & SCTP_DATA_UNORDERED)
 				SCTP_INC_STATS(SctpOutUnorderChunks);
 			else
@@ -201,7 +229,7 @@ int sctp_outq_tail(struct sctp_outq *q, sctp_chunk_t *chunk)
 			break;
 		};
 	} else {
-		skb_queue_tail(&q->control, (struct sk_buff *) chunk);
+		__skb_queue_tail(&q->control, (struct sk_buff *) chunk);
 		SCTP_INC_STATS(SctpOutCtrlChunks);
 	}
 
@@ -351,7 +379,7 @@ void sctp_retransmit(struct sctp_outq *q, struct sctp_transport *transport,
  *
  * The return value is a normal kernel error return value.
  */
-static int sctp_outq_flush_rtx(struct sctp_outq *q, sctp_packet_t *pkt,
+static int sctp_outq_flush_rtx(struct sctp_outq *q, struct sctp_packet *pkt,
 			       int rtx_timeout, int *start_timer)
 {
 	struct list_head *lqueue;
@@ -385,17 +413,6 @@ static int sctp_outq_flush_rtx(struct sctp_outq *q, sctp_packet_t *pkt,
 
 	while (lchunk) {
 		chunk = list_entry(lchunk, sctp_chunk_t, transmitted_list);
-#if 0
-		/* If a chunk has been tried for more than SCTP_DEF_MAX_SEND
-		 * times, discard it, and check the empty flag of the outqueue.
-		 *
-		 * --xguo
-		 */
-		if (chunk->snd_count > SCTP_DEF_MAX_SEND) {
-			sctp_free_chunk(chunk);
-			continue;
-		}
-#endif
 
 		/* Make sure that Gap Acked TSNs are not retransmitted.  A
 		 * simple approach is just to move such TSNs out of the
@@ -462,7 +479,7 @@ static int sctp_outq_flush_rtx(struct sctp_outq *q, sctp_packet_t *pkt,
  * chunk that is currently in the process of fragmentation.
  */
 void sctp_xmit_frag(struct sctp_outq *q, struct sk_buff *pos,
-		    sctp_packet_t *packet, sctp_chunk_t *frag, __u32 tsn)
+		    struct sctp_packet *packet, sctp_chunk_t *frag, __u32 tsn)
 {
 	struct sctp_transport *transport = packet->transport;
 	struct sk_buff_head *queue = &q->out;
@@ -480,11 +497,11 @@ void sctp_xmit_frag(struct sctp_outq *q, struct sk_buff *pos,
 		SCTP_DEBUG_PRINTK("sctp_xmit_frag: q not empty. "
 				  "adding 0x%x to outqueue\n",
 				  ntohl(frag->subh.data_hdr->tsn));
-		if (pos) {
-			skb_insert(pos, (struct sk_buff *) frag);
-		} else {
-			skb_queue_tail(queue, (struct sk_buff *) frag);
-		}
+		if (pos)
+			__skb_insert((struct sk_buff *)frag, pos->prev,
+				     pos, pos->list);
+		else
+			__skb_queue_tail(queue, (struct sk_buff *) frag);
 		return;
 	}
 
@@ -496,11 +513,11 @@ void sctp_xmit_frag(struct sctp_outq *q, struct sk_buff *pos,
 		SCTP_DEBUG_PRINTK("sctp_xmit_frag: rwnd full. "
 				  "adding 0x%x to outqueue\n",
 				  ntohl(frag->subh.data_hdr->tsn));
-		if (pos) {
-			skb_insert(pos, (struct sk_buff *) frag);
-		} else {
-			skb_queue_tail(queue, (struct sk_buff *) frag);
-		}
+		if (pos)
+			__skb_insert((struct sk_buff *)frag, pos->prev,
+				     pos, pos->list);
+		else
+			__skb_queue_tail(queue, (struct sk_buff *)frag);
 		break;
 
 	case SCTP_XMIT_OK:
@@ -512,11 +529,11 @@ void sctp_xmit_frag(struct sctp_outq *q, struct sk_buff *pos,
 			SCTP_DEBUG_PRINTK("sctp_xmit_frag: force output "
 					  "failed. adding 0x%x to outqueue\n",
 					  ntohl(frag->subh.data_hdr->tsn));
-			if (pos) {
-				skb_insert(pos, (struct sk_buff *) frag);
-			} else {
-				skb_queue_tail(queue, (struct sk_buff *) frag);
-			}
+			if (pos)
+				__skb_insert((struct sk_buff *)frag, pos->prev,
+					     pos, pos->list);
+			else
+				__skb_queue_tail(queue,(struct sk_buff *)frag);
 		} else {
 			SCTP_DEBUG_PRINTK("sctp_xmit_frag: force output "
 					  "success. 0x%x sent\n",
@@ -537,7 +554,7 @@ void sctp_xmit_frag(struct sctp_outq *q, struct sk_buff *pos,
  * The argument 'frag' point to the first fragment and it holds the list
  * of all the other fragments in the 'frag_list' field.
  */
-void sctp_xmit_fragmented_chunks(struct sctp_outq *q, sctp_packet_t *packet,
+void sctp_xmit_fragmented_chunks(struct sctp_outq *q, struct sctp_packet *pkt,
 				 sctp_chunk_t *frag)
 {
 	sctp_association_t *asoc = frag->asoc;
@@ -557,13 +574,13 @@ void sctp_xmit_fragmented_chunks(struct sctp_outq *q, sctp_packet_t *packet,
 
 	pos = skb_peek(&q->out);
 	/* Transmit the first fragment. */
-	sctp_xmit_frag(q, pos, packet, frag, tsn++);
+	sctp_xmit_frag(q, pos, pkt, frag, tsn++);
 
 	/* Transmit the rest of fragments. */
 	frag_list = &frag->frag_list;
 	list_for_each(lfrag, frag_list) {
 		frag = list_entry(lfrag, sctp_chunk_t, frag_list);
-		sctp_xmit_frag(q, pos, packet, frag, tsn++);
+		sctp_xmit_frag(q, pos, pkt, frag, tsn++);
 	}
 }
 
@@ -672,15 +689,14 @@ sctp_chunk_t *sctp_fragment_chunk(sctp_chunk_t *chunk,
  *
  * Description: Send everything in q which we legally can, subject to
  * congestion limitations.
- *
- * Note: This function can be called from multiple contexts so appropriate
+ * * Note: This function can be called from multiple contexts so appropriate
  * locking concerns must be made.  Today we use the sock lock to protect
  * this function.
  */
 int sctp_outq_flush(struct sctp_outq *q, int rtx_timeout)
 {
-	sctp_packet_t *packet;
-	sctp_packet_t singleton;
+	struct sctp_packet *packet;
+	struct sctp_packet singleton;
 	sctp_association_t *asoc = q->asoc;
 	int ecn_capable = asoc->peer.ecn_capable;
 	__u16 sport = asoc->base.bind_addr.port;
@@ -852,7 +868,8 @@ int sctp_outq_flush(struct sctp_outq *q, int rtx_timeout)
 		/* Finally, transmit new packets.  */
 		start_timer = 0;
 		queue = &q->out;
-		while (NULL != (chunk = (sctp_chunk_t *) skb_dequeue(queue))) {
+
+		while (NULL != (chunk = sctp_outq_dequeue_data(q))) {
 			/* RFC 2960 6.5 Every DATA chunk MUST carry a valid
 			 * stream identifier.
 			 */
@@ -925,6 +942,7 @@ int sctp_outq_flush(struct sctp_outq *q, int rtx_timeout)
 			switch (status) {
 			case SCTP_XMIT_PMTU_FULL:
 			case SCTP_XMIT_RWND_FULL:
+			case SCTP_XMIT_NAGLE_DELAY:
 				/* We could not append this chunk, so put
 				 * the chunk back on the output queue.
 				 */
@@ -932,7 +950,7 @@ int sctp_outq_flush(struct sctp_outq *q, int rtx_timeout)
 					"not transmit TSN: 0x%x, status: %d\n",
 					ntohl(chunk->subh.data_hdr->tsn),
 					status);
-				skb_queue_head(queue, (struct sk_buff *)chunk);
+				sctp_outq_head_data(q, chunk);
 				goto sctp_flush_out;
 				break;
 
@@ -994,6 +1012,7 @@ int sctp_outq_flush(struct sctp_outq *q, int rtx_timeout)
 	}
 
 sctp_flush_out:
+
 	/* Before returning, examine all the transports touched in
 	 * this call.  Right now, we bluntly force clear all the
 	 * transports.  Things might change after we implement Nagle.
@@ -1163,11 +1182,10 @@ int sctp_outq_sack(struct sctp_outq *q, sctp_sackhdr_t *sack)
 	sack_a_rwnd = ntohl(sack->a_rwnd);
 	outstanding = q->outstanding_bytes;
 
-	if (outstanding < sack_a_rwnd) {
+	if (outstanding < sack_a_rwnd)
 		sack_a_rwnd -= outstanding;
-	} else {
+	else
 		sack_a_rwnd = 0;
-	}
 
 	asoc->peer.rwnd = sack_a_rwnd;
 

net/sctp/protocol.c

@@ -557,10 +557,13 @@ static int sctp_inetaddr_event(struct notifier_block *this, unsigned long event,
  */
 int sctp_ctl_sock_init(void)
 {
-	int err = 0;
-	int family = PF_INET;
+	int err;
+	sa_family_t family;
 
-	SCTP_V6(family = PF_INET6;)
+	if (sctp_get_pf_specific(PF_INET6))
+		family = PF_INET6;
+	else 
+		family = PF_INET;
 
 	err = sock_create(family, SOCK_SEQPACKET, IPPROTO_SCTP,
 			  &sctp_ctl_socket);

net/sctp/sm_sideeffect.c

@@ -256,7 +256,7 @@ int sctp_cmd_interpreter(sctp_event_t event_type, sctp_subtype_t subtype,
 	sctp_cmd_t *cmd;
 	sctp_chunk_t *new_obj;
 	sctp_chunk_t *chunk = NULL;
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 	struct list_head *pos;
 	struct timer_list *timer;
 	unsigned long timeout;

net/sctp/sm_statefuns.c

@@ -189,7 +189,7 @@ sctp_disposition_t sctp_sf_do_5_1B_init(const sctp_endpoint_t *ep,
 	sctp_chunk_t *repl;
 	sctp_association_t *new_asoc;
 	sctp_chunk_t *err_chunk;
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 	sctp_unrecognized_param_t *unk_param;
 	int len;
 
@@ -354,10 +354,9 @@ sctp_disposition_t sctp_sf_do_5_1C_ack(const sctp_endpoint_t *ep,
 	sctp_init_chunk_t *initchunk;
 	__u32 init_tag;
 	sctp_chunk_t *err_chunk;
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 	sctp_disposition_t ret;
 
-
 	/* 6.10 Bundling
 	 * An endpoint MUST NOT bundle INIT, INIT ACK or
 	 * SHUTDOWN COMPLETE with any other chunks.
@@ -912,14 +911,14 @@ static int sctp_sf_send_restart_abort(union sctp_addr *ssa,
 				      sctp_cmd_seq_t *commands)
 {
 	int len;
-	sctp_packet_t *pkt;
+	struct sctp_packet *pkt;
 	sctp_addr_param_t *addrparm;
 	sctp_errhdr_t *errhdr;
 	sctp_endpoint_t *ep;
 	char buffer[sizeof(sctp_errhdr_t) + sizeof(sctp_addr_param_t)];
 
-	/* Build the error on the stack.   We are way to malloc
-	 * malloc crazy throughout the code today.
+	/* Build the error on the stack.   We are way to malloc crazy
+	 * throughout the code today.
 	 */
 	errhdr = (sctp_errhdr_t *)buffer;
 	addrparm = (sctp_addr_param_t *)errhdr->variable;
@@ -1105,11 +1104,10 @@ static sctp_disposition_t sctp_sf_do_unexpected_init(
 	sctp_chunk_t *repl;
 	sctp_association_t *new_asoc;
 	sctp_chunk_t *err_chunk;
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 	sctp_unrecognized_param_t *unk_param;
 	int len;
 
-
 	/* 6.10 Bundling
 	 * An endpoint MUST NOT bundle INIT, INIT ACK or
 	 * SHUTDOWN COMPLETE with any other chunks.
@@ -2751,7 +2749,7 @@ sctp_disposition_t sctp_sf_tabort_8_4_8(const sctp_endpoint_t *ep,
 					void *arg,
 					sctp_cmd_seq_t *commands)
 {
-	sctp_packet_t *packet = NULL;
+	struct sctp_packet *packet = NULL;
 	sctp_chunk_t *chunk = arg;
 	sctp_chunk_t *abort;
 
@@ -2953,7 +2951,7 @@ sctp_disposition_t sctp_sf_shut_8_4_5(const sctp_endpoint_t *ep,
 				      void *arg,
 				      sctp_cmd_seq_t *commands)
 {
-	sctp_packet_t *packet = NULL;
+	struct sctp_packet *packet = NULL;
 	sctp_chunk_t *chunk = arg;
 	sctp_chunk_t *shut;
 
@@ -4377,13 +4375,13 @@ sctp_sackhdr_t *sctp_sm_pull_sack(sctp_chunk_t *chunk)
 /* Create an ABORT packet to be sent as a response, with the specified
  * error causes.
  */
-sctp_packet_t *sctp_abort_pkt_new(const sctp_endpoint_t *ep,
+struct sctp_packet *sctp_abort_pkt_new(const sctp_endpoint_t *ep,
 				  const sctp_association_t *asoc,
 				  sctp_chunk_t *chunk,
 				  const void *payload,
 				  size_t paylen)
 {
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 	sctp_chunk_t *abort;
 
 	packet = sctp_ootb_pkt_new(asoc, chunk);
@@ -4413,10 +4411,10 @@ sctp_packet_t *sctp_abort_pkt_new(const sctp_endpoint_t *ep,
 }
 
 /* Allocate a packet for responding in the OOTB conditions.  */
-sctp_packet_t *sctp_ootb_pkt_new(const sctp_association_t *asoc,
+struct sctp_packet *sctp_ootb_pkt_new(const sctp_association_t *asoc,
 				 const sctp_chunk_t *chunk)
 {
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 	struct sctp_transport *transport;
 	__u16 sport;
 	__u16 dport;
@@ -4449,7 +4447,7 @@ sctp_packet_t *sctp_ootb_pkt_new(const sctp_association_t *asoc,
 		goto nomem;
 
 	/* Allocate a new packet for sending the response. */
-	packet = t_new(sctp_packet_t, GFP_ATOMIC);
+	packet = t_new(struct sctp_packet, GFP_ATOMIC);
 	if (!packet)
 		goto nomem_packet;
 
@@ -4471,7 +4469,7 @@ sctp_packet_t *sctp_ootb_pkt_new(const sctp_association_t *asoc,
 }
 
 /* Free the packet allocated earlier for responding in the OOTB condition.  */
-void sctp_ootb_pkt_free(sctp_packet_t *packet)
+void sctp_ootb_pkt_free(struct sctp_packet *packet)
 {
 	sctp_transport_free(packet->transport);
 	sctp_packet_free(packet);
@@ -4484,7 +4482,7 @@ void sctp_send_stale_cookie_err(const sctp_endpoint_t *ep,
 				sctp_cmd_seq_t *commands,
 				sctp_chunk_t *err_chunk)
 {
-	sctp_packet_t *packet;
+	struct sctp_packet *packet;
 
 	if (err_chunk) {
 		packet = sctp_ootb_pkt_new(asoc, chunk);

net/sctp/socket.c

@@ -1231,7 +1231,7 @@ SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr
 	return err;
 }
 
-static inline int sctp_setsockopt_disable_fragments(struct sock *sk,
+static int sctp_setsockopt_disable_fragments(struct sock *sk,
 						    char *optval, int optlen)
 {
 	int val;
@@ -1247,7 +1247,7 @@ static inline int sctp_setsockopt_disable_fragments(struct sock *sk,
 	return 0;
 }
 
-static inline int sctp_setsockopt_events(struct sock *sk, char *optval,
+static int sctp_setsockopt_events(struct sock *sk, char *optval,
 					int optlen)
 {
 	if (optlen != sizeof(struct sctp_event_subscribe))
@@ -1257,7 +1257,7 @@ static inline int sctp_setsockopt_events(struct sock *sk, char *optval,
 	return 0;
 }
 
-static inline int sctp_setsockopt_autoclose(struct sock *sk, char *optval,
+static int sctp_setsockopt_autoclose(struct sock *sk, char *optval,
 					    int optlen)
 {
 	struct sctp_opt *sp = sctp_sk(sk);
@@ -1274,7 +1274,7 @@ static inline int sctp_setsockopt_autoclose(struct sock *sk, char *optval,
 	return 0;
 }
 
-static inline int sctp_setsockopt_peer_addr_params(struct sock *sk,
+static int sctp_setsockopt_peer_addr_params(struct sock *sk,
 						   char *optval, int optlen)
 {
 	struct sctp_paddrparams params;
@@ -1333,7 +1333,7 @@ static inline int sctp_setsockopt_peer_addr_params(struct sock *sk,
 	return 0;
 }
 
-static inline int sctp_setsockopt_initmsg(struct sock *sk, char *optval,
+static int sctp_setsockopt_initmsg(struct sock *sk, char *optval,
 					  int optlen)
 {
 	if (optlen != sizeof(struct sctp_initmsg))
@@ -1358,7 +1358,7 @@ static inline int sctp_setsockopt_initmsg(struct sock *sk, char *optval,
  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
  *   to this call if the caller is using the UDP model.
  */
-static inline int sctp_setsockopt_default_send_param(struct sock *sk,
+static int sctp_setsockopt_default_send_param(struct sock *sk,
 						char *optval, int optlen)
 {
 	struct sctp_sndrcvinfo info;
@@ -1416,6 +1416,31 @@ static int sctp_setsockopt_peer_prim(struct sock *sk, char *optval, int optlen)
 	return 0;
 }
 
+/*
+ *
+ * 7.1.5 SCTP_NODELAY
+ *
+ * Turn on/off any Nagle-like algorithm.  This means that packets are
+ * generally sent as soon as possible and no unnecessary delays are
+ * introduced, at the cost of more packets in the network.  Expects an
+ *  integer boolean flag.
+ */
+static int sctp_setsockopt_nodelay(struct sock *sk, char *optval,
+					int optlen)
+{
+	__u8 val;
+
+	if (optlen < sizeof(__u8))
+		return -EINVAL;
+
+	if (get_user(val, (__u8 *)optval))
+		return -EFAULT;
+
+	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
+
+	return 0;
+}
+
 /* API 6.2 setsockopt(), getsockopt()
  *
  * Applications use setsockopt() and getsockopt() to set or retrieve
@@ -1515,6 +1540,10 @@ SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
 		retval = sctp_setsockopt_peer_prim(sk, optval, optlen);
 		break;
 
+	case SCTP_NODELAY:
+		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
+		break;
+
 	default:
 		retval = -ENOPROTOOPT;
 		break;
@@ -1786,7 +1815,7 @@ SCTP_STATIC int sctp_init_sock(struct sock *sk)
 	sp->disable_fragments = 0;
 
 	/* Turn on/off any Nagle-like algorithm.  */
-	sp->nodelay           = 0;
+	sp->nodelay           = 1;
 
 	/* Auto-close idle associations after the configured
 	 * number of seconds.  A value of 0 disables this
@@ -1905,7 +1934,7 @@ static int sctp_getsockopt_sctp_status(struct sock *sk, int len, char *optval,
 	return (retval);
 }
 
-static inline int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
+static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
 						    char *optval, int *optlen)
 {
 	int val;
@@ -1922,7 +1951,7 @@ static inline int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
 	return 0;
 }
 
-static inline int sctp_getsockopt_set_events(struct sock *sk, int len, char *optval, int *optlen)
+static int sctp_getsockopt_set_events(struct sock *sk, int len, char *optval, int *optlen)
 {
 	if (len != sizeof(struct sctp_event_subscribe))
 		return -EINVAL;
@@ -1931,7 +1960,7 @@ static inline int sctp_getsockopt_set_events(struct sock *sk, int len, char *opt
 	return 0;
 }
 
-static inline int sctp_getsockopt_autoclose(struct sock *sk, int len, char *optval, int *optlen)
+static int sctp_getsockopt_autoclose(struct sock *sk, int len, char *optval, int *optlen)
 {
 	/* Applicable to UDP-style socket only */
 	if (SCTP_SOCKET_TCP == sctp_sk(sk)->type)
@@ -1975,7 +2004,7 @@ SCTP_STATIC int sctp_do_peeloff(sctp_association_t *assoc, struct socket **newso
 	return err;
 }
 
-static inline int sctp_getsockopt_peeloff(struct sock *sk, int len, char *optval, int *optlen)
+static int sctp_getsockopt_peeloff(struct sock *sk, int len, char *optval, int *optlen)
 {
 	sctp_peeloff_arg_t peeloff;
 	struct socket *newsock;
@@ -2018,7 +2047,7 @@ static inline int sctp_getsockopt_peeloff(struct sock *sk, int len, char *optval
 	return retval;
 }
 
-static inline int sctp_getsockopt_peer_addr_params(struct sock *sk, int len, 
+static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len, 
 						char *optval, int *optlen)
 {
 	struct sctp_paddrparams params;
@@ -2062,7 +2091,7 @@ static inline int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
 	return 0;
 }
 
-static inline int sctp_getsockopt_initmsg(struct sock *sk, int len, char *optval, int *optlen)
+static int sctp_getsockopt_initmsg(struct sock *sk, int len, char *optval, int *optlen)
 {
 	if (len != sizeof(struct sctp_initmsg))
 		return -EINVAL;
@@ -2141,7 +2170,7 @@ static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
 	return 0;
 }
 
-static inline int sctp_getsockopt_local_addrs_num(struct sock *sk, int len,
+static int sctp_getsockopt_local_addrs_num(struct sock *sk, int len,
 						char *optval, int *optlen)
 {
 	sctp_assoc_t id;
@@ -2180,7 +2209,7 @@ static inline int sctp_getsockopt_local_addrs_num(struct sock *sk, int len,
 	return 0;
 }
 
-static inline int sctp_getsockopt_local_addrs(struct sock *sk, int len,
+static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
 					char *optval, int *optlen)
 {
 	sctp_bind_addr_t *bp;
@@ -2282,7 +2311,7 @@ static int sctp_getsockopt_peer_prim(struct sock *sk, int len,
  *
  *   For getsockopt, it get the default sctp_sndrcvinfo structure.
  */
-static inline int sctp_getsockopt_default_send_param(struct sock *sk,
+static int sctp_getsockopt_default_send_param(struct sock *sk,
 					int len, char *optval, int *optlen)
 {
 	struct sctp_sndrcvinfo info;
@@ -2309,6 +2338,33 @@ static inline int sctp_getsockopt_default_send_param(struct sock *sk,
 	return 0;
 }
 
+/*
+ *
+ * 7.1.5 SCTP_NODELAY
+ *
+ * Turn on/off any Nagle-like algorithm.  This means that packets are
+ * generally sent as soon as possible and no unnecessary delays are
+ * introduced, at the cost of more packets in the network.  Expects an
+ * integer boolean flag.
+ */
+
+static int sctp_getsockopt_nodelay(struct sock *sk, int len, 
+					char *optval, int *optlen)
+{
+	__u8 val;
+
+	if (len < sizeof(__u8))
+		return -EINVAL;
+
+	len = sizeof(__u8);
+	val = (sctp_sk(sk)->nodelay == 1);
+	if (put_user(len, optlen))
+		return -EFAULT;
+	if (copy_to_user(optval, &val, len))
+		return -EFAULT;
+	return 0;
+}
+
 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
 				char *optval, int *optlen)
 {
@@ -2382,6 +2438,9 @@ SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
 	case SCTP_SET_PEER_PRIMARY_ADDR:
 		retval = sctp_getsockopt_peer_prim(sk, len, optval, optlen);
 		break;
+	case SCTP_NODELAY:
+		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
+		break;
 	default:
 		retval = -ENOPROTOOPT;
 		break;
@@ -2689,7 +2748,6 @@ int sctp_inet_listen(struct socket *sock, int backlog)
 	case SOCK_SEQPACKET:
 		err = sctp_seqpacket_listen(sk, backlog);
 		break;
-
 	case SOCK_STREAM:
 		err = sctp_stream_listen(sk, backlog);
 		break;
@@ -3077,7 +3135,8 @@ static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags, int no
 }
 
 /* Verify that this is a valid address. */
-static int sctp_verify_addr(struct sock *sk, union sctp_addr *addr, int len)
+static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr, 
+				   int len)
 {
 	struct sctp_af *af;