Merge nuts.ninka.net:/home/davem/src/BK/sctp-2.5

into nuts.ninka.net:/home/davem/src/BK/net-2.5

include/net/sctp/sctp.h

@@ -83,6 +83,7 @@
 #include <asm/uaccess.h>
 #include <asm/page.h>
 #include <net/sock.h>
+#include <net/snmp.h>
 #include <net/sctp/structs.h>
 #include <net/sctp/constants.h>
 #include <net/sctp/sm.h>
@@ -201,6 +202,11 @@ extern void sctp_hash_digest(const char *secret, const int secret_len,
 #define SCTP_SOCK_SLEEP_PRE(sk) SOCK_SLEEP_PRE(sk)
 #define SCTP_SOCK_SLEEP_POST(sk) SOCK_SLEEP_POST(sk)
 
+/* SCTP SNMP MIB stats handlers */
+extern struct sctp_mib sctp_statistics[NR_CPUS * 2];
+#define SCTP_INC_STATS(field)		SNMP_INC_STATS(sctp_statistics, field)
+#define SCTP_INC_STATS_BH(field)	SNMP_INC_STATS_BH(sctp_statistics, field)
+#define SCTP_INC_STATS_USER(field)	SNMP_INC_STATS_USER(sctp_statistics, field)
 
 /* Determine if this is a valid kernel address.  */
 static inline int sctp_is_valid_kaddr(unsigned long addr)

include/net/sctp/sm.h

@@ -44,6 +44,7 @@
  *    Dajiang Zhang <dajiang.zhang@nokia.com>
  *    Sridhar Samudrala <sri@us.ibm.com>
  *    Daisy Chang <daisyc@us.ibm.com>
+ *    Ardelle Fan <ardelle.fan@intel.com>
  *
  * Any bugs reported given to us we will try to fix... any fixes shared will
  * be incorporated into the next SCTP release.
@@ -253,6 +254,9 @@ sctp_chunk_t *sctp_make_abort(const sctp_association_t *,
 sctp_chunk_t *sctp_make_abort_no_data(const sctp_association_t *,
 				      const sctp_chunk_t *,
 				      __u32 tsn);
+sctp_chunk_t *sctp_make_abort_user(const sctp_association_t *,
+				   const sctp_chunk_t *,
+				   const struct msghdr *);
 sctp_chunk_t *sctp_make_heartbeat(const sctp_association_t *,
 				  const sctp_transport_t *,
 				  const void *payload,

include/net/sctp/structs.h

@@ -246,7 +246,10 @@ typedef struct sctp_func {
 					 int optname,
 					 char *optval,
 					 int *optlen);
-	int		(*get_dst_mtu)	(const sockaddr_storage_t *address);
+	struct dst_entry *(*get_dst)	(sockaddr_storage_t *daddr,
+					 sockaddr_storage_t *saddr);
+	int		(*cmp_saddr)	(struct dst_entry *dst,
+					 sockaddr_storage_t *saddr);
 	__u16		net_header_len;	
 	int		sockaddr_len;
 	sa_family_t	sa_family;
@@ -476,6 +479,8 @@ struct SCTP_chunk {
 
 	/* What is the origin IP address for this chunk?  */
 	sockaddr_storage_t source;
+	/* Destination address for this chunk. */
+	sockaddr_storage_t dest;
 
 	/* For an inbound chunk, this tells us where it came from.
 	 * For an outbound chunk, it tells us where we'd like it to
@@ -492,7 +497,7 @@ void  *sctp_addto_chunk(sctp_chunk_t *chunk, int len, const void *data);
 int sctp_user_addto_chunk(sctp_chunk_t *chunk, int len, struct iovec *data);
 sctp_chunk_t *sctp_chunkify(struct sk_buff *, const sctp_association_t *,
 			    struct sock *);
-void sctp_init_source(sctp_chunk_t *chunk);
+void sctp_init_addrs(sctp_chunk_t *chunk);
 const sockaddr_storage_t *sctp_source(const sctp_chunk_t *chunk);
 
 /* This is a structure for holding either an IPv6 or an IPv4 address.  */
@@ -655,6 +660,9 @@ struct SCTP_transport {
 	/* PMTU       : The current known path MTU.  */
 	__u32 pmtu;
 
+	/* Destination */
+	struct dst_entry *dst;
+
 	/* When was the last time(in jiffies) that a data packet was sent on
 	 * this transport?  This is used to adjust the cwnd when the transport
 	 * becomes inactive.
@@ -735,6 +743,7 @@ extern sctp_transport_t *sctp_transport_new(const sockaddr_storage_t *, int);
 extern sctp_transport_t *sctp_transport_init(sctp_transport_t *,
 					     const sockaddr_storage_t *, int);
 extern void sctp_transport_set_owner(sctp_transport_t *, sctp_association_t *);
+extern void sctp_transport_route(sctp_transport_t *, sockaddr_storage_t *);
 extern void sctp_transport_free(sctp_transport_t *);
 extern void sctp_transport_destroy(sctp_transport_t *);
 extern void sctp_transport_reset_timers(sctp_transport_t *);

include/net/snmp.h

@@ -189,6 +189,35 @@ struct udp_mib
 	unsigned long   __pad[0];
 } ____cacheline_aligned; 
 
+/* draft-ietf-sigtran-sctp-mib-07.txt */
+struct sctp_mib
+{
+	unsigned long   SctpCurrEstab;
+	unsigned long   SctpActiveEstabs;
+	unsigned long   SctpPassiveEstabs;
+	unsigned long   SctpAborteds;
+	unsigned long   SctpShutdowns;
+	unsigned long   SctpOutOfBlues;
+	unsigned long   SctpChecksumErrors;
+	unsigned long   SctpOutCtrlChunks;
+	unsigned long   SctpOutOrderChunks;
+	unsigned long   SctpOutUnorderChunks;
+	unsigned long   SctpInCtrlChunks;
+	unsigned long   SctpInOrderChunks;
+	unsigned long   SctpInUnorderChunks;
+	unsigned long   SctpFragUsrMsgs;
+	unsigned long   SctpReasmUsrMsgs;
+	unsigned long   SctpOutSCTPPacks;
+	unsigned long   SctpInSCTPPacks;
+	unsigned long   SctpRtoAlgorithm;
+	unsigned long   SctpRtoMin;
+	unsigned long   SctpRtoMax;
+	unsigned long   SctpRtoInitial;
+	unsigned long   SctpValCookieLife;
+	unsigned long   SctpMaxInitRetr;
+	unsigned long   __pad[0];
+} ____cacheline_aligned;
+
 struct linux_mib 
 {
 	unsigned long	SyncookiesSent;

net/sctp/associola.c

@@ -208,7 +208,7 @@ sctp_association_t *sctp_association_init(sctp_association_t *asoc,
 
 	asoc->ctsn_ack_point = asoc->next_tsn - 1;
 	asoc->highest_sacked = asoc->ctsn_ack_point;
-
+	asoc->last_cwr_tsn = asoc->ctsn_ack_point;
 	asoc->unack_data = 0;
 
 	SCTP_DEBUG_PRINTK("myctsnap for %s INIT as 0x%x.\n",
@@ -404,6 +404,9 @@ sctp_transport_t *sctp_assoc_add_peer(sctp_association_t *asoc,
 
 	sctp_transport_set_owner(peer, asoc);
 
+	/* Cache a route for the transport. */
+	sctp_transport_route(peer, NULL);
+
 	/* If this is the first transport addr on this association,
 	 * initialize the association PMTU to the peer's PMTU.
 	 * If not and the current association PMTU is higher than the new

net/sctp/input.c

@@ -41,6 +41,7 @@
  *    Jon Grimm <jgrimm@us.ibm.com>
  *    Hui Huang <hui.huang@nokia.com>
  *    Daisy Chang <daisyc@us.ibm.com>
+ *    Sridhar Samudrala <sri@us.ibm.com>
  *
  * Any bugs reported given to us we will try to fix... any fixes shared will
  * be incorporated into the next SCTP release.
@@ -217,8 +218,8 @@ int sctp_rcv(struct sk_buff *skb)
 	/* Remember the SCTP header. */
 	chunk->sctp_hdr = sh;
 
-	/* Set the source address.  */
-	sctp_init_source(chunk);
+	/* Set the source and destination addresses of the incoming chunk.  */
+	sctp_init_addrs(chunk);
 
 	/* Remember where we came from.  */
 	chunk->transport = transport;

net/sctp/ipv6.c

@@ -158,35 +158,60 @@ static inline int sctp_v6_xmit(struct sk_buff *skb)
 }
 #endif /* TEST_FRAME */
 
-/* Returns the mtu for the given v6 destination address. */
-int sctp_v6_get_dst_mtu(const sockaddr_storage_t *address)
+/* FIXME: This macro needs to be moved to a common header file. */
+#define NIP6(addr) \
+        ntohs((addr)->s6_addr16[0]), \
+        ntohs((addr)->s6_addr16[1]), \
+        ntohs((addr)->s6_addr16[2]), \
+        ntohs((addr)->s6_addr16[3]), \
+        ntohs((addr)->s6_addr16[4]), \
+        ntohs((addr)->s6_addr16[5]), \
+        ntohs((addr)->s6_addr16[6]), \
+        ntohs((addr)->s6_addr16[7])
+
+/* Returns the dst cache entry for the given source and destination ip
+ * addresses.
+ */
+struct dst_entry *sctp_v6_get_dst(sockaddr_storage_t *daddr,
+				  sockaddr_storage_t *saddr)
 {
 	struct dst_entry *dst;
-	struct flowi fl;
-	int dst_mtu = SCTP_DEFAULT_MAXSEGMENT;
+	struct flowi fl = { .nl_u = { .ip6_u = { .daddr = &daddr->v6.sin6_addr,
+					       } } };
 
-	fl.proto = 0;
-	fl.fl6_dst = (struct in6_addr *)&address->v6.sin6_addr;
-	fl.fl6_src = NULL;
-	fl.fl6_flowlabel = 0;
-	fl.oif = 0;
-	fl.uli_u.ports.sport = 0;
-	fl.uli_u.ports.dport = 0;
+
+	SCTP_DEBUG_PRINTK("%s: DST=%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x ",
+			  __FUNCTION__, NIP6(fl.fl6_dst)); 
+
+	if (saddr) {
+		fl.fl6_src = &saddr->v6.sin6_addr;
+		SCTP_DEBUG_PRINTK(
+			"SRC=%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x - ",
+			NIP6(fl.fl6_src));
+	}
 
 	dst = ip6_route_output(NULL, &fl);
 	if (dst) {
-		dst_mtu = dst_pmtu(dst);
-		SCTP_DEBUG_PRINTK("sctp_v6_get_dst_mtu: "
-				  "ip6_route_output: dev:%s pmtu:%d\n",
-				  dst->dev->name, dst_mtu);
-		dst_release(dst);
+		struct rt6_info *rt;
+		rt = (struct rt6_info *)dst;
+		SCTP_DEBUG_PRINTK(
+			"rt6_dst:%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x "
+			"rt6_src:%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n",
+			NIP6(&rt->rt6i_dst.addr), NIP6(&rt->rt6i_src.addr));
 	} else {
-		SCTP_DEBUG_PRINTK("sctp_v6_get_dst_mtu: "
-				  "ip6_route_output failed, returning "
-				  "%d as dst_mtu\n", dst_mtu);
+		SCTP_DEBUG_PRINTK("NO ROUTE\n");
+		return NULL;
 	}
 
-	return dst_mtu;
+	return dst;
+}
+
+/* Check if the dst entry's source addr matches the given source addr. */
+int sctp_v6_cmp_saddr(struct dst_entry *dst, sockaddr_storage_t *saddr)
+{
+	struct rt6_info *rt = (struct rt6_info *)dst;
+
+	return ipv6_addr_cmp(&rt->rt6i_src.addr, &saddr->v6.sin6_addr); 
 }
 
 /* Initialize a PF_INET6 socket msg_name. */
@@ -301,7 +326,8 @@ static sctp_func_t sctp_ipv6_specific = {
 	.queue_xmit      = sctp_v6_xmit,
 	.setsockopt      = ipv6_setsockopt,
 	.getsockopt      = ipv6_getsockopt,
-	.get_dst_mtu     = sctp_v6_get_dst_mtu,
+	.get_dst	 = sctp_v6_get_dst,
+	.cmp_saddr	 = sctp_v6_cmp_saddr,
 	.net_header_len  = sizeof(struct ipv6hdr),
 	.sockaddr_len    = sizeof(struct sockaddr_in6),
 	.sa_family       = AF_INET6,

net/sctp/output.c

@@ -245,6 +245,7 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 	int err = 0;
 	int padding;		/* How much padding do we need?  */
 	__u8 packet_has_data = 0;
+	struct dst_entry *dst;
 
 	/* Do NOT generate a chunkless packet... */
 	if (skb_queue_empty(&packet->chunks))
@@ -410,13 +411,6 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 		asoc->peer.last_sent_to = transport;
 	}
 
-	/* Hey, before Linux changes, here's what we have to
-	 * do to force IP routing to recognize the change of
-	 * dest addr.				--xguo
-	 */
-	if (sk->dst_cache)
-		sk->dst_cache->obsolete = 1;
-
 	if (packet_has_data) {
 		struct timer_list *timer;
 		unsigned long timeout;
@@ -434,6 +428,13 @@ int sctp_packet_transmit(sctp_packet_t *packet)
 		}
 	}
 
+	dst = transport->dst;
+	if (!dst || dst->obsolete) {
+		sctp_transport_route(transport, NULL);
+	}
+
+	nskb->dst = dst_clone(transport->dst);
+
 	SCTP_DEBUG_PRINTK("***sctp_transmit_packet*** skb length %d\n",
 			  nskb->len);
 	(*transport->af_specific->queue_xmit)(nskb);

net/sctp/outqueue.c

@@ -204,18 +204,43 @@ int sctp_push_outqueue(sctp_outqueue_t *q, sctp_chunk_t *chunk)
 	return error;
 }
 
+/* Insert a chunk into the retransmit queue.  Chunks on the retransmit
+ * queue are kept in order, based on the TSNs.
+ */
+void sctp_retransmit_insert(struct list_head *tlchunk, sctp_outqueue_t *q)
+{
+	struct list_head *rlchunk;
+	sctp_chunk_t *tchunk, *rchunk;
+	__u32 ttsn, rtsn;
+	int done = 0;
+
+	tchunk = list_entry(tlchunk, sctp_chunk_t, transmitted_list);
+	ttsn = ntohl(tchunk->subh.data_hdr->tsn);
+
+	list_for_each(rlchunk, &q->retransmit) {
+		rchunk = list_entry(rlchunk, sctp_chunk_t, transmitted_list);
+		rtsn = ntohl(rchunk->subh.data_hdr->tsn);
+		if (TSN_lt(ttsn, rtsn)) {
+			list_add(tlchunk, rlchunk->prev);
+			done = 1;
+			break;
+		}
+	}
+	if (!done) {
+		list_add_tail(tlchunk, &q->retransmit);
+	}
+}
+			
 /* Mark all the eligible packets on a transport for retransmission.  */
 void sctp_retransmit_mark(sctp_outqueue_t *q, sctp_transport_t *transport,
 			  __u8 fast_retransmit)
 {
-	struct list_head *lchunk;
+	struct list_head *lchunk, *ltemp;
 	sctp_chunk_t *chunk;
-	struct list_head tlist;
 
-	INIT_LIST_HEAD(&tlist);
 
-	while (!list_empty(&transport->transmitted)) {
-		lchunk = sctp_list_dequeue(&transport->transmitted);
+	/* Walk through the specified transmitted queue.  */
+	list_for_each_safe(lchunk, ltemp, &transport->transmitted) {
 		chunk = list_entry(lchunk, sctp_chunk_t, transmitted_list);
 
 		/* If we are doing retransmission due to a fast retransmit,
@@ -224,10 +249,8 @@ void sctp_retransmit_mark(sctp_outqueue_t *q, sctp_transport_t *transport,
 		 * retransmission due to a timeout, only the chunks that are
 		 * not yet acked should be added to the retransmit queue.
 		 */
-		if ((fast_retransmit && !chunk->fast_retransmit) ||
-		    (!fast_retransmit && chunk->tsn_gap_acked)) {
-			list_add_tail(lchunk, &tlist);
-		} else {
+		if ((fast_retransmit && chunk->fast_retransmit) ||
+		   (!fast_retransmit && !chunk->tsn_gap_acked)) {
 			/* RFC 2960 6.2.1 Processing a Received SACK
 			 *
 			 * C) Any time a DATA chunk is marked for
@@ -257,16 +280,15 @@ void sctp_retransmit_mark(sctp_outqueue_t *q, sctp_transport_t *transport,
 				chunk->rtt_in_progress = 0;
 				transport->rto_pending = 0;
 			}
-			list_add_tail(lchunk, &q->retransmit);
+
+			/* Move the chunk to the retransmit queue. The chunks
+			 * on the retransmit queue is always kept in order.
+			 */
+			list_del(lchunk);
+			sctp_retransmit_insert(lchunk, q);
 		}
 	}
 
-	/* Reconstruct the transmitted queue with chunks that are not
-	 * eligible for retransmission.
-	 */
-	while (NULL != (lchunk = sctp_list_dequeue(&tlist)))
-		list_add_tail(lchunk, &transport->transmitted);
-
 	SCTP_DEBUG_PRINTK("%s: transport: %p, fast_retransmit: %d, "
 			  "cwnd: %d, ssthresh: %d, flight_size: %d, "
 			  "pba: %d\n", __FUNCTION__,

net/sctp/protocol.c

@@ -59,6 +59,7 @@
 /* Global data structures. */
 sctp_protocol_t sctp_proto;
 struct proc_dir_entry	*proc_net_sctp;
+struct sctp_mib sctp_statistics[NR_CPUS * 2];
 
 /* This is the global socket data structure used for responding to
  * the Out-of-the-blue (OOTB) packets.  A control sock will be created
@@ -255,27 +256,41 @@ int sctp_copy_local_addr_list(sctp_protocol_t *proto, sctp_bind_addr_t *bp,
 	return error;
 }
 
-/* Returns the mtu for the given v4 destination address.  */
-int sctp_v4_get_dst_mtu(const sockaddr_storage_t *address)
+/* Returns the dst cache entry for the given source and destination ip
+ * addresses.
+ */
+struct dst_entry *sctp_v4_get_dst(sockaddr_storage_t *daddr,
+				  sockaddr_storage_t *saddr)
 {
-	int dst_mtu = SCTP_DEFAULT_MAXSEGMENT;
 	struct rtable *rt;
-	struct flowi fl = { .nl_u = { .ip4_u =
-				      { .daddr = address->v4.sin_addr.s_addr } } };
+	struct flowi fl = { .nl_u = { .ip4_u = { .daddr =
+						    daddr->v4.sin_addr.s_addr,
+					       } } };
+
+	if (saddr)
+		fl.fl4_src = saddr->v4.sin_addr.s_addr;
+
+	SCTP_DEBUG_PRINTK("%s: DST:%u.%u.%u.%u, SRC:%u.%u.%u.%u - ",
+			  __FUNCTION__, NIPQUAD(fl.fl4_dst),
+			  NIPQUAD(fl.fl4_src));
 
 	if (ip_route_output_key(&rt, &fl)) {
-		SCTP_DEBUG_PRINTK("sctp_v4_get_dst_mtu:ip_route_output_key"
-				  " failed, returning %d as dst_mtu\n",
-				  dst_mtu);
-	} else {
-		dst_mtu = dst_pmtu(&rt->u.dst);
-		SCTP_DEBUG_PRINTK("sctp_v4_get_dst_mtu: "
-				  "ip_route_output_key: dev:%s pmtu:%d\n",
-				  rt->u.dst.dev->name, dst_mtu);
-		ip_rt_put(rt);
+		SCTP_DEBUG_PRINTK("NO ROUTE\n");
+		return NULL;
 	}
 
-	return dst_mtu;
+	SCTP_DEBUG_PRINTK("rt_dst:%u.%u.%u.%u, rt_src:%u.%u.%u.%u\n",
+			  NIPQUAD(rt->rt_src), NIPQUAD(rt->rt_dst));
+
+	return &rt->u.dst;
+}
+
+/* Check if the dst entry's source addr matches the given source addr. */
+int sctp_v4_cmp_saddr(struct dst_entry *dst, sockaddr_storage_t *saddr)
+{
+	struct rtable *rt = (struct rtable *)dst;
+
+	return (rt->rt_src == saddr->v4.sin_addr.s_addr); 
 }
 
 /* Event handler for inet device events.
@@ -437,7 +452,8 @@ sctp_func_t sctp_ipv4_specific = {
 	.queue_xmit     = ip_queue_xmit,
 	.setsockopt     = ip_setsockopt,
 	.getsockopt     = ip_getsockopt,
-	.get_dst_mtu    = sctp_v4_get_dst_mtu,
+	.get_dst	= sctp_v4_get_dst,
+	.cmp_saddr	= sctp_v4_cmp_saddr,
 	.net_header_len = sizeof(struct iphdr),
 	.sockaddr_len   = sizeof(struct sockaddr_in),
 	.sa_family      = AF_INET,

net/sctp/sm_make_chunk.c

@@ -47,6 +47,7 @@
  *    Dajiang Zhang	    <dajiang.zhang@nokia.com>
  *    Sridhar Samudrala	    <sri@us.ibm.com>
  *    Daisy Chang	    <daisyc@us.ibm.com>
+ *    Ardelle Fan	    <ardelle.fan@intel.com>
  *
  * Any bugs reported given to us we will try to fix... any fixes shared will
  * be incorporated into the next SCTP release.
@@ -838,6 +839,53 @@ sctp_chunk_t *sctp_make_abort_no_data(const sctp_association_t *asoc,
 	return retval;
 }
 
+/* Helper to create ABORT with a SCTP_ERROR_USER_ABORT error.  */
+sctp_chunk_t *sctp_make_abort_user(const sctp_association_t *asoc,
+				   const sctp_chunk_t *chunk,
+				   const struct msghdr *msg)
+{
+	sctp_chunk_t *retval;
+	void *payload = NULL, *payoff;
+	size_t paylen;
+	struct iovec *iov = msg->msg_iov;
+	int iovlen = msg->msg_iovlen;
+
+	paylen = get_user_iov_size(iov, iovlen);
+	retval = sctp_make_abort(asoc, chunk, sizeof(sctp_errhdr_t) + paylen);
+	if (!retval)
+		goto err_chunk;
+
+	if (paylen) {
+		/* Put the msg_iov together into payload.  */
+		payload = kmalloc(paylen, GFP_ATOMIC);
+		if (!payload)
+			goto err_payload;
+		payoff = payload;
+
+		for (; iovlen > 0; --iovlen) {
+			if (copy_from_user(payoff, iov->iov_base, iov->iov_len))
+				goto err_copy;
+			payoff += iov->iov_len;
+			iov++;
+		}
+	}
+
+	sctp_init_cause(retval, SCTP_ERROR_USER_ABORT, payload, paylen);
+
+	if (paylen)
+		kfree(payload);
+
+	return retval;
+
+err_copy:
+	kfree(payload);
+err_payload:
+	sctp_free_chunk(retval);
+	retval = NULL;
+err_chunk:
+	return retval;
+}
+
 /* Make a HEARTBEAT chunk.  */
 sctp_chunk_t *sctp_make_heartbeat(const sctp_association_t *asoc,
 				  const sctp_transport_t *transport,
@@ -982,16 +1030,17 @@ sctp_chunk_t *sctp_chunkify(struct sk_buff *skb, const sctp_association_t *asoc,
 	return retval;
 }
 
-/* Set chunk->source based on the IP header in chunk->skb.  */
-void sctp_init_source(sctp_chunk_t *chunk)
+/* Set chunk->source and dest based on the IP header in chunk->skb.  */
+void sctp_init_addrs(sctp_chunk_t *chunk)
 {
-	sockaddr_storage_t *source;
+	sockaddr_storage_t *source, *dest;
 	struct sk_buff *skb;
 	struct sctphdr *sh;
 	struct iphdr *ih4;
 	struct ipv6hdr *ih6;
 
 	source = &chunk->source;
+	dest = &chunk->dest;
 	skb = chunk->skb;
 	ih4 = skb->nh.iph;
 	ih6 = skb->nh.ipv6h;
@@ -1002,6 +1051,9 @@ void sctp_init_source(sctp_chunk_t *chunk)
 		source->v4.sin_family = AF_INET;
 		source->v4.sin_port = ntohs(sh->source);
 		source->v4.sin_addr.s_addr = ih4->saddr;
+		dest->v4.sin_family = AF_INET;
+		dest->v4.sin_port = ntohs(sh->dest);
+		dest->v4.sin_addr.s_addr = ih4->daddr;
 		break;
 
 	case 6:
@@ -1009,6 +1061,9 @@ void sctp_init_source(sctp_chunk_t *chunk)
 			source->v6.sin6_family = AF_INET6;
 			source->v6.sin6_port = ntohs(sh->source);
 			source->v6.sin6_addr = ih6->saddr;
+			dest->v6.sin6_family = AF_INET6;
+			dest->v6.sin6_port = ntohs(sh->dest);
+			dest->v6.sin6_addr = ih6->daddr;
 			/* FIXME:  What do we do with scope, etc. ? */
 			break;
 		)

net/sctp/sm_sideeffect.c

@@ -66,7 +66,8 @@ static void sctp_do_ecn_cwr_work(sctp_association_t *asoc,
 static void sctp_do_8_2_transport_strike(sctp_association_t *asoc,
 					 sctp_transport_t *transport);
 static void sctp_cmd_init_failed(sctp_cmd_seq_t *, sctp_association_t *asoc);
-static void sctp_cmd_assoc_failed(sctp_cmd_seq_t *, sctp_association_t *asoc);
+static void sctp_cmd_assoc_failed(sctp_cmd_seq_t *, sctp_association_t *asoc,
+				  sctp_event_t event_type, sctp_chunk_t *chunk);
 static void sctp_cmd_process_init(sctp_cmd_seq_t *, sctp_association_t *asoc,
 				  sctp_chunk_t *chunk,
 				  sctp_init_chunk_t *peer_init,
@@ -251,7 +252,7 @@ int sctp_cmd_interpreter(sctp_event_t event_type, sctp_subtype_t subtype,
 	int force;
 	sctp_cmd_t *command;
 	sctp_chunk_t *new_obj;
-	sctp_chunk_t *chunk;
+	sctp_chunk_t *chunk = NULL;
 	sctp_packet_t *packet;
 	struct list_head *pos;
 	struct timer_list *timer;
@@ -259,7 +260,8 @@ int sctp_cmd_interpreter(sctp_event_t event_type, sctp_subtype_t subtype,
 	sctp_transport_t *t;
 	sctp_sackhdr_t sackh;
 
-	chunk = (sctp_chunk_t *) event_arg;
+	if(SCTP_EVENT_T_TIMEOUT != event_type)
+		chunk = (sctp_chunk_t *) event_arg;
 
 	/* Note:  This whole file is a huge candidate for rework.
 	 * For example, each command could either have its own handler, so
@@ -504,7 +506,8 @@ int sctp_cmd_interpreter(sctp_event_t event_type, sctp_subtype_t subtype,
 			break;
 
 		case SCTP_CMD_ASSOC_FAILED:
-			sctp_cmd_assoc_failed(commands, asoc);
+			sctp_cmd_assoc_failed(commands, asoc, event_type,
+					      chunk);
 			break;
 
 		case SCTP_CMD_COUNTER_INC:
@@ -595,10 +598,7 @@ int sctp_cmd_interpreter(sctp_event_t event_type, sctp_subtype_t subtype,
 /* A helper function for delayed processing of INET ECN CE bit. */
 static void sctp_do_ecn_ce_work(sctp_association_t *asoc, __u32 lowest_tsn)
 {
-	/*
-	 * Save the TSN away for comparison when we receive CWR
-	 * Note: dp->TSN is expected in host endian
-	 */
+	/* Save the TSN away for comparison when we receive CWR */
 
 	asoc->last_ecne_tsn = lowest_tsn;
 	asoc->need_ecne = 1;
@@ -621,7 +621,6 @@ static sctp_chunk_t *sctp_do_ecn_ecne_work(sctp_association_t *asoc,
 					   sctp_chunk_t *chunk)
 {
 	sctp_chunk_t *repl;
-	sctp_transport_t *transport;
 
 	/* Our previously transmitted packet ran into some congestion
 	 * so we should take action by reducing cwnd and ssthresh
@@ -629,43 +628,28 @@ static sctp_chunk_t *sctp_do_ecn_ecne_work(sctp_association_t *asoc,
 	 * sending a CWR.
 	 */
 
-	/* Find which transport's congestion variables
-	 * need to be adjusted.
+	/* First, try to determine if we want to actually lower
+	 * our cwnd variables.  Only lower them if the ECNE looks more
+	 * recent than the last response.
 	 */
+	if (TSN_lt(asoc->last_cwr_tsn, lowest_tsn)) {
+		sctp_transport_t *transport;
 
-	transport = sctp_assoc_lookup_tsn(asoc, lowest_tsn);
+		/* Find which transport's congestion variables
+		 * need to be adjusted.
+		 */
+		transport = sctp_assoc_lookup_tsn(asoc, lowest_tsn);
 
-	/* Update the congestion variables. */
-	if (transport)
-		sctp_transport_lower_cwnd(transport, SCTP_LOWER_CWND_ECNE);
+		/* Update the congestion variables. */
+		if (transport)
+			sctp_transport_lower_cwnd(transport,
+						  SCTP_LOWER_CWND_ECNE);
+		asoc->last_cwr_tsn = lowest_tsn;
+	}
 
-	/* Save away a rough idea of when we last sent out a CWR.
-	 * We compare against this value (see above) to decide if
-	 * this is a fairly new request.
-	 * Note that this is not a perfect solution.  We may
-	 * have moved beyond the window (several times) by the
-	 * next time we get an ECNE.  However, it is cute.  This idea
-	 * came from Randy's reference code.
-	 *
-	 * Here's what RFC 2960 has to say about CWR.  This is NOT
-	 * what we do.
-	 *
-	 * RFC 2960 Appendix A
-	 *
-	 *    CWR:
-	 *
-	 *    RFC 2481 details a specific bit for a sender to send in
-	 *    the header of its next outbound TCP segment to indicate
-	 *    to its peer that it has reduced its congestion window.
-	 *    This is termed the CWR bit.  For SCTP the same
-	 *    indication is made by including the CWR chunk.  This
-	 *    chunk contains one data element, i.e. the TSN number
-	 *    that was sent in the ECNE chunk.  This element
-	 *    represents the lowest TSN number in the datagram that
-	 *    was originally marked with the CE bit.
+	/* Always try to quiet the other end.  In case of lost CWR,
+	 * resend last_cwr_tsn.  
 	 */
-	asoc->last_cwr_tsn = asoc->next_tsn - 1;
-
 	repl = sctp_make_cwr(asoc, asoc->last_cwr_tsn, chunk);
 
 	/* If we run out of memory, it will look like a lost CWR.  We'll
@@ -1038,14 +1022,26 @@ static void sctp_cmd_init_failed(sctp_cmd_seq_t *commands,
 
 /* Worker routine to handle SCTP_CMD_ASSOC_FAILED.  */
 static void sctp_cmd_assoc_failed(sctp_cmd_seq_t *commands,
-				  sctp_association_t *asoc)
+				  sctp_association_t *asoc,
+				  sctp_event_t event_type,
+				  sctp_chunk_t *chunk)
 {
 	sctp_ulpevent_t *event;
+	__u16 error = 0;
+
+	if (event_type == SCTP_EVENT_T_PRIMITIVE)
+		error = SCTP_ERROR_USER_ABORT;
+
+	if (chunk && (SCTP_CID_ABORT == chunk->chunk_hdr->type) &&
+	    (ntohs(chunk->chunk_hdr->length) >= (sizeof(struct sctp_chunkhdr) +
+				 		 sizeof(struct sctp_errhdr)))) {
+		error = ((sctp_errhdr_t *)chunk->skb->data)->cause;
+	}
 
 	event = sctp_ulpevent_make_assoc_change(asoc,
 						0,
 						SCTP_COMM_LOST,
-						0, 0, 0,
+						error, 0, 0,
 						GFP_ATOMIC);
 
 	if (event)

net/sctp/sm_statefuns.c

@@ -43,6 +43,7 @@
  *    Hui Huang 	    <hui.huang@nokia.com>
  *    Dajiang Zhang 	    <dajiang.zhang@nokia.com>
  *    Daisy Chang	    <daisyc@us.ibm.com>
+ *    Ardelle Fan	    <ardelle.fan@intel.com>
  *
  * Any bugs reported given to us we will try to fix... any fixes shared will
  * be incorporated into the next SCTP release.
@@ -865,9 +866,9 @@ sctp_disposition_t sctp_sf_backbeat_8_3(const sctp_endpoint_t *ep,
 /* Helper function to send out an abort for the restart
  * condition.
  */
-static int sctp_sf_send_restart_abort(sockaddr_storage_t *ssa, 
+static int sctp_sf_send_restart_abort(sockaddr_storage_t *ssa,
 				      sctp_chunk_t *init,
-				      sctp_cmd_seq_t *commands) 
+				      sctp_cmd_seq_t *commands)
 {
 	int len;
 	sctp_packet_t *pkt;
@@ -881,7 +882,7 @@ static int sctp_sf_send_restart_abort(sockaddr_storage_t *ssa,
 	 */
 	errhdr = (sctp_errhdr_t *)buffer;
 	addrparm = (sctp_addr_param_t *)errhdr->variable;
-	
+
 	/* Copy into a parm format. */
 	len = sockaddr2sctp_addr(ssa, addrparm);
 	len += sizeof(sctp_errhdr_t);
@@ -897,7 +898,7 @@ static int sctp_sf_send_restart_abort(sockaddr_storage_t *ssa,
 	 */
 	pkt = sctp_abort_pkt_new(ep, NULL, init, errhdr, len);
 
-	if (!pkt) 
+	if (!pkt)
 		goto out;
 	sctp_add_cmd_sf(commands, SCTP_CMD_SEND_PKT, SCTP_PACKET(pkt));
 
@@ -906,18 +907,18 @@ static int sctp_sf_send_restart_abort(sockaddr_storage_t *ssa,
 
 out:
 	/* Even if there is no memory, treat as a failure so
-	 * the packet will get dropped. 
+	 * the packet will get dropped.
 	 */
 	return 0;
 }
 
-/* A restart is occuring, check to make sure no new addresses 
+/* A restart is occuring, check to make sure no new addresses
  * are being added as we may be under a takeover attack.
  */
 static int sctp_sf_check_restart_addrs(const sctp_association_t *new_asoc,
 				       const sctp_association_t *asoc,
 				       sctp_chunk_t *init,
-				       sctp_cmd_seq_t *commands) 
+				       sctp_cmd_seq_t *commands)
 {
 	sctp_transport_t *new_addr, *addr;
 	struct list_head *pos, *pos2;
@@ -956,8 +957,8 @@ static int sctp_sf_check_restart_addrs(const sctp_association_t *new_asoc,
 	if (!found && new_addr) {
 		sctp_sf_send_restart_abort(&new_addr->ipaddr, init, commands);
 	}
-	
-	/* Return success if all addresses were found. */   
+
+	/* Return success if all addresses were found. */
 	return found;
 }
 
@@ -1053,7 +1054,7 @@ static char sctp_tietags_compare(sctp_association_t *new_asoc,
  */
 static sctp_disposition_t sctp_sf_do_unexpected_init(
 	const sctp_endpoint_t *ep,
-	const sctp_association_t *asoc, 
+	const sctp_association_t *asoc,
 	const sctp_subtype_t type,
 	void *arg, sctp_cmd_seq_t *commands)
 {
@@ -1130,10 +1131,10 @@ static sctp_disposition_t sctp_sf_do_unexpected_init(
 	/* Make sure no new addresses are being added during the
 	 * restart.   Do not do this check for COOKIE-WAIT state,
 	 * since there are no peer addresses to check against.
-	 * Upon return an ABORT will have been sent if needed.  
+	 * Upon return an ABORT will have been sent if needed.
 	 */
 	if (asoc->state != SCTP_STATE_COOKIE_WAIT) {
-		if (!sctp_sf_check_restart_addrs(new_asoc, asoc, chunk, 
+		if (!sctp_sf_check_restart_addrs(new_asoc, asoc, chunk,
 						 commands)) {
 			retval = SCTP_DISPOSITION_CONSUME;
 			goto cleanup_asoc;
@@ -1333,9 +1334,9 @@ static sctp_disposition_t sctp_sf_do_dupcook_a(const sctp_endpoint_t *ep,
 	 * since you'd have to get inside the cookie.
 	 */
 	if (!sctp_sf_check_restart_addrs(new_asoc, asoc, chunk, commands)) {
-		printk("cookie echo check\n"); 
+		printk("cookie echo check\n");
 		return SCTP_DISPOSITION_CONSUME;
-	}	
+	}
 
 	/* For now, fail any unsent/unacked data.  Consider the optional
 	 * choice of resending of this data.
@@ -1542,7 +1543,7 @@ sctp_disposition_t sctp_sf_do_5_2_4_dupcook(const sctp_endpoint_t *ep,
 	 * are in good shape.
 	 */
         chunk->subh.cookie_hdr = (sctp_signed_cookie_t *)chunk->skb->data;
-	skb_pull(chunk->skb, ntohs(chunk->chunk_hdr->length) - 
+	skb_pull(chunk->skb, ntohs(chunk->chunk_hdr->length) -
 		 sizeof(sctp_chunkhdr_t));
 
 	/* In RFC 2960 5.2.4 3, if both Verification Tags in the State Cookie
@@ -2098,18 +2099,11 @@ sctp_disposition_t sctp_sf_do_ecne(const sctp_endpoint_t *ep,
 
 	ecne = (sctp_ecnehdr_t *) chunk->skb->data;
 	skb_pull(chunk->skb, sizeof(sctp_ecnehdr_t));
-	ecne->lowest_tsn = ntohl(ecne->lowest_tsn);
-
-	/* Casting away the const, as we are just modifying the spinlock,
-	 * not the association itself.   This should go away in the near
-	 * future when we move to an endpoint based lock.
-	 */
 
 	/* If this is a newer ECNE than the last CWR packet we sent out */
-	if (TSN_lt(asoc->last_cwr_tsn, ecne->lowest_tsn)) {
-		sctp_add_cmd_sf(commands, SCTP_CMD_ECN_ECNE,
-				SCTP_U32(ecne->lowest_tsn));
-	}
+	sctp_add_cmd_sf(commands, SCTP_CMD_ECN_ECNE,
+			SCTP_U32(ntohl(ecne->lowest_tsn)));
+
 	return SCTP_DISPOSITION_CONSUME;
 }
 
@@ -2641,7 +2635,7 @@ sctp_disposition_t sctp_sf_operr_notify(const sctp_endpoint_t *ep,
 	sctp_ulpevent_t *ev;
 
 	while (chunk->chunk_end > chunk->skb->data) {
-		ev = sctp_ulpevent_make_remote_error(asoc, chunk, 0, 
+		ev = sctp_ulpevent_make_remote_error(asoc, chunk, 0,
 						     GFP_ATOMIC);
 		if (!ev)
 			goto nomem;
@@ -3401,13 +3395,14 @@ sctp_disposition_t sctp_sf_do_9_1_prm_abort(const sctp_endpoint_t *ep,
 	 * from its upper layer, but retransmits data to the far end
 	 * if necessary to fill gaps.
 	 */
+	struct msghdr *msg = arg;
 	sctp_chunk_t *abort;
 	sctp_disposition_t retval;
 
 	retval = SCTP_DISPOSITION_CONSUME;
 
 	/* Generate ABORT chunk to send the peer.  */
-	abort = sctp_make_abort(asoc, NULL, 0);
+	abort = sctp_make_abort_user(asoc, NULL, msg);
 	if (!abort)
 		retval = SCTP_DISPOSITION_NOMEM;
 	else
@@ -3525,6 +3520,7 @@ sctp_disposition_t sctp_sf_cookie_wait_prm_abort(const sctp_endpoint_t *ep,
 					void *arg,
 					sctp_cmd_seq_t *commands)
 {
+	struct msghdr *msg = arg;
 	sctp_chunk_t *abort;
 	sctp_disposition_t retval;
 
@@ -3534,7 +3530,7 @@ sctp_disposition_t sctp_sf_cookie_wait_prm_abort(const sctp_endpoint_t *ep,
 	retval = SCTP_DISPOSITION_CONSUME;
 
 	/* Generate ABORT chunk to send the peer */
-	abort = sctp_make_abort(asoc, NULL, 0);
+	abort = sctp_make_abort_user(asoc, NULL, msg);
 	if (!abort)
 		retval = SCTP_DISPOSITION_NOMEM;
 	else
@@ -4258,6 +4254,11 @@ sctp_packet_t *sctp_ootb_pkt_new(const sctp_association_t *asoc,
 	if (!packet)
 		goto nomem_packet;
 
+	/* Cache a route for the transport with the chunk's destination as
+	 * the source address.
+	 */
+	sctp_transport_route(transport, (sockaddr_storage_t *)&chunk->dest);
+
 	packet = sctp_packet_init(packet, transport, sport, dport);
 	packet = sctp_packet_config(packet, vtag, 0, NULL);
 

net/sctp/socket.c

@@ -804,10 +804,12 @@ SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
 	SCTP_DEBUG_PRINTK("msg_len: %Zd, sinfo_flags: 0x%x\n",
 			  msg_len, sinfo_flags);
 
-	/* If MSG_EOF|MSG_ABORT is set, no data can be sent.  Disallow
-	 * sending 0-length messages when MSG_EOF|MSG_ABORT is not set.
-	 */
-	if (((sinfo_flags & (MSG_EOF|MSG_ABORT)) && (msg_len > 0)) ||
+	/* If MSG_EOF is set, no data can be sent. Disallow sending zero
+	 * length messages when MSG_EOF|MSG_ABORT is not set.
+	 * If MSG_ABORT is set, the message length could be non zero with
+	 * the msg_iov set to the user abort reason.
+ 	 */
+	if (((sinfo_flags & MSG_EOF) && (msg_len > 0)) ||
 	    (!(sinfo_flags & (MSG_EOF|MSG_ABORT)) && (msg_len == 0))) {
 		err = -EINVAL;
 		goto out_nounlock;
@@ -879,7 +881,7 @@ SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
 		}
 		if (sinfo_flags & MSG_ABORT) {
 			SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
-			sctp_primitive_ABORT(asoc, NULL);
+			sctp_primitive_ABORT(asoc, msg);
 			err = 0;
 			goto out_unlock;
 		}
@@ -1238,6 +1240,9 @@ static inline int sctp_setsockopt_autoclose(struct sock *sk, char *optval,
 {
 	sctp_opt_t *sp = sctp_sk(sk);
 
+	/* Applicable to UDP-style socket only */
+	if (SCTP_SOCKET_TCP == sp->type)
+		return -EOPNOTSUPP;
 	if (optlen != sizeof(int))
 		return -EINVAL;
 	if (copy_from_user(&sp->autoclose, optval, optlen))
@@ -1593,6 +1598,9 @@ static inline int sctp_getsockopt_set_events(struct sock *sk, int len, char *opt
 
 static inline 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)
+		return -EOPNOTSUPP;
 	if (len != sizeof(int))
 		return -EINVAL;
 	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
@@ -1614,10 +1622,10 @@ SCTP_STATIC int sctp_do_peeloff(sctp_association_t *assoc, struct socket **newso
 	int err = 0;
 
 	/* An association cannot be branched off from an already peeled-off
-	 * socket.
+	 * socket, nor is this supported for tcp style sockets.
 	 */
-	if (SCTP_SOCKET_UDP_HIGH_BANDWIDTH == sctp_sk(oldsk)->type)
-		return -EINVAL;
+	if (SCTP_SOCKET_UDP != sctp_sk(oldsk)->type)
+		return -EOPNOTSUPP;
 
 	/* Create a new socket.  */
 	err = sock_create(PF_INET, SOCK_SEQPACKET, IPPROTO_SCTP, &tmpsock);

net/sctp/transport.c

@@ -87,7 +87,6 @@ sctp_transport_t *sctp_transport_init(sctp_transport_t *peer,
 	peer->ipaddr = *addr;
 	peer->af_specific = sctp_get_af_specific(addr);
 	peer->asoc = NULL;
-	peer->pmtu = peer->af_specific->get_dst_mtu(addr);
 
 	/* From 6.3.1 RTO Calculation:
 	 *
@@ -161,6 +160,7 @@ void sctp_transport_destroy(sctp_transport_t *transport)
 	if (transport->asoc)
 		sctp_association_put(transport->asoc);
 
+	dst_release(transport->dst);
 	kfree(transport);
 	SCTP_DBG_OBJCNT_DEC(transport);
 }
@@ -200,6 +200,78 @@ void sctp_transport_set_owner(sctp_transport_t *transport,
 	sctp_association_hold(asoc);
 }
 
+/* Caches the dst entry for a transport's destination address and an optional
+ * souce address.
+ */ 
+void sctp_transport_route(sctp_transport_t *transport,
+			  sockaddr_storage_t *saddr)
+{
+	sctp_association_t *asoc = transport->asoc;
+	sctp_func_t *af = transport->af_specific;
+	sockaddr_storage_t *daddr = &transport->ipaddr;
+	sctp_bind_addr_t *bp;
+	rwlock_t *addr_lock;
+	struct sockaddr_storage_list *laddr;
+	struct list_head *pos;
+	struct dst_entry *dst;
+
+	dst = af->get_dst(daddr, saddr);
+
+	/* If there is no association or if a source address is passed,
+	 * no more validation is required.
+	 */
+	if (!asoc || saddr)
+		goto out;
+
+	if (SCTP_STATE_ESTABLISHED == asoc->state) {
+		bp = &asoc->base.bind_addr;
+		addr_lock = &asoc->base.addr_lock;
+	} else {
+		bp = &asoc->ep->base.bind_addr;
+		addr_lock = &asoc->ep->base.addr_lock;
+	}
+
+	if (dst) {
+		/* Walk through the bind address list and look for a bind
+		 * address that matches the source address of the returned dst.
+		 */
+		sctp_read_lock(addr_lock);
+		list_for_each(pos, &bp->address_list) {
+			laddr = list_entry(pos, struct sockaddr_storage_list,
+					   list);
+			if (af->cmp_saddr(dst, &laddr->a))
+				goto out_unlock;
+		}
+		sctp_read_unlock(addr_lock);
+	
+		/* None of the bound addresses match the source address of the
+		 * dst. So release it.
+		 */	
+		dst_release(dst);
+	}
+
+	/* Walk through the bind address list and try to get a dst that
+	 * matches a bind address as the source address.
+	 */
+	sctp_read_lock(addr_lock);
+	list_for_each(pos, &bp->address_list) {
+		laddr = list_entry(pos, struct sockaddr_storage_list, list);
+
+		dst = af->get_dst(daddr, &laddr->a);
+		if (dst)
+			goto out_unlock;
+	}
+
+out_unlock:
+	sctp_read_unlock(addr_lock);
+out:
+	transport->dst = dst;
+	if (dst)
+		transport->pmtu = dst_pmtu(dst);
+	else
+		transport->pmtu = SCTP_DEFAULT_MAXSEGMENT;
+}
+
 /* Hold a reference to a transport.  */
 void sctp_transport_hold(sctp_transport_t *transport)
 {