tipc: split link outqueue

struct tipc_link contains one single queue for outgoing packets,
where both transmitted and waiting packets are queued.

This infrastructure is hard to maintain, because we need
to keep a number of fields to keep track of which packets are
sent or unsent, and the number of packets in each category.

A lot of code becomes simpler if we split this queue into a transmission
queue, where sent/unacknowledged packets are kept, and a backlog queue,
where we keep the not yet sent packets.

In this commit we do this separation.
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/tipc/bcast.c

@@ -135,9 +135,10 @@ static void bclink_set_last_sent(struct net *net)
 {
 	struct tipc_net *tn = net_generic(net, tipc_net_id);
 	struct tipc_link *bcl = tn->bcl;
+	struct sk_buff *skb = skb_peek(&bcl->backlogq);
 
-	if (bcl->next_out)
-		bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
+	if (skb)
+		bcl->fsm_msg_cnt = mod(buf_seqno(skb) - 1);
 	else
 		bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
 }
@@ -180,7 +181,7 @@ static void bclink_retransmit_pkt(struct tipc_net *tn, u32 after, u32 to)
 	struct sk_buff *skb;
 	struct tipc_link *bcl = tn->bcl;
 
-	skb_queue_walk(&bcl->outqueue, skb) {
+	skb_queue_walk(&bcl->transmq, skb) {
 		if (more(buf_seqno(skb), after)) {
 			tipc_link_retransmit(bcl, skb, mod(to - after));
 			break;
@@ -210,7 +211,6 @@ void tipc_bclink_wakeup_users(struct net *net)
 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 {
 	struct sk_buff *skb, *tmp;
-	struct sk_buff *next;
 	unsigned int released = 0;
 	struct net *net = n_ptr->net;
 	struct tipc_net *tn = net_generic(net, tipc_net_id);
@@ -221,7 +221,7 @@ void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 	tipc_bclink_lock(net);
 
 	/* Bail out if tx queue is empty (no clean up is required) */
-	skb = skb_peek(&tn->bcl->outqueue);
+	skb = skb_peek(&tn->bcl->transmq);
 	if (!skb)
 		goto exit;
 
@@ -248,27 +248,19 @@ void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 	}
 
 	/* Skip over packets that node has previously acknowledged */
-	skb_queue_walk(&tn->bcl->outqueue, skb) {
+	skb_queue_walk(&tn->bcl->transmq, skb) {
 		if (more(buf_seqno(skb), n_ptr->bclink.acked))
 			break;
 	}
 
 	/* Update packets that node is now acknowledging */
-	skb_queue_walk_from_safe(&tn->bcl->outqueue, skb, tmp) {
+	skb_queue_walk_from_safe(&tn->bcl->transmq, skb, tmp) {
 		if (more(buf_seqno(skb), acked))
 			break;
-
-		next = tipc_skb_queue_next(&tn->bcl->outqueue, skb);
-		if (skb != tn->bcl->next_out) {
-			bcbuf_decr_acks(skb);
-		} else {
-			bcbuf_set_acks(skb, 0);
-			tn->bcl->next_out = next;
-			bclink_set_last_sent(net);
-		}
-
+		bcbuf_decr_acks(skb);
+		bclink_set_last_sent(net);
 		if (bcbuf_acks(skb) == 0) {
-			__skb_unlink(skb, &tn->bcl->outqueue);
+			__skb_unlink(skb, &tn->bcl->transmq);
 			kfree_skb(skb);
 			released = 1;
 		}
@@ -276,7 +268,7 @@ void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 	n_ptr->bclink.acked = acked;
 
 	/* Try resolving broadcast link congestion, if necessary */
-	if (unlikely(tn->bcl->next_out)) {
+	if (unlikely(skb_peek(&tn->bcl->backlogq))) {
 		tipc_link_push_packets(tn->bcl);
 		bclink_set_last_sent(net);
 	}
@@ -323,7 +315,7 @@ void tipc_bclink_update_link_state(struct tipc_node *n_ptr,
 	buf = tipc_buf_acquire(INT_H_SIZE);
 	if (buf) {
 		struct tipc_msg *msg = buf_msg(buf);
-		struct sk_buff *skb = skb_peek(&n_ptr->bclink.deferred_queue);
+		struct sk_buff *skb = skb_peek(&n_ptr->bclink.deferdq);
 		u32 to = skb ? buf_seqno(skb) - 1 : n_ptr->bclink.last_sent;
 
 		tipc_msg_init(tn->own_addr, msg, BCAST_PROTOCOL, STATE_MSG,
@@ -398,7 +390,7 @@ int tipc_bclink_xmit(struct net *net, struct sk_buff_head *list)
 		if (likely(bclink->bcast_nodes.count)) {
 			rc = __tipc_link_xmit(net, bcl, list);
 			if (likely(!rc)) {
-				u32 len = skb_queue_len(&bcl->outqueue);
+				u32 len = skb_queue_len(&bcl->transmq);
 
 				bclink_set_last_sent(net);
 				bcl->stats.queue_sz_counts++;
@@ -563,25 +555,25 @@ void tipc_bclink_rcv(struct net *net, struct sk_buff *buf)
 		if (node->bclink.last_in == node->bclink.last_sent)
 			goto unlock;
 
-		if (skb_queue_empty(&node->bclink.deferred_queue)) {
+		if (skb_queue_empty(&node->bclink.deferdq)) {
 			node->bclink.oos_state = 1;
 			goto unlock;
 		}
 
-		msg = buf_msg(skb_peek(&node->bclink.deferred_queue));
+		msg = buf_msg(skb_peek(&node->bclink.deferdq));
 		seqno = msg_seqno(msg);
 		next_in = mod(next_in + 1);
 		if (seqno != next_in)
 			goto unlock;
 
 		/* Take in-sequence message from deferred queue & deliver it */
-		buf = __skb_dequeue(&node->bclink.deferred_queue);
+		buf = __skb_dequeue(&node->bclink.deferdq);
 		goto receive;
 	}
 
 	/* Handle out-of-sequence broadcast message */
 	if (less(next_in, seqno)) {
-		deferred = tipc_link_defer_pkt(&node->bclink.deferred_queue,
+		deferred = tipc_link_defer_pkt(&node->bclink.deferdq,
 					       buf);
 		bclink_update_last_sent(node, seqno);
 		buf = NULL;
@@ -638,7 +630,6 @@ static int tipc_bcbearer_send(struct net *net, struct sk_buff *buf,
 		msg_set_non_seq(msg, 1);
 		msg_set_mc_netid(msg, tn->net_id);
 		tn->bcl->stats.sent_info++;
-
 		if (WARN_ON(!bclink->bcast_nodes.count)) {
 			dump_stack();
 			return 0;
@@ -917,8 +908,9 @@ int tipc_bclink_init(struct net *net)
 	sprintf(bcbearer->media.name, "tipc-broadcast");
 
 	spin_lock_init(&bclink->lock);
-	__skb_queue_head_init(&bcl->outqueue);
-	__skb_queue_head_init(&bcl->deferred_queue);
+	__skb_queue_head_init(&bcl->transmq);
+	__skb_queue_head_init(&bcl->backlogq);
+	__skb_queue_head_init(&bcl->deferdq);
 	skb_queue_head_init(&bcl->wakeupq);
 	bcl->next_out_no = 1;
 	spin_lock_init(&bclink->node.lock);

net/tipc/link.h

@@ -124,7 +124,8 @@ struct tipc_stats {
  * @max_pkt: current maximum packet size for this link
  * @max_pkt_target: desired maximum packet size for this link
  * @max_pkt_probes: # of probes based on current (max_pkt, max_pkt_target)
- * @outqueue: outbound message queue
+ * @transmitq: queue for sent, non-acked messages
+ * @backlogq: queue for messages waiting to be sent
  * @next_out_no: next sequence number to use for outbound messages
  * @last_retransmitted: sequence number of most recently retransmitted message
  * @stale_count: # of identical retransmit requests made by peer
@@ -177,20 +178,21 @@ struct tipc_link {
 	u32 max_pkt_probes;
 
 	/* Sending */
-	struct sk_buff_head outqueue;
+	struct sk_buff_head transmq;
+	struct sk_buff_head backlogq;
 	u32 next_out_no;
+	u32 window;
 	u32 last_retransmitted;
 	u32 stale_count;
 
 	/* Reception */
 	u32 next_in_no;
-	struct sk_buff_head deferred_queue;
-	u32 unacked_window;
+	u32 rcv_unacked;
+	struct sk_buff_head deferdq;
 	struct sk_buff_head inputq;
 	struct sk_buff_head namedq;
 
 	/* Congestion handling */
-	struct sk_buff *next_out;
 	struct sk_buff_head wakeupq;
 
 	/* Fragmentation/reassembly */
@@ -302,9 +304,4 @@ static inline int link_reset_reset(struct tipc_link *l_ptr)
 	return l_ptr->state == RESET_RESET;
 }
 
-static inline int link_congested(struct tipc_link *l_ptr)
-{
-	return skb_queue_len(&l_ptr->outqueue) >= l_ptr->queue_limit[0];
-}
-
 #endif

net/tipc/msg.c

@@ -330,33 +330,36 @@ int tipc_msg_build(struct tipc_msg *mhdr, struct msghdr *m,
 
 /**
  * tipc_msg_bundle(): Append contents of a buffer to tail of an existing one
- * @list: the buffer chain of the existing buffer ("bundle")
+ * @bskb: the buffer to append to ("bundle")
  * @skb:  buffer to be appended
  * @mtu:  max allowable size for the bundle buffer
  * Consumes buffer if successful
  * Returns true if bundling could be performed, otherwise false
  */
-bool tipc_msg_bundle(struct sk_buff_head *list, struct sk_buff *skb, u32 mtu)
+bool tipc_msg_bundle(struct sk_buff *bskb, struct sk_buff *skb, u32 mtu)
 {
-	struct sk_buff *bskb = skb_peek_tail(list);
-	struct tipc_msg *bmsg = buf_msg(bskb);
+	struct tipc_msg *bmsg;
 	struct tipc_msg *msg = buf_msg(skb);
-	unsigned int bsz = msg_size(bmsg);
+	unsigned int bsz;
 	unsigned int msz = msg_size(msg);
-	u32 start = align(bsz);
+	u32 start, pad;
 	u32 max = mtu - INT_H_SIZE;
-	u32 pad = start - bsz;
 
 	if (likely(msg_user(msg) == MSG_FRAGMENTER))
 		return false;
+	if (!bskb)
+		return false;
+	bmsg = buf_msg(bskb);
+	bsz = msg_size(bmsg);
+	start = align(bsz);
+	pad = start - bsz;
+
 	if (unlikely(msg_user(msg) == CHANGEOVER_PROTOCOL))
 		return false;
 	if (unlikely(msg_user(msg) == BCAST_PROTOCOL))
 		return false;
 	if (likely(msg_user(bmsg) != MSG_BUNDLER))
 		return false;
-	if (likely(!TIPC_SKB_CB(bskb)->bundling))
-		return false;
 	if (unlikely(skb_tailroom(bskb) < (pad + msz)))
 		return false;
 	if (unlikely(max < (start + msz)))
@@ -419,12 +422,11 @@ bool tipc_msg_extract(struct sk_buff *skb, struct sk_buff **iskb, int *pos)
  * Replaces buffer if successful
  * Returns true if success, otherwise false
  */
-bool tipc_msg_make_bundle(struct sk_buff_head *list,
-			  struct sk_buff *skb, u32 mtu, u32 dnode)
+bool tipc_msg_make_bundle(struct sk_buff **skb, u32 mtu, u32 dnode)
 {
 	struct sk_buff *bskb;
 	struct tipc_msg *bmsg;
-	struct tipc_msg *msg = buf_msg(skb);
+	struct tipc_msg *msg = buf_msg(*skb);
 	u32 msz = msg_size(msg);
 	u32 max = mtu - INT_H_SIZE;
 
@@ -448,9 +450,9 @@ bool tipc_msg_make_bundle(struct sk_buff_head *list,
 	msg_set_seqno(bmsg, msg_seqno(msg));
 	msg_set_ack(bmsg, msg_ack(msg));
 	msg_set_bcast_ack(bmsg, msg_bcast_ack(msg));
-	TIPC_SKB_CB(bskb)->bundling = true;
-	__skb_queue_tail(list, bskb);
-	return tipc_msg_bundle(list, skb, mtu);
+	tipc_msg_bundle(bskb, *skb, mtu);
+	*skb = bskb;
+	return true;
 }
 
 /**

net/tipc/msg.h

@@ -767,9 +767,9 @@ struct sk_buff *tipc_msg_create(uint user, uint type, uint hdr_sz,
 				uint data_sz, u32 dnode, u32 onode,
 				u32 dport, u32 oport, int errcode);
 int tipc_buf_append(struct sk_buff **headbuf, struct sk_buff **buf);
-bool tipc_msg_bundle(struct sk_buff_head *list, struct sk_buff *skb, u32 mtu);
-bool tipc_msg_make_bundle(struct sk_buff_head *list,
-			  struct sk_buff *skb, u32 mtu, u32 dnode);
+bool tipc_msg_bundle(struct sk_buff *bskb, struct sk_buff *skb, u32 mtu);
+
+bool tipc_msg_make_bundle(struct sk_buff **skb, u32 mtu, u32 dnode);
 bool tipc_msg_extract(struct sk_buff *skb, struct sk_buff **iskb, int *pos);
 int tipc_msg_build(struct tipc_msg *mhdr, struct msghdr *m,
 		   int offset, int dsz, int mtu, struct sk_buff_head *list);

net/tipc/node.c

@@ -111,7 +111,7 @@ struct tipc_node *tipc_node_create(struct net *net, u32 addr)
 	INIT_LIST_HEAD(&n_ptr->list);
 	INIT_LIST_HEAD(&n_ptr->publ_list);
 	INIT_LIST_HEAD(&n_ptr->conn_sks);
-	__skb_queue_head_init(&n_ptr->bclink.deferred_queue);
+	__skb_queue_head_init(&n_ptr->bclink.deferdq);
 	hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
 	list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
 		if (n_ptr->addr < temp_node->addr)
@@ -354,7 +354,7 @@ static void node_lost_contact(struct tipc_node *n_ptr)
 
 	/* Flush broadcast link info associated with lost node */
 	if (n_ptr->bclink.recv_permitted) {
-		__skb_queue_purge(&n_ptr->bclink.deferred_queue);
+		__skb_queue_purge(&n_ptr->bclink.deferdq);
 
 		if (n_ptr->bclink.reasm_buf) {
 			kfree_skb(n_ptr->bclink.reasm_buf);

net/tipc/node.h

@@ -84,7 +84,7 @@ struct tipc_node_bclink {
 	u32 last_sent;
 	u32 oos_state;
 	u32 deferred_size;
-	struct sk_buff_head deferred_queue;
+	struct sk_buff_head deferdq;
 	struct sk_buff *reasm_buf;
 	int inputq_map;
 	bool recv_permitted;