net: sk_buff rbnode reorg

skb->rbnode shares space with skb->next, skb->prev and skb->tstamp

Current uses (TCP receive ofo queue and netem) need to save/restore
tstamp, while skb->dev is either NULL (TCP) or a constant for a given
queue (netem).

Since we plan using an RB tree for TCP retransmit queue to speedup SACK
processing with large BDP, this patch exchanges skb->dev and
skb->tstamp.

This saves some overhead in both TCP and netem.

v2: removes the swtstamp field from struct tcp_skb_cb

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Cc: Wei Wang <weiwan@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 9 insertions, 25 deletions

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index bddf724..db9bb46 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -4266,11 +4266,6 @@ static void tcp_sack_remove(struct tcp_sock *tp)
 	tp->rx_opt.num_sacks = num_sacks;
 }
 
-enum tcp_queue {
-	OOO_QUEUE,
-	RCV_QUEUE,
-};
-
 /**
  * tcp_try_coalesce - try to merge skb to prior one
  * @sk: socket
@@ -4286,7 +4281,6 @@ enum tcp_queue {
  * Returns true if caller should free @from instead of queueing it
  */
 static bool tcp_try_coalesce(struct sock *sk,
-			     enum tcp_queue dest,
 			     struct sk_buff *to,
 			     struct sk_buff *from,
 			     bool *fragstolen)
@@ -4311,10 +4305,7 @@ static bool tcp_try_coalesce(struct sock *sk,
 
 	if (TCP_SKB_CB(from)->has_rxtstamp) {
 		TCP_SKB_CB(to)->has_rxtstamp = true;
-		if (dest == OOO_QUEUE)
-			TCP_SKB_CB(to)->swtstamp = TCP_SKB_CB(from)->swtstamp;
-		else
-			to->tstamp = from->tstamp;
+		to->tstamp = from->tstamp;
 	}
 
 	return true;
@@ -4351,9 +4342,6 @@ static void tcp_ofo_queue(struct sock *sk)
 		}
 		p = rb_next(p);
 		rb_erase(&skb->rbnode, &tp->out_of_order_queue);
-		/* Replace tstamp which was stomped by rbnode */
-		if (TCP_SKB_CB(skb)->has_rxtstamp)
-			skb->tstamp = TCP_SKB_CB(skb)->swtstamp;
 
 		if (unlikely(!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))) {
 			SOCK_DEBUG(sk, "ofo packet was already received\n");
@@ -4365,8 +4353,7 @@ static void tcp_ofo_queue(struct sock *sk)
 			   TCP_SKB_CB(skb)->end_seq);
 
 		tail = skb_peek_tail(&sk->sk_receive_queue);
-		eaten = tail && tcp_try_coalesce(sk, RCV_QUEUE,
-						 tail, skb, &fragstolen);
+		eaten = tail && tcp_try_coalesce(sk, tail, skb, &fragstolen);
 		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
 		fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN;
 		if (!eaten)
@@ -4420,10 +4407,6 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
 		return;
 	}
 
-	/* Stash tstamp to avoid being stomped on by rbnode */
-	if (TCP_SKB_CB(skb)->has_rxtstamp)
-		TCP_SKB_CB(skb)->swtstamp = skb->tstamp;
-
 	/* Disable header prediction. */
 	tp->pred_flags = 0;
 	inet_csk_schedule_ack(sk);
@@ -4451,7 +4434,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
 	/* In the typical case, we are adding an skb to the end of the list.
 	 * Use of ooo_last_skb avoids the O(Log(N)) rbtree lookup.
 	 */
-	if (tcp_try_coalesce(sk, OOO_QUEUE, tp->ooo_last_skb,
+	if (tcp_try_coalesce(sk, tp->ooo_last_skb,
 			     skb, &fragstolen)) {
 coalesce_done:
 		tcp_grow_window(sk, skb);
@@ -4502,7 +4485,7 @@ coalesce_done:
 				__kfree_skb(skb1);
 				goto merge_right;
 			}
-		} else if (tcp_try_coalesce(sk, OOO_QUEUE, skb1,
+		} else if (tcp_try_coalesce(sk, skb1,
 					    skb, &fragstolen)) {
 			goto coalesce_done;
 		}
@@ -4554,7 +4537,7 @@ static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int
 
 	__skb_pull(skb, hdrlen);
 	eaten = (tail &&
-		 tcp_try_coalesce(sk, RCV_QUEUE, tail,
+		 tcp_try_coalesce(sk, tail,
 				  skb, fragstolen)) ? 1 : 0;
 	tcp_rcv_nxt_update(tcp_sk(sk), TCP_SKB_CB(skb)->end_seq);
 	if (!eaten) {
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index b1266e7..063a4bd 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -146,7 +146,6 @@ struct netem_sched_data {
  */
 struct netem_skb_cb {
 	psched_time_t	time_to_send;
-	ktime_t		tstamp_save;
 };
 
 
@@ -561,7 +560,6 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		}
 
 		cb->time_to_send = now + delay;
-		cb->tstamp_save = skb->tstamp;
 		++q->counter;
 		tfifo_enqueue(skb, sch);
 	} else {
@@ -629,7 +627,10 @@ deliver:
 			qdisc_qstats_backlog_dec(sch, skb);
 			skb->next = NULL;
 			skb->prev = NULL;
-			skb->tstamp = netem_skb_cb(skb)->tstamp_save;
+			/* skb->dev shares skb->rbnode area,
+			 * we need to restore its value.
+			 */
+			skb->dev = qdisc_dev(sch);
 
 #ifdef CONFIG_NET_CLS_ACT
 			/*