Merge branch 'simplify-TCP-loss-marking-code'

Yuchung Cheng says:

====================
simplify TCP loss marking code

The TCP loss marking is implemented by a set of intertwined
subroutines. TCP has several loss detection algorithms
(RACK, RFC6675/FACK, NewReno, etc) each calls a subset of
these routines to mark a packet lost. This has led to
various bugs (and fixes and fixes of fixes).

This patch set is to consolidate the loss marking code so
all detection algorithms call the same routine tcp_mark_skb_lost().
====================

Signed-off-by: David S. Miller <[e-mail]>

net/ipv4/tcp_input.c

···
 		      ts ? LINUX_MIB_TCPTSREORDER : LINUX_MIB_TCPSACKREORDER);
 }
 
-/* This must be called before lost_out is incremented */
+ /* This must be called before lost_out or retrans_out are updated
+  * on a new loss, because we want to know if all skbs previously
+  * known to be lost have already been retransmitted, indicating
+  * that this newly lost skb is our next skb to retransmit.
+  */
 static void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb)
 {
 	if ((!tp->retransmit_skb_hint && tp->retrans_out >= tp->lost_out) ||
···
 		tp->retransmit_skb_hint = skb;
 }
 
-/* Sum the number of packets on the wire we have marked as lost.
- * There are two cases we care about here:
- * a) Packet hasn't been marked lost (nor retransmitted),
- *    and this is the first loss.
- * b) Packet has been marked both lost and retransmitted,
- *    and this means we think it was lost again.
- */
-static void tcp_sum_lost(struct tcp_sock *tp, struct sk_buff *skb)
+void tcp_mark_skb_lost(struct sock *sk, struct sk_buff *skb)
 {
 	__u8 sacked = TCP_SKB_CB(skb)->sacked;
+	struct tcp_sock *tp = tcp_sk(sk);
 
-	if (!(sacked & TCPCB_LOST) ||
-	    ((sacked & TCPCB_LOST) && (sacked & TCPCB_SACKED_RETRANS)))
-		tp->lost += tcp_skb_pcount(skb);
-}
+	if (sacked & TCPCB_SACKED_ACKED)
+		return;
 
-static void tcp_skb_mark_lost(struct tcp_sock *tp, struct sk_buff *skb)
-{
-	if (!(TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_ACKED))) {
-		tcp_verify_retransmit_hint(tp, skb);
-
-		tp->lost_out += tcp_skb_pcount(skb);
-		tcp_sum_lost(tp, skb);
-		TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
-	}
-}
-
-void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb)
-{
 	tcp_verify_retransmit_hint(tp, skb);
-
-	tcp_sum_lost(tp, skb);
-	if (!(TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_ACKED))) {
+	if (sacked & TCPCB_LOST) {
+		if (sacked & TCPCB_SACKED_RETRANS) {
+			/* Account for retransmits that are lost again */
+			TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
+			tp->retrans_out -= tcp_skb_pcount(skb);
+			NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPLOSTRETRANSMIT,
+				      tcp_skb_pcount(skb));
+		}
+	} else {
 		tp->lost_out += tcp_skb_pcount(skb);
 		TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
 	}
···
 		if (cnt > packets)
 			break;
 
-		tcp_skb_mark_lost(tp, skb);
+		if (!(TCP_SKB_CB(skb)->sacked & TCPCB_LOST))
+			tcp_mark_skb_lost(sk, skb);
 
 		if (mark_head)
 			break;
···
 	unsigned int mss = tcp_current_mss(sk);
 
 	skb_rbtree_walk(skb, &sk->tcp_rtx_queue) {
-		if (tcp_skb_seglen(skb) > mss &&
-		    !(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)) {
-			if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
-				TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
-				tp->retrans_out -= tcp_skb_pcount(skb);
-			}
-			tcp_skb_mark_lost_uncond_verify(tp, skb);
-		}
+		if (tcp_skb_seglen(skb) > mss)
+			tcp_mark_skb_lost(sk, skb);
 	}
 
 	tcp_clear_retrans_hints_partial(tp);

net/ipv4/tcp_recovery.c

···
 #include <linux/tcp.h>
 #include <net/tcp.h>
 
-void tcp_mark_skb_lost(struct sock *sk, struct sk_buff *skb)
-{
-	struct tcp_sock *tp = tcp_sk(sk);
-
-	tcp_skb_mark_lost_uncond_verify(tp, skb);
-	if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
-		/* Account for retransmits that are lost again */
-		TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
-		tp->retrans_out -= tcp_skb_pcount(skb);
-		NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPLOSTRETRANSMIT,
-			      tcp_skb_pcount(skb));
-	}
-}
-
 static bool tcp_rack_sent_after(u64 t1, u64 t2, u32 seq1, u32 seq2)
 {
 	return t1 > t2 || (t1 == t2 && after(seq1, seq2));
···
 			tcp_fragment(sk, TCP_FRAG_IN_RTX_QUEUE, skb,
 				     mss, mss, GFP_ATOMIC);
 
-		tcp_skb_mark_lost_uncond_verify(tp, skb);
+		tcp_mark_skb_lost(sk, skb);
 	}
 }