OSDN Git Service

tcp: consistently check retransmit hint
authorYuchung Cheng <ycheng@google.com>
Fri, 25 Sep 2020 17:04:28 +0000 (10:04 -0700)
committerDavid S. Miller <davem@davemloft.net>
Sat, 26 Sep 2020 00:17:14 +0000 (17:17 -0700)
tcp_simple_retransmit() used for path MTU discovery may not adjust
the retransmit hint properly by deducting retrans_out before checking
it to adjust the hint. This patch fixes this by a correct routine
tcp_mark_skb_lost() already used by the RACK loss detection.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_input.c
net/ipv4/tcp_recovery.c

index 7489808..f9f4321 100644 (file)
@@ -2675,13 +2675,8 @@ void tcp_simple_retransmit(struct sock *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);
-               }
+                   !(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
+                       tcp_mark_skb_lost(sk, skb);
        }
 
        tcp_clear_retrans_hints_partial(tp);
index fdb715b..26a4228 100644 (file)
@@ -246,6 +246,6 @@ void tcp_newreno_mark_lost(struct sock *sk, bool snd_una_advanced)
                        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);
        }
 }