Commit 179ac35f authored Sep 25, 2020 by Yuchung Cheng Committed by David S. Miller Sep 25, 2020

tcp: consistently check retransmit hint



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>

parent b4f43483

net/ipv4/tcp_input.c

+2 −7

Original line number	Diff line number	Diff line
		@@ -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);

net/ipv4/tcp_recovery.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -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);
		}
		}