Re: [PATCH v2 net-next 01/14] tcp: try to avoid safer when ACKs are thinned

[Paolo Abeni <pabeni@redhat.com>]

On 9/18/25 6:21 PM, chia-yu.chang@nokia-bell-labs.com wrote:
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index 9fdc6ce25eb1..0b25bf03ae6a 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -488,6 +488,10 @@ static void tcp_count_delivered(struct tcp_sock *tp, u32 delivered,
>  		tcp_count_delivered_ce(tp, delivered);
>  }
>  
> +#define PKTS_ACKED_WEIGHT	6
> +#define PKTS_ACKED_PREC		6
> +#define ACK_COMP_THRESH		4
> +
>  /* Returns the ECN CE delta */
>  static u32 __tcp_accecn_process(struct sock *sk, const struct sk_buff *skb,
>  				u32 delivered_pkts, u32 delivered_bytes,
> @@ -507,6 +511,19 @@ static u32 __tcp_accecn_process(struct sock *sk, const struct sk_buff *skb,
>  	opt_deltas_valid = tcp_accecn_process_option(tp, skb,
>  						     delivered_bytes, flag);
>  
> +	if (delivered_pkts) {
> +		if (!tp->pkts_acked_ewma) {
> +			tp->pkts_acked_ewma = delivered_pkts << PKTS_ACKED_PREC;

Can the above statement overflow pkts_acked_ewma in exceptional cases?
Do that need and additional min_t(u32, <init val>, 0xFFFFU)?

/P