netdev - Re: [PATCH net-next v4 09/15] quic: add congestion control

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <43ea4062-75a8-4152-bf19-2eca561036bd@redhat.com>
Date: Tue, 4 Nov 2025 13:02:00 +0100
From: Paolo Abeni <pabeni@...hat.com>
To: Xin Long <lucien.xin@...il.com>, network dev <netdev@...r.kernel.org>,
 quic@...ts.linux.dev
Cc: davem@...emloft.net, kuba@...nel.org, Eric Dumazet <edumazet@...gle.com>,
 Simon Horman <horms@...nel.org>, Stefan Metzmacher <metze@...ba.org>,
 Moritz Buhl <mbuhl@...nbsd.org>, Tyler Fanelli <tfanelli@...hat.com>,
 Pengtao He <hepengtao@...omi.com>, Thomas Dreibholz <dreibh@...ula.no>,
 linux-cifs@...r.kernel.org, Steve French <smfrench@...il.com>,
 Namjae Jeon <linkinjeon@...nel.org>, Paulo Alcantara <pc@...guebit.com>,
 Tom Talpey <tom@...pey.com>, kernel-tls-handshake@...ts.linux.dev,
 Chuck Lever <chuck.lever@...cle.com>, Jeff Layton <jlayton@...nel.org>,
 Benjamin Coddington <bcodding@...hat.com>, Steve Dickson
 <steved@...hat.com>, Hannes Reinecke <hare@...e.de>,
 Alexander Aring <aahringo@...hat.com>, David Howells <dhowells@...hat.com>,
 Matthieu Baerts <matttbe@...nel.org>, John Ericson <mail@...nericson.me>,
 Cong Wang <xiyou.wangcong@...il.com>, "D . Wythe"
 <alibuda@...ux.alibaba.com>, Jason Baron <jbaron@...mai.com>,
 illiliti <illiliti@...tonmail.com>, Sabrina Dubroca <sd@...asysnail.net>,
 Marcelo Ricardo Leitner <marcelo.leitner@...il.com>,
 Daniel Stenberg <daniel@...x.se>,
 Andy Gospodarek <andrew.gospodarek@...adcom.com>
Subject: Re: [PATCH net-next v4 09/15] quic: add congestion control

On 10/29/25 3:35 PM, Xin Long wrote:
> +/* Compute and update the pacing rate based on congestion window and smoothed RTT. */
> +static void quic_cong_pace_update(struct quic_cong *cong, u32 bytes, u32 max_rate)
> +{
> +	u64 rate;
> +
> +	/* rate = N * congestion_window / smoothed_rtt */
> +	rate = (u64)cong->window * USEC_PER_SEC * 2;
> +	if (likely(cong->smoothed_rtt))
> +		rate = div64_ul(rate, cong->smoothed_rtt);
> +
> +	WRITE_ONCE(cong->pacing_rate, min_t(u64, rate, max_rate));
> +	pr_debug("%s: update pacing rate: %u, max rate: %u, srtt: %u\n",
> +		 __func__, cong->pacing_rate, max_rate, cong->smoothed_rtt);

I think you should skip entirely the pacing_rate update when
`smoothed_rtt == 0`

[...]> +/* rfc9002#section-5: Estimating the Round-Trip Time */
> +void quic_cong_rtt_update(struct quic_cong *cong, u32 time, u32 ack_delay)
> +{
> +	u32 adjusted_rtt, rttvar_sample;
> +
> +	/* Ignore RTT sample if ACK delay is suspiciously large. */
> +	if (ack_delay > cong->max_ack_delay * 2)
> +		return;
> +
> +	/* rfc9002#section-5.1: latest_rtt = ack_time - send_time_of_largest_acked */
> +	cong->latest_rtt = cong->time - time;
> +
> +	/* rfc9002#section-5.2: Estimating min_rtt */
> +	if (!cong->min_rtt_valid) {
> +		cong->min_rtt = cong->latest_rtt;
> +		cong->min_rtt_valid = 1;
> +	}
> +	if (cong->min_rtt > cong->latest_rtt)
> +		cong->min_rtt = cong->latest_rtt;
> +
> +	if (!cong->is_rtt_set) {
> +		/* rfc9002#section-5.3:
> +		 *   smoothed_rtt = latest_rtt
> +		 *   rttvar = latest_rtt / 2
> +		 */
> +		cong->smoothed_rtt = cong->latest_rtt;
> +		cong->rttvar = cong->smoothed_rtt / 2;
> +		quic_cong_pto_update(cong);
> +		cong->is_rtt_set = 1;
> +		return;
> +	}
> +
> +	/* rfc9002#section-5.3:
> +	 *   adjusted_rtt = latest_rtt
> +	 *   if (latest_rtt >= min_rtt + ack_delay):
> +	 *     adjusted_rtt = latest_rtt - ack_delay
> +	 *   smoothed_rtt = 7/8 * smoothed_rtt + 1/8 * adjusted_rtt
> +	 *   rttvar_sample = abs(smoothed_rtt - adjusted_rtt)
> +	 *   rttvar = 3/4 * rttvar + 1/4 * rttvar_sample
> +	 */
> +	adjusted_rtt = cong->latest_rtt;
> +	if (cong->latest_rtt >= cong->min_rtt + ack_delay)
> +		adjusted_rtt = cong->latest_rtt - ack_delay;
> +
> +	cong->smoothed_rtt = (cong->smoothed_rtt * 7 + adjusted_rtt) / 8;

Out of sheer curiosity, is the compiler smart enough to use a 'srl 3'
for the above?

/P