netdev - Re: [PATCH 2/2] [PATCH net-next,2/2] Add test for the use of new args in cong

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: <6e134365-b793-49f9-82dd-d148bdd16c85@linux.dev>
Date: Thu, 25 Apr 2024 12:09:57 -0700
From: Martin KaFai Lau <martin.lau@...ux.dev>
To: Miao Xu <miaxu@...a.com>
Cc: Eric Dumazet <edumazet@...gle.com>, "David S . Miller"
 <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>,
 Paolo Abeni <pabeni@...hat.com>, David Ahern <dsahern@...nel.org>,
 Martin Lau <kafai@...a.com>, netdev@...r.kernel.org, bpf@...r.kernel.org
Subject: Re: [PATCH 2/2] [PATCH net-next,2/2] Add test for the use of new args
 in cong_control

On 4/24/24 1:37 PM, Miao Xu wrote:
> +#define USEC_PER_SEC 1000000UL
> +#define TCP_PACING_SS_RATIO (200)
> +#define TCP_PACING_CA_RATIO (120)
> +#define TCP_REORDERING (12)
> +#define likely(x) (__builtin_expect(!!(x), 1))
> +
> +static __always_inline __u64 div64_u64(__u64 dividend, __u64 divisor)
> +{
> +	return dividend / divisor;
> +}
> +
> +static void tcp_update_pacing_rate(struct sock *sk)
> +{
> +	const struct tcp_sock *tp = tcp_sk(sk);
> +	__u64 rate;
> +
> +	/* set sk_pacing_rate to 200 % of current rate (mss * cwnd / srtt) */
> +	rate = (__u64)tp->mss_cache * ((USEC_PER_SEC / 100) << 3);
> +
> +	/* current rate is (cwnd * mss) / srtt
> +	 * In Slow Start [1], set sk_pacing_rate to 200 % the current rate.
> +	 * In Congestion Avoidance phase, set it to 120 % the current rate.
> +	 *
> +	 * [1] : Normal Slow Start condition is (tp->snd_cwnd < tp->snd_ssthresh)
> +	 *	 If snd_cwnd >= (tp->snd_ssthresh / 2), we are approaching
> +	 *	 end of slow start and should slow down.
> +	 */
> +	if (tp->snd_cwnd < tp->snd_ssthresh / 2)
> +		rate *= TCP_PACING_SS_RATIO;
> +	else
> +		rate *= TCP_PACING_CA_RATIO;
> +
> +	rate *= max(tp->snd_cwnd, tp->packets_out);
> +
> +	if (likely(tp->srtt_us))
> +		rate = div64_u64(rate, (__u64)tp->srtt_us);
> +
> +	sk->sk_pacing_rate = min(rate, (__u64)sk->sk_max_pacing_rate);
> +}
> +
> +static __always_inline void tcp_cwnd_reduction(
> +		struct sock *sk,
> +		int newly_acked_sacked,
> +		int newly_lost,
> +		int flag) {
> +	struct tcp_sock *tp = tcp_sk(sk);
> +	int sndcnt = 0;
> +	__u32 pkts_in_flight = tp->packets_out - (tp->sacked_out + tp->lost_out) + tp->retrans_out;
> +	int delta = tp->snd_ssthresh - pkts_in_flight;
> +
> +	if (newly_acked_sacked <= 0 || !tp->prior_cwnd)
> +		return;
> +
> +	__u32 prr_delivered = tp->prr_delivered + newly_acked_sacked;
> +
> +	if (delta < 0) {
> +		__u64 dividend =
> +			(__u64)tp->snd_ssthresh * prr_delivered + tp->prior_cwnd - 1;
> +		sndcnt = (__u32)div64_u64(dividend, (__u64)tp->prior_cwnd) - tp->prr_out;
> +	} else {
> +		sndcnt = max(prr_delivered - tp->prr_out, newly_acked_sacked);
> +		if (flag & FLAG_SND_UNA_ADVANCED && !newly_lost)
> +			sndcnt++;
> +		sndcnt = min(delta, sndcnt);
> +	}
> +	/* Force a fast retransmit upon entering fast recovery */
> +	sndcnt = max(sndcnt, (tp->prr_out ? 0 : 1));
> +	tp->snd_cwnd = pkts_in_flight + sndcnt;
> +}
> +
> +/* Decide wheather to run the increase function of congestion control. */
> +static __always_inline bool tcp_may_raise_cwnd(
> +		const struct sock *sk,
> +		const int flag) {
> +	if (tcp_sk(sk)->reordering > TCP_REORDERING)
> +		return flag & FLAG_FORWARD_PROGRESS;
> +
> +	return flag & FLAG_DATA_ACKED;
> +}
> +
> +void BPF_STRUCT_OPS(bpf_cubic_cong_control, struct sock *sk, __u32 ack, int flag,
> +		const struct rate_sample *rs)
> +{
> +	struct tcp_sock *tp = tcp_sk(sk);
> +
> +	if (((1<<TCP_CA_CWR) | (1<<TCP_CA_Recovery)) &
> +			(1 << inet_csk(sk)->icsk_ca_state)) {
> +		/* Reduce cwnd if state mandates */
> +		tcp_cwnd_reduction(sk, rs->acked_sacked, rs->losses, flag);
> +
> +		if (!before(tp->snd_una, tp->high_seq)) {
> +			/* Reset cwnd to ssthresh in CWR or Recovery (unless it's undone) */
> +			if (tp->snd_ssthresh < TCP_INFINITE_SSTHRESH &&
> +					inet_csk(sk)->icsk_ca_state == TCP_CA_CWR) {
> +				tp->snd_cwnd = tp->snd_ssthresh;
> +				tp->snd_cwnd_stamp = tcp_jiffies32;
> +			}
> +			// __cwnd_event(sk, CA_EVENT_COMPLETE_CWR);
> +		}
> +	} else if (tcp_may_raise_cwnd(sk, flag)) {
> +		/* Advance cwnd if state allows */
> +		cubictcp_cong_avoid(sk, ack, rs->acked_sacked);
> +		tp->snd_cwnd_stamp = tcp_jiffies32;
> +	}
> +
> +	tcp_update_pacing_rate(sk);

It will be useful to highlight what you want to do differently from the kernel's 
tcp_cong_control()+tcp_cong_avoid() here. or it is something that I missed from 
the above example?