netdev - [PATCH net] dctcp: avoid bogus doubling of cwnd after loss

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-next>] [day] [month] [year] [list]

Message-Id: <1477672991-19181-1-git-send-email-fw@strlen.de>
Date:   Fri, 28 Oct 2016 18:43:11 +0200
From:   Florian Westphal <fw@...len.de>
To:     <netdev@...r.kernel.org>
Cc:     daniel@...earbox.net, Florian Westphal <fw@...len.de>,
        Lawrence Brakmo <brakmo@...com>,
        Andrew Shewmaker <agshew@...il.com>,
        Glenn Judd <glenn.judd@...ganstanley.com>
Subject: [PATCH net] dctcp: avoid bogus doubling of cwnd after loss

If a congestion control module doesn't provide .undo_cwnd function,
tcp_undo_cwnd_reduction() will set cwnd to

   tp->snd_cwnd = max(tp->snd_cwnd, tp->snd_ssthresh << 1);

... which makes sense for reno (it sets ssthresh to half the current cwnd),
but it makes no sense for dctcp, which sets ssthresh based on the current
congestion estimate.

This can cause severe growth of cwnd (eventually overflowing u32).

Fix this by saving last cwnd on loss and restore cwnd based on that,
similar to cubic and other algorithms.

Fixes: e3118e8359bb7c ("net: tcp: add DCTCP congestion control algorithm")
Cc: Lawrence Brakmo <brakmo@...com>
Cc: Andrew Shewmaker <agshew@...il.com>
Cc: Glenn Judd <glenn.judd@...ganstanley.com>
Acked-by: Daniel Borkmann <daniel@...earbox.net>
Signed-off-by: Florian Westphal <fw@...len.de>
---
 net/ipv4/tcp_dctcp.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/tcp_dctcp.c b/net/ipv4/tcp_dctcp.c
index 10d728b6804c..ab37c6775630 100644
--- a/net/ipv4/tcp_dctcp.c
+++ b/net/ipv4/tcp_dctcp.c
@@ -56,6 +56,7 @@ struct dctcp {
 	u32 next_seq;
 	u32 ce_state;
 	u32 delayed_ack_reserved;
+	u32 loss_cwnd;
 };
 
 static unsigned int dctcp_shift_g __read_mostly = 4; /* g = 1/2^4 */
@@ -96,6 +97,7 @@ static void dctcp_init(struct sock *sk)
 		ca->dctcp_alpha = min(dctcp_alpha_on_init, DCTCP_MAX_ALPHA);
 
 		ca->delayed_ack_reserved = 0;
+		ca->loss_cwnd = 0;
 		ca->ce_state = 0;
 
 		dctcp_reset(tp, ca);
@@ -111,9 +113,10 @@ static void dctcp_init(struct sock *sk)
 
 static u32 dctcp_ssthresh(struct sock *sk)
 {
-	const struct dctcp *ca = inet_csk_ca(sk);
+	struct dctcp *ca = inet_csk_ca(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
 
+	ca->loss_cwnd = tp->snd_cwnd;
 	return max(tp->snd_cwnd - ((tp->snd_cwnd * ca->dctcp_alpha) >> 11U), 2U);
 }
 
@@ -308,12 +311,20 @@ static size_t dctcp_get_info(struct sock *sk, u32 ext, int *attr,
 	return 0;
 }
 
+static u32 dctcp_cwnd_undo(struct sock *sk)
+{
+	const struct dctcp *ca = inet_csk_ca(sk);
+
+	return max(tcp_sk(sk)->snd_cwnd, ca->loss_cwnd);
+}
+
 static struct tcp_congestion_ops dctcp __read_mostly = {
 	.init		= dctcp_init,
 	.in_ack_event   = dctcp_update_alpha,
 	.cwnd_event	= dctcp_cwnd_event,
 	.ssthresh	= dctcp_ssthresh,
 	.cong_avoid	= tcp_reno_cong_avoid,
+	.undo_cwnd	= dctcp_cwnd_undo,
 	.set_state	= dctcp_state,
 	.get_info	= dctcp_get_info,
 	.flags		= TCP_CONG_NEEDS_ECN,
-- 
2.7.3