dctcp: avoid bogus doubling of cwnd after loss
authorFlorian Westphal <fw@strlen.de>
Fri, 28 Oct 2016 16:43:11 +0000 (18:43 +0200)
committerDavid S. Miller <davem@davemloft.net>
Mon, 31 Oct 2016 19:16:28 +0000 (15:16 -0400)
commitce6dd23329b1ee6a794acf5f7e40f8e89b8317ee
tree133ffe80d56955a590a8f3d1dfad24e34f61aef6
parent19bda36c4299ce3d7e5bce10bebe01764a655a6d
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@fb.com>
Cc: Andrew Shewmaker <agshew@gmail.com>
Cc: Glenn Judd <glenn.judd@morganstanley.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_dctcp.c