From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stephen Hemminger Subject: Re: [PATCH net] tcp: disallow cwnd undo when switching congestion control Date: Wed, 31 May 2017 11:47:32 -0700 Message-ID: <20170531114732.1cfc4d79@xeon-e3> References: <20170531182127.161289-1-ycheng@google.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: davem@davemloft.net, netdev@vger.kernel.org, ncardwell@google.com, edumazet@google.com, soheil@google.com To: Yuchung Cheng Return-path: Received: from mail-pf0-f180.google.com ([209.85.192.180]:35687 "EHLO mail-pf0-f180.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751100AbdEaSrf (ORCPT ); Wed, 31 May 2017 14:47:35 -0400 Received: by mail-pf0-f180.google.com with SMTP id n23so14910984pfb.2 for ; Wed, 31 May 2017 11:47:35 -0700 (PDT) In-Reply-To: <20170531182127.161289-1-ycheng@google.com> Sender: netdev-owner@vger.kernel.org List-ID: On Wed, 31 May 2017 11:21:27 -0700 Yuchung Cheng wrote: > When the sender switches its congestion control during loss > recovery, if the recovery is spurious then it may incorrectly > revert cwnd and ssthresh to the older values set by a previous > congestion control. Consider a congestion control (like BBR) > that does not use ssthresh and keeps it infinite: the connection > may incorrectly revert cwnd to an infinite value when switching > from BBR to another congestion control. > > This patch fixes it by disallowing such cwnd undo operation > upon switching congestion control. Note that undo_marker > is not reset s.t. the packets that were incorrectly marked > lost would be corrected. We only avoid undoing the cwnd in > tcp_undo_cwnd_reduction(). > > Signed-off-by: Yuchung Cheng > Signed-off-by: Soheil Hassas Yeganeh > Signed-off-by: Neal Cardwell > Signed-off-by: Eric Dumazet That looks correct. Are there other values of congestion state that should be reset?