[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-id: <63599554-7243-4E44-BFEF-5B4B36743606@nuim.ie>
Date: Wed, 12 Nov 2008 08:58:18 +0000
From: Douglas Leith <Doug.Leith@...m.ie>
To: David Miller <davem@...emloft.net>
Cc: Netdev <netdev@...r.kernel.org>,
Stephen Hemminger <shemminger@...tta.com>
Subject: Re: [PATCH 2.6.27] tcp_htcp: last_cong bug fix
This patch fixes a minor bug in tcp_htcp.c which has been
highlighted by Lachlan Andrew and Lawrence Stewart. Currently, the
time since the last congestion event, which is stored in variable
last_cong, is reset whenever there is a state change into
TCP_CA_Open. This includes transitions of the type
TCP_CA_Open->TCP_CA_Disorder->TCP_CA_Open which are not associated
with backoff of cwnd. The patch changes last_cong to be updated
only on transitions into TCP_CA_Open that occur after experiencing
the congestion-related states TCP_CA_Loss, TCP_CA_Recovery,
TCP_CA_CWR.
Signed-off-by: Doug Leith <doug.leith@...m.ie>
---
Resending with patch as attachment. Hope this does the trick.
--- net/ipv4/tcp_htcp.c 2008-11-05 15:50:18.000000000 +0000
+++ net/ipv4/tcp_htcp.c.new 2008-11-11 15:20:31.000000000 +0000
@@ -71,5 +71,8 @@ static u32 htcp_cwnd_undo(struct sock *s
- ca->last_cong = ca->undo_last_cong;
- ca->maxRTT = ca->undo_maxRTT;
- ca->old_maxB = ca->undo_old_maxB;
+ if (ca->undo_last_cong) {
+ ca->last_cong = ca->undo_last_cong;
+ ca->maxRTT = ca->undo_maxRTT;
+ ca->old_maxB = ca->undo_old_maxB;
+ ca->undo_last_cong = 0;
+ }
@@ -267,8 +270,11 @@ static void htcp_state(struct sock *sk,
switch (new_state) {
- case TCP_CA_Open:
+ case TCP_CA_Open:
{
struct htcp *ca = inet_csk_ca(sk);
- ca->last_cong = jiffies;
+ if (ca->undo_last_cong) {
+ ca->last_cong = jiffies;
+ ca->undo_last_cong = 0;
+ }
}
- break;
+ break;
case TCP_CA_CWR:
Download attachment "patch.diff" of type "application/octet-stream" (813 bytes)
Powered by blists - more mailing lists