Re: Weird tcp performance differences with 2.0 and 2.2 kernels

Raul Miller (rdm@test.legislate.com)
Fri, 12 Feb 1999 12:12:10 -0500


I spent quite a bit of time working tracking down an instance of this
problem, and in my case it was not NetBSD (as I originally thought)
but a router.

-- 
Raul

David Miller <davem@twiddle.net> wrote: > Date: Fri, 12 Feb 1999 16:29:55 +0100 (CET) > From: Mike Galbraith <mikeg@weiden.de> > > It's the same either on or off. This is something that snuck in > fairly recently. One day I was trying to get to ftp.x.org, and > this happened. I figured they :) were having troubles and ignored > it. When 'their' problem didn't go away in a couple of days, I got > suspicious. When the same exact failure happened with porcupine, I > remembered hearing about /proc/sys/net/ipv4/tcp_timestamps and gave > it a shot. This snuck in after a kernel upgrade, not an isdn > upgrade fwiw. > > It's a bug in the other end. > > How do I know this? :-) Because now that you mention ftp.x.org I have > seen this problem too and watched the traces carefully. What I saw is > that with timestamps enabled, if there is a dropped packet and either > a fast or normal retransmit happena, ftp.x.org backs off for a _long_ > time and takes forever to retransmit. > > I think ftp.x.org need to apply any updates necessary to their (what > appears to be a) BSD derived system. It's definately a bug in the > sender (ftp.x.org) and not in Linux-2.2.x > > Later, > David S. Miller > davem@redhat.com > > - > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.rutgers.edu > Please read the FAQ at http://www.tux.org/lkml/

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/