From owner-freebsd-bugs Fri Apr 28 11:30:58 1995 Return-Path: bugs-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id LAA20444 for bugs-outgoing; Fri, 28 Apr 1995 11:30:58 -0700 Received: from hp.com (hp.com [15.255.152.4]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id LAA20438 for ; Fri, 28 Apr 1995 11:30:50 -0700 Received: from hpisrdq.cup.hp.com (hpindio.cup.hp.com) by hp.com with SMTP (1.37.109.15/15.5+ECS 3.3) id AA074623845; Fri, 28 Apr 1995 11:30:45 -0700 Received: by hpisrdq.cup.hp.com with SMTP (1.38.193.4/15.5+IOS 3.20+cup+OMrelay) id AA05655; Fri, 28 Apr 1995 11:27:30 -0700 Message-Id: <9504281827.AA05655@hpisrdq.cup.hp.com> To: bugs@FreeBSD.org Subject: Possible bug in the 4/12 TCP Date: Fri, 28 Apr 1995 11:27:29 -0700 From: Rick Jones Sender: bugs-owner@FreeBSD.org Precedence: bulk Here is the original message, with my typo... rick jones To: Andras Olah Cc: rstevens@noao.edu (Richard Stevens), bugs@frteebsd.org Subject: Bug?: immediate ACK's in 4/12 TCP... Date: Fri, 28 Apr 1995 11:23:37 -0700 From: Rick Jones Andras - Thanks for the copy of tclient and tserver. They confirm something that I have noticed with netperf (which I have working again with minor tweaks to netperf.c and netserver.c so that a TCP_TRR test calls the tran routines and not the conn routines - what's RCS anyway :) There is (what I think is) a bug in the 4/12 TCP, of unknown origin. It is immediately ACKing *all* traffic. I have analyser traces of netperf TCP_RR, TCP_TRR, TCP_CRR, and TCP_STREAM tests which show this behaviour. I'm going to try to find time to look through the source to see if I can find the bug. I think I'll also ask bugs@freebsd.org if any other changes were added to TCP for the 4/12 snapshot to try and narrow my search. Bugs - what were the changes in the TCP code between the 3/22 and 4/12 snapshots? For TCP_RR, it means a 2x increase in packets per transaction, for TCP_TRR, things go from three packets per transaction to five. For a TCP_STREAM test, it is a 33% increase in packets. I suspect that it can be seen with just about *any* TCP connection - telnet, ftp, etc...I expect to have the 2.0PL1 netperf on ftp.cup.hp.com by Monday. later, rick ------- =_aaaaaaaaaa--