From owner-freebsd-questions  Fri Apr  7 07:26:44 1995
Return-Path: questions-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id HAA08283
          for questions-outgoing; Fri, 7 Apr 1995 07:26:44 -0700
Received: from squid.umd.edu (squid.umd.edu [129.2.40.6])
          by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id HAA08277
          for <questions@FreeBSD.org>; Fri, 7 Apr 1995 07:26:35 -0700
Received: by squid.umd.edu (5.65/Ultrix3.0-C)
	id AA09044; Fri, 7 Apr 1995 10:34:06 -0400
From: fcawth@squid.umd.edu (Fred Cawthorne)
Message-Id: <9504071434.AA09044@squid.umd.edu>
Subject: Re: pppstats shows 5% packet error rate
To: obrien@antares.aero.org (Mike O'Brien)
Date: Fri, 7 Apr 95 10:34:06 EDT
Cc: questions@FreeBSD.org
In-Reply-To: <95Apr7.015514pdt.111113-2@aero.org>; from "Mike O'Brien" at Apr 7, 95 1:55 am
X-Mailer: ELM [version 2.3 PL11]
Sender: questions-owner@FreeBSD.org
Precedence: bulk

> 
> 	I'm running 2.0R, with the one-line PPP fix.
> 
>         My PPP connection appears to run as it should, but it also shows
> about a 5-10% packet error rate.  Running pppstats, I can see batches of 7
> or 8 errors happening at a time during, say, an inbound FTP.  During
> these periods the modem shows as mostly idle.  Possibly it's waiting for a
> TCP timeout/retransmission.  I am running a U.S.  Robotics Sportster
> modem, 14.4Kb, dialup into an Annex box.  I have 16550s in my Pentium 90.
> I'm running a Logitech bus mouse and XFree86 3.1.0 at the time.
> 
>         These errors and timeouts slow things down.  What's the use of
> spending the bucks to run at 28.8 if I'm only going to spend most of my
> time waiting for TCP (or maybe PPP) timeouts?
> 
I get the same sort of timeouts with my 28.8K modem telnetting through
an annex into a FreeBSD box.  The server side is FreeBSD2.0, and the
client is Snap950210.  It seems that sending something (like hitting CR
in a telnet session) causes it to stop waiting.  It also seems like
only things that send a large amount of data in a stream trigger the
problem.  I can run x clients with very little delays, for example.
I haven't tried ftp'ing a file to /dev/null either.  There may be
something lost somewhere when the disk is writing.  
I have tried going at 14.4K, turning off compression, etc... and it
does the same thing.  And I escape the ^S / ^Q  and the telnet 
escape character too.
My home machine is a 486/66 with 8 megs ram and an IDE drive running 
Snap 950210 and my work machine is a Pentium 90 with 32 megs ram and
scsi drives.  
>         Is anybody else seeing this?  Even suggestions of what to tear
> into would be welcome.  I am comfortable with banging on kernel source if
> I know where to start.
> 
I am going to check to see if anything has changed in pppd or the driver
since 2.0.   Perhaps a bug was fixed. (or introduced...)
I may also just try to install the latest snapshot at home to see if it
helps.


Fred.