Path: utzoo!attcan!uunet!lll-winken!ncis.llnl.gov!helios.ee.lbl.gov!pasteur!ucbvax!decwrl!labrea!rutgers!rochester!pt.cs.cmu.edu!k.gp.cs.cmu.edu!dpm
From: dpm@k.gp.cs.cmu.edu (David Maynard)
Newsgroups: comp.protocols.tcp-ip
Subject: SLIP problems
Message-ID: <4094@pt.cs.cmu.edu>
Date: 23 Jan 89 18:50:01 GMT
Reply-To: dpm@k.gp.cs.cmu.edu (David Maynard)
Organization: Carnegie-Mellon University, CS/RI
Lines: 40
Keywords:

I am trying to set up a SLIP connection between two machines.  I had to
slightly modify the protocol (just quoting more characters) because the line
passes through a terminal concentrator that traps some characters.  I am
also unable to use hardware flow control because of the concentrator.

I have things working to the point that 'ping' seems to work in both
directions.  Telnet from host-a to host-b works fairly reliably (hangs very
occasionally).  However, other programs (and telnet in the opposite
direction) hang sometime after the connection is established.  When I look
at netstat on host-a, the hung connections have characters waiting in the
Send-Q.  On host-b, the queues are empty; however, "netstat -s" reports that
the number of ip packets with "data size < data length" has increased.  Even
after these connections hang, I can still ping host-b and establish new
telnet connections from a to b (or continue to use existing telnet
connections).  

I am assuming that some kind of flow control problem is preventing certain
types of packets from being received properly on host-b.  However, I don't
know enough about tcp-ip to know what types of packets to suspect.  I would
suspect long ones, but some of the connections hang with only 2 characters
in the Send-Q.  I also tried reducing the SLMTU for the line to 128 with no
noticeable improvement.  

I would appreciate any hints as to where the problem might be.  I would also
appreciate possible workarounds if I can't find a real solution.  

Host-a is a Sun-2/120 running SunOS3.5.  Host-b is a Sun-3/75 also running
3.5.  Both hosts are running Van Jacobson's tcp software.  

Thanks,
David
 ---
 David P. Maynard (dpm@cs.cmu.edu)
 Dept. of Electrical and Computer Engineering
 Carnegie Mellon University
 Pittsburgh, PA  15213
 ---
 Any opinions expressed are mine only.  I haven't asked the ECE department
 or CMU what they think.
 ---
--