Patch-ID# 101673-01
Keywords: rsh, inetd, timeout, hangs
Synopsis: SunOS 4.1.3 Point Patch: rsh hangs, talking to a heavily loaded machine
Date: Apr/01/94

Solaris Release: 

SunOS Release: 4.1.3, 4.1.3C

Unbundled Product: 

Unbundled Release: 

Relevant Architectures: sparc

BugId's fixed with this patch: 1130726

Changes incorporated in this version: 1130726

Patches accumulated and obsoleted by this patch: 

Patches which conflict with this patch: 

Patches required with this patch: 

Obsoleted by: 

Files included with this patch: 

/usr/ucb/rsh

Problem Description: 

rsh fails intermintently attempting to connect to a heavily loaded machine.
The problem is actually caused by inetd hanging on the machine being
connected to. 

This patch is a point patch that provides an additional argument "-T num" to
rsh. This argument specifies a number of seconds after which that rsh will 
timeout attempting the connection. Upon timeout, rsh will exit with an
exit code of 3.

This point patch is provided as a work-around to prevent rsh from hanging, 
a situation which is primarily caused by "inetd" failure to keep up with rate 
of requests it receives and therefore shutting down the service. 

RSH/INETD PROBLEM DESCRIPTION:
-----------------------------

There are 2 problems related to inetd, and rsh:

1. rate of request high water mark is too low for new machine speeds.
   This problem is fixed with inetd point patch No. 100178-09 for 4.x machines.
   This patch alone should fix the problem when the inetd server is a 
   Sun machine.

2. The second problem relates to both rsh and inetd. 
   The problem with rsh side is that it does blocking I/O instead of 
   non-blocking. 
   The problem with inetd side is that it does not close the last connection
   which caused it to shut down the service.
   As a result of these 2 the rsh on the client side hangs.
   A solution to this problem is provided by PDE. This solution is only
   effective when both sides are Sun machines, in which case solution to
   problem 1 alone can fix the problem.


Therefore, to solve the "rsh hang" problem one should first check the inetd
server side.
If the server side is a Sun machine, provide patch 100178-09 (it's important 
to use this latest version, as the 100178-08 although providing the limit 
settings as command line options, but it has a problem with number of file 
descriptors that it can open which is fixed in 09 version) and if necessary 
by using:
	
		inetd -r num-of-connections num-of-seconds

adjust the high rate of requests num-of-connections/num-of-seconds which tunes
the inetd for the specific server CPU speed.

RSH (ONLY) POINT PATCH:
----------------------

Description:
-----------
  This rsh allows for a time out for the calls to the remote host and is
useful when the inetd of the non-Sun server shuts down the service. In which
case the rsh will time out instead of hanging.  Checking the exit value 3 of 
this rsh is an indicator of inetd server service shutdown.

  If the server side is not a Sun machine, use this point patch rsh for 4.x
machines. 

Life Time:
---------

The rsh/inetd problem is fixed and integrated in 494. Therefore, there 
wouldn't be a need for this work around rsh patch under future Solaris
release, unless similar situation is reported by a customer that would 
necessitates the use of an rsh with a timeout option.

Patch Installation Instructions: 

su root
mv /usr/ucb/rsh /usr/ucb/rsh.FCS
mv <patchdir>/sun4/4.1.3/rsh /usr/ucb/rsh
chown root	/usr/ucb/rsh
chgrp staff	/usr/ucb/rsh
chmod 4755	/usr/ucb/rsh