SCO ODT TCP/IP Setup Problem

[Geoff Scully]

>>/--mpd at anomaly.sbs.com (Michael P. Deignan) said...
>> [rlogin/telnet problem]
>>| 
>>| The error is "Bad login user id".
>>| 
>>| I've RTFM, but can't locate a reference to getting rid of this error. Of course,
>>| I might have missed it at 3am.
>>\--
>>
>>[ wul at sco.com (Wu Liu) replies... ]
>>I'm not sure if it's mentioned in the doc...
>>
>>Your problem is that TCP on anomaly wasn't started by init.  Your
>>inetd process can't have an luid set, or else telnet/rlogin breaks
>>with this error message.  The only way to get a process to not have
>>an luid is if you start it during boot (i.e. have init fork it off).
>
>[Scott Deardorff adds],
>Michael,
>     In addition to Wu's response, I wanted to include an item from
>our information tools database which also addresses this problem.
>This database is available to our supported customers online via
>a machine named sosco.  For more info on accessing this database
>please contact your SCO Customer Service, or Sales Representitive.
>
>Regards,
>    Scott Deardorff
>
>rlogin or telnet give the error message "bad login user id."
>
>PROBLEM:   When using rlogin or telnet to log on to a machine running 
>	   SCO UNIX System V/386 Release 3.2, the error message "bad 
>	   login user id" is displayed.
>
>SOLUTION:  This error is caused by doing a "tcp stop" followed by a 
>	   "tcp start." init(M) has the unique property in that it does 
>	   not have a loginid associated with it.  Therefore, when SCO 
>	   TCP/IP is started by init, there is no loginid associated 
>	   with the inetd process.
>
>	   The proper method to restart SCO TCP/IP follows:
>
>		# init 1 
>
>			(while logged in as root on /dev/tty01)
>
>		< here all the user processes are killed by "init 1" >
>
>		Then type CONTROL-d to proceed with normal startup.
>
>SEE ALSO:  init(M), setluid(S)

Another option that is available that avoids having to bring a possibly
production machine down to single user and back up, is to use inittab's
a b and c 'pseudo state' entries to start single shot processes from init.
I use this on xenitec for restarting the cron if I have had to kill it off,
using the following inittab entry.

cron:a:once:/etc/cron

So you could set-up something like,

slip:b:once:/etc/tcp start

to restart tcp after you have manually done a 'tcp stop'. To use these
pseudo state entries, use the command 'telinit <state>', or 'telinit b'
for the above example for tcp. The nice thing about these pseudo states
is that the processes they start are not killed by changing init states,
except when going to state 0, 1, or S.

Don't forget to also make these changes to /etc/conf/cf.d/init.base if you
want them to stick around through rebuilds of the kernel environment.

---
Geoff Scully -- XeniTec Consulting Services
gws at xenitec.on.ca
{watmath, utai!lsuc}!xenitec!gws