schmidt@dduck.ctt.bellcore.com (Robert O Schmidt) (08/12/89)
the group that i am in has two ti explorer IIs. i have two sets of questions that hopefully can be answered by a guru on the net. first... has anyone heard of explorers causing sun workstations to not boot? if so, is there a known fix? we have clear evidence that this problem exists. a person boots his sun and it dies in the boot process. if the explorer is taken off of the network, the sun will boot. i am not a network guru, but am told that the problem relates to tftp. the booting sun sends out a tftp broadcast packet, trying to find its server. instead, the ti sends it something that makes it quite ill. i have discovered that using the peek window to arrest all processes referring to UDP resolves the problem. (i noted that udp was the the problem since the udp network statistics showed udp ports were talking to the suns that were not booting. udp appears related to tftp). both of our machines run the tcp/ip package. one is an LX system and it is this one that causes the above mentioned problem. the other, a normal explorer it is also in the habit of talking to suns via udp, but does not seem to cause any problems. i am told that sun considers this their problem but a fix that was obtained by the sun administrator did not solve the problem. thus, what i need is a means to make the ti not respond to booting suns. manually arresting the udp processes is a viable temporary fix. a better fix would be to arrest the offensive processes at boot time, or better yet, not start the processes. certainly the best fix would be to fix the running ti software to not respond to booting suns. any ideas? second...... printer problems. how might one print from a symbolics or a unix box to a printer on a ti? how might one print from a ti to a printer on a unix box or a symbolics? in all cases the printers are postscript devices. the ti printer is as omnilaser 2115. the printers on other systems vary, but are all postscript ready. our namespace server is a symbolics. it defines three printers. one is the ti, one on a pyramid unix box, and one on a symbolics. i have figured out how to make both of our tis use the one ti printer. the symbolics can use the other two. both tis however fail in using the other printers. similarly, symbolics users can not use the 2115. this is silly. is there a known configuration that will allow for sharing this resource? it would also be nice to be able to print postscript files from other systems on the ti. any of a variety of unix utilities create postscript files. these files can be copied to the ti and examined. i have however not been able to get them to produce the proper output when printed. i either get nothing or the ascii source. is this something that is doable and if so how? any help would be much appreciated. my net address is schmidt@flash.bellcore.com Robert Schmidt
pf@islington-terrace.csc.ti.com (Paul Fuqua) (08/14/89)
Date: Saturday, August 12, 1989 4:33pm (CDT)
From: dduck!schmidt at bellcore.com (Robert O Schmidt)
Subject: ti problems
the booting sun sends out a tftp broadcast packet, trying to
find its server. instead, the ti sends it something that makes it
quite ill.
I would argue that Sun is in error in sending a TFTP packet as a
broadcast. If you don't boot over the network, beat on Sun to provide a
boot sequence that doesn't do the broadcast.
(... udp appears related to
tftp).
TFTP is an application protocol running on UDP (similarly, FTP runs over
TCP). TCP and UDP are transport protocols running over IP.
any ideas?
(IP:RESET-TFTP-SERVICE NIL) will shut off TFTP service and prevent the
Explorer from responding to the Suns' TFTP requests (or anybody else's).
(IP:RESET-TFTP-SERVICE T) will start it up again (as will (NET:RESET T)
and (IP:RESET T), which call it). You could put the form in the
boot-init file for the machines if you want to shut it off more-or-less
permanently.
Paul Fuqua pf@csc.ti.com
{smu,texsun,cs.utexas.edu,rice}!ti-csl!pf
Texas Instruments Computer Science Center
PO Box 655474 MS 238, Dallas, Texas 75265tjan@umn-cs.cs.umn.edu ("Bosco Siaufung Tjan") (08/14/89)
There seems to have a solution that has been discussed sometimes ago:
--------------------------------------
From: macbeth.usc.edu!venkat@oberon.usc.edu (V. Venkateswar)
Organization: University of Southern California
Subject: Solution to ICMP error on broadcast messages
Message-Id: <15240@oberon.USC.EDU>
Sender: info-ti-explorer-request@sumex-aim.stanford.edu
To: info-ti-explorer@sumex-aim.stanford.edu
Status: OR
Somebody on the net was complaining that he only sees questions on the
net. Never the solutions.
My problem was solved recently ---
The problem
============
I had a network problem with our Explorer-II (TCP/IP protocol).
Whenever it received a broadcast message (eg:with the address 128.125.0.0),
it responded by sending an ICMP error packet. This cluttered our network.
The solution
============
Based on the suggestion form Paul Fuqua
I found that the :INTERNET-SUBNET-MASKS attribute of the :NETWORK entry
in the namespace of my explorer was NIL. I updated it to
("128.125.0.0" "255.255.0.0").
Hotline (512)250-7404 struggled for a week and came up with this patch.
Add the following code to the case statement in
the function "(defun-method portmap-dispatcher (svc-request stream)"
in the file sys: rpc; portmap.lisp
(case
..
..
;;*******************ADD THE CODE BELOW*****************
(#.pmaproc-callit
(ignore))
;;******************ADD THE CODE ABOVE***************
(otherwise
..
..
)
This presumably prevents the explorer from giving out an error
on broadcast messages.
=============
After these two "fixes" I have had no problem with the explorer
at all. More power to Hotline and Paul Fuqua!
V.V.
p.s. Yet to figure out which one of the above fixes actually solved
my problem.
-------------------
end forward message.
Bosco Tjan Phone (612) 625-0876
Lisp Machine Adminstrator Email tjan@umn-cs.cs.umn.edu
University of Minnesota (Twin Cities)