info-3b2@lamc.UUCP (Info-3b2 Mailer) (01/05/89)
Info-3b2 Digest, Number 48
Thursday, January 5th 1989
Today's Topics:
Terminals and baud rates
Intermittent Ports Hanging
Re: Intermittent Ports Hanging
Re: Intermittent Ports Hanging
----------------------------------------------------------------------
Subject: Terminals and baud rates
From: pacbell!att!ocrjd!randy
John LaBadie writes:
|Don't know if all the reported symptoms are due to the same problem,
|but yes, both the ports and eports cards have specified
| MAXIMUM AGGREGATE BAUD RATES
|
|On the ports card, it is 19.2K for the 4 ports.
|On the eports card, it is 38.4K, but I am uncertain if this is for all
|8 ports, or 2 X 19.2 for 2 sets of 4 ports each.
....and the maximum speed for each port on a 195B (OLD ports board) is 9600
baud, for each port on the 195BA (newer ports board) is 19.2K and for the
eports is 38.4K, and each is tested from 300 baud up to the maximum for the
board before it leaves the factory.
So, for a old ports card (max of 9.6K baud per port), only one can run
9.6k on two ports at one time, and the other two are at zero, or all four at
4.8K. But the old ports card would not go insane (at least not for me), it
would merely flow control, and in sometimes drop characters.
In practice, I have been able to make the old ports card run at 19.2K for
short lengths of time. Must be because the throughput was low.
I have very little consumer experience with the eports, so no help there.
|We solved the symptoms, not the defect, by two techniques.
|
|1) Balancing port assignments to mix high speed ports with fixed rate
| low speed ports (Ex. a 1200 baud modem with a 9.6K laser printer)
...on the same board, but different ports. Perfect way of doing it.
|2) Limiting terminal access to 4.8K baud. This proved to be a
| problem, but not a system administration one. It was a user
| education problem. Sure, we set the gettys to 4800 baud,
| even went around resetting terminals. But the users would
| login at 4800 baud, then issue a "stty 9600" or "stty 19200"
| and reset their terminals accordingly.
[.....]
| Any way, by being slowly responsive to the worst offenders,
| reboots in the middle of the day are a pain to productive
| users, everyone now accepts the slower speed and we have not
| experienced a repeat of the problem for about a month. It
| was occuring twice a day.
|
|Definitely not a fix, but perhaps a work-around.
Also not really a problem, as no *intelligent* terminal I know of actually has
a throughput above 9600 baud. In fact, the AT&T, DEC, and HP terminals I have
used all start using flow control at anything over 4800 baud. By doing the
very unscientific method of "cat"ing a large file and timing it, it seems that
most terminals (combined with any system delays) have a throughput in the range
of 6K to 8K baud. Something to do with the fact that these terminals actually
store the characters for more than a page (you know, the terminals you can
scroll up a few screens).
Randy Davis UUCP: ...(att!)ocrjd!randy
...(att!)occrsh!rjd
------------------------------
Subject: Intermittent Ports Hanging
Date: Thu, 05 Jan 89 02:05:55 -0500
From: len@netsys.COM
I think most of us were aware of the baud rate limitations on the older
ports hardware,since it's well documented in the manuals. I believe it
is a software problem however since it does occur on the eports board,
as well as the older ports hardware.
I have long since stopped using the ports hardware for anything but
modems (1200/2400),and only tolerate it for the parallel port. God
bless eports.
Len Rose
len@ames.arc.nasa.gov or len@netsys.com
------------------------------
Subject: Re: Intermittent Ports Hanging
Date: Mon, 2 Jan 89 22:50:20 EST
From: lll-crg!lll-tis!daitc!otishq!guardian (Harry Skelton)
= this happens at least once a week on some machines I use.. Usually
= using cu to access another machine at speeds of 9600 and up.. The
The table in your ports option manual will tell you how fast you can
run a port at any given time. Without handshaking - you may be lucky to
get by with 2400 but with handshaking, you can only run all ports at a
MAX of 9600 with general terminal I/O - NO CU'S DOING ~%take me's. The
table is broken down to show speed vs. character I/O.
Those that are lazy - here is that table:
(page B-4 Enhanced Ports Manual)
Raw mode character i/o
Max I/O at 19200 :
7 Out - 0 in
6 Out - 1 in
. ... . . ..
0 Out - 7 in
you can run 9600 on all ports but at 19200 you get a performance of:
50% (4 ports) at 100% peak with remainder trailing at 95/70/55/50 (output of
characters).
The number of lines that can input without data loss is even more interesting:
(3B2/500 or 400) (even less with a 300 !!!)
speed w/o flow ctr w/flow ctr
of port raw cooked raw cooked
1200 8 8 8 8
9600 8 4 8 8
19200 7 2 8 7
38400 3 1 3 2 (why have it available then?)
We bugged AT&T about this for a while untill someone got the bright idea to
look at the shrink wrapped manuals 8-). After that we have slowed down all
ports to 9600 and spaced our modems over several cards as not to burden them
when we hit the news or archive sites 8-).
In one of my other messages, I stated that the eports command to set hardware
flow control was 'eports -hfc' it's 'epstty -hfc'. ALSO!!!! NOTE NOTE NOTE:
You have to disable any and all flow control before you use this else it will
either ignore you or crash the port.
--
_______ guardian@otishq _ _ (Via: daitc obdient ddsw1 killer novavax )
/ _____/ __ _ ___ ___ _| ||_| ___ _ __ Harry "Highspeed" Skelton
| |__ || | |/ . || .-'/ . || |/ . || '_ \ AAA National Head Quarters
\_____/ |____|\__|||_| \___||_|\__|||_| |_| Falls Church, Virginia USA
------------------------------
Subject: Re: Intermittent Ports Hanging
Date: Mon, 2 Jan 89 22:24:38 EST
From: lll-crg!lll-tis!daitc!otishq!guardian (Harry Skelton)
= I'm glad someone else is having this problem! We are also running 3.1.1
= with eports 1.2, and have the same problem. In our case, it usually
My understanding is there is a 1.2.1 available (real beta test though) and
there is a real hardware problem with eports in general. In short - this is
as good as it gets. AT&T is working on a new board (yeah - what about us
that have the old one?) and is trying to debug current uucp/cu programs to
be a bit better with the eports. I have tried uupc-H on the lines (uupc ported
to unix. w/cu) and found it hanging more and more. There is a include file
that has to be included to help control the port on disconnect. This is
the hardware flow control header. If you don't have source or uupc-H (uupc-H
has not been released public - only beta at present) then use:
eports -hfc < /dev/ttywhatever (as long as it is eports)
This should set up the port for hardware flow control and you'll have to do
a ~%nostop under cu to allow for this. I think you have to run cu and then
do the ~!eports -hfc < /dev/.... to get it to work. Try it. I haven't tried
since I have source. Sorry guys.
Failing this - check your voltage settings from your devices to insure
that there are no lingering hardware signals about. (3B2's are bad about
this and Static (take my tape drive PLEASE 8-) 8-) )).
--
_______ guardian@otishq _ _ (Via: daitc obdient ddsw1 killer novavax )
/ _____/ __ _ ___ ___ _| ||_| ___ _ __ Harry "Highspeed" Skelton
| |__ || | |/ . || .-'/ . || |/ . || '_ \ AAA National Head Quarters
\_____/ |____|\__|||_| \___||_|\__|||_| |_| Falls Church, Virginia USA
-------------------------------------
To join this group or have your thoughts in the next issue, please
send electronic mail to Ken Davis at the following address;
{pacbell,netsys,hoptoad,well}!lamc!info-3b2-request
The views expressed in Info-3b2 Digest are those of the
individual authors only.
*********************
End of Info-3b2 Digest
*********************