[comp.unix.wizards] cu

dave@micropen (David F. Carlson) (03/16/87)

I am runnning some dumb (as in no DMA etc) serial boards on my
SYS V.2 box (Microport on an AT clone with Digicom/8 port boards).
I have a phenomenon which I cannot explain.  Guesses anyone?

Description:
	When there is a high rate of interrupts (UUCICO running) the
	cu(1) between one machine and another identical machine has doubled
	characters, missing characters and garbled characters.  The
	cu line has no parity enabled so overruns and other interrupt
	related problems are immediately suspect.  However, none of the
	other 4 terminals on the same board experience any problems.  They
	are running in full duplex, as is the cu session.  Also, any uucp
	between the two machines during this period is munged too, so it
	is not the cu session, but the serial board to serial board connection
	that is suspect.  Why do only machine to machine connections fail?
	(Assume correctly that the other machine is totally idle with no other
	interrupt traffic.)  Any guesses are appreciated.  Flames to buy
	a "real" computer or ya68kub to /dev/null.


-- 
David F. Carlson, Micropen, Inc.
...!{seismo}!rochester!ur-valhalla!micropen!dave

"The faster I go, the behinder I get." --Lewis Carroll

philip@axis.UUCP (03/19/87)

In article <164@micropen> dave@micropen.UUCP writes:
|I am runnning some dumb (as in no DMA etc) serial boards on my
|SYS V.2 box (Microport on an AT clone with Digicom/8 port boards).
|I have a phenomenon which I cannot explain.  Guesses anyone?
|
|Description:
|	When there is a high rate of interrupts (UUCICO running) the
|	cu(1) between one machine and another identical machine has doubled
|	characters, missing characters and garbled characters.  The
|	cu line has no parity enabled so overruns and other interrupt
|	related problems are immediately suspect.  However, none of the
|	other 4 terminals on the same board experience any problems.  They
|	are running in full duplex, as is the cu session.  Also, any uucp
|	between the two machines during this period is munged too, so it
|	is not the cu session, but the serial board to serial board connection
|	that is suspect.  Why do only machine to machine connections fail?
|	(Assume correctly that the other machine is totally idle with no other
|	interrupt traffic.)  Any guesses are appreciated.

This is ONLY a guess, but ...

Normal users send characters with LONG spaces between them.
CU and uucp send characters in bursts, packed nose to tail.
The baud rate is the same, but without the 'normal' large gaps
between.

Many UNIX systems experience problems when you throw characters at them
like this. The problem could be in the card, or in the driver, its
difficult to say which - personaly, I would suspect the firmware on the
card.

We have a product which does UNIX <-> MS-DOS communications. One of the
things we were FORCED to add was the ability to specify a delay
BETWEEN CHARACTERS - many UNIX systems just refuse to work if you
dont give them time to fully digest the character before the next one
arriveot.ot.oBrit