Emulex TC7000 trouble reports wanted

[Richard Stevens]

Yes, there is a problem with the TC7000.  I posted a query about 4 weeks
ago when we found a problem with the TC7000/Kennedy 9400 combination.
The problem is as follows:  for certain combinations of record lengths
and starting buffer addresses, the TC7000 will return the actual record
length that was read, but the final portion of the buffer is not actually
transferred from the controller to your buffer.  The problem is repeatable
but will change if you move the buffer from, say a page boundary, to some
other address.  I have since had others run my little test program that
duplicates the problem on a real DEC TM03 controller (which the TC7000
emulates) and on an SI controller which also emulates the TM03.  The
results were fine, which showed that the problem was not in the 4.2 BSD
ht driver.  Emulex, in their true to form 3rd party vendor style, said
"we don't have the problem under VMS so it must be a UNIX problem".
Well, I just got off the phone with Emulex Tech Support and they now
admit that it is a problem in the TC7000 and they were able to duplicate
the problem under VMS.  They were sure that it was a timing problem with
the UNIX driver.  I'll post what I find out from Emulex, or if you need
additional information just give me a call.

One additional note - Emulex is finally getting a UNIX source license in-house.
While I like their products a lot (we have their VMaster/SC780 with an Eagle,
a CS21/F, CS21/H, CS32/F, a TM11/9300 and the TC7000/9400) this is the
second time I've been burned with their "it works fine under VMS" attitude.
Maybe now they will take some time to learn UNIX since they must do quite a
volume of business to UNIX sites.  Maybe they'll even join the net !!

	Richard Stevens
	Health Systems International, New Haven, CT
           ihnp4 ! hsi ! stevens