[unix-pc.general] HARDWARE FLOW CONTROL ??BUG?? ON 3B1

mml@magnus.UUCP (Mike Levin) (06/18/88)

I *think* (I'm actually certain, but I'm trying to be tactful) that the
3B1 hardware flow control has a *serious* bug.  I am running a Trailblazer,
with the port frozen at 9600 baud, and *trying* to use hardware flow
control.  For purposes of uucp, all is OK.  Because of the 'g' protocol,
a bit is sent, we wait for a confirmation, we send again.  However, if I
log into my 3B1 from a remote location, with say a 1200 or 2400 baud
modem, things get weird.

What happens is that it *NEVER* garbles characters, but it garbles the
message.  It does this by *repeating* a portion of the text which it has
already sent.  For example:

Now is the time for all good men to come to the aid of their party.  The
Now is the time for all good men toThe
time has come, the walrus said, to speak of many things, of. . .

I made this example up, but you get the idea.  What I *think* (this time
it's really just an educated guess) that the code uses some sort of a FIFO
routine, who's pointers are not wrapped around correctly if the flow control
occurs at the point in memory where the wrapping occurs.  Just an opinion...

I tested the following:

	1).  Using a line monitor, I eliminated the possiblity of the problem
	being in the Trailblazer.  It is *really* repeating right out of the
	port of the computer.

	2).  With nothing else changed, I borrowed a terminal, set it up with
	no flow control, and plugged into the port.  With no changes to the
	configuration, it worked perfectly.

	3).  I compared copies of the kernal, the hfc_ctl driver, etc. with
	those on another 3.51 system, and they are identical.

	4).  I logged into a 7300 system, running 3.51a, which also has a
	Trailblazer running at 9600 with hardware flow control.  The EXACT
	same thing happenned.  When I asked the person it belongs to if he
	ever logged in from a remote location, he said no.  That's probably
	the ONLY reason he hadn't noticed it previously himself.

	5).  I talked to the AT&T hotline folks, they escalated it up a
	level, and told me they had no idea whether they could, or would
	want to, fix it.

If anybody out there is running a similar configuration, you might want to
give it a test.  Please let me know if you find the same problem.  NOTE: it
doesn't do it constantly, just sporadically.  Try catting out some file with
a few screenfuls of data, and you will most assuredly see it happen.  

IF ANYBODY KNOWS WHAT THE PROBLEM IS, OR HAS A FIX,  ***PLEASE*** LET ME
KNOW.  I have 4 3B1's which need to be able to rely on solid hardware flow
control. 

			T H A N K Y O U ! ! !

-- 
+---+  P L E A S E    R E S P O N D   T O: +---+  *  *  *  *  *  *  *  *  *  *
| Mike Levin, Silent Radio Los Angeles (magnus)| I never thought I'd be LOOKING
| Path {csun|kosman|mtune|srhqla}!magnus!levin |    for something to say! ! !
+----------------------------------------------+------------------------------+