From: tytso@ATHENA.MIT.EDU (Theodore Ts'o) Subject: Re: BUGS (minor) Date: Sun, 4 Oct 1992 20:39:51 GMT
From: edmonds@cs.ubc.ca (Brian Edmonds)
Date: 1 Oct 1992 17:20:24 -0700
1) I have a terminal attached to /dev/ttys3 which I use for most of
my work (I guess I just like an ascii interface :). Approximately
every two hours, my shell gets terminated. Usually I'm running
screen, which apparently receives a SIGHUP, and detaches itself, so
it's not a problem except that I have to log back in and reattach. If
it's just a shell, then I get logged out and jobs running under it
(such as kermit) get immediately terminated.
It sounds like the Carrier Detect line on your RS-232 port is flapping
--- every single time the CD line goes from high to low, the kernel will
send a SIGHUP to the current foreground process. (This is a feature,
when you have a modem connected to the line.) I suspect the CD line is
not connected, and so the signal is just floating. Solution number one
is to properly ground it. Solution number two is to disable the modem
control lines by doing a "stty clocal < /dev/ttys3". The second
solution is probably easier, especially if you put it in /etc/rc.
2) I am running with four serial ports, 3 16450s and one 16550A. The
450s work fine, and generally keep up. I would like to use the 550A,
but it is behaving extremely strangely. I start up kermit and talk to
the modem for a while, then I can dial, and receive the connect
message, but no data received by the modem is passed to the computer.
Data generated by the modem itself (connect msgs for example) arrive
no problem. Could this be a problem with the chip itself, and if so,
is there a test to confirm it?
This is bizarre; there should be no distinguishable difference between
data sent by the modem and data sent through the connection. Have you
tried a different modem? It sounds like a modem problem, except that
you said the modem works if you put it on a different port that has a
16450 installed. I'm grasping at straws here, but perhaps one of the
other chips on the 16550A's port is damaged, perhaps the line receiver
chip? Why don't you trip actually swapping the 16450/16550 chips
themselves, and see if the problem follows the chip or not.
- Ted