From: cummings@hammer.prime.com (Kevin Cummings) Subject: Re: 99.5 Kernel breaks serial logins Date: Wed, 17 Feb 1993 22:16:46 GMT
In article <729894051snxwomble@spuddy.UUCP>, sweh.womble@spuddy.UUCP (Stephen Harris) writes:
> In article <6805@falcon.ukc.ac.uk> cjw1@ukc.ac.uk writes:
>
> > Well, I upgraded my kernel to 99.5 a few days ago, and wondered why I couldn't
> > login via the serial port. Symptom? : After typing the login name, /bin/login
> > freezes, not printing "password:". Since I'm using LILO with multiple images,
> > I tried it with 99.4. No Problem. Oh yes, I tried 995 in autoconfigure, and
> > not-autoconfigure modes with the serial driver.
>
> New serial drivers use hardware signals to wake up.
> In theory, the getty should be blocked as well, but some of them were compiled
> with NOBLOCK, and so will open. Then the login process tries to run, and
> freezes.
>
> The solution: make pin 8 high. This will make the serial port believe a modem
> has connected.
>
> Possible alternate method: run the getty on /dev/cua# where the device
> cua# had major 5, and minor the same as ttyS#
NO! Don't do this! the cua* devices are for dialout! The ttyS* devices are the
right ones to use for gettys and logins. If you are trying to use a serial line
for both dialin AND dialout, the getty should run on ttyS* and the dialout software
should use cua*. Each device will block (or fail) if the other is in use.
The best possible workaround is to set the serial port to ignore DCD with STTY,
or rewire the serial connector to jumper DTR to DCD (from the terminal).
=================================================================
Kevin J. Cummings ComputerVision Services
20 Briarwood Road 500 Old Connecticut Path
Framingham, Mass. Framingham, Mass.
Home: cummings@kjc386.framingham.ma.us Work: cummings@primerd.Prime.COM
Std. Disclaimer: "Mr. McKittrick, after careful consideration,
I've come to the conclusion that your new
defense system SUCKS..." -- War Games
=================================================================