From owner-freebsd-hackers  Wed Dec 13 06:16:30 1995
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id GAA13460
          for hackers-outgoing; Wed, 13 Dec 1995 06:16:30 -0800 (PST)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id GAA13440
          for <hackers@FreeBSD.ORG>; Wed, 13 Dec 1995 06:16:25 -0800 (PST)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id BAA23057; Thu, 14 Dec 1995 01:10:41 +1100
Date: Thu, 14 Dec 1995 01:10:41 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199512131410.BAA23057@godzilla.zeta.org.au>
To: eugen@rd.zgik.zaporizhzhe.ua, hackers@FreeBSD.ORG
Subject: Re: bug or feature in kern/tty.c
Cc: yura@rd.zgik.zaporizhzhe.ua
Sender: owner-hackers@FreeBSD.ORG
Precedence: bulk

>I found strange thing in 2.1.0-RELEASE (and -current) in kern/tty.c.
>When carrier was lossed then ttymodem() set TS_ZOMBIE and clear
>TS_CONNECTED in tp->t_state. But TS_ZOMBIE is cleared only in ttioctl()
>TIOCSETA* when line is setted to ignore modem status and when line is closed.

POSIX says that (what I call the) zombie state remains until the line is
closed.  Clearing it when CLOCAL is raised is a special hack that helped
very old (1.1.5) versions of slattach and perhaps other broken programs
work right.

>As result in ttymodem() after lossing carrier TS_CONNECTED  is not setting
>whithout reopening line or changing c_cflag to CLOCAL, and if this line
>used for SLIP connection as result we have 'No route to host' error
>when trying output to this line after modem reconnected.

>Or is this bug in slattach ? slattach never set CLOCAL when it's working
>with modem control and without redial cmd.

Yes.  slattach goes to a lot of trouble to close the line, but only for
the redial_command case.  The !redial_command case in sighup_handler()
can't work for real SIGHUPs.  In 1.1.5, the only way to close the line
was to exit from the process.  slattach goes to a lot of trouble to make
itself the controlling process so that SIGHUPs get sent to it.  The line
can't be fully closed until it stops being a controlling terminal, and in
1.1.5 there was no way to stop it being a controlling terminal except to
exit! :-(  2.x is better (TIOCSCTTY releases the previous control
terminal, if any).  Anyway, slattach still exits (in daemon() in
aquire_line()) for the redial_command case.

According to the man page, slattach aborts if it receives a SIGHUP in the
!redial_command case.  This got broken when complications to get rid of
the controlling terminal were introduced.

Bruce