Linux 1.3.59 / 66 locks solid on modem hangup

Eric Schenk (schenk@cs.toronto.edu)
Wed, 21 Feb 1996 11:36:11 -0500

Oliver Jowett (oliver@sa-search.massey.ac.nz) writes:
> I've just run into a nasty problem that's present in at least
> 1.3.59-mark1, 1.3.59-mark6, and 1.3.66. I run a Linux box as a PPP
> server (it also runs ftp,http,etc) with 2 modems on a standard serial
> card. Intermittently, when one of the modems hangs up, the box will
> lock solid: no VT switching, no ping response, no reboot, no messages
> to syslog. Using the software watchdog doesn't help. It only seems to
> happen when the modem hangs up on its own (e.g. with a noisy line);
> I haven't had a lockup when the modem hangs up in response to a
> dropped DTR.
>
> I haven't been able to track this one down at all. 1.2.13-elf seems
> to run fine. I'm running a fairly vanilla Slackware 3.0,
> mgetty+sendfax 0.98, diald 0.12/pppd 2.2.0e.

I've seen this at least once myself, and had one report of it.

I suspect that the problem is being caused indirectly by diald-0.12,
which is trying to use the bind() call on the SOCK_PACKET socket
it uses for monitoring. I've had a confirmation that removing the bind()
calls from diald causes the problem to go away. Try either backing
off to diald-0.11, or commenting out all the calls to bind() in
diald-0.12 and see if the problem goes away. If it does, then the
problem is definitely related to the bind() calls that diald is making.

-- eric

---------------------------------------------------------------------------
Eric Schenk schenk@cs.toronto.edu
Department of Computer Science, University of Toronto