Hello...

For some time now I see occasions when my nearly always current system
(at most two weeks old at any point) gets stuck solid. So I assume there
is a bug that was in 1.3.x as well as in the current version. I canīt
reproduce it at will, but it occurs at comparable situations. So I try
to describe those. Perhaps someone has a clue...

I am running vgetty (from the mgetty-package) on my Pentium II. Mostly it
works quite well, but it seems there is a problem. From the vgetty-log
I gather that it gets stuck when it answers a call and the caller
hangs up while the message is played. vgetty is logging a Buffer
underrun. The next thing it tells is that a watchdog timer expired and
it exits.

From that moment on the system is doomed. Sooner or later I want to use
my ISDN TA (a ZyXEL, therefor attached to the serial port) with chat and
pppd. Normally vgetty should give up the serial port at that point and
wait for the connection to terminate and the lock file from pppd to
vanish. But here everything stops dead. Whatever command should be 
executed next in whichever shell just stops dead.

At that point I switched with wscons to the console. (Funny, that works 
for a few moments longer.) Here I entered the kernel debugger. As I know
the situation I went for a reboot. (Hmmpf, I should have requested a
core dump at that moment. Sorry, forgot that.) The debugger mentioned:

uvm_vnp_terminate(0x.....): terminating active vnode (refs=2)

and that was the last the machine did before a power cycle.


I canīt reproduce the initial problem, but the story is always the same:
A problem with vgetty, trying to use chat and pppd on the serial port and
being stuck solid.

Any ideas?

Many thanks in advance,

Bjoern

-- 
Bjoern Labitzke  <hermit@cs.tu-berlin.de>
   Use PGP! (Don't you use envelopes for your letters?)