Re: NetBSD 5 hang

[der Mouse <mouse%Rodents-Montreal.ORG@localhost>]

> I suppose my language wasn't clear enough - yes, the hang occurred
> when I ran ps, but all (perhaps interactive?) processes hung -

Well, at least some processes didn't hang; sshd must have got some user
time or you wouldn't've seen the ^T-provoked output at all.

You might be able to get a working shell with something like "ssh
$MACHINE sh -i"; you won't have job control, but it might be enough to
find out something useful.  Or perhaps it will work to do "ssh $MACHINE
$COMMAND"?  A poor substitute for a real shell, but a lot better than
nothing.

I'm having trouble coming up with a failure mode that will let the pty
driver work enough to accept ^T-generated output, pass it to sshd, and
have sshd work enough to get the data out over the network, but have
enough things wedge to give the symptoms you're seeing.  Maybe exit()
is what's wedging?  Depending on your shell startup files, this may
stop even "ssh $MACHINE $COMMAND" from working....

> the other ssh sessions I had open hung, too, and one which accepted
> control-t also showed the tstile state, too.

But also showed ^T output, meaning its sshd was working too.

> What is interesting is that a control-t now, after a few days, shows
> a higher load average as though cron is still running or something
> like that:

> load: 7.00  cmd: ps 12227 [tstile] 0.03u 0.14s 0% 1040k

Hmm, consistent with the "exit wedging" theory, if the trying-to-exit
processes get counted in the load average....

                                        Mouse