Anyone familiar with the sleep sleeps forever bug? Does this sound like
   it?


maybe... sounds like it probably was.


from ddb you can temporarily work around this problem by calling
"softclock()" manually.  fortunately it takes no arguments so you
just drop to ddb, type "call softclock" then "c" and it will should
have processed the current backlog of sleeping processes (but new
ones start appearing 1/hz'th of a second later ...)

i think i've needed two call softclock twice to reboot ... ie, you
may need to drop to ddb multiple times, each time calling softclock.


the reason named worked while the other services didn't is that
DNS is normally a UDP service.  all the others are TCP which happens
to be depend on softclock() running as well (for the TCP timeouts.)

as a *router* it doesn't get nearly so affected because routing
doesn't need timeouts... (but ipnat *does* ... i've had a U1/140
in this state have 14000 ipnat state entries...)



.mrg.