From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <6a19626458e13903ed8dcf4a40eab71e@plan9.bell-labs.com>
From: "Russ Cox" <rsc@plan9.bell-labs.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] Drawterm session close??
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Date: Fri,  8 Mar 2002 20:53:43 -0500
Topicbox-Message-UUID: 621d107a-eaca-11e9-9e20-41e7f4b1d025

> > I'm not sure why the processes are left over.
> > Usually a name space loop, but in this case
> > that would mean rio has fs in its namespace
> > and fs has rio in its namespace, but I don't
> > think you can set things up that way.
> Can you elaborate this point? I have no clue
> what you are referring to.

File servers like rio and fs stick around until
the last references to their file servies go away.
Those references take the form of other processes
that have them mounted or the service files they
post in /srv.  (Actually the ref count is on the
pipe that you are serving 9P over.)

So: if rio has upas/fs mounted, and upas/fs has
rio mounted, then you have a ref count loop and
won't take them down.  You can accomplish an
easier ref count loop by writing a file server
that mounts itself.  This is why file servers
do rfork(RFNOTEG) before entering the service
loop.

It would be helpful to see the stacks of the
processes involved.  You can do

	chmod +rw /proc/pid/ctl
	acid pid
	stk()

to look at them (assuming you are bootes).

Russ