From mboxrd@z Thu Jan  1 00:00:00 1970
Message-Id: <200501261553.j0QFr9w28322@zamenhof.cs.utwente.nl>
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
From: Axel Belinfante <Axel.Belinfante@cs.utwente.nl>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Wed, 26 Jan 2005 16:53:09 +0100
Subject: [9fans] tftpd vs assumptions about user none? (to have or have not
	home dir)
Topicbox-Message-UUID: 36428abc-eace-11e9-9e20-41e7f4b1d025

I noticed that because I had a /usr/none directory  (*)
tftpd would try to serve relative file name requests
from /usr/none instead of from /lib/tftpd.

the reason is that tftpd first does a chdir
and only later (via setuser() a  newns("none", nil)
which in my case unfortuantely succeeded to cd to /usr/none
(because that directory existed, see below (*) until I
 removed it recently)

so,
a) what are the assumptions about none ?
   should not have home dir, I guess,
   other places where that matters?
   other issues?
b) would it make sense to make tftpd a bit
   more robust in this respect?
   It's probably worth noting, though, that
   I saw that in a recent tftpd.c the setuser()
   call was moved even `later' in the tftpd code,
   such that each session (via doserve()) gets its
   own clean namespace)
   from which I gather that moving setuser() _up_
   in the code till before (earlier than) the chdir
   call is likely a bad idea... 
   
Comments welcome, just trying to learn,
Axel.

(*) I added it myself (summer 2003 - history rocks :-)
and I don't remember why :-(