[comp.sys.att] Root equivalence over RFS

eric@ms.uky.edu (Eric Herrin) (09/04/87)

I have a rather unique(?) problem here running RFS over a STARLAN 
network of AT&T 3b2s.  Root cannot be equivalent over all the machines.
This tends to break many of the SUID root programs which use shared
files and directories, since root has no special access to things like a
user's directories which reside on other machines.  Limiting ROOT 
access to foreign machines is a nice security feature, mostly for sites
in which several groups of people own different sets of machines on
the network (ie. it is usually not desired for someone's workstation
to be equivalent).  I really don't care about such situations, simply
because they do not exist here.  I am the only System Admin. of the
RFS.

My solution was to hack the DU module so it allows root equivalence.
I am not sure if other people have experienced similar problems or even
if anyone else actually runs RFS.  If anyone has needed to allow root
equivalence, and has a better solution, I would love to hear about it.
Otherwise, I am willing to talk about my solution to any interested
parties.

			eric


|									      | 
|    Eric Herrin II				     	cbosgd!ukma!eric      |
|    "'tis better to be silent                         	eric@UKMA.BITNET      |
|     and be THOUGHT a fool, than to open              	eric@ms.uky.csnet     |
|     one's mouth and remove all doubt."                eric@ms.uky.edu       |

gwyn@brl-smoke.ARPA (Doug Gwyn ) (09/04/87)

In article <7211@e.ms.uky.edu> eric@ms.uky.edu (Eric Herrin) writes:
>My solution was to hack the DU module so it allows root equivalence.

I thought you could achieve this with existing facilities (e.g. the
rules files used by idload(1M))?  What am I missing?

ekrell@hector..UUCP (Eduardo Krell) (09/05/87)

Have you tried

global
default transparent

in your /usr/nserve/auth.info/uid.rules?. This works for us. We're not
running RFS over STARLAN, but that should make no difference.
By the way, we're running SVR3.1
    
    Eduardo Krell                   AT&T Bell Laboratories, Murray Hill

    {ihnp4,seismo,ucbvax}!ulysses!ekrell

eric@ms.uky.edu (Eric Herrin) (09/05/87)

>Have you tried
>
>global
>default transparent
>
>in your /usr/nserve/auth.info/uid.rules?. This works for us. We're not
>running RFS over STARLAN, but that should make no difference.
>By the way, we're running SVR3.1
>    
>   Eduardo Krell                   AT&T Bell Laboratories, Murray Hill
>
>   {ihnp4,seismo,ucbvax}!ulysses!ekrell


This seems to work just fine for every uid except root..  This was, of
course, the very first thing I tried.  I tried rebooting RFS, the
OS, and resorted to the hack as a last hope (at least I knew it would work).

The uid.rules and gid.rules SHOULD do what you have implied, and if they
actually do for you, maybe there is something else wrong.  I am running
SVR3.1 (with sources of course) with ether and starlan, but only starlan
with RFS.  The uid being returned from the server side when root owns
something is NO_ACCESS or MAXUID+2.  This implied to me that the .rules
files were automatically excluding root.  If you have any further
suggestions, please e-mail me unless there is considerable interest on
the net about this.

				eric

mjp@sfmag.UUCP (M.J.Purdome) (09/08/87)

> In article <7211@e.ms.uky.edu> eric@ms.uky.edu (Eric Herrin) writes:
> >My solution was to hack the DU module so it allows root equivalence.
> 
> I thought you could achieve this with existing facilities (e.g. the
> rules files used by idload(1M))?  What am I missing?

Right.  The ID mapping rules files used by idload allow all sorts of 
mapping between users.  I think the typical default configuration has
files that say something like
	global	
	default transparent
	exclude 0-99
which allows all users to be mapped except 0-99 (the administrative
logins), which become MAXUID+1.  If you delete the last line, all users,
including root, are mapped transparently.

Check the rules files, located in /usr/nserve/auth.info/?id.rules, 
the man page for idload(1M), or the _Administrator's Guide_, which has an
RFS tutorial.
-- 

    Mark Purdome      AT&T UNIX System Development Laboratory
                      190 River Road A-130, Summit, NJ  07901
                      {ihnp4,allegra}!attunix!mjp
    disclaimer: these statements do not represent AT&T policies or opinions