login
Header Space

 
 

Re: NFS mount by non-root

Previous thread: The OpenBSD Command-Line Companion Book delayed (or MIA?) by Austin Hook on Wednesday, April 25, 2007 - 7:38 pm. (8 messages)

Next thread: Erlangen mirror downtime by Alexander von Gernler on Thursday, April 26, 2007 - 4:44 am. (2 messages)
To: <misc@...>
Date: Wednesday, April 25, 2007 - 11:33 pm

Is it possible for users (non-root) to mount NFS exports?
I seem to be able to mount_nfs using sudo, but not as a regular user.
I actually want to allow regular users to mount the NFS share from
another machine/OS (MacOSX), but since I couldn't get a regular user
to do the mount just on the local machine, I thought I'd start with
this problem first.

With these settings on OpenBSD 4.0, (generic+autoraid)
/etc/rc.conf.local
 lockd=YES
 portmap=YES
 nfs_server=YES
/etc/exports
 /home -alldirs -ro -network=10.0.1 -mask=255.255.255.0
/etc/sysctl.conf
 kern.usermount=1

For example
logged is as usera, on machine 10.0.1.201 (the server)
 uid=1000(usera) gid=1000(usera) groups=1000(usera), 0(wheel), 10(users), 20(staff)

$ mount_nfs 10.0.1.201:/home/usera/dir2share /home/usera/private/mnt
fails with
"mount_nfs: bad MNT RPC: RPC: Authentication error; why = Client credential too weak"

but
$ sudo mount_nfs 10.0.1.201:/home/usera/dir2share /home/usera/private/mnt
works fine


Any help would be appreciated.

More details:
I've tried to adhere to:
man mount
"A mount point node must be an existing directory for a mount to succeed
 (except in the special case of /, of course).  Only the superuser may
 mount file systems unless kern.usermount is nonzero (see sysctl(8)), the
 special device is readable and writeable by the user attempting the
 mount, and the mount point node is owned by the user attempting the
 mount."

by setting:
drwxr-xr-x  root   wheel  /home
drwxr-xr-x  usera  usera  /home/usera
drwxr-xr-x  usera  usera  /home/usera/dir2share
drwxr-xr-x  usera  usera  /home/usera/private
drwxr-xr-x  usera  usera  /home/usera/private/mnt


Thanks
To: <dmaus@...>
Cc: <misc@...>
Date: Thursday, April 26, 2007 - 5:04 pm

mountd wants the request to come on a reserved port.
To: <dmaus@...>
Cc: OpenBSD <misc@...>
Date: Thursday, April 26, 2007 - 11:14 am

i've always approached this class of problem with amd:
http://www.openbsd.org/cgi-bin/man.cgi? 
query=amd&amp;apropos=0&amp;sektion=0&amp;manpath=OpenBSD 
+Current&amp;arch=i386&amp;format=html

the daemon runs with sufficient privs to mount the fs, and all the  
user has to do is reference the fs.

Ben
To: Douglas Maus <dmaus@...>
Cc: <misc@...>
Date: Thursday, April 26, 2007 - 6:54 am

Maybe I am being dumb, but wouldn't the obvious thing be to configure 'sudo'
for each user so that they can mount 'their' share (only)?

This would probably be quite onerous if you had a lot of users, mind you.

Tor
To: Tor Houghton <torh@...>
Cc: Douglas Maus <dmaus@...>, <misc@...>
Date: Thursday, April 26, 2007 - 11:19 am

Actually, that's probably the simplest solution; config a user group with
access, then grant sudo access to that group for the mount.

	Lee
To: <misc@...>
Date: Thursday, April 26, 2007 - 9:43 am

I think 'best scripted' is the UNIX way of spelling 'quite onerous', no?

		Joachim 

-- 
TFMotD: gem (4) - GEM 10/100/Gigabit Ethernet device
To: Douglas Maus <dmaus@...>
Cc: <misc@...>
Date: Thursday, April 26, 2007 - 1:04 am

HISTORY
     The -P flag historically informed the kernel to use a reserved port
     when communicating with clients.  In OpenBSD, a reserved port is
     always used.

This means to me that you will always have to be root to use mount_nfs.
Unless I'm missing something.

-ME
Previous thread: The OpenBSD Command-Line Companion Book delayed (or MIA?) by Austin Hook on Wednesday, April 25, 2007 - 7:38 pm. (8 messages)

Next thread: Erlangen mirror downtime by Alexander von Gernler on Thursday, April 26, 2007 - 4:44 am. (2 messages)
speck-geostationary