fai-setup: (1) all_squash,anonuid=500,anongid=500 (2) exportfs -r

malk at sidehack.sat.gweep.net malk at sidehack.sat.gweep.net
Fri Dec 24 15:42:36 CET 2004


Forgive me if I'm stating the obvious, but doesn't exporting it read-only
solve all the problems?  Even if no_root_squash is set (which is desirable
on nfsroot), no one can do anything to the filesystem remotely since it's
read only.

-Eric

> On Thu, Dec 23, 2004 at 12:43:44PM +0100, Thomas Lange wrote:
> > >>>>> On Thu, 23 Dec 2004 01:02:28 +0200, Shaul Karl <shaulk at 013.net> said:
> > 
> >     >   Just wondering why not having fai-setup
> >     > (1) use all_squash,anonuid=500,anongid=500, or some other uid and gid
> >     >     both for $FAI_CONFIGDIR and $NFSROOT when exporting those
> >     >     directories.
> > Please explain in more detail why this makes sense.
> 
> 
>   It restricts the possible misuse of fai slightly more, doesn't it? I
> intended to run fai on a machine that provides other services because:
> 1. I do not want to set a dedicated fai server. 
> 2. The usage I have for fai is sparse. I install only occasionally,
>    with large gaps between installations. I would also like to use fai
>    in case of a client disk breakage.
> Having one more small assurance that the security holes I have to make
> for fai are of limited consequences is desirable, as far as I am 
> concern. For example, I might mistakenly connect the eth interface that
> I set up for fai for other purposes. Therefore, having root_squash for
> that interface will slightly improve my position on such an error. 
> 
> 
> > 
> >     > (2) use exportfs -r instead of guessing which nfs server is running and
> >     >     restarting it.
> > The command exportfs is only available if you are using the nfs-kernel-server.
> > 
> 
> 
>   I didn't knew that. Still, why not using that when nfs-kernel-server
> is available instead of restarting the service? 
> 



More information about the linux-fai mailing list