Re: stacker and dummy

From: Tony Jones (tonyj@private)
Date: Mon Jun 27 2005 - 13:36:38 PDT


On Mon, Jun 27, 2005 at 03:28:44PM -0500, serue@private wrote: 

Just make sure dummy can't be unloaded by /sys/stacker/unload. I didn't 
bother to enforce this.

I was pondering a soln where dummy isn't really loaded per-se, but it's hook 
is called whenever no registered sub-module implements a given hook (all are 
NULL) which would achieve the same purposes as the patch when no sub modules
are loaded.   But I need to go page in the past discussion regarding verify 
and security_fixup_ops().

I'd still like to see dummy go away and capability become the default.

Thanks for putting in a lot of hard work on this Serge, looks like you may
achieve what many thought was impossible!

Tony

> It seems to me the right thing to do is to continue the current
> non-stacker behavior of dummy being loaded until something else is
> loaded, then popping dummy and pushing the new module.  Since dummy
> doesn't store any information, and since dummy_security_ops won't be
> deleted, there shouldn't be any funky locking issues.  I should be able
> to just swap demmy with the new module and unset an no_lsms_loaded
> boolean all under an rcu_read_lock().
> 
> Does anyone disagree?  Should behavior change?  (I'm not sure how else
> it would change other than to make capabilities the default)  If not,
> I'll send out a patch for this tomorrow.
> 
> thanks,
> -serge



This archive was generated by hypermail 2.1.3 : Mon Jun 27 2005 - 13:41:06 PDT