Re: /sbin/reboot and secmodel

[Emmanuel Vadot <manu%bistoufly.org@localhost>]

Christoph Badura wrote:
> So, assuming that we would want to change our policy of signalling init(8)
> to be overridable by different secmodel, why not just implement that?
> I.e. change secmodel_bsd44 to return KAUTH_RESULT_DEFER when a process
> tries to signal pid 1.
>
> Then Emmanuel can create a secmodel that authorizes specific users to do
> that by returing KAUTH_RESULT_ALLOW.
>
> --chris
>   
Wow, I launch a big discuss here :)
To be clear with everyone, what I'm doing right now it's just pleasure 
and for testing the secmodel under NetBSD-4.0, just that.
I'm new in coding in kernelland and I wanted to code a dynamic secmodel 
with a userland program to modify "rules" which apply under the bsd44 
secmodel.
I thought it was a good start cause it involves a lot of thing 
(secmodel, pseudo-device and ioctl for the dialog).
I triggerded some bugs and talk to elad@ to find out what is the best 
way to doing something.
So the question is not how I will doing some thing like reboot, is how 
thing have to be in the kernel I think.
For those who want my code I think I'm gonna be able to release it at 
the end of the week.
I think secmodel is really a big improvment in the unix point of view.
I would kill someone to have an httpd (for exemple) which to not do a 
setgid(xxx) and setuid(xxx) just to run as user xxx and capable of bind 
the port 80 :) .
The possibility are infinite with this kind of design.
Just my two cents ....
--
Manu