[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: depinit


On Mon, 23 May 2005 18:14:22 BST, Luke Kenneth Casson Leighton said:

>  there _is_ no "start" option or "stop" option to the service
>  startup scripts.  if the script dies, the service is assumed to have
>  died: the signal is caught by depinit (all signals are caught by
>  depinit) and an "action" script e.g. /etc/depinit/sig_chld is called.
>  
>  if you _want_ the service to stop, you get depinit to automatically
>  send increasingly aggressive levels of signals to the "service" script.
> 
>  there _is_ no stupid numbering schemes - not the stupid symlinks, and
>  not the stupid runlevels.

This is actually irrelevant (mostly) as far as SELinux is concerned..

>  everything is _named_.  properly.

Again, mostly irrelevant..

>  the present policy - init.te, initrc.te - is therefore going
>  to be mostly irrelevant.

Actually *looking* at the .te's shows that *most* of it will, in fact,
still be applicable for *any* init-replacment.  The *important* part is
that the policy address things like "can the init-process access these
resources?" and "can the init-process launch the MTA process and have that
transition to the proper domain?" (domain-auto-trans does a *lot* of the
heavy lifting here - for instance, domains/program/mta.te contains:

domain_auto_trans(initrc_t, sendmail_exec_t, sendmail_t)

So as long as your depinit starts off in something that can get to initrc_t,
the kernel doesn't *CARE*.  I if an initrc_t fork/exec a sendmail_exec_t, and
it transitions to the needed domain.

Hell, I suspect that if you just put a:

/sbin/depinit --	system_u:object_r:initrc_exec_t

in file_contexts, 98% of it will Just Work.

Attachment: pgpQx9PrpifMY.pgp
Description: PGP signature



This mailing list archive is a service of Copilot Consulting.