On Thu, 2004-01-29 at 23:42, Fernando Luis Vázquez Cao wrote:
> Due to the latter, I have started to split my server into two
> executables so that there's a listening server that creates children in
> the required security context (using "execve").

Good.

> Anyway, I soon found that the resulting code is not as easy to maintain
> as the original one and that I have to make a fairly big amount of
> changes. Besides, I think this is not the right approach, since in this
> case, the use of SELinux is really having a big impact in the projects
> code.

Restructuring applications to support least privilege and isolation is a good thing.

> I was wandering if allowing to fork a child in a security context
> different from the parent's (without using execve) is a good idea or
> not, since I find it useful and, anyway, it can be achieved splitting a
> program in two and using "execv".

It isn't a good idea. Some prior discussions of this issue can be found at the below links:  

http://marc.theaimsgroup.com/?l=selinux&m=107272253121858&w=2
http://marc.theaimsgroup.com/?l=selinux&m=106910931928255&w=2
http://marc.theaimsgroup.com/?l=selinux&m=104939744006086&w=2
http://marc.theaimsgroup.com/?l=selinux&m=102587211414608&w=2
http://marc.theaimsgroup.com/?l=selinux&m=101343633303503&w=2

-- 
Stephen Smalley <sds@epoch.ncsc.mil>
National Security Agency


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to majordomo@tycho.nsa.gov with
the words "unsubscribe selinux" without quotes as the message.