[CentOS] Odd failure of smbd to start from init.d - CentOS 5.4 - it's that fine SELinux

Wed May 26 01:27:20 UTC 2010
Whit Blauvelt <whit at transpect.com>

On Tue, May 25, 2010 at 07:46:56PM -0500, Les Mikesell wrote:

> I would have looked at selinux first for any "odd failure", but I thought it 
> related to the process itself and couldn't see any way that the process would be 
> different when started as "sh /etc/init.d/smb restart" than simply 
> /etc/init.d/smb restart.  Is it?

That selinux would prevent a normal init.d startup of a common daemon like
smbd, but allow the same startup in several other ways ... okay, I've never
studied selinux. I usually run Ubuntu on servers. I've pretty much literally
inherited a bunch of RH-based servers to admin (coworker sadly died), and
we're adding more to run in parallel, so CentOS was obvious (RH-the-firm
being so badly run it took staff days over the phone just to buy a single
new license from them). Of course AppArmour can also get in the way, but at
least it logs such actions, so it's obvious if you need to reconfig or turn
it off.

I'm solidly impressed with this list. Nothing like it for Ubuntu, and back
when Gentoo was my preferred server distro there was more noise surrounding
that too. It shows that the interest in CentOS is entirely professional. So
that's a strong upside.

But if someone can tell me why selinux thinks it's sane to block
"/etc/init.d/smb start" while leaving "sh /etc/init.d/smb start" and even
/some/random/dir/smb start" wide open ... I just can't believe some happy
hacker at NSA thought that would count as a security scheme. Really, I'd
like to know how this is supposed to be useful.