Received-SPF: pass (google.com: best guess record for domain of selinux-refpolicy-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Dominick Grift <dominick.grift@defensec.nl>
To:     Russell Coker <russell@coker.com.au>
Cc:     Chris PeBenito <pebenito@ieee.org>,
        selinux-refpolicy@vger.kernel.org
Subject: Re: new certbot patch
References: <20200405084141.GA177560@xev>
        <5b70567f-d551-ea5f-50e4-5febe2ad9a09@ieee.org>
        <4305733.qMCtAaFjtT@liv>
Date:   Fri, 10 Apr 2020 09:55:16 +0200
In-Reply-To: <4305733.qMCtAaFjtT@liv> (Russell Coker's message of "Fri, 10 Apr
        2020 15:56:26 +1000")
Message-ID: <ypjlblnzn5az.fsf@defensec.nl>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.3 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Sender: selinux-refpolicy-owner@vger.kernel.org
Precedence: bulk

Russell Coker <russell@coker.com.au> writes:

> On Thursday, 9 April 2020 11:23:00 PM AEST Chris PeBenito wrote:
>> > +miscfiles_read_generic_certs(certbot_t)
>> > +miscfiles_manage_generic_tls_privkey_dirs(certbot_t)
>> > +miscfiles_manage_generic_tls_privkey_files(certbot_t)
>> > +miscfiles_manage_generic_tls_privkey_lnk_files(certbot_t)
>> 
>> Perhaps we should be moving towards having a specific label for these
>> private keys instead.  It seems logical that there would be multiple types
>> of private keys.  Then have a miscfiles_private_key() to declare one and
>> have the type in this module to act on directly.
>
> Certbot isn't written to support different runs on the same system.  It might 
> be worth filing an upstream feature request for that as it would be a useful 
> feature.
>
> As for SE Linux policy to support multiple separate private SSL keys on the 
> same system, it seems that there would be many variations on that and trying 
> to write generic policy wouldn't be viable.  Maybe a better solution would be 
> to support different MCS categories for different daemons and then different 
> categories for private keys.  Then the sysadmin would have full control over 
> which daemons could access which private keys.

A more practical approach here in my experience is to not give access to
certs in /etc/letsencrypt but let the hook functionality copy the certs
from the store and then address labeling with "cert_type()" in the
accessible location. Not ideal either but the way letsencrypt maintains its
certs in /etc/letsencrypt is not very usable either.

Eventually one might end up altering/combining the certs anyway's. For
example znc seems to require that you enclose the privkey with the chain.

-- 
gpg --locate-keys dominick.grift@defensec.nl
Key fingerprint = FCD2 3660 5D6B 9D27 7FC6  E0FF DA7E 521F 10F6 4098
https://sks-keyservers.net/pks/lookup?op=get&search=0xDA7E521F10F64098
Dominick Grift