libvirt/docs/tlscerts.html.in

414 lines
13 KiB
XML

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html>
<html xmlns="http://www.w3.org/1999/xhtml">
<body>
<h1>TLS x509 certificate setup</h1>
<ul id="toc"></ul>
<h2>
<a id="Remote_PKI">Public Key Infrastructure set up</a>
</h2>
<p>
If you are unsure how to create TLS certificates, skip to the
next section.
</p>
<table class="top_table">
<tr>
<th> Location </th>
<th> Machine </th>
<th> Description </th>
<th> Required fields </th>
</tr>
<tr>
<td>
<code>/etc/pki/CA/cacert.pem</code>
</td>
<td> Installed on the client and server </td>
<td> CA's certificate (<a href="#Remote_TLS_CA">more info</a>)</td>
<td> n/a </td>
</tr>
<tr>
<td>
<code>$HOME/.pki/cacert.pem</code>
</td>
<td> Installed on the client </td>
<td> CA's certificate (<a href="#Remote_TLS_CA">more info</a>)</td>
<td> n/a </td>
</tr>
<tr>
<td>
<code>/etc/pki/libvirt/private/serverkey.pem</code>
</td>
<td> Installed on the server </td>
<td> Server's private key (<a href="#Remote_TLS_server_certificates">more info</a>)</td>
<td> n/a </td>
</tr>
<tr>
<td>
<code>/etc/pki/libvirt/servercert.pem</code>
</td>
<td> Installed on the server </td>
<td> Server's certificate signed by the CA.
(<a href="#Remote_TLS_server_certificates">more info</a>) </td>
<td> CommonName (CN) must be the hostname of the server as it
is seen by clients. All hostname and IP address variants that might
be used to reach the server should be listed in Subject Alt Name
fields.</td>
</tr>
<tr>
<td>
<code>/etc/pki/libvirt/private/clientkey.pem</code>
</td>
<td> Installed on the client </td>
<td> Client's private key. (<a href="#Remote_TLS_client_certificates">more info</a>) </td>
<td> n/a </td>
</tr>
<tr>
<td>
<code>/etc/pki/libvirt/clientcert.pem</code>
</td>
<td> Installed on the client </td>
<td> Client's certificate signed by the CA
(<a href="#Remote_TLS_client_certificates">more info</a>) </td>
<td> Distinguished Name (DN) can be checked against an access
control list (<code>tls_allowed_dn_list</code>).
</td>
</tr>
<tr>
<td>
<code>$HOME/.pki/libvirt/clientkey.pem</code>
</td>
<td> Installed on the client </td>
<td> Client's private key. (<a href="#Remote_TLS_client_certificates">more info</a>) </td>
<td> n/a </td>
</tr>
<tr>
<td>
<code>$HOME/.pki/libvirt/clientcert.pem</code>
</td>
<td> Installed on the client </td>
<td> Client's certificate signed by the CA
(<a href="#Remote_TLS_client_certificates">more info</a>) </td>
<td> Distinguished Name (DN) can be checked against an access
control list (<code>tls_allowed_dn_list</code>).
</td>
</tr>
</table>
<p>
If 'pkipath' is specified in URI, then all the client
certificates must be found in the path specified, otherwise the
connection will fail with a fatal error. If 'pkipath' is not
specified:
</p>
<ul>
<li> For a non-root user, libvirt tries to find the certificates
in $HOME/.pki/libvirt first. If the required CA certificate cannot
be found, then the global default location
(/etc/pki/CA/cacert.pem) will be used.
Likewise, if either the client certificate
or the client key cannot be found, then the global default
locations (/etc/pki/libvirt/clientcert.pem,
/etc/pki/libvirt/private/clientkey.pem) will be used.
</li>
<li> For the root user, the global default locations will always be used.</li>
</ul>
<h2>
<a id="Remote_TLS_background">Background to TLS certificates</a>
</h2>
<p>
Libvirt supports TLS certificates for verifying the identity
of the server and clients. There are two distinct checks involved:
</p>
<ul>
<li> The client should know that it is connecting to the right
server. Checking done by client by matching the certificate that
the server sends to the server's hostname. May be disabled by adding
<code>?no_verify=1</code> to the
<a href="#Remote_URI_parameters">remote URI</a>.
</li>
<li> The server should know that only permitted clients are
connecting. This can be done based on client's IP address, or on
client's IP address and client's certificate. Checking done by the
server. May be enabled and disabled in the <a href="#Remote_libvirtd_configuration">libvirtd.conf file</a>.
</li>
</ul>
<p>
For full certificate checking you will need to have certificates
issued by a recognised <a href="https://en.wikipedia.org/wiki/Certificate_authority">Certificate
Authority (CA)</a> for your server(s) and all clients. To avoid the
expense of getting certificates from a commercial CA, you can set up
your own CA and tell your server(s) and clients to trust certificates
issues by your own CA. Follow the instructions in the next section.
</p>
<p>
Be aware that the <a href="#Remote_libvirtd_configuration">default
configuration for libvirtd</a> allows any client to connect provided
they have a valid certificate issued by the CA for their own IP
address. You may want to change this to make it less (or more)
permissive, depending on your needs.
</p>
<h2>
<a id="Remote_TLS_CA">Setting up a Certificate Authority (CA)</a>
</h2>
<p>
You will need the <a href="https://www.gnutls.org/manual/html_node/certtool-Invocation.html">GnuTLS
certtool program documented here</a>. In Fedora, it is in the
<code>gnutls-utils</code> package.
</p>
<p>
Create a private key for your CA:
</p>
<pre>
certtool --generate-privkey &gt; cakey.pem
</pre>
<p>
and self-sign it by creating a file with the
signature details called
<code>ca.info</code> containing:
</p>
<pre>
cn = <i>Name of your organization</i>
ca
cert_signing_key
</pre>
<pre>
certtool --generate-self-signed --load-privkey cakey.pem \
--template ca.info --outfile cacert.pem
</pre>
<p>
(You can delete <code>ca.info</code> file now if you
want).
</p>
<p>
Now you have two files which matter:
</p>
<ul>
<li><code>cakey.pem</code> - Your CA's private key (keep this very secret!)
</li>
<li><code>cacert.pem</code> - Your CA's certificate (this is public).
</li>
</ul>
<p><code>cacert.pem</code> has to be installed on clients and
server(s) to let them know that they can trust certificates issued by
your CA.
</p>
<p>
The normal installation directory for <code>cacert.pem</code>
is <code>/etc/pki/CA/cacert.pem</code> on all clients and servers.
</p>
<p>
To see the contents of this file, do:
</p>
<pre><b>certtool -i --infile cacert.pem</b>
X.509 certificate info:
Version: 3
Serial Number (hex): 00
Subject: CN=Libvirt Project
Issuer: CN=Libvirt Project
Signature Algorithm: RSA-SHA
Validity:
Not Before: Mon Jun 18 16:22:18 2007
Not After: Tue Jun 17 16:22:18 2008
<i>[etc]</i>
</pre>
<p>
This is all that is required to set up your CA. Keep the CA's private
key carefully as you will need it when you come to issue certificates
for your clients and servers.
</p>
<h2>
<a id="Remote_TLS_server_certificates">Issuing server certificates</a>
</h2>
<p>
For each server (libvirtd) you need to issue a certificate
containing one or more hostnames and/or IP addresses.
Historically the CommonName (CN) field would contain the
hostname of the server and would match the hostname used
in the URI that clients pass to libvirt. In most TLS implementations
the CN field is considered legacy data. The preferential mechanism
is to use Subject Alt Name (SAN) extension fields to validate
against. In the future use of the CN field for validation may be
discontinued entirely, so it is strongly recommended to
include the SAN fields.
</p>
<p>
In the example below, clients will be connecting to the
server using a <a href="#Remote_URI_reference">URI</a> of
<code>qemu://compute1.libvirt.org/system</code>, so the CN
must be "<code>compute1.libvirt.org</code>".
</p>
<p>
Make a private key for the server:
</p>
<pre>
certtool --generate-privkey &gt; serverkey.pem
</pre>
<p>
and sign that key with the CA's private key by first
creating a template file called <code>server.info</code>.
The template file will contain a number of fields to define
the server as follows:
</p>
<pre>
organization = <i>Name of your organization</i>
cn = compute1.libvirt.org
dns_name = compute1
dns_name = compute1.libvirt.org
ip_address = 10.0.0.74
ip_address = 192.168.1.24
ip_address = 2001:cafe::74
ip_address = fe20::24
tls_www_server
encryption_key
signing_key
</pre>
<p>
The 'cn' field should refer to the fully qualified public
hostname of the server. For the SAN extension data, there
must also be one or more 'dns_name' fields that contain all
possible hostnames that can be reasonably used by clients
to reach the server, both with and without domain name
qualifiers. If clients are likely to connect to the server
by IP address, then one or more 'ip_address' fields should
also be added.
</p>
<p>
Use the template file as input to a <code>certtool</code>
command to sign the server certificate:
</p>
<pre>
certtool --generate-certificate --load-privkey serverkey.pem \
--load-ca-certificate cacert.pem --load-ca-privkey cakey.pem \
--template server.info --outfile servercert.pem
</pre>
<p>
This gives two files:
</p>
<ul>
<li><code>serverkey.pem</code> - The server's private key.
</li>
<li><code>servercert.pem</code> - The server's public key.
</li>
</ul>
<p>
We can examine this certificate and its signature:
</p>
<pre><b>certtool -i --infile servercert.pem</b>
X.509 certificate info:
Version: 3
Serial Number (hex): 00
Subject: O=Libvirt Project,CN=compute1.libvirt.org
Issuer: CN=Libvirt Project
Signature Algorithm: RSA-SHA
Validity:
Not Before: Wed Oct 04 09:09:44 UTC 2017
Not After: Thu Oct 04 09:09:44 UTC 2018
Extensions:
Basic Constraints (critical):
Certificate Authority (CA): FALSE
Subject Alternative Name (not critical):
DNSname: compute1
DNSname: compute1.libvirt.org
IPAddress: 10.0.0.74
IPAddress: 192.168.1.24
IPAddress: 2001:cafe::74
IPAddress: fe20::24
</pre>
<p>
Note the "Issuer" CN is "Libvirt Project" (the CA) and
the "Subject" CN is "compute1.libvirt.org" (the server).
Notice that the hostname listed in the CN must also
be duplicated as a DNSname entry
</p>
<p>
Finally we have two files to install:
</p>
<ul>
<li><code>serverkey.pem</code> is
the server's private key which should be copied to the
server <i>only</i> as
<code>/etc/pki/libvirt/private/serverkey.pem</code>.
</li>
<li><code>servercert.pem</code> is the server's certificate
which can be installed on the server as
<code>/etc/pki/libvirt/servercert.pem</code>.
</li>
</ul>
<h2>
<a id="Remote_TLS_client_certificates">Issuing client certificates</a>
</h2>
<p>
For each client (ie. any program linked with libvirt, such as
<a href="https://virt-manager.org/">virt-manager</a>)
you need to issue a certificate with the X.509 Distinguished Name (DN)
set to a suitable name. You can decide this on a company / organisation
policy. For example:
</p>
<pre>
C=GB,ST=London,L=London,O=Libvirt Project,CN=<i>name_of_client</i>
</pre>
<p>
The process is the same as for
<a href="#Remote_TLS_server_certificates">setting up the
server certificate</a> so here we just briefly cover the
steps.
</p>
<ol>
<li>
Make a private key:
<pre>
certtool --generate-privkey &gt; clientkey.pem
</pre>
</li>
<li>
Act as CA and sign the certificate. Create client.info containing:
<pre>
country = GB
state = London
locality = London
organization = Libvirt Project
cn = client1
tls_www_client
encryption_key
signing_key
</pre>
and sign by doing:
<pre>
certtool --generate-certificate --load-privkey clientkey.pem \
--load-ca-certificate cacert.pem --load-ca-privkey cakey.pem \
--template client.info --outfile clientcert.pem
</pre>
</li>
<li>
Install the certificates on the client machine:
<pre>
cp clientkey.pem /etc/pki/libvirt/private/clientkey.pem
cp clientcert.pem /etc/pki/libvirt/clientcert.pem
</pre>
</li>
</ol>
<h2>
<a id="Remote_TLS_troubleshooting">Troubleshooting TLS certificate problems</a>
</h2>
<dl>
<dt> failed to verify client's certificate </dt>
<dd>
<p>
On the server side, run the libvirtd server with
the '--listen' and '--verbose' options while the
client is connecting. The verbose log messages should
tell you enough to diagnose the problem.
</p>
</dd>
</dl>
<p> You can use the virt-pki-validate shell script
to analyze the setup on the client or server machines, preferably as root.
It will try to point out the possible problems and provide solutions to
fix the set up up to a point where you have secure remote access.</p>
</body>
</html>