<div dir="ltr">On 10 April 2013 02:44, Guillaume Thouvenin <span dir="ltr"><<a href="mailto:guillaume.thouvenin@bull.net" target="_blank">guillaume.thouvenin@bull.net</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div>On 10/04/2013 11:32, Daniel P. Berrange wrote:<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Oh, well that is LXC, rather than KVM so you have pretty<br>
much zero security in anything.<br>
</blockquote>
<br></div>
Yes sure. In my mind the idea was to start one container for one physical machine. So the solution is more comparable to bare metal provisioning than hypervisors based solution. But when IOMMU is not available maybe it can be interesting.<br>
</blockquote><div><br></div><div>There's two things here.<br><br></div><div>Firstly, it's increasingly rare to find new server hardware missing IOMMUs, so you may be solving a problem that most people don't have. That's up to you...<br>
<br></div><div>Secondly, have you passed intel_iommu=on to your kernel (or whatever the right flag is)? I know the Ubuntu kernel, at least, needs the extra boot parameter to enable it.<br></div></div></div></div>