Xan wrote:
Does this remove that entropy from the host, or is the same host entropy sent to every Linode?
Could open up some interesting intra-host Linode attacks. I doubt if any would ever be practical though.
It removes entropy from the host -- there's no sharing or duplicating of identical entropy between Linodes.
Xan wrote:
Does a single host have enough entropy for 32 (or whatever) machines?
That's to be seen. The reasoning behind this patch is that the host, in general, will generate much more entropy because of all the activity that happens on the host, compared to the little entropy generate by UML's drivers. I also set the pool size very high.
-Chris