On Mon, Jun 04, 2001 at 03:14:17PM -0700, Michael K. Sanders wrote:

 > This sounds great for applications such as raw Oracle databases, 
 > Coda RVM, etc.  But for those applications, the data must be persistent,
 > and you don't really want that for swap.

Well, for swap, you'd obviously want to use a random key, at very least.
I haven't read the paper yet, so I don't know how many keys they use for
swap... but you could do it per-swapdev, or even something like per-thing-
that-is-backed-by-swap, i.e. generate a random key whenever you assign
a "swapslot" to an object/anon.

 > >Now, this was supposedly implemented by one of the TNF developers, then
 > >"lost in the ether"...?
 > 
 > Pity.

A fresh implementation of a cipher block device should take all of an
evening or two to write.  Maybe I'll write one this week.

-- 
        -- Jason R. Thorpe <thorpej@zembu.com>