Total system freeze when using geli

[Robert Simmons]

I've run into a total freeze with 9.0.  I'm using geli encrypted swap
and / with the following:
geli init -b -v -a hmac/sha256 -l 256 -s 4096 /dev/ada0p4
and
geli_swap_flags="-d -l 256 -s 4096"

The following specs: 128M memory and 4G hard disk in a Xen HMV virtual
server environment.  I have allocated 384M of swap space.

After install when I run the freebsd-update fetch command the server
freezes at the following stage:
Applying patches...

I've been trying to diagnose the problem by running top and watching
what happens during this stage.  I noticed the following:

1) the box runs out of physical memory at this stage (totally
expected, that's why there is sufficient swap space).
2) All the processes except 2 sleep:
31 processes:  1 running, 29 sleeping, 1 waiting
3) the box is responsive to hitting enter at the console (it produces
another login: prompt)
4) sshd is asleep, so I can't ssh into the box
5) if I try to login to the console, it lets me enter a username then
locks up totally, it does not present me with a password: prompt.
6) it has not run out of swap, nowhere close:
Mem: 54M Active, 9524K Inact, 41M Wired, 24K Cache, 21M Buf, 32K Free
Swap: 384M Total, 6452K Used, 378M Free, 1% Inuse
7) the moment it runs out of physical memory it begins being unresponsive

I've brought this thread from freebsd-hackers because after testing
this exact same environment and conditions but without using geli, I
am unable to reproduce the freeze.  With geli, I'm able to produce the
freeze 100% of the time.  It was suggested in hackers that there is a
logjam in the buffer cache.

Any thoughts?