On 20/03/13 20:04, Christopher Covington wrote:

Hi Marc,

On 03/13/2013 03:59 PM, Christopher Covington wrote:

[...]

quoted

Alternatively, you could consider storing the host registers in a
slimmed-down vcpu structure for hosts, rather than on the stack.

I am actively implementing this (I'm turning the vfp_host pointer into a
full blown CPU context). It looks promising so far, stay tuned.

One potential argument for storing the host in the same sort of vcpu structure
as the guest rather than on the hypervisor stack is that snapshot and
migration support initially intended for guests might more easily be extended
to work for hosts as well.

Not sure I'm following you here. Are you thinking of snapshoting both
host and guests, and migrating the whole thing? Ambitious... ;-)

	M.
-- 
Jazz is not dead. It just smells funny...