Anthony,

If there's read and write access, it needs to be part of the savevm state.  Is it per-cpu?  If it's write only, then it doesn't need to be persisted in savevm.

I did implement and test the savevm/loadvm functionality.

-Jordan

On Mon, Jun 29, 2009 at 11:39 AM, Anthony Liguori <anthony@codemonkey.ws> wrote:
Avi Kivity wrote:
I still think the read/write port route would
* better emulate systems,
* be more flexible (allowing software the option to read it),
* and, be easier to implement :)

But, I think the most important part is to make the data accessible somehow.  So, the monitor access method would work fine as well.

I don't object to read/write access.

If there's read and write access, it needs to be part of the savevm state.  Is it per-cpu?  If it's write only, then it doesn't need to be persisted in savevm.

I'd lean toward write-only unless there was a compelling reason to make it read/write.

Regards,

Anthony Liguori