Re: [Qemu-devel] Re: [PATCH 0/4] Machine config files

[Anthony Liguori]

Avi Kivity wrote:
> > That alone will *not* provide the position in the device tree 
> > though.  And I suspect making the transition without breaking 
> > backward compatibility could be quite tricky too.  So I'm not sure 
> > how useful that actually is, although it surely would be much cleaner 
> > from a design point of view.
>
> I think we need to break the format.  It's semi-broken anyway wrt 
> hotplug and the use of ram addresses.  It may be possible to add 
> transition code but I don't think it's worth it.

The format is fine.  ram v3 uses ram_addr_t because it's was the only 
sane thing to do at the time.  v4 could easily use target_phys_t.

Most importantly, machine config should not be sent until non-live 
migration kicks in.  At this point, the device tree is stable because 
the guest isn't running.  The device tree should be the first section 
saved for non-live migration.

This means you need to avoid initializing the device model until after 
you get to the second step in migration.  You really just need to 
register any live migration handlers outside of the device model 
creation which pretty much already happens today.

Regards,

Anthony Liguori