On 12/14/2015 01:05 PM, John Snow wrote:

> I was actually contemplating re-spinning this for a v3:
> 
> Instead of having a "typeA" and "typeB" properties of the FDC, I'll just
> spin the properties in such a way that they write directly to
> FDCtrl.drives[n].drive, which avoids the need for two new members and a
> method designed to "fetch" the default from the controller.
> 
> I also want to add a "fallback" member to the FDC as a CLI configurable
> parameter such that when using "auto" you can configure directly what
> type of drive you'll get.
> 
> This way the default behavior will be "auto" (but configurable) and the
> fallback if the drive is empty or it cannot make a confident guess will
> be whatever the user chose as "fallback" -- presumed "144" before this
> series and "288" afterwards.
> 
> I believe this also opens up the possibility for having "fallback"
> default to "144" for machine types created prior to 2.6 ... I think.
> (I'm less familiar with machine compat code as of yet.)
> 
> Sound reasonable?

On paper, yes that sounds reasonable. I'm also not familiar enough with
machine types to know if it will let you keep 2.5 and earlier machines
with a fallback of 144, and newer machines with a fallback of 288, but
sounds promising.


-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org