On 02/11/2014 12:19 AM, Paolo Bonzini wrote: > Il 11/02/2014 03:47, Luiz Capitulino ha scritto: >> On Tue, 11 Feb 2014 00:34:37 +0100 >> Paolo Bonzini wrote: >> >>> Il 11/02/2014 00:30, Laszlo Ersek ha scritto: >>>>>> However, I don't see the point in having the "format-option" >>>>>> field. What >>>>>> about: >>>>>> >>>>>> -> { "execute": "query-dump-guest-memory-capabilities" } >>>>>> <- { "return": { "formats": >>>>>> ["elf", "kdump-zlib", "kdump-lzo", "kdump-snappy"] } >>>> Technically you might be right. However, this partial introspection >>>> feature is entirely a wart >> Yes, I can. But what's the problem with patch 13? For me having Eric's >> ACK is enough for applying it. Anything else will be done by QMP >> introspection. > > I'm suggesting another, more streamlined format; see above. I like Paolo's suggestion of a more streamlined format. With migration introspection, we could do two things - learn the set of knobs available (query-migrate-capabilities), AND change those knobs (migrate-set-capabilities). Thus, each knob had to be listed as a struct that showed both the name of the knob and the current status. But here, we are not changing the set of format options on the fly. The simpler array of string names supported is sufficient; it's less work for libvirt to get an array of strings than it is to get an array of structs only to then determine the name from the struct and the redundant boolean that all such names are supported. From libvirt's perspective, I think patch 13/13 needs to be respun into the simpler format. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org