Re: [PATCH v5 1/3] x86/viridian: Re-purpose the HVM parameter to be a feature mask

[Ian Campbell <ian.campbell@citrix.com>]

On Tue, 2014-08-05 at 14:07 +0100, Paul Durrant wrote:

Sorry for the delay replying, I've been on vacation and travelling.

> diff --git a/tools/libxl/libxl_types.idl b/tools/libxl/libxl_types.idl
> index a412f9c..6c5d7e0 100644
> --- a/tools/libxl/libxl_types.idl
> +++ b/tools/libxl/libxl_types.idl
> @@ -363,7 +363,7 @@ libxl_domain_build_info = Struct("domain_build_info",[
>                                         ("acpi_s3",          libxl_defbool),
>                                         ("acpi_s4",          libxl_defbool),
>                                         ("nx",               libxl_defbool),
> -                                       ("viridian",         libxl_defbool),
> +                                       ("viridian",         libxl_string_list),

Unfortunately this will break the API. specifically existing  apps which
call libxl_defbool_set on this field will now get a type error.

You need to keep the existing viridian field and add a new one with the
new semantics and a new name (e.g. viridian_features, or a set of new
defbools as I suggested in my previous mail) and define the precedence.
See usbdevice vs. usbdevice_list or the (very new) serial one for
existing examples of this sort of thing.

For your case though you could continue to treat the viridian defbool as
a gate on a bunch of viridian_feature_X defbools, IOW if viridian ==
false the others are ignored, if it is true then the others have the
usual defbool behaviour of defaults and overrides. I think that would
fall out fairly naturally in the setdefault function.

Ian.