Re: [PATCH v2 2/3] qdev-properties: Add DEFINE_PROP_ON_OFF_AUTO_BIT64()

["Daniel P. Berrangé" <berrange@redhat.com>]

On Thu, Oct 31, 2024 at 04:21:53PM +0900, Akihiko Odaki wrote:
> On 2024/10/29 1:50, Daniel P. Berrangé wrote:
> > On Tue, Oct 22, 2024 at 01:50:39PM +0900, Akihiko Odaki wrote:
> > > DEFINE_PROP_ON_OFF_AUTO_BIT64() corresponds to DEFINE_PROP_ON_OFF_AUTO()
> > > as DEFINE_PROP_BIT64() corresponds to DEFINE_PROP_BOOL(). The difference
> > > is that DEFINE_PROP_ON_OFF_AUTO_BIT64() exposes OnOffAuto instead of
> > > bool.
> > 
> > Again, same feedback as last time.
> > 
> > With this design, existing users of DEFINE_PROP_BIT64 that
> > get converted to DEFINE_PROP_ON_OFF_AUTO_BIT64, in addition
> > to gaining the desired "auto" value, also gain redundant
> > 'on' and 'off' values as side-effects.
> > 
> > In the next patch, the stated problem is that virtio code
> > needs to distinguish between bits that are user set, and
> > bits that are set based on available host backend features.
> > 
> > That doesn't seem to require us to accept any new values
> > from the user. It should be sufficient to track user
> > specified features, separately from user specified values.
> > ie when parsing user input for bitfields, we need to parse
> > into a pair of fields
> > 
> >    uint64_t has_user_features;  /* which bits were specified */
> >    uint64_t user_features;      /* values of specified bits*/
> 
> Properties also have getters. We don't know what to return with them without
> the new value.

The virtio changes in the next patch are just accessing the bitsets
directly. A getter could just be made to return false for unset
values, on the assumption that any caller should be checking the
has_user_features bits beforehand.

With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|