On Wed, 11 Jul 2018 18:48:21 +0200 Paolo Bonzini <pbonz...@redhat.com> wrote:
> On 11/07/2018 17:22, Igor Mammedov wrote: > > It also seems wrong to call _plug handler on maybe partially > > initialized device so perhaps we should first finish devices/children > > realization then do reset and only after that call _plug() handler > > I agree but this is too dangerous until we look at plug() handlers and I'll put on my TODO list for 3.1 > remove the usage of Error **. considering we have _per_plug now with all checks, we probably would be able to remove Error** from _plug but only after splitting/auditing existing _plug callbacks. > Introducing a new callback helps the > transition. sure, we can drop it later > Paolo >