On Tue, 24 Sep 2019 07:22:51 +0200 Cédric Le Goater <c...@kaod.org> wrote:
> On 24/09/2019 06:59, David Gibson wrote: > > The XICS interrupt controller device used to have separate subtypes > > for the KVM and non-KVM variant of the device. That was a bad idea, > > because it leaked information that should be entirely host-side > > implementation specific to the kinda-sorta guest visible QOM class > > names. > > > > We eliminated the KVM specific class some time ago, but it's left > > behind a distinction between the TYPE_ICS_BASE abstract class and > > TYPE_ICS_SIMPLE subtype which no longer serves any purpose. > > > > This series collapses the two types back into one. > > OK. Is it migration compatible ? because this is why we have kept Yeah, the types themselves don't matter, only the format of the binary stream described in the VMStateDescription does. > this subclass. I am glad we can remove it finally. > I was thinking we were keeping that for pnv... > Thanks, > > C. > > > > > > David Gibson (4): > > xics: Eliminate 'reject', 'resend' and 'eoi' class hooks > > xics: Merge reset and realize hooks > > xics: Rename misleading ics_simple_*() functions > > xics: Merge TYPE_ICS_BASE and TYPE_ICS_SIMPLE classes > > > > hw/intc/trace-events | 10 +- > > hw/intc/xics.c | 211 ++++++++++++++---------------------------- > > hw/intc/xics_spapr.c | 12 +-- > > hw/ppc/pnv_psi.c | 6 +- > > hw/ppc/spapr_irq.c | 6 +- > > include/hw/ppc/xics.h | 30 +----- > > 6 files changed, 92 insertions(+), 183 deletions(-) > > >