On Wed, 25 Nov 2020 12:34:25 +0100 Cédric Le Goater <c...@kaod.org> wrote:
> > >>> This complexifies migration because we have to guess at > >>> post load if we should claim the IPI in KVM or not. The > >>> simple presence of the vCPU isn't enough : we need to > >>> guess if the guest actually configured the IPI or not. > >> > >> The EAT will be transferred from the source and the call to > >> kvmppc_xive_source_reset_one() should initialize the KVM > >> device correctly on the target for all interrupts. > >> > > > > Except that the EAS appears as valid for all IPIs, even > > though the source didn't claim them at the KVM level. > > why ? we would stop claiming IPIs in spapr_irq_init() and so > they won't appear as being valid anymore, at boot time or > restore time. > If we don't claim the IPIs in spapr_irq_init() anymore then we must at least claim them on the path of H_INT_GET_SOURCE_INFO otherwise it will fail with H_P2 and the guest won't even try to setup the IPI. Even if we do that, we still have a window where the source is valid in QEMU but not yet at the KVM level. > > C.