On 07/23/2018 03:06 PM, Janosch Frank wrote: > On 23.07.2018 14:42, Christian Borntraeger wrote: >> Makes a lot of sense and it also improves the skey performane for the hlp patch series. >> Does it make sense to let this be part of the next version of the host large pages >> series or shall I take this patch directly? > > I'd prefer to have it separately. > >>> if (kvm_s390_test_cpuflags(vcpu, CPUSTAT_KSS)) >>> kvm_s390_clear_cpuflags(vcpu, CPUSTAT_KSS); >>> - if (!vcpu->kvm->arch.use_skf) >>> - sie_block->ictl |= ICTL_ISKE | ICTL_SSKE | ICTL_RRBE; >>> - else >>> - sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE); > > I'm not allowed to remove both lines here. > If we have kss but !skf, we'd never get any intercepts as we either set > ICTLS or KSS in the vcpu setup... I.e. hpage with KSS would be broken. So you are going to send a v3? > > >>> + if (vcpu->kvm->arch.use_skf) >>> + vcpu->arch.sie_block->ictl &= ~(ICTL_ISKE | ICTL_SSKE | ICTL_RRBE); >>> + vcpu->arch.skey_enabled = true; >>> return 0; >>> } >>> >>> > >
