On Fri, Jul 27, 2012 at 10:32 AM, Marc Zyngier <marc.zyngier@xxxxxxx> wrote: > On 26/07/12 19:50, Christoffer Dall wrote: >> On Thu, Jul 26, 2012 at 5:30 PM, Marc Zyngier <marc.zyngier@xxxxxxx> wrote: >>> On 25/07/12 15:17, Christoffer Dall wrote: >>>> list rename, shot myself in the foot. >>>> >>>> >>>>>> + } >>>>>> + >>>>>> + /************************************************************** >>>>>> + * Enter the guest >>>>>> + */ >>>>>> + trace_kvm_entry(vcpu->arch.regs.pc); >>>>>> + kvm_guest_enter(); >>>>>> + vcpu->mode = IN_GUEST_MODE; >>>>>> + >>>>>> + ret = __kvm_vcpu_run(vcpu); >>>>>> + >>>>>> + vcpu->mode = OUTSIDE_GUEST_MODE; >>>>>> + vcpu->stat.exits++; >>>>> >>>>> The tracepoint above should be sufficient for statistics. >>>>> >>>> >>>> Marc added this one, so I assume there's a valid need for the separate >>>> counter. Marc? >>> >>> Probably because I've been too lazy to setup a trace infrastructure on >>> my test boxes. Stats are easier to access, but seeing now that they've >>> been deprecated and scheduled for removal, we probably shouldn't bother >>> with them. >>> >> you want to take them all out then? > > Looks like all stats will be dropped from KVM by 2013, if I trust > Documentation/feature-removal-schedule.txt: > > What: KVM debugfs statistics > When: 2013 > Why: KVM tracepoints provide mostly equivalent information in a much more > flexible fashion. > ok, they will go then. _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
