On Mon, 18 Feb 2019 12:03:05 +0000
Andrew Murray <andrew.murray at arm.com> wrote:
> On Thu, Feb 14, 2019 at 11:42:15AM +0000, Suzuki K Poulose wrote:
> >
> >
> > On 05/02/2019 14:33, Julien Thierry wrote:
> > > Hi Andrew,
> > >
> > > On 04/02/2019 16:53, Andrew Murray wrote:
> > > > Emulate chained PMU counters by creating a single 64 bit event counter
> > > > for a pair of chained KVM counters.
> > > >
> > > > Signed-off-by: Andrew Murray <andrew.murray at arm.com>
> > > > ---
> > > > include/kvm/arm_pmu.h | 1 +
> > > > virt/kvm/arm/pmu.c | 321 +++++++++++++++++++++++++++++++++++++++++---------
> > > > 2 files changed, 269 insertions(+), 53 deletions(-)
> > > >
> > > > diff --git a/include/kvm/arm_pmu.h b/include/kvm/arm_pmu.h
> > > > index b73f31b..8e691ee 100644
> > > > --- a/include/kvm/arm_pmu.h
> > > > +++ b/include/kvm/arm_pmu.h
> > > > @@ -29,6 +29,7 @@ struct kvm_pmc {
> > > > u8 idx; /* index into the pmu->pmc array */
> > > > struct perf_event *perf_event;
> > > > u64 bitmask;
> > > > + u64 overflow_count;
> > > > };
> > > > struct kvm_pmu {
> > > > diff --git a/virt/kvm/arm/pmu.c b/virt/kvm/arm/pmu.c
> > > > index a64aeb2..9318130 100644
> > > > --- a/virt/kvm/arm/pmu.c
> > > > +++ b/virt/kvm/arm/pmu.c
> > > > @@ -24,9 +24,25 @@
> > > > #include <kvm/arm_pmu.h>
> > > > #include <kvm/arm_vgic.h>
> > > > +#define ARMV8_PMUV3_PERFCTR_CHAIN 0x1E
> > >
> > > I find it a bit awkward to have this redefined here.
> > >
> > > Maybe we could define a helper in kvm_host.h:
> > > bool kvm_pmu_typer_is_chain(u64 typer);
> > >
> > > That would always return false for arm32?
> >
> > We don't support ARMv7 host, so that doesn't matter. But
> > it is a good idea to wrap it in a function here.
>
> I'm not sure kvm_host.h is the right place for this as this really relates
> to the ARM PMU drivers. Seeing that armv8pmu_filter_match in
> arch/arm64/kernel/perf_event.c would also benefit from this new function I'll
> add it to arch/arm64/include/asm/perf_event.h and stub it out in arm_pmu.c
> for the ARM32 case.
Doing that is going to create a dependency with the perf subsystem,
and will need an Ack from the maintainer (Will).
Thanks,
M.
--
Without deviation from the norm, progress is not possible.
