Re: [PATCH V7 03/18] perf/x86/intel: Handle guest PEBS overflow PMI for KVM guest

[Zhu, Lingshan]

On 7/2/2021 7:28 PM, Peter Zijlstra wrote:
> On Tue, Jun 22, 2021 at 05:42:51PM +0800, Zhu Lingshan wrote:
> > +DECLARE_STATIC_CALL(x86_guest_state, *(perf_guest_cbs->state));
> > +
> > +/*
> > + * We may be running with guest PEBS events created by KVM, and the
> > + * PEBS records are logged into the guest's DS and invisible to host.
> > + *
> > + * In the case of guest PEBS overflow, we only trigger a fake event
> > + * to emulate the PEBS overflow PMI for guest PBES counters in KVM.
> > + * The guest will then vm-entry and check the guest DS area to read
> > + * the guest PEBS records.
> > + *
> > + * The contents and other behavior of the guest event do not matter.
> > + */
> > +static void x86_pmu_handle_guest_pebs(struct pt_regs *regs,
> > +				      struct perf_sample_data *data)
> > +{
> > +	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
> > +	u64 guest_pebs_idxs = cpuc->pebs_enabled & ~cpuc->intel_ctrl_host_mask;
> > +	struct perf_event *event = NULL;
> > +	unsigned int guest = 0;
> > +	int bit;
> > +
> > +	if (!x86_pmu.pebs_vmx || !x86_pmu.pebs_active ||
> > +	    !(cpuc->pebs_enabled & ~cpuc->intel_ctrl_host_mask))
> > +		return;
> > +
> > +	guest = static_call(x86_guest_state)();
> > +	if (!(guest & PERF_GUEST_ACTIVE))
> > +		return;
> I think you've got the branches the wrong way around here; nobody runs a
> VM so this branch will get you out without a load.
>
> Only if you're one of those daft people running a VM, are you interested
> in any of the other conditions that are required.
>
> Also, I think both pebs_active and pebs_vmx can he a static_branch, but
> that can be done later I suppose.
Hi Peter,

If I understand this correctly, are you suggesting we put "if (!(guest & 
PERF_GUEST_ACTIVE))" first because this is a lower cost branch?

Thanks,
Zhu Lingshan