Re: [PATCH v4 3/9] KVM: selftests: Add pmu.h for PMU events and common masks

[Sean Christopherson]

Shortlog should call out lib/pmu.c (or just say "a library", but naming files is
fine).

On Mon, Sep 11, 2023, Jinrong Liang wrote:
> +#define PMU_CAP_FW_WRITES				BIT_ULL(13)
> +#define PMU_CAP_LBR_FMT				0x3f
> +
> +enum intel_pmu_architectural_events {
> +	/*
> +	 * The order of the architectural events matters as support for each
> +	 * event is enumerated via CPUID using the index of the event.
> +	 */
> +	INTEL_ARCH_CPU_CYCLES,
> +	INTEL_ARCH_INSTRUCTIONS_RETIRED,
> +	INTEL_ARCH_REFERENCE_CYCLES,
> +	INTEL_ARCH_LLC_REFERENCES,
> +	INTEL_ARCH_LLC_MISSES,
> +	INTEL_ARCH_BRANCHES_RETIRED,
> +	INTEL_ARCH_BRANCHES_MISPREDICTED,
> +
> +	NR_REAL_INTEL_ARCH_EVENTS,
> +
> +	/*
> +	 * Pseudo-architectural event used to implement IA32_FIXED_CTR2, a.k.a.
> +	 * TSC reference cycles. The architectural reference cycles event may
> +	 * or may not actually use the TSC as the reference, e.g. might use the
> +	 * core crystal clock or the bus clock (yeah, "architectural").
> +	 */
> +	PSEUDO_ARCH_REFERENCE_CYCLES = NR_REAL_INTEL_ARCH_EVENTS,
> +	NR_INTEL_ARCH_EVENTS,
> +};

...

> +/* Definitions for Architectural Performance Events */
> +#define ARCH_EVENT(select, umask) (((select) & 0xff) | ((umask) & 0xff) << 8)
> +
> +const uint64_t intel_pmu_arch_events[] = {
> +	[INTEL_ARCH_CPU_CYCLES]			= ARCH_EVENT(0x3c, 0x0),
> +	[INTEL_ARCH_INSTRUCTIONS_RETIRED]	= ARCH_EVENT(0xc0, 0x0),
> +	[INTEL_ARCH_REFERENCE_CYCLES]		= ARCH_EVENT(0x3c, 0x1),
> +	[INTEL_ARCH_LLC_REFERENCES]		= ARCH_EVENT(0x2e, 0x4f),
> +	[INTEL_ARCH_LLC_MISSES]			= ARCH_EVENT(0x2e, 0x41),
> +	[INTEL_ARCH_BRANCHES_RETIRED]		= ARCH_EVENT(0xc4, 0x0),
> +	[INTEL_ARCH_BRANCHES_MISPREDICTED]	= ARCH_EVENT(0xc5, 0x0),
> +	[PSEUDO_ARCH_REFERENCE_CYCLES]		= ARCH_EVENT(0xa4, 0x1),

Argh, seriously, why!?!?!  0xa4, 0x1 is Topdown Slots, not TSC reference cycles.
There is _zero_ reason to carry over KVM's godawful pseudo-architectural event
crud that exists purely because KVM uses perf events to virtualized PMCs.

And peeking forward at the usage, the whole thing is completely nonsensical, as
the selftest will set expectations on the TSC reference cycles behavior based
on whether or not the Topdown Slots event is supported.

Grr.