[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20241203193220.1070811-1-oliver.upton@linux.dev>
Date: Tue, 3 Dec 2024 11:32:06 -0800
From: Oliver Upton <oliver.upton@...ux.dev>
To: kvmarm@...ts.linux.dev
Cc: Marc Zyngier <maz@...nel.org>,
Joey Gouly <joey.gouly@....com>,
Suzuki K Poulose <suzuki.poulose@....com>,
Zenghui Yu <yuzenghui@...wei.com>,
Mingwei Zhang <mizhang@...gle.com>,
Colton Lewis <coltonlewis@...gle.com>,
Raghavendra Rao Ananta <rananta@...gle.com>,
Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>,
Mark Rutland <mark.rutland@....com>,
linux-arm-kernel@...ts.infradead.org,
linux-kernel@...r.kernel.org,
Oliver Upton <oliver.upton@...ux.dev>
Subject: [RFC PATCH 00/14] KVM: arm64: Support FEAT_PMUv3 on Apple hardware
One of the interesting features of some Apple M* parts is an IMPDEF trap
that routes EL1/EL0 accesses of the PMUv3 registers to EL2. This allows
a hypervisor to emulate an architectural PMUv3 on top of the IMPDEF PMU
hardware present in the CPU.
And if you squint, this _might_ look like a CPU erratum :-)
This series takes advantage of these IMPDEF traps to provide PMUv3 to
KVM guests. As a starting point, only expose the fixed CPU cycle counter
and no event counters. Conveniently, this is enough to get Windows
running as a KVM guest on Apple hardware.
I've tried to keep the deviation to a minimum by refactoring some of the
flows used for PMUv3, e.g. computing PMCEID from the arm_pmu bitmap
instead of reading hardware directly.
Sending this as an RFC as there are some obvious open Qs:
- Does enabling PMUv3 meet the bar for allowing IMPDEF crap to slide
into KVM?
I certainly think the answer is 'yes', especially considering this
enables Windows guests.
- Do we want to support programmable event counters?
I'm sitting on some extra patches to do this, which maps a few
architectural event IDs into the PERF_TYPE_HARDWARE space, allowing
the perf driver to select the right hardware event ID.
Deciding on a sensible number of counters in this case is a bit of a
mess, as the M* PMCs aren't entirely fungible (some events only work
on specific counters).
Applies to 6.13-rc1. I've only enabled this for M2 as it is the hardware
I have, but it is extremely likely this feature works on other M* parts.
It is also very possible I've broken something for true PMUv3 hardware,
as I've only tested on the M2.
Oliver Upton (14):
drivers/perf: apple_m1: Refactor event select/filter configuration
drivers/perf: apple_m1: Support host/guest event filtering
drivers/perf: apple_m1: Map generic branch events
KVM: arm64: Compute PMCEID from arm_pmu's event bitmaps
KVM: arm64: Always allow fixed cycle counter
KVM: arm64: Use PERF_COUNT_HW_CPU_CYCLES for fixed cycle counter
KVM: arm64: Use a cpucap to determine if system supports FEAT_PMUv3
KVM: arm64: Drop kvm_arm_pmu_available static key
KVM: arm64: Use guard() to cleanup usage of arm_pmus_lock
KVM: arm64: Move PMUVer filtering into KVM code
KVM: arm64: Compute synthetic sysreg ESR for Apple PMUv3 traps
KVM: arm64: Advertise PMUv3 if IMPDEF traps are present
KVM: arm64: Advertise 0 event counters for IMPDEF PMU
arm64: Enable IMP DEF PMUv3 traps on Apple M2
arch/arm64/include/asm/apple_m1_pmu.h | 1 +
arch/arm64/include/asm/cpufeature.h | 28 +----
arch/arm64/kernel/cpu_errata.c | 38 ++++++
arch/arm64/kernel/cpufeature.c | 19 +++
arch/arm64/kernel/image-vars.h | 5 -
arch/arm64/kvm/arm.c | 4 +-
arch/arm64/kvm/hyp/include/hyp/switch.h | 4 +-
arch/arm64/kvm/hyp/vhe/switch.c | 22 ++++
arch/arm64/kvm/pmu-emul.c | 159 +++++++++++++++---------
arch/arm64/kvm/pmu.c | 10 +-
arch/arm64/tools/cpucaps | 2 +
drivers/perf/apple_m1_cpu_pmu.c | 68 ++++++----
include/kvm/arm_pmu.h | 15 +--
13 files changed, 248 insertions(+), 127 deletions(-)
base-commit: 40384c840ea1944d7c5a392e8975ed088ecf0b37
--
2.39.5
Powered by blists - more mailing lists