KVM: x86/pmu: Fix and isolate TSX-specific performance event logic
authorLike Xu <likexu@tencent.com>
Wed, 9 Mar 2022 08:42:57 +0000 (16:42 +0800)
committerPaolo Bonzini <pbonzini@redhat.com>
Sat, 2 Apr 2022 09:34:46 +0000 (05:34 -0400)
commite644896f5106aa3f6d7e8c7adf2e4dc0fce53555
tree1c66609ea0983dae121d991e27fac7765efe166d
parent5959ff4ae96eece2f0c3dfde5d27bff70ab1ce56
KVM: x86/pmu: Fix and isolate TSX-specific performance event logic

HSW_IN_TX* bits are used in generic code which are not supported on
AMD. Worse, these bits overlap with AMD EventSelect[11:8] and hence
using HSW_IN_TX* bits unconditionally in generic code is resulting in
unintentional pmu behavior on AMD. For example, if EventSelect[11:8]
is 0x2, pmc_reprogram_counter() wrongly assumes that
HSW_IN_TX_CHECKPOINTED is set and thus forces sampling period to be 0.

Also per the SDM, both bits 32 and 33 "may only be set if the processor
supports HLE or RTM" and for "IN_TXCP (bit 33): this bit may only be set
for IA32_PERFEVTSEL2."

Opportunistically eliminate code redundancy, because if the HSW_IN_TX*
bit is set in pmc->eventsel, it is already set in attr.config.

Reported-by: Ravi Bangoria <ravi.bangoria@amd.com>
Reported-by: Jim Mattson <jmattson@google.com>
Fixes: 103af0a98788 ("perf, kvm: Support the in_tx/in_tx_cp modifiers in KVM arch perfmon emulation v5")
Co-developed-by: Ravi Bangoria <ravi.bangoria@amd.com>
Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
Signed-off-by: Like Xu <likexu@tencent.com>
Message-Id: <20220309084257.88931-1-likexu@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/pmu.c
arch/x86/kvm/vmx/pmu_intel.c