[PATCH 00/26] KVM: x86: Purge kvm_x86_ops->*_supported()
From: Sean Christopherson
Date: Wed Jan 29 2020 - 18:48:23 EST
Our benevolent dictator decreed that "all *_supported() should be removed,
and the code moved from __do_cpuid_func() to set_supported_cpuid"[*].
To make that happen, move CPUID and MSR checks from x86 into SVM/VMX via
the existing ->set_supported_cpuid() and a ->new has_virtualized_msr() ops
respectively.
As usual, there's a fair amount of cleanup in between the mechanical
changes. Most notable is the introduction of cpuid entry accessors and
mutators to replace all of the code to manipulate individual feature bits
in cpuid entries, which is error prone and annoying. MPX (*sigh*) also
gets a healthly dose of cleanup.
I don't love every patch in this series. Specifically, adding an extra
call to ->set_supported_cpuid() to handle XSAVES is ugly. But, I do like
that it purges all ->*_supported() hooks. And practically speaking, odds
are good that CPUID 0xD.1 will get more feature bits, i.e. keeping
->xsaves_supported() would likely lead to another ->*_supported() hook.
Paolo also expressed a dislike for clearing bits in set_supported_cpuid().
I don't have a strong opinion regarding clearing bits, but the alternative
approach, i.e. leave the bits clear and then set them in vendor code,
gets quite kludgy because the vendor code (mostly VMX) would need to
manually recheck boot_cpu_data to ensure it wasn't advertising a feature
that the user/kernel expressly disabled. IMO, forcing manual checks is
more likely to introduce errors and provides less insight into why VMX
needs to adjust the advertised CPUID values (VMCS != CPUID).
Tested on Intel by verifying the output of KVM_GET_SUPPORTED_CPUID is
identical before and after (on almost every patch) on a Haswell and Coffee
Lake. The big untested pieces are PKU and PT on Intel, and everything AMD.
[*] https://lkml.kernel.org/r/8a77e3b9-049e-e622-9332-9bebb829bc3d@xxxxxxxxxx
Sean Christopherson (26):
KVM: x86: Remove superfluous brackets from case statement
KVM: x86: Take an unsigned 32-bit int for has_emulated_msr()'s index
KVM: x86: Snapshot MSR index in a local variable when processing lists
KVM: x86: Add a kvm_x86_ops hook to query virtualized MSR support
KVM: x86: Move MSR_TSC_AUX existence checks into vendor code
KVM: x86: Move MSR_IA32_BNDCFGS existence checks into vendor code
KVM: VMX: Add helpers to query Intel PT mode
KVM: x86: Move RTIT (Intel PT) MSR existence checks into vendor code
KVM: x86: Calculate the supported xcr0 mask at load time
KVM: x86: Use supported_xcr0 to detect MPX support
KVM: x86: Make kvm_mpx_supported() an inline function
KVM: x86: Drop explicit @func param from ->set_supported_cpuid()
KVM: x86: Use u32 for holding CPUID register value in helpers
KVM: x86: Introduce cpuid_entry_{get,has}() accessors
KVM: x86: Introduce cpuid_entry_{change,set,clear}() mutators
KVM: x86: Add Kconfig-controlled auditing of reverse CPUID lookups
KVM: x86: Handle MPX CPUID adjustment in vendor code
KVM: x86: Handle INVPCID CPUID adjustment in vendor code
KVM: x86: Handle UMIP emulation CPUID adjustment in VMX code
KVM: x86: Handle PKU CPUID adjustment in SVM code
KVM: x86: Handle RDTSCP CPUID adjustment in VMX code
KVM: x86: Handle XSAVES CPUID adjustment in VMX code
KVM: x86: Handle Intel PT CPUID adjustment in vendor code
KVM: x86: Clear output regs for CPUID 0x14 if PT isn't exposed to
guest
KVM: x86: Handle main Intel PT CPUID leaf in vendor code
KVM: VMX: Directly query Intel PT mode when refreshing PMUs
arch/x86/include/asm/kvm_host.h | 12 +--
arch/x86/kvm/Kconfig | 10 +++
arch/x86/kvm/cpuid.c | 147 +++++++++++++-------------------
arch/x86/kvm/cpuid.h | 85 +++++++++++++++---
arch/x86/kvm/svm.c | 88 ++++++++++---------
arch/x86/kvm/vmx/capabilities.h | 25 ++++--
arch/x86/kvm/vmx/nested.c | 2 +-
arch/x86/kvm/vmx/pmu_intel.c | 2 +-
arch/x86/kvm/vmx/vmx.c | 119 ++++++++++++++++++--------
arch/x86/kvm/vmx/vmx.h | 4 +-
arch/x86/kvm/x86.c | 76 +++++++----------
arch/x86/kvm/x86.h | 10 +--
12 files changed, 331 insertions(+), 249 deletions(-)
--
2.24.1