Date:   Tue,  7 Jun 2022 21:35:49 +0000
From:   Sean Christopherson <seanjc@...gle.com>
To:     Paolo Bonzini <pbonzini@...hat.com>
Cc:     Sean Christopherson <seanjc@...gle.com>,
        Vitaly Kuznetsov <vkuznets@...hat.com>,
        Wanpeng Li <wanpengli@...cent.com>,
        Jim Mattson <jmattson@...gle.com>,
        Joerg Roedel <joro@...tes.org>, kvm@...r.kernel.org,
        linux-kernel@...r.kernel.org, Eric Li <ercli@...avis.edu>,
        David Matlack <dmatlack@...gle.com>,
        Oliver Upton <oupton@...gle.com>
Subject: [PATCH v5 00/15]  KVM: nVMX: VMX MSR quirk+fixes, CR4 fixes

Resurrecting Oliver's series to quirk nVMX's manipulation of the VMX MSRs.

KVM has a quirk where nVMX overwrites select VMX MSR bits in response to
CPUID updates.  Specifically, KVM forces CR{0,4}_FIXED1 bits and the
VM-Entry/VM-Exit control bits for BNDCFGS and PERF_GLOBAL_CTRL to align
with the guest vCPU model.  Add a quirk to (a) allow userspace to opt out
of the existing behavior and (b) make it clear the the existing behavior
should not be propagated to new features.

Patches 0-4 are tangentially related fixes for correctly handling CR4
checks on nested VMXON and VM-Enter.  They are included here because there
is a subtle dependency created by the fix in patch 02, as it changes the
resulting behavior of patch 10, "Extend VMX MSRs quirk to CR0/4 fixed1 bits".

Patch 05 fixes a bug where KVM forces incoming VMX MSR values to be a
subset of _current_ value, not of KVM's support valu.  E.g. if userspace
clears an allowed-1 bit, it can never set that bit back to the original
value.

Patch 06, "Keep KVM updates to BNDCFGS ctrl bits across MSR write", fixes
a related bug where KVM's original quirky behavior kept the VMX MSRs
up-to-date (almost) all the time.

This series is technically based on my selftests overhaul[*], but practically
speaking that only affects the selftests.  The KVM should apply cleanly on
kvm/queue, 55371f1d0c01 ("KVM: x86/pmu: Update global ...")

I have a KUT test for patch 3 (VMXON fixes) that I'l post separately (yet
more cleanup involved, ugh).  I spot tested patch 2 by fudging PKU in guest
CPUID, but I'm not planning on submitting an official test anywhere (though
it could be done without too much pain in selftests).

v5:
 - Rebase (see above).
 - Fix "CR4 valid for nVMX" bugs.
 - Modify PERF_GLOBAL_CTRL bits iff the MSR exists.
 - Fix a bug where KVM doesn't allow userspace to restore VMX MSRs to
   _KVM's_ allowed values.
 - Fix the UMIP emulation goof.
 - Extend the quirk to CR0/4_FIXED1 MSRs.
 - Add a helper to identify if the vCPU has a vPMU.
 - Rewrote the selftests to more exhaustively test combos, and to test
   the aforementioned bugs fixed in v5.

v4:
 - https://lore.kernel.org/all/20220301060351.442881-1-oupton@google.com
 - Rebased to kvm/queue. Avoids conflicts with new CAPs and commit
   0bcd556e15f9 ("KVM: nVMX: Refactor PMU refresh to avoid referencing
   kvm_x86_ops.pmu_ops") on kvm/queue.
 - Grabbed KVM_CAP_DISABLE_QUIRKS2 patch, since this series also
   introduces a quirk.
 - Fix typo in KVM_CAP_DISABLE_QUIRKS2 documentation (Sean)
 - Eliminated the need to refresh 'load IA32_PGC' bits from PMU refresh.
 - Use consistent formatting to make test cases more easily readable
   (David Dunn)
 - Use correct 'Fixes: ' tag and correct a typo in Patch 2 changelog.

Oliver Upton (4):
  KVM: nVMX: Keep KVM updates to BNDCFGS ctrl bits across MSR write
  KVM: nVMX: Keep KVM updates to PERF_GLOBAL_CTRL ctrl bits across MSR
    write
  KVM: nVMX: Drop nested_vmx_pmu_refresh()
  KVM: nVMX: Add a quirk for KVM tweaks to VMX MSRs

Sean Christopherson (11):
  KVM: x86: Split kvm_is_valid_cr4() and export only the non-vendor bits
  KVM: nVMX: Account for KVM reserved CR4 bits in consistency checks
  KVM: nVMX: Inject #UD if VMXON is attempted with incompatible CR0/CR4
  KVM: nVMX: Rename handle_vm{on,off}() to handle_vmx{on,off}()
  KVM: nVMX: Let userspace set nVMX MSR to any _host_ supported value
  KVM: VMX: Add helper to check if the guest PMU has PERF_GLOBAL_CTRL
  KVM: nVMX: Set UMIP bit CR4_FIXED1 MSR when emulating UMIP
  KVM: nVMX: Extend VMX MSRs quirk to CR0/4 fixed1 bits
  KVM: selftests: Add test to verify KVM's VMX MSRs quirk for controls
  KVM: selftests: Extend VMX MSRs test to cover CR4_FIXED1 (and its
    quirks)
  KVM: selftests: Verify VMX MSRs can be restored to KVM-supported
    values

 Documentation/virt/kvm/api.rst                |  29 ++
 arch/x86/include/asm/kvm_host.h               |   3 +-
 arch/x86/include/uapi/asm/kvm.h               |   1 +
 arch/x86/kvm/svm/nested.c                     |   3 +-
 arch/x86/kvm/vmx/nested.c                     | 199 ++++++------
 arch/x86/kvm/vmx/nested.h                     |   5 +-
 arch/x86/kvm/vmx/pmu_intel.c                  |   7 +-
 arch/x86/kvm/vmx/vmx.c                        |  23 +-
 arch/x86/kvm/vmx/vmx.h                        |  14 +
 arch/x86/kvm/x86.c                            |  12 +-
 arch/x86/kvm/x86.h                            |   2 +-
 tools/testing/selftests/kvm/.gitignore        |   1 +
 tools/testing/selftests/kvm/Makefile          |   1 +
 .../selftests/kvm/include/x86_64/processor.h  |   8 +
 .../selftests/kvm/include/x86_64/vmx.h        |   2 +
 .../selftests/kvm/x86_64/vmx_msrs_test.c      | 287 ++++++++++++++++++
 16 files changed, 493 insertions(+), 104 deletions(-)
 create mode 100644 tools/testing/selftests/kvm/x86_64/vmx_msrs_test.c


base-commit: 081ad4bbae8d503c79fae45f463766d28b2f3241
-- 
2.36.1.255.ge46751e96f-goog

Powered by blists - more mailing lists