Currently ignore_msrs and report_ignored_msrs have a few issues:
- Errors could be dumped to dmesg even if the msr access is triggered inside
kvm itself (e.g., kvm_cpuid), while what we really want to trap should be
either guest msr accesses, or KVM_SET_MSRS.
- These two parameters didn't apply to feature msrs.
Each of the patch in this series tries to handle one of the issues.
Here KVM_MSR_RET_INVALID is introduced. Ideally it can be an enum with both
0/1 defined too, but I'll see whether there's any feedback first about this
version.
This originates from a discussion between Paolo and me on an unexpected warning
msr access message that triggered on a RT system, which seemed to have caused
some system jitters.
Please have a look, thanks.
Peter Xu (2):
KVM: X86: Move ignore_msrs handling upper the stack
KVM: X86: Do the same ignore_msrs check for feature msrs
arch/x86/kvm/svm/svm.c | 2 +-
arch/x86/kvm/vmx/vmx.c | 2 +-
arch/x86/kvm/x86.c | 90 +++++++++++++++++++++++++++++-------------
arch/x86/kvm/x86.h | 2 +
4 files changed, 66 insertions(+), 30 deletions(-)
--
2.26.2
