2018-05-24 10:29-0700, Jim Mattson:
> KVM_GET_SUPPORTED_CPUID should return the x86 cpuid features which are
> supported by both the hardware and kvm. Since commit 4d5422cea3b6
> ("KVM: X86: Provide a capability to disable MWAIT intercepts"), it is
> possible to configure a VM so that MONITOR/MWAIT are fully supported
> in the guest. Userspace still has the option to mask these features
> off (e.g. for VMs that don't disable MONITOR/MWAIT exits.)
>
> Fixes: 4d5422cea3b6 ("KVM: X86: Provide a capability to disable MWAIT intercepts")
> Signed-off-by: Jim Mattson <jmattson@xxxxxxxxxx>
> ---
QEMU has a peculiar '-cpu host' mode that enables everything that is
reported as supported and this change would regress old userspaces by
tricking the guest OS into using a VM exit busy-loop instead of HLT.
I think it would be better to say that the userspace should enable
MONITOR in CPUID when using KVM_X86_DISABLE_EXITS_MWAIT.
> arch/x86/kvm/cpuid.c | 4 +---
> 1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index 82055b90a8b3..794e8c6adc11 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -359,9 +359,7 @@ static inline int __do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
> 0 /* Reserved */ | f_lm | F(3DNOWEXT) | F(3DNOW);
> /* cpuid 1.ecx */
> const u32 kvm_cpuid_1_ecx_x86_features =
> - /* NOTE: MONITOR (and MWAIT) are emulated as NOP,
> - * but *not* advertised to guests via CPUID ! */
> - F(XMM3) | F(PCLMULQDQ) | 0 /* DTES64, MONITOR */ |
> + F(XMM3) | F(PCLMULQDQ) | 0 /* DTES64 */ | F(MONITOR) |
> 0 /* DS-CPL, VMX, SMX, EST */ |
> 0 /* TM2 */ | F(SSSE3) | 0 /* CNXT-ID */ | 0 /* Reserved */ |
> F(FMA) | F(CX16) | 0 /* xTPR Update, PDCM */ |
> --
> 2.17.0.441.gb46fe60e1d-goog
>
