Re: [PATCH v6 2/7] x86/cpu, kvm: Add the NO_NESTED_DATA_BP feature

[Sean Christopherson]

On Tue, Jan 10, 2023, Kim Phillips wrote:
> The "Processor ignores nested data breakpoints" feature was being
> open-coded for KVM in __do_cpuid_func().  Add it to its newly added
> CPUID leaf 0x80000021 EAX proper, and propagate it in kvm_set_cpu_caps()
> instead.
> 
> Also drop the bit description comments now it's more self-describing.
> 
> Signed-off-by: Kim Phillips <kim.phillips@xxxxxxx>
> ---
>  arch/x86/include/asm/cpufeatures.h | 3 +++
>  arch/x86/kvm/cpuid.c               | 8 ++++++--
>  2 files changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
> index d53e13048d2e..0cd7b4afd528 100644
> --- a/arch/x86/include/asm/cpufeatures.h
> +++ b/arch/x86/include/asm/cpufeatures.h
> @@ -426,6 +426,9 @@
>  #define X86_FEATURE_V_TSC_AUX		(19*32+ 9) /* "" Virtual TSC_AUX */
>  #define X86_FEATURE_SME_COHERENT	(19*32+10) /* "" AMD hardware-enforced cache coherency */
>  
> +/* AMD-defined Extended Feature 2 EAX, CPUID level 0x80000021 (EAX), word 20 */
> +#define X86_FEATURE_NO_NESTED_DATA_BP	(20*32+ 0) /* "" AMD No Nested Data Breakpoints */
> +
>  /*
>   * BUG word(s)
>   */
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index b14653b61470..69e433e4e9ff 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -741,6 +741,10 @@ void kvm_set_cpu_caps(void)
>  		0 /* SME */ | F(SEV) | 0 /* VM_PAGE_FLUSH */ | F(SEV_ES) |
>  		F(SME_COHERENT));
>  
> +	kvm_cpu_cap_mask(CPUID_8000_0021_EAX,
> +		F(NO_NESTED_DATA_BP)
> +	);
> +
>  	kvm_cpu_cap_mask(CPUID_C000_0001_EDX,
>  		F(XSTORE) | F(XSTORE_EN) | F(XCRYPT) | F(XCRYPT_EN) |
>  		F(ACE2) | F(ACE2_EN) | F(PHE) | F(PHE_EN) |
> @@ -1222,9 +1226,9 @@ static inline int __do_cpuid_func(struct kvm_cpuid_array *array, u32 function)
>  		break;
>  	case 0x80000021:
>  		entry->ebx = entry->ecx = entry->edx = 0;
> +		cpuid_entry_override(entry, CPUID_8000_0021_EAX);
>  		/*
>  		 * Pass down these bits:
> -		 *    EAX      0      NNDBP, Processor ignores nested data breakpoints
>  		 *    EAX      2      LAS, LFENCE always serializing
>  		 *    EAX      6      NSCB, Null selector clear base
>  		 *
> @@ -1235,7 +1239,7 @@ static inline int __do_cpuid_func(struct kvm_cpuid_array *array, u32 function)
>  		 * KVM doesn't support SMM_CTL.
>  		 *   EAX       9     SMM_CTL MSR is not supported
>  		 */
> -		entry->eax &= BIT(0) | BIT(2) | BIT(6);
> +		entry->eax &= BIT(2) | BIT(6);

This is broken.  It gets fixed by the end of the series, but between here and
commit b1366f515fd6 ("x86/cpu, kvm: Add the Null Selector Clears Base feature"),
the AND with open coded bits means any bits preserved/set by cpuid_entry_override()
are wiped out.  E.g. NO_NESTED_DATA_BP will never be advertised as of this patch.

The proper way to do this is to first convert all supported bits away from magic
numbers in a single patch, and then introduce newly supported bits one by one.
That one patch will be larger, but I don't see a better approach.

Is it too late to back this out?  Not a huge deal, but it seems easy enough to
clean up.