Re: [PATCH v7 4/6] target/i386: Add couple of feature bits in CPUID_Fn80000021_EAX

2025-06-16 Thread Dr. David Alan Gilbert
* Moger, Babu ([email protected]) wrote:
> Hi Dave,
> 
> On 6/6/25 07:29, Dr. David Alan Gilbert wrote:
> > * Babu Moger ([email protected]) wrote:
> >> Add CPUID bit indicates that a WRMSR to MSR_FS_BASE, MSR_GS_BASE, or
> >> MSR_KERNEL_GS_BASE is non-serializing amd PREFETCHI that the indicates
> >> support for IC prefetch.
> >>
> >> CPUID_Fn8021_EAX
> >> BitFeature description
> >> 20 Indicates support for IC prefetch.
> >> 1  FsGsKernelGsBaseNonSerializing.
> > 
> > I'm curious about this:
> >   a) Is this new CPUs are non-serialising on that write?
> >   b) If so, what happens if you run existing kernels/firmware on them?
> >   c) Bonus migration question; what happens if you live migrate from a host
> >  that claims to be serialising to one that has the extra non-serialising
> >  flag but is disabled in the emulated CPU model.
> 
> Good question. After looking at the AMD64 Architecture Programmer’s Manual
> again, these writes have always been non-serializing. Behavior has not
> changed.

Ah OK, then nothing to worry about.

> We're just reporting it through CPUID now. This information
> likely isn’t being used anywhere.

Seems curious to add it then!

> Let me know if you have any questions.

No, thanks for the reply.

Dave

> -- 
> Thanks
> Babu Moger
-- 
 -Open up your eyes, open up your mind, open up your code ---   
/ Dr. David Alan Gilbert|   Running GNU/Linux   | Happy  \ 
\dave @ treblig.org |   | In Hex /
 \ _|_ http://www.treblig.org   |___/



Re: [PATCH v7 4/6] target/i386: Add couple of feature bits in CPUID_Fn80000021_EAX

2025-06-16 Thread Moger, Babu
Hi Dave,

On 6/6/25 07:29, Dr. David Alan Gilbert wrote:
> * Babu Moger ([email protected]) wrote:
>> Add CPUID bit indicates that a WRMSR to MSR_FS_BASE, MSR_GS_BASE, or
>> MSR_KERNEL_GS_BASE is non-serializing amd PREFETCHI that the indicates
>> support for IC prefetch.
>>
>> CPUID_Fn8021_EAX
>> BitFeature description
>> 20 Indicates support for IC prefetch.
>> 1  FsGsKernelGsBaseNonSerializing.
> 
> I'm curious about this:
>   a) Is this new CPUs are non-serialising on that write?
>   b) If so, what happens if you run existing kernels/firmware on them?
>   c) Bonus migration question; what happens if you live migrate from a host
>  that claims to be serialising to one that has the extra non-serialising
>  flag but is disabled in the emulated CPU model.

Good question. After looking at the AMD64 Architecture Programmer’s Manual
again, these writes have always been non-serializing. Behavior has not
changed. We're just reporting it through CPUID now. This information
likely isn’t being used anywhere. Let me know if you have any questions.
-- 
Thanks
Babu Moger



Re: [PATCH v7 4/6] target/i386: Add couple of feature bits in CPUID_Fn80000021_EAX

2025-06-06 Thread Dr. David Alan Gilbert
* Babu Moger ([email protected]) wrote:
> Add CPUID bit indicates that a WRMSR to MSR_FS_BASE, MSR_GS_BASE, or
> MSR_KERNEL_GS_BASE is non-serializing amd PREFETCHI that the indicates
> support for IC prefetch.
> 
> CPUID_Fn8021_EAX
> BitFeature description
> 20 Indicates support for IC prefetch.
> 1  FsGsKernelGsBaseNonSerializing.

I'm curious about this:
  a) Is this new CPUs are non-serialising on that write?
  b) If so, what happens if you run existing kernels/firmware on them?
  c) Bonus migration question; what happens if you live migrate from a host
 that claims to be serialising to one that has the extra non-serialising
 flag but is disabled in the emulated CPU model.

Dave

>WRMSR to FS_BASE, GS_BASE and KernelGSbase are non-serializing.
> 
> Link: 
> https://www.amd.com/content/dam/amd/en/documents/epyc-technical-docs/programmer-references/57238.zip
> Signed-off-by: Babu Moger 
> Reviewed-by: Maksim Davydov 
> Reviewed-by: Zhao Liu 
> ---
>  target/i386/cpu.c | 4 ++--
>  target/i386/cpu.h | 4 
>  2 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/target/i386/cpu.c b/target/i386/cpu.c
> index 98fad3a2f9..741be0eaa8 100644
> --- a/target/i386/cpu.c
> +++ b/target/i386/cpu.c
> @@ -1239,12 +1239,12 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
>  [FEAT_8000_0021_EAX] = {
>  .type = CPUID_FEATURE_WORD,
>  .feat_names = {
> -"no-nested-data-bp", NULL, "lfence-always-serializing", NULL,
> +"no-nested-data-bp", "fs-gs-base-ns", 
> "lfence-always-serializing", NULL,
>  NULL, NULL, "null-sel-clr-base", NULL,
>  "auto-ibrs", NULL, NULL, NULL,
>  NULL, NULL, NULL, NULL,
>  NULL, NULL, NULL, NULL,
> -NULL, NULL, NULL, NULL,
> +"prefetchi", NULL, NULL, NULL,
>  "eraps", NULL, NULL, "sbpb",
>  "ibpb-brtype", "srso-no", "srso-user-kernel-no", NULL,
>  },
> diff --git a/target/i386/cpu.h b/target/i386/cpu.h
> index 4f8ed8868e..d251e32ae9 100644
> --- a/target/i386/cpu.h
> +++ b/target/i386/cpu.h
> @@ -1070,12 +1070,16 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU 
> *cpu, FeatureWord w);
>  
>  /* Processor ignores nested data breakpoints */
>  #define CPUID_8000_0021_EAX_NO_NESTED_DATA_BP(1U << 0)
> +/* WRMSR to FS_BASE, GS_BASE, or KERNEL_GS_BASE is non-serializing */
> +#define CPUID_8000_0021_EAX_FS_GS_BASE_NS(1U << 1)
>  /* LFENCE is always serializing */
>  #define CPUID_8000_0021_EAX_LFENCE_ALWAYS_SERIALIZING(1U << 2)
>  /* Null Selector Clears Base */
>  #define CPUID_8000_0021_EAX_NULL_SEL_CLR_BASE(1U << 6)
>  /* Automatic IBRS */
>  #define CPUID_8000_0021_EAX_AUTO_IBRS(1U << 8)
> +/* Indicates support for IC prefetch */
> +#define CPUID_8000_0021_EAX_PREFETCHI(1U << 20)
>  /* Enhanced Return Address Predictor Scurity */
>  #define CPUID_8000_0021_EAX_ERAPS(1U << 24)
>  /* Selective Branch Predictor Barrier */
> -- 
> 2.34.1
> 
> 
-- 
 -Open up your eyes, open up your mind, open up your code ---   
/ Dr. David Alan Gilbert|   Running GNU/Linux   | Happy  \ 
\dave @ treblig.org |   | In Hex /
 \ _|_ http://www.treblig.org   |___/