Re: [PATCH] x86/cpufeatures: extend CPUID leaf 0xc0000001 support for Zhaoxin

[Tony W Wang-oc]

On 4/14/23 18:48, Borislav Petkov wrote:
> On Fri, Apr 14, 2023 at 05:53:34PM +0800, Tony W Wang-oc wrote:
>> Extend CPUID leaf 0xc0000001 to support SM2, SM3, SM4, PARALLAX, TM3,
>> RNG2, PHE2, RSA.
>>
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 0]  SM2
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 1]  SM2_EN
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 4]  SM3 SM4
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 5]  SM3_EN SM4_EN
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 16] PARALLAX
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 17] PARALLAX_EN
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 20] TM3
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 21] TM3_EN
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 22] RNG2
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 23] RNG2_EN
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 25] PHE2
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 26] PHE2_EN
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 27] RSA
>> CPUID.(EAX=0xc0000001,ECX=0):EDX[bit 28] RSA_EN
> 
> None of those flags are used in code, why do we need this patch?
> 
 The instructions about these flags can be executed at any privilege
level. I think using these flags in kernel mode is a case. This patch
shows the statement of these flags to the user mode explicitly. So users
can see and use these CPU features conveniently.

> If you want to dump them on the hardware to know what's set or not,
> there's tools/arch/x86/kcpuid/ for that.
> 
>> SM2/SM3/SM4 imply the instructions support for Chinese cipher security
>> algorithm generations 2/3/4.
>> PARALLAX is the feature of Zhaoxin CPU that automatically adjusts
>> processors's voltage as a function of temperature.
>> TM3 is the abbreviation of Thermal Monitor version 3.
>> RNG2 is the abbreviation of Random Number Generation version 2.
>> PHE2 is the abbreviation of Padlock Hash Engine version 2.
>> RSA implies Zhaoxin hardware support for RSA algorithm.
>>
>> All these features have two relative CPUID bits, one bit implies the
>> existence of the feature and the other bit with postfix "EN" implies
>> the availability of this feature.
> 
> That's a lot of waste of CPUID bits but that's your decision.
> 
Sorry about that.
This implementation aimed to control or statement these features
independently.

Sincerely
TonyWWangoc