Re: [PATCH v3 4/6] x86/sev: Add interface to re-enable RMP optimizations.

[Kalra, Ashish]

Hello Dave,

On 3/30/2026 6:33 PM, Dave Hansen wrote:

>> +int snp_perform_rmp_optimization(void)
>> +{
>> +	if (!cpu_feature_enabled(X86_FEATURE_RMPOPT))
>> +		return -EINVAL;
>> +
>> +	if (!cc_platform_has(CC_ATTR_HOST_SEV_SNP))
>> +		return -EINVAL;
>> +
>> +	if (!(rmp_cfg & MSR_AMD64_SEG_RMP_ENABLED))
>> +		return -EINVAL;
> 
> This seems wrong. How about we just make 'X86_FEATURE_RMPOPT' the one
> true source of RMP support?
> 
> If you don't have CC_ATTR_HOST_SEV_SNP you:
> 
> 	setup_clear_cpu_cap(X86_FEATURE_RMPOPT)
> 
> Ditto for MSR_AMD64_SEG_RMP_ENABLED.
> 
> It could also potentially replace the 'rmpopt_wq' checks.
> 

Following up on this ...

It is straightforward to clear X86_FEATURE_RMPOPT if the RMPOPT setup
function (that is, configure and enable RMPOPT function) gets called, but 
if CC_ATTR_HOST_SEV_SNP is not set, then __sev_snp_init_locked() (CCP module)
does not invoke the RMPOPT setup function. 

And then as this function snp_perform_rmp_optimization() is an external
API, it needs to check for both CC_ATTR_HOST_SEV_SNP and MSR_AMD64_SEG_RMP_ENABLED.

Otherwise, we will need to clear X86_FEATURE_RMPOPT, wherever CC_ATTR_HOST_SEV_SNP
is cleared all across call sites like the AMD IOMMU driver, 
AMD SVM-SEV command line parsing support code and AMD CPU detection and BSP init
code.

And for clearing X86_FEATURE_RMPOPT, if MSR_AMD64_SEG_RMP_ENABLED is not set, 
the support will need to be added in setup_rmptable().

It is much more straightforward to check for both CC_ATTR_HOST_SEV_SNP and
MSR_AMD64_SEG_RMP_ENABLED in this API function itself.

Thanks,
Ashish