On Wed, Jan 20, 2021 at 01:02:15PM -0800, Dave Hansen wrote: > On 1/17/21 7:27 PM, Kai Huang wrote: > > - enable_sgx = cpu_has(c, X86_FEATURE_SGX) && > > - cpu_has(c, X86_FEATURE_SGX_LC) && > > - IS_ENABLED(CONFIG_X86_SGX); > > + enable_sgx_driver = cpu_has(c, X86_FEATURE_SGX) && > > + cpu_has(c, X86_FEATURE_SGX1) && > > + IS_ENABLED(CONFIG_X86_SGX) && > > + cpu_has(c, X86_FEATURE_SGX_LC); > > + enable_sgx_virt = cpu_has(c, X86_FEATURE_SGX) && > > + cpu_has(c, X86_FEATURE_SGX1) && > > + IS_ENABLED(CONFIG_X86_SGX) && > > + IS_ENABLED(CONFIG_X86_SGX_VIRTUALIZATION) && > > + enable_vmx; > > Would it be too much to ask that the SGX/SGX1 checks not be duplicated? > Perhaps: > > enable_sgx_any = cpu_feature_enabled(CONFIG_X86_SGX) && > cpu_feature_enabled(CONFIG_X86_SGX1); > > enable_sgx_driver = enable_sgx_any && > cpu_has(c, X86_FEATURE_SGX_LC); > > enable_sgx_virt = enable_sgx_any && > enable_vmx && > IS_ENABLED(CONFIG_X86_SGX_VIRTUALIZATION) > > BTW, CONFIG_X86_SGX_VIRTUALIZATION is a pretty porky name. Maybe just > CONFIG_X86_SGX_VIRT? If my /dev/sgx_vepc naming gets acceptance, then IMHO the best name ought to be CONFIG_X86_VEPC. /Jarkko