Re: [PATCH 1/2] x86: Use CPUID 0x8000001f to confirm SME availability

[Tom Lendacky]

On 2/10/22 23:36, Mario Limonciello wrote:
> An upcoming change will disable the X86 SME feature flag when the
> kernel hasn't activated SME.  Avoid relying upon that when determining
> whether to call `native_wbinvd` by directly calling the CPUID that
> indicates it.
>
> Suggested-by: Borislav Petkov <bp@xxxxxxxxx>
> Signed-off-by: Mario Limonciello <mario.limonciello@xxxxxxx>
> ---
>   arch/x86/kernel/process.c | 5 ++++-
>   1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
> index 81d8ef036637..e131d71b3cae 100644
> --- a/arch/x86/kernel/process.c
> +++ b/arch/x86/kernel/process.c
> @@ -765,8 +765,11 @@ void stop_this_cpu(void *dummy)
>   	 * without the encryption bit, they don't race each other when flushed
>   	 * and potentially end up with the wrong entry being committed to
>   	 * memory.
> +	 *
> +	 * Test the CPUID bit directly because the machine might've cleared
> +	 * X86_FEATURE_SME due to cmdline options.
>   	 */
> -	if (boot_cpu_has(X86_FEATURE_SME))
> +	if (cpuid_eax(0x8000001f) & BIT(0))

Please test this by running kexec and alternating between mem_encrypt=on 
boots of the kexec kernel and mem_encrypt=off boots of the kexec kernel to 
ensure there is no memory corruption in the newly booted kernel. This 
testing needs to be done on hardware that doesn't have the 
X86_FEATURE_SME_COHERENT feature.

Or if you post the generated instructions in this area I should be able to 
determine if the change is safe.

Thanks,
Tom

>   		native_wbinvd();
>   	for (;;) {
>   		/*