From: Borislav Petkov <bp@alien8.de> Sent: Monday, January 9, 2023 8:39 AM

[snip]

quoted

diff --git a/arch/x86/coco/core.c b/arch/x86/coco/core.c
index 49b44f8..c361c52 100644
--- a/arch/x86/coco/core.c
+++ b/arch/x86/coco/core.c

@@ -44,6 +44,24 @@ static bool intel_cc_platform_has(enum cc_attr attr)
 static bool amd_cc_platform_has(enum cc_attr attr)
 {
 #ifdef CONFIG_AMD_MEM_ENCRYPT
+
+	/*
+	 * Handle the SEV-SNP vTOM case where sme_me_mask must be zero,
+	 * and the other levels of SME/SEV functionality, including C-bit
+	 * based SEV-SNP, must not be enabled.
+	 */
+	if (sev_status & MSR_AMD64_SNP_VTOM_ENABLED) {

		return amd_cc_platform_vtom();

or so and then stick that switch in there.

This way it looks kinda grafted in front and with a function call with a telling
name it says it is a special case...

quoted

+		switch (attr) {
+		case CC_ATTR_GUEST_MEM_ENCRYPT:
+		case CC_ATTR_MEM_ENCRYPT:
+		case CC_ATTR_ACCESS_IOAPIC_ENCRYPTED:
+			return true;
+		default:
+			return false;
+		}
+	}

The rest looks kinda nice, I gotta say. I can't complain. :)

OK.  I have no objection to putting that code in a separate "helper"
function.  The only slight messiness is that the helper function must
be separately wrapped in #ifdef CONFIG_AMD_MEM_ENCRYPT, or
marked __maybe_unused.

Michael