Hello Sean/Paolo,

On 5/12/2025 11:46 PM, Gupta, Pankaj wrote:

On 5/13/2025 12:16 AM, Ashish Kalra wrote:

...

From: Ashish Kalra ashish.kalra@amd.com

During platform init, SNP initialization may fail for several reasons, such as firmware command failures and incompatible versions. However, the KVM capability may continue to advertise support for it.

The platform may have SNP enabled but if SNP_INIT fails then SNP is not supported by KVM.

During KVM module initialization query the SNP platform status to obtain the SNP initialization state and use it as an additional condition to determine support for SEV-SNP.

Co-developed-by: Sean Christopherson seanjc@google.com Signed-off-by: Sean Christopherson seanjc@google.com Co-developed-by: Pratik R. Sampat prsampat@amd.com Signed-off-by: Pratik R. Sampat prsampat@amd.com Reviewed-by: Tom Lendacky thomas.lendacky@amd.com Signed-off-by: Ashish Kalra ashish.kalra@amd.com

LGTM

Reviewed-by: Pankaj Gupta pankaj.gupta@amd.com

Can you please pull in this patch.

Thanks, Ashish

...

---

arch/x86/kvm/svm/sev.c | 44 +++++++++++++++++++++++++++++++++---------   1 file changed, 35 insertions(+), 9 deletions(-)

diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c index dea9480b9ff6..8c3b12e3de8c 100644 --- a/arch/x86/kvm/svm/sev.c +++ b/arch/x86/kvm/svm/sev.c @@ -2935,6 +2935,33 @@ void __init sev_set_cpu_caps(void)       }   }   +static bool is_sev_snp_initialized(void) +{ +    struct sev_user_data_snp_status *status; +    struct sev_data_snp_addr buf; +    bool initialized = false; +    int ret, error = 0;

+    status = snp_alloc_firmware_page(GFP_KERNEL | __GFP_ZERO); +    if (!status) +        return false;

+    buf.address = __psp_pa(status); +    ret = sev_do_cmd(SEV_CMD_SNP_PLATFORM_STATUS, &buf, &error); +    if (ret) { +        pr_err("SEV: SNP_PLATFORM_STATUS failed ret=%d, fw_error=%d (%#x)\n", +               ret, error, error); +        goto out; +    }

+    initialized = !!status->state;

+out: +    snp_free_firmware_page(status);

+    return initialized; +}

void __init sev_hardware_setup(void)   {       unsigned int eax, ebx, ecx, edx, sev_asid_count, sev_es_asid_count; @@ -3039,6 +3066,14 @@ void __init sev_hardware_setup(void)       sev_snp_supported = sev_snp_enabled && cc_platform_has(CC_ATTR_HOST_SEV_SNP);     out: +    if (sev_enabled) { +        init_args.probe = true; +        if (sev_platform_init(&init_args)) +            sev_supported = sev_es_supported = sev_snp_supported = false; +        else if (sev_snp_supported) +            sev_snp_supported = is_sev_snp_initialized(); +    }

if (boot_cpu_has(X86_FEATURE_SEV))           pr_info("SEV %s (ASIDs %u - %u)\n",               sev_supported ? min_sev_asid <= max_sev_asid ? "enabled" : @@ -3065,15 +3100,6 @@ void __init sev_hardware_setup(void)       sev_supported_vmsa_features = 0;       if (sev_es_debug_swap_enabled)           sev_supported_vmsa_features |= SVM_SEV_FEAT_DEBUG_SWAP;

-    if (!sev_enabled) -        return;

-    /* -     * Do both SNP and SEV initialization at KVM module load. -     */ -    init_args.probe = true; -    sev_platform_init(&init_args);   }     void sev_hardware_unsetup(void)