Re: [PATCH 6.12 111/181] KVM: SVM: Add support to initialize SEV/SNP functionality in KVM

Hi Jack, Greg, Sasha,


On 17/08/26 7:03 pm, Greg Kroah-Hartman wrote:
> 6.12-stable review patch.  If anyone has any objections, please let me know.
> 
> ------------------
> 
> From: Ashish Kalra <[email protected]>
> 
> commit 6f1d5a3513c2370bbd6115dd430906bc2f4bdc53 upstream.
> 
> Move platform initialization of SEV/SNP from CCP driver probe time to
> KVM module load time so that KVM can do SEV/SNP platform initialization
> explicitly if it actually wants to use SEV/SNP functionality.
> 
> Add support for KVM to explicitly call into the CCP driver at load time
> to initialize SEV/SNP. If required, this behavior can be altered with KVM
> module parameters to not do SEV/SNP platform initialization at module load
> time. Additionally, a corresponding SEV/SNP platform shutdown is invoked
> during KVM module unload time.
> 
> Continue to support SEV deferred initialization as the user may have the
> file containing SEV persistent data for SEV INIT_EX available only later
> after module load/init.
> 
> Suggested-by: Sean Christopherson <[email protected]>
> Reviewed-by: Tom Lendacky <[email protected]>
> Signed-off-by: Ashish Kalra <[email protected]>
> Signed-off-by: Herbert Xu <[email protected]>
> Signed-off-by: Jack Wang <[email protected]>
> Signed-off-by: Sasha Levin <[email protected]>
> ---
>   arch/x86/kvm/svm/sev.c | 12 ++++++++++++
>   1 file changed, 12 insertions(+)
> 
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index 6cec696461da2..bd0344bacaf1e 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -2972,6 +2972,7 @@ void __init sev_set_cpu_caps(void)
>   void __init sev_hardware_setup(void)
>   {
>   	unsigned int eax, ebx, ecx, edx, sev_asid_count, sev_es_asid_count;
> +	struct sev_platform_init_args init_args = {0};
>   	bool sev_snp_supported = false;
>   	bool sev_es_supported = false;
>   	bool sev_supported = false;
> @@ -3088,6 +3089,15 @@ 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);
>   }
>   

I ran an AI-assisted backport review and checked the finding against
6.12.y. The backport appears to be missing the ordering used upstream
when kvm_amd is built-in.

Upstream ensures that the PSP driver is ready before KVM continues:

	if (IS_BUILTIN(CONFIG_KVM_AMD) && sev_module_init())
		goto out;

It also initializes the SNP RMP table from AMD IOMMU initialization:

	if (snp_rmptable_init())
		goto disable_snp;

6.12.y has no sev_module_init() call and still initializes the RMP table
through a later initcall:

	device_initcall(snp_rmptable_init);

The backport can therefore reach:

	init_args.probe = true;
	sev_platform_init(&init_args);

before the PSP device exists. The call returns -ENODEV and its result is
ignored. KVM can then enable SVM before SNP/RMP setup has completed. 
Since bb1c84647025 removed probe-time platform initialization, this can 
leave encrypted guests unable to start.

The relevant upstream commits are:

435b344a7042 ("crypto: ccp: Add external API interface for PSP module 
initialization")
44e70718df4f ("KVM: SVM: Ensure PSP module is initialized if KVM module 
is built-in")
409f45387c93 ("x86/sev: Fix broken SNP support with KVM module built-in")


https://lore.kernel.org/stable/[email protected]/


Could the KVM maintainers confirm whether 6.12.y needs this full series 
? Probably could be done after this cycle, but still noting it here.

thanks,
Harshit
>   void sev_hardware_unsetup(void)
> @@ -3103,6 +3113,8 @@ void sev_hardware_unsetup(void)
>   
>   	misc_cg_set_capacity(MISC_CG_RES_SEV, 0);
>   	misc_cg_set_capacity(MISC_CG_RES_SEV_ES, 0);
> +
> +	sev_platform_shutdown();
>   }
>   
>   int sev_cpu_init(struct svm_cpu_data *sd)