Re: [PATCH v5] x86/sev: Add SEV-SNP guest feature negotiation support

["Nikunj A. Dadhania" <nikunj@amd.com>]

On 13/01/23 17:23, Zhi Wang wrote:
> On Thu, 12 Jan 2023 14:11:39 +0530
> Nikunj A Dadhania <nikunj@amd.com> wrote:
> 
>> The hypervisor can enable various new features (SEV_FEATURES[1:63])
>> and start the SNP guest. Some of these features need guest side
>> implementation. If any of these features are enabled without guest
>> side implementation, the behavior of the SNP guest will be undefined.
>> The SNP guest boot may fail in a non-obvious way making it difficult
>> to debug.
>>
>> Instead of allowing the guest to continue and have it fail randomly
>> later, detect this early and fail gracefully.
>>
>> SEV_STATUS MSR indicates features which the hypervisor has enabled.
>> While booting, SNP guests should ascertain that all the enabled
>> features have guest side implementation. In case any feature is not
>> implemented in the guest, the guest terminates booting with GHCB
>> protocol Non-Automatic Exit(NAE) termination request event[1]. Populate
>> SW_EXITINFO2 with mask of unsupported features that the hypervisor
>> can easily report to the user.
>>
>> More details in AMD64 APM[2] Vol 2: 15.34.10 SEV_STATUS MSR
>>
>> [1] https://developer.amd.com/wp-content/resources/56421.pdf
>>     4.1.13 Termination Request
>>
>> [2] https://www.amd.com/system/files/TechDocs/40332_4.05.pdf
>>
> 
> The link of [2] is broken. Better update it.

That is strange, I will fix that.

>> +
>> +void snp_check_features(void)
>> +{
>> +	u64 unsupported_features;
>> +
>> +	if (!(sev_status & MSR_AMD64_SEV_SNP_ENABLED))
>> +		return;
>> +
>> +	/*
>> +	 * Terminate the boot if hypervisor has enabled any feature
>> +	 * lacking guest side implementation.
>> +	 */
>> +	unsupported_features = sev_status & SNP_FEATURES_IMPL_REQ &
>> ~SNP_FEATURES_PRESENT;
>> +	if (unsupported_features) {
>> +		if (sev_es_get_ghcb_version() < 2 ||
>> +		    (!boot_ghcb && !early_setup_ghcb()))
>> +			sev_es_terminate(SEV_TERM_SET_GEN,
>> GHCB_SNP_UNSUPPORTED); +
> 
> ===
>> +		u64 exit_info_1 =
>> SVM_VMGEXIT_TERM_REASON(SEV_TERM_SET_GEN, GHCB_SNP_UNSUPPORTED); +
>> +		vc_ghcb_invalidate(boot_ghcb);
>> +		ghcb_set_sw_exit_code(boot_ghcb,
>> SVM_VMGEXIT_TERM_REQUEST);
>> +		ghcb_set_sw_exit_info_1(boot_ghcb, exit_info_1);
>> +		ghcb_set_sw_exit_info_2(boot_ghcb,
>> unsupported_features); +
>> +		sev_es_wr_ghcb_msr(__pa(boot_ghcb));
>> +		VMGEXIT();
>> +
>> +		while (true)
>> +			asm volatile("hlt\n" : : : "memory");
> ===
> 
> This seems another approach to terminate the guest which can bring extra
> reason info compared to sev_es_terminate(). It would be better to wrap
> the above snippet into a function and call it here.

Right, I will add it as part of the new function in my next revision:

static void __noreturn sev_es_ghcb_terminate(struct ghcb *ghcb, u64 exit_info_1, u64 exit_info_2)

Regards
Nikunj