Re: [RFC PATCH v2 05/22] crypto: ccp: Enable SEV-TIO feature in the PSP when supported

[Alexey Kardashevskiy <aik@amd.com>]

On 22/3/25 22:50, Francesco Lavra wrote:
> On 2025-02-18 at 11:09, Alexey Kardashevskiy wrote:
>> @@ -601,6 +603,25 @@ struct sev_data_snp_addr {
>>   	u64 address;				/* In/Out */
>>   } __packed;
>>   
>> +/**
>> + * struct sev_data_snp_feature_info - SEV_CMD_SNP_FEATURE_INFO
>> command params
>> + *
>> + * @len: length of this struct
>> + * @ecx_in: subfunction index of CPUID Fn8000_0024
>> + * @feature_info_paddr: physical address of a page with
>> sev_snp_feature_info
>> + */
>> +#define SNP_FEATURE_FN8000_0024_EBX_X00_SEVTIO	1
> 
> According to the SNP firmware ABI spec, support for SEV TIO commands is
> indicated by bit 1 (bit 0 is for SEV legacy commands).

well, I wanted a bit number (which is 1) but this is wrong nevertheless:

present = (fi.ebx & SNP_FEATURE_FN8000_0024_EBX_X00_SEVTIO) != 0;

should be:

present = (fi.ebx & BIT(SNP_FEATURE_FN8000_0024_EBX_X00_SEVTIO)) != 0;

good spotting!

> 
>> +static int snp_get_feature_info(struct sev_device *sev, u32 ecx,
>> struct sev_snp_feature_info *fi)
>> +{
>> +	struct sev_user_data_snp_status status = { 0 };
>> +	int psp_ret = 0, ret;
>> +
>> +	ret = snp_platform_status_locked(sev, &status, &psp_ret);
>> +	if (ret)
>> +		return ret;
>> +	if (ret != SEV_RET_SUCCESS)
> 
> s/ret/psp_ret/

yeah I noticed this after posting. Thanks,

> 
>> +		return -EFAULT;
>> +	if (!status.feature_info)
>> +		return -ENOENT;
>> +
>> +	ret = snp_feature_info_locked(sev, ecx, fi, &psp_ret);
>> +	if (ret)
>> +		return ret;
>> +	if (ret != SEV_RET_SUCCESS)
> 
> Same here

-- 
Alexey