On 24/02/2022 10:59, Daniel P. Berrangé wrote:
> On Thu, Feb 24, 2022 at 06:14:05AM +0000, Dov Murik wrote:
>> Add a new field 'cpu0-id' to the response of query-sev-capabilities QMP
>> command. The value of the field is the base64-encoded 64-byte unique ID
>> of the CPU0 (socket 0), which can be used to retrieve the signed CEK of
>> the CPU from AMD's Key Distribution Service (KDS).
>>
>> Signed-off-by: Dov Murik <dovmu...@linux.ibm.com>
>>
>> ---
>>
>> v2:
>> - change encoding to Base64 (thanks Daniel)
>> - rename constant to SEV_CPU_UNIQUE_ID_LEN
>> ---
>> qapi/misc-target.json | 4 ++++
>> target/i386/sev.c | 27 +++++++++++++++++++++++++++
>> 2 files changed, 31 insertions(+)
>>
>> diff --git a/qapi/misc-target.json b/qapi/misc-target.json
>> index 4bc45d2474..c6d9ad69e1 100644
>> --- a/qapi/misc-target.json
>> +++ b/qapi/misc-target.json
>> @@ -177,6 +177,8 @@
>> #
>> # @cert-chain: PDH certificate chain (base64 encoded)
>> #
>> +# @cpu0-id: 64-byte unique ID of CPU0 (base64 encoded) (since 7.0)
>> +#
>> # @cbitpos: C-bit location in page table entry
>> #
>> # @reduced-phys-bits: Number of physical Address bit reduction when SEV is
>> @@ -187,6 +189,7 @@
>> { 'struct': 'SevCapability',
>> 'data': { 'pdh': 'str',
>> 'cert-chain': 'str',
>> + 'cpu0-id': 'str',
>> 'cbitpos': 'int',
>> 'reduced-phys-bits': 'int'},
>> 'if': 'TARGET_I386' }
>> @@ -205,6 +208,7 @@
>> #
>> # -> { "execute": "query-sev-capabilities" }
>> # <- { "return": { "pdh": "8CCDD8DDD", "cert-chain": "888CCCDDDEE",
>> +# "cpu0-id": "2lvmGwo+...61iEinw==",
>> # "cbitpos": 47, "reduced-phys-bits": 5}}
>> #
>> ##
>> diff --git a/target/i386/sev.c b/target/i386/sev.c
>> index 025ff7a6f8..d3d2680e16 100644
>> --- a/target/i386/sev.c
>> +++ b/target/i386/sev.c
>> @@ -82,6 +82,8 @@ struct SevGuestState {
>> #define DEFAULT_GUEST_POLICY 0x1 /* disable debug */
>> #define DEFAULT_SEV_DEVICE "/dev/sev"
>>
>> +#define SEV_CPU_UNIQUE_ID_LEN 64
>
> Is this fixed size actually guaranteed by AMD ? In reading the spec
> for "GET_ID" it feels like they're intentionally not specifying a
> fixed length, pushing towards querying once and then re-trying with
> the reported buffer size:
>
> "If the value of the ID_LEN field is too small, an
> INVALID_LENGTH error is returned and the minimum
> required length is written to the field"
>
> I didn't find where it says 64 bytes exactly.
>
OK.
I'll change it to dynamically allocate the ID buffer based on the length
returned by the first query, similar to sev_get_pdh_info().
I'll remove the explicit lengths from the qapi description and the
commit message.
-Dov
>> +
>> #define SEV_INFO_BLOCK_GUID "00f771de-1a7e-4fcb-890e-68c77e2fb44e"
>> typedef struct __attribute__((__packed__)) SevInfoBlock {
>> /* SEV-ES Reset Vector Address */
>> @@ -531,11 +533,31 @@ e_free:
>> return 1;
>> }
>>
>> +static int
>> +sev_get_id(int fd, guchar *id_buf, size_t id_buf_len, Error **errp)
>> +{
>> + struct sev_user_data_get_id2 id = {
>> + .address = (unsigned long)id_buf,
>> + .length = id_buf_len
>> + };
>> + int err, r;
>> +
>> + r = sev_platform_ioctl(fd, SEV_GET_ID2, &id, &err);
>> + if (r < 0) {
>> + error_setg(errp, "SEV: Failed to get ID ret=%d fw_err=%d (%s)",
>> + r, err, fw_error_to_str(err));
>> + return 1;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> static SevCapability *sev_get_capabilities(Error **errp)
>> {
>> SevCapability *cap = NULL;
>> guchar *pdh_data = NULL;
>> guchar *cert_chain_data = NULL;
>> + guchar cpu0_id[SEV_CPU_UNIQUE_ID_LEN];
>> size_t pdh_len = 0, cert_chain_len = 0;
>> uint32_t ebx;
>> int fd;
>> @@ -561,9 +583,14 @@ static SevCapability *sev_get_capabilities(Error **errp)
>> goto out;
>> }
>>
>> + if (sev_get_id(fd, cpu0_id, sizeof(cpu0_id), errp)) {
>> + goto out;
>> + }
>> +
>> cap = g_new0(SevCapability, 1);
>> cap->pdh = g_base64_encode(pdh_data, pdh_len);
>> cap->cert_chain = g_base64_encode(cert_chain_data, cert_chain_len);
>> + cap->cpu0_id = g_base64_encode(cpu0_id, sizeof(cpu0_id));
>>
>> host_cpuid(0x8000001F, 0, NULL, &ebx, NULL, NULL);
>> cap->cbitpos = ebx & 0x3f;
>> --
>> 2.25.1
>>
>
> Regards,
> Daniel
