JannePeltonen replied on github web page:
platform/linux-generic/odp_crypto.c
line 139
@@ -165,24 +192,51 @@ null_crypto_routine(odp_crypto_op_param_t *param
ODP_UNUSED,
return ODP_CRYPTO_ALG_ERR_NONE;
}
+/* Mimic new OpenSSL 1.1.y API */
+#if OPENSSL_VERSION_NUMBER < 0x10100000L
+static HMAC_CTX *HMAC_CTX_new(void)
+{
+ HMAC_CTX *ctx = malloc(sizeof(*ctx));
+
+ HMAC_CTX_init(ctx);
+ return ctx;
+}
+
+static void HMAC_CTX_free(HMAC_CTX *ctx)
+{
+ HMAC_CTX_cleanup(ctx);
+ free(ctx);
+}
+#endif
+
static
-void packet_hmac_calculate(HMAC_CTX *ctx,
- odp_crypto_op_param_t *param,
- odp_crypto_generic_session_t *session,
- uint8_t *hash)
+void packet_hmac(odp_crypto_op_param_t *param,
+ odp_crypto_generic_session_t *session,
+ uint8_t *hash)
{
+ HMAC_CTX *ctx = local.ctx[session->idx].hmac;
odp_packet_t pkt = param->out_pkt;
uint32_t offset = param->auth_range.offset;
uint32_t len = param->auth_range.length;
ODP_ASSERT(offset + len <= odp_packet_len(pkt));
- HMAC_Init_ex(ctx,
- session->auth.key,
- session->auth.key_length,
- session->auth.evp_md,
- NULL);
+ if (!IS_VALID(session, hmac, local.id)) {
+ SET_VALID(session, hmac, local.id);
+ HMAC_Init_ex(ctx,
+ session->auth.key,
+ session->auth.key_length,
+ session->auth.evp_md,
+ NULL);
+ } else {
+ HMAC_Init_ex(ctx,
+ NULL,
+ session->auth.key_length,
Comment:
Why is this needed here? The key length was already set. If this is because of
some OpenSSL weirdness, an explanatory comment would be nice.
> JannePeltonen wrote
> The cipher_valid and hmac_valid bits are not cleared at session allocation,
> so they can be left to what they were in an old and freed session that
> occupied the same memory slot.
>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>> Then I'd follow "Maxim's rule" and do the renames as a separate follow-on PR.
>>> muvarov wrote
>>> I prefer the rule that new code goes in in clean way. Even if rest of the
>>> code used old rules.
>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>> That makes sense to me.
>>>>> Dmitry Eremin-Solenikov(lumag) wrote:
>>>>> Then I would prefer to use names w/o underscore for now and to change all
>>>>> of them later.
>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>> I don't see the need to do that at this point as these are internal
>>>>>> routines anyway. That type of restructure might be something to consider
>>>>>> for Caterpillar.
>>>>>>> Dmitry Eremin-Solenikov(lumag) wrote:
>>>>>>> Hmm. The rest of init/term functions do not have such prefix. Should we
>>>>>>> change them?
>>>>>>>> Dmitry Eremin-Solenikov(lumag) wrote:
>>>>>>>> No, `sizeof uint32_t`
>>>>>>>>> muvarov wrote
>>>>>>>>> `_odp_crypto_term_local`
>>>>>>>>>> muvarov wrote
>>>>>>>>>> `_odp_crypto_init_local`
>>>>>>>>>>> muvarov wrote
>>>>>>>>>>> is 32 everywhere MAX_SESSIONS?
https://github.com/Linaro/odp/pull/342#discussion_r157490855
updated_at 2017-12-18 14:00:46
