psa_cipher_finish: treat status and output length as sensitive
In `psa_cipher_finish()` and in the corresponding function in our built-in
implementation `mbedtls_psa_cipher_finish()`, treat `status` and
`*output_length` as sensitive variables whose value must not leak through a
timing side channel. This is important when doing decryption with unpadding,
where leaking the validity or amount of padding can enable a padding oracle
attack.
With this change, `psa_cipher_finish()` should be constant-time if the
underlying legacy function (including the cipher implementation) is.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
diff --git a/library/psa_crypto.c b/library/psa_crypto.c
index 9c28609..51ec72f 100644
--- a/library/psa_crypto.c
+++ b/library/psa_crypto.c
@@ -73,6 +73,8 @@
#include "mbedtls/psa_util.h"
#include "mbedtls/threading.h"
+#include "constant_time_internal.h"
+
#if defined(MBEDTLS_PSA_BUILTIN_ALG_HKDF) || \
defined(MBEDTLS_PSA_BUILTIN_ALG_HKDF_EXTRACT) || \
defined(MBEDTLS_PSA_BUILTIN_ALG_HKDF_EXPAND)
@@ -4692,13 +4694,27 @@
output_length);
exit:
- if (status == PSA_SUCCESS) {
- status = psa_cipher_abort(operation);
- } else {
- *output_length = 0;
- (void) psa_cipher_abort(operation);
+ /* C99 doesn't allow a declaration to follow a label */;
+ psa_status_t abort_status = psa_cipher_abort(operation);
+ /* Normally abort shouldn't fail unless the operation is in a bad
+ * state, in which case we'd expect finish to fail with the same error.
+ * So it doesn't matter much which call's error code we pick when both
+ * fail. However, in unauthenticated decryption specifically, the
+ * distinction between PSA_SUCCESS and PSA_ERROR_INVALID_PADDING is
+ * security-sensitive (risk of a padding oracle attack), so here we
+ * must not have a code path that depends on the value of status. */
+ if (abort_status != PSA_SUCCESS) {
+ status = abort_status;
}
+ /* Set *output_length to 0 if status != PSA_SUCCESS, without
+ * leaking the value of status through a timing side channel
+ * (status == PSA_ERROR_INVALID_PADDING is sensitive when doing
+ * unpadded decryption, due to the risk of padding oracle attack). */
+ mbedtls_ct_condition_t success =
+ mbedtls_ct_bool_not(mbedtls_ct_bool(status));
+ *output_length = mbedtls_ct_size_if_else_0(success, *output_length);
+
LOCAL_OUTPUT_FREE(output_external, output);
return status;