int blake2sp_init_key( blake2sp_state *S, const uint8_t outlen, const void *key, const uint8_t keylen )
{
if( !outlen || outlen > BLAKE2S_OUTBYTES ) return -1;
if( !key || !keylen || keylen > BLAKE2S_KEYBYTES ) return -1;
memset( S->buf, 0, sizeof( S->buf ) );
S->buflen = 0;
if( blake2sp_init_root( S->R, outlen, keylen ) < 0 )
return -1;
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
if( blake2sp_init_leaf( S->S[i], outlen, keylen, i ) < 0 ) return -1;
S->R->last_node = 1;
S->S[PARALLELISM_DEGREE - 1]->last_node = 1;
{
uint8_t block[BLAKE2S_BLOCKBYTES];
memset( block, 0, BLAKE2S_BLOCKBYTES );
memcpy( block, key, keylen );
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
blake2s_update( S->S[i], block, BLAKE2S_BLOCKBYTES );
secure_zero_memory( block, BLAKE2S_BLOCKBYTES ); /* Burn the key from stack */
}
return 0;
}
int blake2sp_init( blake2sp_state *S, const uint8_t outlen )
{
if( !outlen || outlen > BLAKE2S_OUTBYTES ) return -1;
memset( S->buf, 0, sizeof( S->buf ) );
S->buflen = 0;
if( blake2sp_init_root( S->R, outlen, 0 ) < 0 )
return -1;
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
if( blake2sp_init_leaf( S->S[i], outlen, 0, i ) < 0 ) return -1;
S->R->last_node = 1;
S->S[PARALLELISM_DEGREE - 1]->last_node = 1;
return 0;
}
ERL_NIF_TERM blake2sp_hash(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
{
uint8_t hash[PARALLELISM_DEGREE][BLAKE2S_OUTBYTES];
blake2s_state S[PARALLELISM_DEGREE][1];
blake2s_state FS[1];
ErlNifBinary input, key, salt, personal;
uint8_t out[BLAKE2S_OUTBYTES] = {0};
unsigned int outlen;
int i;
ERL_NIF_TERM tmphash[BLAKE2S_OUTBYTES];
if (argc != 5 || !enif_inspect_binary(env, argv[0], &input) ||
!enif_inspect_binary(env, argv[1], &key) ||
!enif_get_uint(env, argv[2], &outlen) ||
!enif_inspect_binary(env, argv[3], &salt) ||
!enif_inspect_binary(env, argv[4], &personal))
return enif_make_badarg(env);
if (!outlen || outlen > BLAKE2S_OUTBYTES) return -1;
if( key.size > BLAKE2S_KEYBYTES ) return -1;
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
if( blake2sp_init_leaf( S[i], outlen, key.size, i, salt.data,
personal.data, salt.size, personal.size) < 0 )
return -1;
S[PARALLELISM_DEGREE - 1]->last_node = 1; // mark last node
if( key.size > 0 )
{
uint8_t block[BLAKE2S_BLOCKBYTES];
memset( block, 0, BLAKE2S_BLOCKBYTES );
memcpy( block, key.data, key.size );
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
blake2s_update( S[i], block, BLAKE2S_BLOCKBYTES );
secure_zero_memory( block, BLAKE2S_BLOCKBYTES ); /* Burn the key from stack */
}
#if defined(_OPENMP)
#pragma omp parallel shared(S,hash), num_threads(PARALLELISM_DEGREE)
#else
for( size_t id__ = 0; id__ < PARALLELISM_DEGREE; ++id__ )
#endif
{
#if defined(_OPENMP)
size_t id__ = omp_get_thread_num();
#endif
uint64_t inlen__ = input.size;
const uint8_t *in__ = ( const uint8_t * )input.data;
in__ += id__ * BLAKE2S_BLOCKBYTES;
while( inlen__ >= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES )
{
blake2s_update( S[id__], in__, BLAKE2S_BLOCKBYTES );
in__ += PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
inlen__ -= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
}
if( inlen__ > id__ * BLAKE2S_BLOCKBYTES )
{
const size_t left = inlen__ - id__ * BLAKE2S_BLOCKBYTES;
const size_t len = left <= BLAKE2S_BLOCKBYTES ? left : BLAKE2S_BLOCKBYTES;
blake2s_update( S[id__], in__, len );
}
blake2s_final( S[id__], hash[id__], BLAKE2S_OUTBYTES );
}
if( blake2sp_init_root( FS, outlen, key.size, salt.data, personal.data, salt.size, personal.size) < 0 )
return -1;
FS->last_node = 1; // Mark as last node
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
blake2s_update( FS, hash[i], BLAKE2S_OUTBYTES );
blake2s_final( FS, out, outlen );;
for (i = 0; i < outlen; i++) {
tmphash[i] = enif_make_uint(env, out[i]);
}
return enif_make_list_from_array(env, tmphash, outlen);
}
int blake2sp( uint8_t *out, const void *in, const void *key, uint8_t outlen, uint64_t inlen, uint8_t keylen )
{
uint8_t hash[PARALLELISM_DEGREE][BLAKE2S_OUTBYTES];
blake2s_state S[PARALLELISM_DEGREE][1];
blake2s_state FS[1];
/* Verify parameters */
if ( NULL == in && inlen > 0 ) return -1;
if ( NULL == out ) return -1;
if ( NULL == key && keylen > 0) return -1;
if( !outlen || outlen > BLAKE2S_OUTBYTES ) return -1;
if( keylen > BLAKE2S_KEYBYTES ) return -1;
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
if( blake2sp_init_leaf( S[i], outlen, keylen, i ) < 0 ) return -1;
S[PARALLELISM_DEGREE - 1]->last_node = 1; // mark last node
if( keylen > 0 )
{
uint8_t block[BLAKE2S_BLOCKBYTES];
memset( block, 0, BLAKE2S_BLOCKBYTES );
memcpy( block, key, keylen );
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
blake2s_update( S[i], block, BLAKE2S_BLOCKBYTES );
secure_zero_memory( block, BLAKE2S_BLOCKBYTES ); /* Burn the key from stack */
}
#if defined(_OPENMP)
#pragma omp parallel shared(S,hash), num_threads(PARALLELISM_DEGREE)
#else
for( size_t id__ = 0; id__ < PARALLELISM_DEGREE; ++id__ )
#endif
{
#if defined(_OPENMP)
size_t id__ = omp_get_thread_num();
#endif
uint64_t inlen__ = inlen;
const uint8_t *in__ = ( const uint8_t * )in;
in__ += id__ * BLAKE2S_BLOCKBYTES;
while( inlen__ >= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES )
{
blake2s_update( S[id__], in__, BLAKE2S_BLOCKBYTES );
in__ += PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
inlen__ -= PARALLELISM_DEGREE * BLAKE2S_BLOCKBYTES;
}
if( inlen__ > id__ * BLAKE2S_BLOCKBYTES )
{
const size_t left = inlen__ - id__ * BLAKE2S_BLOCKBYTES;
const size_t len = left <= BLAKE2S_BLOCKBYTES ? left : BLAKE2S_BLOCKBYTES;
blake2s_update( S[id__], in__, len );
}
blake2s_final( S[id__], hash[id__], BLAKE2S_OUTBYTES );
}
if( blake2sp_init_root( FS, outlen, keylen ) < 0 )
return -1;
FS->last_node = 1;
for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
blake2s_update( FS, hash[i], BLAKE2S_OUTBYTES );
return blake2s_final( FS, out, outlen );
}