Only copy W from mem on C variant

the inline assembly version doesn't need to allocate an additional
buffer in order to work with W data as it uses hardware vectors
instead. However the C variant will overwrite the input buffer if it's
used directly as W.

sha256_compress.c

@@ -379,7 +379,7 @@
       "memory"                                                              \
   ); } while (0)
 
-void sha2_transform(base_type* _h, base_type* w) {
+void sha2_transform(base_type* _h, unsigned char* w) {
   vector_base_type a, b, c, d, e, f, g, h;
   vector int vRb;
 

sha2_common.c

@@ -152,25 +152,14 @@ int sha2(unsigned char *input, size_t size, size_t padded_size) {
   // Concatenate '1' to input.
   input[size] = (unsigned char)(1 << 7);
 
-  // Swap bytes due to endianess .
-  unsigned char* input_swapped = (unsigned char *) calloc(padded_size, sizeof(unsigned char));
-  if (input_swapped == NULL) {
-    fprintf(stderr, "%s\n.", strerror(errno));
-    return errno;
-  }
-  //swap_bytes(input, input_swapped, padded_size);
-
   // write total message size at the end (2 base_types).
   write_size(input, size, padded_size - 2 * base_type_size);
   swap_bytes(input + padded_size - 2 * base_type_size,
       input + padded_size - 2 * base_type_size, 2 * base_type_size);
 
   // Sha compression process.
-  for (size_t i = 0; i < padded_size; i = i + BLOCK_SIZE) {
-    base_type w[W_SIZE];
-    memcpy(w, input + i, 16 * sizeof(base_type));
-    sha2_transform(_h, w);
-  }
+  for (size_t i = 0; i < padded_size; i = i + BLOCK_SIZE)
+    sha2_transform(_h, input + i);
 
   printf(
 #if SHA_BITS == 256

sha2_compress.h

@@ -3,6 +3,6 @@
 
 #include "base-types.h"
 
-void sha2_transform(base_type* _h, base_type* w);
+void sha2_transform(base_type* _h, unsigned char* w);
 
 #endif // _PPC64_LE_SHA2_COMPRESS_H_

sha2_compress_c.c

@@ -55,9 +55,11 @@ void static inline sha2_round(base_type* a, base_type* b, base_type* c,
   *a = tmp1 + tmp2;
 }
 
-void sha2_transform(base_type* _h, base_type* w) {
+void sha2_transform(base_type* _h, unsigned char* w_in) {
   base_type a, b, c, d, e, f, g, h;
   int i;
+  // using a local w as it's going to be modified
+  base_type w[W_SIZE];
 
   a = _h[0];
   b = _h[1];
@@ -68,7 +70,8 @@ void sha2_transform(base_type* _h, base_type* w) {
   g = _h[6];
   h = _h[7];
 
-  swap_bytes((unsigned char*)w, (unsigned char*)w, BLOCK_SIZE);
+  // Only first 16 bytes needs to be copied as the ones above will be defined
+  swap_bytes(w_in, (unsigned char*)w, BLOCK_SIZE);
 
   // Loop unrolling, from 0 to 15
   for (i = 0; i < 16; i++) {

sha512_compress.c

@@ -449,7 +449,7 @@
       "memory"                                                              \
   ); } while (0)
 
-void sha2_transform(base_type* _h, base_type* w) {
+void sha2_transform(base_type* _h, unsigned char* w) {
   vector_base_type a, b, c, d, e, f, g, h;
   vector int vRb;