upstream: github.com/robur-coop/kdf
1#include <stdint.h>
2
3#define CAML_NAME_SPACE
4#include <caml/mlvalues.h>
5#include <caml/bigarray.h>
6
7static inline uint32_t r(uint32_t a, int b) {
8 int rs = 32 - b;
9 return (a << b) | (a >> rs);
10}
11
12static inline uint32_t combine(uint32_t y0, uint32_t y1, uint32_t y2, int shift) {
13 return r(y1 + y2, shift) ^ y0;
14}
15
16static inline void quarterround(uint32_t *x, int y0, int y1, int y2, int y3) {
17 x[y1] = combine(x[y1], x[y0], x[y3], 7);
18 x[y2] = combine(x[y2], x[y1], x[y0], 9);
19 x[y3] = combine(x[y3], x[y2], x[y1], 13);
20 x[y0] = combine(x[y0], x[y3], x[y2], 18);
21}
22
23static inline uint32_t get_u32_le(const uint8_t *input, int offset) {
24 return input[offset]
25 | (input[offset + 1] << 8)
26 | (input[offset + 2] << 16)
27 | (input[offset + 3] << 24);
28}
29
30static inline void set_u32_le(uint8_t *input, int offset, uint32_t value) {
31 input[offset] = (uint8_t) value;
32 input[offset + 1] = (uint8_t) (value >> 8);
33 input[offset + 2] = (uint8_t) (value >> 16);
34 input[offset + 3] = (uint8_t) (value >> 24);
35}
36
37static void salsa_core(int count, const uint8_t *src, uint8_t *dst) {
38 uint32_t x[16];
39 for (int i = 0; i < 16; i++) {
40 x[i] = get_u32_le(src, i * 4);
41 }
42 for (int i = 0; i < count; i++) {
43 quarterround(x, 0, 4, 8, 12);
44 quarterround(x, 5, 9, 13, 1);
45 quarterround(x, 10, 14, 2, 6);
46 quarterround(x, 15, 3, 7, 11);
47
48 quarterround(x, 0, 1, 2, 3);
49 quarterround(x, 5, 6, 7, 4);
50 quarterround(x, 10, 11, 8, 9);
51 quarterround(x, 15, 12, 13, 14);
52 }
53 for (int i = 0; i < 16; i++) {
54 uint32_t xi = x[i];
55 uint32_t hj = get_u32_le(src, i * 4);
56 set_u32_le(dst, i * 4, xi + hj);
57 }
58}
59
60CAMLprim value
61caml_salsa_core(value count, value src, value dst)
62{
63 salsa_core(Int_val(count), (const uint8_t*)(String_val(src)), Bytes_val(dst));
64 return Val_unit;
65}