/**
* blockmix_salsa8(Bin, Bout, X, r):
* Compute Bout = BlockMix_{salsa20/8, r}(Bin). The input Bin must be 128r
* bytes in length; the output Bout must also be the same size. The
* temporary space X must be 64 bytes.
*/
static void
blockmix_salsa8(__m128i * Bin, __m128i * Bout, __m128i * X, size_t r)
{
size_t i;
/* 1: X <-- B_{2r - 1} */
blkcpy(X, &Bin[8 * r - 4], 64);
/* 2: for i = 0 to 2r - 1 do */
for (i = 0; i < r; i++) {
/* 3: X <-- H(X \xor B_i) */
blkxor(X, &Bin[i * 8], 64);
salsa20_8(X);
/* 4: Y_i <-- X */
/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
blkcpy(&Bout[i * 4], X, 64);
/* 3: X <-- H(X \xor B_i) */
blkxor(X, &Bin[i * 8 + 4], 64);
salsa20_8(X);
/* 4: Y_i <-- X */
/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
blkcpy(&Bout[(r + i) * 4], X, 64);
}
}
/**
* blockmix_salsa8(Bin, Bout, X, r):
* Compute Bout = BlockMix_{salsa20/8, r}(Bin). The input Bin must be 128r
* bytes in length; the output Bout must also be the same size. The
* temporary space X must be 64 bytes.
*/
static void
blockmix_salsa8(uint32_t * Bin, uint32_t * Bout, uint32_t * X, size_t r)
{
size_t i;
/* 1: X <-- B_{2r - 1} */
blkcpy(X, &Bin[(2 * r - 1) * 16], 64);
/* 2: for i = 0 to 2r - 1 do */
for (i = 0; i < 2 * r; i += 2) {
/* 3: X <-- H(X \xor B_i) */
blkxor(X, &Bin[i * 16], 64);
salsa20_8(X);
/* 4: Y_i <-- X */
/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
blkcpy(&Bout[i * 8], X, 64);
/* 3: X <-- H(X \xor B_i) */
blkxor(X, &Bin[i * 16 + 16], 64);
salsa20_8(X);
/* 4: Y_i <-- X */
/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
blkcpy(&Bout[i * 8 + r * 16], X, 64);
}
}
/**
* smix(B, r, N, V, XY):
* Compute B = SMix_r(B, N). The input B must be 128r bytes in length;
* the temporary storage V must be 128rN bytes in length; the temporary
* storage XY must be 256r + 64 bytes in length. The value N must be a
* power of 2 greater than 1. The arrays B, V, and XY must be aligned to a
* multiple of 64 bytes.
*/
static void
smix(uint8_t * B, size_t r, uint64_t N, void * V, void * XY)
{
__m128i * X = XY;
__m128i * Y = (void *)((uintptr_t)(XY) + 128 * r);
__m128i * Z = (void *)((uintptr_t)(XY) + 256 * r);
uint32_t * X32 = (void *)X;
uint64_t i, j;
size_t k;
/* 1: X <-- B */
for (k = 0; k < 2 * r; k++) {
for (i = 0; i < 16; i++) {
X32[k * 16 + i] =
le32dec(&B[(k * 16 + (i * 5 % 16)) * 4]);
}
}
/* 2: for i = 0 to N - 1 do */
for (i = 0; i < N; i += 2) {
/* 3: V_i <-- X */
blkcpy((void *)((uintptr_t)(V) + i * 128 * r), X, 128 * r);
/* 4: X <-- H(X) */
blockmix_salsa8(X, Y, Z, r);
/* 3: V_i <-- X */
blkcpy((void *)((uintptr_t)(V) + (i + 1) * 128 * r),
Y, 128 * r);
/* 4: X <-- H(X) */
blockmix_salsa8(Y, X, Z, r);
}
/* 6: for i = 0 to N - 1 do */
for (i = 0; i < N; i += 2) {
/* 7: j <-- Integerify(X) mod N */
j = integerify(X, r) & (N - 1);
/* 8: X <-- H(X \xor V_j) */
blkxor(X, (void *)((uintptr_t)(V) + j * 128 * r), 128 * r);
blockmix_salsa8(X, Y, Z, r);
/* 7: j <-- Integerify(X) mod N */
j = integerify(Y, r) & (N - 1);
/* 8: X <-- H(X \xor V_j) */
blkxor(Y, (void *)((uintptr_t)(V) + j * 128 * r), 128 * r);
blockmix_salsa8(Y, X, Z, r);
}
/* 10: B' <-- X */
for (k = 0; k < 2 * r; k++) {
for (i = 0; i < 16; i++) {
le32enc(&B[(k * 16 + (i * 5 % 16)) * 4],
X32[k * 16 + i]);
}
}
}
/**
* crypto_scrypt_smix(B, r, N, V, XY):
* Compute B = SMix_r(B, N). The input B must be 128r bytes in length;
* the temporary storage V must be 128rN bytes in length; the temporary
* storage XY must be 256r + 64 bytes in length. The value N must be a
* power of 2 greater than 1. The arrays B, V, and XY must be aligned to a
* multiple of 64 bytes.
*/
void
crypto_scrypt_smix(uint8_t * B, size_t r, uint64_t N, void * _V, void * XY)
{
uint32_t * X = XY;
uint32_t * Y = (void *)((uint8_t *)(XY) + 128 * r);
uint32_t * Z = (void *)((uint8_t *)(XY) + 256 * r);
uint32_t * V = _V;
uint64_t i;
uint64_t j;
size_t k;
/* 1: X <-- B */
for (k = 0; k < 32 * r; k++)
X[k] = le32dec(&B[4 * k]);
/* 2: for i = 0 to N - 1 do */
for (i = 0; i < N; i += 2) {
/* 3: V_i <-- X */
blkcpy(&V[i * (32 * r)], X, 128 * r);
/* 4: X <-- H(X) */
blockmix_salsa8(X, Y, Z, r);
/* 3: V_i <-- X */
blkcpy(&V[(i + 1) * (32 * r)], Y, 128 * r);
/* 4: X <-- H(X) */
blockmix_salsa8(Y, X, Z, r);
}
/* 6: for i = 0 to N - 1 do */
for (i = 0; i < N; i += 2) {
/* 7: j <-- Integerify(X) mod N */
j = integerify(X, r) & (N - 1);
/* 8: X <-- H(X \xor V_j) */
blkxor(X, &V[j * (32 * r)], 128 * r);
blockmix_salsa8(X, Y, Z, r);
/* 7: j <-- Integerify(X) mod N */
j = integerify(Y, r) & (N - 1);
/* 8: X <-- H(X \xor V_j) */
blkxor(Y, &V[j * (32 * r)], 128 * r);
blockmix_salsa8(Y, X, Z, r);
}
/* 10: B' <-- X */
for (k = 0; k < 32 * r; k++)
le32enc(&B[4 * k], X[k]);
}
/**
* smix(B, r, N, V, XY):
* Compute B = SMix_r(B, N). The input B must be 128r bytes in length; the
* temporary storage V must be 128rN bytes in length; the temporary storage
* XY must be 256r bytes in length. The value N must be a power of 2.
*/
static void
smix(uint8_t * B, size_t r, uint64_t N, uint8_t * V, uint8_t * XY)
{
uint8_t * X = XY;
uint8_t * Y = &XY[128 * r];
uint64_t i;
uint64_t j;
/* 1: X <-- B */
blkcpy(X, B, 128 * r);
/* 2: for i = 0 to N - 1 do */
for (i = 0; i < N; i++) {
/* 3: V_i <-- X */
blkcpy(&V[i * (128 * r)], X, 128 * r);
/* 4: X <-- H(X) */
blockmix_salsa8(X, Y, r);
}
/* 6: for i = 0 to N - 1 do */
for (i = 0; i < N; i++) {
/* 7: j <-- Integerify(X) mod N */
j = integerify(X, r) & (N - 1);
/* 8: X <-- H(X \xor V_j) */
blkxor(X, &V[j * (128 * r)], 128 * r);
blockmix_salsa8(X, Y, r);
}
/* 10: B' <-- X */
blkcpy(B, X, 128 * r);
}
/**
* blockmix_salsa8(B, Y, r):
* Compute B = BlockMix_{salsa20/8, r}(B). The input B must be 128r bytes in
* length; the temporary space Y must also be the same size.
*/
static void
blockmix_salsa8(uint8_t * B, uint8_t * Y, size_t r)
{
uint8_t X[64];
size_t i;
/* 1: X <-- B_{2r - 1} */
blkcpy(X, &B[(2 * r - 1) * 64], 64);
/* 2: for i = 0 to 2r - 1 do */
for (i = 0; i < 2 * r; i++) {
/* 3: X <-- H(X \xor B_i) */
blkxor(X, &B[i * 64], 64);
salsa20_8(X);
/* 4: Y_i <-- X */
blkcpy(&Y[i * 64], X, 64);
}
/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
for (i = 0; i < r; i++)
blkcpy(&B[i * 64], &Y[(i * 2) * 64], 64);
for (i = 0; i < r; i++)
blkcpy(&B[(i + r) * 64], &Y[(i * 2 + 1) * 64], 64);
}
bool
ocb::decrypt (void *_ptext, u_int64_t nonce, const void *_ctext,
const blk *tag, size_t len) const
{
char *ptext = static_cast <char *> (_ptext);
const char *ctext = static_cast <const char *> (_ctext);
blk r;
blkclear (&r);
puthyper (r.c + (r.nc - 8), nonce);
blkxor (&r, l[0]);
k.encipher_bytes (r.c);
blk s;
blkclear (&s);
size_t i = 1;
blk tmp;
while (len > blk::nc) {
blkxor (&r, l[ffs (i) - 1]);
tmp.get (ctext);
blkxor (&tmp, r);
k.decipher_bytes (tmp.c);
blkxor (&tmp, r);
tmp.put (ptext);
blkxor (&s, tmp);
ptext += blk::nc;
ctext += blk::nc;
len -= blk::nc;
i++;
};
blkxor (&r, l[ffs (i) - 1]);
blkxor (&tmp, l[-1], r);
tmp.c[tmp.nc - 1] ^= len << 3;
k.encipher_bytes (tmp.c);
blkxor (&s, tmp);
for (u_int b = 0; b < len; b++) {
s.c[b] ^= ctext[b];
ptext[b] = tmp.c[b] ^ ctext[b];
}
blkxor (&tmp, s, r);
k.encipher_bytes (tmp.c);
return !memcmp (tag->c, tmp.c, tag->nc);
}