1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * The "hash function" used as the core of the ChaCha stream cipher (RFC7539)
4 *
5 * Copyright (C) 2015 Martin Willi
6 */
7
8#include <linux/bug.h>
9#include <linux/kernel.h>
10#include <linux/export.h>
11#include <linux/bitops.h>
12#include <linux/string.h>
13#include <asm/unaligned.h>
14#include <crypto/chacha.h>
15
16static void chacha_permute(u32 *x, int nrounds)
17{
18	int i;
19
20	/* whitelist the allowed round counts */
21	WARN_ON_ONCE(nrounds != 20 && nrounds != 12);
22
23	for (i = 0; i < nrounds; i += 2) {
24		x[0]  += x[4];    x[12] = rol32(x[12] ^ x[0],  16);
25		x[1]  += x[5];    x[13] = rol32(x[13] ^ x[1],  16);
26		x[2]  += x[6];    x[14] = rol32(x[14] ^ x[2],  16);
27		x[3]  += x[7];    x[15] = rol32(x[15] ^ x[3],  16);
28
29		x[8]  += x[12];   x[4]  = rol32(x[4]  ^ x[8],  12);
30		x[9]  += x[13];   x[5]  = rol32(x[5]  ^ x[9],  12);
31		x[10] += x[14];   x[6]  = rol32(x[6]  ^ x[10], 12);
32		x[11] += x[15];   x[7]  = rol32(x[7]  ^ x[11], 12);
33
34		x[0]  += x[4];    x[12] = rol32(x[12] ^ x[0],   8);
35		x[1]  += x[5];    x[13] = rol32(x[13] ^ x[1],   8);
36		x[2]  += x[6];    x[14] = rol32(x[14] ^ x[2],   8);
37		x[3]  += x[7];    x[15] = rol32(x[15] ^ x[3],   8);
38
39		x[8]  += x[12];   x[4]  = rol32(x[4]  ^ x[8],   7);
40		x[9]  += x[13];   x[5]  = rol32(x[5]  ^ x[9],   7);
41		x[10] += x[14];   x[6]  = rol32(x[6]  ^ x[10],  7);
42		x[11] += x[15];   x[7]  = rol32(x[7]  ^ x[11],  7);
43
44		x[0]  += x[5];    x[15] = rol32(x[15] ^ x[0],  16);
45		x[1]  += x[6];    x[12] = rol32(x[12] ^ x[1],  16);
46		x[2]  += x[7];    x[13] = rol32(x[13] ^ x[2],  16);
47		x[3]  += x[4];    x[14] = rol32(x[14] ^ x[3],  16);
48
49		x[10] += x[15];   x[5]  = rol32(x[5]  ^ x[10], 12);
50		x[11] += x[12];   x[6]  = rol32(x[6]  ^ x[11], 12);
51		x[8]  += x[13];   x[7]  = rol32(x[7]  ^ x[8],  12);
52		x[9]  += x[14];   x[4]  = rol32(x[4]  ^ x[9],  12);
53
54		x[0]  += x[5];    x[15] = rol32(x[15] ^ x[0],   8);
55		x[1]  += x[6];    x[12] = rol32(x[12] ^ x[1],   8);
56		x[2]  += x[7];    x[13] = rol32(x[13] ^ x[2],   8);
57		x[3]  += x[4];    x[14] = rol32(x[14] ^ x[3],   8);
58
59		x[10] += x[15];   x[5]  = rol32(x[5]  ^ x[10],  7);
60		x[11] += x[12];   x[6]  = rol32(x[6]  ^ x[11],  7);
61		x[8]  += x[13];   x[7]  = rol32(x[7]  ^ x[8],   7);
62		x[9]  += x[14];   x[4]  = rol32(x[4]  ^ x[9],   7);
63	}
64}
65
66/**
67 * chacha_block_generic - generate one keystream block and increment block counter
68 * @state: input state matrix (16 32-bit words)
69 * @stream: output keystream block (64 bytes)
70 * @nrounds: number of rounds (20 or 12; 20 is recommended)
71 *
72 * This is the ChaCha core, a function from 64-byte strings to 64-byte strings.
73 * The caller has already converted the endianness of the input.  This function
74 * also handles incrementing the block counter in the input matrix.
75 */
76void chacha_block_generic(u32 *state, u8 *stream, int nrounds)
77{
78	u32 x[16];
79	int i;
80
81	memcpy(x, state, 64);
82
83	chacha_permute(x, nrounds);
84
85	for (i = 0; i < ARRAY_SIZE(x); i++)
86		put_unaligned_le32(x[i] + state[i], &stream[i * sizeof(u32)]);
87
88	state[12]++;
89}
90EXPORT_SYMBOL(chacha_block_generic);
91
92/**
93 * hchacha_block_generic - abbreviated ChaCha core, for XChaCha
94 * @state: input state matrix (16 32-bit words)
95 * @stream: output (8 32-bit words)
96 * @nrounds: number of rounds (20 or 12; 20 is recommended)
97 *
98 * HChaCha is the ChaCha equivalent of HSalsa and is an intermediate step
99 * towards XChaCha (see https://cr.yp.to/snuffle/xsalsa-20081128.pdf).  HChaCha
100 * skips the final addition of the initial state, and outputs only certain words
101 * of the state.  It should not be used for streaming directly.
102 */
103void hchacha_block_generic(const u32 *state, u32 *stream, int nrounds)
104{
105	u32 x[16];
106
107	memcpy(x, state, 64);
108
109	chacha_permute(x, nrounds);
110
111	memcpy(&stream[0], &x[0], 16);
112	memcpy(&stream[4], &x[12], 16);
113}
114EXPORT_SYMBOL(hchacha_block_generic);
115