blake2b.c

  1// SPDX-License-Identifier: GPL-2.0 OR MIT
  2/*
  3 * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
  4 * Copyright 2025 Google LLC
  5 *
  6 * This is an implementation of the BLAKE2b hash and PRF functions.
  7 *
  8 * Information: https://blake2.net/
  9 */
 10
 11#include <crypto/blake2b.h>
 12#include <linux/bug.h>
 13#include <linux/export.h>
 14#include <linux/kernel.h>
 15#include <linux/module.h>
 16#include <linux/string.h>
 17#include <linux/unroll.h>
 18#include <linux/types.h>
 19
 20static const u8 blake2b_sigma[12][16] = {
 21	{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
 22	{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
 23	{ 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
 24	{ 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
 25	{ 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
 26	{ 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 },
 27	{ 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 },
 28	{ 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 },
 29	{ 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 },
 30	{ 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0 },
 31	{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
 32	{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 }
 33};
 34
 35static inline void blake2b_increment_counter(struct blake2b_ctx *ctx, u32 inc)
 36{
 37	ctx->t[0] += inc;
 38	ctx->t[1] += (ctx->t[0] < inc);
 39}
 40
 41static void __maybe_unused
 42blake2b_compress_generic(struct blake2b_ctx *ctx,
 43			 const u8 *data, size_t nblocks, u32 inc)
 44{
 45	u64 m[16];
 46	u64 v[16];
 47	int i;
 48
 49	WARN_ON(IS_ENABLED(DEBUG) &&
 50		(nblocks > 1 && inc != BLAKE2B_BLOCK_SIZE));
 51
 52	while (nblocks > 0) {
 53		blake2b_increment_counter(ctx, inc);
 54		memcpy(m, data, BLAKE2B_BLOCK_SIZE);
 55		le64_to_cpu_array(m, ARRAY_SIZE(m));
 56		memcpy(v, ctx->h, 64);
 57		v[ 8] = BLAKE2B_IV0;
 58		v[ 9] = BLAKE2B_IV1;
 59		v[10] = BLAKE2B_IV2;
 60		v[11] = BLAKE2B_IV3;
 61		v[12] = BLAKE2B_IV4 ^ ctx->t[0];
 62		v[13] = BLAKE2B_IV5 ^ ctx->t[1];
 63		v[14] = BLAKE2B_IV6 ^ ctx->f[0];
 64		v[15] = BLAKE2B_IV7 ^ ctx->f[1];
 65
 66#define G(r, i, a, b, c, d) do { \
 67	a += b + m[blake2b_sigma[r][2 * i + 0]]; \
 68	d = ror64(d ^ a, 32); \
 69	c += d; \
 70	b = ror64(b ^ c, 24); \
 71	a += b + m[blake2b_sigma[r][2 * i + 1]]; \
 72	d = ror64(d ^ a, 16); \
 73	c += d; \
 74	b = ror64(b ^ c, 63); \
 75} while (0)
 76
 77#ifdef CONFIG_64BIT
 78		/*
 79		 * Unroll the rounds loop to enable constant-folding of the
 80		 * blake2b_sigma values.  Seems worthwhile on 64-bit kernels.
 81		 * Not worthwhile on 32-bit kernels because the code size is
 82		 * already so large there due to BLAKE2b using 64-bit words.
 83		 */
 84		unrolled_full
 85#endif
 86		for (int r = 0; r < 12; r++) {
 87			G(r, 0, v[0], v[4], v[8], v[12]);
 88			G(r, 1, v[1], v[5], v[9], v[13]);
 89			G(r, 2, v[2], v[6], v[10], v[14]);
 90			G(r, 3, v[3], v[7], v[11], v[15]);
 91			G(r, 4, v[0], v[5], v[10], v[15]);
 92			G(r, 5, v[1], v[6], v[11], v[12]);
 93			G(r, 6, v[2], v[7], v[8], v[13]);
 94			G(r, 7, v[3], v[4], v[9], v[14]);
 95		}
 96#undef G
 97
 98		for (i = 0; i < 8; ++i)
 99			ctx->h[i] ^= v[i] ^ v[i + 8];
100
101		data += BLAKE2B_BLOCK_SIZE;
102		--nblocks;
103	}
104}
105
106#ifdef CONFIG_CRYPTO_LIB_BLAKE2B_ARCH
107#include "blake2b.h" /* $(SRCARCH)/blake2b.h */
108#else
109#define blake2b_compress blake2b_compress_generic
110#endif
111
112static inline void blake2b_set_lastblock(struct blake2b_ctx *ctx)
113{
114	ctx->f[0] = -1;
115}
116
117void blake2b_update(struct blake2b_ctx *ctx, const u8 *in, size_t inlen)
118{
119	const size_t fill = BLAKE2B_BLOCK_SIZE - ctx->buflen;
120
121	if (unlikely(!inlen))
122		return;
123	if (inlen > fill) {
124		memcpy(ctx->buf + ctx->buflen, in, fill);
125		blake2b_compress(ctx, ctx->buf, 1, BLAKE2B_BLOCK_SIZE);
126		ctx->buflen = 0;
127		in += fill;
128		inlen -= fill;
129	}
130	if (inlen > BLAKE2B_BLOCK_SIZE) {
131		const size_t nblocks = DIV_ROUND_UP(inlen, BLAKE2B_BLOCK_SIZE);
132
133		blake2b_compress(ctx, in, nblocks - 1, BLAKE2B_BLOCK_SIZE);
134		in += BLAKE2B_BLOCK_SIZE * (nblocks - 1);
135		inlen -= BLAKE2B_BLOCK_SIZE * (nblocks - 1);
136	}
137	memcpy(ctx->buf + ctx->buflen, in, inlen);
138	ctx->buflen += inlen;
139}
140EXPORT_SYMBOL(blake2b_update);
141
142void blake2b_final(struct blake2b_ctx *ctx, u8 *out)
143{
144	WARN_ON(IS_ENABLED(DEBUG) && !out);
145	blake2b_set_lastblock(ctx);
146	memset(ctx->buf + ctx->buflen, 0,
147	       BLAKE2B_BLOCK_SIZE - ctx->buflen); /* Padding */
148	blake2b_compress(ctx, ctx->buf, 1, ctx->buflen);
149	cpu_to_le64_array(ctx->h, ARRAY_SIZE(ctx->h));
150	memcpy(out, ctx->h, ctx->outlen);
151	memzero_explicit(ctx, sizeof(*ctx));
152}
153EXPORT_SYMBOL(blake2b_final);
154
155#ifdef blake2b_mod_init_arch
156static int __init blake2b_mod_init(void)
157{
158	blake2b_mod_init_arch();
159	return 0;
160}
161subsys_initcall(blake2b_mod_init);
162
163static void __exit blake2b_mod_exit(void)
164{
165}
166module_exit(blake2b_mod_exit);
167#endif
168
169MODULE_DESCRIPTION("BLAKE2b hash function");
170MODULE_LICENSE("GPL");