sha3.c

  1// SPDX-License-Identifier: GPL-2.0-or-later
  2/*
  3 * SHA-3, as specified in
  4 * https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf
  5 *
  6 * SHA-3 code by Jeff Garzik <jeff@garzik.org>
  7 *               Ard Biesheuvel <ard.biesheuvel@linaro.org>
  8 *               David Howells <dhowells@redhat.com>
  9 *
 10 * See also Documentation/crypto/sha3.rst
 11 */
 12
 13#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 14#include <crypto/sha3.h>
 15#include <crypto/utils.h>
 16#include <linux/export.h>
 17#include <linux/kernel.h>
 18#include <linux/module.h>
 19#include <linux/unaligned.h>
 20#include "fips.h"
 21
 22/*
 23 * On some 32-bit architectures, such as h8300, GCC ends up using over 1 KB of
 24 * stack if the round calculation gets inlined into the loop in
 25 * sha3_keccakf_generic().  On the other hand, on 64-bit architectures with
 26 * plenty of [64-bit wide] general purpose registers, not inlining it severely
 27 * hurts performance.  So let's use 64-bitness as a heuristic to decide whether
 28 * to inline or not.
 29 */
 30#ifdef CONFIG_64BIT
 31#define SHA3_INLINE inline
 32#else
 33#define SHA3_INLINE noinline
 34#endif
 35
 36#define SHA3_KECCAK_ROUNDS 24
 37
 38static const u64 sha3_keccakf_rndc[SHA3_KECCAK_ROUNDS] = {
 39	0x0000000000000001ULL, 0x0000000000008082ULL, 0x800000000000808aULL,
 40	0x8000000080008000ULL, 0x000000000000808bULL, 0x0000000080000001ULL,
 41	0x8000000080008081ULL, 0x8000000000008009ULL, 0x000000000000008aULL,
 42	0x0000000000000088ULL, 0x0000000080008009ULL, 0x000000008000000aULL,
 43	0x000000008000808bULL, 0x800000000000008bULL, 0x8000000000008089ULL,
 44	0x8000000000008003ULL, 0x8000000000008002ULL, 0x8000000000000080ULL,
 45	0x000000000000800aULL, 0x800000008000000aULL, 0x8000000080008081ULL,
 46	0x8000000000008080ULL, 0x0000000080000001ULL, 0x8000000080008008ULL
 47};
 48
 49/*
 50 * Perform a single round of Keccak mixing.
 51 */
 52static SHA3_INLINE void sha3_keccakf_one_round_generic(u64 st[25], int round)
 53{
 54	u64 t[5], tt, bc[5];
 55
 56	/* Theta */
 57	bc[0] = st[0] ^ st[5] ^ st[10] ^ st[15] ^ st[20];
 58	bc[1] = st[1] ^ st[6] ^ st[11] ^ st[16] ^ st[21];
 59	bc[2] = st[2] ^ st[7] ^ st[12] ^ st[17] ^ st[22];
 60	bc[3] = st[3] ^ st[8] ^ st[13] ^ st[18] ^ st[23];
 61	bc[4] = st[4] ^ st[9] ^ st[14] ^ st[19] ^ st[24];
 62
 63	t[0] = bc[4] ^ rol64(bc[1], 1);
 64	t[1] = bc[0] ^ rol64(bc[2], 1);
 65	t[2] = bc[1] ^ rol64(bc[3], 1);
 66	t[3] = bc[2] ^ rol64(bc[4], 1);
 67	t[4] = bc[3] ^ rol64(bc[0], 1);
 68
 69	st[0] ^= t[0];
 70
 71	/* Rho Pi */
 72	tt = st[1];
 73	st[ 1] = rol64(st[ 6] ^ t[1], 44);
 74	st[ 6] = rol64(st[ 9] ^ t[4], 20);
 75	st[ 9] = rol64(st[22] ^ t[2], 61);
 76	st[22] = rol64(st[14] ^ t[4], 39);
 77	st[14] = rol64(st[20] ^ t[0], 18);
 78	st[20] = rol64(st[ 2] ^ t[2], 62);
 79	st[ 2] = rol64(st[12] ^ t[2], 43);
 80	st[12] = rol64(st[13] ^ t[3], 25);
 81	st[13] = rol64(st[19] ^ t[4],  8);
 82	st[19] = rol64(st[23] ^ t[3], 56);
 83	st[23] = rol64(st[15] ^ t[0], 41);
 84	st[15] = rol64(st[ 4] ^ t[4], 27);
 85	st[ 4] = rol64(st[24] ^ t[4], 14);
 86	st[24] = rol64(st[21] ^ t[1],  2);
 87	st[21] = rol64(st[ 8] ^ t[3], 55);
 88	st[ 8] = rol64(st[16] ^ t[1], 45);
 89	st[16] = rol64(st[ 5] ^ t[0], 36);
 90	st[ 5] = rol64(st[ 3] ^ t[3], 28);
 91	st[ 3] = rol64(st[18] ^ t[3], 21);
 92	st[18] = rol64(st[17] ^ t[2], 15);
 93	st[17] = rol64(st[11] ^ t[1], 10);
 94	st[11] = rol64(st[ 7] ^ t[2],  6);
 95	st[ 7] = rol64(st[10] ^ t[0],  3);
 96	st[10] = rol64(    tt ^ t[1],  1);
 97
 98	/* Chi */
 99	bc[ 0] = ~st[ 1] & st[ 2];
100	bc[ 1] = ~st[ 2] & st[ 3];
101	bc[ 2] = ~st[ 3] & st[ 4];
102	bc[ 3] = ~st[ 4] & st[ 0];
103	bc[ 4] = ~st[ 0] & st[ 1];
104	st[ 0] ^= bc[ 0];
105	st[ 1] ^= bc[ 1];
106	st[ 2] ^= bc[ 2];
107	st[ 3] ^= bc[ 3];
108	st[ 4] ^= bc[ 4];
109
110	bc[ 0] = ~st[ 6] & st[ 7];
111	bc[ 1] = ~st[ 7] & st[ 8];
112	bc[ 2] = ~st[ 8] & st[ 9];
113	bc[ 3] = ~st[ 9] & st[ 5];
114	bc[ 4] = ~st[ 5] & st[ 6];
115	st[ 5] ^= bc[ 0];
116	st[ 6] ^= bc[ 1];
117	st[ 7] ^= bc[ 2];
118	st[ 8] ^= bc[ 3];
119	st[ 9] ^= bc[ 4];
120
121	bc[ 0] = ~st[11] & st[12];
122	bc[ 1] = ~st[12] & st[13];
123	bc[ 2] = ~st[13] & st[14];
124	bc[ 3] = ~st[14] & st[10];
125	bc[ 4] = ~st[10] & st[11];
126	st[10] ^= bc[ 0];
127	st[11] ^= bc[ 1];
128	st[12] ^= bc[ 2];
129	st[13] ^= bc[ 3];
130	st[14] ^= bc[ 4];
131
132	bc[ 0] = ~st[16] & st[17];
133	bc[ 1] = ~st[17] & st[18];
134	bc[ 2] = ~st[18] & st[19];
135	bc[ 3] = ~st[19] & st[15];
136	bc[ 4] = ~st[15] & st[16];
137	st[15] ^= bc[ 0];
138	st[16] ^= bc[ 1];
139	st[17] ^= bc[ 2];
140	st[18] ^= bc[ 3];
141	st[19] ^= bc[ 4];
142
143	bc[ 0] = ~st[21] & st[22];
144	bc[ 1] = ~st[22] & st[23];
145	bc[ 2] = ~st[23] & st[24];
146	bc[ 3] = ~st[24] & st[20];
147	bc[ 4] = ~st[20] & st[21];
148	st[20] ^= bc[ 0];
149	st[21] ^= bc[ 1];
150	st[22] ^= bc[ 2];
151	st[23] ^= bc[ 3];
152	st[24] ^= bc[ 4];
153
154	/* Iota */
155	st[0] ^= sha3_keccakf_rndc[round];
156}
157
158/* Generic implementation of the Keccak-f[1600] permutation */
159static void sha3_keccakf_generic(struct sha3_state *state)
160{
161	/*
162	 * Temporarily convert the state words from little-endian to native-
163	 * endian so that they can be operated on.  Note that on little-endian
164	 * machines this conversion is a no-op and is optimized out.
165	 */
166
167	for (int i = 0; i < ARRAY_SIZE(state->words); i++)
168		state->native_words[i] = le64_to_cpu(state->words[i]);
169
170	for (int round = 0; round < SHA3_KECCAK_ROUNDS; round++)
171		sha3_keccakf_one_round_generic(state->native_words, round);
172
173	for (int i = 0; i < ARRAY_SIZE(state->words); i++)
174		state->words[i] = cpu_to_le64(state->native_words[i]);
175}
176
177/*
178 * Generic implementation of absorbing the given nonzero number of full blocks
179 * into the sponge function Keccak[r=8*block_size, c=1600-8*block_size].
180 */
181static void __maybe_unused
182sha3_absorb_blocks_generic(struct sha3_state *state, const u8 *data,
183			   size_t nblocks, size_t block_size)
184{
185	do {
186		for (size_t i = 0; i < block_size; i += 8)
187			state->words[i / 8] ^= get_unaligned((__le64 *)&data[i]);
188		sha3_keccakf_generic(state);
189		data += block_size;
190	} while (--nblocks);
191}
192
193#ifdef CONFIG_CRYPTO_LIB_SHA3_ARCH
194#include "sha3.h" /* $(SRCARCH)/sha3.h */
195#else
196#define sha3_keccakf		sha3_keccakf_generic
197#define sha3_absorb_blocks	sha3_absorb_blocks_generic
198#endif
199
200void __sha3_update(struct __sha3_ctx *ctx, const u8 *in, size_t in_len)
201{
202	const size_t block_size = ctx->block_size;
203	size_t absorb_offset = ctx->absorb_offset;
204
205	/* Warn if squeezing has already begun. */
206	WARN_ON_ONCE(absorb_offset >= block_size);
207
208	if (absorb_offset && absorb_offset + in_len >= block_size) {
209		crypto_xor(&ctx->state.bytes[absorb_offset], in,
210			   block_size - absorb_offset);
211		in += block_size - absorb_offset;
212		in_len -= block_size - absorb_offset;
213		sha3_keccakf(&ctx->state);
214		absorb_offset = 0;
215	}
216
217	if (in_len >= block_size) {
218		size_t nblocks = in_len / block_size;
219
220		sha3_absorb_blocks(&ctx->state, in, nblocks, block_size);
221		in += nblocks * block_size;
222		in_len -= nblocks * block_size;
223	}
224
225	if (in_len) {
226		crypto_xor(&ctx->state.bytes[absorb_offset], in, in_len);
227		absorb_offset += in_len;
228	}
229	ctx->absorb_offset = absorb_offset;
230}
231EXPORT_SYMBOL_GPL(__sha3_update);
232
233void sha3_final(struct sha3_ctx *sha3_ctx, u8 *out)
234{
235	struct __sha3_ctx *ctx = &sha3_ctx->ctx;
236
237	ctx->state.bytes[ctx->absorb_offset] ^= 0x06;
238	ctx->state.bytes[ctx->block_size - 1] ^= 0x80;
239	sha3_keccakf(&ctx->state);
240	memcpy(out, ctx->state.bytes, ctx->digest_size);
241	sha3_zeroize_ctx(sha3_ctx);
242}
243EXPORT_SYMBOL_GPL(sha3_final);
244
245void shake_squeeze(struct shake_ctx *shake_ctx, u8 *out, size_t out_len)
246{
247	struct __sha3_ctx *ctx = &shake_ctx->ctx;
248	const size_t block_size = ctx->block_size;
249	size_t squeeze_offset = ctx->squeeze_offset;
250
251	if (ctx->absorb_offset < block_size) {
252		/* First squeeze: */
253
254		/* Add the domain separation suffix and padding. */
255		ctx->state.bytes[ctx->absorb_offset] ^= 0x1f;
256		ctx->state.bytes[block_size - 1] ^= 0x80;
257
258		/* Indicate that squeezing has begun. */
259		ctx->absorb_offset = block_size;
260
261		/*
262		 * Indicate that no output is pending yet, i.e. sha3_keccakf()
263		 * will need to be called before the first copy.
264		 */
265		squeeze_offset = block_size;
266	}
267	while (out_len) {
268		if (squeeze_offset == block_size) {
269			sha3_keccakf(&ctx->state);
270			squeeze_offset = 0;
271		}
272		size_t copy = min(out_len, block_size - squeeze_offset);
273
274		memcpy(out, &ctx->state.bytes[squeeze_offset], copy);
275		out += copy;
276		out_len -= copy;
277		squeeze_offset += copy;
278	}
279	ctx->squeeze_offset = squeeze_offset;
280}
281EXPORT_SYMBOL_GPL(shake_squeeze);
282
283#ifndef sha3_224_arch
284static inline bool sha3_224_arch(const u8 *in, size_t in_len,
285				 u8 out[SHA3_224_DIGEST_SIZE])
286{
287	return false;
288}
289#endif
290#ifndef sha3_256_arch
291static inline bool sha3_256_arch(const u8 *in, size_t in_len,
292				 u8 out[SHA3_256_DIGEST_SIZE])
293{
294	return false;
295}
296#endif
297#ifndef sha3_384_arch
298static inline bool sha3_384_arch(const u8 *in, size_t in_len,
299				 u8 out[SHA3_384_DIGEST_SIZE])
300{
301	return false;
302}
303#endif
304#ifndef sha3_512_arch
305static inline bool sha3_512_arch(const u8 *in, size_t in_len,
306				 u8 out[SHA3_512_DIGEST_SIZE])
307{
308	return false;
309}
310#endif
311
312void sha3_224(const u8 *in, size_t in_len, u8 out[SHA3_224_DIGEST_SIZE])
313{
314	struct sha3_ctx ctx;
315
316	if (sha3_224_arch(in, in_len, out))
317		return;
318	sha3_224_init(&ctx);
319	sha3_update(&ctx, in, in_len);
320	sha3_final(&ctx, out);
321}
322EXPORT_SYMBOL_GPL(sha3_224);
323
324void sha3_256(const u8 *in, size_t in_len, u8 out[SHA3_256_DIGEST_SIZE])
325{
326	struct sha3_ctx ctx;
327
328	if (sha3_256_arch(in, in_len, out))
329		return;
330	sha3_256_init(&ctx);
331	sha3_update(&ctx, in, in_len);
332	sha3_final(&ctx, out);
333}
334EXPORT_SYMBOL_GPL(sha3_256);
335
336void sha3_384(const u8 *in, size_t in_len, u8 out[SHA3_384_DIGEST_SIZE])
337{
338	struct sha3_ctx ctx;
339
340	if (sha3_384_arch(in, in_len, out))
341		return;
342	sha3_384_init(&ctx);
343	sha3_update(&ctx, in, in_len);
344	sha3_final(&ctx, out);
345}
346EXPORT_SYMBOL_GPL(sha3_384);
347
348void sha3_512(const u8 *in, size_t in_len, u8 out[SHA3_512_DIGEST_SIZE])
349{
350	struct sha3_ctx ctx;
351
352	if (sha3_512_arch(in, in_len, out))
353		return;
354	sha3_512_init(&ctx);
355	sha3_update(&ctx, in, in_len);
356	sha3_final(&ctx, out);
357}
358EXPORT_SYMBOL_GPL(sha3_512);
359
360void shake128(const u8 *in, size_t in_len, u8 *out, size_t out_len)
361{
362	struct shake_ctx ctx;
363
364	shake128_init(&ctx);
365	shake_update(&ctx, in, in_len);
366	shake_squeeze(&ctx, out, out_len);
367	shake_zeroize_ctx(&ctx);
368}
369EXPORT_SYMBOL_GPL(shake128);
370
371void shake256(const u8 *in, size_t in_len, u8 *out, size_t out_len)
372{
373	struct shake_ctx ctx;
374
375	shake256_init(&ctx);
376	shake_update(&ctx, in, in_len);
377	shake_squeeze(&ctx, out, out_len);
378	shake_zeroize_ctx(&ctx);
379}
380EXPORT_SYMBOL_GPL(shake256);
381
382#if defined(sha3_mod_init_arch) || defined(CONFIG_CRYPTO_FIPS)
383static int __init sha3_mod_init(void)
384{
385#ifdef sha3_mod_init_arch
386	sha3_mod_init_arch();
387#endif
388	if (fips_enabled) {
389		/*
390		 * FIPS cryptographic algorithm self-test.  As per the FIPS
391		 * Implementation Guidance, testing any SHA-3 algorithm
392		 * satisfies the test requirement for all of them.
393		 */
394		u8 hash[SHA3_256_DIGEST_SIZE];
395
396		sha3_256(fips_test_data, sizeof(fips_test_data), hash);
397		if (memcmp(fips_test_sha3_256_value, hash, sizeof(hash)) != 0)
398			panic("sha3: FIPS self-test failed\n");
399	}
400	return 0;
401}
402subsys_initcall(sha3_mod_init);
403
404static void __exit sha3_mod_exit(void)
405{
406}
407module_exit(sha3_mod_exit);
408#endif
409
410MODULE_DESCRIPTION("SHA-3 library functions");
411MODULE_LICENSE("GPL");