1 /*-
2  * Copyright 2009 Colin Percival
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * This file was originally written by Colin Percival as part of the Tarsnap
27  * online backup system.
28  */
29 #include "scrypt_platform.h"
30 
31 #include <sys/types.h>
32 #include <sys/mman.h>
33 
34 #include <emmintrin.h>
35 #include <errno.h>
36 #include <stdint.h>
37 #include <stdlib.h>
38 #include <string.h>
39 
40 #ifdef USE_OPENSSL_PBKDF2
41 #include <openssl/evp.h>
42 #else
43 #include "sha256.h"
44 #endif
45 #include "sysendian.h"
46 
47 #include "crypto_scrypt.h"
48 
49 static void blkcpy(void *, void *, size_t);
50 static void blkxor(void *, void *, size_t);
51 static void salsa20_8(__m128i *);
52 static void blockmix_salsa8(__m128i *, __m128i *, __m128i *, size_t);
53 static uint64_t integerify(void *, size_t);
54 static void smix(uint8_t *, size_t, uint64_t, void *, void *);
55 
56 static void
blkcpy(void * dest,void * src,size_t len)57 blkcpy(void * dest, void * src, size_t len)
58 {
59 	__m128i * D = dest;
60 	__m128i * S = src;
61 	size_t L = len / 16;
62 	size_t i;
63 
64 	for (i = 0; i < L; i++)
65 		D[i] = S[i];
66 }
67 
68 static void
blkxor(void * dest,void * src,size_t len)69 blkxor(void * dest, void * src, size_t len)
70 {
71 	__m128i * D = dest;
72 	__m128i * S = src;
73 	size_t L = len / 16;
74 	size_t i;
75 
76 	for (i = 0; i < L; i++)
77 		D[i] = _mm_xor_si128(D[i], S[i]);
78 }
79 
80 /**
81  * salsa20_8(B):
82  * Apply the salsa20/8 core to the provided block.
83  */
84 static void
salsa20_8(__m128i B[4])85 salsa20_8(__m128i B[4])
86 {
87 	__m128i X0, X1, X2, X3;
88 	__m128i T;
89 	size_t i;
90 
91 	X0 = B[0];
92 	X1 = B[1];
93 	X2 = B[2];
94 	X3 = B[3];
95 
96 	for (i = 0; i < 8; i += 2) {
97 		/* Operate on "columns". */
98 		T = _mm_add_epi32(X0, X3);
99 		X1 = _mm_xor_si128(X1, _mm_slli_epi32(T, 7));
100 		X1 = _mm_xor_si128(X1, _mm_srli_epi32(T, 25));
101 		T = _mm_add_epi32(X1, X0);
102 		X2 = _mm_xor_si128(X2, _mm_slli_epi32(T, 9));
103 		X2 = _mm_xor_si128(X2, _mm_srli_epi32(T, 23));
104 		T = _mm_add_epi32(X2, X1);
105 		X3 = _mm_xor_si128(X3, _mm_slli_epi32(T, 13));
106 		X3 = _mm_xor_si128(X3, _mm_srli_epi32(T, 19));
107 		T = _mm_add_epi32(X3, X2);
108 		X0 = _mm_xor_si128(X0, _mm_slli_epi32(T, 18));
109 		X0 = _mm_xor_si128(X0, _mm_srli_epi32(T, 14));
110 
111 		/* Rearrange data. */
112 		X1 = _mm_shuffle_epi32(X1, 0x93);
113 		X2 = _mm_shuffle_epi32(X2, 0x4E);
114 		X3 = _mm_shuffle_epi32(X3, 0x39);
115 
116 		/* Operate on "rows". */
117 		T = _mm_add_epi32(X0, X1);
118 		X3 = _mm_xor_si128(X3, _mm_slli_epi32(T, 7));
119 		X3 = _mm_xor_si128(X3, _mm_srli_epi32(T, 25));
120 		T = _mm_add_epi32(X3, X0);
121 		X2 = _mm_xor_si128(X2, _mm_slli_epi32(T, 9));
122 		X2 = _mm_xor_si128(X2, _mm_srli_epi32(T, 23));
123 		T = _mm_add_epi32(X2, X3);
124 		X1 = _mm_xor_si128(X1, _mm_slli_epi32(T, 13));
125 		X1 = _mm_xor_si128(X1, _mm_srli_epi32(T, 19));
126 		T = _mm_add_epi32(X1, X2);
127 		X0 = _mm_xor_si128(X0, _mm_slli_epi32(T, 18));
128 		X0 = _mm_xor_si128(X0, _mm_srli_epi32(T, 14));
129 
130 		/* Rearrange data. */
131 		X1 = _mm_shuffle_epi32(X1, 0x39);
132 		X2 = _mm_shuffle_epi32(X2, 0x4E);
133 		X3 = _mm_shuffle_epi32(X3, 0x93);
134 	}
135 
136 	B[0] = _mm_add_epi32(B[0], X0);
137 	B[1] = _mm_add_epi32(B[1], X1);
138 	B[2] = _mm_add_epi32(B[2], X2);
139 	B[3] = _mm_add_epi32(B[3], X3);
140 }
141 
142 /**
143  * blockmix_salsa8(Bin, Bout, X, r):
144  * Compute Bout = BlockMix_{salsa20/8, r}(Bin).  The input Bin must be 128r
145  * bytes in length; the output Bout must also be the same size.  The
146  * temporary space X must be 64 bytes.
147  */
148 static void
blockmix_salsa8(__m128i * Bin,__m128i * Bout,__m128i * X,size_t r)149 blockmix_salsa8(__m128i * Bin, __m128i * Bout, __m128i * X, size_t r)
150 {
151 	size_t i;
152 
153 	/* 1: X <-- B_{2r - 1} */
154 	blkcpy(X, &Bin[8 * r - 4], 64);
155 
156 	/* 2: for i = 0 to 2r - 1 do */
157 	for (i = 0; i < r; i++) {
158 		/* 3: X <-- H(X \xor B_i) */
159 		blkxor(X, &Bin[i * 8], 64);
160 		salsa20_8(X);
161 
162 		/* 4: Y_i <-- X */
163 		/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
164 		blkcpy(&Bout[i * 4], X, 64);
165 
166 		/* 3: X <-- H(X \xor B_i) */
167 		blkxor(X, &Bin[i * 8 + 4], 64);
168 		salsa20_8(X);
169 
170 		/* 4: Y_i <-- X */
171 		/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
172 		blkcpy(&Bout[(r + i) * 4], X, 64);
173 	}
174 }
175 
176 /**
177  * integerify(B, r):
178  * Return the result of parsing B_{2r-1} as a little-endian integer.
179  */
180 static uint64_t
integerify(void * B,size_t r)181 integerify(void * B, size_t r)
182 {
183 	uint32_t * X = (void *)((uintptr_t)(B) + (2 * r - 1) * 64);
184 
185 	return (((uint64_t)(X[13]) << 32) + X[0]);
186 }
187 
188 /**
189  * smix(B, r, N, V, XY):
190  * Compute B = SMix_r(B, N).  The input B must be 128r bytes in length;
191  * the temporary storage V must be 128rN bytes in length; the temporary
192  * storage XY must be 256r + 64 bytes in length.  The value N must be a
193  * power of 2 greater than 1.  The arrays B, V, and XY must be aligned to a
194  * multiple of 64 bytes.
195  */
196 static void
smix(uint8_t * B,size_t r,uint64_t N,void * V,void * XY)197 smix(uint8_t * B, size_t r, uint64_t N, void * V, void * XY)
198 {
199 	__m128i * X = XY;
200 	__m128i * Y = (void *)((uintptr_t)(XY) + 128 * r);
201 	__m128i * Z = (void *)((uintptr_t)(XY) + 256 * r);
202 	uint32_t * X32 = (void *)X;
203 	uint64_t i, j;
204 	size_t k;
205 
206 	/* 1: X <-- B */
207 	for (k = 0; k < 2 * r; k++) {
208 		for (i = 0; i < 16; i++) {
209 			X32[k * 16 + i] =
210 			    le32dec(&B[(k * 16 + (i * 5 % 16)) * 4]);
211 		}
212 	}
213 
214 	/* 2: for i = 0 to N - 1 do */
215 	for (i = 0; i < N; i += 2) {
216 		/* 3: V_i <-- X */
217 		blkcpy((void *)((uintptr_t)(V) + i * 128 * r), X, 128 * r);
218 
219 		/* 4: X <-- H(X) */
220 		blockmix_salsa8(X, Y, Z, r);
221 
222 		/* 3: V_i <-- X */
223 		blkcpy((void *)((uintptr_t)(V) + (i + 1) * 128 * r),
224 		    Y, 128 * r);
225 
226 		/* 4: X <-- H(X) */
227 		blockmix_salsa8(Y, X, Z, r);
228 	}
229 
230 	/* 6: for i = 0 to N - 1 do */
231 	for (i = 0; i < N; i += 2) {
232 		/* 7: j <-- Integerify(X) mod N */
233 		j = integerify(X, r) & (N - 1);
234 
235 		/* 8: X <-- H(X \xor V_j) */
236 		blkxor(X, (void *)((uintptr_t)(V) + j * 128 * r), 128 * r);
237 		blockmix_salsa8(X, Y, Z, r);
238 
239 		/* 7: j <-- Integerify(X) mod N */
240 		j = integerify(Y, r) & (N - 1);
241 
242 		/* 8: X <-- H(X \xor V_j) */
243 		blkxor(Y, (void *)((uintptr_t)(V) + j * 128 * r), 128 * r);
244 		blockmix_salsa8(Y, X, Z, r);
245 	}
246 
247 	/* 10: B' <-- X */
248 	for (k = 0; k < 2 * r; k++) {
249 		for (i = 0; i < 16; i++) {
250 			le32enc(&B[(k * 16 + (i * 5 % 16)) * 4],
251 			    X32[k * 16 + i]);
252 		}
253 	}
254 }
255 
256 /**
257  * crypto_scrypt(passwd, passwdlen, salt, saltlen, N, r, p, buf, buflen):
258  * Compute scrypt(passwd[0 .. passwdlen - 1], salt[0 .. saltlen - 1], N, r,
259  * p, buflen) and write the result into buf.  The parameters r, p, and buflen
260  * must satisfy r * p < 2^30 and buflen <= (2^32 - 1) * 32.  The parameter N
261  * must be a power of 2 greater than 1.
262  *
263  * Return 0 on success; or -1 on error.
264  */
265 int
crypto_scrypt(const uint8_t * passwd,size_t passwdlen,const uint8_t * salt,size_t saltlen,uint64_t N,uint32_t r,uint32_t p,uint8_t * buf,size_t buflen)266 crypto_scrypt(const uint8_t * passwd, size_t passwdlen,
267     const uint8_t * salt, size_t saltlen, uint64_t N, uint32_t r, uint32_t p,
268     uint8_t * buf, size_t buflen)
269 {
270 	void * B0, * V0, * XY0;
271 	uint8_t * B;
272 	uint32_t * V;
273 	uint32_t * XY;
274 	uint32_t i;
275 
276 	/* Sanity-check parameters. */
277 #if SIZE_MAX > UINT32_MAX
278 	if (buflen > (((uint64_t)(1) << 32) - 1) * 32) {
279 		errno = EFBIG;
280 		goto err0;
281 	}
282 #endif
283 	if ((uint64_t)(r) * (uint64_t)(p) >= (1 << 30)) {
284 		errno = EFBIG;
285 		goto err0;
286 	}
287 	if (((N & (N - 1)) != 0) || (N == 0)) {
288 		errno = EINVAL;
289 		goto err0;
290 	}
291 	if ((r > SIZE_MAX / 128 / p) ||
292 #if SIZE_MAX / 256 <= UINT32_MAX
293 	    (r > (SIZE_MAX - 64) / 256) ||
294 #endif
295 	    (N > SIZE_MAX / 128 / r)) {
296 		errno = ENOMEM;
297 		goto err0;
298 	}
299 
300 	/* Allocate memory. */
301 #ifdef HAVE_POSIX_MEMALIGN
302 	if ((errno = posix_memalign(&B0, 64, 128 * r * p)) != 0)
303 		goto err0;
304 	B = (uint8_t *)(B0);
305 	if ((errno = posix_memalign(&XY0, 64, 256 * r + 64)) != 0)
306 		goto err1;
307 	XY = (uint32_t *)(XY0);
308 #ifndef MAP_ANON
309 	if ((errno = posix_memalign(&V0, 64, 128 * r * N)) != 0)
310 		goto err2;
311 	V = (uint32_t *)(V0);
312 #endif
313 #else
314 	if ((B0 = malloc(128 * r * p + 63)) == NULL)
315 		goto err0;
316 	B = (uint8_t *)(((uintptr_t)(B0) + 63) & ~ (uintptr_t)(63));
317 	if ((XY0 = malloc(256 * r + 64 + 63)) == NULL)
318 		goto err1;
319 	XY = (uint32_t *)(((uintptr_t)(XY0) + 63) & ~ (uintptr_t)(63));
320 #ifndef MAP_ANON
321 	if ((V0 = malloc(128 * r * N + 63)) == NULL)
322 		goto err2;
323 	V = (uint32_t *)(((uintptr_t)(V0) + 63) & ~ (uintptr_t)(63));
324 #endif
325 #endif
326 #ifdef MAP_ANON
327 	if ((V0 = mmap(NULL, 128 * r * N, PROT_READ | PROT_WRITE,
328 #ifdef MAP_NOCORE
329 	    MAP_ANON | MAP_PRIVATE | MAP_NOCORE,
330 #else
331 	    MAP_ANON | MAP_PRIVATE,
332 #endif
333 	    -1, 0)) == MAP_FAILED)
334 		goto err2;
335 	V = (uint32_t *)(V0);
336 #endif
337 
338 	/* 1: (B_0 ... B_{p-1}) <-- PBKDF2(P, S, 1, p * MFLen) */
339 #ifdef USE_OPENSSL_PBKDF2
340 	PKCS5_PBKDF2_HMAC((const char *)passwd, passwdlen, salt, saltlen, 1, EVP_sha256(), p * 128 * r, B);
341 #else
342 	PBKDF2_SHA256(passwd, passwdlen, salt, saltlen, 1, B, p * 128 * r);
343 #endif
344 
345 	/* 2: for i = 0 to p - 1 do */
346 	for (i = 0; i < p; i++) {
347 		/* 3: B_i <-- MF(B_i, N) */
348 		smix(&B[i * 128 * r], r, N, V, XY);
349 	}
350 
351 	/* 5: DK <-- PBKDF2(P, B, 1, dkLen) */
352 #ifdef USE_OPENSSL_PBKDF2
353 	PKCS5_PBKDF2_HMAC((const char *)passwd, passwdlen, B, p * 128 * r, 1, EVP_sha256(), buflen, buf);
354 #else
355 	PBKDF2_SHA256(passwd, passwdlen, B, p * 128 * r, 1, buf, buflen);
356 #endif
357 
358 	/* Free memory. */
359 #ifdef MAP_ANON
360 	if (munmap(V0, 128 * r * N))
361 		goto err2;
362 #else
363 	free(V0);
364 #endif
365 	free(XY0);
366 	free(B0);
367 
368 	/* Success! */
369 	return (0);
370 
371 err2:
372 	free(XY0);
373 err1:
374 	free(B0);
375 err0:
376 	/* Failure! */
377 	return (-1);
378 }
379