Mercurial > trustbridge > nss-cmake-static

comparison patches/nss-chacha20-poly1305.patch @ 0:1e5118fa0cb1

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
This is NSS with a Cmake Buildsyste To compile a static NSS library for Windows we've used the Chromium-NSS fork and added a Cmake buildsystem to compile it statically for Windows. See README.chromium for chromium changes and README.trustbridge for our modifications.
author Andre Heinecke <andre.heinecke@intevation.de>
date Mon, 28 Jul 2014 10:47:06 +0200
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:1e5118fa0cb1
1 diff -r c3565a90b8c4 lib/freebl/blapi.h
2 --- a/lib/freebl/blapi.h Fri Jan 03 20:59:10 2014 +0100
3 +++ b/lib/freebl/blapi.h Tue Jan 07 12:11:36 2014 -0800
4 @@ -986,6 +986,38 @@
5 unsigned int *outputLen, unsigned int maxOutputLen,
6 const unsigned char *input, unsigned int inputLen);
7
8 +/******************************************/
9 +/*
10 +** ChaCha20+Poly1305 AEAD
11 +*/
12 +
13 +extern SECStatus
14 +ChaCha20Poly1305_InitContext(ChaCha20Poly1305Context *ctx,
15 + const unsigned char *key, unsigned int keyLen,
16 + unsigned int tagLen);
17 +
18 +extern ChaCha20Poly1305Context *
19 +ChaCha20Poly1305_CreateContext(const unsigned char *key, unsigned int keyLen,
20 + unsigned int tagLen);
21 +
22 +extern void
23 +ChaCha20Poly1305_DestroyContext(ChaCha20Poly1305Context *ctx, PRBool freeit);
24 +
25 +extern SECStatus
26 +ChaCha20Poly1305_Seal(const ChaCha20Poly1305Context *ctx,
27 + unsigned char *output, unsigned int *outputLen,
28 + unsigned int maxOutputLen,
29 + const unsigned char *input, unsigned int inputLen,
30 + const unsigned char *nonce, unsigned int nonceLen,
31 + const unsigned char *ad, unsigned int adLen);
32 +
33 +extern SECStatus
34 +ChaCha20Poly1305_Open(const ChaCha20Poly1305Context *ctx,
35 + unsigned char *output, unsigned int *outputLen,
36 + unsigned int maxOutputLen,
37 + const unsigned char *input, unsigned int inputLen,
38 + const unsigned char *nonce, unsigned int nonceLen,
39 + const unsigned char *ad, unsigned int adLen);
40
41 /******************************************/
42 /*
43 diff -r c3565a90b8c4 lib/freebl/blapit.h
44 --- a/lib/freebl/blapit.h Fri Jan 03 20:59:10 2014 +0100
45 +++ b/lib/freebl/blapit.h Tue Jan 07 12:11:36 2014 -0800
46 @@ -222,6 +222,7 @@
47 struct SHA512ContextStr ;
48 struct AESKeyWrapContextStr ;
49 struct SEEDContextStr ;
50 +struct ChaCha20Poly1305ContextStr;
51
52 typedef struct DESContextStr DESContext;
53 typedef struct RC2ContextStr RC2Context;
54 @@ -240,6 +241,7 @@
55 typedef struct SHA512ContextStr SHA384Context;
56 typedef struct AESKeyWrapContextStr AESKeyWrapContext;
57 typedef struct SEEDContextStr SEEDContext;
58 +typedef struct ChaCha20Poly1305ContextStr ChaCha20Poly1305Context;
59
60 /***************************************************************************
61 ** RSA Public and Private Key structures
62 diff -r c3565a90b8c4 lib/freebl/chacha20/chacha20.c
63 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
64 +++ b/lib/freebl/chacha20/chacha20.c Tue Jan 07 12:11:36 2014 -0800
65 @@ -0,0 +1,108 @@
66 +/* This Source Code Form is subject to the terms of the Mozilla Public
67 + * License, v. 2.0. If a copy of the MPL was not distributed with this
68 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
69 +
70 +/* Adopted from the public domain code in NaCl by djb. */
71 +
72 +#include <string.h>
73 +#include <stdio.h>
74 +
75 +#include "prtypes.h"
76 +#include "chacha20.h"
77 +
78 +#define ROTL32(v, n) (((v) << (n)) | ((v) >> (32 - (n))))
79 +#define ROTATE(v, c) ROTL32((v), (c))
80 +#define XOR(v, w) ((v) ^ (w))
81 +#define PLUS(x, y) ((x) + (y))
82 +
83 +#define U32TO8_LITTLE(p, v) \
84 + { (p)[0] = ((v) ) & 0xff; (p)[1] = ((v) >> 8) & 0xff; \
85 + (p)[2] = ((v) >> 16) & 0xff; (p)[3] = ((v) >> 24) & 0xff; }
86 +#define U8TO32_LITTLE(p) \
87 + (((PRUint32)((p)[0]) ) | ((PRUint32)((p)[1]) << 8) | \
88 + ((PRUint32)((p)[2]) << 16) | ((PRUint32)((p)[3]) << 24) )
89 +
90 +#define QUARTERROUND(a,b,c,d) \
91 + x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]),16); \
92 + x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]),12); \
93 + x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]), 8); \
94 + x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]), 7);
95 +
96 +static void ChaChaCore(unsigned char output[64], const PRUint32 input[16],
97 + int num_rounds) {
98 + PRUint32 x[16];
99 + int i;
100 +
101 + memcpy(x, input, sizeof(PRUint32) * 16);
102 + for (i = num_rounds; i > 0; i -= 2) {
103 + QUARTERROUND( 0, 4, 8,12)
104 + QUARTERROUND( 1, 5, 9,13)
105 + QUARTERROUND( 2, 6,10,14)
106 + QUARTERROUND( 3, 7,11,15)
107 + QUARTERROUND( 0, 5,10,15)
108 + QUARTERROUND( 1, 6,11,12)
109 + QUARTERROUND( 2, 7, 8,13)
110 + QUARTERROUND( 3, 4, 9,14)
111 + }
112 +
113 + for (i = 0; i < 16; ++i) {
114 + x[i] = PLUS(x[i], input[i]);
115 + }
116 + for (i = 0; i < 16; ++i) {
117 + U32TO8_LITTLE(output + 4 * i, x[i]);
118 + }
119 +}
120 +
121 +static const unsigned char sigma[16] = "expand 32-byte k";
122 +
123 +void ChaCha20XOR(unsigned char *out, const unsigned char *in, unsigned int inLen,
124 + const unsigned char key[32], const unsigned char nonce[8],
125 + uint64_t counter) {
126 + unsigned char block[64];
127 + PRUint32 input[16];
128 + unsigned int u;
129 + unsigned int i;
130 +
131 + input[4] = U8TO32_LITTLE(key + 0);
132 + input[5] = U8TO32_LITTLE(key + 4);
133 + input[6] = U8TO32_LITTLE(key + 8);
134 + input[7] = U8TO32_LITTLE(key + 12);
135 +
136 + input[8] = U8TO32_LITTLE(key + 16);
137 + input[9] = U8TO32_LITTLE(key + 20);
138 + input[10] = U8TO32_LITTLE(key + 24);
139 + input[11] = U8TO32_LITTLE(key + 28);
140 +
141 + input[0] = U8TO32_LITTLE(sigma + 0);
142 + input[1] = U8TO32_LITTLE(sigma + 4);
143 + input[2] = U8TO32_LITTLE(sigma + 8);
144 + input[3] = U8TO32_LITTLE(sigma + 12);
145 +
146 + input[12] = counter;
147 + input[13] = counter >> 32;
148 + input[14] = U8TO32_LITTLE(nonce + 0);
149 + input[15] = U8TO32_LITTLE(nonce + 4);
150 +
151 + while (inLen >= 64) {
152 + ChaChaCore(block, input, 20);
153 + for (i = 0; i < 64; i++) {
154 + out[i] = in[i] ^ block[i];
155 + }
156 +
157 + input[12]++;
158 + if (input[12] == 0) {
159 + input[13]++;
160 + }
161 +
162 + inLen -= 64;
163 + in += 64;
164 + out += 64;
165 + }
166 +
167 + if (inLen > 0) {
168 + ChaChaCore(block, input, 20);
169 + for (i = 0; i < inLen; i++) {
170 + out[i] = in[i] ^ block[i];
171 + }
172 + }
173 +}
174 diff -r c3565a90b8c4 lib/freebl/chacha20/chacha20.h
175 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
176 +++ b/lib/freebl/chacha20/chacha20.h Tue Jan 07 12:11:36 2014 -0800
177 @@ -0,0 +1,22 @@
178 +/*
179 + * chacha20.h - header file for ChaCha20 implementation.
180 + *
181 + * This Source Code Form is subject to the terms of the Mozilla Public
182 + * License, v. 2.0. If a copy of the MPL was not distributed with this
183 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
184 +
185 +#ifndef FREEBL_CHACHA20_H_
186 +#define FREEBL_CHACHA20_H_
187 +
188 +#include <stdint.h>
189 +
190 +/* ChaCha20XOR encrypts |inLen| bytes from |in| with the given key and
191 + * nonce and writes the result to |out|, which may be equal to |in|. The
192 + * initial block counter is specified by |counter|. */
193 +extern void ChaCha20XOR(unsigned char *out,
194 + const unsigned char *in, unsigned int inLen,
195 + const unsigned char key[32],
196 + const unsigned char nonce[8],
197 + uint64_t counter);
198 +
199 +#endif /* FREEBL_CHACHA20_H_ */
200 diff -r c3565a90b8c4 lib/freebl/chacha20/chacha20_vec.c
201 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
202 +++ b/lib/freebl/chacha20/chacha20_vec.c Tue Jan 07 12:11:36 2014 -0800
203 @@ -0,0 +1,281 @@
204 +/* This Source Code Form is subject to the terms of the Mozilla Public
205 + * License, v. 2.0. If a copy of the MPL was not distributed with this
206 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
207 +
208 +/* This implementation is by Ted Krovetz and was submitted to SUPERCOP and
209 + * marked as public domain. It was been altered to allow for non-aligned inputs
210 + * and to allow the block counter to be passed in specifically. */
211 +
212 +#include <string.h>
213 +
214 +#include "chacha20.h"
215 +
216 +#ifndef CHACHA_RNDS
217 +#define CHACHA_RNDS 20 /* 8 (high speed), 20 (conservative), 12 (middle) */
218 +#endif
219 +
220 +/* Architecture-neutral way to specify 16-byte vector of ints */
221 +typedef unsigned vec __attribute__ ((vector_size (16)));
222 +
223 +/* This implementation is designed for Neon, SSE and AltiVec machines. The
224 + * following specify how to do certain vector operations efficiently on
225 + * each architecture, using intrinsics.
226 + * This implementation supports parallel processing of multiple blocks,
227 + * including potentially using general-purpose registers.
228 + */
229 +#if __ARM_NEON__
230 +#include <arm_neon.h>
231 +#define GPR_TOO 1
232 +#define VBPI 2
233 +#define ONE (vec)vsetq_lane_u32(1,vdupq_n_u32(0),0)
234 +#define LOAD(m) (vec)(*((vec*)(m)))
235 +#define STORE(m,r) (*((vec*)(m))) = (r)
236 +#define ROTV1(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,1)
237 +#define ROTV2(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,2)
238 +#define ROTV3(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,3)
239 +#define ROTW16(x) (vec)vrev32q_u16((uint16x8_t)x)
240 +#if __clang__
241 +#define ROTW7(x) (x << ((vec){ 7, 7, 7, 7})) ^ (x >> ((vec){25,25,25,25}))
242 +#define ROTW8(x) (x << ((vec){ 8, 8, 8, 8})) ^ (x >> ((vec){24,24,24,24}))
243 +#define ROTW12(x) (x << ((vec){12,12,12,12})) ^ (x >> ((vec){20,20,20,20}))
244 +#else
245 +#define ROTW7(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,7),(uint32x4_t)x,25)
246 +#define ROTW8(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,8),(uint32x4_t)x,24)
247 +#define ROTW12(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,12),(uint32x4_t)x,20)
248 +#endif
249 +#elif __SSE2__
250 +#include <emmintrin.h>
251 +#define GPR_TOO 0
252 +#if __clang__
253 +#define VBPI 4
254 +#else
255 +#define VBPI 3
256 +#endif
257 +#define ONE (vec)_mm_set_epi32(0,0,0,1)
258 +#define LOAD(m) (vec)_mm_loadu_si128((__m128i*)(m))
259 +#define STORE(m,r) _mm_storeu_si128((__m128i*)(m), (__m128i) (r))
260 +#define ROTV1(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(0,3,2,1))
261 +#define ROTV2(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(1,0,3,2))
262 +#define ROTV3(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(2,1,0,3))
263 +#define ROTW7(x) (vec)(_mm_slli_epi32((__m128i)x, 7) ^ _mm_srli_epi32((__m128i)x,25))
264 +#define ROTW12(x) (vec)(_mm_slli_epi32((__m128i)x,12) ^ _mm_srli_epi32((__m128i)x,20))
265 +#if __SSSE3__
266 +#include <tmmintrin.h>
267 +#define ROTW8(x) (vec)_mm_shuffle_epi8((__m128i)x,_mm_set_epi8(14,13,12,15,10,9,8,11,6,5,4,7,2,1,0,3))
268 +#define ROTW16(x) (vec)_mm_shuffle_epi8((__m128i)x,_mm_set_epi8(13,12,15,14,9,8,11,10,5,4,7,6,1,0,3,2))
269 +#else
270 +#define ROTW8(x) (vec)(_mm_slli_epi32((__m128i)x, 8) ^ _mm_srli_epi32((__m128i)x,24))
271 +#define ROTW16(x) (vec)(_mm_slli_epi32((__m128i)x,16) ^ _mm_srli_epi32((__m128i)x,16))
272 +#endif
273 +#else
274 +#error -- Implementation supports only machines with neon or SSE2
275 +#endif
276 +
277 +#ifndef REVV_BE
278 +#define REVV_BE(x) (x)
279 +#endif
280 +
281 +#ifndef REVW_BE
282 +#define REVW_BE(x) (x)
283 +#endif
284 +
285 +#define BPI (VBPI + GPR_TOO) /* Blocks computed per loop iteration */
286 +
287 +#define DQROUND_VECTORS(a,b,c,d) \
288 + a += b; d ^= a; d = ROTW16(d); \
289 + c += d; b ^= c; b = ROTW12(b); \
290 + a += b; d ^= a; d = ROTW8(d); \
291 + c += d; b ^= c; b = ROTW7(b); \
292 + b = ROTV1(b); c = ROTV2(c); d = ROTV3(d); \
293 + a += b; d ^= a; d = ROTW16(d); \
294 + c += d; b ^= c; b = ROTW12(b); \
295 + a += b; d ^= a; d = ROTW8(d); \
296 + c += d; b ^= c; b = ROTW7(b); \
297 + b = ROTV3(b); c = ROTV2(c); d = ROTV1(d);
298 +
299 +#define QROUND_WORDS(a,b,c,d) \
300 + a = a+b; d ^= a; d = d<<16 | d>>16; \
301 + c = c+d; b ^= c; b = b<<12 | b>>20; \
302 + a = a+b; d ^= a; d = d<< 8 | d>>24; \
303 + c = c+d; b ^= c; b = b<< 7 | b>>25;
304 +
305 +#define WRITE_XOR(in, op, d, v0, v1, v2, v3) \
306 + STORE(op + d + 0, LOAD(in + d + 0) ^ REVV_BE(v0)); \
307 + STORE(op + d + 4, LOAD(in + d + 4) ^ REVV_BE(v1)); \
308 + STORE(op + d + 8, LOAD(in + d + 8) ^ REVV_BE(v2)); \
309 + STORE(op + d +12, LOAD(in + d +12) ^ REVV_BE(v3));
310 +
311 +void ChaCha20XOR(
312 + unsigned char *out,
313 + const unsigned char *in,
314 + unsigned int inlen,
315 + const unsigned char key[32],
316 + const unsigned char nonce[8],
317 + uint64_t counter)
318 +{
319 + unsigned iters, i, *op=(unsigned *)out, *ip=(unsigned *)in, *kp;
320 +#if defined(__ARM_NEON__)
321 + unsigned *np;
322 +#endif
323 + vec s0, s1, s2, s3;
324 +#if !defined(__ARM_NEON__) && !defined(__SSE2__)
325 + __attribute__ ((aligned (16))) unsigned key[8], nonce[4];
326 +#endif
327 + __attribute__ ((aligned (16))) unsigned chacha_const[] =
328 + {0x61707865,0x3320646E,0x79622D32,0x6B206574};
329 +#if defined(__ARM_NEON__) || defined(__SSE2__)
330 + kp = (unsigned *)key;
331 +#else
332 + ((vec *)key)[0] = REVV_BE(((vec *)key)[0]);
333 + ((vec *)key)[1] = REVV_BE(((vec *)key)[1]);
334 + nonce[0] = REVW_BE(((unsigned *)nonce)[0]);
335 + nonce[1] = REVW_BE(((unsigned *)nonce)[1]);
336 + nonce[2] = REVW_BE(((unsigned *)nonce)[2]);
337 + nonce[3] = REVW_BE(((unsigned *)nonce)[3]);
338 + kp = (unsigned *)key;
339 + np = (unsigned *)nonce;
340 +#endif
341 +#if defined(__ARM_NEON__)
342 + np = (unsigned*) nonce;
343 +#endif
344 + s0 = LOAD(chacha_const);
345 + s1 = LOAD(&((vec*)kp)[0]);
346 + s2 = LOAD(&((vec*)kp)[1]);
347 + s3 = (vec) {
348 + counter & 0xffffffff,
349 + counter >> 32,
350 + ((uint32_t*)nonce)[0],
351 + ((uint32_t*)nonce)[1]
352 + };
353 +
354 + for (iters = 0; iters < inlen/(BPI*64); iters++) {
355 +#if GPR_TOO
356 + register unsigned x0, x1, x2, x3, x4, x5, x6, x7, x8,
357 + x9, x10, x11, x12, x13, x14, x15;
358 +#endif
359 +#if VBPI > 2
360 + vec v8,v9,v10,v11;
361 +#endif
362 +#if VBPI > 3
363 + vec v12,v13,v14,v15;
364 +#endif
365 +
366 + vec v0,v1,v2,v3,v4,v5,v6,v7;
367 + v4 = v0 = s0; v5 = v1 = s1; v6 = v2 = s2; v3 = s3;
368 + v7 = v3 + ONE;
369 +#if VBPI > 2
370 + v8 = v4; v9 = v5; v10 = v6;
371 + v11 = v7 + ONE;
372 +#endif
373 +#if VBPI > 3
374 + v12 = v8; v13 = v9; v14 = v10;
375 + v15 = v11 + ONE;
376 +#endif
377 +#if GPR_TOO
378 + x0 = chacha_const[0]; x1 = chacha_const[1];
379 + x2 = chacha_const[2]; x3 = chacha_const[3];
380 + x4 = kp[0]; x5 = kp[1]; x6 = kp[2]; x7 = kp[3];
381 + x8 = kp[4]; x9 = kp[5]; x10 = kp[6]; x11 = kp[7];
382 + x12 = (counter & 0xffffffff)+BPI*iters+(BPI-1); x13 = counter >> 32;
383 + x14 = np[0]; x15 = np[1];
384 +#endif
385 + for (i = CHACHA_RNDS/2; i; i--) {
386 + DQROUND_VECTORS(v0,v1,v2,v3)
387 + DQROUND_VECTORS(v4,v5,v6,v7)
388 +#if VBPI > 2
389 + DQROUND_VECTORS(v8,v9,v10,v11)
390 +#endif
391 +#if VBPI > 3
392 + DQROUND_VECTORS(v12,v13,v14,v15)
393 +#endif
394 +#if GPR_TOO
395 + QROUND_WORDS( x0, x4, x8,x12)
396 + QROUND_WORDS( x1, x5, x9,x13)
397 + QROUND_WORDS( x2, x6,x10,x14)
398 + QROUND_WORDS( x3, x7,x11,x15)
399 + QROUND_WORDS( x0, x5,x10,x15)
400 + QROUND_WORDS( x1, x6,x11,x12)
401 + QROUND_WORDS( x2, x7, x8,x13)
402 + QROUND_WORDS( x3, x4, x9,x14)
403 +#endif
404 + }
405 +
406 + WRITE_XOR(ip, op, 0, v0+s0, v1+s1, v2+s2, v3+s3)
407 + s3 += ONE;
408 + WRITE_XOR(ip, op, 16, v4+s0, v5+s1, v6+s2, v7+s3)
409 + s3 += ONE;
410 +#if VBPI > 2
411 + WRITE_XOR(ip, op, 32, v8+s0, v9+s1, v10+s2, v11+s3)
412 + s3 += ONE;
413 +#endif
414 +#if VBPI > 3
415 + WRITE_XOR(ip, op, 48, v12+s0, v13+s1, v14+s2, v15+s3)
416 + s3 += ONE;
417 +#endif
418 + ip += VBPI*16;
419 + op += VBPI*16;
420 +#if GPR_TOO
421 + op[0] = REVW_BE(REVW_BE(ip[0]) ^ (x0 + chacha_const[0]));
422 + op[1] = REVW_BE(REVW_BE(ip[1]) ^ (x1 + chacha_const[1]));
423 + op[2] = REVW_BE(REVW_BE(ip[2]) ^ (x2 + chacha_const[2]));
424 + op[3] = REVW_BE(REVW_BE(ip[3]) ^ (x3 + chacha_const[3]));
425 + op[4] = REVW_BE(REVW_BE(ip[4]) ^ (x4 + kp[0]));
426 + op[5] = REVW_BE(REVW_BE(ip[5]) ^ (x5 + kp[1]));
427 + op[6] = REVW_BE(REVW_BE(ip[6]) ^ (x6 + kp[2]));
428 + op[7] = REVW_BE(REVW_BE(ip[7]) ^ (x7 + kp[3]));
429 + op[8] = REVW_BE(REVW_BE(ip[8]) ^ (x8 + kp[4]));
430 + op[9] = REVW_BE(REVW_BE(ip[9]) ^ (x9 + kp[5]));
431 + op[10] = REVW_BE(REVW_BE(ip[10]) ^ (x10 + kp[6]));
432 + op[11] = REVW_BE(REVW_BE(ip[11]) ^ (x11 + kp[7]));
433 + op[12] = REVW_BE(REVW_BE(ip[12]) ^ (x12 + (counter & 0xffffffff)+BPI*iters+(BPI-1)));
434 + op[13] = REVW_BE(REVW_BE(ip[13]) ^ (x13 + (counter >> 32)));
435 + op[14] = REVW_BE(REVW_BE(ip[14]) ^ (x14 + np[0]));
436 + op[15] = REVW_BE(REVW_BE(ip[15]) ^ (x15 + np[1]));
437 + s3 += ONE;
438 + ip += 16;
439 + op += 16;
440 +#endif
441 + }
442 +
443 + for (iters = inlen%(BPI*64)/64; iters != 0; iters--) {
444 + vec v0 = s0, v1 = s1, v2 = s2, v3 = s3;
445 + for (i = CHACHA_RNDS/2; i; i--) {
446 + DQROUND_VECTORS(v0,v1,v2,v3);
447 + }
448 + WRITE_XOR(ip, op, 0, v0+s0, v1+s1, v2+s2, v3+s3)
449 + s3 += ONE;
450 + ip += 16;
451 + op += 16;
452 + }
453 +
454 + inlen = inlen % 64;
455 + if (inlen) {
456 + __attribute__ ((aligned (16))) vec buf[4];
457 + vec v0,v1,v2,v3;
458 + v0 = s0; v1 = s1; v2 = s2; v3 = s3;
459 + for (i = CHACHA_RNDS/2; i; i--) {
460 + DQROUND_VECTORS(v0,v1,v2,v3);
461 + }
462 +
463 + if (inlen >= 16) {
464 + STORE(op + 0, LOAD(ip + 0) ^ REVV_BE(v0 + s0));
465 + if (inlen >= 32) {
466 + STORE(op + 4, LOAD(ip + 4) ^ REVV_BE(v1 + s1));
467 + if (inlen >= 48) {
468 + STORE(op + 8, LOAD(ip + 8) ^ REVV_BE(v2 + s2));
469 + buf[3] = REVV_BE(v3 + s3);
470 + } else {
471 + buf[2] = REVV_BE(v2 + s2);
472 + }
473 + } else {
474 + buf[1] = REVV_BE(v1 + s1);
475 + }
476 + } else {
477 + buf[0] = REVV_BE(v0 + s0);
478 + }
479 +
480 + for (i=inlen & ~15; i<inlen; i++) {
481 + ((char *)op)[i] = ((char *)ip)[i] ^ ((char *)buf)[i];
482 + }
483 + }
484 +}
485 diff -r c3565a90b8c4 lib/freebl/chacha20poly1305.c
486 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
487 +++ b/lib/freebl/chacha20poly1305.c Tue Jan 07 12:11:36 2014 -0800
488 @@ -0,0 +1,169 @@
489 +/* This Source Code Form is subject to the terms of the Mozilla Public
490 + * License, v. 2.0. If a copy of the MPL was not distributed with this
491 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
492 +
493 +#ifdef FREEBL_NO_DEPEND
494 +#include "stubs.h"
495 +#endif
496 +
497 +#include <string.h>
498 +#include <stdio.h>
499 +
500 +#include "seccomon.h"
501 +#include "secerr.h"
502 +#include "blapit.h"
503 +#include "poly1305/poly1305.h"
504 +#include "chacha20/chacha20.h"
505 +#include "chacha20poly1305.h"
506 +
507 +/* Poly1305Do writes the Poly1305 authenticator of the given additional data
508 + * and ciphertext to |out|. */
509 +static void
510 +Poly1305Do(unsigned char *out,
511 + const unsigned char *ad, unsigned int adLen,
512 + const unsigned char *ciphertext, unsigned int ciphertextLen,
513 + const unsigned char key[32])
514 +{
515 + poly1305_state state;
516 + unsigned int j;
517 + unsigned char lengthBytes[8];
518 + unsigned int i;
519 +
520 + Poly1305Init(&state, key);
521 + j = adLen;
522 + for (i = 0; i < sizeof(lengthBytes); i++) {
523 + lengthBytes[i] = j;
524 + j >>= 8;
525 + }
526 + Poly1305Update(&state, ad, adLen);
527 + Poly1305Update(&state, lengthBytes, sizeof(lengthBytes));
528 + j = ciphertextLen;
529 + for (i = 0; i < sizeof(lengthBytes); i++) {
530 + lengthBytes[i] = j;
531 + j >>= 8;
532 + }
533 + Poly1305Update(&state, ciphertext, ciphertextLen);
534 + Poly1305Update(&state, lengthBytes, sizeof(lengthBytes));
535 + Poly1305Finish(&state, out);
536 +}
537 +
538 +SECStatus
539 +ChaCha20Poly1305_InitContext(ChaCha20Poly1305Context *ctx,
540 + const unsigned char *key, unsigned int keyLen,
541 + unsigned int tagLen)
542 +{
543 + if (keyLen != 32) {
544 + PORT_SetError(SEC_ERROR_BAD_KEY);
545 + return SECFailure;
546 + }
547 + if (tagLen == 0 || tagLen > 16) {
548 + PORT_SetError(SEC_ERROR_INPUT_LEN);
549 + return SECFailure;
550 + }
551 +
552 + memcpy(ctx->key, key, sizeof(ctx->key));
553 + ctx->tagLen = tagLen;
554 +
555 + return SECSuccess;
556 +}
557 +
558 +ChaCha20Poly1305Context *
559 +ChaCha20Poly1305_CreateContext(const unsigned char *key, unsigned int keyLen,
560 + unsigned int tagLen)
561 +{
562 + ChaCha20Poly1305Context *ctx;
563 +
564 + ctx = PORT_New(ChaCha20Poly1305Context);
565 + if (ctx == NULL) {
566 + return NULL;
567 + }
568 +
569 + if (ChaCha20Poly1305_InitContext(ctx, key, keyLen, tagLen) != SECSuccess) {
570 + PORT_Free(ctx);
571 + ctx = NULL;
572 + }
573 +
574 + return ctx;
575 +}
576 +
577 +void
578 +ChaCha20Poly1305_DestroyContext(ChaCha20Poly1305Context *ctx, PRBool freeit)
579 +{
580 + memset(ctx, 0, sizeof(*ctx));
581 + if (freeit) {
582 + PORT_Free(ctx);
583 + }
584 +}
585 +
586 +SECStatus
587 +ChaCha20Poly1305_Seal(const ChaCha20Poly1305Context *ctx,
588 + unsigned char *output, unsigned int *outputLen,
589 + unsigned int maxOutputLen,
590 + const unsigned char *input, unsigned int inputLen,
591 + const unsigned char *nonce, unsigned int nonceLen,
592 + const unsigned char *ad, unsigned int adLen)
593 +{
594 + unsigned char block[64];
595 + unsigned char tag[16];
596 +
597 + if (nonceLen != 8) {
598 + PORT_SetError(SEC_ERROR_INPUT_LEN);
599 + return SECFailure;
600 + }
601 + *outputLen = inputLen + ctx->tagLen;
602 + if (maxOutputLen < *outputLen) {
603 + PORT_SetError(SEC_ERROR_OUTPUT_LEN);
604 + return SECFailure;
605 + }
606 +
607 + memset(block, 0, sizeof(block));
608 + // Generate a block of keystream. The first 32 bytes will be the poly1305
609 + // key. The remainder of the block is discarded.
610 + ChaCha20XOR(block, block, sizeof(block), ctx->key, nonce, 0);
611 + ChaCha20XOR(output, input, inputLen, ctx->key, nonce, 1);
612 +
613 + Poly1305Do(tag, ad, adLen, output, inputLen, block);
614 + memcpy(output + inputLen, tag, ctx->tagLen);
615 +
616 + return SECSuccess;
617 +}
618 +
619 +SECStatus
620 +ChaCha20Poly1305_Open(const ChaCha20Poly1305Context *ctx,
621 + unsigned char *output, unsigned int *outputLen,
622 + unsigned int maxOutputLen,
623 + const unsigned char *input, unsigned int inputLen,
624 + const unsigned char *nonce, unsigned int nonceLen,
625 + const unsigned char *ad, unsigned int adLen)
626 +{
627 + unsigned char block[64];
628 + unsigned char tag[16];
629 +
630 + if (nonceLen != 8) {
631 + PORT_SetError(SEC_ERROR_INPUT_LEN);
632 + return SECFailure;
633 + }
634 + if (inputLen < ctx->tagLen) {
635 + PORT_SetError(SEC_ERROR_INPUT_LEN);
636 + return SECFailure;
637 + }
638 + *outputLen = inputLen - ctx->tagLen;
639 + if (maxOutputLen < *outputLen) {
640 + PORT_SetError(SEC_ERROR_OUTPUT_LEN);
641 + return SECFailure;
642 + }
643 +
644 + memset(block, 0, sizeof(block));
645 + // Generate a block of keystream. The first 32 bytes will be the poly1305
646 + // key. The remainder of the block is discarded.
647 + ChaCha20XOR(block, block, sizeof(block), ctx->key, nonce, 0);
648 + Poly1305Do(tag, ad, adLen, input, inputLen - ctx->tagLen, block);
649 + if (NSS_SecureMemcmp(tag, &input[inputLen - ctx->tagLen], ctx->tagLen) != 0) {
650 + PORT_SetError(SEC_ERROR_BAD_DATA);
651 + return SECFailure;
652 + }
653 +
654 + ChaCha20XOR(output, input, inputLen - ctx->tagLen, ctx->key, nonce, 1);
655 +
656 + return SECSuccess;
657 +}
658 diff -r c3565a90b8c4 lib/freebl/chacha20poly1305.h
659 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
660 +++ b/lib/freebl/chacha20poly1305.h Tue Jan 07 12:11:36 2014 -0800
661 @@ -0,0 +1,15 @@
662 +/* This Source Code Form is subject to the terms of the Mozilla Public
663 + * License, v. 2.0. If a copy of the MPL was not distributed with this
664 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
665 +
666 +#ifndef _CHACHA20_POLY1305_H_
667 +#define _CHACHA20_POLY1305_H_ 1
668 +
669 +/* ChaCha20Poly1305ContextStr saves the key and tag length for a
670 + * ChaCha20+Poly1305 AEAD operation. */
671 +struct ChaCha20Poly1305ContextStr {
672 + unsigned char key[32];
673 + unsigned char tagLen;
674 +};
675 +
676 +#endif /* _CHACHA20_POLY1305_H_ */
677 diff -r c3565a90b8c4 lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c
678 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
679 +++ b/lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c Tue Jan 07 12:11:36 2014 -0800
680 @@ -0,0 +1,623 @@
681 +/* This Source Code Form is subject to the terms of the Mozilla Public
682 + * License, v. 2.0. If a copy of the MPL was not distributed with this
683 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
684 +
685 +/* This implementation of poly1305 is by Andrew Moon
686 + * (https://github.com/floodyberry/poly1305-donna) and released as public
687 + * domain. It implements SIMD vectorization based on the algorithm described in
688 + * http://cr.yp.to/papers.html#neoncrypto. Unrolled to 2 powers, i.e. 64 byte
689 + * block size. */
690 +
691 +#include <emmintrin.h>
692 +#include <stdint.h>
693 +
694 +#include "poly1305.h"
695 +
696 +#define ALIGN(x) __attribute__((aligned(x)))
697 +#define INLINE inline
698 +#define U8TO64_LE(m) (*(uint64_t*)(m))
699 +#define U8TO32_LE(m) (*(uint32_t*)(m))
700 +#define U64TO8_LE(m,v) (*(uint64_t*)(m)) = v
701 +
702 +typedef __m128i xmmi;
703 +typedef unsigned __int128 uint128_t;
704 +
705 +static const uint32_t ALIGN(16) poly1305_x64_sse2_message_mask[4] = {(1 << 26) - 1, 0, (1 << 26) - 1, 0};
706 +static const uint32_t ALIGN(16) poly1305_x64_sse2_5[4] = {5, 0, 5, 0};
707 +static const uint32_t ALIGN(16) poly1305_x64_sse2_1shl128[4] = {(1 << 24), 0, (1 << 24), 0};
708 +
709 +static uint128_t INLINE
710 +add128(uint128_t a, uint128_t b) {
711 + return a + b;
712 +}
713 +
714 +static uint128_t INLINE
715 +add128_64(uint128_t a, uint64_t b) {
716 + return a + b;
717 +}
718 +
719 +static uint128_t INLINE
720 +mul64x64_128(uint64_t a, uint64_t b) {
721 + return (uint128_t)a * b;
722 +}
723 +
724 +static uint64_t INLINE
725 +lo128(uint128_t a) {
726 + return (uint64_t)a;
727 +}
728 +
729 +static uint64_t INLINE
730 +shr128(uint128_t v, const int shift) {
731 + return (uint64_t)(v >> shift);
732 +}
733 +
734 +static uint64_t INLINE
735 +shr128_pair(uint64_t hi, uint64_t lo, const int shift) {
736 + return (uint64_t)((((uint128_t)hi << 64) | lo) >> shift);
737 +}
738 +
739 +typedef struct poly1305_power_t {
740 + union {
741 + xmmi v;
742 + uint64_t u[2];
743 + uint32_t d[4];
744 + } R20,R21,R22,R23,R24,S21,S22,S23,S24;
745 +} poly1305_power;
746 +
747 +typedef struct poly1305_state_internal_t {
748 + poly1305_power P[2]; /* 288 bytes, top 32 bit halves unused = 144 bytes of free storage */
749 + union {
750 + xmmi H[5]; /* 80 bytes */
751 + uint64_t HH[10];
752 + };
753 + /* uint64_t r0,r1,r2; [24 bytes] */
754 + /* uint64_t pad0,pad1; [16 bytes] */
755 + uint64_t started; /* 8 bytes */
756 + uint64_t leftover; /* 8 bytes */
757 + uint8_t buffer[64]; /* 64 bytes */
758 +} poly1305_state_internal; /* 448 bytes total + 63 bytes for alignment = 511 bytes raw */
759 +
760 +static poly1305_state_internal INLINE
761 +*poly1305_aligned_state(poly1305_state *state) {
762 + return (poly1305_state_internal *)(((uint64_t)state + 63) & ~63);
763 +}
764 +
765 +/* copy 0-63 bytes */
766 +static void INLINE
767 +poly1305_block_copy(uint8_t *dst, const uint8_t *src, size_t bytes) {
768 + size_t offset = src - dst;
769 + if (bytes & 32) {
770 + _mm_storeu_si128((xmmi *)(dst + 0), _mm_loadu_si128((xmmi *)(dst + offset + 0)));
771 + _mm_storeu_si128((xmmi *)(dst + 16), _mm_loadu_si128((xmmi *)(dst + offset + 16)));
772 + dst += 32;
773 + }
774 + if (bytes & 16) { _mm_storeu_si128((xmmi *)dst, _mm_loadu_si128((xmmi *)(dst + offset))); dst += 16; }
775 + if (bytes & 8) { *(uint64_t *)dst = *(uint64_t *)(dst + offset); dst += 8; }
776 + if (bytes & 4) { *(uint32_t *)dst = *(uint32_t *)(dst + offset); dst += 4; }
777 + if (bytes & 2) { *(uint16_t *)dst = *(uint16_t *)(dst + offset); dst += 2; }
778 + if (bytes & 1) { *( uint8_t *)dst = *( uint8_t *)(dst + offset); }
779 +}
780 +
781 +/* zero 0-15 bytes */
782 +static void INLINE
783 +poly1305_block_zero(uint8_t *dst, size_t bytes) {
784 + if (bytes & 8) { *(uint64_t *)dst = 0; dst += 8; }
785 + if (bytes & 4) { *(uint32_t *)dst = 0; dst += 4; }
786 + if (bytes & 2) { *(uint16_t *)dst = 0; dst += 2; }
787 + if (bytes & 1) { *( uint8_t *)dst = 0; }
788 +}
789 +
790 +static size_t INLINE
791 +poly1305_min(size_t a, size_t b) {
792 + return (a < b) ? a : b;
793 +}
794 +
795 +void
796 +Poly1305Init(poly1305_state *state, const unsigned char key[32]) {
797 + poly1305_state_internal *st = poly1305_aligned_state(state);
798 + poly1305_power *p;
799 + uint64_t r0,r1,r2;
800 + uint64_t t0,t1;
801 +
802 + /* clamp key */
803 + t0 = U8TO64_LE(key + 0);
804 + t1 = U8TO64_LE(key + 8);
805 + r0 = t0 & 0xffc0fffffff; t0 >>= 44; t0 |= t1 << 20;
806 + r1 = t0 & 0xfffffc0ffff; t1 >>= 24;
807 + r2 = t1 & 0x00ffffffc0f;
808 +
809 + /* store r in un-used space of st->P[1] */
810 + p = &st->P[1];
811 + p->R20.d[1] = (uint32_t)(r0 );
812 + p->R20.d[3] = (uint32_t)(r0 >> 32);
813 + p->R21.d[1] = (uint32_t)(r1 );
814 + p->R21.d[3] = (uint32_t)(r1 >> 32);
815 + p->R22.d[1] = (uint32_t)(r2 );
816 + p->R22.d[3] = (uint32_t)(r2 >> 32);
817 +
818 + /* store pad */
819 + p->R23.d[1] = U8TO32_LE(key + 16);
820 + p->R23.d[3] = U8TO32_LE(key + 20);
821 + p->R24.d[1] = U8TO32_LE(key + 24);
822 + p->R24.d[3] = U8TO32_LE(key + 28);
823 +
824 + /* H = 0 */
825 + st->H[0] = _mm_setzero_si128();
826 + st->H[1] = _mm_setzero_si128();
827 + st->H[2] = _mm_setzero_si128();
828 + st->H[3] = _mm_setzero_si128();
829 + st->H[4] = _mm_setzero_si128();
830 +
831 + st->started = 0;
832 + st->leftover = 0;
833 +}
834 +
835 +static void
836 +poly1305_first_block(poly1305_state_internal *st, const uint8_t *m) {
837 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask);
838 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5);
839 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128);
840 + xmmi T5,T6;
841 + poly1305_power *p;
842 + uint128_t d[3];
843 + uint64_t r0,r1,r2;
844 + uint64_t r20,r21,r22,s22;
845 + uint64_t pad0,pad1;
846 + uint64_t c;
847 + uint64_t i;
848 +
849 + /* pull out stored info */
850 + p = &st->P[1];
851 +
852 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1];
853 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1];
854 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1];
855 + pad0 = ((uint64_t)p->R23.d[3] << 32) | (uint64_t)p->R23.d[1];
856 + pad1 = ((uint64_t)p->R24.d[3] << 32) | (uint64_t)p->R24.d[1];
857 +
858 + /* compute powers r^2,r^4 */
859 + r20 = r0;
860 + r21 = r1;
861 + r22 = r2;
862 + for (i = 0; i < 2; i++) {
863 + s22 = r22 * (5 << 2);
864 +
865 + d[0] = add128(mul64x64_128(r20, r20), mul64x64_128(r21 * 2, s22));
866 + d[1] = add128(mul64x64_128(r22, s22), mul64x64_128(r20 * 2, r21));
867 + d[2] = add128(mul64x64_128(r21, r21), mul64x64_128(r22 * 2, r20));
868 +
869 + r20 = lo128(d[0]) & 0xfffffffffff; c = shr128(d[0], 44);
870 + d[1] = add128_64(d[1], c); r21 = lo128(d[1]) & 0xfffffffffff; c = shr128(d[1], 44);
871 + d[2] = add128_64(d[2], c); r22 = lo128(d[2]) & 0x3ffffffffff; c = shr128(d[2], 42);
872 + r20 += c * 5; c = (r20 >> 44); r20 = r20 & 0xfffffffffff;
873 + r21 += c;
874 +
875 + p->R20.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)( r20 ) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
876 + p->R21.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r20 >> 26) | (r21 << 18)) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
877 + p->R22.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r21 >> 8) ) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
878 + p->R23.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r21 >> 34) | (r22 << 10)) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
879 + p->R24.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r22 >> 16) ) ), _MM_SHUFFLE(1,0,1,0));
880 + p->S21.v = _mm_mul_epu32(p->R21.v, FIVE);
881 + p->S22.v = _mm_mul_epu32(p->R22.v, FIVE);
882 + p->S23.v = _mm_mul_epu32(p->R23.v, FIVE);
883 + p->S24.v = _mm_mul_epu32(p->R24.v, FIVE);
884 + p--;
885 + }
886 +
887 + /* put saved info back */
888 + p = &st->P[1];
889 + p->R20.d[1] = (uint32_t)(r0 );
890 + p->R20.d[3] = (uint32_t)(r0 >> 32);
891 + p->R21.d[1] = (uint32_t)(r1 );
892 + p->R21.d[3] = (uint32_t)(r1 >> 32);
893 + p->R22.d[1] = (uint32_t)(r2 );
894 + p->R22.d[3] = (uint32_t)(r2 >> 32);
895 + p->R23.d[1] = (uint32_t)(pad0 );
896 + p->R23.d[3] = (uint32_t)(pad0 >> 32);
897 + p->R24.d[1] = (uint32_t)(pad1 );
898 + p->R24.d[3] = (uint32_t)(pad1 >> 32);
899 +
900 + /* H = [Mx,My] */
901 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_loadl_epi64((xmmi *)(m + 16)));
902 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_loadl_epi64((xmmi *)(m + 24)));
903 + st->H[0] = _mm_and_si128(MMASK, T5);
904 + st->H[1] = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
905 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12));
906 + st->H[2] = _mm_and_si128(MMASK, T5);
907 + st->H[3] = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
908 + st->H[4] = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
909 +}
910 +
911 +static void
912 +poly1305_blocks(poly1305_state_internal *st, const uint8_t *m, size_t bytes) {
913 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask);
914 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5);
915 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128);
916 +
917 + poly1305_power *p;
918 + xmmi H0,H1,H2,H3,H4;
919 + xmmi T0,T1,T2,T3,T4,T5,T6;
920 + xmmi M0,M1,M2,M3,M4;
921 + xmmi C1,C2;
922 +
923 + H0 = st->H[0];
924 + H1 = st->H[1];
925 + H2 = st->H[2];
926 + H3 = st->H[3];
927 + H4 = st->H[4];
928 +
929 + while (bytes >= 64) {
930 + /* H *= [r^4,r^4] */
931 + p = &st->P[0];
932 + T0 = _mm_mul_epu32(H0, p->R20.v);
933 + T1 = _mm_mul_epu32(H0, p->R21.v);
934 + T2 = _mm_mul_epu32(H0, p->R22.v);
935 + T3 = _mm_mul_epu32(H0, p->R23.v);
936 + T4 = _mm_mul_epu32(H0, p->R24.v);
937 + T5 = _mm_mul_epu32(H1, p->S24.v); T6 = _mm_mul_epu32(H1, p->R20.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
938 + T5 = _mm_mul_epu32(H2, p->S23.v); T6 = _mm_mul_epu32(H2, p->S24.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
939 + T5 = _mm_mul_epu32(H3, p->S22.v); T6 = _mm_mul_epu32(H3, p->S23.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
940 + T5 = _mm_mul_epu32(H4, p->S21.v); T6 = _mm_mul_epu32(H4, p->S22.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
941 + T5 = _mm_mul_epu32(H1, p->R21.v); T6 = _mm_mul_epu32(H1, p->R22.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
942 + T5 = _mm_mul_epu32(H2, p->R20.v); T6 = _mm_mul_epu32(H2, p->R21.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
943 + T5 = _mm_mul_epu32(H3, p->S24.v); T6 = _mm_mul_epu32(H3, p->R20.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
944 + T5 = _mm_mul_epu32(H4, p->S23.v); T6 = _mm_mul_epu32(H4, p->S24.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
945 + T5 = _mm_mul_epu32(H1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
946 + T5 = _mm_mul_epu32(H2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
947 + T5 = _mm_mul_epu32(H3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
948 + T5 = _mm_mul_epu32(H4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
949 +
950 + /* H += [Mx,My]*[r^2,r^2] */
951 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_loadl_epi64((xmmi *)(m + 16)));
952 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_loadl_epi64((xmmi *)(m + 24)));
953 + M0 = _mm_and_si128(MMASK, T5);
954 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
955 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12));
956 + M2 = _mm_and_si128(MMASK, T5);
957 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
958 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
959 +
960 + p = &st->P[1];
961 + T5 = _mm_mul_epu32(M0, p->R20.v); T6 = _mm_mul_epu32(M0, p->R21.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
962 + T5 = _mm_mul_epu32(M1, p->S24.v); T6 = _mm_mul_epu32(M1, p->R20.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
963 + T5 = _mm_mul_epu32(M2, p->S23.v); T6 = _mm_mul_epu32(M2, p->S24.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
964 + T5 = _mm_mul_epu32(M3, p->S22.v); T6 = _mm_mul_epu32(M3, p->S23.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
965 + T5 = _mm_mul_epu32(M4, p->S21.v); T6 = _mm_mul_epu32(M4, p->S22.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
966 + T5 = _mm_mul_epu32(M0, p->R22.v); T6 = _mm_mul_epu32(M0, p->R23.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
967 + T5 = _mm_mul_epu32(M1, p->R21.v); T6 = _mm_mul_epu32(M1, p->R22.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
968 + T5 = _mm_mul_epu32(M2, p->R20.v); T6 = _mm_mul_epu32(M2, p->R21.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
969 + T5 = _mm_mul_epu32(M3, p->S24.v); T6 = _mm_mul_epu32(M3, p->R20.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
970 + T5 = _mm_mul_epu32(M4, p->S23.v); T6 = _mm_mul_epu32(M4, p->S24.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
971 + T5 = _mm_mul_epu32(M0, p->R24.v); T4 = _mm_add_epi64(T4, T5);
972 + T5 = _mm_mul_epu32(M1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
973 + T5 = _mm_mul_epu32(M2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
974 + T5 = _mm_mul_epu32(M3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
975 + T5 = _mm_mul_epu32(M4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
976 +
977 + /* H += [Mx,My] */
978 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 32)), _mm_loadl_epi64((xmmi *)(m + 48)));
979 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 40)), _mm_loadl_epi64((xmmi *)(m + 56)));
980 + M0 = _mm_and_si128(MMASK, T5);
981 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
982 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12));
983 + M2 = _mm_and_si128(MMASK, T5);
984 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
985 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
986 +
987 + T0 = _mm_add_epi64(T0, M0);
988 + T1 = _mm_add_epi64(T1, M1);
989 + T2 = _mm_add_epi64(T2, M2);
990 + T3 = _mm_add_epi64(T3, M3);
991 + T4 = _mm_add_epi64(T4, M4);
992 +
993 + /* reduce */
994 + C1 = _mm_srli_epi64(T0, 26); C2 = _mm_srli_epi64(T3, 26); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_and_si128(T3, MMASK); T1 = _mm_add_epi64(T1, C1); T4 = _mm_add_epi64(T4, C2);
995 + C1 = _mm_srli_epi64(T1, 26); C2 = _mm_srli_epi64(T4, 26); T1 = _mm_and_si128(T1, MMASK); T4 = _mm_and_si128(T4, MMASK); T2 = _mm_add_epi64(T2, C1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE));
996 + C1 = _mm_srli_epi64(T2, 26); C2 = _mm_srli_epi64(T0, 26); T2 = _mm_and_si128(T2, MMASK); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_add_epi64(T3, C1); T1 = _mm_add_epi64(T1, C2);
997 + C1 = _mm_srli_epi64(T3, 26); T3 = _mm_and_si128(T3, MMASK); T4 = _mm_add_epi64(T4, C1);
998 +
999 + /* H = (H*[r^4,r^4] + [Mx,My]*[r^2,r^2] + [Mx,My]) */
1000 + H0 = T0;
1001 + H1 = T1;
1002 + H2 = T2;
1003 + H3 = T3;
1004 + H4 = T4;
1005 +
1006 + m += 64;
1007 + bytes -= 64;
1008 + }
1009 +
1010 + st->H[0] = H0;
1011 + st->H[1] = H1;
1012 + st->H[2] = H2;
1013 + st->H[3] = H3;
1014 + st->H[4] = H4;
1015 +}
1016 +
1017 +static size_t
1018 +poly1305_combine(poly1305_state_internal *st, const uint8_t *m, size_t bytes) {
1019 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask);
1020 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128);
1021 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5);
1022 +
1023 + poly1305_power *p;
1024 + xmmi H0,H1,H2,H3,H4;
1025 + xmmi M0,M1,M2,M3,M4;
1026 + xmmi T0,T1,T2,T3,T4,T5,T6;
1027 + xmmi C1,C2;
1028 +
1029 + uint64_t r0,r1,r2;
1030 + uint64_t t0,t1,t2,t3,t4;
1031 + uint64_t c;
1032 + size_t consumed = 0;
1033 +
1034 + H0 = st->H[0];
1035 + H1 = st->H[1];
1036 + H2 = st->H[2];
1037 + H3 = st->H[3];
1038 + H4 = st->H[4];
1039 +
1040 + /* p = [r^2,r^2] */
1041 + p = &st->P[1];
1042 +
1043 + if (bytes >= 32) {
1044 + /* H *= [r^2,r^2] */
1045 + T0 = _mm_mul_epu32(H0, p->R20.v);
1046 + T1 = _mm_mul_epu32(H0, p->R21.v);
1047 + T2 = _mm_mul_epu32(H0, p->R22.v);
1048 + T3 = _mm_mul_epu32(H0, p->R23.v);
1049 + T4 = _mm_mul_epu32(H0, p->R24.v);
1050 + T5 = _mm_mul_epu32(H1, p->S24.v); T6 = _mm_mul_epu32(H1, p->R20.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1051 + T5 = _mm_mul_epu32(H2, p->S23.v); T6 = _mm_mul_epu32(H2, p->S24.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1052 + T5 = _mm_mul_epu32(H3, p->S22.v); T6 = _mm_mul_epu32(H3, p->S23.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1053 + T5 = _mm_mul_epu32(H4, p->S21.v); T6 = _mm_mul_epu32(H4, p->S22.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1054 + T5 = _mm_mul_epu32(H1, p->R21.v); T6 = _mm_mul_epu32(H1, p->R22.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1055 + T5 = _mm_mul_epu32(H2, p->R20.v); T6 = _mm_mul_epu32(H2, p->R21.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1056 + T5 = _mm_mul_epu32(H3, p->S24.v); T6 = _mm_mul_epu32(H3, p->R20.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1057 + T5 = _mm_mul_epu32(H4, p->S23.v); T6 = _mm_mul_epu32(H4, p->S24.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1058 + T5 = _mm_mul_epu32(H1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
1059 + T5 = _mm_mul_epu32(H2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
1060 + T5 = _mm_mul_epu32(H3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
1061 + T5 = _mm_mul_epu32(H4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
1062 +
1063 + /* H += [Mx,My] */
1064 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_loadl_epi64((xmmi *)(m + 16)));
1065 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_loadl_epi64((xmmi *)(m + 24)));
1066 + M0 = _mm_and_si128(MMASK, T5);
1067 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
1068 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12));
1069 + M2 = _mm_and_si128(MMASK, T5);
1070 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
1071 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
1072 +
1073 + T0 = _mm_add_epi64(T0, M0);
1074 + T1 = _mm_add_epi64(T1, M1);
1075 + T2 = _mm_add_epi64(T2, M2);
1076 + T3 = _mm_add_epi64(T3, M3);
1077 + T4 = _mm_add_epi64(T4, M4);
1078 +
1079 + /* reduce */
1080 + C1 = _mm_srli_epi64(T0, 26); C2 = _mm_srli_epi64(T3, 26); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_and_si128(T3, MMASK); T1 = _mm_add_epi64(T1, C1); T4 = _mm_add_epi64(T4, C2);
1081 + C1 = _mm_srli_epi64(T1, 26); C2 = _mm_srli_epi64(T4, 26); T1 = _mm_and_si128(T1, MMASK); T4 = _mm_and_si128(T4, MMASK); T2 = _mm_add_epi64(T2, C1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE));
1082 + C1 = _mm_srli_epi64(T2, 26); C2 = _mm_srli_epi64(T0, 26); T2 = _mm_and_si128(T2, MMASK); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_add_epi64(T3, C1); T1 = _mm_add_epi64(T1, C2);
1083 + C1 = _mm_srli_epi64(T3, 26); T3 = _mm_and_si128(T3, MMASK); T4 = _mm_add_epi64(T4, C1);
1084 +
1085 + /* H = (H*[r^2,r^2] + [Mx,My]) */
1086 + H0 = T0;
1087 + H1 = T1;
1088 + H2 = T2;
1089 + H3 = T3;
1090 + H4 = T4;
1091 +
1092 + consumed = 32;
1093 + }
1094 +
1095 + /* finalize, H *= [r^2,r] */
1096 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1];
1097 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1];
1098 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1];
1099 +
1100 + p->R20.d[2] = (uint32_t)( r0 ) & 0x3ffffff;
1101 + p->R21.d[2] = (uint32_t)((r0 >> 26) | (r1 << 18)) & 0x3ffffff;
1102 + p->R22.d[2] = (uint32_t)((r1 >> 8) ) & 0x3ffffff;
1103 + p->R23.d[2] = (uint32_t)((r1 >> 34) | (r2 << 10)) & 0x3ffffff;
1104 + p->R24.d[2] = (uint32_t)((r2 >> 16) ) ;
1105 + p->S21.d[2] = p->R21.d[2] * 5;
1106 + p->S22.d[2] = p->R22.d[2] * 5;
1107 + p->S23.d[2] = p->R23.d[2] * 5;
1108 + p->S24.d[2] = p->R24.d[2] * 5;
1109 +
1110 + /* H *= [r^2,r] */
1111 + T0 = _mm_mul_epu32(H0, p->R20.v);
1112 + T1 = _mm_mul_epu32(H0, p->R21.v);
1113 + T2 = _mm_mul_epu32(H0, p->R22.v);
1114 + T3 = _mm_mul_epu32(H0, p->R23.v);
1115 + T4 = _mm_mul_epu32(H0, p->R24.v);
1116 + T5 = _mm_mul_epu32(H1, p->S24.v); T6 = _mm_mul_epu32(H1, p->R20.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1117 + T5 = _mm_mul_epu32(H2, p->S23.v); T6 = _mm_mul_epu32(H2, p->S24.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1118 + T5 = _mm_mul_epu32(H3, p->S22.v); T6 = _mm_mul_epu32(H3, p->S23.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1119 + T5 = _mm_mul_epu32(H4, p->S21.v); T6 = _mm_mul_epu32(H4, p->S22.v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6);
1120 + T5 = _mm_mul_epu32(H1, p->R21.v); T6 = _mm_mul_epu32(H1, p->R22.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1121 + T5 = _mm_mul_epu32(H2, p->R20.v); T6 = _mm_mul_epu32(H2, p->R21.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1122 + T5 = _mm_mul_epu32(H3, p->S24.v); T6 = _mm_mul_epu32(H3, p->R20.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1123 + T5 = _mm_mul_epu32(H4, p->S23.v); T6 = _mm_mul_epu32(H4, p->S24.v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6);
1124 + T5 = _mm_mul_epu32(H1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
1125 + T5 = _mm_mul_epu32(H2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
1126 + T5 = _mm_mul_epu32(H3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
1127 + T5 = _mm_mul_epu32(H4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
1128 +
1129 + C1 = _mm_srli_epi64(T0, 26); C2 = _mm_srli_epi64(T3, 26); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_and_si128(T3, MMASK); T1 = _mm_add_epi64(T1, C1); T4 = _mm_add_epi64(T4, C2);
1130 + C1 = _mm_srli_epi64(T1, 26); C2 = _mm_srli_epi64(T4, 26); T1 = _mm_and_si128(T1, MMASK); T4 = _mm_and_si128(T4, MMASK); T2 = _mm_add_epi64(T2, C1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE));
1131 + C1 = _mm_srli_epi64(T2, 26); C2 = _mm_srli_epi64(T0, 26); T2 = _mm_and_si128(T2, MMASK); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_add_epi64(T3, C1); T1 = _mm_add_epi64(T1, C2);
1132 + C1 = _mm_srli_epi64(T3, 26); T3 = _mm_and_si128(T3, MMASK); T4 = _mm_add_epi64(T4, C1);
1133 +
1134 + /* H = H[0]+H[1] */
1135 + H0 = _mm_add_epi64(T0, _mm_srli_si128(T0, 8));
1136 + H1 = _mm_add_epi64(T1, _mm_srli_si128(T1, 8));
1137 + H2 = _mm_add_epi64(T2, _mm_srli_si128(T2, 8));
1138 + H3 = _mm_add_epi64(T3, _mm_srli_si128(T3, 8));
1139 + H4 = _mm_add_epi64(T4, _mm_srli_si128(T4, 8));
1140 +
1141 + t0 = _mm_cvtsi128_si32(H0) ; c = (t0 >> 26); t0 &= 0x3ffffff;
1142 + t1 = _mm_cvtsi128_si32(H1) + c; c = (t1 >> 26); t1 &= 0x3ffffff;
1143 + t2 = _mm_cvtsi128_si32(H2) + c; c = (t2 >> 26); t2 &= 0x3ffffff;
1144 + t3 = _mm_cvtsi128_si32(H3) + c; c = (t3 >> 26); t3 &= 0x3ffffff;
1145 + t4 = _mm_cvtsi128_si32(H4) + c; c = (t4 >> 26); t4 &= 0x3ffffff;
1146 + t0 = t0 + (c * 5); c = (t0 >> 26); t0 &= 0x3ffffff;
1147 + t1 = t1 + c;
1148 +
1149 + st->HH[0] = ((t0 ) | (t1 << 26) ) & 0xfffffffffffull;
1150 + st->HH[1] = ((t1 >> 18) | (t2 << 8) | (t3 << 34)) & 0xfffffffffffull;
1151 + st->HH[2] = ((t3 >> 10) | (t4 << 16) ) & 0x3ffffffffffull;
1152 +
1153 + return consumed;
1154 +}
1155 +
1156 +void
1157 +Poly1305Update(poly1305_state *state, const unsigned char *m, size_t bytes) {
1158 + poly1305_state_internal *st = poly1305_aligned_state(state);
1159 + size_t want;
1160 +
1161 + /* need at least 32 initial bytes to start the accelerated branch */
1162 + if (!st->started) {
1163 + if ((st->leftover == 0) && (bytes > 32)) {
1164 + poly1305_first_block(st, m);
1165 + m += 32;
1166 + bytes -= 32;
1167 + } else {
1168 + want = poly1305_min(32 - st->leftover, bytes);
1169 + poly1305_block_copy(st->buffer + st->leftover, m, want);
1170 + bytes -= want;
1171 + m += want;
1172 + st->leftover += want;
1173 + if ((st->leftover < 32) || (bytes == 0))
1174 + return;
1175 + poly1305_first_block(st, st->buffer);
1176 + st->leftover = 0;
1177 + }
1178 + st->started = 1;
1179 + }
1180 +
1181 + /* handle leftover */
1182 + if (st->leftover) {
1183 + want = poly1305_min(64 - st->leftover, bytes);
1184 + poly1305_block_copy(st->buffer + st->leftover, m, want);
1185 + bytes -= want;
1186 + m += want;
1187 + st->leftover += want;
1188 + if (st->leftover < 64)
1189 + return;
1190 + poly1305_blocks(st, st->buffer, 64);
1191 + st->leftover = 0;
1192 + }
1193 +
1194 + /* process 64 byte blocks */
1195 + if (bytes >= 64) {
1196 + want = (bytes & ~63);
1197 + poly1305_blocks(st, m, want);
1198 + m += want;
1199 + bytes -= want;
1200 + }
1201 +
1202 + if (bytes) {
1203 + poly1305_block_copy(st->buffer + st->leftover, m, bytes);
1204 + st->leftover += bytes;
1205 + }
1206 +}
1207 +
1208 +void
1209 +Poly1305Finish(poly1305_state *state, unsigned char mac[16]) {
1210 + poly1305_state_internal *st = poly1305_aligned_state(state);
1211 + size_t leftover = st->leftover;
1212 + uint8_t *m = st->buffer;
1213 + uint128_t d[3];
1214 + uint64_t h0,h1,h2;
1215 + uint64_t t0,t1;
1216 + uint64_t g0,g1,g2,c,nc;
1217 + uint64_t r0,r1,r2,s1,s2;
1218 + poly1305_power *p;
1219 +
1220 + if (st->started) {
1221 + size_t consumed = poly1305_combine(st, m, leftover);
1222 + leftover -= consumed;
1223 + m += consumed;
1224 + }
1225 +
1226 + /* st->HH will either be 0 or have the combined result */
1227 + h0 = st->HH[0];
1228 + h1 = st->HH[1];
1229 + h2 = st->HH[2];
1230 +
1231 + p = &st->P[1];
1232 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1];
1233 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1];
1234 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1];
1235 + s1 = r1 * (5 << 2);
1236 + s2 = r2 * (5 << 2);
1237 +
1238 + if (leftover < 16)
1239 + goto poly1305_donna_atmost15bytes;
1240 +
1241 +poly1305_donna_atleast16bytes:
1242 + t0 = U8TO64_LE(m + 0);
1243 + t1 = U8TO64_LE(m + 8);
1244 + h0 += t0 & 0xfffffffffff;
1245 + t0 = shr128_pair(t1, t0, 44);
1246 + h1 += t0 & 0xfffffffffff;
1247 + h2 += (t1 >> 24) | ((uint64_t)1 << 40);
1248 +
1249 +poly1305_donna_mul:
1250 + d[0] = add128(add128(mul64x64_128(h0, r0), mul64x64_128(h1, s2)), mul64x64_128(h2, s1));
1251 + d[1] = add128(add128(mul64x64_128(h0, r1), mul64x64_128(h1, r0)), mul64x64_128(h2, s2));
1252 + d[2] = add128(add128(mul64x64_128(h0, r2), mul64x64_128(h1, r1)), mul64x64_128(h2, r0));
1253 + h0 = lo128(d[0]) & 0xfffffffffff; c = shr128(d[0], 44);
1254 + d[1] = add128_64(d[1], c); h1 = lo128(d[1]) & 0xfffffffffff; c = shr128(d[1], 44);
1255 + d[2] = add128_64(d[2], c); h2 = lo128(d[2]) & 0x3ffffffffff; c = shr128(d[2], 42);
1256 + h0 += c * 5;
1257 +
1258 + m += 16;
1259 + leftover -= 16;
1260 + if (leftover >= 16) goto poly1305_donna_atleast16bytes;
1261 +
1262 + /* final bytes */
1263 +poly1305_donna_atmost15bytes:
1264 + if (!leftover) goto poly1305_donna_finish;
1265 +
1266 + m[leftover++] = 1;
1267 + poly1305_block_zero(m + leftover, 16 - leftover);
1268 + leftover = 16;
1269 +
1270 + t0 = U8TO64_LE(m+0);
1271 + t1 = U8TO64_LE(m+8);
1272 + h0 += t0 & 0xfffffffffff; t0 = shr128_pair(t1, t0, 44);
1273 + h1 += t0 & 0xfffffffffff;
1274 + h2 += (t1 >> 24);
1275 +
1276 + goto poly1305_donna_mul;
1277 +
1278 +poly1305_donna_finish:
1279 + c = (h0 >> 44); h0 &= 0xfffffffffff;
1280 + h1 += c; c = (h1 >> 44); h1 &= 0xfffffffffff;
1281 + h2 += c; c = (h2 >> 42); h2 &= 0x3ffffffffff;
1282 + h0 += c * 5;
1283 +
1284 + g0 = h0 + 5; c = (g0 >> 44); g0 &= 0xfffffffffff;
1285 + g1 = h1 + c; c = (g1 >> 44); g1 &= 0xfffffffffff;
1286 + g2 = h2 + c - ((uint64_t)1 << 42);
1287 +
1288 + c = (g2 >> 63) - 1;
1289 + nc = ~c;
1290 + h0 = (h0 & nc) | (g0 & c);
1291 + h1 = (h1 & nc) | (g1 & c);
1292 + h2 = (h2 & nc) | (g2 & c);
1293 +
1294 + /* pad */
1295 + t0 = ((uint64_t)p->R23.d[3] << 32) | (uint64_t)p->R23.d[1];
1296 + t1 = ((uint64_t)p->R24.d[3] << 32) | (uint64_t)p->R24.d[1];
1297 + h0 += (t0 & 0xfffffffffff) ; c = (h0 >> 44); h0 &= 0xfffffffffff; t0 = shr128_pair(t1, t0, 44);
1298 + h1 += (t0 & 0xfffffffffff) + c; c = (h1 >> 44); h1 &= 0xfffffffffff; t1 = (t1 >> 24);
1299 + h2 += (t1 ) + c;
1300 +
1301 + U64TO8_LE(mac + 0, ((h0 ) | (h1 << 44)));
1302 + U64TO8_LE(mac + 8, ((h1 >> 20) | (h2 << 24)));
1303 +}
1304 diff -r c3565a90b8c4 lib/freebl/poly1305/poly1305.c
1305 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1306 +++ b/lib/freebl/poly1305/poly1305.c Tue Jan 07 12:11:36 2014 -0800
1307 @@ -0,0 +1,254 @@
1308 +/* This Source Code Form is subject to the terms of the Mozilla Public
1309 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1310 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1311 +
1312 +/* This implementation of poly1305 is by Andrew Moon
1313 + * (https://github.com/floodyberry/poly1305-donna) and released as public
1314 + * domain. */
1315 +
1316 +#include <string.h>
1317 +#include <stdint.h>
1318 +
1319 +#include "poly1305.h"
1320 +
1321 +#if defined(NSS_X86) || defined(NSS_X64)
1322 +/* We can assume little-endian. */
1323 +static uint32_t U8TO32_LE(const unsigned char *m) {
1324 + uint32_t r;
1325 + memcpy(&r, m, sizeof(r));
1326 + return r;
1327 +}
1328 +
1329 +static void U32TO8_LE(unsigned char *m, uint32_t v) {
1330 + memcpy(m, &v, sizeof(v));
1331 +}
1332 +#else
1333 +static uint32_t U8TO32_LE(const unsigned char *m) {
1334 + return (uint32_t)m[0] |
1335 + (uint32_t)m[1] << 8 |
1336 + (uint32_t)m[2] << 16 |
1337 + (uint32_t)m[3] << 24;
1338 +}
1339 +
1340 +static void U32TO8_LE(unsigned char *m, uint32_t v) {
1341 + m[0] = v;
1342 + m[1] = v >> 8;
1343 + m[2] = v >> 16;
1344 + m[3] = v >> 24;
1345 +}
1346 +#endif
1347 +
1348 +static uint64_t
1349 +mul32x32_64(uint32_t a, uint32_t b) {
1350 + return (uint64_t)a * b;
1351 +}
1352 +
1353 +struct poly1305_state_st {
1354 + uint32_t r0,r1,r2,r3,r4;
1355 + uint32_t s1,s2,s3,s4;
1356 + uint32_t h0,h1,h2,h3,h4;
1357 + unsigned char buf[16];
1358 + unsigned int buf_used;
1359 + unsigned char key[16];
1360 +};
1361 +
1362 +/* update updates |state| given some amount of input data. This function may
1363 + * only be called with a |len| that is not a multiple of 16 at the end of the
1364 + * data. Otherwise the input must be buffered into 16 byte blocks. */
1365 +static void update(struct poly1305_state_st *state, const unsigned char *in,
1366 + size_t len) {
1367 + uint32_t t0,t1,t2,t3;
1368 + uint64_t t[5];
1369 + uint32_t b;
1370 + uint64_t c;
1371 + size_t j;
1372 + unsigned char mp[16];
1373 +
1374 + if (len < 16)
1375 + goto poly1305_donna_atmost15bytes;
1376 +
1377 +poly1305_donna_16bytes:
1378 + t0 = U8TO32_LE(in);
1379 + t1 = U8TO32_LE(in+4);
1380 + t2 = U8TO32_LE(in+8);
1381 + t3 = U8TO32_LE(in+12);
1382 +
1383 + in += 16;
1384 + len -= 16;
1385 +
1386 + state->h0 += t0 & 0x3ffffff;
1387 + state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
1388 + state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
1389 + state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
1390 + state->h4 += (t3 >> 8) | (1 << 24);
1391 +
1392 +poly1305_donna_mul:
1393 + t[0] = mul32x32_64(state->h0,state->r0) +
1394 + mul32x32_64(state->h1,state->s4) +
1395 + mul32x32_64(state->h2,state->s3) +
1396 + mul32x32_64(state->h3,state->s2) +
1397 + mul32x32_64(state->h4,state->s1);
1398 + t[1] = mul32x32_64(state->h0,state->r1) +
1399 + mul32x32_64(state->h1,state->r0) +
1400 + mul32x32_64(state->h2,state->s4) +
1401 + mul32x32_64(state->h3,state->s3) +
1402 + mul32x32_64(state->h4,state->s2);
1403 + t[2] = mul32x32_64(state->h0,state->r2) +
1404 + mul32x32_64(state->h1,state->r1) +
1405 + mul32x32_64(state->h2,state->r0) +
1406 + mul32x32_64(state->h3,state->s4) +
1407 + mul32x32_64(state->h4,state->s3);
1408 + t[3] = mul32x32_64(state->h0,state->r3) +
1409 + mul32x32_64(state->h1,state->r2) +
1410 + mul32x32_64(state->h2,state->r1) +
1411 + mul32x32_64(state->h3,state->r0) +
1412 + mul32x32_64(state->h4,state->s4);
1413 + t[4] = mul32x32_64(state->h0,state->r4) +
1414 + mul32x32_64(state->h1,state->r3) +
1415 + mul32x32_64(state->h2,state->r2) +
1416 + mul32x32_64(state->h3,state->r1) +
1417 + mul32x32_64(state->h4,state->r0);
1418 +
1419 + state->h0 = (uint32_t)t[0] & 0x3ffffff; c = (t[0] >> 26);
1420 + t[1] += c; state->h1 = (uint32_t)t[1] & 0x3ffffff; b = (uint32_t)(t[1] >> 26);
1421 + t[2] += b; state->h2 = (uint32_t)t[2] & 0x3ffffff; b = (uint32_t)(t[2] >> 26);
1422 + t[3] += b; state->h3 = (uint32_t)t[3] & 0x3ffffff; b = (uint32_t)(t[3] >> 26);
1423 + t[4] += b; state->h4 = (uint32_t)t[4] & 0x3ffffff; b = (uint32_t)(t[4] >> 26);
1424 + state->h0 += b * 5;
1425 +
1426 + if (len >= 16)
1427 + goto poly1305_donna_16bytes;
1428 +
1429 + /* final bytes */
1430 +poly1305_donna_atmost15bytes:
1431 + if (!len)
1432 + return;
1433 +
1434 + for (j = 0; j < len; j++)
1435 + mp[j] = in[j];
1436 + mp[j++] = 1;
1437 + for (; j < 16; j++)
1438 + mp[j] = 0;
1439 + len = 0;
1440 +
1441 + t0 = U8TO32_LE(mp+0);
1442 + t1 = U8TO32_LE(mp+4);
1443 + t2 = U8TO32_LE(mp+8);
1444 + t3 = U8TO32_LE(mp+12);
1445 +
1446 + state->h0 += t0 & 0x3ffffff;
1447 + state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
1448 + state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
1449 + state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
1450 + state->h4 += (t3 >> 8);
1451 +
1452 + goto poly1305_donna_mul;
1453 +}
1454 +
1455 +void Poly1305Init(poly1305_state *statep, const unsigned char key[32]) {
1456 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep;
1457 + uint32_t t0,t1,t2,t3;
1458 +
1459 + t0 = U8TO32_LE(key+0);
1460 + t1 = U8TO32_LE(key+4);
1461 + t2 = U8TO32_LE(key+8);
1462 + t3 = U8TO32_LE(key+12);
1463 +
1464 + /* precompute multipliers */
1465 + state->r0 = t0 & 0x3ffffff; t0 >>= 26; t0 |= t1 << 6;
1466 + state->r1 = t0 & 0x3ffff03; t1 >>= 20; t1 |= t2 << 12;
1467 + state->r2 = t1 & 0x3ffc0ff; t2 >>= 14; t2 |= t3 << 18;
1468 + state->r3 = t2 & 0x3f03fff; t3 >>= 8;
1469 + state->r4 = t3 & 0x00fffff;
1470 +
1471 + state->s1 = state->r1 * 5;
1472 + state->s2 = state->r2 * 5;
1473 + state->s3 = state->r3 * 5;
1474 + state->s4 = state->r4 * 5;
1475 +
1476 + /* init state */
1477 + state->h0 = 0;
1478 + state->h1 = 0;
1479 + state->h2 = 0;
1480 + state->h3 = 0;
1481 + state->h4 = 0;
1482 +
1483 + state->buf_used = 0;
1484 + memcpy(state->key, key + 16, sizeof(state->key));
1485 +}
1486 +
1487 +void Poly1305Update(poly1305_state *statep, const unsigned char *in,
1488 + size_t in_len) {
1489 + unsigned int i;
1490 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep;
1491 +
1492 + if (state->buf_used) {
1493 + unsigned int todo = 16 - state->buf_used;
1494 + if (todo > in_len)
1495 + todo = in_len;
1496 + for (i = 0; i < todo; i++)
1497 + state->buf[state->buf_used + i] = in[i];
1498 + state->buf_used += todo;
1499 + in_len -= todo;
1500 + in += todo;
1501 +
1502 + if (state->buf_used == 16) {
1503 + update(state, state->buf, 16);
1504 + state->buf_used = 0;
1505 + }
1506 + }
1507 +
1508 + if (in_len >= 16) {
1509 + size_t todo = in_len & ~0xf;
1510 + update(state, in, todo);
1511 + in += todo;
1512 + in_len &= 0xf;
1513 + }
1514 +
1515 + if (in_len) {
1516 + for (i = 0; i < in_len; i++)
1517 + state->buf[i] = in[i];
1518 + state->buf_used = in_len;
1519 + }
1520 +}
1521 +
1522 +void Poly1305Finish(poly1305_state *statep, unsigned char mac[16]) {
1523 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep;
1524 + uint64_t f0,f1,f2,f3;
1525 + uint32_t g0,g1,g2,g3,g4;
1526 + uint32_t b, nb;
1527 +
1528 + if (state->buf_used)
1529 + update(state, state->buf, state->buf_used);
1530 +
1531 + b = state->h0 >> 26; state->h0 = state->h0 & 0x3ffffff;
1532 + state->h1 += b; b = state->h1 >> 26; state->h1 = state->h1 & 0x3ffffff;
1533 + state->h2 += b; b = state->h2 >> 26; state->h2 = state->h2 & 0x3ffffff;
1534 + state->h3 += b; b = state->h3 >> 26; state->h3 = state->h3 & 0x3ffffff;
1535 + state->h4 += b; b = state->h4 >> 26; state->h4 = state->h4 & 0x3ffffff;
1536 + state->h0 += b * 5;
1537 +
1538 + g0 = state->h0 + 5; b = g0 >> 26; g0 &= 0x3ffffff;
1539 + g1 = state->h1 + b; b = g1 >> 26; g1 &= 0x3ffffff;
1540 + g2 = state->h2 + b; b = g2 >> 26; g2 &= 0x3ffffff;
1541 + g3 = state->h3 + b; b = g3 >> 26; g3 &= 0x3ffffff;
1542 + g4 = state->h4 + b - (1 << 26);
1543 +
1544 + b = (g4 >> 31) - 1;
1545 + nb = ~b;
1546 + state->h0 = (state->h0 & nb) | (g0 & b);
1547 + state->h1 = (state->h1 & nb) | (g1 & b);
1548 + state->h2 = (state->h2 & nb) | (g2 & b);
1549 + state->h3 = (state->h3 & nb) | (g3 & b);
1550 + state->h4 = (state->h4 & nb) | (g4 & b);
1551 +
1552 + f0 = ((state->h0 ) | (state->h1 << 26)) + (uint64_t)U8TO32_LE(&state->key[0]);
1553 + f1 = ((state->h1 >> 6) | (state->h2 << 20)) + (uint64_t)U8TO32_LE(&state->key[4]);
1554 + f2 = ((state->h2 >> 12) | (state->h3 << 14)) + (uint64_t)U8TO32_LE(&state->key[8]);
1555 + f3 = ((state->h3 >> 18) | (state->h4 << 8)) + (uint64_t)U8TO32_LE(&state->key[12]);
1556 +
1557 + U32TO8_LE(&mac[ 0], f0); f1 += (f0 >> 32);
1558 + U32TO8_LE(&mac[ 4], f1); f2 += (f1 >> 32);
1559 + U32TO8_LE(&mac[ 8], f2); f3 += (f2 >> 32);
1560 + U32TO8_LE(&mac[12], f3);
1561 +}
1562 diff -r c3565a90b8c4 lib/freebl/poly1305/poly1305.h
1563 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1564 +++ b/lib/freebl/poly1305/poly1305.h Tue Jan 07 12:11:36 2014 -0800
1565 @@ -0,0 +1,31 @@
1566 +/*
1567 + * poly1305.h - header file for Poly1305 implementation.
1568 + *
1569 + * This Source Code Form is subject to the terms of the Mozilla Public
1570 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1571 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1572 +
1573 +#ifndef FREEBL_POLY1305_H_
1574 +#define FREEBL_POLY1305_H_
1575 +
1576 +typedef unsigned char poly1305_state[512];
1577 +
1578 +/* Poly1305Init sets up |state| so that it can be used to calculate an
1579 + * authentication tag with the one-time key |key|. Note that |key| is a
1580 + * one-time key and therefore there is no `reset' method because that would
1581 + * enable several messages to be authenticated with the same key. */
1582 +extern void Poly1305Init(poly1305_state* state,
1583 + const unsigned char key[32]);
1584 +
1585 +/* Poly1305Update processes |in_len| bytes from |in|. It can be called zero or
1586 + * more times after poly1305_init. */
1587 +extern void Poly1305Update(poly1305_state* state,
1588 + const unsigned char *in,
1589 + size_t inLen);
1590 +
1591 +/* Poly1305Finish completes the poly1305 calculation and writes a 16 byte
1592 + * authentication tag to |mac|. */
1593 +extern void Poly1305Finish(poly1305_state* state,
1594 + unsigned char mac[16]);
1595 +
1596 +#endif /* FREEBL_POLY1305_H_ */
1597 diff -r c3565a90b8c4 lib/pk11wrap/pk11mech.c
1598 --- a/lib/pk11wrap/pk11mech.c Fri Jan 03 20:59:10 2014 +0100
1599 +++ b/lib/pk11wrap/pk11mech.c Tue Jan 07 12:11:36 2014 -0800
1600 @@ -152,6 +152,8 @@
1601 return CKM_SEED_CBC;
1602 case CKK_CAMELLIA:
1603 return CKM_CAMELLIA_CBC;
1604 + case CKK_NSS_CHACHA20:
1605 + return CKM_NSS_CHACHA20_POLY1305;
1606 case CKK_AES:
1607 return CKM_AES_CBC;
1608 case CKK_DES:
1609 @@ -219,6 +221,8 @@
1610 case CKM_CAMELLIA_CBC_PAD:
1611 case CKM_CAMELLIA_KEY_GEN:
1612 return CKK_CAMELLIA;
1613 + case CKM_NSS_CHACHA20_POLY1305:
1614 + return CKK_NSS_CHACHA20;
1615 case CKM_AES_ECB:
1616 case CKM_AES_CBC:
1617 case CKM_AES_CCM:
1618 @@ -429,6 +433,8 @@
1619 case CKM_CAMELLIA_CBC_PAD:
1620 case CKM_CAMELLIA_KEY_GEN:
1621 return CKM_CAMELLIA_KEY_GEN;
1622 + case CKM_NSS_CHACHA20_POLY1305:
1623 + return CKM_NSS_CHACHA20_KEY_GEN;
1624 case CKM_AES_ECB:
1625 case CKM_AES_CBC:
1626 case CKM_AES_CCM:
1627 diff -r c3565a90b8c4 lib/softoken/pkcs11.c
1628 --- a/lib/softoken/pkcs11.c Fri Jan 03 20:59:10 2014 +0100
1629 +++ b/lib/softoken/pkcs11.c Tue Jan 07 12:11:36 2014 -0800
1630 @@ -368,6 +368,9 @@
1631 {CKM_SEED_MAC, {16, 16, CKF_SN_VR}, PR_TRUE},
1632 {CKM_SEED_MAC_GENERAL, {16, 16, CKF_SN_VR}, PR_TRUE},
1633 {CKM_SEED_CBC_PAD, {16, 16, CKF_EN_DE_WR_UN}, PR_TRUE},
1634 + /* ------------------------- ChaCha20 Operations ---------------------- */
1635 + {CKM_NSS_CHACHA20_KEY_GEN, {32, 32, CKF_GENERATE}, PR_TRUE},
1636 + {CKM_NSS_CHACHA20_POLY1305,{32, 32, CKF_EN_DE}, PR_TRUE},
1637 /* ------------------------- Hashing Operations ----------------------- */
1638 {CKM_MD2, {0, 0, CKF_DIGEST}, PR_FALSE},
1639 {CKM_MD2_HMAC, {1, 128, CKF_SN_VR}, PR_TRUE},
1640 diff -r c3565a90b8c4 lib/softoken/pkcs11c.c
1641 --- a/lib/softoken/pkcs11c.c Fri Jan 03 20:59:10 2014 +0100
1642 +++ b/lib/softoken/pkcs11c.c Tue Jan 07 12:11:36 2014 -0800
1643 @@ -632,6 +632,97 @@
1644 return rv;
1645 }
1646
1647 +static SFTKChaCha20Poly1305Info *
1648 +sftk_ChaCha20Poly1305_CreateContext(const unsigned char *key,
1649 + unsigned int keyLen,
1650 + const CK_NSS_AEAD_PARAMS* params)
1651 +{
1652 + SFTKChaCha20Poly1305Info *ctx;
1653 +
1654 + if (params->ulIvLen != sizeof(ctx->nonce)) {
1655 + PORT_SetError(SEC_ERROR_INPUT_LEN);
1656 + return NULL;
1657 + }
1658 +
1659 + ctx = PORT_New(SFTKChaCha20Poly1305Info);
1660 + if (ctx == NULL) {
1661 + return NULL;
1662 + }
1663 +
1664 + if (ChaCha20Poly1305_InitContext(&ctx->freeblCtx, key, keyLen,
1665 + params->ulTagLen) != SECSuccess) {
1666 + PORT_Free(ctx);
1667 + return NULL;
1668 + }
1669 +
1670 + memcpy(ctx->nonce, params->pIv, sizeof(ctx->nonce));
1671 +
1672 + if (params->ulAADLen > sizeof(ctx->ad)) {
1673 + /* Need to allocate an overflow buffer for the additional data. */
1674 + ctx->adOverflow = (unsigned char *)PORT_Alloc(params->ulAADLen);
1675 + if (!ctx->adOverflow) {
1676 + PORT_Free(ctx);
1677 + return NULL;
1678 + }
1679 + memcpy(ctx->adOverflow, params->pAAD, params->ulAADLen);
1680 + } else {
1681 + ctx->adOverflow = NULL;
1682 + memcpy(ctx->ad, params->pAAD, params->ulAADLen);
1683 + }
1684 + ctx->adLen = params->ulAADLen;
1685 +
1686 + return ctx;
1687 +}
1688 +
1689 +static void
1690 +sftk_ChaCha20Poly1305_DestroyContext(SFTKChaCha20Poly1305Info *ctx,
1691 + PRBool freeit)
1692 +{
1693 + ChaCha20Poly1305_DestroyContext(&ctx->freeblCtx, PR_FALSE);
1694 + if (ctx->adOverflow != NULL) {
1695 + PORT_Free(ctx->adOverflow);
1696 + ctx->adOverflow = NULL;
1697 + }
1698 + ctx->adLen = 0;
1699 + if (freeit) {
1700 + PORT_Free(ctx);
1701 + }
1702 +}
1703 +
1704 +static SECStatus
1705 +sftk_ChaCha20Poly1305_Encrypt(const SFTKChaCha20Poly1305Info *ctx,
1706 + unsigned char *output, unsigned int *outputLen,
1707 + unsigned int maxOutputLen,
1708 + const unsigned char *input, unsigned int inputLen)
1709 +{
1710 + const unsigned char *ad = ctx->adOverflow;
1711 +
1712 + if (ad == NULL) {
1713 + ad = ctx->ad;
1714 + }
1715 +
1716 + return ChaCha20Poly1305_Seal(&ctx->freeblCtx, output, outputLen,
1717 + maxOutputLen, input, inputLen, ctx->nonce,
1718 + sizeof(ctx->nonce), ad, ctx->adLen);
1719 +}
1720 +
1721 +static SECStatus
1722 +sftk_ChaCha20Poly1305_Decrypt(const SFTKChaCha20Poly1305Info *ctx,
1723 + unsigned char *output, unsigned int *outputLen,
1724 + unsigned int maxOutputLen,
1725 + const unsigned char *input, unsigned int inputLen)
1726 +{
1727 + const unsigned char *ad = ctx->adOverflow;
1728 +
1729 + if (ad == NULL) {
1730 + ad = ctx->ad;
1731 + }
1732 +
1733 + return ChaCha20Poly1305_Open(&ctx->freeblCtx, output, outputLen,
1734 + maxOutputLen, input, inputLen, ctx->nonce,
1735 + sizeof(ctx->nonce), ad, ctx->adLen);
1736 +}
1737 +
1738 /** NSC_CryptInit initializes an encryption/Decryption operation.
1739 *
1740 * Always called by NSC_EncryptInit, NSC_DecryptInit, NSC_WrapKey,NSC_UnwrapKey.
1741 @@ -1027,6 +1118,35 @@
1742 context->destroy = (SFTKDestroy) AES_DestroyContext;
1743 break;
1744
1745 + case CKM_NSS_CHACHA20_POLY1305:
1746 + if (pMechanism->ulParameterLen != sizeof(CK_NSS_AEAD_PARAMS)) {
1747 + crv = CKR_MECHANISM_PARAM_INVALID;
1748 + break;
1749 + }
1750 + context->multi = PR_FALSE;
1751 + if (key_type != CKK_NSS_CHACHA20) {
1752 + crv = CKR_KEY_TYPE_INCONSISTENT;
1753 + break;
1754 + }
1755 + att = sftk_FindAttribute(key,CKA_VALUE);
1756 + if (att == NULL) {
1757 + crv = CKR_KEY_HANDLE_INVALID;
1758 + break;
1759 + }
1760 + context->cipherInfo = sftk_ChaCha20Poly1305_CreateContext(
1761 + (unsigned char*) att->attrib.pValue, att->attrib.ulValueLen,
1762 + (CK_NSS_AEAD_PARAMS*) pMechanism->pParameter);
1763 + sftk_FreeAttribute(att);
1764 + if (context->cipherInfo == NULL) {
1765 + crv = sftk_MapCryptError(PORT_GetError());
1766 + break;
1767 + }
1768 + context->update = (SFTKCipher) (isEncrypt ?
1769 + sftk_ChaCha20Poly1305_Encrypt :
1770 + sftk_ChaCha20Poly1305_Decrypt);
1771 + context->destroy = (SFTKDestroy) sftk_ChaCha20Poly1305_DestroyContext;
1772 + break;
1773 +
1774 case CKM_NETSCAPE_AES_KEY_WRAP_PAD:
1775 context->doPad = PR_TRUE;
1776 /* fall thru */
1777 @@ -3601,6 +3721,10 @@
1778 *key_type = CKK_AES;
1779 if (*key_length == 0) crv = CKR_TEMPLATE_INCOMPLETE;
1780 break;
1781 + case CKM_NSS_CHACHA20_KEY_GEN:
1782 + *key_type = CKK_NSS_CHACHA20;
1783 + if (*key_length == 0) crv = CKR_TEMPLATE_INCOMPLETE;
1784 + break;
1785 default:
1786 PORT_Assert(0);
1787 crv = CKR_MECHANISM_INVALID;
1788 @@ -3846,6 +3970,7 @@
1789 case CKM_SEED_KEY_GEN:
1790 case CKM_CAMELLIA_KEY_GEN:
1791 case CKM_AES_KEY_GEN:
1792 + case CKM_NSS_CHACHA20_KEY_GEN:
1793 #if NSS_SOFTOKEN_DOES_RC5
1794 case CKM_RC5_KEY_GEN:
1795 #endif
1796 diff -r c3565a90b8c4 lib/softoken/pkcs11i.h
1797 --- a/lib/softoken/pkcs11i.h Fri Jan 03 20:59:10 2014 +0100
1798 +++ b/lib/softoken/pkcs11i.h Tue Jan 07 12:11:36 2014 -0800
1799 @@ -14,6 +14,7 @@
1800 #include "pkcs11t.h"
1801
1802 #include "sftkdbt.h"
1803 +#include "chacha20poly1305.h"
1804 #include "hasht.h"
1805
1806 /*
1807 @@ -104,6 +105,7 @@
1808 typedef struct SFTKOAEPEncryptInfoStr SFTKOAEPEncryptInfo;
1809 typedef struct SFTKOAEPDecryptInfoStr SFTKOAEPDecryptInfo;
1810 typedef struct SFTKSSLMACInfoStr SFTKSSLMACInfo;
1811 +typedef struct SFTKChaCha20Poly1305InfoStr SFTKChaCha20Poly1305Info;
1812 typedef struct SFTKItemTemplateStr SFTKItemTemplate;
1813
1814 /* define function pointer typdefs for pointer tables */
1815 @@ -399,6 +401,16 @@
1816 unsigned int keySize;
1817 };
1818
1819 +/* SFTKChaCha20Poly1305Info saves the key, tag length, nonce, and additional
1820 + * data for a ChaCha20+Poly1305 AEAD operation. */
1821 +struct SFTKChaCha20Poly1305InfoStr {
1822 + ChaCha20Poly1305Context freeblCtx;
1823 + unsigned char nonce[8];
1824 + unsigned char ad[16];
1825 + unsigned char *adOverflow;
1826 + unsigned int adLen;
1827 +};
1828 +
1829 /*
1830 * Template based on SECItems, suitable for passing as arrays
1831 */
1832 diff -r c3565a90b8c4 lib/util/pkcs11n.h
1833 --- a/lib/util/pkcs11n.h Fri Jan 03 20:59:10 2014 +0100
1834 +++ b/lib/util/pkcs11n.h Tue Jan 07 12:11:36 2014 -0800
1835 @@ -51,6 +51,8 @@
1836 #define CKK_NSS_JPAKE_ROUND1 (CKK_NSS + 2)
1837 #define CKK_NSS_JPAKE_ROUND2 (CKK_NSS + 3)
1838
1839 +#define CKK_NSS_CHACHA20 (CKK_NSS + 4)
1840 +
1841 /*
1842 * NSS-defined certificate types
1843 *
1844 @@ -214,6 +216,9 @@
1845 #define CKM_NSS_TLS_KEY_AND_MAC_DERIVE_SHA256 (CKM_NSS + 23)
1846 #define CKM_NSS_TLS_MASTER_KEY_DERIVE_DH_SHA256 (CKM_NSS + 24)
1847
1848 +#define CKM_NSS_CHACHA20_KEY_GEN (CKM_NSS + 25)
1849 +#define CKM_NSS_CHACHA20_POLY1305 (CKM_NSS + 26)
1850 +
1851 /*
1852 * HISTORICAL:
1853 * Do not attempt to use these. They are only used by NETSCAPE's internal
1854 @@ -281,6 +286,14 @@
1855 CK_ULONG ulHeaderLen; /* in */
1856 } CK_NSS_MAC_CONSTANT_TIME_PARAMS;
1857
1858 +typedef struct CK_NSS_AEAD_PARAMS {
1859 + CK_BYTE_PTR pIv; /* This is the nonce. */
1860 + CK_ULONG ulIvLen;
1861 + CK_BYTE_PTR pAAD;
1862 + CK_ULONG ulAADLen;
1863 + CK_ULONG ulTagLen;
1864 +} CK_NSS_AEAD_PARAMS;
1865 +
1866 /*
1867 * NSS-defined return values
1868 *
This site is hosted by Intevation GmbH (Datenschutzerklärung und Impressum | Privacy Policy and Imprint)