Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(24)

Issues Search
    My Issues | Starred     Open | Closed | All

Side by Side Diff: patches/nss-chacha20-poly1305.patch

Issue 1843333003: Update NSPR to 4.12 and NSS to 3.23 on iOS (Closed) Base URL: https://chromium.googlesource.com/chromium/deps/nss.git@master
Patch Set: Created 4 years, 8 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« nss/lib/util/secoid.c ('K') | « patches/nspr-unused-variable.patch ('k') | patches/nss-static.patch » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
(Empty)
1 diff --git a/lib/freebl/blapi.h b/lib/freebl/blapi.h
2 index 8324714..682be76 100644
3 --- a/lib/freebl/blapi.h
4 +++ b/lib/freebl/blapi.h
5 @@ -986,6 +986,38 @@ Camellia_Decrypt(CamelliaContext *cx, unsigned char *output ,
6 unsigned int *outputLen, unsigned int maxOutputLen,
7 const unsigned char *input, unsigned int inputLen);
8
9 +/******************************************/
10 +/*
11 +** ChaCha20+Poly1305 AEAD
12 +*/
13 +
14 +extern SECStatus
15 +ChaCha20Poly1305_InitContext(ChaCha20Poly1305Context *ctx,
16 + const unsigned char *key, unsigned int keyLen,
17 + unsigned int tagLen);
18 +
19 +extern ChaCha20Poly1305Context *
20 +ChaCha20Poly1305_CreateContext(const unsigned char *key, unsigned int keyLen,
21 + unsigned int tagLen);
22 +
23 +extern void
24 +ChaCha20Poly1305_DestroyContext(ChaCha20Poly1305Context *ctx, PRBool freeit);
25 +
26 +extern SECStatus
27 +ChaCha20Poly1305_Seal(const ChaCha20Poly1305Context *ctx,
28 + unsigned char *output, unsigned int *outputLen,
29 + unsigned int maxOutputLen,
30 + const unsigned char *input, unsigned int inputLen,
31 + const unsigned char *nonce, unsigned int nonceLen,
32 + const unsigned char *ad, unsigned int adLen);
33 +
34 +extern SECStatus
35 +ChaCha20Poly1305_Open(const ChaCha20Poly1305Context *ctx,
36 + unsigned char *output, unsigned int *outputLen,
37 + unsigned int maxOutputLen,
38 + const unsigned char *input, unsigned int inputLen,
39 + const unsigned char *nonce, unsigned int nonceLen,
40 + const unsigned char *ad, unsigned int adLen);
41
42 /******************************************/
43 /*
44 diff --git a/lib/freebl/blapit.h b/lib/freebl/blapit.h
45 index 8e172d4..5726dc7 100644
46 --- a/lib/freebl/blapit.h
47 +++ b/lib/freebl/blapit.h
48 @@ -222,6 +222,7 @@ struct SHA256ContextStr ;
49 struct SHA512ContextStr ;
50 struct AESKeyWrapContextStr ;
51 struct SEEDContextStr ;
52 +struct ChaCha20Poly1305ContextStr;
53
54 typedef struct DESContextStr DESContext;
55 typedef struct RC2ContextStr RC2Context;
56 @@ -240,6 +241,7 @@ typedef struct SHA512ContextStr SHA512Context;
57 typedef struct SHA512ContextStr SHA384Context;
58 typedef struct AESKeyWrapContextStr AESKeyWrapContext;
59 typedef struct SEEDContextStr SEEDContext;
60 +typedef struct ChaCha20Poly1305ContextStr ChaCha20Poly1305Context;
61
62 /***************************************************************************
63 ** RSA Public and Private Key structures
64 diff --git a/lib/freebl/chacha20/chacha20.c b/lib/freebl/chacha20/chacha20.c
65 new file mode 100644
66 index 0000000..ca0b1ff
67 --- /dev/null
68 +++ b/lib/freebl/chacha20/chacha20.c
69 @@ -0,0 +1,108 @@
70 +/* This Source Code Form is subject to the terms of the Mozilla Public
71 + * License, v. 2.0. If a copy of the MPL was not distributed with this
72 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
73 +
74 +/* Adopted from the public domain code in NaCl by djb. */
75 +
76 +#include <string.h>
77 +#include <stdio.h>
78 +
79 +#include "prtypes.h"
80 +#include "chacha20.h"
81 +
82 +#define ROTL32(v, n) (((v) << (n)) | ((v) >> (32 - (n))))
83 +#define ROTATE(v, c) ROTL32((v), (c))
84 +#define XOR(v, w) ((v) ^ (w))
85 +#define PLUS(x, y) ((x) + (y))
86 +
87 +#define U32TO8_LITTLE(p, v) \
88 + { (p)[0] = ((v) ) & 0xff; (p)[1] = ((v) >> 8) & 0xff; \
89 + (p)[2] = ((v) >> 16) & 0xff; (p)[3] = ((v) >> 24) & 0xff; }
90 +#define U8TO32_LITTLE(p) \
91 + (((PRUint32)((p)[0]) ) | ((PRUint32)((p)[1]) << 8) | \
92 + ((PRUint32)((p)[2]) << 16) | ((PRUint32)((p)[3]) << 24) )
93 +
94 +#define QUARTERROUND(a,b,c,d) \
95 + x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]),16); \
96 + x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]),12); \
97 + x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]), 8); \
98 + x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]), 7);
99 +
100 +static void ChaChaCore(unsigned char output[64], const PRUint32 input[16],
101 + int num_rounds) {
102 + PRUint32 x[16];
103 + int i;
104 +
105 + memcpy(x, input, sizeof(PRUint32) * 16);
106 + for (i = num_rounds; i > 0; i -= 2) {
107 + QUARTERROUND( 0, 4, 8,12)
108 + QUARTERROUND( 1, 5, 9,13)
109 + QUARTERROUND( 2, 6,10,14)
110 + QUARTERROUND( 3, 7,11,15)
111 + QUARTERROUND( 0, 5,10,15)
112 + QUARTERROUND( 1, 6,11,12)
113 + QUARTERROUND( 2, 7, 8,13)
114 + QUARTERROUND( 3, 4, 9,14)
115 + }
116 +
117 + for (i = 0; i < 16; ++i) {
118 + x[i] = PLUS(x[i], input[i]);
119 + }
120 + for (i = 0; i < 16; ++i) {
121 + U32TO8_LITTLE(output + 4 * i, x[i]);
122 + }
123 +}
124 +
125 +static const unsigned char sigma[16] = "expand 32-byte k";
126 +
127 +void ChaCha20XOR(unsigned char *out, const unsigned char *in, unsigned int inLe n,
128 + const unsigned char key[32], const unsigned char nonce[8],
129 + uint64_t counter) {
130 + unsigned char block[64];
131 + PRUint32 input[16];
132 + unsigned int u;
133 + unsigned int i;
134 +
135 + input[4] = U8TO32_LITTLE(key + 0);
136 + input[5] = U8TO32_LITTLE(key + 4);
137 + input[6] = U8TO32_LITTLE(key + 8);
138 + input[7] = U8TO32_LITTLE(key + 12);
139 +
140 + input[8] = U8TO32_LITTLE(key + 16);
141 + input[9] = U8TO32_LITTLE(key + 20);
142 + input[10] = U8TO32_LITTLE(key + 24);
143 + input[11] = U8TO32_LITTLE(key + 28);
144 +
145 + input[0] = U8TO32_LITTLE(sigma + 0);
146 + input[1] = U8TO32_LITTLE(sigma + 4);
147 + input[2] = U8TO32_LITTLE(sigma + 8);
148 + input[3] = U8TO32_LITTLE(sigma + 12);
149 +
150 + input[12] = (PRUint32)counter;
151 + input[13] = (PRUint32)(counter >> 32);
152 + input[14] = U8TO32_LITTLE(nonce + 0);
153 + input[15] = U8TO32_LITTLE(nonce + 4);
154 +
155 + while (inLen >= 64) {
156 + ChaChaCore(block, input, 20);
157 + for (i = 0; i < 64; i++) {
158 + out[i] = in[i] ^ block[i];
159 + }
160 +
161 + input[12]++;
162 + if (input[12] == 0) {
163 + input[13]++;
164 + }
165 +
166 + inLen -= 64;
167 + in += 64;
168 + out += 64;
169 + }
170 +
171 + if (inLen > 0) {
172 + ChaChaCore(block, input, 20);
173 + for (i = 0; i < inLen; i++) {
174 + out[i] = in[i] ^ block[i];
175 + }
176 + }
177 +}
178 diff --git a/lib/freebl/chacha20/chacha20.h b/lib/freebl/chacha20/chacha20.h
179 new file mode 100644
180 index 0000000..6336ba7
181 --- /dev/null
182 +++ b/lib/freebl/chacha20/chacha20.h
183 @@ -0,0 +1,22 @@
184 +/*
185 + * chacha20.h - header file for ChaCha20 implementation.
186 + *
187 + * This Source Code Form is subject to the terms of the Mozilla Public
188 + * License, v. 2.0. If a copy of the MPL was not distributed with this
189 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
190 +
191 +#ifndef FREEBL_CHACHA20_H_
192 +#define FREEBL_CHACHA20_H_
193 +
194 +#include <stdint.h>
195 +
196 +/* ChaCha20XOR encrypts |inLen| bytes from |in| with the given key and
197 + * nonce and writes the result to |out|, which may be equal to |in|. The
198 + * initial block counter is specified by |counter|. */
199 +extern void ChaCha20XOR(unsigned char *out,
200 + const unsigned char *in, unsigned int inLen,
201 + const unsigned char key[32],
202 + const unsigned char nonce[8],
203 + uint64_t counter);
204 +
205 +#endif /* FREEBL_CHACHA20_H_ */
206 diff --git a/lib/freebl/chacha20/chacha20_vec.c b/lib/freebl/chacha20/chacha20_v ec.c
207 new file mode 100644
208 index 0000000..c3573b3
209 --- /dev/null
210 +++ b/lib/freebl/chacha20/chacha20_vec.c
211 @@ -0,0 +1,281 @@
212 +/* This Source Code Form is subject to the terms of the Mozilla Public
213 + * License, v. 2.0. If a copy of the MPL was not distributed with this
214 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
215 +
216 +/* This implementation is by Ted Krovetz and was submitted to SUPERCOP and
217 + * marked as public domain. It was been altered to allow for non-aligned inputs
218 + * and to allow the block counter to be passed in specifically. */
219 +
220 +#include <string.h>
221 +
222 +#include "chacha20.h"
223 +
224 +#ifndef CHACHA_RNDS
225 +#define CHACHA_RNDS 20 /* 8 (high speed), 20 (conservative), 12 (middle) */
226 +#endif
227 +
228 +/* Architecture-neutral way to specify 16-byte vector of ints */
229 +typedef unsigned vec __attribute__ ((vector_size (16)));
230 +
231 +/* This implementation is designed for Neon, SSE and AltiVec machines. The
232 + * following specify how to do certain vector operations efficiently on
233 + * each architecture, using intrinsics.
234 + * This implementation supports parallel processing of multiple blocks,
235 + * including potentially using general-purpose registers.
236 + */
237 +#if __ARM_NEON__
238 +#include <arm_neon.h>
239 +#define GPR_TOO 1
240 +#define VBPI 2
241 +#define ONE (vec)vsetq_lane_u32(1,vdupq_n_u32(0),0)
242 +#define LOAD(m) (vec)(*((vec*)(m)))
243 +#define STORE(m,r) (*((vec*)(m))) = (r)
244 +#define ROTV1(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,1)
245 +#define ROTV2(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,2)
246 +#define ROTV3(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,3)
247 +#define ROTW16(x) (vec)vrev32q_u16((uint16x8_t)x)
248 +#if __clang__
249 +#define ROTW7(x) (x << ((vec){ 7, 7, 7, 7})) ^ (x >> ((vec){25,25,25,25}))
250 +#define ROTW8(x) (x << ((vec){ 8, 8, 8, 8})) ^ (x >> ((vec){24,24,24,24}))
251 +#define ROTW12(x) (x << ((vec){12,12,12,12})) ^ (x >> ((vec){20,20,20,20}))
252 +#else
253 +#define ROTW7(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,7),(uint32x4_t)x,2 5)
254 +#define ROTW8(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,8),(uint32x4_t)x,2 4)
255 +#define ROTW12(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,12),(uint32x4_t)x, 20)
256 +#endif
257 +#elif __SSE2__
258 +#include <emmintrin.h>
259 +#define GPR_TOO 0
260 +#if __clang__
261 +#define VBPI 4
262 +#else
263 +#define VBPI 3
264 +#endif
265 +#define ONE (vec)_mm_set_epi32(0,0,0,1)
266 +#define LOAD(m) (vec)_mm_loadu_si128((__m128i*)(m))
267 +#define STORE(m,r) _mm_storeu_si128((__m128i*)(m), (__m128i) (r))
268 +#define ROTV1(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(0,3,2,1))
269 +#define ROTV2(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(1,0,3,2))
270 +#define ROTV3(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(2,1,0,3))
271 +#define ROTW7(x) (vec)(_mm_slli_epi32((__m128i)x, 7) ^ _mm_srli_epi32((__m128i )x,25))
272 +#define ROTW12(x) (vec)(_mm_slli_epi32((__m128i)x,12) ^ _mm_srli_epi32((__m128i )x,20))
273 +#if __SSSE3__
274 +#include <tmmintrin.h>
275 +#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))
276 +#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))
277 +#else
278 +#define ROTW8(x) (vec)(_mm_slli_epi32((__m128i)x, 8) ^ _mm_srli_epi32((__m128i )x,24))
279 +#define ROTW16(x) (vec)(_mm_slli_epi32((__m128i)x,16) ^ _mm_srli_epi32((__m128i )x,16))
280 +#endif
281 +#else
282 +#error -- Implementation supports only machines with neon or SSE2
283 +#endif
284 +
285 +#ifndef REVV_BE
286 +#define REVV_BE(x) (x)
287 +#endif
288 +
289 +#ifndef REVW_BE
290 +#define REVW_BE(x) (x)
291 +#endif
292 +
293 +#define BPI (VBPI + GPR_TOO) /* Blocks computed per loop iteration */
294 +
295 +#define DQROUND_VECTORS(a,b,c,d) \
296 + a += b; d ^= a; d = ROTW16(d); \
297 + c += d; b ^= c; b = ROTW12(b); \
298 + a += b; d ^= a; d = ROTW8(d); \
299 + c += d; b ^= c; b = ROTW7(b); \
300 + b = ROTV1(b); c = ROTV2(c); d = ROTV3(d); \
301 + a += b; d ^= a; d = ROTW16(d); \
302 + c += d; b ^= c; b = ROTW12(b); \
303 + a += b; d ^= a; d = ROTW8(d); \
304 + c += d; b ^= c; b = ROTW7(b); \
305 + b = ROTV3(b); c = ROTV2(c); d = ROTV1(d);
306 +
307 +#define QROUND_WORDS(a,b,c,d) \
308 + a = a+b; d ^= a; d = d<<16 | d>>16; \
309 + c = c+d; b ^= c; b = b<<12 | b>>20; \
310 + a = a+b; d ^= a; d = d<< 8 | d>>24; \
311 + c = c+d; b ^= c; b = b<< 7 | b>>25;
312 +
313 +#define WRITE_XOR(in, op, d, v0, v1, v2, v3) \
314 + STORE(op + d + 0, LOAD(in + d + 0) ^ REVV_BE(v0)); \
315 + STORE(op + d + 4, LOAD(in + d + 4) ^ REVV_BE(v1)); \
316 + STORE(op + d + 8, LOAD(in + d + 8) ^ REVV_BE(v2)); \
317 + STORE(op + d +12, LOAD(in + d +12) ^ REVV_BE(v3));
318 +
319 +void ChaCha20XOR(
320 + unsigned char *out,
321 + const unsigned char *in,
322 + unsigned int inlen,
323 + const unsigned char key[32],
324 + const unsigned char nonce[8],
325 + uint64_t counter)
326 +{
327 + unsigned iters, i, *op=(unsigned *)out, *ip=(unsigned *)in, *kp;
328 +#if defined(__ARM_NEON__)
329 + unsigned *np;
330 +#endif
331 + vec s0, s1, s2, s3;
332 +#if !defined(__ARM_NEON__) && !defined(__SSE2__)
333 + __attribute__ ((aligned (16))) unsigned key[8], nonce[4];
334 +#endif
335 + __attribute__ ((aligned (16))) unsigned chacha_const[] =
336 + {0x61707865,0x3320646E,0x79622D32,0x6B206574};
337 +#if defined(__ARM_NEON__) || defined(__SSE2__)
338 + kp = (unsigned *)key;
339 +#else
340 + ((vec *)key)[0] = REVV_BE(((vec *)key)[0]);
341 + ((vec *)key)[1] = REVV_BE(((vec *)key)[1]);
342 + nonce[0] = REVW_BE(((unsigned *)nonce)[0]);
343 + nonce[1] = REVW_BE(((unsigned *)nonce)[1]);
344 + nonce[2] = REVW_BE(((unsigned *)nonce)[2]);
345 + nonce[3] = REVW_BE(((unsigned *)nonce)[3]);
346 + kp = (unsigned *)key;
347 + np = (unsigned *)nonce;
348 +#endif
349 +#if defined(__ARM_NEON__)
350 + np = (unsigned*) nonce;
351 +#endif
352 + s0 = LOAD(chacha_const);
353 + s1 = LOAD(&((vec*)kp)[0]);
354 + s2 = LOAD(&((vec*)kp)[1]);
355 + s3 = (vec) {
356 + counter & 0xffffffff,
357 + counter >> 32,
358 + ((uint32_t*)nonce)[0],
359 + ((uint32_t*)nonce)[1]
360 + };
361 +
362 + for (iters = 0; iters < inlen/(BPI*64); iters++) {
363 +#if GPR_TOO
364 + register unsigned x0, x1, x2, x3, x4, x5, x6, x7, x8,
365 + x9, x10, x11, x12, x13, x14, x15;
366 +#endif
367 +#if VBPI > 2
368 + vec v8,v9,v10,v11;
369 +#endif
370 +#if VBPI > 3
371 + vec v12,v13,v14,v15;
372 +#endif
373 +
374 + vec v0,v1,v2,v3,v4,v5,v6,v7;
375 + v4 = v0 = s0; v5 = v1 = s1; v6 = v2 = s2; v3 = s3;
376 + v7 = v3 + ONE;
377 +#if VBPI > 2
378 + v8 = v4; v9 = v5; v10 = v6;
379 + v11 = v7 + ONE;
380 +#endif
381 +#if VBPI > 3
382 + v12 = v8; v13 = v9; v14 = v10;
383 + v15 = v11 + ONE;
384 +#endif
385 +#if GPR_TOO
386 + x0 = chacha_const[0]; x1 = chacha_const[1];
387 + x2 = chacha_const[2]; x3 = chacha_const[3];
388 + x4 = kp[0]; x5 = kp[1]; x6 = kp[2]; x7 = kp[3];
389 + x8 = kp[4]; x9 = kp[5]; x10 = kp[6]; x11 = kp[7];
390 + x12 = (counter & 0xffffffff)+BPI*iters+(BPI-1); x13 = counter >> 32;
391 + x14 = np[0]; x15 = np[1];
392 +#endif
393 + for (i = CHACHA_RNDS/2; i; i--) {
394 + DQROUND_VECTORS(v0,v1,v2,v3)
395 + DQROUND_VECTORS(v4,v5,v6,v7)
396 +#if VBPI > 2
397 + DQROUND_VECTORS(v8,v9,v10,v11)
398 +#endif
399 +#if VBPI > 3
400 + DQROUND_VECTORS(v12,v13,v14,v15)
401 +#endif
402 +#if GPR_TOO
403 + QROUND_WORDS( x0, x4, x8,x12)
404 + QROUND_WORDS( x1, x5, x9,x13)
405 + QROUND_WORDS( x2, x6,x10,x14)
406 + QROUND_WORDS( x3, x7,x11,x15)
407 + QROUND_WORDS( x0, x5,x10,x15)
408 + QROUND_WORDS( x1, x6,x11,x12)
409 + QROUND_WORDS( x2, x7, x8,x13)
410 + QROUND_WORDS( x3, x4, x9,x14)
411 +#endif
412 + }
413 +
414 + WRITE_XOR(ip, op, 0, v0+s0, v1+s1, v2+s2, v3+s3)
415 + s3 += ONE;
416 + WRITE_XOR(ip, op, 16, v4+s0, v5+s1, v6+s2, v7+s3)
417 + s3 += ONE;
418 +#if VBPI > 2
419 + WRITE_XOR(ip, op, 32, v8+s0, v9+s1, v10+s2, v11+s3)
420 + s3 += ONE;
421 +#endif
422 +#if VBPI > 3
423 + WRITE_XOR(ip, op, 48, v12+s0, v13+s1, v14+s2, v15+s3)
424 + s3 += ONE;
425 +#endif
426 + ip += VBPI*16;
427 + op += VBPI*16;
428 +#if GPR_TOO
429 + op[0] = REVW_BE(REVW_BE(ip[0]) ^ (x0 + chacha_const[0]));
430 + op[1] = REVW_BE(REVW_BE(ip[1]) ^ (x1 + chacha_const[1]));
431 + op[2] = REVW_BE(REVW_BE(ip[2]) ^ (x2 + chacha_const[2]));
432 + op[3] = REVW_BE(REVW_BE(ip[3]) ^ (x3 + chacha_const[3]));
433 + op[4] = REVW_BE(REVW_BE(ip[4]) ^ (x4 + kp[0]));
434 + op[5] = REVW_BE(REVW_BE(ip[5]) ^ (x5 + kp[1]));
435 + op[6] = REVW_BE(REVW_BE(ip[6]) ^ (x6 + kp[2]));
436 + op[7] = REVW_BE(REVW_BE(ip[7]) ^ (x7 + kp[3]));
437 + op[8] = REVW_BE(REVW_BE(ip[8]) ^ (x8 + kp[4]));
438 + op[9] = REVW_BE(REVW_BE(ip[9]) ^ (x9 + kp[5]));
439 + op[10] = REVW_BE(REVW_BE(ip[10]) ^ (x10 + kp[6]));
440 + op[11] = REVW_BE(REVW_BE(ip[11]) ^ (x11 + kp[7]));
441 + op[12] = REVW_BE(REVW_BE(ip[12]) ^ (x12 + (counter & 0xffffffff)+BPI*ite rs+(BPI-1)));
442 + op[13] = REVW_BE(REVW_BE(ip[13]) ^ (x13 + (counter >> 32)));
443 + op[14] = REVW_BE(REVW_BE(ip[14]) ^ (x14 + np[0]));
444 + op[15] = REVW_BE(REVW_BE(ip[15]) ^ (x15 + np[1]));
445 + s3 += ONE;
446 + ip += 16;
447 + op += 16;
448 +#endif
449 + }
450 +
451 + for (iters = inlen%(BPI*64)/64; iters != 0; iters--) {
452 + vec v0 = s0, v1 = s1, v2 = s2, v3 = s3;
453 + for (i = CHACHA_RNDS/2; i; i--) {
454 + DQROUND_VECTORS(v0,v1,v2,v3);
455 + }
456 + WRITE_XOR(ip, op, 0, v0+s0, v1+s1, v2+s2, v3+s3)
457 + s3 += ONE;
458 + ip += 16;
459 + op += 16;
460 + }
461 +
462 + inlen = inlen % 64;
463 + if (inlen) {
464 + __attribute__ ((aligned (16))) vec buf[4];
465 + vec v0,v1,v2,v3;
466 + v0 = s0; v1 = s1; v2 = s2; v3 = s3;
467 + for (i = CHACHA_RNDS/2; i; i--) {
468 + DQROUND_VECTORS(v0,v1,v2,v3);
469 + }
470 +
471 + if (inlen >= 16) {
472 + STORE(op + 0, LOAD(ip + 0) ^ REVV_BE(v0 + s0));
473 + if (inlen >= 32) {
474 + STORE(op + 4, LOAD(ip + 4) ^ REVV_BE(v1 + s1));
475 + if (inlen >= 48) {
476 + STORE(op + 8, LOAD(ip + 8) ^ REVV_BE(v2 + s2));
477 + buf[3] = REVV_BE(v3 + s3);
478 + } else {
479 + buf[2] = REVV_BE(v2 + s2);
480 + }
481 + } else {
482 + buf[1] = REVV_BE(v1 + s1);
483 + }
484 + } else {
485 + buf[0] = REVV_BE(v0 + s0);
486 + }
487 +
488 + for (i=inlen & ~15; i<inlen; i++) {
489 + ((char *)op)[i] = ((char *)ip)[i] ^ ((char *)buf)[i];
490 + }
491 + }
492 +}
493 diff --git a/lib/freebl/chacha20poly1305.c b/lib/freebl/chacha20poly1305.c
494 new file mode 100644
495 index 0000000..6fa5c4b
496 --- /dev/null
497 +++ b/lib/freebl/chacha20poly1305.c
498 @@ -0,0 +1,169 @@
499 +/* This Source Code Form is subject to the terms of the Mozilla Public
500 + * License, v. 2.0. If a copy of the MPL was not distributed with this
501 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
502 +
503 +#ifdef FREEBL_NO_DEPEND
504 +#include "stubs.h"
505 +#endif
506 +
507 +#include <string.h>
508 +#include <stdio.h>
509 +
510 +#include "seccomon.h"
511 +#include "secerr.h"
512 +#include "blapit.h"
513 +#include "poly1305/poly1305.h"
514 +#include "chacha20/chacha20.h"
515 +#include "chacha20poly1305.h"
516 +
517 +/* Poly1305Do writes the Poly1305 authenticator of the given additional data
518 + * and ciphertext to |out|. */
519 +static void
520 +Poly1305Do(unsigned char *out,
521 + const unsigned char *ad, unsigned int adLen,
522 + const unsigned char *ciphertext, unsigned int ciphertextLen,
523 + const unsigned char key[32])
524 +{
525 + poly1305_state state;
526 + unsigned int j;
527 + unsigned char lengthBytes[8];
528 + unsigned int i;
529 +
530 + Poly1305Init(&state, key);
531 + j = adLen;
532 + for (i = 0; i < sizeof(lengthBytes); i++) {
533 + lengthBytes[i] = j;
534 + j >>= 8;
535 + }
536 + Poly1305Update(&state, ad, adLen);
537 + Poly1305Update(&state, lengthBytes, sizeof(lengthBytes));
538 + j = ciphertextLen;
539 + for (i = 0; i < sizeof(lengthBytes); i++) {
540 + lengthBytes[i] = j;
541 + j >>= 8;
542 + }
543 + Poly1305Update(&state, ciphertext, ciphertextLen);
544 + Poly1305Update(&state, lengthBytes, sizeof(lengthBytes));
545 + Poly1305Finish(&state, out);
546 +}
547 +
548 +SECStatus
549 +ChaCha20Poly1305_InitContext(ChaCha20Poly1305Context *ctx,
550 + const unsigned char *key, unsigned int keyLen,
551 + unsigned int tagLen)
552 +{
553 + if (keyLen != 32) {
554 + PORT_SetError(SEC_ERROR_BAD_KEY);
555 + return SECFailure;
556 + }
557 + if (tagLen == 0 || tagLen > 16) {
558 + PORT_SetError(SEC_ERROR_INPUT_LEN);
559 + return SECFailure;
560 + }
561 +
562 + memcpy(ctx->key, key, sizeof(ctx->key));
563 + ctx->tagLen = tagLen;
564 +
565 + return SECSuccess;
566 +}
567 +
568 +ChaCha20Poly1305Context *
569 +ChaCha20Poly1305_CreateContext(const unsigned char *key, unsigned int keyLen,
570 + unsigned int tagLen)
571 +{
572 + ChaCha20Poly1305Context *ctx;
573 +
574 + ctx = PORT_New(ChaCha20Poly1305Context);
575 + if (ctx == NULL) {
576 + return NULL;
577 + }
578 +
579 + if (ChaCha20Poly1305_InitContext(ctx, key, keyLen, tagLen) != SECSuccess) {
580 + PORT_Free(ctx);
581 + ctx = NULL;
582 + }
583 +
584 + return ctx;
585 +}
586 +
587 +void
588 +ChaCha20Poly1305_DestroyContext(ChaCha20Poly1305Context *ctx, PRBool freeit)
589 +{
590 + memset(ctx, 0, sizeof(*ctx));
591 + if (freeit) {
592 + PORT_Free(ctx);
593 + }
594 +}
595 +
596 +SECStatus
597 +ChaCha20Poly1305_Seal(const ChaCha20Poly1305Context *ctx,
598 + unsigned char *output, unsigned int *outputLen,
599 + unsigned int maxOutputLen,
600 + const unsigned char *input, unsigned int inputLen,
601 + const unsigned char *nonce, unsigned int nonceLen,
602 + const unsigned char *ad, unsigned int adLen)
603 +{
604 + unsigned char block[64];
605 + unsigned char tag[16];
606 +
607 + if (nonceLen != 8) {
608 + PORT_SetError(SEC_ERROR_INPUT_LEN);
609 + return SECFailure;
610 + }
611 + *outputLen = inputLen + ctx->tagLen;
612 + if (maxOutputLen < *outputLen) {
613 + PORT_SetError(SEC_ERROR_OUTPUT_LEN);
614 + return SECFailure;
615 + }
616 +
617 + memset(block, 0, sizeof(block));
618 + // Generate a block of keystream. The first 32 bytes will be the poly1305
619 + // key. The remainder of the block is discarded.
620 + ChaCha20XOR(block, block, sizeof(block), ctx->key, nonce, 0);
621 + ChaCha20XOR(output, input, inputLen, ctx->key, nonce, 1);
622 +
623 + Poly1305Do(tag, ad, adLen, output, inputLen, block);
624 + memcpy(output + inputLen, tag, ctx->tagLen);
625 +
626 + return SECSuccess;
627 +}
628 +
629 +SECStatus
630 +ChaCha20Poly1305_Open(const ChaCha20Poly1305Context *ctx,
631 + unsigned char *output, unsigned int *outputLen,
632 + unsigned int maxOutputLen,
633 + const unsigned char *input, unsigned int inputLen,
634 + const unsigned char *nonce, unsigned int nonceLen,
635 + const unsigned char *ad, unsigned int adLen)
636 +{
637 + unsigned char block[64];
638 + unsigned char tag[16];
639 +
640 + if (nonceLen != 8) {
641 + PORT_SetError(SEC_ERROR_INPUT_LEN);
642 + return SECFailure;
643 + }
644 + if (inputLen < ctx->tagLen) {
645 + PORT_SetError(SEC_ERROR_INPUT_LEN);
646 + return SECFailure;
647 + }
648 + *outputLen = inputLen - ctx->tagLen;
649 + if (maxOutputLen < *outputLen) {
650 + PORT_SetError(SEC_ERROR_OUTPUT_LEN);
651 + return SECFailure;
652 + }
653 +
654 + memset(block, 0, sizeof(block));
655 + // Generate a block of keystream. The first 32 bytes will be the poly1305
656 + // key. The remainder of the block is discarded.
657 + ChaCha20XOR(block, block, sizeof(block), ctx->key, nonce, 0);
658 + Poly1305Do(tag, ad, adLen, input, inputLen - ctx->tagLen, block);
659 + if (NSS_SecureMemcmp(tag, &input[inputLen - ctx->tagLen], ctx->tagLen) != 0 ) {
660 + PORT_SetError(SEC_ERROR_BAD_DATA);
661 + return SECFailure;
662 + }
663 +
664 + ChaCha20XOR(output, input, inputLen - ctx->tagLen, ctx->key, nonce, 1);
665 +
666 + return SECSuccess;
667 +}
668 diff --git a/lib/freebl/chacha20poly1305.h b/lib/freebl/chacha20poly1305.h
669 new file mode 100644
670 index 0000000..c77632a
671 --- /dev/null
672 +++ b/lib/freebl/chacha20poly1305.h
673 @@ -0,0 +1,15 @@
674 +/* This Source Code Form is subject to the terms of the Mozilla Public
675 + * License, v. 2.0. If a copy of the MPL was not distributed with this
676 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
677 +
678 +#ifndef _CHACHA20_POLY1305_H_
679 +#define _CHACHA20_POLY1305_H_ 1
680 +
681 +/* ChaCha20Poly1305ContextStr saves the key and tag length for a
682 + * ChaCha20+Poly1305 AEAD operation. */
683 +struct ChaCha20Poly1305ContextStr {
684 + unsigned char key[32];
685 + unsigned char tagLen;
686 +};
687 +
688 +#endif /* _CHACHA20_POLY1305_H_ */
689 diff --git a/lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c b/ lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c
690 new file mode 100644
691 index 0000000..38cbf35
692 --- /dev/null
693 +++ b/lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c
694 @@ -0,0 +1,623 @@
695 +/* This Source Code Form is subject to the terms of the Mozilla Public
696 + * License, v. 2.0. If a copy of the MPL was not distributed with this
697 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
698 +
699 +/* This implementation of poly1305 is by Andrew Moon
700 + * (https://github.com/floodyberry/poly1305-donna) and released as public
701 + * domain. It implements SIMD vectorization based on the algorithm described in
702 + * http://cr.yp.to/papers.html#neoncrypto. Unrolled to 2 powers, i.e. 64 byte
703 + * block size. */
704 +
705 +#include <emmintrin.h>
706 +#include <stdint.h>
707 +
708 +#include "poly1305.h"
709 +
710 +#define ALIGN(x) __attribute__((aligned(x)))
711 +#define INLINE inline
712 +#define U8TO64_LE(m) (*(uint64_t*)(m))
713 +#define U8TO32_LE(m) (*(uint32_t*)(m))
714 +#define U64TO8_LE(m,v) (*(uint64_t*)(m)) = v
715 +
716 +typedef __m128i xmmi;
717 +typedef unsigned __int128 uint128_t;
718 +
719 +static const uint32_t ALIGN(16) poly1305_x64_sse2_message_mask[4] = {(1 << 26) - 1, 0, (1 << 26) - 1, 0};
720 +static const uint32_t ALIGN(16) poly1305_x64_sse2_5[4] = {5, 0, 5, 0};
721 +static const uint32_t ALIGN(16) poly1305_x64_sse2_1shl128[4] = {(1 << 24), 0, ( 1 << 24), 0};
722 +
723 +static uint128_t INLINE
724 +add128(uint128_t a, uint128_t b) {
725 + return a + b;
726 +}
727 +
728 +static uint128_t INLINE
729 +add128_64(uint128_t a, uint64_t b) {
730 + return a + b;
731 +}
732 +
733 +static uint128_t INLINE
734 +mul64x64_128(uint64_t a, uint64_t b) {
735 + return (uint128_t)a * b;
736 +}
737 +
738 +static uint64_t INLINE
739 +lo128(uint128_t a) {
740 + return (uint64_t)a;
741 +}
742 +
743 +static uint64_t INLINE
744 +shr128(uint128_t v, const int shift) {
745 + return (uint64_t)(v >> shift);
746 +}
747 +
748 +static uint64_t INLINE
749 +shr128_pair(uint64_t hi, uint64_t lo, const int shift) {
750 + return (uint64_t)((((uint128_t)hi << 64) | lo) >> shift);
751 +}
752 +
753 +typedef struct poly1305_power_t {
754 + union {
755 + xmmi v;
756 + uint64_t u[2];
757 + uint32_t d[4];
758 + } R20,R21,R22,R23,R24,S21,S22,S23,S24;
759 +} poly1305_power;
760 +
761 +typedef struct poly1305_state_internal_t {
762 + poly1305_power P[2]; /* 288 bytes, top 32 bit halves unused = 144 by tes of free storage */
763 + union {
764 + xmmi H[5]; /* 80 bytes */
765 + uint64_t HH[10];
766 + };
767 + /* uint64_t r0,r1,r2; [24 bytes] */
768 + /* uint64_t pad0,pad1; [16 bytes] */
769 + uint64_t started; /* 8 bytes */
770 + uint64_t leftover; /* 8 bytes */
771 + uint8_t buffer[64]; /* 64 bytes */
772 +} poly1305_state_internal; /* 448 bytes total + 63 bytes for alignment = 511 bytes raw */
773 +
774 +static poly1305_state_internal INLINE
775 +*poly1305_aligned_state(poly1305_state *state) {
776 + return (poly1305_state_internal *)(((uint64_t)state + 63) & ~63);
777 +}
778 +
779 +/* copy 0-63 bytes */
780 +static void INLINE
781 +poly1305_block_copy(uint8_t *dst, const uint8_t *src, size_t bytes) {
782 + size_t offset = src - dst;
783 + if (bytes & 32) {
784 + _mm_storeu_si128((xmmi *)(dst + 0), _mm_loadu_si128((xmmi *)(dst + offset + 0)));
785 + _mm_storeu_si128((xmmi *)(dst + 16), _mm_loadu_si128((xmmi *)(ds t + offset + 16)));
786 + dst += 32;
787 + }
788 + if (bytes & 16) { _mm_storeu_si128((xmmi *)dst, _mm_loadu_si128((xmmi *) (dst + offset))); dst += 16; }
789 + if (bytes & 8) { *(uint64_t *)dst = *(uint64_t *)(dst + offset); dst += 8; }
790 + if (bytes & 4) { *(uint32_t *)dst = *(uint32_t *)(dst + offset); dst += 4; }
791 + if (bytes & 2) { *(uint16_t *)dst = *(uint16_t *)(dst + offset); dst += 2; }
792 + if (bytes & 1) { *( uint8_t *)dst = *( uint8_t *)(dst + offset); }
793 +}
794 +
795 +/* zero 0-15 bytes */
796 +static void INLINE
797 +poly1305_block_zero(uint8_t *dst, size_t bytes) {
798 + if (bytes & 8) { *(uint64_t *)dst = 0; dst += 8; }
799 + if (bytes & 4) { *(uint32_t *)dst = 0; dst += 4; }
800 + if (bytes & 2) { *(uint16_t *)dst = 0; dst += 2; }
801 + if (bytes & 1) { *( uint8_t *)dst = 0; }
802 +}
803 +
804 +static size_t INLINE
805 +poly1305_min(size_t a, size_t b) {
806 + return (a < b) ? a : b;
807 +}
808 +
809 +void
810 +Poly1305Init(poly1305_state *state, const unsigned char key[32]) {
811 + poly1305_state_internal *st = poly1305_aligned_state(state);
812 + poly1305_power *p;
813 + uint64_t r0,r1,r2;
814 + uint64_t t0,t1;
815 +
816 + /* clamp key */
817 + t0 = U8TO64_LE(key + 0);
818 + t1 = U8TO64_LE(key + 8);
819 + r0 = t0 & 0xffc0fffffff; t0 >>= 44; t0 |= t1 << 20;
820 + r1 = t0 & 0xfffffc0ffff; t1 >>= 24;
821 + r2 = t1 & 0x00ffffffc0f;
822 +
823 + /* store r in un-used space of st->P[1] */
824 + p = &st->P[1];
825 + p->R20.d[1] = (uint32_t)(r0 );
826 + p->R20.d[3] = (uint32_t)(r0 >> 32);
827 + p->R21.d[1] = (uint32_t)(r1 );
828 + p->R21.d[3] = (uint32_t)(r1 >> 32);
829 + p->R22.d[1] = (uint32_t)(r2 );
830 + p->R22.d[3] = (uint32_t)(r2 >> 32);
831 +
832 + /* store pad */
833 + p->R23.d[1] = U8TO32_LE(key + 16);
834 + p->R23.d[3] = U8TO32_LE(key + 20);
835 + p->R24.d[1] = U8TO32_LE(key + 24);
836 + p->R24.d[3] = U8TO32_LE(key + 28);
837 +
838 + /* H = 0 */
839 + st->H[0] = _mm_setzero_si128();
840 + st->H[1] = _mm_setzero_si128();
841 + st->H[2] = _mm_setzero_si128();
842 + st->H[3] = _mm_setzero_si128();
843 + st->H[4] = _mm_setzero_si128();
844 +
845 + st->started = 0;
846 + st->leftover = 0;
847 +}
848 +
849 +static void
850 +poly1305_first_block(poly1305_state_internal *st, const uint8_t *m) {
851 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask );
852 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5);
853 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128);
854 + xmmi T5,T6;
855 + poly1305_power *p;
856 + uint128_t d[3];
857 + uint64_t r0,r1,r2;
858 + uint64_t r20,r21,r22,s22;
859 + uint64_t pad0,pad1;
860 + uint64_t c;
861 + uint64_t i;
862 +
863 + /* pull out stored info */
864 + p = &st->P[1];
865 +
866 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1];
867 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1];
868 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1];
869 + pad0 = ((uint64_t)p->R23.d[3] << 32) | (uint64_t)p->R23.d[1];
870 + pad1 = ((uint64_t)p->R24.d[3] << 32) | (uint64_t)p->R24.d[1];
871 +
872 + /* compute powers r^2,r^4 */
873 + r20 = r0;
874 + r21 = r1;
875 + r22 = r2;
876 + for (i = 0; i < 2; i++) {
877 + s22 = r22 * (5 << 2);
878 +
879 + d[0] = add128(mul64x64_128(r20, r20), mul64x64_128(r21 * 2, s22) );
880 + d[1] = add128(mul64x64_128(r22, s22), mul64x64_128(r20 * 2, r21) );
881 + d[2] = add128(mul64x64_128(r21, r21), mul64x64_128(r22 * 2, r20) );
882 +
883 + r20 = lo128(d[0]) & 0xfffffffffff; c = shr128(d[0], 44);
884 + d[1] = add128_64(d[1], c); r21 = lo128(d[1]) & 0xfffffffffff; c = shr128(d[1], 44);
885 + d[2] = add128_64(d[2], c); r22 = lo128(d[2]) & 0x3ffffffffff; c = shr128(d[2], 42);
886 + r20 += c * 5; c = (r20 >> 44); r20 = r20 & 0xfffffffffff;
887 + r21 += c;
888 +
889 + p->R20.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)( r20 ) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
890 + p->R21.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r20 > > 26) | (r21 << 18)) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
891 + p->R22.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r21 > > 8) ) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
892 + p->R23.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r21 > > 34) | (r22 << 10)) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0));
893 + p->R24.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r22 > > 16) ) ), _MM_SHUFFLE(1,0,1,0));
894 + p->S21.v = _mm_mul_epu32(p->R21.v, FIVE);
895 + p->S22.v = _mm_mul_epu32(p->R22.v, FIVE);
896 + p->S23.v = _mm_mul_epu32(p->R23.v, FIVE);
897 + p->S24.v = _mm_mul_epu32(p->R24.v, FIVE);
898 + p--;
899 + }
900 +
901 + /* put saved info back */
902 + p = &st->P[1];
903 + p->R20.d[1] = (uint32_t)(r0 );
904 + p->R20.d[3] = (uint32_t)(r0 >> 32);
905 + p->R21.d[1] = (uint32_t)(r1 );
906 + p->R21.d[3] = (uint32_t)(r1 >> 32);
907 + p->R22.d[1] = (uint32_t)(r2 );
908 + p->R22.d[3] = (uint32_t)(r2 >> 32);
909 + p->R23.d[1] = (uint32_t)(pad0 );
910 + p->R23.d[3] = (uint32_t)(pad0 >> 32);
911 + p->R24.d[1] = (uint32_t)(pad1 );
912 + p->R24.d[3] = (uint32_t)(pad1 >> 32);
913 +
914 + /* H = [Mx,My] */
915 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_loadl_epi6 4((xmmi *)(m + 16)));
916 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_loadl_epi6 4((xmmi *)(m + 24)));
917 + st->H[0] = _mm_and_si128(MMASK, T5);
918 + st->H[1] = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
919 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12));
920 + st->H[2] = _mm_and_si128(MMASK, T5);
921 + st->H[3] = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
922 + st->H[4] = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
923 +}
924 +
925 +static void
926 +poly1305_blocks(poly1305_state_internal *st, const uint8_t *m, size_t bytes) {
927 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask );
928 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5);
929 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128);
930 +
931 + poly1305_power *p;
932 + xmmi H0,H1,H2,H3,H4;
933 + xmmi T0,T1,T2,T3,T4,T5,T6;
934 + xmmi M0,M1,M2,M3,M4;
935 + xmmi C1,C2;
936 +
937 + H0 = st->H[0];
938 + H1 = st->H[1];
939 + H2 = st->H[2];
940 + H3 = st->H[3];
941 + H4 = st->H[4];
942 +
943 + while (bytes >= 64) {
944 + /* H *= [r^4,r^4] */
945 + p = &st->P[0];
946 + T0 = _mm_mul_epu32(H0, p->R20.v);
947 + T1 = _mm_mul_epu32(H0, p->R21.v);
948 + T2 = _mm_mul_epu32(H0, p->R22.v);
949 + T3 = _mm_mul_epu32(H0, p->R23.v);
950 + T4 = _mm_mul_epu32(H0, p->R24.v);
951 + 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);
952 + 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);
953 + 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);
954 + 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);
955 + 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);
956 + 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);
957 + 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);
958 + 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);
959 + T5 = _mm_mul_epu32(H1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
960 + T5 = _mm_mul_epu32(H2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
961 + T5 = _mm_mul_epu32(H3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
962 + T5 = _mm_mul_epu32(H4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
963 +
964 + /* H += [Mx,My]*[r^2,r^2] */
965 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_lo adl_epi64((xmmi *)(m + 16)));
966 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_lo adl_epi64((xmmi *)(m + 24)));
967 + M0 = _mm_and_si128(MMASK, T5);
968 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
969 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12) );
970 + M2 = _mm_and_si128(MMASK, T5);
971 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
972 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
973 +
974 + p = &st->P[1];
975 + 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);
976 + 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);
977 + 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);
978 + 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);
979 + 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);
980 + 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);
981 + 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);
982 + 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);
983 + 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);
984 + 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);
985 + T5 = _mm_mul_epu32(M0, p->R24.v); T4 = _mm_add_epi64(T4, T5);
986 + T5 = _mm_mul_epu32(M1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
987 + T5 = _mm_mul_epu32(M2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
988 + T5 = _mm_mul_epu32(M3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
989 + T5 = _mm_mul_epu32(M4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
990 +
991 + /* H += [Mx,My] */
992 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 32)), _mm_l oadl_epi64((xmmi *)(m + 48)));
993 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 40)), _mm_l oadl_epi64((xmmi *)(m + 56)));
994 + M0 = _mm_and_si128(MMASK, T5);
995 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
996 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12) );
997 + M2 = _mm_and_si128(MMASK, T5);
998 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
999 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
1000 +
1001 + T0 = _mm_add_epi64(T0, M0);
1002 + T1 = _mm_add_epi64(T1, M1);
1003 + T2 = _mm_add_epi64(T2, M2);
1004 + T3 = _mm_add_epi64(T3, M3);
1005 + T4 = _mm_add_epi64(T4, M4);
1006 +
1007 + /* reduce */
1008 + 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, C 1); T4 = _mm_add_epi64(T4, C2);
1009 + 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, C 1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE));
1010 + 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, C 1); T1 = _mm_add_epi64(T1, C2);
1011 + C1 = _mm_srli_epi64(T3, 26); T3 = _ mm_and_si128(T3, MMASK); T4 = _mm_add_epi64(T4, C 1);
1012 +
1013 + /* H = (H*[r^4,r^4] + [Mx,My]*[r^2,r^2] + [Mx,My]) */
1014 + H0 = T0;
1015 + H1 = T1;
1016 + H2 = T2;
1017 + H3 = T3;
1018 + H4 = T4;
1019 +
1020 + m += 64;
1021 + bytes -= 64;
1022 + }
1023 +
1024 + st->H[0] = H0;
1025 + st->H[1] = H1;
1026 + st->H[2] = H2;
1027 + st->H[3] = H3;
1028 + st->H[4] = H4;
1029 +}
1030 +
1031 +static size_t
1032 +poly1305_combine(poly1305_state_internal *st, const uint8_t *m, size_t bytes) {
1033 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask );
1034 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128);
1035 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5);
1036 +
1037 + poly1305_power *p;
1038 + xmmi H0,H1,H2,H3,H4;
1039 + xmmi M0,M1,M2,M3,M4;
1040 + xmmi T0,T1,T2,T3,T4,T5,T6;
1041 + xmmi C1,C2;
1042 +
1043 + uint64_t r0,r1,r2;
1044 + uint64_t t0,t1,t2,t3,t4;
1045 + uint64_t c;
1046 + size_t consumed = 0;
1047 +
1048 + H0 = st->H[0];
1049 + H1 = st->H[1];
1050 + H2 = st->H[2];
1051 + H3 = st->H[3];
1052 + H4 = st->H[4];
1053 +
1054 + /* p = [r^2,r^2] */
1055 + p = &st->P[1];
1056 +
1057 + if (bytes >= 32) {
1058 + /* H *= [r^2,r^2] */
1059 + T0 = _mm_mul_epu32(H0, p->R20.v);
1060 + T1 = _mm_mul_epu32(H0, p->R21.v);
1061 + T2 = _mm_mul_epu32(H0, p->R22.v);
1062 + T3 = _mm_mul_epu32(H0, p->R23.v);
1063 + T4 = _mm_mul_epu32(H0, p->R24.v);
1064 + 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);
1065 + 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);
1066 + 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);
1067 + 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);
1068 + 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);
1069 + 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);
1070 + 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);
1071 + 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);
1072 + T5 = _mm_mul_epu32(H1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
1073 + T5 = _mm_mul_epu32(H2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
1074 + T5 = _mm_mul_epu32(H3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
1075 + T5 = _mm_mul_epu32(H4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
1076 +
1077 + /* H += [Mx,My] */
1078 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_lo adl_epi64((xmmi *)(m + 16)));
1079 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_lo adl_epi64((xmmi *)(m + 24)));
1080 + M0 = _mm_and_si128(MMASK, T5);
1081 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
1082 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12) );
1083 + M2 = _mm_and_si128(MMASK, T5);
1084 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26));
1085 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT);
1086 +
1087 + T0 = _mm_add_epi64(T0, M0);
1088 + T1 = _mm_add_epi64(T1, M1);
1089 + T2 = _mm_add_epi64(T2, M2);
1090 + T3 = _mm_add_epi64(T3, M3);
1091 + T4 = _mm_add_epi64(T4, M4);
1092 +
1093 + /* reduce */
1094 + 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, C 1); T4 = _mm_add_epi64(T4, C2);
1095 + 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, C 1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE));
1096 + 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, C 1); T1 = _mm_add_epi64(T1, C2);
1097 + C1 = _mm_srli_epi64(T3, 26); T3 = _ mm_and_si128(T3, MMASK); T4 = _mm_add_epi64(T4, C 1);
1098 +
1099 + /* H = (H*[r^2,r^2] + [Mx,My]) */
1100 + H0 = T0;
1101 + H1 = T1;
1102 + H2 = T2;
1103 + H3 = T3;
1104 + H4 = T4;
1105 +
1106 + consumed = 32;
1107 + }
1108 +
1109 + /* finalize, H *= [r^2,r] */
1110 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1];
1111 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1];
1112 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1];
1113 +
1114 + p->R20.d[2] = (uint32_t)( r0 ) & 0x3ffffff;
1115 + p->R21.d[2] = (uint32_t)((r0 >> 26) | (r1 << 18)) & 0x3ffffff;
1116 + p->R22.d[2] = (uint32_t)((r1 >> 8) ) & 0x3ffffff;
1117 + p->R23.d[2] = (uint32_t)((r1 >> 34) | (r2 << 10)) & 0x3ffffff;
1118 + p->R24.d[2] = (uint32_t)((r2 >> 16) ) ;
1119 + p->S21.d[2] = p->R21.d[2] * 5;
1120 + p->S22.d[2] = p->R22.d[2] * 5;
1121 + p->S23.d[2] = p->R23.d[2] * 5;
1122 + p->S24.d[2] = p->R24.d[2] * 5;
1123 +
1124 + /* H *= [r^2,r] */
1125 + T0 = _mm_mul_epu32(H0, p->R20.v);
1126 + T1 = _mm_mul_epu32(H0, p->R21.v);
1127 + T2 = _mm_mul_epu32(H0, p->R22.v);
1128 + T3 = _mm_mul_epu32(H0, p->R23.v);
1129 + T4 = _mm_mul_epu32(H0, p->R24.v);
1130 + 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);
1131 + 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);
1132 + 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);
1133 + 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);
1134 + 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);
1135 + 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);
1136 + 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);
1137 + 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);
1138 + T5 = _mm_mul_epu32(H1, p->R23.v); T4 = _mm_add_epi64(T4, T5);
1139 + T5 = _mm_mul_epu32(H2, p->R22.v); T4 = _mm_add_epi64(T4, T5);
1140 + T5 = _mm_mul_epu32(H3, p->R21.v); T4 = _mm_add_epi64(T4, T5);
1141 + T5 = _mm_mul_epu32(H4, p->R20.v); T4 = _mm_add_epi64(T4, T5);
1142 +
1143 + C1 = _mm_srli_epi64(T0, 26); C2 = _mm_srli_epi64(T3, 26); T0 = _mm_and_s i128(T0, MMASK); T3 = _mm_and_si128(T3, MMASK); T1 = _mm_add_epi64(T1, C1); T4 = _mm_add_epi64(T4, C2);
1144 + C1 = _mm_srli_epi64(T1, 26); C2 = _mm_srli_epi64(T4, 26); T1 = _mm_and_s i128(T1, MMASK); T4 = _mm_and_si128(T4, MMASK); T2 = _mm_add_epi64(T2, C1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE));
1145 + C1 = _mm_srli_epi64(T2, 26); C2 = _mm_srli_epi64(T0, 26); T2 = _mm_and_s i128(T2, MMASK); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_add_epi64(T3, C1); T1 = _mm_add_epi64(T1, C2);
1146 + C1 = _mm_srli_epi64(T3, 26); T3 = _mm_and_s i128(T3, MMASK); T4 = _mm_add_epi64(T4, C1);
1147 +
1148 + /* H = H[0]+H[1] */
1149 + H0 = _mm_add_epi64(T0, _mm_srli_si128(T0, 8));
1150 + H1 = _mm_add_epi64(T1, _mm_srli_si128(T1, 8));
1151 + H2 = _mm_add_epi64(T2, _mm_srli_si128(T2, 8));
1152 + H3 = _mm_add_epi64(T3, _mm_srli_si128(T3, 8));
1153 + H4 = _mm_add_epi64(T4, _mm_srli_si128(T4, 8));
1154 +
1155 + t0 = _mm_cvtsi128_si32(H0) ; c = (t0 >> 26); t0 &= 0x3ffffff;
1156 + t1 = _mm_cvtsi128_si32(H1) + c; c = (t1 >> 26); t1 &= 0x3ffffff;
1157 + t2 = _mm_cvtsi128_si32(H2) + c; c = (t2 >> 26); t2 &= 0x3ffffff;
1158 + t3 = _mm_cvtsi128_si32(H3) + c; c = (t3 >> 26); t3 &= 0x3ffffff;
1159 + t4 = _mm_cvtsi128_si32(H4) + c; c = (t4 >> 26); t4 &= 0x3ffffff;
1160 + t0 = t0 + (c * 5); c = (t0 >> 26); t0 &= 0x3ffffff;
1161 + t1 = t1 + c;
1162 +
1163 + st->HH[0] = ((t0 ) | (t1 << 26) ) & 0xfffffffffffull;
1164 + st->HH[1] = ((t1 >> 18) | (t2 << 8) | (t3 << 34)) & 0xfffffffffffull;
1165 + st->HH[2] = ((t3 >> 10) | (t4 << 16) ) & 0x3ffffffffffull;
1166 +
1167 + return consumed;
1168 +}
1169 +
1170 +void
1171 +Poly1305Update(poly1305_state *state, const unsigned char *m, size_t bytes) {
1172 + poly1305_state_internal *st = poly1305_aligned_state(state);
1173 + size_t want;
1174 +
1175 + /* need at least 32 initial bytes to start the accelerated branch */
1176 + if (!st->started) {
1177 + if ((st->leftover == 0) && (bytes > 32)) {
1178 + poly1305_first_block(st, m);
1179 + m += 32;
1180 + bytes -= 32;
1181 + } else {
1182 + want = poly1305_min(32 - st->leftover, bytes);
1183 + poly1305_block_copy(st->buffer + st->leftover, m, want);
1184 + bytes -= want;
1185 + m += want;
1186 + st->leftover += want;
1187 + if ((st->leftover < 32) || (bytes == 0))
1188 + return;
1189 + poly1305_first_block(st, st->buffer);
1190 + st->leftover = 0;
1191 + }
1192 + st->started = 1;
1193 + }
1194 +
1195 + /* handle leftover */
1196 + if (st->leftover) {
1197 + want = poly1305_min(64 - st->leftover, bytes);
1198 + poly1305_block_copy(st->buffer + st->leftover, m, want);
1199 + bytes -= want;
1200 + m += want;
1201 + st->leftover += want;
1202 + if (st->leftover < 64)
1203 + return;
1204 + poly1305_blocks(st, st->buffer, 64);
1205 + st->leftover = 0;
1206 + }
1207 +
1208 + /* process 64 byte blocks */
1209 + if (bytes >= 64) {
1210 + want = (bytes & ~63);
1211 + poly1305_blocks(st, m, want);
1212 + m += want;
1213 + bytes -= want;
1214 + }
1215 +
1216 + if (bytes) {
1217 + poly1305_block_copy(st->buffer + st->leftover, m, bytes);
1218 + st->leftover += bytes;
1219 + }
1220 +}
1221 +
1222 +void
1223 +Poly1305Finish(poly1305_state *state, unsigned char mac[16]) {
1224 + poly1305_state_internal *st = poly1305_aligned_state(state);
1225 + size_t leftover = st->leftover;
1226 + uint8_t *m = st->buffer;
1227 + uint128_t d[3];
1228 + uint64_t h0,h1,h2;
1229 + uint64_t t0,t1;
1230 + uint64_t g0,g1,g2,c,nc;
1231 + uint64_t r0,r1,r2,s1,s2;
1232 + poly1305_power *p;
1233 +
1234 + if (st->started) {
1235 + size_t consumed = poly1305_combine(st, m, leftover);
1236 + leftover -= consumed;
1237 + m += consumed;
1238 + }
1239 +
1240 + /* st->HH will either be 0 or have the combined result */
1241 + h0 = st->HH[0];
1242 + h1 = st->HH[1];
1243 + h2 = st->HH[2];
1244 +
1245 + p = &st->P[1];
1246 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1];
1247 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1];
1248 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1];
1249 + s1 = r1 * (5 << 2);
1250 + s2 = r2 * (5 << 2);
1251 +
1252 + if (leftover < 16)
1253 + goto poly1305_donna_atmost15bytes;
1254 +
1255 +poly1305_donna_atleast16bytes:
1256 + t0 = U8TO64_LE(m + 0);
1257 + t1 = U8TO64_LE(m + 8);
1258 + h0 += t0 & 0xfffffffffff;
1259 + t0 = shr128_pair(t1, t0, 44);
1260 + h1 += t0 & 0xfffffffffff;
1261 + h2 += (t1 >> 24) | ((uint64_t)1 << 40);
1262 +
1263 +poly1305_donna_mul:
1264 + d[0] = add128(add128(mul64x64_128(h0, r0), mul64x64_128(h1, s2)), mul64x 64_128(h2, s1));
1265 + d[1] = add128(add128(mul64x64_128(h0, r1), mul64x64_128(h1, r0)), mul64x 64_128(h2, s2));
1266 + d[2] = add128(add128(mul64x64_128(h0, r2), mul64x64_128(h1, r1)), mul64x 64_128(h2, r0));
1267 + h0 = lo128(d[0]) & 0xfffffffffff; c = shr128( d[0], 44);
1268 + d[1] = add128_64(d[1], c); h1 = lo128(d[1]) & 0xfffffffffff; c = shr128( d[1], 44);
1269 + d[2] = add128_64(d[2], c); h2 = lo128(d[2]) & 0x3ffffffffff; c = shr128( d[2], 42);
1270 + h0 += c * 5;
1271 +
1272 + m += 16;
1273 + leftover -= 16;
1274 + if (leftover >= 16) goto poly1305_donna_atleast16bytes;
1275 +
1276 + /* final bytes */
1277 +poly1305_donna_atmost15bytes:
1278 + if (!leftover) goto poly1305_donna_finish;
1279 +
1280 + m[leftover++] = 1;
1281 + poly1305_block_zero(m + leftover, 16 - leftover);
1282 + leftover = 16;
1283 +
1284 + t0 = U8TO64_LE(m+0);
1285 + t1 = U8TO64_LE(m+8);
1286 + h0 += t0 & 0xfffffffffff; t0 = shr128_pair(t1, t0, 44);
1287 + h1 += t0 & 0xfffffffffff;
1288 + h2 += (t1 >> 24);
1289 +
1290 + goto poly1305_donna_mul;
1291 +
1292 +poly1305_donna_finish:
1293 + c = (h0 >> 44); h0 &= 0xfffffffffff;
1294 + h1 += c; c = (h1 >> 44); h1 &= 0xfffffffffff;
1295 + h2 += c; c = (h2 >> 42); h2 &= 0x3ffffffffff;
1296 + h0 += c * 5;
1297 +
1298 + g0 = h0 + 5; c = (g0 >> 44); g0 &= 0xfffffffffff;
1299 + g1 = h1 + c; c = (g1 >> 44); g1 &= 0xfffffffffff;
1300 + g2 = h2 + c - ((uint64_t)1 << 42);
1301 +
1302 + c = (g2 >> 63) - 1;
1303 + nc = ~c;
1304 + h0 = (h0 & nc) | (g0 & c);
1305 + h1 = (h1 & nc) | (g1 & c);
1306 + h2 = (h2 & nc) | (g2 & c);
1307 +
1308 + /* pad */
1309 + t0 = ((uint64_t)p->R23.d[3] << 32) | (uint64_t)p->R23.d[1];
1310 + t1 = ((uint64_t)p->R24.d[3] << 32) | (uint64_t)p->R24.d[1];
1311 + h0 += (t0 & 0xfffffffffff) ; c = (h0 >> 44); h0 &= 0xfffffffffff; t0 = shr128_pair(t1, t0, 44);
1312 + h1 += (t0 & 0xfffffffffff) + c; c = (h1 >> 44); h1 &= 0xfffffffffff; t1 = (t1 >> 24);
1313 + h2 += (t1 ) + c;
1314 +
1315 + U64TO8_LE(mac + 0, ((h0 ) | (h1 << 44)));
1316 + U64TO8_LE(mac + 8, ((h1 >> 20) | (h2 << 24)));
1317 +}
1318 diff --git a/lib/freebl/poly1305/poly1305.c b/lib/freebl/poly1305/poly1305.c
1319 new file mode 100644
1320 index 0000000..d86048a
1321 --- /dev/null
1322 +++ b/lib/freebl/poly1305/poly1305.c
1323 @@ -0,0 +1,254 @@
1324 +/* This Source Code Form is subject to the terms of the Mozilla Public
1325 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1326 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1327 +
1328 +/* This implementation of poly1305 is by Andrew Moon
1329 + * (https://github.com/floodyberry/poly1305-donna) and released as public
1330 + * domain. */
1331 +
1332 +#include <string.h>
1333 +#include <stdint.h>
1334 +
1335 +#include "poly1305.h"
1336 +
1337 +#if defined(NSS_X86) || defined(NSS_X64)
1338 +/* We can assume little-endian. */
1339 +static uint32_t U8TO32_LE(const unsigned char *m) {
1340 + uint32_t r;
1341 + memcpy(&r, m, sizeof(r));
1342 + return r;
1343 +}
1344 +
1345 +static void U32TO8_LE(unsigned char *m, uint32_t v) {
1346 + memcpy(m, &v, sizeof(v));
1347 +}
1348 +#else
1349 +static uint32_t U8TO32_LE(const unsigned char *m) {
1350 + return (uint32_t)m[0] |
1351 + (uint32_t)m[1] << 8 |
1352 + (uint32_t)m[2] << 16 |
1353 + (uint32_t)m[3] << 24;
1354 +}
1355 +
1356 +static void U32TO8_LE(unsigned char *m, uint32_t v) {
1357 + m[0] = v;
1358 + m[1] = v >> 8;
1359 + m[2] = v >> 16;
1360 + m[3] = v >> 24;
1361 +}
1362 +#endif
1363 +
1364 +static uint64_t
1365 +mul32x32_64(uint32_t a, uint32_t b) {
1366 + return (uint64_t)a * b;
1367 +}
1368 +
1369 +struct poly1305_state_st {
1370 + uint32_t r0,r1,r2,r3,r4;
1371 + uint32_t s1,s2,s3,s4;
1372 + uint32_t h0,h1,h2,h3,h4;
1373 + unsigned char buf[16];
1374 + unsigned int buf_used;
1375 + unsigned char key[16];
1376 +};
1377 +
1378 +/* update updates |state| given some amount of input data. This function may
1379 + * only be called with a |len| that is not a multiple of 16 at the end of the
1380 + * data. Otherwise the input must be buffered into 16 byte blocks. */
1381 +static void update(struct poly1305_state_st *state, const unsigned char *in,
1382 + size_t len) {
1383 + uint32_t t0,t1,t2,t3;
1384 + uint64_t t[5];
1385 + uint32_t b;
1386 + uint64_t c;
1387 + size_t j;
1388 + unsigned char mp[16];
1389 +
1390 + if (len < 16)
1391 + goto poly1305_donna_atmost15bytes;
1392 +
1393 +poly1305_donna_16bytes:
1394 + t0 = U8TO32_LE(in);
1395 + t1 = U8TO32_LE(in+4);
1396 + t2 = U8TO32_LE(in+8);
1397 + t3 = U8TO32_LE(in+12);
1398 +
1399 + in += 16;
1400 + len -= 16;
1401 +
1402 + state->h0 += t0 & 0x3ffffff;
1403 + state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
1404 + state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
1405 + state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
1406 + state->h4 += (t3 >> 8) | (1 << 24);
1407 +
1408 +poly1305_donna_mul:
1409 + t[0] = mul32x32_64(state->h0,state->r0) +
1410 + mul32x32_64(state->h1,state->s4) +
1411 + mul32x32_64(state->h2,state->s3) +
1412 + mul32x32_64(state->h3,state->s2) +
1413 + mul32x32_64(state->h4,state->s1);
1414 + t[1] = mul32x32_64(state->h0,state->r1) +
1415 + mul32x32_64(state->h1,state->r0) +
1416 + mul32x32_64(state->h2,state->s4) +
1417 + mul32x32_64(state->h3,state->s3) +
1418 + mul32x32_64(state->h4,state->s2);
1419 + t[2] = mul32x32_64(state->h0,state->r2) +
1420 + mul32x32_64(state->h1,state->r1) +
1421 + mul32x32_64(state->h2,state->r0) +
1422 + mul32x32_64(state->h3,state->s4) +
1423 + mul32x32_64(state->h4,state->s3);
1424 + t[3] = mul32x32_64(state->h0,state->r3) +
1425 + mul32x32_64(state->h1,state->r2) +
1426 + mul32x32_64(state->h2,state->r1) +
1427 + mul32x32_64(state->h3,state->r0) +
1428 + mul32x32_64(state->h4,state->s4);
1429 + t[4] = mul32x32_64(state->h0,state->r4) +
1430 + mul32x32_64(state->h1,state->r3) +
1431 + mul32x32_64(state->h2,state->r2) +
1432 + mul32x32_64(state->h3,state->r1) +
1433 + mul32x32_64(state->h4,state->r0);
1434 +
1435 + state->h0 = (uint32_t)t[0] & 0x3ffffff; c = (t[0] > > 26);
1436 + t[1] += c; state->h1 = (uint32_t)t[1] & 0x3ffffff; b = (uint32_t)(t[1] > > 26);
1437 + t[2] += b; state->h2 = (uint32_t)t[2] & 0x3ffffff; b = (uint32_t)(t[2] > > 26);
1438 + t[3] += b; state->h3 = (uint32_t)t[3] & 0x3ffffff; b = (uint32_t)(t[3] > > 26);
1439 + t[4] += b; state->h4 = (uint32_t)t[4] & 0x3ffffff; b = (uint32_t)(t[4] > > 26);
1440 + state->h0 += b * 5;
1441 +
1442 + if (len >= 16)
1443 + goto poly1305_donna_16bytes;
1444 +
1445 + /* final bytes */
1446 +poly1305_donna_atmost15bytes:
1447 + if (!len)
1448 + return;
1449 +
1450 + for (j = 0; j < len; j++)
1451 + mp[j] = in[j];
1452 + mp[j++] = 1;
1453 + for (; j < 16; j++)
1454 + mp[j] = 0;
1455 + len = 0;
1456 +
1457 + t0 = U8TO32_LE(mp+0);
1458 + t1 = U8TO32_LE(mp+4);
1459 + t2 = U8TO32_LE(mp+8);
1460 + t3 = U8TO32_LE(mp+12);
1461 +
1462 + state->h0 += t0 & 0x3ffffff;
1463 + state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
1464 + state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
1465 + state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
1466 + state->h4 += (t3 >> 8);
1467 +
1468 + goto poly1305_donna_mul;
1469 +}
1470 +
1471 +void Poly1305Init(poly1305_state *statep, const unsigned char key[32]) {
1472 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep;
1473 + uint32_t t0,t1,t2,t3;
1474 +
1475 + t0 = U8TO32_LE(key+0);
1476 + t1 = U8TO32_LE(key+4);
1477 + t2 = U8TO32_LE(key+8);
1478 + t3 = U8TO32_LE(key+12);
1479 +
1480 + /* precompute multipliers */
1481 + state->r0 = t0 & 0x3ffffff; t0 >>= 26; t0 |= t1 << 6;
1482 + state->r1 = t0 & 0x3ffff03; t1 >>= 20; t1 |= t2 << 12;
1483 + state->r2 = t1 & 0x3ffc0ff; t2 >>= 14; t2 |= t3 << 18;
1484 + state->r3 = t2 & 0x3f03fff; t3 >>= 8;
1485 + state->r4 = t3 & 0x00fffff;
1486 +
1487 + state->s1 = state->r1 * 5;
1488 + state->s2 = state->r2 * 5;
1489 + state->s3 = state->r3 * 5;
1490 + state->s4 = state->r4 * 5;
1491 +
1492 + /* init state */
1493 + state->h0 = 0;
1494 + state->h1 = 0;
1495 + state->h2 = 0;
1496 + state->h3 = 0;
1497 + state->h4 = 0;
1498 +
1499 + state->buf_used = 0;
1500 + memcpy(state->key, key + 16, sizeof(state->key));
1501 +}
1502 +
1503 +void Poly1305Update(poly1305_state *statep, const unsigned char *in,
1504 + size_t in_len) {
1505 + unsigned int i;
1506 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep;
1507 +
1508 + if (state->buf_used) {
1509 + unsigned int todo = 16 - state->buf_used;
1510 + if (todo > in_len)
1511 + todo = in_len;
1512 + for (i = 0; i < todo; i++)
1513 + state->buf[state->buf_used + i] = in[i];
1514 + state->buf_used += todo;
1515 + in_len -= todo;
1516 + in += todo;
1517 +
1518 + if (state->buf_used == 16) {
1519 + update(state, state->buf, 16);
1520 + state->buf_used = 0;
1521 + }
1522 + }
1523 +
1524 + if (in_len >= 16) {
1525 + size_t todo = in_len & ~0xf;
1526 + update(state, in, todo);
1527 + in += todo;
1528 + in_len &= 0xf;
1529 + }
1530 +
1531 + if (in_len) {
1532 + for (i = 0; i < in_len; i++)
1533 + state->buf[i] = in[i];
1534 + state->buf_used = in_len;
1535 + }
1536 +}
1537 +
1538 +void Poly1305Finish(poly1305_state *statep, unsigned char mac[16]) {
1539 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep;
1540 + uint64_t f0,f1,f2,f3;
1541 + uint32_t g0,g1,g2,g3,g4;
1542 + uint32_t b, nb;
1543 +
1544 + if (state->buf_used)
1545 + update(state, state->buf, state->buf_used);
1546 +
1547 + b = state->h0 >> 26; state->h0 = state->h0 & 0x3ffff ff;
1548 + state->h1 += b; b = state->h1 >> 26; state->h1 = state->h1 & 0x3ffff ff;
1549 + state->h2 += b; b = state->h2 >> 26; state->h2 = state->h2 & 0x3ffff ff;
1550 + state->h3 += b; b = state->h3 >> 26; state->h3 = state->h3 & 0x3ffff ff;
1551 + state->h4 += b; b = state->h4 >> 26; state->h4 = state->h4 & 0x3ffff ff;
1552 + state->h0 += b * 5;
1553 +
1554 + g0 = state->h0 + 5; b = g0 >> 26; g0 &= 0x3ffffff;
1555 + g1 = state->h1 + b; b = g1 >> 26; g1 &= 0x3ffffff;
1556 + g2 = state->h2 + b; b = g2 >> 26; g2 &= 0x3ffffff;
1557 + g3 = state->h3 + b; b = g3 >> 26; g3 &= 0x3ffffff;
1558 + g4 = state->h4 + b - (1 << 26);
1559 +
1560 + b = (g4 >> 31) - 1;
1561 + nb = ~b;
1562 + state->h0 = (state->h0 & nb) | (g0 & b);
1563 + state->h1 = (state->h1 & nb) | (g1 & b);
1564 + state->h2 = (state->h2 & nb) | (g2 & b);
1565 + state->h3 = (state->h3 & nb) | (g3 & b);
1566 + state->h4 = (state->h4 & nb) | (g4 & b);
1567 +
1568 + f0 = ((state->h0 ) | (state->h1 << 26)) + (uint64_t)U8TO32_LE(&stat e->key[0]);
1569 + f1 = ((state->h1 >> 6) | (state->h2 << 20)) + (uint64_t)U8TO32_LE(&stat e->key[4]);
1570 + f2 = ((state->h2 >> 12) | (state->h3 << 14)) + (uint64_t)U8TO32_LE(&stat e->key[8]);
1571 + f3 = ((state->h3 >> 18) | (state->h4 << 8)) + (uint64_t)U8TO32_LE(&stat e->key[12]);
1572 +
1573 + U32TO8_LE(&mac[ 0], (uint32_t)f0); f1 += (f0 >> 32);
1574 + U32TO8_LE(&mac[ 4], (uint32_t)f1); f2 += (f1 >> 32);
1575 + U32TO8_LE(&mac[ 8], (uint32_t)f2); f3 += (f2 >> 32);
1576 + U32TO8_LE(&mac[12], (uint32_t)f3);
1577 +}
1578 diff --git a/lib/freebl/poly1305/poly1305.h b/lib/freebl/poly1305/poly1305.h
1579 new file mode 100644
1580 index 0000000..4beb172
1581 --- /dev/null
1582 +++ b/lib/freebl/poly1305/poly1305.h
1583 @@ -0,0 +1,31 @@
1584 +/*
1585 + * poly1305.h - header file for Poly1305 implementation.
1586 + *
1587 + * This Source Code Form is subject to the terms of the Mozilla Public
1588 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1589 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1590 +
1591 +#ifndef FREEBL_POLY1305_H_
1592 +#define FREEBL_POLY1305_H_
1593 +
1594 +typedef unsigned char poly1305_state[512];
1595 +
1596 +/* Poly1305Init sets up |state| so that it can be used to calculate an
1597 + * authentication tag with the one-time key |key|. Note that |key| is a
1598 + * one-time key and therefore there is no `reset' method because that would
1599 + * enable several messages to be authenticated with the same key. */
1600 +extern void Poly1305Init(poly1305_state* state,
1601 + const unsigned char key[32]);
1602 +
1603 +/* Poly1305Update processes |in_len| bytes from |in|. It can be called zero or
1604 + * more times after poly1305_init. */
1605 +extern void Poly1305Update(poly1305_state* state,
1606 + const unsigned char *in,
1607 + size_t inLen);
1608 +
1609 +/* Poly1305Finish completes the poly1305 calculation and writes a 16 byte
1610 + * authentication tag to |mac|. */
1611 +extern void Poly1305Finish(poly1305_state* state,
1612 + unsigned char mac[16]);
1613 +
1614 +#endif /* FREEBL_POLY1305_H_ */
1615 diff --git a/lib/pk11wrap/pk11mech.c b/lib/pk11wrap/pk11mech.c
1616 index 29e86e6..0ebb075 100644
1617 --- a/lib/pk11wrap/pk11mech.c
1618 +++ b/lib/pk11wrap/pk11mech.c
1619 @@ -152,6 +152,8 @@ PK11_GetKeyMechanism(CK_KEY_TYPE type)
1620 return CKM_SEED_CBC;
1621 case CKK_CAMELLIA:
1622 return CKM_CAMELLIA_CBC;
1623 + case CKK_NSS_CHACHA20:
1624 + return CKM_NSS_CHACHA20_POLY1305;
1625 case CKK_AES:
1626 return CKM_AES_CBC;
1627 case CKK_DES:
1628 @@ -219,6 +221,8 @@ PK11_GetKeyType(CK_MECHANISM_TYPE type,unsigned long len)
1629 case CKM_CAMELLIA_CBC_PAD:
1630 case CKM_CAMELLIA_KEY_GEN:
1631 return CKK_CAMELLIA;
1632 + case CKM_NSS_CHACHA20_POLY1305:
1633 + return CKK_NSS_CHACHA20;
1634 case CKM_AES_ECB:
1635 case CKM_AES_CBC:
1636 case CKM_AES_CCM:
1637 @@ -431,6 +435,8 @@ PK11_GetKeyGenWithSize(CK_MECHANISM_TYPE type, int size)
1638 case CKM_CAMELLIA_CBC_PAD:
1639 case CKM_CAMELLIA_KEY_GEN:
1640 return CKM_CAMELLIA_KEY_GEN;
1641 + case CKM_NSS_CHACHA20_POLY1305:
1642 + return CKM_NSS_CHACHA20_KEY_GEN;
1643 case CKM_AES_ECB:
1644 case CKM_AES_CBC:
1645 case CKM_AES_CCM:
1646 diff --git a/lib/softoken/pkcs11.c b/lib/softoken/pkcs11.c
1647 index 97d6d3f..75c9e8e 100644
1648 --- a/lib/softoken/pkcs11.c
1649 +++ b/lib/softoken/pkcs11.c
1650 @@ -370,6 +370,9 @@ static const struct mechanismList mechanisms[] = {
1651 {CKM_SEED_MAC, {16, 16, CKF_SN_VR}, PR_TRUE},
1652 {CKM_SEED_MAC_GENERAL, {16, 16, CKF_SN_VR}, PR_TRUE},
1653 {CKM_SEED_CBC_PAD, {16, 16, CKF_EN_DE_WR_UN}, PR_TRUE} ,
1654 + /* ------------------------- ChaCha20 Operations ---------------------- */
1655 + {CKM_NSS_CHACHA20_KEY_GEN, {32, 32, CKF_GENERATE}, PR_TRUE} ,
1656 + {CKM_NSS_CHACHA20_POLY1305,{32, 32, CKF_EN_DE}, PR_TRUE},
1657 /* ------------------------- Hashing Operations ----------------------- */
1658 {CKM_MD2, {0, 0, CKF_DIGEST}, PR_FALSE},
1659 {CKM_MD2_HMAC, {1, 128, CKF_SN_VR}, PR_TRUE},
1660 diff --git a/lib/softoken/pkcs11c.c b/lib/softoken/pkcs11c.c
1661 index 8755f24..992fba4 100644
1662 --- a/lib/softoken/pkcs11c.c
1663 +++ b/lib/softoken/pkcs11c.c
1664 @@ -664,6 +664,97 @@ sftk_RSADecryptOAEP(SFTKOAEPDecryptInfo *info, unsigned cha r *output,
1665 return rv;
1666 }
1667
1668 +static SFTKChaCha20Poly1305Info *
1669 +sftk_ChaCha20Poly1305_CreateContext(const unsigned char *key,
1670 + unsigned int keyLen,
1671 + const CK_NSS_AEAD_PARAMS* params)
1672 +{
1673 + SFTKChaCha20Poly1305Info *ctx;
1674 +
1675 + if (params->ulIvLen != sizeof(ctx->nonce)) {
1676 + PORT_SetError(SEC_ERROR_INPUT_LEN);
1677 + return NULL;
1678 + }
1679 +
1680 + ctx = PORT_New(SFTKChaCha20Poly1305Info);
1681 + if (ctx == NULL) {
1682 + return NULL;
1683 + }
1684 +
1685 + if (ChaCha20Poly1305_InitContext(&ctx->freeblCtx, key, keyLen,
1686 + params->ulTagLen) != SECSuccess) {
1687 + PORT_Free(ctx);
1688 + return NULL;
1689 + }
1690 +
1691 + memcpy(ctx->nonce, params->pIv, sizeof(ctx->nonce));
1692 +
1693 + if (params->ulAADLen > sizeof(ctx->ad)) {
1694 + /* Need to allocate an overflow buffer for the additional data. */
1695 + ctx->adOverflow = (unsigned char *)PORT_Alloc(params->ulAADLen);
1696 + if (!ctx->adOverflow) {
1697 + PORT_Free(ctx);
1698 + return NULL;
1699 + }
1700 + memcpy(ctx->adOverflow, params->pAAD, params->ulAADLen);
1701 + } else {
1702 + ctx->adOverflow = NULL;
1703 + memcpy(ctx->ad, params->pAAD, params->ulAADLen);
1704 + }
1705 + ctx->adLen = params->ulAADLen;
1706 +
1707 + return ctx;
1708 +}
1709 +
1710 +static void
1711 +sftk_ChaCha20Poly1305_DestroyContext(SFTKChaCha20Poly1305Info *ctx,
1712 + PRBool freeit)
1713 +{
1714 + ChaCha20Poly1305_DestroyContext(&ctx->freeblCtx, PR_FALSE);
1715 + if (ctx->adOverflow != NULL) {
1716 + PORT_Free(ctx->adOverflow);
1717 + ctx->adOverflow = NULL;
1718 + }
1719 + ctx->adLen = 0;
1720 + if (freeit) {
1721 + PORT_Free(ctx);
1722 + }
1723 +}
1724 +
1725 +static SECStatus
1726 +sftk_ChaCha20Poly1305_Encrypt(const SFTKChaCha20Poly1305Info *ctx,
1727 + unsigned char *output, unsigned int *outputLen,
1728 + unsigned int maxOutputLen,
1729 + const unsigned char *input, unsigned int inputLen)
1730 +{
1731 + const unsigned char *ad = ctx->adOverflow;
1732 +
1733 + if (ad == NULL) {
1734 + ad = ctx->ad;
1735 + }
1736 +
1737 + return ChaCha20Poly1305_Seal(&ctx->freeblCtx, output, outputLen,
1738 + maxOutputLen, input, inputLen, ctx->nonce,
1739 + sizeof(ctx->nonce), ad, ctx->adLen);
1740 +}
1741 +
1742 +static SECStatus
1743 +sftk_ChaCha20Poly1305_Decrypt(const SFTKChaCha20Poly1305Info *ctx,
1744 + unsigned char *output, unsigned int *outputLen,
1745 + unsigned int maxOutputLen,
1746 + const unsigned char *input, unsigned int inputLen)
1747 +{
1748 + const unsigned char *ad = ctx->adOverflow;
1749 +
1750 + if (ad == NULL) {
1751 + ad = ctx->ad;
1752 + }
1753 +
1754 + return ChaCha20Poly1305_Open(&ctx->freeblCtx, output, outputLen,
1755 + maxOutputLen, input, inputLen, ctx->nonce,
1756 + sizeof(ctx->nonce), ad, ctx->adLen);
1757 +}
1758 +
1759 /** NSC_CryptInit initializes an encryption/Decryption operation.
1760 *
1761 * Always called by NSC_EncryptInit, NSC_DecryptInit, NSC_WrapKey,NSC_UnwrapKey .
1762 @@ -1057,6 +1148,35 @@ finish_des:
1763 context->destroy = (SFTKDestroy) AES_DestroyContext;
1764 break;
1765
1766 + case CKM_NSS_CHACHA20_POLY1305:
1767 + if (pMechanism->ulParameterLen != sizeof(CK_NSS_AEAD_PARAMS)) {
1768 + crv = CKR_MECHANISM_PARAM_INVALID;
1769 + break;
1770 + }
1771 + context->multi = PR_FALSE;
1772 + if (key_type != CKK_NSS_CHACHA20) {
1773 + crv = CKR_KEY_TYPE_INCONSISTENT;
1774 + break;
1775 + }
1776 + att = sftk_FindAttribute(key,CKA_VALUE);
1777 + if (att == NULL) {
1778 + crv = CKR_KEY_HANDLE_INVALID;
1779 + break;
1780 + }
1781 + context->cipherInfo = sftk_ChaCha20Poly1305_CreateContext(
1782 + (unsigned char*) att->attrib.pValue, att->attrib.ulValueLen,
1783 + (CK_NSS_AEAD_PARAMS*) pMechanism->pParameter);
1784 + sftk_FreeAttribute(att);
1785 + if (context->cipherInfo == NULL) {
1786 + crv = sftk_MapCryptError(PORT_GetError());
1787 + break;
1788 + }
1789 + context->update = (SFTKCipher) (isEncrypt ?
1790 + sftk_ChaCha20Poly1305_Encrypt :
1791 + sftk_ChaCha20Poly1305_Decrypt);
1792 + context->destroy = (SFTKDestroy) sftk_ChaCha20Poly1305_DestroyContext;
1793 + break;
1794 +
1795 case CKM_NETSCAPE_AES_KEY_WRAP_PAD:
1796 context->doPad = PR_TRUE;
1797 /* fall thru */
1798 @@ -3654,6 +3774,10 @@ nsc_SetupBulkKeyGen(CK_MECHANISM_TYPE mechanism, CK_KEY_T YPE *key_type,
1799 *key_type = CKK_AES;
1800 if (*key_length == 0) crv = CKR_TEMPLATE_INCOMPLETE;
1801 break;
1802 + case CKM_NSS_CHACHA20_KEY_GEN:
1803 + *key_type = CKK_NSS_CHACHA20;
1804 + if (*key_length == 0) crv = CKR_TEMPLATE_INCOMPLETE;
1805 + break;
1806 default:
1807 PORT_Assert(0);
1808 crv = CKR_MECHANISM_INVALID;
1809 @@ -3900,6 +4024,7 @@ CK_RV NSC_GenerateKey(CK_SESSION_HANDLE hSession,
1810 case CKM_SEED_KEY_GEN:
1811 case CKM_CAMELLIA_KEY_GEN:
1812 case CKM_AES_KEY_GEN:
1813 + case CKM_NSS_CHACHA20_KEY_GEN:
1814 #if NSS_SOFTOKEN_DOES_RC5
1815 case CKM_RC5_KEY_GEN:
1816 #endif
1817 diff --git a/lib/softoken/pkcs11i.h b/lib/softoken/pkcs11i.h
1818 index 1023a00..4e8601b 100644
1819 --- a/lib/softoken/pkcs11i.h
1820 +++ b/lib/softoken/pkcs11i.h
1821 @@ -14,6 +14,7 @@
1822 #include "pkcs11t.h"
1823
1824 #include "sftkdbt.h"
1825 +#include "chacha20poly1305.h"
1826 #include "hasht.h"
1827
1828 /*
1829 @@ -104,6 +105,7 @@ typedef struct SFTKHashSignInfoStr SFTKHashSignInfo;
1830 typedef struct SFTKOAEPEncryptInfoStr SFTKOAEPEncryptInfo;
1831 typedef struct SFTKOAEPDecryptInfoStr SFTKOAEPDecryptInfo;
1832 typedef struct SFTKSSLMACInfoStr SFTKSSLMACInfo;
1833 +typedef struct SFTKChaCha20Poly1305InfoStr SFTKChaCha20Poly1305Info;
1834 typedef struct SFTKItemTemplateStr SFTKItemTemplate;
1835
1836 /* define function pointer typdefs for pointer tables */
1837 @@ -399,6 +401,16 @@ struct SFTKSSLMACInfoStr {
1838 unsigned int keySize;
1839 };
1840
1841 +/* SFTKChaCha20Poly1305Info saves the key, tag length, nonce, and additional
1842 + * data for a ChaCha20+Poly1305 AEAD operation. */
1843 +struct SFTKChaCha20Poly1305InfoStr {
1844 + ChaCha20Poly1305Context freeblCtx;
1845 + unsigned char nonce[8];
1846 + unsigned char ad[16];
1847 + unsigned char *adOverflow;
1848 + unsigned int adLen;
1849 +};
1850 +
1851 /*
1852 * Template based on SECItems, suitable for passing as arrays
1853 */
1854 diff --git a/lib/util/pkcs11n.h b/lib/util/pkcs11n.h
1855 index 5e13784..86a396f 100644
1856 --- a/lib/util/pkcs11n.h
1857 +++ b/lib/util/pkcs11n.h
1858 @@ -51,6 +51,8 @@
1859 #define CKK_NSS_JPAKE_ROUND1 (CKK_NSS + 2)
1860 #define CKK_NSS_JPAKE_ROUND2 (CKK_NSS + 3)
1861
1862 +#define CKK_NSS_CHACHA20 (CKK_NSS + 4)
1863 +
1864 /*
1865 * NSS-defined certificate types
1866 *
1867 @@ -218,6 +220,9 @@
1868 #define CKM_NSS_TLS_EXTENDED_MASTER_KEY_DERIVE (CKM_NSS + 25)
1869 #define CKM_NSS_TLS_EXTENDED_MASTER_KEY_DERIVE_DH (CKM_NSS + 26)
1870
1871 +#define CKM_NSS_CHACHA20_KEY_GEN (CKM_NSS + 27)
1872 +#define CKM_NSS_CHACHA20_POLY1305 (CKM_NSS + 28)
1873 +
1874 /*
1875 * HISTORICAL:
1876 * Do not attempt to use these. They are only used by NETSCAPE's internal
1877 @@ -285,6 +290,14 @@ typedef struct CK_NSS_MAC_CONSTANT_TIME_PARAMS {
1878 CK_ULONG ulHeaderLen; /* in */
1879 } CK_NSS_MAC_CONSTANT_TIME_PARAMS;
1880
1881 +typedef struct CK_NSS_AEAD_PARAMS {
1882 + CK_BYTE_PTR pIv; /* This is the nonce. */
1883 + CK_ULONG ulIvLen;
1884 + CK_BYTE_PTR pAAD;
1885 + CK_ULONG ulAADLen;
1886 + CK_ULONG ulTagLen;
1887 +} CK_NSS_AEAD_PARAMS;
1888 +
1889 /*
1890 * NSS-defined return values
1891 *
OLDNEW
« nss/lib/util/secoid.c ('K') | « patches/nspr-unused-variable.patch ('k') | patches/nss-static.patch » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698