Line data Source code
1 : /* ====================================================================
2 : * Copyright (c) 2011-2013 The OpenSSL Project. All rights reserved.
3 : *
4 : * Redistribution and use in source and binary forms, with or without
5 : * modification, are permitted provided that the following conditions
6 : * are met:
7 : *
8 : * 1. Redistributions of source code must retain the above copyright
9 : * notice, this list of conditions and the following disclaimer.
10 : *
11 : * 2. Redistributions in binary form must reproduce the above copyright
12 : * notice, this list of conditions and the following disclaimer in
13 : * the documentation and/or other materials provided with the
14 : * distribution.
15 : *
16 : * 3. All advertising materials mentioning features or use of this
17 : * software must display the following acknowledgment:
18 : * "This product includes software developed by the OpenSSL Project
19 : * for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)"
20 : *
21 : * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
22 : * endorse or promote products derived from this software without
23 : * prior written permission. For written permission, please contact
24 : * licensing@OpenSSL.org.
25 : *
26 : * 5. Products derived from this software may not be called "OpenSSL"
27 : * nor may "OpenSSL" appear in their names without prior written
28 : * permission of the OpenSSL Project.
29 : *
30 : * 6. Redistributions of any form whatsoever must retain the following
31 : * acknowledgment:
32 : * "This product includes software developed by the OpenSSL Project
33 : * for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)"
34 : *
35 : * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
36 : * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
37 : * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
38 : * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR
39 : * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
40 : * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
41 : * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
42 : * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
43 : * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
44 : * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
45 : * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
46 : * OF THE POSSIBILITY OF SUCH DAMAGE.
47 : * ====================================================================
48 : */
49 :
50 : #include <openssl/opensslconf.h>
51 :
52 : #include <stdio.h>
53 : #include <string.h>
54 :
55 : #if !defined(OPENSSL_NO_AES) && !defined(OPENSSL_NO_SHA256)
56 :
57 : # include <openssl/evp.h>
58 : # include <openssl/objects.h>
59 : # include <openssl/aes.h>
60 : # include <openssl/sha.h>
61 : # include <openssl/rand.h>
62 : # include "modes_lcl.h"
63 :
64 : # ifndef EVP_CIPH_FLAG_AEAD_CIPHER
65 : # define EVP_CIPH_FLAG_AEAD_CIPHER 0x200000
66 : # define EVP_CTRL_AEAD_TLS1_AAD 0x16
67 : # define EVP_CTRL_AEAD_SET_MAC_KEY 0x17
68 : # endif
69 :
70 : # if !defined(EVP_CIPH_FLAG_DEFAULT_ASN1)
71 : # define EVP_CIPH_FLAG_DEFAULT_ASN1 0
72 : # endif
73 :
74 : # if !defined(EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK)
75 : # define EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK 0
76 : # endif
77 :
78 : # define TLS1_1_VERSION 0x0302
79 :
80 : typedef struct {
81 : AES_KEY ks;
82 : SHA256_CTX head, tail, md;
83 : size_t payload_length; /* AAD length in decrypt case */
84 : union {
85 : unsigned int tls_ver;
86 : unsigned char tls_aad[16]; /* 13 used */
87 : } aux;
88 : } EVP_AES_HMAC_SHA256;
89 :
90 : # define NO_PAYLOAD_LENGTH ((size_t)-1)
91 :
92 : # if defined(AES_ASM) && ( \
93 : defined(__x86_64) || defined(__x86_64__) || \
94 : defined(_M_AMD64) || defined(_M_X64) || \
95 : defined(__INTEL__) )
96 :
97 : extern unsigned int OPENSSL_ia32cap_P[];
98 : # define AESNI_CAPABLE (1<<(57-32))
99 :
100 : int aesni_set_encrypt_key(const unsigned char *userKey, int bits,
101 : AES_KEY *key);
102 : int aesni_set_decrypt_key(const unsigned char *userKey, int bits,
103 : AES_KEY *key);
104 :
105 : void aesni_cbc_encrypt(const unsigned char *in,
106 : unsigned char *out,
107 : size_t length,
108 : const AES_KEY *key, unsigned char *ivec, int enc);
109 :
110 : int aesni_cbc_sha256_enc(const void *inp, void *out, size_t blocks,
111 : const AES_KEY *key, unsigned char iv[16],
112 : SHA256_CTX *ctx, const void *in0);
113 :
114 : # define data(ctx) ((EVP_AES_HMAC_SHA256 *)(ctx)->cipher_data)
115 :
116 : static int aesni_cbc_hmac_sha256_init_key(EVP_CIPHER_CTX *ctx,
117 : const unsigned char *inkey,
118 : const unsigned char *iv, int enc)
119 : {
120 : EVP_AES_HMAC_SHA256 *key = data(ctx);
121 : int ret;
122 :
123 : if (enc)
124 : memset(&key->ks, 0, sizeof(key->ks.rd_key)),
125 : ret = aesni_set_encrypt_key(inkey, ctx->key_len * 8, &key->ks);
126 : else
127 : ret = aesni_set_decrypt_key(inkey, ctx->key_len * 8, &key->ks);
128 :
129 : SHA256_Init(&key->head); /* handy when benchmarking */
130 : key->tail = key->head;
131 : key->md = key->head;
132 :
133 : key->payload_length = NO_PAYLOAD_LENGTH;
134 :
135 : return ret < 0 ? 0 : 1;
136 : }
137 :
138 : # define STITCHED_CALL
139 :
140 : # if !defined(STITCHED_CALL)
141 : # define aes_off 0
142 : # endif
143 :
144 : void sha256_block_data_order(void *c, const void *p, size_t len);
145 :
146 : static void sha256_update(SHA256_CTX *c, const void *data, size_t len)
147 : {
148 : const unsigned char *ptr = data;
149 : size_t res;
150 :
151 : if ((res = c->num)) {
152 : res = SHA256_CBLOCK - res;
153 : if (len < res)
154 : res = len;
155 : SHA256_Update(c, ptr, res);
156 : ptr += res;
157 : len -= res;
158 : }
159 :
160 : res = len % SHA256_CBLOCK;
161 : len -= res;
162 :
163 : if (len) {
164 : sha256_block_data_order(c, ptr, len / SHA256_CBLOCK);
165 :
166 : ptr += len;
167 : c->Nh += len >> 29;
168 : c->Nl += len <<= 3;
169 : if (c->Nl < (unsigned int)len)
170 : c->Nh++;
171 : }
172 :
173 : if (res)
174 : SHA256_Update(c, ptr, res);
175 : }
176 :
177 : # ifdef SHA256_Update
178 : # undef SHA256_Update
179 : # endif
180 : # define SHA256_Update sha256_update
181 :
182 : # if !defined(OPENSSL_NO_MULTIBLOCK) && EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK
183 :
184 : typedef struct {
185 : unsigned int A[8], B[8], C[8], D[8], E[8], F[8], G[8], H[8];
186 : } SHA256_MB_CTX;
187 : typedef struct {
188 : const unsigned char *ptr;
189 : int blocks;
190 : } HASH_DESC;
191 :
192 : void sha256_multi_block(SHA256_MB_CTX *, const HASH_DESC *, int);
193 :
194 : typedef struct {
195 : const unsigned char *inp;
196 : unsigned char *out;
197 : int blocks;
198 : u64 iv[2];
199 : } CIPH_DESC;
200 :
201 : void aesni_multi_cbc_encrypt(CIPH_DESC *, void *, int);
202 :
203 : static size_t tls1_1_multi_block_encrypt(EVP_AES_HMAC_SHA256 *key,
204 : unsigned char *out,
205 : const unsigned char *inp,
206 : size_t inp_len, int n4x)
207 : { /* n4x is 1 or 2 */
208 : HASH_DESC hash_d[8], edges[8];
209 : CIPH_DESC ciph_d[8];
210 : unsigned char storage[sizeof(SHA256_MB_CTX) + 32];
211 : union {
212 : u64 q[16];
213 : u32 d[32];
214 : u8 c[128];
215 : } blocks[8];
216 : SHA256_MB_CTX *ctx;
217 : unsigned int frag, last, packlen, i, x4 = 4 * n4x, minblocks, processed =
218 : 0;
219 : size_t ret = 0;
220 : u8 *IVs;
221 : # if defined(BSWAP8)
222 : u64 seqnum;
223 : # endif
224 :
225 : /* ask for IVs in bulk */
226 : if (RAND_bytes((IVs = blocks[0].c), 16 * x4) <= 0)
227 : return 0;
228 :
229 : /* align */
230 : ctx = (SHA256_MB_CTX *) (storage + 32 - ((size_t)storage % 32));
231 :
232 : frag = (unsigned int)inp_len >> (1 + n4x);
233 : last = (unsigned int)inp_len + frag - (frag << (1 + n4x));
234 : if (last > frag && ((last + 13 + 9) % 64) < (x4 - 1)) {
235 : frag++;
236 : last -= x4 - 1;
237 : }
238 :
239 : packlen = 5 + 16 + ((frag + 32 + 16) & -16);
240 :
241 : /* populate descriptors with pointers and IVs */
242 : hash_d[0].ptr = inp;
243 : ciph_d[0].inp = inp;
244 : /* 5+16 is place for header and explicit IV */
245 : ciph_d[0].out = out + 5 + 16;
246 : memcpy(ciph_d[0].out - 16, IVs, 16);
247 : memcpy(ciph_d[0].iv, IVs, 16);
248 : IVs += 16;
249 :
250 : for (i = 1; i < x4; i++) {
251 : ciph_d[i].inp = hash_d[i].ptr = hash_d[i - 1].ptr + frag;
252 : ciph_d[i].out = ciph_d[i - 1].out + packlen;
253 : memcpy(ciph_d[i].out - 16, IVs, 16);
254 : memcpy(ciph_d[i].iv, IVs, 16);
255 : IVs += 16;
256 : }
257 :
258 : # if defined(BSWAP8)
259 : memcpy(blocks[0].c, key->md.data, 8);
260 : seqnum = BSWAP8(blocks[0].q[0]);
261 : # endif
262 : for (i = 0; i < x4; i++) {
263 : unsigned int len = (i == (x4 - 1) ? last : frag);
264 : # if !defined(BSWAP8)
265 : unsigned int carry, j;
266 : # endif
267 :
268 : ctx->A[i] = key->md.h[0];
269 : ctx->B[i] = key->md.h[1];
270 : ctx->C[i] = key->md.h[2];
271 : ctx->D[i] = key->md.h[3];
272 : ctx->E[i] = key->md.h[4];
273 : ctx->F[i] = key->md.h[5];
274 : ctx->G[i] = key->md.h[6];
275 : ctx->H[i] = key->md.h[7];
276 :
277 : /* fix seqnum */
278 : # if defined(BSWAP8)
279 : blocks[i].q[0] = BSWAP8(seqnum + i);
280 : # else
281 : for (carry = i, j = 8; j--;) {
282 : blocks[i].c[j] = ((u8 *)key->md.data)[j] + carry;
283 : carry = (blocks[i].c[j] - carry) >> (sizeof(carry) * 8 - 1);
284 : }
285 : # endif
286 : blocks[i].c[8] = ((u8 *)key->md.data)[8];
287 : blocks[i].c[9] = ((u8 *)key->md.data)[9];
288 : blocks[i].c[10] = ((u8 *)key->md.data)[10];
289 : /* fix length */
290 : blocks[i].c[11] = (u8)(len >> 8);
291 : blocks[i].c[12] = (u8)(len);
292 :
293 : memcpy(blocks[i].c + 13, hash_d[i].ptr, 64 - 13);
294 : hash_d[i].ptr += 64 - 13;
295 : hash_d[i].blocks = (len - (64 - 13)) / 64;
296 :
297 : edges[i].ptr = blocks[i].c;
298 : edges[i].blocks = 1;
299 : }
300 :
301 : /* hash 13-byte headers and first 64-13 bytes of inputs */
302 : sha256_multi_block(ctx, edges, n4x);
303 : /* hash bulk inputs */
304 : # define MAXCHUNKSIZE 2048
305 : # if MAXCHUNKSIZE%64
306 : # error "MAXCHUNKSIZE is not divisible by 64"
307 : # elif MAXCHUNKSIZE
308 : /*
309 : * goal is to minimize pressure on L1 cache by moving in shorter steps,
310 : * so that hashed data is still in the cache by the time we encrypt it
311 : */
312 : minblocks = ((frag <= last ? frag : last) - (64 - 13)) / 64;
313 : if (minblocks > MAXCHUNKSIZE / 64) {
314 : for (i = 0; i < x4; i++) {
315 : edges[i].ptr = hash_d[i].ptr;
316 : edges[i].blocks = MAXCHUNKSIZE / 64;
317 : ciph_d[i].blocks = MAXCHUNKSIZE / 16;
318 : }
319 : do {
320 : sha256_multi_block(ctx, edges, n4x);
321 : aesni_multi_cbc_encrypt(ciph_d, &key->ks, n4x);
322 :
323 : for (i = 0; i < x4; i++) {
324 : edges[i].ptr = hash_d[i].ptr += MAXCHUNKSIZE;
325 : hash_d[i].blocks -= MAXCHUNKSIZE / 64;
326 : edges[i].blocks = MAXCHUNKSIZE / 64;
327 : ciph_d[i].inp += MAXCHUNKSIZE;
328 : ciph_d[i].out += MAXCHUNKSIZE;
329 : ciph_d[i].blocks = MAXCHUNKSIZE / 16;
330 : memcpy(ciph_d[i].iv, ciph_d[i].out - 16, 16);
331 : }
332 : processed += MAXCHUNKSIZE;
333 : minblocks -= MAXCHUNKSIZE / 64;
334 : } while (minblocks > MAXCHUNKSIZE / 64);
335 : }
336 : # endif
337 : # undef MAXCHUNKSIZE
338 : sha256_multi_block(ctx, hash_d, n4x);
339 :
340 : memset(blocks, 0, sizeof(blocks));
341 : for (i = 0; i < x4; i++) {
342 : unsigned int len = (i == (x4 - 1) ? last : frag),
343 : off = hash_d[i].blocks * 64;
344 : const unsigned char *ptr = hash_d[i].ptr + off;
345 :
346 : off = (len - processed) - (64 - 13) - off; /* remainder actually */
347 : memcpy(blocks[i].c, ptr, off);
348 : blocks[i].c[off] = 0x80;
349 : len += 64 + 13; /* 64 is HMAC header */
350 : len *= 8; /* convert to bits */
351 : if (off < (64 - 8)) {
352 : # ifdef BSWAP4
353 : blocks[i].d[15] = BSWAP4(len);
354 : # else
355 : PUTU32(blocks[i].c + 60, len);
356 : # endif
357 : edges[i].blocks = 1;
358 : } else {
359 : # ifdef BSWAP4
360 : blocks[i].d[31] = BSWAP4(len);
361 : # else
362 : PUTU32(blocks[i].c + 124, len);
363 : # endif
364 : edges[i].blocks = 2;
365 : }
366 : edges[i].ptr = blocks[i].c;
367 : }
368 :
369 : /* hash input tails and finalize */
370 : sha256_multi_block(ctx, edges, n4x);
371 :
372 : memset(blocks, 0, sizeof(blocks));
373 : for (i = 0; i < x4; i++) {
374 : # ifdef BSWAP4
375 : blocks[i].d[0] = BSWAP4(ctx->A[i]);
376 : ctx->A[i] = key->tail.h[0];
377 : blocks[i].d[1] = BSWAP4(ctx->B[i]);
378 : ctx->B[i] = key->tail.h[1];
379 : blocks[i].d[2] = BSWAP4(ctx->C[i]);
380 : ctx->C[i] = key->tail.h[2];
381 : blocks[i].d[3] = BSWAP4(ctx->D[i]);
382 : ctx->D[i] = key->tail.h[3];
383 : blocks[i].d[4] = BSWAP4(ctx->E[i]);
384 : ctx->E[i] = key->tail.h[4];
385 : blocks[i].d[5] = BSWAP4(ctx->F[i]);
386 : ctx->F[i] = key->tail.h[5];
387 : blocks[i].d[6] = BSWAP4(ctx->G[i]);
388 : ctx->G[i] = key->tail.h[6];
389 : blocks[i].d[7] = BSWAP4(ctx->H[i]);
390 : ctx->H[i] = key->tail.h[7];
391 : blocks[i].c[32] = 0x80;
392 : blocks[i].d[15] = BSWAP4((64 + 32) * 8);
393 : # else
394 : PUTU32(blocks[i].c + 0, ctx->A[i]);
395 : ctx->A[i] = key->tail.h[0];
396 : PUTU32(blocks[i].c + 4, ctx->B[i]);
397 : ctx->B[i] = key->tail.h[1];
398 : PUTU32(blocks[i].c + 8, ctx->C[i]);
399 : ctx->C[i] = key->tail.h[2];
400 : PUTU32(blocks[i].c + 12, ctx->D[i]);
401 : ctx->D[i] = key->tail.h[3];
402 : PUTU32(blocks[i].c + 16, ctx->E[i]);
403 : ctx->E[i] = key->tail.h[4];
404 : PUTU32(blocks[i].c + 20, ctx->F[i]);
405 : ctx->F[i] = key->tail.h[5];
406 : PUTU32(blocks[i].c + 24, ctx->G[i]);
407 : ctx->G[i] = key->tail.h[6];
408 : PUTU32(blocks[i].c + 28, ctx->H[i]);
409 : ctx->H[i] = key->tail.h[7];
410 : blocks[i].c[32] = 0x80;
411 : PUTU32(blocks[i].c + 60, (64 + 32) * 8);
412 : # endif
413 : edges[i].ptr = blocks[i].c;
414 : edges[i].blocks = 1;
415 : }
416 :
417 : /* finalize MACs */
418 : sha256_multi_block(ctx, edges, n4x);
419 :
420 : for (i = 0; i < x4; i++) {
421 : unsigned int len = (i == (x4 - 1) ? last : frag), pad, j;
422 : unsigned char *out0 = out;
423 :
424 : memcpy(ciph_d[i].out, ciph_d[i].inp, len - processed);
425 : ciph_d[i].inp = ciph_d[i].out;
426 :
427 : out += 5 + 16 + len;
428 :
429 : /* write MAC */
430 : PUTU32(out + 0, ctx->A[i]);
431 : PUTU32(out + 4, ctx->B[i]);
432 : PUTU32(out + 8, ctx->C[i]);
433 : PUTU32(out + 12, ctx->D[i]);
434 : PUTU32(out + 16, ctx->E[i]);
435 : PUTU32(out + 20, ctx->F[i]);
436 : PUTU32(out + 24, ctx->G[i]);
437 : PUTU32(out + 28, ctx->H[i]);
438 : out += 32;
439 : len += 32;
440 :
441 : /* pad */
442 : pad = 15 - len % 16;
443 : for (j = 0; j <= pad; j++)
444 : *(out++) = pad;
445 : len += pad + 1;
446 :
447 : ciph_d[i].blocks = (len - processed) / 16;
448 : len += 16; /* account for explicit iv */
449 :
450 : /* arrange header */
451 : out0[0] = ((u8 *)key->md.data)[8];
452 : out0[1] = ((u8 *)key->md.data)[9];
453 : out0[2] = ((u8 *)key->md.data)[10];
454 : out0[3] = (u8)(len >> 8);
455 : out0[4] = (u8)(len);
456 :
457 : ret += len + 5;
458 : inp += frag;
459 : }
460 :
461 : aesni_multi_cbc_encrypt(ciph_d, &key->ks, n4x);
462 :
463 : OPENSSL_cleanse(blocks, sizeof(blocks));
464 : OPENSSL_cleanse(ctx, sizeof(*ctx));
465 :
466 : return ret;
467 : }
468 : # endif
469 :
470 : static int aesni_cbc_hmac_sha256_cipher(EVP_CIPHER_CTX *ctx,
471 : unsigned char *out,
472 : const unsigned char *in, size_t len)
473 : {
474 : EVP_AES_HMAC_SHA256 *key = data(ctx);
475 : unsigned int l;
476 : size_t plen = key->payload_length, iv = 0, /* explicit IV in TLS 1.1 and
477 : * later */
478 : sha_off = 0;
479 : # if defined(STITCHED_CALL)
480 : size_t aes_off = 0, blocks;
481 :
482 : sha_off = SHA256_CBLOCK - key->md.num;
483 : # endif
484 :
485 : key->payload_length = NO_PAYLOAD_LENGTH;
486 :
487 : if (len % AES_BLOCK_SIZE)
488 : return 0;
489 :
490 : if (ctx->encrypt) {
491 : if (plen == NO_PAYLOAD_LENGTH)
492 : plen = len;
493 : else if (len !=
494 : ((plen + SHA256_DIGEST_LENGTH +
495 : AES_BLOCK_SIZE) & -AES_BLOCK_SIZE))
496 : return 0;
497 : else if (key->aux.tls_ver >= TLS1_1_VERSION)
498 : iv = AES_BLOCK_SIZE;
499 :
500 : # if defined(STITCHED_CALL)
501 : if (OPENSSL_ia32cap_P[1] & (1 << (60 - 32)) && /* AVX? */
502 : plen > (sha_off + iv) &&
503 : (blocks = (plen - (sha_off + iv)) / SHA256_CBLOCK)) {
504 : SHA256_Update(&key->md, in + iv, sha_off);
505 :
506 : (void)aesni_cbc_sha256_enc(in, out, blocks, &key->ks,
507 : ctx->iv, &key->md, in + iv + sha_off);
508 : blocks *= SHA256_CBLOCK;
509 : aes_off += blocks;
510 : sha_off += blocks;
511 : key->md.Nh += blocks >> 29;
512 : key->md.Nl += blocks <<= 3;
513 : if (key->md.Nl < (unsigned int)blocks)
514 : key->md.Nh++;
515 : } else {
516 : sha_off = 0;
517 : }
518 : # endif
519 : sha_off += iv;
520 : SHA256_Update(&key->md, in + sha_off, plen - sha_off);
521 :
522 : if (plen != len) { /* "TLS" mode of operation */
523 : if (in != out)
524 : memcpy(out + aes_off, in + aes_off, plen - aes_off);
525 :
526 : /* calculate HMAC and append it to payload */
527 : SHA256_Final(out + plen, &key->md);
528 : key->md = key->tail;
529 : SHA256_Update(&key->md, out + plen, SHA256_DIGEST_LENGTH);
530 : SHA256_Final(out + plen, &key->md);
531 :
532 : /* pad the payload|hmac */
533 : plen += SHA256_DIGEST_LENGTH;
534 : for (l = len - plen - 1; plen < len; plen++)
535 : out[plen] = l;
536 : /* encrypt HMAC|padding at once */
537 : aesni_cbc_encrypt(out + aes_off, out + aes_off, len - aes_off,
538 : &key->ks, ctx->iv, 1);
539 : } else {
540 : aesni_cbc_encrypt(in + aes_off, out + aes_off, len - aes_off,
541 : &key->ks, ctx->iv, 1);
542 : }
543 : } else {
544 : union {
545 : unsigned int u[SHA256_DIGEST_LENGTH / sizeof(unsigned int)];
546 : unsigned char c[64 + SHA256_DIGEST_LENGTH];
547 : } mac, *pmac;
548 :
549 : /* arrange cache line alignment */
550 : pmac = (void *)(((size_t)mac.c + 63) & ((size_t)0 - 64));
551 :
552 : /* decrypt HMAC|padding at once */
553 : aesni_cbc_encrypt(in, out, len, &key->ks, ctx->iv, 0);
554 :
555 : if (plen != NO_PAYLOAD_LENGTH) { /* "TLS" mode of operation */
556 : size_t inp_len, mask, j, i;
557 : unsigned int res, maxpad, pad, bitlen;
558 : int ret = 1;
559 : union {
560 : unsigned int u[SHA_LBLOCK];
561 : unsigned char c[SHA256_CBLOCK];
562 : } *data = (void *)key->md.data;
563 :
564 : if ((key->aux.tls_aad[plen - 4] << 8 | key->aux.tls_aad[plen - 3])
565 : >= TLS1_1_VERSION)
566 : iv = AES_BLOCK_SIZE;
567 :
568 : if (len < (iv + SHA256_DIGEST_LENGTH + 1))
569 : return 0;
570 :
571 : /* omit explicit iv */
572 : out += iv;
573 : len -= iv;
574 :
575 : /* figure out payload length */
576 : pad = out[len - 1];
577 : maxpad = len - (SHA256_DIGEST_LENGTH + 1);
578 : maxpad |= (255 - maxpad) >> (sizeof(maxpad) * 8 - 8);
579 : maxpad &= 255;
580 :
581 : inp_len = len - (SHA256_DIGEST_LENGTH + pad + 1);
582 : mask = (0 - ((inp_len - len) >> (sizeof(inp_len) * 8 - 1)));
583 : inp_len &= mask;
584 : ret &= (int)mask;
585 :
586 : key->aux.tls_aad[plen - 2] = inp_len >> 8;
587 : key->aux.tls_aad[plen - 1] = inp_len;
588 :
589 : /* calculate HMAC */
590 : key->md = key->head;
591 : SHA256_Update(&key->md, key->aux.tls_aad, plen);
592 :
593 : # if 1
594 : len -= SHA256_DIGEST_LENGTH; /* amend mac */
595 : if (len >= (256 + SHA256_CBLOCK)) {
596 : j = (len - (256 + SHA256_CBLOCK)) & (0 - SHA256_CBLOCK);
597 : j += SHA256_CBLOCK - key->md.num;
598 : SHA256_Update(&key->md, out, j);
599 : out += j;
600 : len -= j;
601 : inp_len -= j;
602 : }
603 :
604 : /* but pretend as if we hashed padded payload */
605 : bitlen = key->md.Nl + (inp_len << 3); /* at most 18 bits */
606 : # ifdef BSWAP4
607 : bitlen = BSWAP4(bitlen);
608 : # else
609 : mac.c[0] = 0;
610 : mac.c[1] = (unsigned char)(bitlen >> 16);
611 : mac.c[2] = (unsigned char)(bitlen >> 8);
612 : mac.c[3] = (unsigned char)bitlen;
613 : bitlen = mac.u[0];
614 : # endif
615 :
616 : pmac->u[0] = 0;
617 : pmac->u[1] = 0;
618 : pmac->u[2] = 0;
619 : pmac->u[3] = 0;
620 : pmac->u[4] = 0;
621 : pmac->u[5] = 0;
622 : pmac->u[6] = 0;
623 : pmac->u[7] = 0;
624 :
625 : for (res = key->md.num, j = 0; j < len; j++) {
626 : size_t c = out[j];
627 : mask = (j - inp_len) >> (sizeof(j) * 8 - 8);
628 : c &= mask;
629 : c |= 0x80 & ~mask & ~((inp_len - j) >> (sizeof(j) * 8 - 8));
630 : data->c[res++] = (unsigned char)c;
631 :
632 : if (res != SHA256_CBLOCK)
633 : continue;
634 :
635 : /* j is not incremented yet */
636 : mask = 0 - ((inp_len + 7 - j) >> (sizeof(j) * 8 - 1));
637 : data->u[SHA_LBLOCK - 1] |= bitlen & mask;
638 : sha256_block_data_order(&key->md, data, 1);
639 : mask &= 0 - ((j - inp_len - 72) >> (sizeof(j) * 8 - 1));
640 : pmac->u[0] |= key->md.h[0] & mask;
641 : pmac->u[1] |= key->md.h[1] & mask;
642 : pmac->u[2] |= key->md.h[2] & mask;
643 : pmac->u[3] |= key->md.h[3] & mask;
644 : pmac->u[4] |= key->md.h[4] & mask;
645 : pmac->u[5] |= key->md.h[5] & mask;
646 : pmac->u[6] |= key->md.h[6] & mask;
647 : pmac->u[7] |= key->md.h[7] & mask;
648 : res = 0;
649 : }
650 :
651 : for (i = res; i < SHA256_CBLOCK; i++, j++)
652 : data->c[i] = 0;
653 :
654 : if (res > SHA256_CBLOCK - 8) {
655 : mask = 0 - ((inp_len + 8 - j) >> (sizeof(j) * 8 - 1));
656 : data->u[SHA_LBLOCK - 1] |= bitlen & mask;
657 : sha256_block_data_order(&key->md, data, 1);
658 : mask &= 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1));
659 : pmac->u[0] |= key->md.h[0] & mask;
660 : pmac->u[1] |= key->md.h[1] & mask;
661 : pmac->u[2] |= key->md.h[2] & mask;
662 : pmac->u[3] |= key->md.h[3] & mask;
663 : pmac->u[4] |= key->md.h[4] & mask;
664 : pmac->u[5] |= key->md.h[5] & mask;
665 : pmac->u[6] |= key->md.h[6] & mask;
666 : pmac->u[7] |= key->md.h[7] & mask;
667 :
668 : memset(data, 0, SHA256_CBLOCK);
669 : j += 64;
670 : }
671 : data->u[SHA_LBLOCK - 1] = bitlen;
672 : sha256_block_data_order(&key->md, data, 1);
673 : mask = 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1));
674 : pmac->u[0] |= key->md.h[0] & mask;
675 : pmac->u[1] |= key->md.h[1] & mask;
676 : pmac->u[2] |= key->md.h[2] & mask;
677 : pmac->u[3] |= key->md.h[3] & mask;
678 : pmac->u[4] |= key->md.h[4] & mask;
679 : pmac->u[5] |= key->md.h[5] & mask;
680 : pmac->u[6] |= key->md.h[6] & mask;
681 : pmac->u[7] |= key->md.h[7] & mask;
682 :
683 : # ifdef BSWAP4
684 : pmac->u[0] = BSWAP4(pmac->u[0]);
685 : pmac->u[1] = BSWAP4(pmac->u[1]);
686 : pmac->u[2] = BSWAP4(pmac->u[2]);
687 : pmac->u[3] = BSWAP4(pmac->u[3]);
688 : pmac->u[4] = BSWAP4(pmac->u[4]);
689 : pmac->u[5] = BSWAP4(pmac->u[5]);
690 : pmac->u[6] = BSWAP4(pmac->u[6]);
691 : pmac->u[7] = BSWAP4(pmac->u[7]);
692 : # else
693 : for (i = 0; i < 8; i++) {
694 : res = pmac->u[i];
695 : pmac->c[4 * i + 0] = (unsigned char)(res >> 24);
696 : pmac->c[4 * i + 1] = (unsigned char)(res >> 16);
697 : pmac->c[4 * i + 2] = (unsigned char)(res >> 8);
698 : pmac->c[4 * i + 3] = (unsigned char)res;
699 : }
700 : # endif
701 : len += SHA256_DIGEST_LENGTH;
702 : # else
703 : SHA256_Update(&key->md, out, inp_len);
704 : res = key->md.num;
705 : SHA256_Final(pmac->c, &key->md);
706 :
707 : {
708 : unsigned int inp_blocks, pad_blocks;
709 :
710 : /* but pretend as if we hashed padded payload */
711 : inp_blocks =
712 : 1 + ((SHA256_CBLOCK - 9 - res) >> (sizeof(res) * 8 - 1));
713 : res += (unsigned int)(len - inp_len);
714 : pad_blocks = res / SHA256_CBLOCK;
715 : res %= SHA256_CBLOCK;
716 : pad_blocks +=
717 : 1 + ((SHA256_CBLOCK - 9 - res) >> (sizeof(res) * 8 - 1));
718 : for (; inp_blocks < pad_blocks; inp_blocks++)
719 : sha1_block_data_order(&key->md, data, 1);
720 : }
721 : # endif
722 : key->md = key->tail;
723 : SHA256_Update(&key->md, pmac->c, SHA256_DIGEST_LENGTH);
724 : SHA256_Final(pmac->c, &key->md);
725 :
726 : /* verify HMAC */
727 : out += inp_len;
728 : len -= inp_len;
729 : # if 1
730 : {
731 : unsigned char *p =
732 : out + len - 1 - maxpad - SHA256_DIGEST_LENGTH;
733 : size_t off = out - p;
734 : unsigned int c, cmask;
735 :
736 : maxpad += SHA256_DIGEST_LENGTH;
737 : for (res = 0, i = 0, j = 0; j < maxpad; j++) {
738 : c = p[j];
739 : cmask =
740 : ((int)(j - off - SHA256_DIGEST_LENGTH)) >>
741 : (sizeof(int) * 8 - 1);
742 : res |= (c ^ pad) & ~cmask; /* ... and padding */
743 : cmask &= ((int)(off - 1 - j)) >> (sizeof(int) * 8 - 1);
744 : res |= (c ^ pmac->c[i]) & cmask;
745 : i += 1 & cmask;
746 : }
747 : maxpad -= SHA256_DIGEST_LENGTH;
748 :
749 : res = 0 - ((0 - res) >> (sizeof(res) * 8 - 1));
750 : ret &= (int)~res;
751 : }
752 : # else
753 : for (res = 0, i = 0; i < SHA256_DIGEST_LENGTH; i++)
754 : res |= out[i] ^ pmac->c[i];
755 : res = 0 - ((0 - res) >> (sizeof(res) * 8 - 1));
756 : ret &= (int)~res;
757 :
758 : /* verify padding */
759 : pad = (pad & ~res) | (maxpad & res);
760 : out = out + len - 1 - pad;
761 : for (res = 0, i = 0; i < pad; i++)
762 : res |= out[i] ^ pad;
763 :
764 : res = (0 - res) >> (sizeof(res) * 8 - 1);
765 : ret &= (int)~res;
766 : # endif
767 : return ret;
768 : } else {
769 : SHA256_Update(&key->md, out, len);
770 : }
771 : }
772 :
773 : return 1;
774 : }
775 :
776 : static int aesni_cbc_hmac_sha256_ctrl(EVP_CIPHER_CTX *ctx, int type, int arg,
777 : void *ptr)
778 : {
779 : EVP_AES_HMAC_SHA256 *key = data(ctx);
780 :
781 : switch (type) {
782 : case EVP_CTRL_AEAD_SET_MAC_KEY:
783 : {
784 : unsigned int i;
785 : unsigned char hmac_key[64];
786 :
787 : memset(hmac_key, 0, sizeof(hmac_key));
788 :
789 : if (arg > (int)sizeof(hmac_key)) {
790 : SHA256_Init(&key->head);
791 : SHA256_Update(&key->head, ptr, arg);
792 : SHA256_Final(hmac_key, &key->head);
793 : } else {
794 : memcpy(hmac_key, ptr, arg);
795 : }
796 :
797 : for (i = 0; i < sizeof(hmac_key); i++)
798 : hmac_key[i] ^= 0x36; /* ipad */
799 : SHA256_Init(&key->head);
800 : SHA256_Update(&key->head, hmac_key, sizeof(hmac_key));
801 :
802 : for (i = 0; i < sizeof(hmac_key); i++)
803 : hmac_key[i] ^= 0x36 ^ 0x5c; /* opad */
804 : SHA256_Init(&key->tail);
805 : SHA256_Update(&key->tail, hmac_key, sizeof(hmac_key));
806 :
807 : OPENSSL_cleanse(hmac_key, sizeof(hmac_key));
808 :
809 : return 1;
810 : }
811 : case EVP_CTRL_AEAD_TLS1_AAD:
812 : {
813 : unsigned char *p = ptr;
814 : unsigned int len = p[arg - 2] << 8 | p[arg - 1];
815 :
816 : if (arg != EVP_AEAD_TLS1_AAD_LEN)
817 : return -1;
818 :
819 : len = p[arg - 2] << 8 | p[arg - 1];
820 :
821 : if (ctx->encrypt) {
822 : key->payload_length = len;
823 : if ((key->aux.tls_ver =
824 : p[arg - 4] << 8 | p[arg - 3]) >= TLS1_1_VERSION) {
825 : len -= AES_BLOCK_SIZE;
826 : p[arg - 2] = len >> 8;
827 : p[arg - 1] = len;
828 : }
829 : key->md = key->head;
830 : SHA256_Update(&key->md, p, arg);
831 :
832 : return (int)(((len + SHA256_DIGEST_LENGTH +
833 : AES_BLOCK_SIZE) & -AES_BLOCK_SIZE)
834 : - len);
835 : } else {
836 : memcpy(key->aux.tls_aad, ptr, arg);
837 : key->payload_length = arg;
838 :
839 : return SHA256_DIGEST_LENGTH;
840 : }
841 : }
842 : # if !defined(OPENSSL_NO_MULTIBLOCK) && EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK
843 : case EVP_CTRL_TLS1_1_MULTIBLOCK_MAX_BUFSIZE:
844 : return (int)(5 + 16 + ((arg + 32 + 16) & -16));
845 : case EVP_CTRL_TLS1_1_MULTIBLOCK_AAD:
846 : {
847 : EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param =
848 : (EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *) ptr;
849 : unsigned int n4x = 1, x4;
850 : unsigned int frag, last, packlen, inp_len;
851 :
852 : if (arg < (int)sizeof(EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM))
853 : return -1;
854 :
855 : inp_len = param->inp[11] << 8 | param->inp[12];
856 :
857 : if (ctx->encrypt) {
858 : if ((param->inp[9] << 8 | param->inp[10]) < TLS1_1_VERSION)
859 : return -1;
860 :
861 : if (inp_len) {
862 : if (inp_len < 4096)
863 : return 0; /* too short */
864 :
865 : if (inp_len >= 8192 && OPENSSL_ia32cap_P[2] & (1 << 5))
866 : n4x = 2; /* AVX2 */
867 : } else if ((n4x = param->interleave / 4) && n4x <= 2)
868 : inp_len = param->len;
869 : else
870 : return -1;
871 :
872 : key->md = key->head;
873 : SHA256_Update(&key->md, param->inp, 13);
874 :
875 : x4 = 4 * n4x;
876 : n4x += 1;
877 :
878 : frag = inp_len >> n4x;
879 : last = inp_len + frag - (frag << n4x);
880 : if (last > frag && ((last + 13 + 9) % 64 < (x4 - 1))) {
881 : frag++;
882 : last -= x4 - 1;
883 : }
884 :
885 : packlen = 5 + 16 + ((frag + 32 + 16) & -16);
886 : packlen = (packlen << n4x) - packlen;
887 : packlen += 5 + 16 + ((last + 32 + 16) & -16);
888 :
889 : param->interleave = x4;
890 :
891 : return (int)packlen;
892 : } else
893 : return -1; /* not yet */
894 : }
895 : case EVP_CTRL_TLS1_1_MULTIBLOCK_ENCRYPT:
896 : {
897 : EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param =
898 : (EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *) ptr;
899 :
900 : return (int)tls1_1_multi_block_encrypt(key, param->out,
901 : param->inp, param->len,
902 : param->interleave / 4);
903 : }
904 : case EVP_CTRL_TLS1_1_MULTIBLOCK_DECRYPT:
905 : # endif
906 : default:
907 : return -1;
908 : }
909 : }
910 :
911 : static EVP_CIPHER aesni_128_cbc_hmac_sha256_cipher = {
912 : # ifdef NID_aes_128_cbc_hmac_sha256
913 : NID_aes_128_cbc_hmac_sha256,
914 : # else
915 : NID_undef,
916 : # endif
917 : 16, 16, 16,
918 : EVP_CIPH_CBC_MODE | EVP_CIPH_FLAG_DEFAULT_ASN1 |
919 : EVP_CIPH_FLAG_AEAD_CIPHER | EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK,
920 : aesni_cbc_hmac_sha256_init_key,
921 : aesni_cbc_hmac_sha256_cipher,
922 : NULL,
923 : sizeof(EVP_AES_HMAC_SHA256),
924 : EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_set_asn1_iv,
925 : EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_get_asn1_iv,
926 : aesni_cbc_hmac_sha256_ctrl,
927 : NULL
928 : };
929 :
930 : static EVP_CIPHER aesni_256_cbc_hmac_sha256_cipher = {
931 : # ifdef NID_aes_256_cbc_hmac_sha256
932 : NID_aes_256_cbc_hmac_sha256,
933 : # else
934 : NID_undef,
935 : # endif
936 : 16, 32, 16,
937 : EVP_CIPH_CBC_MODE | EVP_CIPH_FLAG_DEFAULT_ASN1 |
938 : EVP_CIPH_FLAG_AEAD_CIPHER | EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK,
939 : aesni_cbc_hmac_sha256_init_key,
940 : aesni_cbc_hmac_sha256_cipher,
941 : NULL,
942 : sizeof(EVP_AES_HMAC_SHA256),
943 : EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_set_asn1_iv,
944 : EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_get_asn1_iv,
945 : aesni_cbc_hmac_sha256_ctrl,
946 : NULL
947 : };
948 :
949 : const EVP_CIPHER *EVP_aes_128_cbc_hmac_sha256(void)
950 : {
951 : return ((OPENSSL_ia32cap_P[1] & AESNI_CAPABLE) &&
952 : aesni_cbc_sha256_enc(NULL, NULL, 0, NULL, NULL, NULL, NULL) ?
953 : &aesni_128_cbc_hmac_sha256_cipher : NULL);
954 : }
955 :
956 : const EVP_CIPHER *EVP_aes_256_cbc_hmac_sha256(void)
957 : {
958 : return ((OPENSSL_ia32cap_P[1] & AESNI_CAPABLE) &&
959 : aesni_cbc_sha256_enc(NULL, NULL, 0, NULL, NULL, NULL, NULL) ?
960 : &aesni_256_cbc_hmac_sha256_cipher : NULL);
961 : }
962 : # else
963 242 : const EVP_CIPHER *EVP_aes_128_cbc_hmac_sha256(void)
964 : {
965 242 : return NULL;
966 : }
967 :
968 242 : const EVP_CIPHER *EVP_aes_256_cbc_hmac_sha256(void)
969 : {
970 242 : return NULL;
971 : }
972 : # endif
973 : #endif
|