]>
Commit | Line | Data |
---|---|---|
0d2bfe52 SL |
1 | /* |
2 | * Copyright 2011-2019 The OpenSSL Project Authors. All Rights Reserved. | |
3 | * | |
4 | * Licensed under the Apache License 2.0 (the "License"). You may not use | |
5 | * this file except in compliance with the License. You can obtain a copy | |
6 | * in the file LICENSE in the source distribution or at | |
7 | * https://www.openssl.org/source/license.html | |
8 | */ | |
9 | ||
0081ce9b RL |
10 | /* |
11 | * AES low level APIs are deprecated for public use, but still ok for internal | |
12 | * use where we're using them to implement the higher level EVP interface, as is | |
13 | * the case here. | |
14 | */ | |
15 | #include "internal/deprecated.h" | |
16 | ||
0d2bfe52 SL |
17 | #include "cipher_aes_cbc_hmac_sha.h" |
18 | ||
19 | #ifndef AES_CBC_HMAC_SHA_CAPABLE | |
20 | int cipher_capable_aes_cbc_hmac_sha1(void) | |
21 | { | |
22 | return 0; | |
23 | } | |
24 | #else | |
25 | ||
26 | # include "crypto/rand.h" | |
27 | # include "crypto/evp.h" | |
28 | # include "internal/constant_time.h" | |
29 | ||
30 | void sha1_block_data_order(void *c, const void *p, size_t len); | |
31 | void aesni_cbc_sha1_enc(const void *inp, void *out, size_t blocks, | |
32 | const AES_KEY *key, unsigned char iv[16], | |
33 | SHA_CTX *ctx, const void *in0); | |
34 | ||
35 | int cipher_capable_aes_cbc_hmac_sha1(void) | |
36 | { | |
37 | return AESNI_CBC_HMAC_SHA_CAPABLE; | |
38 | } | |
39 | ||
40 | static int aesni_cbc_hmac_sha1_init_key(PROV_CIPHER_CTX *vctx, | |
41 | const unsigned char *key, size_t keylen) | |
42 | { | |
43 | int ret; | |
44 | PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx; | |
45 | PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx; | |
46 | ||
47 | if (ctx->base.enc) | |
48 | ret = aesni_set_encrypt_key(key, keylen * 8, &ctx->ks); | |
49 | else | |
50 | ret = aesni_set_decrypt_key(key, keylen * 8, &ctx->ks); | |
51 | ||
52 | SHA1_Init(&sctx->head); /* handy when benchmarking */ | |
53 | sctx->tail = sctx->head; | |
54 | sctx->md = sctx->head; | |
55 | ||
56 | ctx->payload_length = NO_PAYLOAD_LENGTH; | |
57 | ||
58 | return ret < 0 ? 0 : 1; | |
59 | } | |
60 | ||
61 | static void sha1_update(SHA_CTX *c, const void *data, size_t len) | |
62 | { | |
63 | const unsigned char *ptr = data; | |
64 | size_t res; | |
65 | ||
66 | if ((res = c->num)) { | |
67 | res = SHA_CBLOCK - res; | |
68 | if (len < res) | |
69 | res = len; | |
70 | SHA1_Update(c, ptr, res); | |
71 | ptr += res; | |
72 | len -= res; | |
73 | } | |
74 | ||
75 | res = len % SHA_CBLOCK; | |
76 | len -= res; | |
77 | ||
78 | if (len) { | |
79 | sha1_block_data_order(c, ptr, len / SHA_CBLOCK); | |
80 | ||
81 | ptr += len; | |
82 | c->Nh += len >> 29; | |
83 | c->Nl += len <<= 3; | |
84 | if (c->Nl < (unsigned int)len) | |
85 | c->Nh++; | |
86 | } | |
87 | ||
88 | if (res) | |
89 | SHA1_Update(c, ptr, res); | |
90 | } | |
91 | ||
92 | # if !defined(OPENSSL_NO_MULTIBLOCK) | |
93 | ||
94 | typedef struct { | |
95 | unsigned int A[8], B[8], C[8], D[8], E[8]; | |
96 | } SHA1_MB_CTX; | |
97 | ||
98 | typedef struct { | |
99 | const unsigned char *ptr; | |
100 | int blocks; | |
101 | } HASH_DESC; | |
102 | ||
103 | typedef struct { | |
104 | const unsigned char *inp; | |
105 | unsigned char *out; | |
106 | int blocks; | |
107 | u64 iv[2]; | |
108 | } CIPH_DESC; | |
109 | ||
110 | void sha1_multi_block(SHA1_MB_CTX *, const HASH_DESC *, int); | |
111 | void aesni_multi_cbc_encrypt(CIPH_DESC *, void *, int); | |
112 | ||
113 | static size_t tls1_multi_block_encrypt(void *vctx, | |
114 | unsigned char *out, | |
115 | const unsigned char *inp, | |
116 | size_t inp_len, int n4x) | |
117 | { /* n4x is 1 or 2 */ | |
118 | PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx; | |
119 | PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx; | |
120 | HASH_DESC hash_d[8], edges[8]; | |
121 | CIPH_DESC ciph_d[8]; | |
122 | unsigned char storage[sizeof(SHA1_MB_CTX) + 32]; | |
123 | union { | |
124 | u64 q[16]; | |
125 | u32 d[32]; | |
126 | u8 c[128]; | |
127 | } blocks[8]; | |
128 | SHA1_MB_CTX *mctx; | |
129 | unsigned int frag, last, packlen, i; | |
130 | unsigned int x4 = 4 * n4x, minblocks, processed = 0; | |
131 | size_t ret = 0; | |
132 | u8 *IVs; | |
133 | # if defined(BSWAP8) | |
134 | u64 seqnum; | |
135 | # endif | |
136 | ||
137 | /* ask for IVs in bulk */ | |
138 | if (rand_bytes_ex(ctx->base.libctx, (IVs = blocks[0].c), 16 * x4) <= 0) | |
139 | return 0; | |
140 | ||
141 | mctx = (SHA1_MB_CTX *) (storage + 32 - ((size_t)storage % 32)); /* align */ | |
142 | ||
143 | frag = (unsigned int)inp_len >> (1 + n4x); | |
144 | last = (unsigned int)inp_len + frag - (frag << (1 + n4x)); | |
145 | if (last > frag && ((last + 13 + 9) % 64) < (x4 - 1)) { | |
146 | frag++; | |
147 | last -= x4 - 1; | |
148 | } | |
149 | ||
150 | packlen = 5 + 16 + ((frag + 20 + 16) & -16); | |
151 | ||
152 | /* populate descriptors with pointers and IVs */ | |
153 | hash_d[0].ptr = inp; | |
154 | ciph_d[0].inp = inp; | |
155 | /* 5+16 is place for header and explicit IV */ | |
156 | ciph_d[0].out = out + 5 + 16; | |
157 | memcpy(ciph_d[0].out - 16, IVs, 16); | |
158 | memcpy(ciph_d[0].iv, IVs, 16); | |
159 | IVs += 16; | |
160 | ||
161 | for (i = 1; i < x4; i++) { | |
162 | ciph_d[i].inp = hash_d[i].ptr = hash_d[i - 1].ptr + frag; | |
163 | ciph_d[i].out = ciph_d[i - 1].out + packlen; | |
164 | memcpy(ciph_d[i].out - 16, IVs, 16); | |
165 | memcpy(ciph_d[i].iv, IVs, 16); | |
166 | IVs += 16; | |
167 | } | |
168 | ||
169 | # if defined(BSWAP8) | |
170 | memcpy(blocks[0].c, sctx->md.data, 8); | |
171 | seqnum = BSWAP8(blocks[0].q[0]); | |
172 | # endif | |
173 | for (i = 0; i < x4; i++) { | |
174 | unsigned int len = (i == (x4 - 1) ? last : frag); | |
175 | # if !defined(BSWAP8) | |
176 | unsigned int carry, j; | |
177 | # endif | |
178 | ||
179 | mctx->A[i] = sctx->md.h0; | |
180 | mctx->B[i] = sctx->md.h1; | |
181 | mctx->C[i] = sctx->md.h2; | |
182 | mctx->D[i] = sctx->md.h3; | |
183 | mctx->E[i] = sctx->md.h4; | |
184 | ||
185 | /* fix seqnum */ | |
186 | # if defined(BSWAP8) | |
187 | blocks[i].q[0] = BSWAP8(seqnum + i); | |
188 | # else | |
189 | for (carry = i, j = 8; j--;) { | |
190 | blocks[i].c[j] = ((u8 *)sctx->md.data)[j] + carry; | |
191 | carry = (blocks[i].c[j] - carry) >> (sizeof(carry) * 8 - 1); | |
192 | } | |
193 | # endif | |
194 | blocks[i].c[8] = ((u8 *)sctx->md.data)[8]; | |
195 | blocks[i].c[9] = ((u8 *)sctx->md.data)[9]; | |
196 | blocks[i].c[10] = ((u8 *)sctx->md.data)[10]; | |
197 | /* fix length */ | |
198 | blocks[i].c[11] = (u8)(len >> 8); | |
199 | blocks[i].c[12] = (u8)(len); | |
200 | ||
201 | memcpy(blocks[i].c + 13, hash_d[i].ptr, 64 - 13); | |
202 | hash_d[i].ptr += 64 - 13; | |
203 | hash_d[i].blocks = (len - (64 - 13)) / 64; | |
204 | ||
205 | edges[i].ptr = blocks[i].c; | |
206 | edges[i].blocks = 1; | |
207 | } | |
208 | ||
209 | /* hash 13-byte headers and first 64-13 bytes of inputs */ | |
210 | sha1_multi_block(mctx, edges, n4x); | |
211 | /* hash bulk inputs */ | |
212 | # define MAXCHUNKSIZE 2048 | |
213 | # if MAXCHUNKSIZE%64 | |
214 | # error "MAXCHUNKSIZE is not divisible by 64" | |
215 | # elif MAXCHUNKSIZE | |
216 | /* | |
217 | * goal is to minimize pressure on L1 cache by moving in shorter steps, | |
218 | * so that hashed data is still in the cache by the time we encrypt it | |
219 | */ | |
220 | minblocks = ((frag <= last ? frag : last) - (64 - 13)) / 64; | |
221 | if (minblocks > MAXCHUNKSIZE / 64) { | |
222 | for (i = 0; i < x4; i++) { | |
223 | edges[i].ptr = hash_d[i].ptr; | |
224 | edges[i].blocks = MAXCHUNKSIZE / 64; | |
225 | ciph_d[i].blocks = MAXCHUNKSIZE / 16; | |
226 | } | |
227 | do { | |
228 | sha1_multi_block(mctx, edges, n4x); | |
229 | aesni_multi_cbc_encrypt(ciph_d, &ctx->ks, n4x); | |
230 | ||
231 | for (i = 0; i < x4; i++) { | |
232 | edges[i].ptr = hash_d[i].ptr += MAXCHUNKSIZE; | |
233 | hash_d[i].blocks -= MAXCHUNKSIZE / 64; | |
234 | edges[i].blocks = MAXCHUNKSIZE / 64; | |
235 | ciph_d[i].inp += MAXCHUNKSIZE; | |
236 | ciph_d[i].out += MAXCHUNKSIZE; | |
237 | ciph_d[i].blocks = MAXCHUNKSIZE / 16; | |
238 | memcpy(ciph_d[i].iv, ciph_d[i].out - 16, 16); | |
239 | } | |
240 | processed += MAXCHUNKSIZE; | |
241 | minblocks -= MAXCHUNKSIZE / 64; | |
242 | } while (minblocks > MAXCHUNKSIZE / 64); | |
243 | } | |
244 | # endif | |
245 | # undef MAXCHUNKSIZE | |
246 | sha1_multi_block(mctx, hash_d, n4x); | |
247 | ||
248 | memset(blocks, 0, sizeof(blocks)); | |
249 | for (i = 0; i < x4; i++) { | |
250 | unsigned int len = (i == (x4 - 1) ? last : frag), | |
251 | off = hash_d[i].blocks * 64; | |
252 | const unsigned char *ptr = hash_d[i].ptr + off; | |
253 | ||
254 | off = (len - processed) - (64 - 13) - off; /* remainder actually */ | |
255 | memcpy(blocks[i].c, ptr, off); | |
256 | blocks[i].c[off] = 0x80; | |
257 | len += 64 + 13; /* 64 is HMAC header */ | |
258 | len *= 8; /* convert to bits */ | |
259 | if (off < (64 - 8)) { | |
260 | # ifdef BSWAP4 | |
261 | blocks[i].d[15] = BSWAP4(len); | |
262 | # else | |
263 | PUTU32(blocks[i].c + 60, len); | |
264 | # endif | |
265 | edges[i].blocks = 1; | |
266 | } else { | |
267 | # ifdef BSWAP4 | |
268 | blocks[i].d[31] = BSWAP4(len); | |
269 | # else | |
270 | PUTU32(blocks[i].c + 124, len); | |
271 | # endif | |
272 | edges[i].blocks = 2; | |
273 | } | |
274 | edges[i].ptr = blocks[i].c; | |
275 | } | |
276 | ||
277 | /* hash input tails and finalize */ | |
278 | sha1_multi_block(mctx, edges, n4x); | |
279 | ||
280 | memset(blocks, 0, sizeof(blocks)); | |
281 | for (i = 0; i < x4; i++) { | |
282 | # ifdef BSWAP4 | |
283 | blocks[i].d[0] = BSWAP4(mctx->A[i]); | |
284 | mctx->A[i] = sctx->tail.h0; | |
285 | blocks[i].d[1] = BSWAP4(mctx->B[i]); | |
286 | mctx->B[i] = sctx->tail.h1; | |
287 | blocks[i].d[2] = BSWAP4(mctx->C[i]); | |
288 | mctx->C[i] = sctx->tail.h2; | |
289 | blocks[i].d[3] = BSWAP4(mctx->D[i]); | |
290 | mctx->D[i] = sctx->tail.h3; | |
291 | blocks[i].d[4] = BSWAP4(mctx->E[i]); | |
292 | mctx->E[i] = sctx->tail.h4; | |
293 | blocks[i].c[20] = 0x80; | |
294 | blocks[i].d[15] = BSWAP4((64 + 20) * 8); | |
295 | # else | |
296 | PUTU32(blocks[i].c + 0, mctx->A[i]); | |
297 | mctx->A[i] = sctx->tail.h0; | |
298 | PUTU32(blocks[i].c + 4, mctx->B[i]); | |
299 | mctx->B[i] = sctx->tail.h1; | |
300 | PUTU32(blocks[i].c + 8, mctx->C[i]); | |
301 | mctx->C[i] = sctx->tail.h2; | |
302 | PUTU32(blocks[i].c + 12, mctx->D[i]); | |
303 | mctx->D[i] = sctx->tail.h3; | |
304 | PUTU32(blocks[i].c + 16, mctx->E[i]); | |
305 | mctx->E[i] = sctx->tail.h4; | |
306 | blocks[i].c[20] = 0x80; | |
307 | PUTU32(blocks[i].c + 60, (64 + 20) * 8); | |
308 | # endif /* BSWAP */ | |
309 | edges[i].ptr = blocks[i].c; | |
310 | edges[i].blocks = 1; | |
311 | } | |
312 | ||
313 | /* finalize MACs */ | |
314 | sha1_multi_block(mctx, edges, n4x); | |
315 | ||
316 | for (i = 0; i < x4; i++) { | |
317 | unsigned int len = (i == (x4 - 1) ? last : frag), pad, j; | |
318 | unsigned char *out0 = out; | |
319 | ||
320 | memcpy(ciph_d[i].out, ciph_d[i].inp, len - processed); | |
321 | ciph_d[i].inp = ciph_d[i].out; | |
322 | ||
323 | out += 5 + 16 + len; | |
324 | ||
325 | /* write MAC */ | |
326 | PUTU32(out + 0, mctx->A[i]); | |
327 | PUTU32(out + 4, mctx->B[i]); | |
328 | PUTU32(out + 8, mctx->C[i]); | |
329 | PUTU32(out + 12, mctx->D[i]); | |
330 | PUTU32(out + 16, mctx->E[i]); | |
331 | out += 20; | |
332 | len += 20; | |
333 | ||
334 | /* pad */ | |
335 | pad = 15 - len % 16; | |
336 | for (j = 0; j <= pad; j++) | |
337 | *(out++) = pad; | |
338 | len += pad + 1; | |
339 | ||
340 | ciph_d[i].blocks = (len - processed) / 16; | |
341 | len += 16; /* account for explicit iv */ | |
342 | ||
343 | /* arrange header */ | |
344 | out0[0] = ((u8 *)sctx->md.data)[8]; | |
345 | out0[1] = ((u8 *)sctx->md.data)[9]; | |
346 | out0[2] = ((u8 *)sctx->md.data)[10]; | |
347 | out0[3] = (u8)(len >> 8); | |
348 | out0[4] = (u8)(len); | |
349 | ||
350 | ret += len + 5; | |
351 | inp += frag; | |
352 | } | |
353 | ||
354 | aesni_multi_cbc_encrypt(ciph_d, &ctx->ks, n4x); | |
355 | ||
356 | OPENSSL_cleanse(blocks, sizeof(blocks)); | |
357 | OPENSSL_cleanse(mctx, sizeof(*mctx)); | |
358 | ||
359 | ctx->multiblock_encrypt_len = ret; | |
360 | return ret; | |
361 | } | |
362 | # endif /* OPENSSL_NO_MULTIBLOCK */ | |
363 | ||
364 | static int aesni_cbc_hmac_sha1_cipher(PROV_CIPHER_CTX *vctx, | |
365 | unsigned char *out, | |
366 | const unsigned char *in, size_t len) | |
367 | { | |
368 | PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx; | |
369 | PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx; | |
370 | unsigned int l; | |
371 | size_t plen = ctx->payload_length; | |
372 | size_t iv = 0; /* explicit IV in TLS 1.1 and later */ | |
373 | size_t aes_off = 0, blocks; | |
374 | size_t sha_off = SHA_CBLOCK - sctx->md.num; | |
375 | ||
376 | ctx->payload_length = NO_PAYLOAD_LENGTH; | |
377 | ||
378 | if (len % AES_BLOCK_SIZE) | |
379 | return 0; | |
380 | ||
381 | if (ctx->base.enc) { | |
382 | if (plen == NO_PAYLOAD_LENGTH) | |
383 | plen = len; | |
384 | else if (len != | |
385 | ((plen + SHA_DIGEST_LENGTH + | |
386 | AES_BLOCK_SIZE) & -AES_BLOCK_SIZE)) | |
387 | return 0; | |
388 | else if (ctx->aux.tls_ver >= TLS1_1_VERSION) | |
389 | iv = AES_BLOCK_SIZE; | |
390 | ||
391 | if (plen > (sha_off + iv) | |
392 | && (blocks = (plen - (sha_off + iv)) / SHA_CBLOCK)) { | |
393 | sha1_update(&sctx->md, in + iv, sha_off); | |
394 | ||
395 | aesni_cbc_sha1_enc(in, out, blocks, &ctx->ks, ctx->base.iv, | |
396 | &sctx->md, in + iv + sha_off); | |
397 | blocks *= SHA_CBLOCK; | |
398 | aes_off += blocks; | |
399 | sha_off += blocks; | |
400 | sctx->md.Nh += blocks >> 29; | |
401 | sctx->md.Nl += blocks <<= 3; | |
402 | if (sctx->md.Nl < (unsigned int)blocks) | |
403 | sctx->md.Nh++; | |
404 | } else { | |
405 | sha_off = 0; | |
406 | } | |
407 | sha_off += iv; | |
408 | sha1_update(&sctx->md, in + sha_off, plen - sha_off); | |
409 | ||
410 | if (plen != len) { /* "TLS" mode of operation */ | |
411 | if (in != out) | |
412 | memcpy(out + aes_off, in + aes_off, plen - aes_off); | |
413 | ||
414 | /* calculate HMAC and append it to payload */ | |
415 | SHA1_Final(out + plen, &sctx->md); | |
416 | sctx->md = sctx->tail; | |
417 | sha1_update(&sctx->md, out + plen, SHA_DIGEST_LENGTH); | |
418 | SHA1_Final(out + plen, &sctx->md); | |
419 | ||
420 | /* pad the payload|hmac */ | |
421 | plen += SHA_DIGEST_LENGTH; | |
422 | for (l = len - plen - 1; plen < len; plen++) | |
423 | out[plen] = l; | |
424 | /* encrypt HMAC|padding at once */ | |
425 | aesni_cbc_encrypt(out + aes_off, out + aes_off, len - aes_off, | |
426 | &ctx->ks, ctx->base.iv, 1); | |
427 | } else { | |
428 | aesni_cbc_encrypt(in + aes_off, out + aes_off, len - aes_off, | |
429 | &ctx->ks, ctx->base.iv, 1); | |
430 | } | |
431 | } else { | |
432 | union { | |
433 | unsigned int u[SHA_DIGEST_LENGTH / sizeof(unsigned int)]; | |
434 | unsigned char c[32 + SHA_DIGEST_LENGTH]; | |
435 | } mac, *pmac; | |
436 | ||
437 | /* arrange cache line alignment */ | |
438 | pmac = (void *)(((size_t)mac.c + 31) & ((size_t)0 - 32)); | |
439 | ||
440 | if (plen != NO_PAYLOAD_LENGTH) { /* "TLS" mode of operation */ | |
441 | size_t inp_len, mask, j, i; | |
442 | unsigned int res, maxpad, pad, bitlen; | |
443 | int ret = 1; | |
444 | union { | |
445 | unsigned int u[SHA_LBLOCK]; | |
446 | unsigned char c[SHA_CBLOCK]; | |
447 | } *data = (void *)sctx->md.data; | |
448 | ||
449 | if ((ctx->aux.tls_aad[plen - 4] << 8 | ctx->aux.tls_aad[plen - 3]) | |
450 | >= TLS1_1_VERSION) { | |
451 | if (len < (AES_BLOCK_SIZE + SHA_DIGEST_LENGTH + 1)) | |
452 | return 0; | |
453 | ||
454 | /* omit explicit iv */ | |
455 | memcpy(ctx->base.iv, in, AES_BLOCK_SIZE); | |
456 | ||
457 | in += AES_BLOCK_SIZE; | |
458 | out += AES_BLOCK_SIZE; | |
459 | len -= AES_BLOCK_SIZE; | |
460 | } else if (len < (SHA_DIGEST_LENGTH + 1)) | |
461 | return 0; | |
462 | ||
463 | /* decrypt HMAC|padding at once */ | |
464 | aesni_cbc_encrypt(in, out, len, &ctx->ks, ctx->base.iv, 0); | |
465 | ||
466 | /* figure out payload length */ | |
467 | pad = out[len - 1]; | |
468 | maxpad = len - (SHA_DIGEST_LENGTH + 1); | |
469 | maxpad |= (255 - maxpad) >> (sizeof(maxpad) * 8 - 8); | |
470 | maxpad &= 255; | |
471 | ||
472 | mask = constant_time_ge(maxpad, pad); | |
473 | ret &= mask; | |
474 | /* | |
475 | * If pad is invalid then we will fail the above test but we must | |
476 | * continue anyway because we are in constant time code. However, | |
477 | * we'll use the maxpad value instead of the supplied pad to make | |
478 | * sure we perform well defined pointer arithmetic. | |
479 | */ | |
480 | pad = constant_time_select(mask, pad, maxpad); | |
481 | ||
482 | inp_len = len - (SHA_DIGEST_LENGTH + pad + 1); | |
483 | ||
484 | ctx->aux.tls_aad[plen - 2] = inp_len >> 8; | |
485 | ctx->aux.tls_aad[plen - 1] = inp_len; | |
486 | ||
487 | /* calculate HMAC */ | |
488 | sctx->md = sctx->head; | |
489 | sha1_update(&sctx->md, ctx->aux.tls_aad, plen); | |
490 | ||
491 | /* code containing lucky-13 fix */ | |
492 | len -= SHA_DIGEST_LENGTH; /* amend mac */ | |
493 | if (len >= (256 + SHA_CBLOCK)) { | |
494 | j = (len - (256 + SHA_CBLOCK)) & (0 - SHA_CBLOCK); | |
495 | j += SHA_CBLOCK - sctx->md.num; | |
496 | sha1_update(&sctx->md, out, j); | |
497 | out += j; | |
498 | len -= j; | |
499 | inp_len -= j; | |
500 | } | |
501 | ||
502 | /* but pretend as if we hashed padded payload */ | |
503 | bitlen = sctx->md.Nl + (inp_len << 3); /* at most 18 bits */ | |
504 | # ifdef BSWAP4 | |
505 | bitlen = BSWAP4(bitlen); | |
506 | # else | |
507 | mac.c[0] = 0; | |
508 | mac.c[1] = (unsigned char)(bitlen >> 16); | |
509 | mac.c[2] = (unsigned char)(bitlen >> 8); | |
510 | mac.c[3] = (unsigned char)bitlen; | |
511 | bitlen = mac.u[0]; | |
512 | # endif /* BSWAP */ | |
513 | ||
514 | pmac->u[0] = 0; | |
515 | pmac->u[1] = 0; | |
516 | pmac->u[2] = 0; | |
517 | pmac->u[3] = 0; | |
518 | pmac->u[4] = 0; | |
519 | ||
520 | for (res = sctx->md.num, j = 0; j < len; j++) { | |
521 | size_t c = out[j]; | |
522 | mask = (j - inp_len) >> (sizeof(j) * 8 - 8); | |
523 | c &= mask; | |
524 | c |= 0x80 & ~mask & ~((inp_len - j) >> (sizeof(j) * 8 - 8)); | |
525 | data->c[res++] = (unsigned char)c; | |
526 | ||
527 | if (res != SHA_CBLOCK) | |
528 | continue; | |
529 | ||
530 | /* j is not incremented yet */ | |
531 | mask = 0 - ((inp_len + 7 - j) >> (sizeof(j) * 8 - 1)); | |
532 | data->u[SHA_LBLOCK - 1] |= bitlen & mask; | |
533 | sha1_block_data_order(&sctx->md, data, 1); | |
534 | mask &= 0 - ((j - inp_len - 72) >> (sizeof(j) * 8 - 1)); | |
535 | pmac->u[0] |= sctx->md.h0 & mask; | |
536 | pmac->u[1] |= sctx->md.h1 & mask; | |
537 | pmac->u[2] |= sctx->md.h2 & mask; | |
538 | pmac->u[3] |= sctx->md.h3 & mask; | |
539 | pmac->u[4] |= sctx->md.h4 & mask; | |
540 | res = 0; | |
541 | } | |
542 | ||
543 | for (i = res; i < SHA_CBLOCK; i++, j++) | |
544 | data->c[i] = 0; | |
545 | ||
546 | if (res > SHA_CBLOCK - 8) { | |
547 | mask = 0 - ((inp_len + 8 - j) >> (sizeof(j) * 8 - 1)); | |
548 | data->u[SHA_LBLOCK - 1] |= bitlen & mask; | |
549 | sha1_block_data_order(&sctx->md, data, 1); | |
550 | mask &= 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1)); | |
551 | pmac->u[0] |= sctx->md.h0 & mask; | |
552 | pmac->u[1] |= sctx->md.h1 & mask; | |
553 | pmac->u[2] |= sctx->md.h2 & mask; | |
554 | pmac->u[3] |= sctx->md.h3 & mask; | |
555 | pmac->u[4] |= sctx->md.h4 & mask; | |
556 | ||
557 | memset(data, 0, SHA_CBLOCK); | |
558 | j += 64; | |
559 | } | |
560 | data->u[SHA_LBLOCK - 1] = bitlen; | |
561 | sha1_block_data_order(&sctx->md, data, 1); | |
562 | mask = 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1)); | |
563 | pmac->u[0] |= sctx->md.h0 & mask; | |
564 | pmac->u[1] |= sctx->md.h1 & mask; | |
565 | pmac->u[2] |= sctx->md.h2 & mask; | |
566 | pmac->u[3] |= sctx->md.h3 & mask; | |
567 | pmac->u[4] |= sctx->md.h4 & mask; | |
568 | ||
569 | # ifdef BSWAP4 | |
570 | pmac->u[0] = BSWAP4(pmac->u[0]); | |
571 | pmac->u[1] = BSWAP4(pmac->u[1]); | |
572 | pmac->u[2] = BSWAP4(pmac->u[2]); | |
573 | pmac->u[3] = BSWAP4(pmac->u[3]); | |
574 | pmac->u[4] = BSWAP4(pmac->u[4]); | |
575 | # else | |
576 | for (i = 0; i < 5; i++) { | |
577 | res = pmac->u[i]; | |
578 | pmac->c[4 * i + 0] = (unsigned char)(res >> 24); | |
579 | pmac->c[4 * i + 1] = (unsigned char)(res >> 16); | |
580 | pmac->c[4 * i + 2] = (unsigned char)(res >> 8); | |
581 | pmac->c[4 * i + 3] = (unsigned char)res; | |
582 | } | |
583 | # endif /* BSWAP4 */ | |
584 | len += SHA_DIGEST_LENGTH; | |
585 | sctx->md = sctx->tail; | |
586 | sha1_update(&sctx->md, pmac->c, SHA_DIGEST_LENGTH); | |
587 | SHA1_Final(pmac->c, &sctx->md); | |
588 | ||
589 | /* verify HMAC */ | |
590 | out += inp_len; | |
591 | len -= inp_len; | |
592 | /* version of code with lucky-13 fix */ | |
593 | { | |
594 | unsigned char *p = out + len - 1 - maxpad - SHA_DIGEST_LENGTH; | |
595 | size_t off = out - p; | |
596 | unsigned int c, cmask; | |
597 | ||
598 | maxpad += SHA_DIGEST_LENGTH; | |
599 | for (res = 0, i = 0, j = 0; j < maxpad; j++) { | |
600 | c = p[j]; | |
601 | cmask = | |
602 | ((int)(j - off - SHA_DIGEST_LENGTH)) >> (sizeof(int) * | |
603 | 8 - 1); | |
604 | res |= (c ^ pad) & ~cmask; /* ... and padding */ | |
605 | cmask &= ((int)(off - 1 - j)) >> (sizeof(int) * 8 - 1); | |
606 | res |= (c ^ pmac->c[i]) & cmask; | |
607 | i += 1 & cmask; | |
608 | } | |
609 | maxpad -= SHA_DIGEST_LENGTH; | |
610 | ||
611 | res = 0 - ((0 - res) >> (sizeof(res) * 8 - 1)); | |
612 | ret &= (int)~res; | |
613 | } | |
614 | return ret; | |
615 | } else { | |
616 | /* decrypt HMAC|padding at once */ | |
617 | aesni_cbc_encrypt(in, out, len, &ctx->ks, ctx->base.iv, 0); | |
618 | sha1_update(&sctx->md, out, len); | |
619 | } | |
620 | } | |
621 | ||
622 | return 1; | |
623 | } | |
624 | ||
625 | /* EVP_CTRL_AEAD_SET_MAC_KEY */ | |
626 | static void aesni_cbc_hmac_sha1_set_mac_key(void *vctx, | |
627 | const unsigned char *mac, size_t len) | |
628 | { | |
629 | PROV_AES_HMAC_SHA1_CTX *ctx = (PROV_AES_HMAC_SHA1_CTX *)vctx; | |
630 | unsigned int i; | |
631 | unsigned char hmac_key[64]; | |
632 | ||
633 | memset(hmac_key, 0, sizeof(hmac_key)); | |
634 | ||
635 | if (len > (int)sizeof(hmac_key)) { | |
636 | SHA1_Init(&ctx->head); | |
637 | sha1_update(&ctx->head, mac, len); | |
638 | SHA1_Final(hmac_key, &ctx->head); | |
639 | } else { | |
640 | memcpy(hmac_key, mac, len); | |
641 | } | |
642 | ||
643 | for (i = 0; i < sizeof(hmac_key); i++) | |
644 | hmac_key[i] ^= 0x36; /* ipad */ | |
645 | SHA1_Init(&ctx->head); | |
646 | sha1_update(&ctx->head, hmac_key, sizeof(hmac_key)); | |
647 | ||
648 | for (i = 0; i < sizeof(hmac_key); i++) | |
649 | hmac_key[i] ^= 0x36 ^ 0x5c; /* opad */ | |
650 | SHA1_Init(&ctx->tail); | |
651 | sha1_update(&ctx->tail, hmac_key, sizeof(hmac_key)); | |
652 | ||
653 | OPENSSL_cleanse(hmac_key, sizeof(hmac_key)); | |
654 | } | |
655 | ||
656 | /* EVP_CTRL_AEAD_TLS1_AAD */ | |
657 | static int aesni_cbc_hmac_sha1_set_tls1_aad(void *vctx, | |
658 | unsigned char *aad_rec, int aad_len) | |
659 | { | |
660 | PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx; | |
661 | PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx; | |
662 | unsigned char *p = aad_rec; | |
663 | unsigned int len; | |
664 | ||
665 | if (aad_len != EVP_AEAD_TLS1_AAD_LEN) | |
666 | return -1; | |
667 | ||
668 | len = p[aad_len - 2] << 8 | p[aad_len - 1]; | |
669 | ||
670 | if (ctx->base.enc) { | |
671 | ctx->payload_length = len; | |
672 | if ((ctx->aux.tls_ver = | |
673 | p[aad_len - 4] << 8 | p[aad_len - 3]) >= TLS1_1_VERSION) { | |
674 | if (len < AES_BLOCK_SIZE) | |
675 | return 0; | |
676 | len -= AES_BLOCK_SIZE; | |
677 | p[aad_len - 2] = len >> 8; | |
678 | p[aad_len - 1] = len; | |
679 | } | |
680 | sctx->md = sctx->head; | |
681 | sha1_update(&sctx->md, p, aad_len); | |
682 | ctx->tls_aad_pad = (int)(((len + SHA_DIGEST_LENGTH + | |
683 | AES_BLOCK_SIZE) & -AES_BLOCK_SIZE) | |
684 | - len); | |
685 | return 1; | |
686 | } else { | |
687 | memcpy(ctx->aux.tls_aad, aad_rec, aad_len); | |
688 | ctx->payload_length = aad_len; | |
689 | ctx->tls_aad_pad = SHA_DIGEST_LENGTH; | |
690 | return 1; | |
691 | } | |
692 | } | |
693 | ||
694 | # if !defined(OPENSSL_NO_MULTIBLOCK) | |
695 | ||
696 | /* EVP_CTRL_TLS1_1_MULTIBLOCK_MAX_BUFSIZE */ | |
697 | static int aesni_cbc_hmac_sha1_tls1_multiblock_max_bufsize(void *vctx) | |
698 | { | |
699 | PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx; | |
700 | ||
701 | OPENSSL_assert(ctx->multiblock_max_send_fragment != 0); | |
702 | return (int)(5 + 16 | |
703 | + (((int)ctx->multiblock_max_send_fragment + 20 + 16) & -16)); | |
704 | } | |
705 | ||
706 | /* EVP_CTRL_TLS1_1_MULTIBLOCK_AAD */ | |
707 | static int aesni_cbc_hmac_sha1_tls1_multiblock_aad( | |
708 | void *vctx, EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param) | |
709 | { | |
710 | PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx; | |
711 | PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx; | |
712 | unsigned int n4x = 1, x4; | |
713 | unsigned int frag, last, packlen, inp_len; | |
714 | ||
715 | inp_len = param->inp[11] << 8 | param->inp[12]; | |
716 | ctx->multiblock_interleave = param->interleave; | |
717 | ||
718 | if (ctx->base.enc) { | |
719 | if ((param->inp[9] << 8 | param->inp[10]) < TLS1_1_VERSION) | |
720 | return -1; | |
721 | ||
722 | if (inp_len) { | |
723 | if (inp_len < 4096) | |
724 | return 0; /* too short */ | |
725 | ||
726 | if (inp_len >= 8192 && OPENSSL_ia32cap_P[2] & (1 << 5)) | |
727 | n4x = 2; /* AVX2 */ | |
728 | } else if ((n4x = param->interleave / 4) && n4x <= 2) | |
729 | inp_len = param->len; | |
730 | else | |
731 | return -1; | |
732 | ||
733 | sctx->md = sctx->head; | |
734 | sha1_update(&sctx->md, param->inp, 13); | |
735 | ||
736 | x4 = 4 * n4x; | |
737 | n4x += 1; | |
738 | ||
739 | frag = inp_len >> n4x; | |
740 | last = inp_len + frag - (frag << n4x); | |
741 | if (last > frag && ((last + 13 + 9) % 64 < (x4 - 1))) { | |
742 | frag++; | |
743 | last -= x4 - 1; | |
744 | } | |
745 | ||
746 | packlen = 5 + 16 + ((frag + 20 + 16) & -16); | |
747 | packlen = (packlen << n4x) - packlen; | |
748 | packlen += 5 + 16 + ((last + 20 + 16) & -16); | |
749 | ||
750 | param->interleave = x4; | |
751 | /* The returned values used by get need to be stored */ | |
752 | ctx->multiblock_interleave = x4; | |
753 | ctx->multiblock_aad_packlen = packlen; | |
754 | return 1; | |
755 | } | |
756 | return -1; /* not yet */ | |
757 | } | |
758 | ||
759 | /* EVP_CTRL_TLS1_1_MULTIBLOCK_ENCRYPT */ | |
760 | static int aesni_cbc_hmac_sha1_tls1_multiblock_encrypt( | |
761 | void *ctx, EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param) | |
762 | { | |
763 | return (int)tls1_multi_block_encrypt(ctx, param->out, | |
764 | param->inp, param->len, | |
765 | param->interleave / 4); | |
766 | } | |
767 | ||
768 | #endif /* OPENSSL_NO_MULTIBLOCK */ | |
769 | ||
770 | static const PROV_CIPHER_HW_AES_HMAC_SHA cipher_hw_aes_hmac_sha1 = { | |
771 | { | |
772 | aesni_cbc_hmac_sha1_init_key, | |
773 | aesni_cbc_hmac_sha1_cipher | |
774 | }, | |
775 | aesni_cbc_hmac_sha1_set_mac_key, | |
776 | aesni_cbc_hmac_sha1_set_tls1_aad, | |
777 | # if !defined(OPENSSL_NO_MULTIBLOCK) | |
778 | aesni_cbc_hmac_sha1_tls1_multiblock_max_bufsize, | |
779 | aesni_cbc_hmac_sha1_tls1_multiblock_aad, | |
780 | aesni_cbc_hmac_sha1_tls1_multiblock_encrypt | |
781 | # endif | |
782 | }; | |
783 | ||
784 | const PROV_CIPHER_HW_AES_HMAC_SHA *PROV_CIPHER_HW_aes_cbc_hmac_sha1(void) | |
785 | { | |
786 | return &cipher_hw_aes_hmac_sha1; | |
787 | } | |
788 | ||
789 | #endif /* AES_CBC_HMAC_SHA_CAPABLE */ |