1 /*
2  * Copyright 2011-2021 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 
10 /*
11  * All 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 
17 #include "cipher_aes_cbc_hmac_sha.h"
18 
19 #if !defined(AES_CBC_HMAC_SHA_CAPABLE) || !defined(AESNI_CAPABLE)
ossl_cipher_capable_aes_cbc_hmac_sha1(void)20 int ossl_cipher_capable_aes_cbc_hmac_sha1(void)
21 {
22     return 0;
23 }
24 
ossl_prov_cipher_hw_aes_cbc_hmac_sha1(void)25 const PROV_CIPHER_HW_AES_HMAC_SHA *ossl_prov_cipher_hw_aes_cbc_hmac_sha1(void)
26 {
27     return NULL;
28 }
29 #else
30 
31 # include <openssl/rand.h>
32 # include "crypto/evp.h"
33 # include "internal/constant_time.h"
34 
35 void sha1_block_data_order(void *c, const void *p, size_t len);
36 void aesni_cbc_sha1_enc(const void *inp, void *out, size_t blocks,
37                         const AES_KEY *key, unsigned char iv[16],
38                         SHA_CTX *ctx, const void *in0);
39 
ossl_cipher_capable_aes_cbc_hmac_sha1(void)40 int ossl_cipher_capable_aes_cbc_hmac_sha1(void)
41 {
42     return AESNI_CBC_HMAC_SHA_CAPABLE;
43 }
44 
aesni_cbc_hmac_sha1_init_key(PROV_CIPHER_CTX * vctx,const unsigned char * key,size_t keylen)45 static int aesni_cbc_hmac_sha1_init_key(PROV_CIPHER_CTX *vctx,
46                                         const unsigned char *key, size_t keylen)
47 {
48     int ret;
49     PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx;
50     PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx;
51 
52     if (ctx->base.enc)
53         ret = aesni_set_encrypt_key(key, keylen * 8, &ctx->ks);
54     else
55         ret = aesni_set_decrypt_key(key, keylen * 8, &ctx->ks);
56 
57     SHA1_Init(&sctx->head);      /* handy when benchmarking */
58     sctx->tail = sctx->head;
59     sctx->md = sctx->head;
60 
61     ctx->payload_length = NO_PAYLOAD_LENGTH;
62 
63     vctx->removetlspad = 1;
64     vctx->removetlsfixed = SHA_DIGEST_LENGTH + AES_BLOCK_SIZE;
65 
66     return ret < 0 ? 0 : 1;
67 }
68 
sha1_update(SHA_CTX * c,const void * data,size_t len)69 static void sha1_update(SHA_CTX *c, const void *data, size_t len)
70 {
71     const unsigned char *ptr = data;
72     size_t res;
73 
74     if ((res = c->num)) {
75         res = SHA_CBLOCK - res;
76         if (len < res)
77             res = len;
78         SHA1_Update(c, ptr, res);
79         ptr += res;
80         len -= res;
81     }
82 
83     res = len % SHA_CBLOCK;
84     len -= res;
85 
86     if (len) {
87         sha1_block_data_order(c, ptr, len / SHA_CBLOCK);
88 
89         ptr += len;
90         c->Nh += len >> 29;
91         c->Nl += len <<= 3;
92         if (c->Nl < (unsigned int)len)
93             c->Nh++;
94     }
95 
96     if (res)
97         SHA1_Update(c, ptr, res);
98 }
99 
100 # if !defined(OPENSSL_NO_MULTIBLOCK)
101 
102 typedef struct {
103     unsigned int A[8], B[8], C[8], D[8], E[8];
104 } SHA1_MB_CTX;
105 
106 typedef struct {
107     const unsigned char *ptr;
108     int blocks;
109 } HASH_DESC;
110 
111 typedef struct {
112     const unsigned char *inp;
113     unsigned char *out;
114     int blocks;
115     u64 iv[2];
116 } CIPH_DESC;
117 
118 void sha1_multi_block(SHA1_MB_CTX *, const HASH_DESC *, int);
119 void aesni_multi_cbc_encrypt(CIPH_DESC *, void *, int);
120 
tls1_multi_block_encrypt(void * vctx,unsigned char * out,const unsigned char * inp,size_t inp_len,int n4x)121 static size_t tls1_multi_block_encrypt(void *vctx,
122                                        unsigned char *out,
123                                        const unsigned char *inp,
124                                        size_t inp_len, int n4x)
125 {                               /* n4x is 1 or 2 */
126     PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx;
127     PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx;
128     HASH_DESC hash_d[8], edges[8];
129     CIPH_DESC ciph_d[8];
130     unsigned char storage[sizeof(SHA1_MB_CTX) + 32];
131     union {
132         u64 q[16];
133         u32 d[32];
134         u8 c[128];
135     } blocks[8];
136     SHA1_MB_CTX *mctx;
137     unsigned int frag, last, packlen, i;
138     unsigned int x4 = 4 * n4x, minblocks, processed = 0;
139     size_t ret = 0;
140     u8 *IVs;
141 #  if defined(BSWAP8)
142     u64 seqnum;
143 #  endif
144 
145     /* ask for IVs in bulk */
146     if (RAND_bytes_ex(ctx->base.libctx, (IVs = blocks[0].c), 16 * x4, 0) <= 0)
147         return 0;
148 
149     mctx = (SHA1_MB_CTX *) (storage + 32 - ((size_t)storage % 32)); /* align */
150 
151     frag = (unsigned int)inp_len >> (1 + n4x);
152     last = (unsigned int)inp_len + frag - (frag << (1 + n4x));
153     if (last > frag && ((last + 13 + 9) % 64) < (x4 - 1)) {
154         frag++;
155         last -= x4 - 1;
156     }
157 
158     packlen = 5 + 16 + ((frag + 20 + 16) & -16);
159 
160     /* populate descriptors with pointers and IVs */
161     hash_d[0].ptr = inp;
162     ciph_d[0].inp = inp;
163     /* 5+16 is place for header and explicit IV */
164     ciph_d[0].out = out + 5 + 16;
165     memcpy(ciph_d[0].out - 16, IVs, 16);
166     memcpy(ciph_d[0].iv, IVs, 16);
167     IVs += 16;
168 
169     for (i = 1; i < x4; i++) {
170         ciph_d[i].inp = hash_d[i].ptr = hash_d[i - 1].ptr + frag;
171         ciph_d[i].out = ciph_d[i - 1].out + packlen;
172         memcpy(ciph_d[i].out - 16, IVs, 16);
173         memcpy(ciph_d[i].iv, IVs, 16);
174         IVs += 16;
175     }
176 
177 #  if defined(BSWAP8)
178     memcpy(blocks[0].c, sctx->md.data, 8);
179     seqnum = BSWAP8(blocks[0].q[0]);
180 #  endif
181     for (i = 0; i < x4; i++) {
182         unsigned int len = (i == (x4 - 1) ? last : frag);
183 #  if !defined(BSWAP8)
184         unsigned int carry, j;
185 #  endif
186 
187         mctx->A[i] = sctx->md.h0;
188         mctx->B[i] = sctx->md.h1;
189         mctx->C[i] = sctx->md.h2;
190         mctx->D[i] = sctx->md.h3;
191         mctx->E[i] = sctx->md.h4;
192 
193         /* fix seqnum */
194 #  if defined(BSWAP8)
195         blocks[i].q[0] = BSWAP8(seqnum + i);
196 #  else
197         for (carry = i, j = 8; j--;) {
198             blocks[i].c[j] = ((u8 *)sctx->md.data)[j] + carry;
199             carry = (blocks[i].c[j] - carry) >> (sizeof(carry) * 8 - 1);
200         }
201 #  endif
202         blocks[i].c[8] = ((u8 *)sctx->md.data)[8];
203         blocks[i].c[9] = ((u8 *)sctx->md.data)[9];
204         blocks[i].c[10] = ((u8 *)sctx->md.data)[10];
205         /* fix length */
206         blocks[i].c[11] = (u8)(len >> 8);
207         blocks[i].c[12] = (u8)(len);
208 
209         memcpy(blocks[i].c + 13, hash_d[i].ptr, 64 - 13);
210         hash_d[i].ptr += 64 - 13;
211         hash_d[i].blocks = (len - (64 - 13)) / 64;
212 
213         edges[i].ptr = blocks[i].c;
214         edges[i].blocks = 1;
215     }
216 
217     /* hash 13-byte headers and first 64-13 bytes of inputs */
218     sha1_multi_block(mctx, edges, n4x);
219     /* hash bulk inputs */
220 #  define MAXCHUNKSIZE    2048
221 #  if     MAXCHUNKSIZE%64
222 #   error  "MAXCHUNKSIZE is not divisible by 64"
223 #  elif   MAXCHUNKSIZE
224     /*
225      * goal is to minimize pressure on L1 cache by moving in shorter steps,
226      * so that hashed data is still in the cache by the time we encrypt it
227      */
228     minblocks = ((frag <= last ? frag : last) - (64 - 13)) / 64;
229     if (minblocks > MAXCHUNKSIZE / 64) {
230         for (i = 0; i < x4; i++) {
231             edges[i].ptr = hash_d[i].ptr;
232             edges[i].blocks = MAXCHUNKSIZE / 64;
233             ciph_d[i].blocks = MAXCHUNKSIZE / 16;
234         }
235         do {
236             sha1_multi_block(mctx, edges, n4x);
237             aesni_multi_cbc_encrypt(ciph_d, &ctx->ks, n4x);
238 
239             for (i = 0; i < x4; i++) {
240                 edges[i].ptr = hash_d[i].ptr += MAXCHUNKSIZE;
241                 hash_d[i].blocks -= MAXCHUNKSIZE / 64;
242                 edges[i].blocks = MAXCHUNKSIZE / 64;
243                 ciph_d[i].inp += MAXCHUNKSIZE;
244                 ciph_d[i].out += MAXCHUNKSIZE;
245                 ciph_d[i].blocks = MAXCHUNKSIZE / 16;
246                 memcpy(ciph_d[i].iv, ciph_d[i].out - 16, 16);
247             }
248             processed += MAXCHUNKSIZE;
249             minblocks -= MAXCHUNKSIZE / 64;
250         } while (minblocks > MAXCHUNKSIZE / 64);
251     }
252 #  endif
253 #  undef  MAXCHUNKSIZE
254     sha1_multi_block(mctx, hash_d, n4x);
255 
256     memset(blocks, 0, sizeof(blocks));
257     for (i = 0; i < x4; i++) {
258         unsigned int len = (i == (x4 - 1) ? last : frag),
259             off = hash_d[i].blocks * 64;
260         const unsigned char *ptr = hash_d[i].ptr + off;
261 
262         off = (len - processed) - (64 - 13) - off; /* remainder actually */
263         memcpy(blocks[i].c, ptr, off);
264         blocks[i].c[off] = 0x80;
265         len += 64 + 13;         /* 64 is HMAC header */
266         len *= 8;               /* convert to bits */
267         if (off < (64 - 8)) {
268 #  ifdef BSWAP4
269             blocks[i].d[15] = BSWAP4(len);
270 #  else
271             PUTU32(blocks[i].c + 60, len);
272 #  endif
273             edges[i].blocks = 1;
274         } else {
275 #  ifdef BSWAP4
276             blocks[i].d[31] = BSWAP4(len);
277 #  else
278             PUTU32(blocks[i].c + 124, len);
279 #  endif
280             edges[i].blocks = 2;
281         }
282         edges[i].ptr = blocks[i].c;
283     }
284 
285     /* hash input tails and finalize */
286     sha1_multi_block(mctx, edges, n4x);
287 
288     memset(blocks, 0, sizeof(blocks));
289     for (i = 0; i < x4; i++) {
290 #  ifdef BSWAP4
291         blocks[i].d[0] = BSWAP4(mctx->A[i]);
292         mctx->A[i] = sctx->tail.h0;
293         blocks[i].d[1] = BSWAP4(mctx->B[i]);
294         mctx->B[i] = sctx->tail.h1;
295         blocks[i].d[2] = BSWAP4(mctx->C[i]);
296         mctx->C[i] = sctx->tail.h2;
297         blocks[i].d[3] = BSWAP4(mctx->D[i]);
298         mctx->D[i] = sctx->tail.h3;
299         blocks[i].d[4] = BSWAP4(mctx->E[i]);
300         mctx->E[i] = sctx->tail.h4;
301         blocks[i].c[20] = 0x80;
302         blocks[i].d[15] = BSWAP4((64 + 20) * 8);
303 #  else
304         PUTU32(blocks[i].c + 0, mctx->A[i]);
305         mctx->A[i] = sctx->tail.h0;
306         PUTU32(blocks[i].c + 4, mctx->B[i]);
307         mctx->B[i] = sctx->tail.h1;
308         PUTU32(blocks[i].c + 8, mctx->C[i]);
309         mctx->C[i] = sctx->tail.h2;
310         PUTU32(blocks[i].c + 12, mctx->D[i]);
311         mctx->D[i] = sctx->tail.h3;
312         PUTU32(blocks[i].c + 16, mctx->E[i]);
313         mctx->E[i] = sctx->tail.h4;
314         blocks[i].c[20] = 0x80;
315         PUTU32(blocks[i].c + 60, (64 + 20) * 8);
316 #  endif /* BSWAP */
317         edges[i].ptr = blocks[i].c;
318         edges[i].blocks = 1;
319     }
320 
321     /* finalize MACs */
322     sha1_multi_block(mctx, edges, n4x);
323 
324     for (i = 0; i < x4; i++) {
325         unsigned int len = (i == (x4 - 1) ? last : frag), pad, j;
326         unsigned char *out0 = out;
327 
328         memcpy(ciph_d[i].out, ciph_d[i].inp, len - processed);
329         ciph_d[i].inp = ciph_d[i].out;
330 
331         out += 5 + 16 + len;
332 
333         /* write MAC */
334         PUTU32(out + 0, mctx->A[i]);
335         PUTU32(out + 4, mctx->B[i]);
336         PUTU32(out + 8, mctx->C[i]);
337         PUTU32(out + 12, mctx->D[i]);
338         PUTU32(out + 16, mctx->E[i]);
339         out += 20;
340         len += 20;
341 
342         /* pad */
343         pad = 15 - len % 16;
344         for (j = 0; j <= pad; j++)
345             *(out++) = pad;
346         len += pad + 1;
347 
348         ciph_d[i].blocks = (len - processed) / 16;
349         len += 16;              /* account for explicit iv */
350 
351         /* arrange header */
352         out0[0] = ((u8 *)sctx->md.data)[8];
353         out0[1] = ((u8 *)sctx->md.data)[9];
354         out0[2] = ((u8 *)sctx->md.data)[10];
355         out0[3] = (u8)(len >> 8);
356         out0[4] = (u8)(len);
357 
358         ret += len + 5;
359         inp += frag;
360     }
361 
362     aesni_multi_cbc_encrypt(ciph_d, &ctx->ks, n4x);
363 
364     OPENSSL_cleanse(blocks, sizeof(blocks));
365     OPENSSL_cleanse(mctx, sizeof(*mctx));
366 
367     ctx->multiblock_encrypt_len = ret;
368     return ret;
369 }
370 # endif /* OPENSSL_NO_MULTIBLOCK */
371 
aesni_cbc_hmac_sha1_cipher(PROV_CIPHER_CTX * vctx,unsigned char * out,const unsigned char * in,size_t len)372 static int aesni_cbc_hmac_sha1_cipher(PROV_CIPHER_CTX *vctx,
373                                       unsigned char *out,
374                                       const unsigned char *in, size_t len)
375 {
376     PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx;
377     PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx;
378     unsigned int l;
379     size_t plen = ctx->payload_length;
380     size_t iv = 0; /* explicit IV in TLS 1.1 and later */
381     size_t aes_off = 0, blocks;
382     size_t sha_off = SHA_CBLOCK - sctx->md.num;
383 
384     ctx->payload_length = NO_PAYLOAD_LENGTH;
385 
386     if (len % AES_BLOCK_SIZE)
387         return 0;
388 
389     if (ctx->base.enc) {
390         if (plen == NO_PAYLOAD_LENGTH)
391             plen = len;
392         else if (len !=
393                  ((plen + SHA_DIGEST_LENGTH +
394                    AES_BLOCK_SIZE) & -AES_BLOCK_SIZE))
395             return 0;
396         else if (ctx->aux.tls_ver >= TLS1_1_VERSION)
397             iv = AES_BLOCK_SIZE;
398 
399         if (plen > (sha_off + iv)
400                 && (blocks = (plen - (sha_off + iv)) / SHA_CBLOCK)) {
401             sha1_update(&sctx->md, in + iv, sha_off);
402 
403             aesni_cbc_sha1_enc(in, out, blocks, &ctx->ks, ctx->base.iv,
404                                &sctx->md, in + iv + sha_off);
405             blocks *= SHA_CBLOCK;
406             aes_off += blocks;
407             sha_off += blocks;
408             sctx->md.Nh += blocks >> 29;
409             sctx->md.Nl += blocks <<= 3;
410             if (sctx->md.Nl < (unsigned int)blocks)
411                 sctx->md.Nh++;
412         } else {
413             sha_off = 0;
414         }
415         sha_off += iv;
416         sha1_update(&sctx->md, in + sha_off, plen - sha_off);
417 
418         if (plen != len) {      /* "TLS" mode of operation */
419             if (in != out)
420                 memcpy(out + aes_off, in + aes_off, plen - aes_off);
421 
422             /* calculate HMAC and append it to payload */
423             SHA1_Final(out + plen, &sctx->md);
424             sctx->md = sctx->tail;
425             sha1_update(&sctx->md, out + plen, SHA_DIGEST_LENGTH);
426             SHA1_Final(out + plen, &sctx->md);
427 
428             /* pad the payload|hmac */
429             plen += SHA_DIGEST_LENGTH;
430             for (l = len - plen - 1; plen < len; plen++)
431                 out[plen] = l;
432             /* encrypt HMAC|padding at once */
433             aesni_cbc_encrypt(out + aes_off, out + aes_off, len - aes_off,
434                               &ctx->ks, ctx->base.iv, 1);
435         } else {
436             aesni_cbc_encrypt(in + aes_off, out + aes_off, len - aes_off,
437                               &ctx->ks, ctx->base.iv, 1);
438         }
439     } else {
440         union {
441             unsigned int u[SHA_DIGEST_LENGTH / sizeof(unsigned int)];
442             unsigned char c[32 + SHA_DIGEST_LENGTH];
443         } mac, *pmac;
444 
445         /* arrange cache line alignment */
446         pmac = (void *)(((size_t)mac.c + 31) & ((size_t)0 - 32));
447 
448         if (plen != NO_PAYLOAD_LENGTH) { /* "TLS" mode of operation */
449             size_t inp_len, mask, j, i;
450             unsigned int res, maxpad, pad, bitlen;
451             int ret = 1;
452             union {
453                 unsigned int u[SHA_LBLOCK];
454                 unsigned char c[SHA_CBLOCK];
455             } *data = (void *)sctx->md.data;
456 
457             if ((ctx->aux.tls_aad[plen - 4] << 8 | ctx->aux.tls_aad[plen - 3])
458                 >= TLS1_1_VERSION) {
459                 if (len < (AES_BLOCK_SIZE + SHA_DIGEST_LENGTH + 1))
460                     return 0;
461 
462                 /* omit explicit iv */
463                 memcpy(ctx->base.iv, in, AES_BLOCK_SIZE);
464 
465                 in += AES_BLOCK_SIZE;
466                 out += AES_BLOCK_SIZE;
467                 len -= AES_BLOCK_SIZE;
468             } else if (len < (SHA_DIGEST_LENGTH + 1))
469                 return 0;
470 
471             /* decrypt HMAC|padding at once */
472             aesni_cbc_encrypt(in, out, len, &ctx->ks, ctx->base.iv, 0);
473 
474             /* figure out payload length */
475             pad = out[len - 1];
476             maxpad = len - (SHA_DIGEST_LENGTH + 1);
477             maxpad |= (255 - maxpad) >> (sizeof(maxpad) * 8 - 8);
478             maxpad &= 255;
479 
480             mask = constant_time_ge(maxpad, pad);
481             ret &= mask;
482             /*
483              * If pad is invalid then we will fail the above test but we must
484              * continue anyway because we are in constant time code. However,
485              * we'll use the maxpad value instead of the supplied pad to make
486              * sure we perform well defined pointer arithmetic.
487              */
488             pad = constant_time_select(mask, pad, maxpad);
489 
490             inp_len = len - (SHA_DIGEST_LENGTH + pad + 1);
491 
492             ctx->aux.tls_aad[plen - 2] = inp_len >> 8;
493             ctx->aux.tls_aad[plen - 1] = inp_len;
494 
495             /* calculate HMAC */
496             sctx->md = sctx->head;
497             sha1_update(&sctx->md, ctx->aux.tls_aad, plen);
498 
499             /* code containing lucky-13 fix */
500             len -= SHA_DIGEST_LENGTH; /* amend mac */
501             if (len >= (256 + SHA_CBLOCK)) {
502                 j = (len - (256 + SHA_CBLOCK)) & (0 - SHA_CBLOCK);
503                 j += SHA_CBLOCK - sctx->md.num;
504                 sha1_update(&sctx->md, out, j);
505                 out += j;
506                 len -= j;
507                 inp_len -= j;
508             }
509 
510             /* but pretend as if we hashed padded payload */
511             bitlen = sctx->md.Nl + (inp_len << 3); /* at most 18 bits */
512 # ifdef BSWAP4
513             bitlen = BSWAP4(bitlen);
514 # else
515             mac.c[0] = 0;
516             mac.c[1] = (unsigned char)(bitlen >> 16);
517             mac.c[2] = (unsigned char)(bitlen >> 8);
518             mac.c[3] = (unsigned char)bitlen;
519             bitlen = mac.u[0];
520 # endif /* BSWAP */
521 
522             pmac->u[0] = 0;
523             pmac->u[1] = 0;
524             pmac->u[2] = 0;
525             pmac->u[3] = 0;
526             pmac->u[4] = 0;
527 
528             for (res = sctx->md.num, j = 0; j < len; j++) {
529                 size_t c = out[j];
530                 mask = (j - inp_len) >> (sizeof(j) * 8 - 8);
531                 c &= mask;
532                 c |= 0x80 & ~mask & ~((inp_len - j) >> (sizeof(j) * 8 - 8));
533                 data->c[res++] = (unsigned char)c;
534 
535                 if (res != SHA_CBLOCK)
536                     continue;
537 
538                 /* j is not incremented yet */
539                 mask = 0 - ((inp_len + 7 - j) >> (sizeof(j) * 8 - 1));
540                 data->u[SHA_LBLOCK - 1] |= bitlen & mask;
541                 sha1_block_data_order(&sctx->md, data, 1);
542                 mask &= 0 - ((j - inp_len - 72) >> (sizeof(j) * 8 - 1));
543                 pmac->u[0] |= sctx->md.h0 & mask;
544                 pmac->u[1] |= sctx->md.h1 & mask;
545                 pmac->u[2] |= sctx->md.h2 & mask;
546                 pmac->u[3] |= sctx->md.h3 & mask;
547                 pmac->u[4] |= sctx->md.h4 & mask;
548                 res = 0;
549             }
550 
551             for (i = res; i < SHA_CBLOCK; i++, j++)
552                 data->c[i] = 0;
553 
554             if (res > SHA_CBLOCK - 8) {
555                 mask = 0 - ((inp_len + 8 - j) >> (sizeof(j) * 8 - 1));
556                 data->u[SHA_LBLOCK - 1] |= bitlen & mask;
557                 sha1_block_data_order(&sctx->md, data, 1);
558                 mask &= 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1));
559                 pmac->u[0] |= sctx->md.h0 & mask;
560                 pmac->u[1] |= sctx->md.h1 & mask;
561                 pmac->u[2] |= sctx->md.h2 & mask;
562                 pmac->u[3] |= sctx->md.h3 & mask;
563                 pmac->u[4] |= sctx->md.h4 & mask;
564 
565                 memset(data, 0, SHA_CBLOCK);
566                 j += 64;
567             }
568             data->u[SHA_LBLOCK - 1] = bitlen;
569             sha1_block_data_order(&sctx->md, data, 1);
570             mask = 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1));
571             pmac->u[0] |= sctx->md.h0 & mask;
572             pmac->u[1] |= sctx->md.h1 & mask;
573             pmac->u[2] |= sctx->md.h2 & mask;
574             pmac->u[3] |= sctx->md.h3 & mask;
575             pmac->u[4] |= sctx->md.h4 & mask;
576 
577 # ifdef BSWAP4
578             pmac->u[0] = BSWAP4(pmac->u[0]);
579             pmac->u[1] = BSWAP4(pmac->u[1]);
580             pmac->u[2] = BSWAP4(pmac->u[2]);
581             pmac->u[3] = BSWAP4(pmac->u[3]);
582             pmac->u[4] = BSWAP4(pmac->u[4]);
583 # else
584             for (i = 0; i < 5; i++) {
585                 res = pmac->u[i];
586                 pmac->c[4 * i + 0] = (unsigned char)(res >> 24);
587                 pmac->c[4 * i + 1] = (unsigned char)(res >> 16);
588                 pmac->c[4 * i + 2] = (unsigned char)(res >> 8);
589                 pmac->c[4 * i + 3] = (unsigned char)res;
590             }
591 # endif /* BSWAP4 */
592             len += SHA_DIGEST_LENGTH;
593             sctx->md = sctx->tail;
594             sha1_update(&sctx->md, pmac->c, SHA_DIGEST_LENGTH);
595             SHA1_Final(pmac->c, &sctx->md);
596 
597             /* verify HMAC */
598             out += inp_len;
599             len -= inp_len;
600             /* version of code with lucky-13 fix */
601             {
602                 unsigned char *p = out + len - 1 - maxpad - SHA_DIGEST_LENGTH;
603                 size_t off = out - p;
604                 unsigned int c, cmask;
605 
606                 for (res = 0, i = 0, j = 0; j < maxpad + SHA_DIGEST_LENGTH; j++) {
607                     c = p[j];
608                     cmask =
609                         ((int)(j - off - SHA_DIGEST_LENGTH)) >> (sizeof(int) *
610                                                                  8 - 1);
611                     res |= (c ^ pad) & ~cmask; /* ... and padding */
612                     cmask &= ((int)(off - 1 - j)) >> (sizeof(int) * 8 - 1);
613                     res |= (c ^ pmac->c[i]) & cmask;
614                     i += 1 & cmask;
615                 }
616 
617                 res = 0 - ((0 - res) >> (sizeof(res) * 8 - 1));
618                 ret &= (int)~res;
619             }
620             return ret;
621         } else {
622             /* decrypt HMAC|padding at once */
623             aesni_cbc_encrypt(in, out, len, &ctx->ks, ctx->base.iv, 0);
624             sha1_update(&sctx->md, out, len);
625         }
626     }
627 
628     return 1;
629 }
630 
631 /* EVP_CTRL_AEAD_SET_MAC_KEY */
aesni_cbc_hmac_sha1_set_mac_key(void * vctx,const unsigned char * mac,size_t len)632 static void aesni_cbc_hmac_sha1_set_mac_key(void *vctx,
633                                             const unsigned char *mac, size_t len)
634 {
635     PROV_AES_HMAC_SHA1_CTX *ctx = (PROV_AES_HMAC_SHA1_CTX *)vctx;
636     unsigned int i;
637     unsigned char hmac_key[64];
638 
639     memset(hmac_key, 0, sizeof(hmac_key));
640 
641     if (len > (int)sizeof(hmac_key)) {
642         SHA1_Init(&ctx->head);
643         sha1_update(&ctx->head, mac, len);
644         SHA1_Final(hmac_key, &ctx->head);
645     } else {
646         memcpy(hmac_key, mac, len);
647     }
648 
649     for (i = 0; i < sizeof(hmac_key); i++)
650         hmac_key[i] ^= 0x36; /* ipad */
651     SHA1_Init(&ctx->head);
652     sha1_update(&ctx->head, hmac_key, sizeof(hmac_key));
653 
654     for (i = 0; i < sizeof(hmac_key); i++)
655         hmac_key[i] ^= 0x36 ^ 0x5c; /* opad */
656     SHA1_Init(&ctx->tail);
657     sha1_update(&ctx->tail, hmac_key, sizeof(hmac_key));
658 
659     OPENSSL_cleanse(hmac_key, sizeof(hmac_key));
660 }
661 
662 /* EVP_CTRL_AEAD_TLS1_AAD */
aesni_cbc_hmac_sha1_set_tls1_aad(void * vctx,unsigned char * aad_rec,int aad_len)663 static int aesni_cbc_hmac_sha1_set_tls1_aad(void *vctx,
664                                             unsigned char *aad_rec, int aad_len)
665 {
666     PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx;
667     PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx;
668     unsigned char *p = aad_rec;
669     unsigned int len;
670 
671     if (aad_len != EVP_AEAD_TLS1_AAD_LEN)
672         return -1;
673 
674     len = p[aad_len - 2] << 8 | p[aad_len - 1];
675 
676     if (ctx->base.enc) {
677         ctx->payload_length = len;
678         if ((ctx->aux.tls_ver =
679              p[aad_len - 4] << 8 | p[aad_len - 3]) >= TLS1_1_VERSION) {
680             if (len < AES_BLOCK_SIZE)
681                 return 0;
682             len -= AES_BLOCK_SIZE;
683             p[aad_len - 2] = len >> 8;
684             p[aad_len - 1] = len;
685         }
686         sctx->md = sctx->head;
687         sha1_update(&sctx->md, p, aad_len);
688         ctx->tls_aad_pad = (int)(((len + SHA_DIGEST_LENGTH +
689                        AES_BLOCK_SIZE) & -AES_BLOCK_SIZE)
690                      - len);
691         return 1;
692     } else {
693         memcpy(ctx->aux.tls_aad, aad_rec, aad_len);
694         ctx->payload_length = aad_len;
695         ctx->tls_aad_pad = SHA_DIGEST_LENGTH;
696         return 1;
697     }
698 }
699 
700 # if !defined(OPENSSL_NO_MULTIBLOCK)
701 
702 /* EVP_CTRL_TLS1_1_MULTIBLOCK_MAX_BUFSIZE */
aesni_cbc_hmac_sha1_tls1_multiblock_max_bufsize(void * vctx)703 static int aesni_cbc_hmac_sha1_tls1_multiblock_max_bufsize(void *vctx)
704 {
705     PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx;
706 
707     OPENSSL_assert(ctx->multiblock_max_send_fragment != 0);
708     return (int)(5 + 16
709                  + (((int)ctx->multiblock_max_send_fragment + 20 + 16) & -16));
710 }
711 
712 /* EVP_CTRL_TLS1_1_MULTIBLOCK_AAD */
aesni_cbc_hmac_sha1_tls1_multiblock_aad(void * vctx,EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM * param)713 static int aesni_cbc_hmac_sha1_tls1_multiblock_aad(
714     void *vctx, EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param)
715 {
716     PROV_AES_HMAC_SHA_CTX *ctx = (PROV_AES_HMAC_SHA_CTX *)vctx;
717     PROV_AES_HMAC_SHA1_CTX *sctx = (PROV_AES_HMAC_SHA1_CTX *)vctx;
718     unsigned int n4x = 1, x4;
719     unsigned int frag, last, packlen, inp_len;
720 
721     inp_len = param->inp[11] << 8 | param->inp[12];
722     ctx->multiblock_interleave = param->interleave;
723 
724     if (ctx->base.enc) {
725         if ((param->inp[9] << 8 | param->inp[10]) < TLS1_1_VERSION)
726             return -1;
727 
728         if (inp_len) {
729             if (inp_len < 4096)
730                 return 0; /* too short */
731 
732             if (inp_len >= 8192 && OPENSSL_ia32cap_P[2] & (1 << 5))
733                 n4x = 2; /* AVX2 */
734         } else if ((n4x = param->interleave / 4) && n4x <= 2)
735             inp_len = param->len;
736         else
737             return -1;
738 
739         sctx->md = sctx->head;
740         sha1_update(&sctx->md, param->inp, 13);
741 
742         x4 = 4 * n4x;
743         n4x += 1;
744 
745         frag = inp_len >> n4x;
746         last = inp_len + frag - (frag << n4x);
747         if (last > frag && ((last + 13 + 9) % 64 < (x4 - 1))) {
748             frag++;
749             last -= x4 - 1;
750         }
751 
752         packlen = 5 + 16 + ((frag + 20 + 16) & -16);
753         packlen = (packlen << n4x) - packlen;
754         packlen += 5 + 16 + ((last + 20 + 16) & -16);
755 
756         param->interleave = x4;
757         /* The returned values used by get need to be stored */
758         ctx->multiblock_interleave = x4;
759         ctx->multiblock_aad_packlen = packlen;
760         return 1;
761     }
762     return -1;      /* not yet */
763 }
764 
765 /* EVP_CTRL_TLS1_1_MULTIBLOCK_ENCRYPT */
aesni_cbc_hmac_sha1_tls1_multiblock_encrypt(void * ctx,EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM * param)766 static int aesni_cbc_hmac_sha1_tls1_multiblock_encrypt(
767     void *ctx, EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param)
768 {
769     return (int)tls1_multi_block_encrypt(ctx, param->out,
770                                          param->inp, param->len,
771                                          param->interleave / 4);
772 }
773 
774 # endif /* OPENSSL_NO_MULTIBLOCK */
775 
776 static const PROV_CIPHER_HW_AES_HMAC_SHA cipher_hw_aes_hmac_sha1 = {
777     {
778       aesni_cbc_hmac_sha1_init_key,
779       aesni_cbc_hmac_sha1_cipher
780     },
781     aesni_cbc_hmac_sha1_set_mac_key,
782     aesni_cbc_hmac_sha1_set_tls1_aad,
783 # if !defined(OPENSSL_NO_MULTIBLOCK)
784     aesni_cbc_hmac_sha1_tls1_multiblock_max_bufsize,
785     aesni_cbc_hmac_sha1_tls1_multiblock_aad,
786     aesni_cbc_hmac_sha1_tls1_multiblock_encrypt
787 # endif
788 };
789 
ossl_prov_cipher_hw_aes_cbc_hmac_sha1(void)790 const PROV_CIPHER_HW_AES_HMAC_SHA *ossl_prov_cipher_hw_aes_cbc_hmac_sha1(void)
791 {
792     return &cipher_hw_aes_hmac_sha1;
793 }
794 
795 #endif /* !defined(AES_CBC_HMAC_SHA_CAPABLE) || !defined(AESNI_CAPABLE) */
796