1 /* ====================================================================
2 * Copyright (c) 2010 The OpenSSL Project. All rights reserved.
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
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
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/)"
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 * openssl-core@openssl.org.
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.
30 * 6. Redistributions of any form whatsoever must retain the following
32 * "This product includes software developed by the OpenSSL Project
33 * for use in the OpenSSL Toolkit (http://www.openssl.org/)"
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 * ====================================================================
50 #define OPENSSL_FIPSAPI
52 #include <openssl/crypto.h>
53 #include "modes_lcl.h"
63 #if defined(BSWAP4) && defined(STRICT_ALIGNMENT)
64 /* redefine, because alignment is ensured */
66 # define GETU32(p) BSWAP4(*(const u32 *)(p))
68 # define PUTU32(p,v) *(u32 *)(p) = BSWAP4(v)
71 #define PACK(s) ((size_t)(s)<<(sizeof(size_t)*8-16))
72 #define REDUCE1BIT(V) do { \
73 if (sizeof(size_t)==8) { \
74 u64 T = U64(0xe100000000000000) & (0-(V.lo&1)); \
75 V.lo = (V.hi<<63)|(V.lo>>1); \
76 V.hi = (V.hi>>1 )^T; \
79 u32 T = 0xe1000000U & (0-(u32)(V.lo&1)); \
80 V.lo = (V.hi<<63)|(V.lo>>1); \
81 V.hi = (V.hi>>1 )^((u64)T<<32); \
86 * Even though permitted values for TABLE_BITS are 8, 4 and 1, it should
87 * never be set to 8. 8 is effectively reserved for testing purposes.
88 * TABLE_BITS>1 are lookup-table-driven implementations referred to as
89 * "Shoup's" in GCM specification. In other words OpenSSL does not cover
90 * whole spectrum of possible table driven implementations. Why? In
91 * non-"Shoup's" case memory access pattern is segmented in such manner,
92 * that it's trivial to see that cache timing information can reveal
93 * fair portion of intermediate hash value. Given that ciphertext is
94 * always available to attacker, it's possible for him to attempt to
95 * deduce secret parameter H and if successful, tamper with messages
96 * [which is nothing but trivial in CTR mode]. In "Shoup's" case it's
97 * not as trivial, but there is no reason to believe that it's resistant
98 * to cache-timing attack. And the thing about "8-bit" implementation is
99 * that it consumes 16 (sixteen) times more memory, 4KB per individual
100 * key + 1KB shared. Well, on pros side it should be twice as fast as
101 * "4-bit" version. And for gcc-generated x86[_64] code, "8-bit" version
102 * was observed to run ~75% faster, closer to 100% for commercial
103 * compilers... Yet "4-bit" procedure is preferred, because it's
104 * believed to provide better security-performance balance and adequate
105 * all-round performance. "All-round" refers to things like:
107 * - shorter setup time effectively improves overall timing for
108 * handling short messages;
109 * - larger table allocation can become unbearable because of VM
110 * subsystem penalties (for example on Windows large enough free
111 * results in VM working set trimming, meaning that consequent
112 * malloc would immediately incur working set expansion);
113 * - larger table has larger cache footprint, which can affect
114 * performance of other code paths (not necessarily even from same
115 * thread in Hyper-Threading world);
117 * Value of 1 is not appropriate for performance reasons.
121 static void gcm_init_8bit(u128 Htable[256], u64 H[2])
131 for (Htable[128] = V, i = 64; i > 0; i >>= 1) {
136 for (i = 2; i < 256; i <<= 1) {
137 u128 *Hi = Htable + i, H0 = *Hi;
138 for (j = 1; j < i; ++j) {
139 Hi[j].hi = H0.hi ^ Htable[j].hi;
140 Hi[j].lo = H0.lo ^ Htable[j].lo;
145 static void gcm_gmult_8bit(u64 Xi[2], const u128 Htable[256])
148 const u8 *xi = (const u8 *)Xi + 15;
156 static const size_t rem_8bit[256] = {
157 PACK(0x0000), PACK(0x01C2), PACK(0x0384), PACK(0x0246),
158 PACK(0x0708), PACK(0x06CA), PACK(0x048C), PACK(0x054E),
159 PACK(0x0E10), PACK(0x0FD2), PACK(0x0D94), PACK(0x0C56),
160 PACK(0x0918), PACK(0x08DA), PACK(0x0A9C), PACK(0x0B5E),
161 PACK(0x1C20), PACK(0x1DE2), PACK(0x1FA4), PACK(0x1E66),
162 PACK(0x1B28), PACK(0x1AEA), PACK(0x18AC), PACK(0x196E),
163 PACK(0x1230), PACK(0x13F2), PACK(0x11B4), PACK(0x1076),
164 PACK(0x1538), PACK(0x14FA), PACK(0x16BC), PACK(0x177E),
165 PACK(0x3840), PACK(0x3982), PACK(0x3BC4), PACK(0x3A06),
166 PACK(0x3F48), PACK(0x3E8A), PACK(0x3CCC), PACK(0x3D0E),
167 PACK(0x3650), PACK(0x3792), PACK(0x35D4), PACK(0x3416),
168 PACK(0x3158), PACK(0x309A), PACK(0x32DC), PACK(0x331E),
169 PACK(0x2460), PACK(0x25A2), PACK(0x27E4), PACK(0x2626),
170 PACK(0x2368), PACK(0x22AA), PACK(0x20EC), PACK(0x212E),
171 PACK(0x2A70), PACK(0x2BB2), PACK(0x29F4), PACK(0x2836),
172 PACK(0x2D78), PACK(0x2CBA), PACK(0x2EFC), PACK(0x2F3E),
173 PACK(0x7080), PACK(0x7142), PACK(0x7304), PACK(0x72C6),
174 PACK(0x7788), PACK(0x764A), PACK(0x740C), PACK(0x75CE),
175 PACK(0x7E90), PACK(0x7F52), PACK(0x7D14), PACK(0x7CD6),
176 PACK(0x7998), PACK(0x785A), PACK(0x7A1C), PACK(0x7BDE),
177 PACK(0x6CA0), PACK(0x6D62), PACK(0x6F24), PACK(0x6EE6),
178 PACK(0x6BA8), PACK(0x6A6A), PACK(0x682C), PACK(0x69EE),
179 PACK(0x62B0), PACK(0x6372), PACK(0x6134), PACK(0x60F6),
180 PACK(0x65B8), PACK(0x647A), PACK(0x663C), PACK(0x67FE),
181 PACK(0x48C0), PACK(0x4902), PACK(0x4B44), PACK(0x4A86),
182 PACK(0x4FC8), PACK(0x4E0A), PACK(0x4C4C), PACK(0x4D8E),
183 PACK(0x46D0), PACK(0x4712), PACK(0x4554), PACK(0x4496),
184 PACK(0x41D8), PACK(0x401A), PACK(0x425C), PACK(0x439E),
185 PACK(0x54E0), PACK(0x5522), PACK(0x5764), PACK(0x56A6),
186 PACK(0x53E8), PACK(0x522A), PACK(0x506C), PACK(0x51AE),
187 PACK(0x5AF0), PACK(0x5B32), PACK(0x5974), PACK(0x58B6),
188 PACK(0x5DF8), PACK(0x5C3A), PACK(0x5E7C), PACK(0x5FBE),
189 PACK(0xE100), PACK(0xE0C2), PACK(0xE284), PACK(0xE346),
190 PACK(0xE608), PACK(0xE7CA), PACK(0xE58C), PACK(0xE44E),
191 PACK(0xEF10), PACK(0xEED2), PACK(0xEC94), PACK(0xED56),
192 PACK(0xE818), PACK(0xE9DA), PACK(0xEB9C), PACK(0xEA5E),
193 PACK(0xFD20), PACK(0xFCE2), PACK(0xFEA4), PACK(0xFF66),
194 PACK(0xFA28), PACK(0xFBEA), PACK(0xF9AC), PACK(0xF86E),
195 PACK(0xF330), PACK(0xF2F2), PACK(0xF0B4), PACK(0xF176),
196 PACK(0xF438), PACK(0xF5FA), PACK(0xF7BC), PACK(0xF67E),
197 PACK(0xD940), PACK(0xD882), PACK(0xDAC4), PACK(0xDB06),
198 PACK(0xDE48), PACK(0xDF8A), PACK(0xDDCC), PACK(0xDC0E),
199 PACK(0xD750), PACK(0xD692), PACK(0xD4D4), PACK(0xD516),
200 PACK(0xD058), PACK(0xD19A), PACK(0xD3DC), PACK(0xD21E),
201 PACK(0xC560), PACK(0xC4A2), PACK(0xC6E4), PACK(0xC726),
202 PACK(0xC268), PACK(0xC3AA), PACK(0xC1EC), PACK(0xC02E),
203 PACK(0xCB70), PACK(0xCAB2), PACK(0xC8F4), PACK(0xC936),
204 PACK(0xCC78), PACK(0xCDBA), PACK(0xCFFC), PACK(0xCE3E),
205 PACK(0x9180), PACK(0x9042), PACK(0x9204), PACK(0x93C6),
206 PACK(0x9688), PACK(0x974A), PACK(0x950C), PACK(0x94CE),
207 PACK(0x9F90), PACK(0x9E52), PACK(0x9C14), PACK(0x9DD6),
208 PACK(0x9898), PACK(0x995A), PACK(0x9B1C), PACK(0x9ADE),
209 PACK(0x8DA0), PACK(0x8C62), PACK(0x8E24), PACK(0x8FE6),
210 PACK(0x8AA8), PACK(0x8B6A), PACK(0x892C), PACK(0x88EE),
211 PACK(0x83B0), PACK(0x8272), PACK(0x8034), PACK(0x81F6),
212 PACK(0x84B8), PACK(0x857A), PACK(0x873C), PACK(0x86FE),
213 PACK(0xA9C0), PACK(0xA802), PACK(0xAA44), PACK(0xAB86),
214 PACK(0xAEC8), PACK(0xAF0A), PACK(0xAD4C), PACK(0xAC8E),
215 PACK(0xA7D0), PACK(0xA612), PACK(0xA454), PACK(0xA596),
216 PACK(0xA0D8), PACK(0xA11A), PACK(0xA35C), PACK(0xA29E),
217 PACK(0xB5E0), PACK(0xB422), PACK(0xB664), PACK(0xB7A6),
218 PACK(0xB2E8), PACK(0xB32A), PACK(0xB16C), PACK(0xB0AE),
219 PACK(0xBBF0), PACK(0xBA32), PACK(0xB874), PACK(0xB9B6),
220 PACK(0xBCF8), PACK(0xBD3A), PACK(0xBF7C), PACK(0xBEBE)
224 Z.hi ^= Htable[n].hi;
225 Z.lo ^= Htable[n].lo;
232 rem = (size_t)Z.lo & 0xff;
233 Z.lo = (Z.hi << 56) | (Z.lo >> 8);
235 if (sizeof(size_t) == 8)
236 Z.hi ^= rem_8bit[rem];
238 Z.hi ^= (u64)rem_8bit[rem] << 32;
241 if (is_endian.little) {
243 Xi[0] = BSWAP8(Z.hi);
244 Xi[1] = BSWAP8(Z.lo);
248 v = (u32)(Z.hi >> 32);
252 v = (u32)(Z.lo >> 32);
263 # define GCM_MUL(ctx,Xi) gcm_gmult_8bit(ctx->Xi.u,ctx->Htable)
267 static void gcm_init_4bit(u128 Htable[16], u64 H[2])
270 # if defined(OPENSSL_SMALL_FOOTPRINT)
279 # if defined(OPENSSL_SMALL_FOOTPRINT)
280 for (Htable[8] = V, i = 4; i > 0; i >>= 1) {
285 for (i = 2; i < 16; i <<= 1) {
286 u128 *Hi = Htable + i;
288 for (V = *Hi, j = 1; j < i; ++j) {
289 Hi[j].hi = V.hi ^ Htable[j].hi;
290 Hi[j].lo = V.lo ^ Htable[j].lo;
301 Htable[3].hi = V.hi ^ Htable[2].hi, Htable[3].lo = V.lo ^ Htable[2].lo;
303 Htable[5].hi = V.hi ^ Htable[1].hi, Htable[5].lo = V.lo ^ Htable[1].lo;
304 Htable[6].hi = V.hi ^ Htable[2].hi, Htable[6].lo = V.lo ^ Htable[2].lo;
305 Htable[7].hi = V.hi ^ Htable[3].hi, Htable[7].lo = V.lo ^ Htable[3].lo;
307 Htable[9].hi = V.hi ^ Htable[1].hi, Htable[9].lo = V.lo ^ Htable[1].lo;
308 Htable[10].hi = V.hi ^ Htable[2].hi, Htable[10].lo = V.lo ^ Htable[2].lo;
309 Htable[11].hi = V.hi ^ Htable[3].hi, Htable[11].lo = V.lo ^ Htable[3].lo;
310 Htable[12].hi = V.hi ^ Htable[4].hi, Htable[12].lo = V.lo ^ Htable[4].lo;
311 Htable[13].hi = V.hi ^ Htable[5].hi, Htable[13].lo = V.lo ^ Htable[5].lo;
312 Htable[14].hi = V.hi ^ Htable[6].hi, Htable[14].lo = V.lo ^ Htable[6].lo;
313 Htable[15].hi = V.hi ^ Htable[7].hi, Htable[15].lo = V.lo ^ Htable[7].lo;
315 # if defined(GHASH_ASM) && (defined(__arm__) || defined(__arm))
317 * ARM assembler expects specific dword order in Htable.
328 if (is_endian.little)
329 for (j = 0; j < 16; ++j) {
334 for (j = 0; j < 16; ++j) {
336 Htable[j].hi = V.lo << 32 | V.lo >> 32;
337 Htable[j].lo = V.hi << 32 | V.hi >> 32;
344 static const size_t rem_4bit[16] = {
345 PACK(0x0000), PACK(0x1C20), PACK(0x3840), PACK(0x2460),
346 PACK(0x7080), PACK(0x6CA0), PACK(0x48C0), PACK(0x54E0),
347 PACK(0xE100), PACK(0xFD20), PACK(0xD940), PACK(0xC560),
348 PACK(0x9180), PACK(0x8DA0), PACK(0xA9C0), PACK(0xB5E0)
351 static void gcm_gmult_4bit(u64 Xi[2], const u128 Htable[16])
355 size_t rem, nlo, nhi;
363 nlo = ((const u8 *)Xi)[15];
367 Z.hi = Htable[nlo].hi;
368 Z.lo = Htable[nlo].lo;
371 rem = (size_t)Z.lo & 0xf;
372 Z.lo = (Z.hi << 60) | (Z.lo >> 4);
374 if (sizeof(size_t) == 8)
375 Z.hi ^= rem_4bit[rem];
377 Z.hi ^= (u64)rem_4bit[rem] << 32;
379 Z.hi ^= Htable[nhi].hi;
380 Z.lo ^= Htable[nhi].lo;
385 nlo = ((const u8 *)Xi)[cnt];
389 rem = (size_t)Z.lo & 0xf;
390 Z.lo = (Z.hi << 60) | (Z.lo >> 4);
392 if (sizeof(size_t) == 8)
393 Z.hi ^= rem_4bit[rem];
395 Z.hi ^= (u64)rem_4bit[rem] << 32;
397 Z.hi ^= Htable[nlo].hi;
398 Z.lo ^= Htable[nlo].lo;
401 if (is_endian.little) {
403 Xi[0] = BSWAP8(Z.hi);
404 Xi[1] = BSWAP8(Z.lo);
408 v = (u32)(Z.hi >> 32);
412 v = (u32)(Z.lo >> 32);
423 # if !defined(OPENSSL_SMALL_FOOTPRINT)
425 * Streamed gcm_mult_4bit, see CRYPTO_gcm128_[en|de]crypt for
426 * details... Compiler-generated code doesn't seem to give any
427 * performance improvement, at least not on x86[_64]. It's here
428 * mostly as reference and a placeholder for possible future
429 * non-trivial optimization[s]...
431 static void gcm_ghash_4bit(u64 Xi[2], const u128 Htable[16],
432 const u8 *inp, size_t len)
436 size_t rem, nlo, nhi;
447 nlo = ((const u8 *)Xi)[15];
452 Z.hi = Htable[nlo].hi;
453 Z.lo = Htable[nlo].lo;
456 rem = (size_t)Z.lo & 0xf;
457 Z.lo = (Z.hi << 60) | (Z.lo >> 4);
459 if (sizeof(size_t) == 8)
460 Z.hi ^= rem_4bit[rem];
462 Z.hi ^= (u64)rem_4bit[rem] << 32;
464 Z.hi ^= Htable[nhi].hi;
465 Z.lo ^= Htable[nhi].lo;
470 nlo = ((const u8 *)Xi)[cnt];
475 rem = (size_t)Z.lo & 0xf;
476 Z.lo = (Z.hi << 60) | (Z.lo >> 4);
478 if (sizeof(size_t) == 8)
479 Z.hi ^= rem_4bit[rem];
481 Z.hi ^= (u64)rem_4bit[rem] << 32;
483 Z.hi ^= Htable[nlo].hi;
484 Z.lo ^= Htable[nlo].lo;
488 * Extra 256+16 bytes per-key plus 512 bytes shared tables
489 * [should] give ~50% improvement... One could have PACK()-ed
490 * the rem_8bit even here, but the priority is to minimize
493 u128 Hshr4[16]; /* Htable shifted right by 4 bits */
494 u8 Hshl4[16]; /* Htable shifted left by 4 bits */
495 static const unsigned short rem_8bit[256] = {
496 0x0000, 0x01C2, 0x0384, 0x0246, 0x0708, 0x06CA, 0x048C, 0x054E,
497 0x0E10, 0x0FD2, 0x0D94, 0x0C56, 0x0918, 0x08DA, 0x0A9C, 0x0B5E,
498 0x1C20, 0x1DE2, 0x1FA4, 0x1E66, 0x1B28, 0x1AEA, 0x18AC, 0x196E,
499 0x1230, 0x13F2, 0x11B4, 0x1076, 0x1538, 0x14FA, 0x16BC, 0x177E,
500 0x3840, 0x3982, 0x3BC4, 0x3A06, 0x3F48, 0x3E8A, 0x3CCC, 0x3D0E,
501 0x3650, 0x3792, 0x35D4, 0x3416, 0x3158, 0x309A, 0x32DC, 0x331E,
502 0x2460, 0x25A2, 0x27E4, 0x2626, 0x2368, 0x22AA, 0x20EC, 0x212E,
503 0x2A70, 0x2BB2, 0x29F4, 0x2836, 0x2D78, 0x2CBA, 0x2EFC, 0x2F3E,
504 0x7080, 0x7142, 0x7304, 0x72C6, 0x7788, 0x764A, 0x740C, 0x75CE,
505 0x7E90, 0x7F52, 0x7D14, 0x7CD6, 0x7998, 0x785A, 0x7A1C, 0x7BDE,
506 0x6CA0, 0x6D62, 0x6F24, 0x6EE6, 0x6BA8, 0x6A6A, 0x682C, 0x69EE,
507 0x62B0, 0x6372, 0x6134, 0x60F6, 0x65B8, 0x647A, 0x663C, 0x67FE,
508 0x48C0, 0x4902, 0x4B44, 0x4A86, 0x4FC8, 0x4E0A, 0x4C4C, 0x4D8E,
509 0x46D0, 0x4712, 0x4554, 0x4496, 0x41D8, 0x401A, 0x425C, 0x439E,
510 0x54E0, 0x5522, 0x5764, 0x56A6, 0x53E8, 0x522A, 0x506C, 0x51AE,
511 0x5AF0, 0x5B32, 0x5974, 0x58B6, 0x5DF8, 0x5C3A, 0x5E7C, 0x5FBE,
512 0xE100, 0xE0C2, 0xE284, 0xE346, 0xE608, 0xE7CA, 0xE58C, 0xE44E,
513 0xEF10, 0xEED2, 0xEC94, 0xED56, 0xE818, 0xE9DA, 0xEB9C, 0xEA5E,
514 0xFD20, 0xFCE2, 0xFEA4, 0xFF66, 0xFA28, 0xFBEA, 0xF9AC, 0xF86E,
515 0xF330, 0xF2F2, 0xF0B4, 0xF176, 0xF438, 0xF5FA, 0xF7BC, 0xF67E,
516 0xD940, 0xD882, 0xDAC4, 0xDB06, 0xDE48, 0xDF8A, 0xDDCC, 0xDC0E,
517 0xD750, 0xD692, 0xD4D4, 0xD516, 0xD058, 0xD19A, 0xD3DC, 0xD21E,
518 0xC560, 0xC4A2, 0xC6E4, 0xC726, 0xC268, 0xC3AA, 0xC1EC, 0xC02E,
519 0xCB70, 0xCAB2, 0xC8F4, 0xC936, 0xCC78, 0xCDBA, 0xCFFC, 0xCE3E,
520 0x9180, 0x9042, 0x9204, 0x93C6, 0x9688, 0x974A, 0x950C, 0x94CE,
521 0x9F90, 0x9E52, 0x9C14, 0x9DD6, 0x9898, 0x995A, 0x9B1C, 0x9ADE,
522 0x8DA0, 0x8C62, 0x8E24, 0x8FE6, 0x8AA8, 0x8B6A, 0x892C, 0x88EE,
523 0x83B0, 0x8272, 0x8034, 0x81F6, 0x84B8, 0x857A, 0x873C, 0x86FE,
524 0xA9C0, 0xA802, 0xAA44, 0xAB86, 0xAEC8, 0xAF0A, 0xAD4C, 0xAC8E,
525 0xA7D0, 0xA612, 0xA454, 0xA596, 0xA0D8, 0xA11A, 0xA35C, 0xA29E,
526 0xB5E0, 0xB422, 0xB664, 0xB7A6, 0xB2E8, 0xB32A, 0xB16C, 0xB0AE,
527 0xBBF0, 0xBA32, 0xB874, 0xB9B6, 0xBCF8, 0xBD3A, 0xBF7C, 0xBEBE
530 * This pre-processing phase slows down procedure by approximately
531 * same time as it makes each loop spin faster. In other words
532 * single block performance is approximately same as straightforward
533 * "4-bit" implementation, and then it goes only faster...
535 for (cnt = 0; cnt < 16; ++cnt) {
536 Z.hi = Htable[cnt].hi;
537 Z.lo = Htable[cnt].lo;
538 Hshr4[cnt].lo = (Z.hi << 60) | (Z.lo >> 4);
539 Hshr4[cnt].hi = (Z.hi >> 4);
540 Hshl4[cnt] = (u8)(Z.lo << 4);
544 for (Z.lo = 0, Z.hi = 0, cnt = 15; cnt; --cnt) {
545 nlo = ((const u8 *)Xi)[cnt];
550 Z.hi ^= Htable[nlo].hi;
551 Z.lo ^= Htable[nlo].lo;
553 rem = (size_t)Z.lo & 0xff;
555 Z.lo = (Z.hi << 56) | (Z.lo >> 8);
558 Z.hi ^= Hshr4[nhi].hi;
559 Z.lo ^= Hshr4[nhi].lo;
560 Z.hi ^= (u64)rem_8bit[rem ^ Hshl4[nhi]] << 48;
563 nlo = ((const u8 *)Xi)[0];
568 Z.hi ^= Htable[nlo].hi;
569 Z.lo ^= Htable[nlo].lo;
571 rem = (size_t)Z.lo & 0xf;
573 Z.lo = (Z.hi << 60) | (Z.lo >> 4);
576 Z.hi ^= Htable[nhi].hi;
577 Z.lo ^= Htable[nhi].lo;
578 Z.hi ^= ((u64)rem_8bit[rem << 4]) << 48;
581 if (is_endian.little) {
583 Xi[0] = BSWAP8(Z.hi);
584 Xi[1] = BSWAP8(Z.lo);
588 v = (u32)(Z.hi >> 32);
592 v = (u32)(Z.lo >> 32);
601 } while (inp += 16, len -= 16);
605 void gcm_gmult_4bit(u64 Xi[2], const u128 Htable[16]);
606 void gcm_ghash_4bit(u64 Xi[2], const u128 Htable[16], const u8 *inp,
610 # define GCM_MUL(ctx,Xi) gcm_gmult_4bit(ctx->Xi.u,ctx->Htable)
611 # if defined(GHASH_ASM) || !defined(OPENSSL_SMALL_FOOTPRINT)
612 # define GHASH(ctx,in,len) gcm_ghash_4bit((ctx)->Xi.u,(ctx)->Htable,in,len)
614 * GHASH_CHUNK is "stride parameter" missioned to mitigate cache trashing
615 * effect. In other words idea is to hash data while it's still in L1 cache
616 * after encryption pass...
618 # define GHASH_CHUNK (3*1024)
621 #else /* TABLE_BITS */
623 static void gcm_gmult_1bit(u64 Xi[2], const u64 H[2])
625 u128 V, Z = { 0, 0 };
628 const long *xi = (const long *)Xi;
636 V.hi = H[0]; /* H is in host byte order, no byte swapping */
639 for (j = 0; j < 16 / sizeof(long); ++j) {
640 if (is_endian.little) {
641 if (sizeof(long) == 8) {
643 X = (long)(BSWAP8(xi[j]));
645 const u8 *p = (const u8 *)(xi + j);
646 X = (long)((u64)GETU32(p) << 32 | GETU32(p + 4));
649 const u8 *p = (const u8 *)(xi + j);
655 for (i = 0; i < 8 * sizeof(long); ++i, X <<= 1) {
656 u64 M = (u64)(X >> (8 * sizeof(long) - 1));
664 if (is_endian.little) {
666 Xi[0] = BSWAP8(Z.hi);
667 Xi[1] = BSWAP8(Z.lo);
671 v = (u32)(Z.hi >> 32);
675 v = (u32)(Z.lo >> 32);
686 # define GCM_MUL(ctx,Xi) gcm_gmult_1bit(ctx->Xi.u,ctx->H.u)
690 #if TABLE_BITS==4 && defined(GHASH_ASM)
691 # if !defined(I386_ONLY) && \
692 (defined(__i386) || defined(__i386__) || \
693 defined(__x86_64) || defined(__x86_64__) || \
694 defined(_M_IX86) || defined(_M_AMD64) || defined(_M_X64))
695 # define GHASH_ASM_X86_OR_64
696 # define GCM_FUNCREF_4BIT
697 extern unsigned int OPENSSL_ia32cap_P[2];
699 void gcm_init_clmul(u128 Htable[16], const u64 Xi[2]);
700 void gcm_gmult_clmul(u64 Xi[2], const u128 Htable[16]);
701 void gcm_ghash_clmul(u64 Xi[2], const u128 Htable[16], const u8 *inp,
704 # if defined(__i386) || defined(__i386__) || defined(_M_IX86)
705 # define GHASH_ASM_X86
706 void gcm_gmult_4bit_mmx(u64 Xi[2], const u128 Htable[16]);
707 void gcm_ghash_4bit_mmx(u64 Xi[2], const u128 Htable[16], const u8 *inp,
710 void gcm_gmult_4bit_x86(u64 Xi[2], const u128 Htable[16]);
711 void gcm_ghash_4bit_x86(u64 Xi[2], const u128 Htable[16], const u8 *inp,
714 # elif defined(__arm__) || defined(__arm)
715 # include "arm_arch.h"
717 # define GHASH_ASM_ARM
718 # define GCM_FUNCREF_4BIT
719 void gcm_gmult_neon(u64 Xi[2], const u128 Htable[16]);
720 void gcm_ghash_neon(u64 Xi[2], const u128 Htable[16], const u8 *inp,
726 #ifdef GCM_FUNCREF_4BIT
728 # define GCM_MUL(ctx,Xi) (*gcm_gmult_p)(ctx->Xi.u,ctx->Htable)
731 # define GHASH(ctx,in,len) (*gcm_ghash_p)(ctx->Xi.u,ctx->Htable,in,len)
735 void CRYPTO_gcm128_init(GCM128_CONTEXT *ctx, void *key, block128_f block)
744 memset(ctx, 0, sizeof(*ctx));
748 (*block) (ctx->H.c, ctx->H.c, key);
750 if (is_endian.little) {
751 /* H is stored in host byte order */
753 ctx->H.u[0] = BSWAP8(ctx->H.u[0]);
754 ctx->H.u[1] = BSWAP8(ctx->H.u[1]);
758 hi = (u64)GETU32(p) << 32 | GETU32(p + 4);
759 lo = (u64)GETU32(p + 8) << 32 | GETU32(p + 12);
765 gcm_init_8bit(ctx->Htable, ctx->H.u);
767 # if defined(GHASH_ASM_X86_OR_64)
768 # if !defined(GHASH_ASM_X86) || defined(OPENSSL_IA32_SSE2)
769 if (OPENSSL_ia32cap_P[0] & (1 << 24) && /* check FXSR bit */
770 OPENSSL_ia32cap_P[1] & (1 << 1)) { /* check PCLMULQDQ bit */
771 gcm_init_clmul(ctx->Htable, ctx->H.u);
772 ctx->gmult = gcm_gmult_clmul;
773 ctx->ghash = gcm_ghash_clmul;
777 gcm_init_4bit(ctx->Htable, ctx->H.u);
778 # if defined(GHASH_ASM_X86) /* x86 only */
779 # if defined(OPENSSL_IA32_SSE2)
780 if (OPENSSL_ia32cap_P[0] & (1 << 25)) { /* check SSE bit */
782 if (OPENSSL_ia32cap_P[0] & (1 << 23)) { /* check MMX bit */
784 ctx->gmult = gcm_gmult_4bit_mmx;
785 ctx->ghash = gcm_ghash_4bit_mmx;
787 ctx->gmult = gcm_gmult_4bit_x86;
788 ctx->ghash = gcm_ghash_4bit_x86;
791 ctx->gmult = gcm_gmult_4bit;
792 ctx->ghash = gcm_ghash_4bit;
794 # elif defined(GHASH_ASM_ARM)
795 if (OPENSSL_armcap_P & ARMV7_NEON) {
796 ctx->gmult = gcm_gmult_neon;
797 ctx->ghash = gcm_ghash_neon;
799 gcm_init_4bit(ctx->Htable, ctx->H.u);
800 ctx->gmult = gcm_gmult_4bit;
801 ctx->ghash = gcm_ghash_4bit;
804 gcm_init_4bit(ctx->Htable, ctx->H.u);
809 void CRYPTO_gcm128_setiv(GCM128_CONTEXT *ctx, const unsigned char *iv,
819 #ifdef GCM_FUNCREF_4BIT
820 void (*gcm_gmult_p) (u64 Xi[2], const u128 Htable[16]) = ctx->gmult;
827 ctx->len.u[0] = 0; /* AAD length */
828 ctx->len.u[1] = 0; /* message length */
833 memcpy(ctx->Yi.c, iv, 12);
841 for (i = 0; i < 16; ++i)
842 ctx->Yi.c[i] ^= iv[i];
848 for (i = 0; i < len; ++i)
849 ctx->Yi.c[i] ^= iv[i];
853 if (is_endian.little) {
855 ctx->Yi.u[1] ^= BSWAP8(len0);
857 ctx->Yi.c[8] ^= (u8)(len0 >> 56);
858 ctx->Yi.c[9] ^= (u8)(len0 >> 48);
859 ctx->Yi.c[10] ^= (u8)(len0 >> 40);
860 ctx->Yi.c[11] ^= (u8)(len0 >> 32);
861 ctx->Yi.c[12] ^= (u8)(len0 >> 24);
862 ctx->Yi.c[13] ^= (u8)(len0 >> 16);
863 ctx->Yi.c[14] ^= (u8)(len0 >> 8);
864 ctx->Yi.c[15] ^= (u8)(len0);
867 ctx->Yi.u[1] ^= len0;
871 if (is_endian.little)
873 ctr = BSWAP4(ctx->Yi.d[3]);
875 ctr = GETU32(ctx->Yi.c + 12);
881 (*ctx->block) (ctx->Yi.c, ctx->EK0.c, ctx->key);
883 if (is_endian.little)
885 ctx->Yi.d[3] = BSWAP4(ctr);
887 PUTU32(ctx->Yi.c + 12, ctr);
893 int CRYPTO_gcm128_aad(GCM128_CONTEXT *ctx, const unsigned char *aad,
898 u64 alen = ctx->len.u[0];
899 #ifdef GCM_FUNCREF_4BIT
900 void (*gcm_gmult_p) (u64 Xi[2], const u128 Htable[16]) = ctx->gmult;
902 void (*gcm_ghash_p) (u64 Xi[2], const u128 Htable[16],
903 const u8 *inp, size_t len) = ctx->ghash;
911 if (alen > (U64(1) << 61) || (sizeof(len) == 8 && alen < len))
913 ctx->len.u[0] = alen;
918 ctx->Xi.c[n] ^= *(aad++);
930 if ((i = (len & (size_t)-16))) {
937 for (i = 0; i < 16; ++i)
938 ctx->Xi.c[i] ^= aad[i];
945 n = (unsigned int)len;
946 for (i = 0; i < len; ++i)
947 ctx->Xi.c[i] ^= aad[i];
954 int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx,
955 const unsigned char *in, unsigned char *out,
966 u64 mlen = ctx->len.u[1];
967 block128_f block = ctx->block;
968 void *key = ctx->key;
969 #ifdef GCM_FUNCREF_4BIT
970 void (*gcm_gmult_p) (u64 Xi[2], const u128 Htable[16]) = ctx->gmult;
972 void (*gcm_ghash_p) (u64 Xi[2], const u128 Htable[16],
973 const u8 *inp, size_t len) = ctx->ghash;
978 n = (unsigned int)mlen % 16; /* alternative to ctx->mres */
981 if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len))
983 ctx->len.u[1] = mlen;
986 /* First call to encrypt finalizes GHASH(AAD) */
991 if (is_endian.little)
993 ctr = BSWAP4(ctx->Yi.d[3]);
995 ctr = GETU32(ctx->Yi.c + 12);
1001 #if !defined(OPENSSL_SMALL_FOOTPRINT)
1002 if (16 % sizeof(size_t) == 0) { /* always true actually */
1006 ctx->Xi.c[n] ^= *(out++) = *(in++) ^ ctx->EKi.c[n];
1017 # if defined(STRICT_ALIGNMENT)
1018 if (((size_t)in | (size_t)out) % sizeof(size_t) != 0)
1021 # if defined(GHASH) && defined(GHASH_CHUNK)
1022 while (len >= GHASH_CHUNK) {
1023 size_t j = GHASH_CHUNK;
1026 size_t *out_t = (size_t *)out;
1027 const size_t *in_t = (const size_t *)in;
1029 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1031 if (is_endian.little)
1033 ctx->Yi.d[3] = BSWAP4(ctr);
1035 PUTU32(ctx->Yi.c + 12, ctr);
1039 for (i = 0; i < 16 / sizeof(size_t); ++i)
1040 out_t[i] = in_t[i] ^ ctx->EKi.t[i];
1045 GHASH(ctx, out - GHASH_CHUNK, GHASH_CHUNK);
1048 if ((i = (len & (size_t)-16))) {
1052 size_t *out_t = (size_t *)out;
1053 const size_t *in_t = (const size_t *)in;
1055 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1057 if (is_endian.little)
1059 ctx->Yi.d[3] = BSWAP4(ctr);
1061 PUTU32(ctx->Yi.c + 12, ctr);
1065 for (i = 0; i < 16 / sizeof(size_t); ++i)
1066 out_t[i] = in_t[i] ^ ctx->EKi.t[i];
1071 GHASH(ctx, out - j, j);
1075 size_t *out_t = (size_t *)out;
1076 const size_t *in_t = (const size_t *)in;
1078 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1080 if (is_endian.little)
1082 ctx->Yi.d[3] = BSWAP4(ctr);
1084 PUTU32(ctx->Yi.c + 12, ctr);
1088 for (i = 0; i < 16 / sizeof(size_t); ++i)
1089 ctx->Xi.t[i] ^= out_t[i] = in_t[i] ^ ctx->EKi.t[i];
1097 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1099 if (is_endian.little)
1101 ctx->Yi.d[3] = BSWAP4(ctr);
1103 PUTU32(ctx->Yi.c + 12, ctr);
1108 ctx->Xi.c[n] ^= out[n] = in[n] ^ ctx->EKi.c[n];
1118 for (i = 0; i < len; ++i) {
1120 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1122 if (is_endian.little)
1124 ctx->Yi.d[3] = BSWAP4(ctr);
1126 PUTU32(ctx->Yi.c + 12, ctr);
1131 ctx->Xi.c[n] ^= out[i] = in[i] ^ ctx->EKi.c[n];
1141 int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx,
1142 const unsigned char *in, unsigned char *out,
1151 unsigned int n, ctr;
1153 u64 mlen = ctx->len.u[1];
1154 block128_f block = ctx->block;
1155 void *key = ctx->key;
1156 #ifdef GCM_FUNCREF_4BIT
1157 void (*gcm_gmult_p) (u64 Xi[2], const u128 Htable[16]) = ctx->gmult;
1159 void (*gcm_ghash_p) (u64 Xi[2], const u128 Htable[16],
1160 const u8 *inp, size_t len) = ctx->ghash;
1165 if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len))
1167 ctx->len.u[1] = mlen;
1170 /* First call to decrypt finalizes GHASH(AAD) */
1175 if (is_endian.little)
1177 ctr = BSWAP4(ctx->Yi.d[3]);
1179 ctr = GETU32(ctx->Yi.c + 12);
1185 #if !defined(OPENSSL_SMALL_FOOTPRINT)
1186 if (16 % sizeof(size_t) == 0) { /* always true actually */
1191 *(out++) = c ^ ctx->EKi.c[n];
1203 # if defined(STRICT_ALIGNMENT)
1204 if (((size_t)in | (size_t)out) % sizeof(size_t) != 0)
1207 # if defined(GHASH) && defined(GHASH_CHUNK)
1208 while (len >= GHASH_CHUNK) {
1209 size_t j = GHASH_CHUNK;
1211 GHASH(ctx, in, GHASH_CHUNK);
1213 size_t *out_t = (size_t *)out;
1214 const size_t *in_t = (const size_t *)in;
1216 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1218 if (is_endian.little)
1220 ctx->Yi.d[3] = BSWAP4(ctr);
1222 PUTU32(ctx->Yi.c + 12, ctr);
1226 for (i = 0; i < 16 / sizeof(size_t); ++i)
1227 out_t[i] = in_t[i] ^ ctx->EKi.t[i];
1234 if ((i = (len & (size_t)-16))) {
1237 size_t *out_t = (size_t *)out;
1238 const size_t *in_t = (const size_t *)in;
1240 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1242 if (is_endian.little)
1244 ctx->Yi.d[3] = BSWAP4(ctr);
1246 PUTU32(ctx->Yi.c + 12, ctr);
1250 for (i = 0; i < 16 / sizeof(size_t); ++i)
1251 out_t[i] = in_t[i] ^ ctx->EKi.t[i];
1259 size_t *out_t = (size_t *)out;
1260 const size_t *in_t = (const size_t *)in;
1262 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1264 if (is_endian.little)
1266 ctx->Yi.d[3] = BSWAP4(ctr);
1268 PUTU32(ctx->Yi.c + 12, ctr);
1272 for (i = 0; i < 16 / sizeof(size_t); ++i) {
1274 out[i] = c ^ ctx->EKi.t[i];
1284 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1286 if (is_endian.little)
1288 ctx->Yi.d[3] = BSWAP4(ctr);
1290 PUTU32(ctx->Yi.c + 12, ctr);
1297 out[n] = c ^ ctx->EKi.c[n];
1307 for (i = 0; i < len; ++i) {
1310 (*block) (ctx->Yi.c, ctx->EKi.c, key);
1312 if (is_endian.little)
1314 ctx->Yi.d[3] = BSWAP4(ctr);
1316 PUTU32(ctx->Yi.c + 12, ctr);
1322 out[i] = c ^ ctx->EKi.c[n];
1333 int CRYPTO_gcm128_encrypt_ctr32(GCM128_CONTEXT *ctx,
1334 const unsigned char *in, unsigned char *out,
1335 size_t len, ctr128_f stream)
1343 unsigned int n, ctr;
1345 u64 mlen = ctx->len.u[1];
1346 void *key = ctx->key;
1347 #ifdef GCM_FUNCREF_4BIT
1348 void (*gcm_gmult_p) (u64 Xi[2], const u128 Htable[16]) = ctx->gmult;
1350 void (*gcm_ghash_p) (u64 Xi[2], const u128 Htable[16],
1351 const u8 *inp, size_t len) = ctx->ghash;
1356 if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len))
1358 ctx->len.u[1] = mlen;
1361 /* First call to encrypt finalizes GHASH(AAD) */
1366 if (is_endian.little)
1368 ctr = BSWAP4(ctx->Yi.d[3]);
1370 ctr = GETU32(ctx->Yi.c + 12);
1378 ctx->Xi.c[n] ^= *(out++) = *(in++) ^ ctx->EKi.c[n];
1389 #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT)
1390 while (len >= GHASH_CHUNK) {
1391 (*stream) (in, out, GHASH_CHUNK / 16, key, ctx->Yi.c);
1392 ctr += GHASH_CHUNK / 16;
1393 if (is_endian.little)
1395 ctx->Yi.d[3] = BSWAP4(ctr);
1397 PUTU32(ctx->Yi.c + 12, ctr);
1401 GHASH(ctx, out, GHASH_CHUNK);
1407 if ((i = (len & (size_t)-16))) {
1410 (*stream) (in, out, j, key, ctx->Yi.c);
1411 ctr += (unsigned int)j;
1412 if (is_endian.little)
1414 ctx->Yi.d[3] = BSWAP4(ctr);
1416 PUTU32(ctx->Yi.c + 12, ctr);
1427 for (i = 0; i < 16; ++i)
1428 ctx->Xi.c[i] ^= out[i];
1435 (*ctx->block) (ctx->Yi.c, ctx->EKi.c, key);
1437 if (is_endian.little)
1439 ctx->Yi.d[3] = BSWAP4(ctr);
1441 PUTU32(ctx->Yi.c + 12, ctr);
1446 ctx->Xi.c[n] ^= out[n] = in[n] ^ ctx->EKi.c[n];
1455 int CRYPTO_gcm128_decrypt_ctr32(GCM128_CONTEXT *ctx,
1456 const unsigned char *in, unsigned char *out,
1457 size_t len, ctr128_f stream)
1465 unsigned int n, ctr;
1467 u64 mlen = ctx->len.u[1];
1468 void *key = ctx->key;
1469 #ifdef GCM_FUNCREF_4BIT
1470 void (*gcm_gmult_p) (u64 Xi[2], const u128 Htable[16]) = ctx->gmult;
1472 void (*gcm_ghash_p) (u64 Xi[2], const u128 Htable[16],
1473 const u8 *inp, size_t len) = ctx->ghash;
1478 if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len))
1480 ctx->len.u[1] = mlen;
1483 /* First call to decrypt finalizes GHASH(AAD) */
1488 if (is_endian.little)
1490 ctr = BSWAP4(ctx->Yi.d[3]);
1492 ctr = GETU32(ctx->Yi.c + 12);
1501 *(out++) = c ^ ctx->EKi.c[n];
1513 #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT)
1514 while (len >= GHASH_CHUNK) {
1515 GHASH(ctx, in, GHASH_CHUNK);
1516 (*stream) (in, out, GHASH_CHUNK / 16, key, ctx->Yi.c);
1517 ctr += GHASH_CHUNK / 16;
1518 if (is_endian.little)
1520 ctx->Yi.d[3] = BSWAP4(ctr);
1522 PUTU32(ctx->Yi.c + 12, ctr);
1531 if ((i = (len & (size_t)-16))) {
1539 for (k = 0; k < 16; ++k)
1540 ctx->Xi.c[k] ^= in[k];
1547 (*stream) (in, out, j, key, ctx->Yi.c);
1548 ctr += (unsigned int)j;
1549 if (is_endian.little)
1551 ctx->Yi.d[3] = BSWAP4(ctr);
1553 PUTU32(ctx->Yi.c + 12, ctr);
1562 (*ctx->block) (ctx->Yi.c, ctx->EKi.c, key);
1564 if (is_endian.little)
1566 ctx->Yi.d[3] = BSWAP4(ctr);
1568 PUTU32(ctx->Yi.c + 12, ctr);
1575 out[n] = c ^ ctx->EKi.c[n];
1584 int CRYPTO_gcm128_finish(GCM128_CONTEXT *ctx, const unsigned char *tag,
1593 u64 alen = ctx->len.u[0] << 3;
1594 u64 clen = ctx->len.u[1] << 3;
1595 #ifdef GCM_FUNCREF_4BIT
1596 void (*gcm_gmult_p) (u64 Xi[2], const u128 Htable[16]) = ctx->gmult;
1599 if (ctx->mres || ctx->ares)
1602 if (is_endian.little) {
1604 alen = BSWAP8(alen);
1605 clen = BSWAP8(clen);
1609 ctx->len.u[0] = alen;
1610 ctx->len.u[1] = clen;
1612 alen = (u64)GETU32(p) << 32 | GETU32(p + 4);
1613 clen = (u64)GETU32(p + 8) << 32 | GETU32(p + 12);
1617 ctx->Xi.u[0] ^= alen;
1618 ctx->Xi.u[1] ^= clen;
1621 ctx->Xi.u[0] ^= ctx->EK0.u[0];
1622 ctx->Xi.u[1] ^= ctx->EK0.u[1];
1624 if (tag && len <= sizeof(ctx->Xi))
1625 return CRYPTO_memcmp(ctx->Xi.c, tag, len);
1630 void CRYPTO_gcm128_tag(GCM128_CONTEXT *ctx, unsigned char *tag, size_t len)
1632 CRYPTO_gcm128_finish(ctx, NULL, 0);
1633 memcpy(tag, ctx->Xi.c,
1634 len <= sizeof(ctx->Xi.c) ? len : sizeof(ctx->Xi.c));
1637 GCM128_CONTEXT *CRYPTO_gcm128_new(void *key, block128_f block)
1639 GCM128_CONTEXT *ret;
1641 if ((ret = (GCM128_CONTEXT *)OPENSSL_malloc(sizeof(GCM128_CONTEXT))))
1642 CRYPTO_gcm128_init(ret, key, block);
1647 void CRYPTO_gcm128_release(GCM128_CONTEXT *ctx)
1650 OPENSSL_cleanse(ctx, sizeof(*ctx));
1655 #if defined(SELFTEST)
1657 # include <openssl/aes.h>
1660 static const u8 K1[16], *P1 = NULL, *A1 = NULL, IV1[12], *C1 = NULL;
1661 static const u8 T1[] = {
1662 0x58, 0xe2, 0xfc, 0xce, 0xfa, 0x7e, 0x30, 0x61,
1663 0x36, 0x7f, 0x1d, 0x57, 0xa4, 0xe7, 0x45, 0x5a
1670 static const u8 P2[16];
1671 static const u8 C2[] = {
1672 0x03, 0x88, 0xda, 0xce, 0x60, 0xb6, 0xa3, 0x92,
1673 0xf3, 0x28, 0xc2, 0xb9, 0x71, 0xb2, 0xfe, 0x78
1676 static const u8 T2[] = {
1677 0xab, 0x6e, 0x47, 0xd4, 0x2c, 0xec, 0x13, 0xbd,
1678 0xf5, 0x3a, 0x67, 0xb2, 0x12, 0x57, 0xbd, 0xdf
1683 static const u8 K3[] = {
1684 0xfe, 0xff, 0xe9, 0x92, 0x86, 0x65, 0x73, 0x1c,
1685 0x6d, 0x6a, 0x8f, 0x94, 0x67, 0x30, 0x83, 0x08
1688 static const u8 P3[] = {
1689 0xd9, 0x31, 0x32, 0x25, 0xf8, 0x84, 0x06, 0xe5,
1690 0xa5, 0x59, 0x09, 0xc5, 0xaf, 0xf5, 0x26, 0x9a,
1691 0x86, 0xa7, 0xa9, 0x53, 0x15, 0x34, 0xf7, 0xda,
1692 0x2e, 0x4c, 0x30, 0x3d, 0x8a, 0x31, 0x8a, 0x72,
1693 0x1c, 0x3c, 0x0c, 0x95, 0x95, 0x68, 0x09, 0x53,
1694 0x2f, 0xcf, 0x0e, 0x24, 0x49, 0xa6, 0xb5, 0x25,
1695 0xb1, 0x6a, 0xed, 0xf5, 0xaa, 0x0d, 0xe6, 0x57,
1696 0xba, 0x63, 0x7b, 0x39, 0x1a, 0xaf, 0xd2, 0x55
1699 static const u8 IV3[] = {
1700 0xca, 0xfe, 0xba, 0xbe, 0xfa, 0xce, 0xdb, 0xad,
1701 0xde, 0xca, 0xf8, 0x88
1704 static const u8 C3[] = {
1705 0x42, 0x83, 0x1e, 0xc2, 0x21, 0x77, 0x74, 0x24,
1706 0x4b, 0x72, 0x21, 0xb7, 0x84, 0xd0, 0xd4, 0x9c,
1707 0xe3, 0xaa, 0x21, 0x2f, 0x2c, 0x02, 0xa4, 0xe0,
1708 0x35, 0xc1, 0x7e, 0x23, 0x29, 0xac, 0xa1, 0x2e,
1709 0x21, 0xd5, 0x14, 0xb2, 0x54, 0x66, 0x93, 0x1c,
1710 0x7d, 0x8f, 0x6a, 0x5a, 0xac, 0x84, 0xaa, 0x05,
1711 0x1b, 0xa3, 0x0b, 0x39, 0x6a, 0x0a, 0xac, 0x97,
1712 0x3d, 0x58, 0xe0, 0x91, 0x47, 0x3f, 0x59, 0x85
1715 static const u8 T3[] = {
1716 0x4d, 0x5c, 0x2a, 0xf3, 0x27, 0xcd, 0x64, 0xa6,
1717 0x2c, 0xf3, 0x5a, 0xbd, 0x2b, 0xa6, 0xfa, 0xb4
1723 static const u8 P4[] = {
1724 0xd9, 0x31, 0x32, 0x25, 0xf8, 0x84, 0x06, 0xe5,
1725 0xa5, 0x59, 0x09, 0xc5, 0xaf, 0xf5, 0x26, 0x9a,
1726 0x86, 0xa7, 0xa9, 0x53, 0x15, 0x34, 0xf7, 0xda,
1727 0x2e, 0x4c, 0x30, 0x3d, 0x8a, 0x31, 0x8a, 0x72,
1728 0x1c, 0x3c, 0x0c, 0x95, 0x95, 0x68, 0x09, 0x53,
1729 0x2f, 0xcf, 0x0e, 0x24, 0x49, 0xa6, 0xb5, 0x25,
1730 0xb1, 0x6a, 0xed, 0xf5, 0xaa, 0x0d, 0xe6, 0x57,
1731 0xba, 0x63, 0x7b, 0x39
1734 static const u8 A4[] = {
1735 0xfe, 0xed, 0xfa, 0xce, 0xde, 0xad, 0xbe, 0xef,
1736 0xfe, 0xed, 0xfa, 0xce, 0xde, 0xad, 0xbe, 0xef,
1737 0xab, 0xad, 0xda, 0xd2
1740 static const u8 C4[] = {
1741 0x42, 0x83, 0x1e, 0xc2, 0x21, 0x77, 0x74, 0x24,
1742 0x4b, 0x72, 0x21, 0xb7, 0x84, 0xd0, 0xd4, 0x9c,
1743 0xe3, 0xaa, 0x21, 0x2f, 0x2c, 0x02, 0xa4, 0xe0,
1744 0x35, 0xc1, 0x7e, 0x23, 0x29, 0xac, 0xa1, 0x2e,
1745 0x21, 0xd5, 0x14, 0xb2, 0x54, 0x66, 0x93, 0x1c,
1746 0x7d, 0x8f, 0x6a, 0x5a, 0xac, 0x84, 0xaa, 0x05,
1747 0x1b, 0xa3, 0x0b, 0x39, 0x6a, 0x0a, 0xac, 0x97,
1748 0x3d, 0x58, 0xe0, 0x91
1751 static const u8 T4[] = {
1752 0x5b, 0xc9, 0x4f, 0xbc, 0x32, 0x21, 0xa5, 0xdb,
1753 0x94, 0xfa, 0xe9, 0x5a, 0xe7, 0x12, 0x1a, 0x47
1760 static const u8 IV5[] = {
1761 0xca, 0xfe, 0xba, 0xbe, 0xfa, 0xce, 0xdb, 0xad
1764 static const u8 C5[] = {
1765 0x61, 0x35, 0x3b, 0x4c, 0x28, 0x06, 0x93, 0x4a,
1766 0x77, 0x7f, 0xf5, 0x1f, 0xa2, 0x2a, 0x47, 0x55,
1767 0x69, 0x9b, 0x2a, 0x71, 0x4f, 0xcd, 0xc6, 0xf8,
1768 0x37, 0x66, 0xe5, 0xf9, 0x7b, 0x6c, 0x74, 0x23,
1769 0x73, 0x80, 0x69, 0x00, 0xe4, 0x9f, 0x24, 0xb2,
1770 0x2b, 0x09, 0x75, 0x44, 0xd4, 0x89, 0x6b, 0x42,
1771 0x49, 0x89, 0xb5, 0xe1, 0xeb, 0xac, 0x0f, 0x07,
1772 0xc2, 0x3f, 0x45, 0x98
1775 static const u8 T5[] = {
1776 0x36, 0x12, 0xd2, 0xe7, 0x9e, 0x3b, 0x07, 0x85,
1777 0x56, 0x1b, 0xe1, 0x4a, 0xac, 0xa2, 0xfc, 0xcb
1784 static const u8 IV6[] = {
1785 0x93, 0x13, 0x22, 0x5d, 0xf8, 0x84, 0x06, 0xe5,
1786 0x55, 0x90, 0x9c, 0x5a, 0xff, 0x52, 0x69, 0xaa,
1787 0x6a, 0x7a, 0x95, 0x38, 0x53, 0x4f, 0x7d, 0xa1,
1788 0xe4, 0xc3, 0x03, 0xd2, 0xa3, 0x18, 0xa7, 0x28,
1789 0xc3, 0xc0, 0xc9, 0x51, 0x56, 0x80, 0x95, 0x39,
1790 0xfc, 0xf0, 0xe2, 0x42, 0x9a, 0x6b, 0x52, 0x54,
1791 0x16, 0xae, 0xdb, 0xf5, 0xa0, 0xde, 0x6a, 0x57,
1792 0xa6, 0x37, 0xb3, 0x9b
1795 static const u8 C6[] = {
1796 0x8c, 0xe2, 0x49, 0x98, 0x62, 0x56, 0x15, 0xb6,
1797 0x03, 0xa0, 0x33, 0xac, 0xa1, 0x3f, 0xb8, 0x94,
1798 0xbe, 0x91, 0x12, 0xa5, 0xc3, 0xa2, 0x11, 0xa8,
1799 0xba, 0x26, 0x2a, 0x3c, 0xca, 0x7e, 0x2c, 0xa7,
1800 0x01, 0xe4, 0xa9, 0xa4, 0xfb, 0xa4, 0x3c, 0x90,
1801 0xcc, 0xdc, 0xb2, 0x81, 0xd4, 0x8c, 0x7c, 0x6f,
1802 0xd6, 0x28, 0x75, 0xd2, 0xac, 0xa4, 0x17, 0x03,
1803 0x4c, 0x34, 0xae, 0xe5
1806 static const u8 T6[] = {
1807 0x61, 0x9c, 0xc5, 0xae, 0xff, 0xfe, 0x0b, 0xfa,
1808 0x46, 0x2a, 0xf4, 0x3c, 0x16, 0x99, 0xd0, 0x50
1812 static const u8 K7[24], *P7 = NULL, *A7 = NULL, IV7[12], *C7 = NULL;
1813 static const u8 T7[] = {
1814 0xcd, 0x33, 0xb2, 0x8a, 0xc7, 0x73, 0xf7, 0x4b,
1815 0xa0, 0x0e, 0xd1, 0xf3, 0x12, 0x57, 0x24, 0x35
1822 static const u8 P8[16];
1823 static const u8 C8[] = {
1824 0x98, 0xe7, 0x24, 0x7c, 0x07, 0xf0, 0xfe, 0x41,
1825 0x1c, 0x26, 0x7e, 0x43, 0x84, 0xb0, 0xf6, 0x00
1828 static const u8 T8[] = {
1829 0x2f, 0xf5, 0x8d, 0x80, 0x03, 0x39, 0x27, 0xab,
1830 0x8e, 0xf4, 0xd4, 0x58, 0x75, 0x14, 0xf0, 0xfb
1835 static const u8 K9[] = {
1836 0xfe, 0xff, 0xe9, 0x92, 0x86, 0x65, 0x73, 0x1c,
1837 0x6d, 0x6a, 0x8f, 0x94, 0x67, 0x30, 0x83, 0x08,
1838 0xfe, 0xff, 0xe9, 0x92, 0x86, 0x65, 0x73, 0x1c
1841 static const u8 P9[] = {
1842 0xd9, 0x31, 0x32, 0x25, 0xf8, 0x84, 0x06, 0xe5,
1843 0xa5, 0x59, 0x09, 0xc5, 0xaf, 0xf5, 0x26, 0x9a,
1844 0x86, 0xa7, 0xa9, 0x53, 0x15, 0x34, 0xf7, 0xda,
1845 0x2e, 0x4c, 0x30, 0x3d, 0x8a, 0x31, 0x8a, 0x72,
1846 0x1c, 0x3c, 0x0c, 0x95, 0x95, 0x68, 0x09, 0x53,
1847 0x2f, 0xcf, 0x0e, 0x24, 0x49, 0xa6, 0xb5, 0x25,
1848 0xb1, 0x6a, 0xed, 0xf5, 0xaa, 0x0d, 0xe6, 0x57,
1849 0xba, 0x63, 0x7b, 0x39, 0x1a, 0xaf, 0xd2, 0x55
1852 static const u8 IV9[] = {
1853 0xca, 0xfe, 0xba, 0xbe, 0xfa, 0xce, 0xdb, 0xad,
1854 0xde, 0xca, 0xf8, 0x88
1857 static const u8 C9[] = {
1858 0x39, 0x80, 0xca, 0x0b, 0x3c, 0x00, 0xe8, 0x41,
1859 0xeb, 0x06, 0xfa, 0xc4, 0x87, 0x2a, 0x27, 0x57,
1860 0x85, 0x9e, 0x1c, 0xea, 0xa6, 0xef, 0xd9, 0x84,
1861 0x62, 0x85, 0x93, 0xb4, 0x0c, 0xa1, 0xe1, 0x9c,
1862 0x7d, 0x77, 0x3d, 0x00, 0xc1, 0x44, 0xc5, 0x25,
1863 0xac, 0x61, 0x9d, 0x18, 0xc8, 0x4a, 0x3f, 0x47,
1864 0x18, 0xe2, 0x44, 0x8b, 0x2f, 0xe3, 0x24, 0xd9,
1865 0xcc, 0xda, 0x27, 0x10, 0xac, 0xad, 0xe2, 0x56
1868 static const u8 T9[] = {
1869 0x99, 0x24, 0xa7, 0xc8, 0x58, 0x73, 0x36, 0xbf,
1870 0xb1, 0x18, 0x02, 0x4d, 0xb8, 0x67, 0x4a, 0x14
1876 static const u8 P10[] = {
1877 0xd9, 0x31, 0x32, 0x25, 0xf8, 0x84, 0x06, 0xe5,
1878 0xa5, 0x59, 0x09, 0xc5, 0xaf, 0xf5, 0x26, 0x9a,
1879 0x86, 0xa7, 0xa9, 0x53, 0x15, 0x34, 0xf7, 0xda,
1880 0x2e, 0x4c, 0x30, 0x3d, 0x8a, 0x31, 0x8a, 0x72,
1881 0x1c, 0x3c, 0x0c, 0x95, 0x95, 0x68, 0x09, 0x53,
1882 0x2f, 0xcf, 0x0e, 0x24, 0x49, 0xa6, 0xb5, 0x25,
1883 0xb1, 0x6a, 0xed, 0xf5, 0xaa, 0x0d, 0xe6, 0x57,
1884 0xba, 0x63, 0x7b, 0x39
1887 static const u8 A10[] = {
1888 0xfe, 0xed, 0xfa, 0xce, 0xde, 0xad, 0xbe, 0xef,
1889 0xfe, 0xed, 0xfa, 0xce, 0xde, 0xad, 0xbe, 0xef,
1890 0xab, 0xad, 0xda, 0xd2
1893 static const u8 C10[] = {
1894 0x39, 0x80, 0xca, 0x0b, 0x3c, 0x00, 0xe8, 0x41,
1895 0xeb, 0x06, 0xfa, 0xc4, 0x87, 0x2a, 0x27, 0x57,
1896 0x85, 0x9e, 0x1c, 0xea, 0xa6, 0xef, 0xd9, 0x84,
1897 0x62, 0x85, 0x93, 0xb4, 0x0c, 0xa1, 0xe1, 0x9c,
1898 0x7d, 0x77, 0x3d, 0x00, 0xc1, 0x44, 0xc5, 0x25,
1899 0xac, 0x61, 0x9d, 0x18, 0xc8, 0x4a, 0x3f, 0x47,
1900 0x18, 0xe2, 0x44, 0x8b, 0x2f, 0xe3, 0x24, 0xd9,
1901 0xcc, 0xda, 0x27, 0x10
1904 static const u8 T10[] = {
1905 0x25, 0x19, 0x49, 0x8e, 0x80, 0xf1, 0x47, 0x8f,
1906 0x37, 0xba, 0x55, 0xbd, 0x6d, 0x27, 0x61, 0x8c
1913 static const u8 IV11[] = { 0xca, 0xfe, 0xba, 0xbe, 0xfa, 0xce, 0xdb, 0xad };
1915 static const u8 C11[] = {
1916 0x0f, 0x10, 0xf5, 0x99, 0xae, 0x14, 0xa1, 0x54,
1917 0xed, 0x24, 0xb3, 0x6e, 0x25, 0x32, 0x4d, 0xb8,
1918 0xc5, 0x66, 0x63, 0x2e, 0xf2, 0xbb, 0xb3, 0x4f,
1919 0x83, 0x47, 0x28, 0x0f, 0xc4, 0x50, 0x70, 0x57,
1920 0xfd, 0xdc, 0x29, 0xdf, 0x9a, 0x47, 0x1f, 0x75,
1921 0xc6, 0x65, 0x41, 0xd4, 0xd4, 0xda, 0xd1, 0xc9,
1922 0xe9, 0x3a, 0x19, 0xa5, 0x8e, 0x8b, 0x47, 0x3f,
1923 0xa0, 0xf0, 0x62, 0xf7
1926 static const u8 T11[] = {
1927 0x65, 0xdc, 0xc5, 0x7f, 0xcf, 0x62, 0x3a, 0x24,
1928 0x09, 0x4f, 0xcc, 0xa4, 0x0d, 0x35, 0x33, 0xf8
1935 static const u8 IV12[] = {
1936 0x93, 0x13, 0x22, 0x5d, 0xf8, 0x84, 0x06, 0xe5,
1937 0x55, 0x90, 0x9c, 0x5a, 0xff, 0x52, 0x69, 0xaa,
1938 0x6a, 0x7a, 0x95, 0x38, 0x53, 0x4f, 0x7d, 0xa1,
1939 0xe4, 0xc3, 0x03, 0xd2, 0xa3, 0x18, 0xa7, 0x28,
1940 0xc3, 0xc0, 0xc9, 0x51, 0x56, 0x80, 0x95, 0x39,
1941 0xfc, 0xf0, 0xe2, 0x42, 0x9a, 0x6b, 0x52, 0x54,
1942 0x16, 0xae, 0xdb, 0xf5, 0xa0, 0xde, 0x6a, 0x57,
1943 0xa6, 0x37, 0xb3, 0x9b
1946 static const u8 C12[] = {
1947 0xd2, 0x7e, 0x88, 0x68, 0x1c, 0xe3, 0x24, 0x3c,
1948 0x48, 0x30, 0x16, 0x5a, 0x8f, 0xdc, 0xf9, 0xff,
1949 0x1d, 0xe9, 0xa1, 0xd8, 0xe6, 0xb4, 0x47, 0xef,
1950 0x6e, 0xf7, 0xb7, 0x98, 0x28, 0x66, 0x6e, 0x45,
1951 0x81, 0xe7, 0x90, 0x12, 0xaf, 0x34, 0xdd, 0xd9,
1952 0xe2, 0xf0, 0x37, 0x58, 0x9b, 0x29, 0x2d, 0xb3,
1953 0xe6, 0x7c, 0x03, 0x67, 0x45, 0xfa, 0x22, 0xe7,
1954 0xe9, 0xb7, 0x37, 0x3b
1957 static const u8 T12[] = {
1958 0xdc, 0xf5, 0x66, 0xff, 0x29, 0x1c, 0x25, 0xbb,
1959 0xb8, 0x56, 0x8f, 0xc3, 0xd3, 0x76, 0xa6, 0xd9
1963 static const u8 K13[32], *P13 = NULL, *A13 = NULL, IV13[12], *C13 = NULL;
1964 static const u8 T13[] = {
1965 0x53, 0x0f, 0x8a, 0xfb, 0xc7, 0x45, 0x36, 0xb9,
1966 0xa9, 0x63, 0xb4, 0xf1, 0xc4, 0xcb, 0x73, 0x8b
1972 static const u8 P14[16], IV14[12];
1973 static const u8 C14[] = {
1974 0xce, 0xa7, 0x40, 0x3d, 0x4d, 0x60, 0x6b, 0x6e,
1975 0x07, 0x4e, 0xc5, 0xd3, 0xba, 0xf3, 0x9d, 0x18
1978 static const u8 T14[] = {
1979 0xd0, 0xd1, 0xc8, 0xa7, 0x99, 0x99, 0x6b, 0xf0,
1980 0x26, 0x5b, 0x98, 0xb5, 0xd4, 0x8a, 0xb9, 0x19
1985 static const u8 K15[] = {
1986 0xfe, 0xff, 0xe9, 0x92, 0x86, 0x65, 0x73, 0x1c,
1987 0x6d, 0x6a, 0x8f, 0x94, 0x67, 0x30, 0x83, 0x08,
1988 0xfe, 0xff, 0xe9, 0x92, 0x86, 0x65, 0x73, 0x1c,
1989 0x6d, 0x6a, 0x8f, 0x94, 0x67, 0x30, 0x83, 0x08
1992 static const u8 P15[] = {
1993 0xd9, 0x31, 0x32, 0x25, 0xf8, 0x84, 0x06, 0xe5,
1994 0xa5, 0x59, 0x09, 0xc5, 0xaf, 0xf5, 0x26, 0x9a,
1995 0x86, 0xa7, 0xa9, 0x53, 0x15, 0x34, 0xf7, 0xda,
1996 0x2e, 0x4c, 0x30, 0x3d, 0x8a, 0x31, 0x8a, 0x72,
1997 0x1c, 0x3c, 0x0c, 0x95, 0x95, 0x68, 0x09, 0x53,
1998 0x2f, 0xcf, 0x0e, 0x24, 0x49, 0xa6, 0xb5, 0x25,
1999 0xb1, 0x6a, 0xed, 0xf5, 0xaa, 0x0d, 0xe6, 0x57,
2000 0xba, 0x63, 0x7b, 0x39, 0x1a, 0xaf, 0xd2, 0x55
2003 static const u8 IV15[] = {
2004 0xca, 0xfe, 0xba, 0xbe, 0xfa, 0xce, 0xdb, 0xad,
2005 0xde, 0xca, 0xf8, 0x88
2008 static const u8 C15[] = {
2009 0x52, 0x2d, 0xc1, 0xf0, 0x99, 0x56, 0x7d, 0x07,
2010 0xf4, 0x7f, 0x37, 0xa3, 0x2a, 0x84, 0x42, 0x7d,
2011 0x64, 0x3a, 0x8c, 0xdc, 0xbf, 0xe5, 0xc0, 0xc9,
2012 0x75, 0x98, 0xa2, 0xbd, 0x25, 0x55, 0xd1, 0xaa,
2013 0x8c, 0xb0, 0x8e, 0x48, 0x59, 0x0d, 0xbb, 0x3d,
2014 0xa7, 0xb0, 0x8b, 0x10, 0x56, 0x82, 0x88, 0x38,
2015 0xc5, 0xf6, 0x1e, 0x63, 0x93, 0xba, 0x7a, 0x0a,
2016 0xbc, 0xc9, 0xf6, 0x62, 0x89, 0x80, 0x15, 0xad
2019 static const u8 T15[] = {
2020 0xb0, 0x94, 0xda, 0xc5, 0xd9, 0x34, 0x71, 0xbd,
2021 0xec, 0x1a, 0x50, 0x22, 0x70, 0xe3, 0xcc, 0x6c
2027 static const u8 P16[] = {
2028 0xd9, 0x31, 0x32, 0x25, 0xf8, 0x84, 0x06, 0xe5,
2029 0xa5, 0x59, 0x09, 0xc5, 0xaf, 0xf5, 0x26, 0x9a,
2030 0x86, 0xa7, 0xa9, 0x53, 0x15, 0x34, 0xf7, 0xda,
2031 0x2e, 0x4c, 0x30, 0x3d, 0x8a, 0x31, 0x8a, 0x72,
2032 0x1c, 0x3c, 0x0c, 0x95, 0x95, 0x68, 0x09, 0x53,
2033 0x2f, 0xcf, 0x0e, 0x24, 0x49, 0xa6, 0xb5, 0x25,
2034 0xb1, 0x6a, 0xed, 0xf5, 0xaa, 0x0d, 0xe6, 0x57,
2035 0xba, 0x63, 0x7b, 0x39
2038 static const u8 A16[] = {
2039 0xfe, 0xed, 0xfa, 0xce, 0xde, 0xad, 0xbe, 0xef,
2040 0xfe, 0xed, 0xfa, 0xce, 0xde, 0xad, 0xbe, 0xef,
2041 0xab, 0xad, 0xda, 0xd2
2044 static const u8 C16[] = {
2045 0x52, 0x2d, 0xc1, 0xf0, 0x99, 0x56, 0x7d, 0x07,
2046 0xf4, 0x7f, 0x37, 0xa3, 0x2a, 0x84, 0x42, 0x7d,
2047 0x64, 0x3a, 0x8c, 0xdc, 0xbf, 0xe5, 0xc0, 0xc9,
2048 0x75, 0x98, 0xa2, 0xbd, 0x25, 0x55, 0xd1, 0xaa,
2049 0x8c, 0xb0, 0x8e, 0x48, 0x59, 0x0d, 0xbb, 0x3d,
2050 0xa7, 0xb0, 0x8b, 0x10, 0x56, 0x82, 0x88, 0x38,
2051 0xc5, 0xf6, 0x1e, 0x63, 0x93, 0xba, 0x7a, 0x0a,
2052 0xbc, 0xc9, 0xf6, 0x62
2055 static const u8 T16[] = {
2056 0x76, 0xfc, 0x6e, 0xce, 0x0f, 0x4e, 0x17, 0x68,
2057 0xcd, 0xdf, 0x88, 0x53, 0xbb, 0x2d, 0x55, 0x1b
2064 static const u8 IV17[] = { 0xca, 0xfe, 0xba, 0xbe, 0xfa, 0xce, 0xdb, 0xad };
2066 static const u8 C17[] = {
2067 0xc3, 0x76, 0x2d, 0xf1, 0xca, 0x78, 0x7d, 0x32,
2068 0xae, 0x47, 0xc1, 0x3b, 0xf1, 0x98, 0x44, 0xcb,
2069 0xaf, 0x1a, 0xe1, 0x4d, 0x0b, 0x97, 0x6a, 0xfa,
2070 0xc5, 0x2f, 0xf7, 0xd7, 0x9b, 0xba, 0x9d, 0xe0,
2071 0xfe, 0xb5, 0x82, 0xd3, 0x39, 0x34, 0xa4, 0xf0,
2072 0x95, 0x4c, 0xc2, 0x36, 0x3b, 0xc7, 0x3f, 0x78,
2073 0x62, 0xac, 0x43, 0x0e, 0x64, 0xab, 0xe4, 0x99,
2074 0xf4, 0x7c, 0x9b, 0x1f
2077 static const u8 T17[] = {
2078 0x3a, 0x33, 0x7d, 0xbf, 0x46, 0xa7, 0x92, 0xc4,
2079 0x5e, 0x45, 0x49, 0x13, 0xfe, 0x2e, 0xa8, 0xf2
2086 static const u8 IV18[] = {
2087 0x93, 0x13, 0x22, 0x5d, 0xf8, 0x84, 0x06, 0xe5,
2088 0x55, 0x90, 0x9c, 0x5a, 0xff, 0x52, 0x69, 0xaa,
2089 0x6a, 0x7a, 0x95, 0x38, 0x53, 0x4f, 0x7d, 0xa1,
2090 0xe4, 0xc3, 0x03, 0xd2, 0xa3, 0x18, 0xa7, 0x28,
2091 0xc3, 0xc0, 0xc9, 0x51, 0x56, 0x80, 0x95, 0x39,
2092 0xfc, 0xf0, 0xe2, 0x42, 0x9a, 0x6b, 0x52, 0x54,
2093 0x16, 0xae, 0xdb, 0xf5, 0xa0, 0xde, 0x6a, 0x57,
2094 0xa6, 0x37, 0xb3, 0x9b
2097 static const u8 C18[] = {
2098 0x5a, 0x8d, 0xef, 0x2f, 0x0c, 0x9e, 0x53, 0xf1,
2099 0xf7, 0x5d, 0x78, 0x53, 0x65, 0x9e, 0x2a, 0x20,
2100 0xee, 0xb2, 0xb2, 0x2a, 0xaf, 0xde, 0x64, 0x19,
2101 0xa0, 0x58, 0xab, 0x4f, 0x6f, 0x74, 0x6b, 0xf4,
2102 0x0f, 0xc0, 0xc3, 0xb7, 0x80, 0xf2, 0x44, 0x45,
2103 0x2d, 0xa3, 0xeb, 0xf1, 0xc5, 0xd8, 0x2c, 0xde,
2104 0xa2, 0x41, 0x89, 0x97, 0x20, 0x0e, 0xf8, 0x2e,
2105 0x44, 0xae, 0x7e, 0x3f
2108 static const u8 T18[] = {
2109 0xa4, 0x4a, 0x82, 0x66, 0xee, 0x1c, 0x8e, 0xb0,
2110 0xc8, 0xb5, 0xd4, 0xcf, 0x5a, 0xe9, 0xf1, 0x9a
2118 static const u8 A19[] = {
2119 0xd9, 0x31, 0x32, 0x25, 0xf8, 0x84, 0x06, 0xe5,
2120 0xa5, 0x59, 0x09, 0xc5, 0xaf, 0xf5, 0x26, 0x9a,
2121 0x86, 0xa7, 0xa9, 0x53, 0x15, 0x34, 0xf7, 0xda,
2122 0x2e, 0x4c, 0x30, 0x3d, 0x8a, 0x31, 0x8a, 0x72,
2123 0x1c, 0x3c, 0x0c, 0x95, 0x95, 0x68, 0x09, 0x53,
2124 0x2f, 0xcf, 0x0e, 0x24, 0x49, 0xa6, 0xb5, 0x25,
2125 0xb1, 0x6a, 0xed, 0xf5, 0xaa, 0x0d, 0xe6, 0x57,
2126 0xba, 0x63, 0x7b, 0x39, 0x1a, 0xaf, 0xd2, 0x55,
2127 0x52, 0x2d, 0xc1, 0xf0, 0x99, 0x56, 0x7d, 0x07,
2128 0xf4, 0x7f, 0x37, 0xa3, 0x2a, 0x84, 0x42, 0x7d,
2129 0x64, 0x3a, 0x8c, 0xdc, 0xbf, 0xe5, 0xc0, 0xc9,
2130 0x75, 0x98, 0xa2, 0xbd, 0x25, 0x55, 0xd1, 0xaa,
2131 0x8c, 0xb0, 0x8e, 0x48, 0x59, 0x0d, 0xbb, 0x3d,
2132 0xa7, 0xb0, 0x8b, 0x10, 0x56, 0x82, 0x88, 0x38,
2133 0xc5, 0xf6, 0x1e, 0x63, 0x93, 0xba, 0x7a, 0x0a,
2134 0xbc, 0xc9, 0xf6, 0x62, 0x89, 0x80, 0x15, 0xad
2137 static const u8 T19[] = {
2138 0x5f, 0xea, 0x79, 0x3a, 0x2d, 0x6f, 0x97, 0x4d,
2139 0x37, 0xe6, 0x8e, 0x0c, 0xb8, 0xff, 0x94, 0x92
2145 /* this results in 0xff in counter LSB */
2146 static const u8 IV20[64] = { 0xff, 0xff, 0xff, 0xff };
2148 static const u8 P20[288];
2149 static const u8 C20[] = {
2150 0x56, 0xb3, 0x37, 0x3c, 0xa9, 0xef, 0x6e, 0x4a,
2151 0x2b, 0x64, 0xfe, 0x1e, 0x9a, 0x17, 0xb6, 0x14,
2152 0x25, 0xf1, 0x0d, 0x47, 0xa7, 0x5a, 0x5f, 0xce,
2153 0x13, 0xef, 0xc6, 0xbc, 0x78, 0x4a, 0xf2, 0x4f,
2154 0x41, 0x41, 0xbd, 0xd4, 0x8c, 0xf7, 0xc7, 0x70,
2155 0x88, 0x7a, 0xfd, 0x57, 0x3c, 0xca, 0x54, 0x18,
2156 0xa9, 0xae, 0xff, 0xcd, 0x7c, 0x5c, 0xed, 0xdf,
2157 0xc6, 0xa7, 0x83, 0x97, 0xb9, 0xa8, 0x5b, 0x49,
2158 0x9d, 0xa5, 0x58, 0x25, 0x72, 0x67, 0xca, 0xab,
2159 0x2a, 0xd0, 0xb2, 0x3c, 0xa4, 0x76, 0xa5, 0x3c,
2160 0xb1, 0x7f, 0xb4, 0x1c, 0x4b, 0x8b, 0x47, 0x5c,
2161 0xb4, 0xf3, 0xf7, 0x16, 0x50, 0x94, 0xc2, 0x29,
2162 0xc9, 0xe8, 0xc4, 0xdc, 0x0a, 0x2a, 0x5f, 0xf1,
2163 0x90, 0x3e, 0x50, 0x15, 0x11, 0x22, 0x13, 0x76,
2164 0xa1, 0xcd, 0xb8, 0x36, 0x4c, 0x50, 0x61, 0xa2,
2165 0x0c, 0xae, 0x74, 0xbc, 0x4a, 0xcd, 0x76, 0xce,
2166 0xb0, 0xab, 0xc9, 0xfd, 0x32, 0x17, 0xef, 0x9f,
2167 0x8c, 0x90, 0xbe, 0x40, 0x2d, 0xdf, 0x6d, 0x86,
2168 0x97, 0xf4, 0xf8, 0x80, 0xdf, 0xf1, 0x5b, 0xfb,
2169 0x7a, 0x6b, 0x28, 0x24, 0x1e, 0xc8, 0xfe, 0x18,
2170 0x3c, 0x2d, 0x59, 0xe3, 0xf9, 0xdf, 0xff, 0x65,
2171 0x3c, 0x71, 0x26, 0xf0, 0xac, 0xb9, 0xe6, 0x42,
2172 0x11, 0xf4, 0x2b, 0xae, 0x12, 0xaf, 0x46, 0x2b,
2173 0x10, 0x70, 0xbe, 0xf1, 0xab, 0x5e, 0x36, 0x06,
2174 0x87, 0x2c, 0xa1, 0x0d, 0xee, 0x15, 0xb3, 0x24,
2175 0x9b, 0x1a, 0x1b, 0x95, 0x8f, 0x23, 0x13, 0x4c,
2176 0x4b, 0xcc, 0xb7, 0xd0, 0x32, 0x00, 0xbc, 0xe4,
2177 0x20, 0xa2, 0xf8, 0xeb, 0x66, 0xdc, 0xf3, 0x64,
2178 0x4d, 0x14, 0x23, 0xc1, 0xb5, 0x69, 0x90, 0x03,
2179 0xc1, 0x3e, 0xce, 0xf4, 0xbf, 0x38, 0xa3, 0xb6,
2180 0x0e, 0xed, 0xc3, 0x40, 0x33, 0xba, 0xc1, 0x90,
2181 0x27, 0x83, 0xdc, 0x6d, 0x89, 0xe2, 0xe7, 0x74,
2182 0x18, 0x8a, 0x43, 0x9c, 0x7e, 0xbc, 0xc0, 0x67,
2183 0x2d, 0xbd, 0xa4, 0xdd, 0xcf, 0xb2, 0x79, 0x46,
2184 0x13, 0xb0, 0xbe, 0x41, 0x31, 0x5e, 0xf7, 0x78,
2185 0x70, 0x8a, 0x70, 0xee, 0x7d, 0x75, 0x16, 0x5c
2188 static const u8 T20[] = {
2189 0x8b, 0x30, 0x7f, 0x6b, 0x33, 0x28, 0x6d, 0x0a,
2190 0xb0, 0x26, 0xa9, 0xed, 0x3f, 0xe1, 0xe8, 0x5f
2193 # define TEST_CASE(n) do { \
2194 u8 out[sizeof(P##n)]; \
2195 AES_set_encrypt_key(K##n,sizeof(K##n)*8,&key); \
2196 CRYPTO_gcm128_init(&ctx,&key,(block128_f)AES_encrypt); \
2197 CRYPTO_gcm128_setiv(&ctx,IV##n,sizeof(IV##n)); \
2198 memset(out,0,sizeof(out)); \
2199 if (A##n) CRYPTO_gcm128_aad(&ctx,A##n,sizeof(A##n)); \
2200 if (P##n) CRYPTO_gcm128_encrypt(&ctx,P##n,out,sizeof(out)); \
2201 if (CRYPTO_gcm128_finish(&ctx,T##n,16) || \
2202 (C##n && memcmp(out,C##n,sizeof(out)))) \
2203 ret++, printf ("encrypt test#%d failed.\n",n); \
2204 CRYPTO_gcm128_setiv(&ctx,IV##n,sizeof(IV##n)); \
2205 memset(out,0,sizeof(out)); \
2206 if (A##n) CRYPTO_gcm128_aad(&ctx,A##n,sizeof(A##n)); \
2207 if (C##n) CRYPTO_gcm128_decrypt(&ctx,C##n,out,sizeof(out)); \
2208 if (CRYPTO_gcm128_finish(&ctx,T##n,16) || \
2209 (P##n && memcmp(out,P##n,sizeof(out)))) \
2210 ret++, printf ("decrypt test#%d failed.\n",n); \
2240 # ifdef OPENSSL_CPUID_OBJ
2242 size_t start, stop, gcm_t, ctr_t, OPENSSL_rdtsc();
2249 AES_set_encrypt_key(K1, sizeof(K1) * 8, &key);
2250 CRYPTO_gcm128_init(&ctx, &key, (block128_f) AES_encrypt);
2251 CRYPTO_gcm128_setiv(&ctx, IV1, sizeof(IV1));
2253 CRYPTO_gcm128_encrypt(&ctx, buf.c, buf.c, sizeof(buf));
2254 start = OPENSSL_rdtsc();
2255 CRYPTO_gcm128_encrypt(&ctx, buf.c, buf.c, sizeof(buf));
2256 gcm_t = OPENSSL_rdtsc() - start;
2258 CRYPTO_ctr128_encrypt(buf.c, buf.c, sizeof(buf),
2259 &key, ctx.Yi.c, ctx.EKi.c, &ctx.mres,
2260 (block128_f) AES_encrypt);
2261 start = OPENSSL_rdtsc();
2262 CRYPTO_ctr128_encrypt(buf.c, buf.c, sizeof(buf),
2263 &key, ctx.Yi.c, ctx.EKi.c, &ctx.mres,
2264 (block128_f) AES_encrypt);
2265 ctr_t = OPENSSL_rdtsc() - start;
2267 printf("%.2f-%.2f=%.2f\n",
2268 gcm_t / (double)sizeof(buf),
2269 ctr_t / (double)sizeof(buf),
2270 (gcm_t - ctr_t) / (double)sizeof(buf));
2273 void (*gcm_ghash_p) (u64 Xi[2], const u128 Htable[16],
2274 const u8 *inp, size_t len) = ctx.ghash;
2276 GHASH((&ctx), buf.c, sizeof(buf));
2277 start = OPENSSL_rdtsc();
2278 for (i = 0; i < 100; ++i)
2279 GHASH((&ctx), buf.c, sizeof(buf));
2280 gcm_t = OPENSSL_rdtsc() - start;
2281 printf("%.2f\n", gcm_t / (double)sizeof(buf) / (double)i);