]>
Commit | Line | Data |
---|---|---|
4f22f405 | 1 | /* |
999376dc | 2 | * Copyright 2010-2022 The OpenSSL Project Authors. All Rights Reserved. |
e7f5b1cd | 3 | * |
81cae8ce | 4 | * Licensed under the Apache License 2.0 (the "License"). You may not use |
4f22f405 RS |
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 | |
e7f5b1cd AP |
8 | */ |
9 | ||
e7f5b1cd | 10 | #include <string.h> |
459b15d4 | 11 | #include <openssl/crypto.h> |
24fd8541 | 12 | #include "internal/cryptlib.h" |
e23d850f | 13 | #include "internal/endian.h" |
25f2138b | 14 | #include "crypto/modes.h" |
e7f5b1cd | 15 | |
77286fe3 BE |
16 | #if defined(__GNUC__) && !defined(STRICT_ALIGNMENT) |
17 | typedef size_t size_t_aX __attribute((__aligned__(1))); | |
18 | #else | |
19 | typedef size_t size_t_aX; | |
20 | #endif | |
21 | ||
f472ec8c AP |
22 | #if defined(BSWAP4) && defined(STRICT_ALIGNMENT) |
23 | /* redefine, because alignment is ensured */ | |
0f113f3e MC |
24 | # undef GETU32 |
25 | # define GETU32(p) BSWAP4(*(const u32 *)(p)) | |
26 | # undef PUTU32 | |
27 | # define PUTU32(p,v) *(u32 *)(p) = BSWAP4(v) | |
28 | #endif | |
29 | ||
f3fed0d5 | 30 | /* RISC-V uses C implementation as a fallback. */ |
999376dc HB |
31 | #if defined(__riscv) |
32 | # define INCLUDE_C_GMULT_4BIT | |
f3fed0d5 | 33 | # define INCLUDE_C_GHASH_4BIT |
999376dc HB |
34 | #endif |
35 | ||
0f113f3e MC |
36 | #define PACK(s) ((size_t)(s)<<(sizeof(size_t)*8-16)) |
37 | #define REDUCE1BIT(V) do { \ | |
38 | if (sizeof(size_t)==8) { \ | |
39 | u64 T = U64(0xe100000000000000) & (0-(V.lo&1)); \ | |
40 | V.lo = (V.hi<<63)|(V.lo>>1); \ | |
41 | V.hi = (V.hi>>1 )^T; \ | |
42 | } \ | |
43 | else { \ | |
44 | u32 T = 0xe1000000U & (0-(u32)(V.lo&1)); \ | |
45 | V.lo = (V.hi<<63)|(V.lo>>1); \ | |
46 | V.hi = (V.hi>>1 )^((u64)T<<32); \ | |
47 | } \ | |
c1f092d1 AP |
48 | } while(0) |
49 | ||
1d97c843 | 50 | /*- |
7b6e19fc TS |
51 | * |
52 | * NOTE: TABLE_BITS and all non-4bit implmentations have been removed in 3.1. | |
53 | * | |
d8d95832 AP |
54 | * Even though permitted values for TABLE_BITS are 8, 4 and 1, it should |
55 | * never be set to 8. 8 is effectively reserved for testing purposes. | |
56 | * TABLE_BITS>1 are lookup-table-driven implementations referred to as | |
57 | * "Shoup's" in GCM specification. In other words OpenSSL does not cover | |
58 | * whole spectrum of possible table driven implementations. Why? In | |
59 | * non-"Shoup's" case memory access pattern is segmented in such manner, | |
60 | * that it's trivial to see that cache timing information can reveal | |
61 | * fair portion of intermediate hash value. Given that ciphertext is | |
62 | * always available to attacker, it's possible for him to attempt to | |
63 | * deduce secret parameter H and if successful, tamper with messages | |
64 | * [which is nothing but trivial in CTR mode]. In "Shoup's" case it's | |
65 | * not as trivial, but there is no reason to believe that it's resistant | |
66 | * to cache-timing attack. And the thing about "8-bit" implementation is | |
67 | * that it consumes 16 (sixteen) times more memory, 4KB per individual | |
68 | * key + 1KB shared. Well, on pros side it should be twice as fast as | |
69 | * "4-bit" version. And for gcc-generated x86[_64] code, "8-bit" version | |
70 | * was observed to run ~75% faster, closer to 100% for commercial | |
71 | * compilers... Yet "4-bit" procedure is preferred, because it's | |
72 | * believed to provide better security-performance balance and adequate | |
73 | * all-round performance. "All-round" refers to things like: | |
74 | * | |
75 | * - shorter setup time effectively improves overall timing for | |
76 | * handling short messages; | |
77 | * - larger table allocation can become unbearable because of VM | |
78 | * subsystem penalties (for example on Windows large enough free | |
79 | * results in VM working set trimming, meaning that consequent | |
80 | * malloc would immediately incur working set expansion); | |
81 | * - larger table has larger cache footprint, which can affect | |
82 | * performance of other code paths (not necessarily even from same | |
83 | * thread in Hyper-Threading world); | |
84 | * | |
85 | * Value of 1 is not appropriate for performance reasons. | |
86 | */ | |
a595baff | 87 | |
92c9086e | 88 | static void gcm_init_4bit(u128 Htable[16], const u64 H[2]) |
e7f5b1cd | 89 | { |
0f113f3e MC |
90 | u128 V; |
91 | # if defined(OPENSSL_SMALL_FOOTPRINT) | |
92 | int i; | |
93 | # endif | |
e7f5b1cd | 94 | |
0f113f3e MC |
95 | Htable[0].hi = 0; |
96 | Htable[0].lo = 0; | |
97 | V.hi = H[0]; | |
98 | V.lo = H[1]; | |
99 | ||
100 | # if defined(OPENSSL_SMALL_FOOTPRINT) | |
101 | for (Htable[8] = V, i = 4; i > 0; i >>= 1) { | |
102 | REDUCE1BIT(V); | |
103 | Htable[i] = V; | |
104 | } | |
105 | ||
106 | for (i = 2; i < 16; i <<= 1) { | |
107 | u128 *Hi = Htable + i; | |
108 | int j; | |
109 | for (V = *Hi, j = 1; j < i; ++j) { | |
110 | Hi[j].hi = V.hi ^ Htable[j].hi; | |
111 | Hi[j].lo = V.lo ^ Htable[j].lo; | |
112 | } | |
113 | } | |
114 | # else | |
115 | Htable[8] = V; | |
116 | REDUCE1BIT(V); | |
117 | Htable[4] = V; | |
118 | REDUCE1BIT(V); | |
119 | Htable[2] = V; | |
120 | REDUCE1BIT(V); | |
121 | Htable[1] = V; | |
122 | Htable[3].hi = V.hi ^ Htable[2].hi, Htable[3].lo = V.lo ^ Htable[2].lo; | |
123 | V = Htable[4]; | |
124 | Htable[5].hi = V.hi ^ Htable[1].hi, Htable[5].lo = V.lo ^ Htable[1].lo; | |
125 | Htable[6].hi = V.hi ^ Htable[2].hi, Htable[6].lo = V.lo ^ Htable[2].lo; | |
126 | Htable[7].hi = V.hi ^ Htable[3].hi, Htable[7].lo = V.lo ^ Htable[3].lo; | |
127 | V = Htable[8]; | |
128 | Htable[9].hi = V.hi ^ Htable[1].hi, Htable[9].lo = V.lo ^ Htable[1].lo; | |
129 | Htable[10].hi = V.hi ^ Htable[2].hi, Htable[10].lo = V.lo ^ Htable[2].lo; | |
130 | Htable[11].hi = V.hi ^ Htable[3].hi, Htable[11].lo = V.lo ^ Htable[3].lo; | |
131 | Htable[12].hi = V.hi ^ Htable[4].hi, Htable[12].lo = V.lo ^ Htable[4].lo; | |
132 | Htable[13].hi = V.hi ^ Htable[5].hi, Htable[13].lo = V.lo ^ Htable[5].lo; | |
133 | Htable[14].hi = V.hi ^ Htable[6].hi, Htable[14].lo = V.lo ^ Htable[6].lo; | |
134 | Htable[15].hi = V.hi ^ Htable[7].hi, Htable[15].lo = V.lo ^ Htable[7].lo; | |
135 | # endif | |
136 | # if defined(GHASH_ASM) && (defined(__arm__) || defined(__arm)) | |
137 | /* | |
138 | * ARM assembler expects specific dword order in Htable. | |
139 | */ | |
140 | { | |
141 | int j; | |
e23d850f | 142 | DECLARE_IS_ENDIAN; |
0f113f3e | 143 | |
e23d850f | 144 | if (IS_LITTLE_ENDIAN) |
0f113f3e MC |
145 | for (j = 0; j < 16; ++j) { |
146 | V = Htable[j]; | |
147 | Htable[j].hi = V.lo; | |
148 | Htable[j].lo = V.hi; | |
149 | } else | |
150 | for (j = 0; j < 16; ++j) { | |
151 | V = Htable[j]; | |
152 | Htable[j].hi = V.lo << 32 | V.lo >> 32; | |
153 | Htable[j].lo = V.hi << 32 | V.hi >> 32; | |
154 | } | |
155 | } | |
156 | # endif | |
e7f5b1cd AP |
157 | } |
158 | ||
999376dc | 159 | # if !defined(GHASH_ASM) || defined(INCLUDE_C_GMULT_4BIT) |
2262beef | 160 | static const size_t rem_4bit[16] = { |
0f113f3e MC |
161 | PACK(0x0000), PACK(0x1C20), PACK(0x3840), PACK(0x2460), |
162 | PACK(0x7080), PACK(0x6CA0), PACK(0x48C0), PACK(0x54E0), | |
163 | PACK(0xE100), PACK(0xFD20), PACK(0xD940), PACK(0xC560), | |
164 | PACK(0x9180), PACK(0x8DA0), PACK(0xA9C0), PACK(0xB5E0) | |
165 | }; | |
2262beef | 166 | |
4f39edbf | 167 | static void gcm_gmult_4bit(u64 Xi[2], const u128 Htable[16]) |
e7f5b1cd | 168 | { |
0f113f3e MC |
169 | u128 Z; |
170 | int cnt = 15; | |
171 | size_t rem, nlo, nhi; | |
e23d850f | 172 | DECLARE_IS_ENDIAN; |
0f113f3e MC |
173 | |
174 | nlo = ((const u8 *)Xi)[15]; | |
175 | nhi = nlo >> 4; | |
176 | nlo &= 0xf; | |
177 | ||
178 | Z.hi = Htable[nlo].hi; | |
179 | Z.lo = Htable[nlo].lo; | |
180 | ||
181 | while (1) { | |
182 | rem = (size_t)Z.lo & 0xf; | |
183 | Z.lo = (Z.hi << 60) | (Z.lo >> 4); | |
184 | Z.hi = (Z.hi >> 4); | |
185 | if (sizeof(size_t) == 8) | |
186 | Z.hi ^= rem_4bit[rem]; | |
187 | else | |
188 | Z.hi ^= (u64)rem_4bit[rem] << 32; | |
189 | ||
190 | Z.hi ^= Htable[nhi].hi; | |
191 | Z.lo ^= Htable[nhi].lo; | |
192 | ||
193 | if (--cnt < 0) | |
194 | break; | |
195 | ||
196 | nlo = ((const u8 *)Xi)[cnt]; | |
197 | nhi = nlo >> 4; | |
198 | nlo &= 0xf; | |
199 | ||
200 | rem = (size_t)Z.lo & 0xf; | |
201 | Z.lo = (Z.hi << 60) | (Z.lo >> 4); | |
202 | Z.hi = (Z.hi >> 4); | |
203 | if (sizeof(size_t) == 8) | |
204 | Z.hi ^= rem_4bit[rem]; | |
205 | else | |
206 | Z.hi ^= (u64)rem_4bit[rem] << 32; | |
207 | ||
208 | Z.hi ^= Htable[nlo].hi; | |
209 | Z.lo ^= Htable[nlo].lo; | |
210 | } | |
211 | ||
e23d850f | 212 | if (IS_LITTLE_ENDIAN) { |
0f113f3e MC |
213 | # ifdef BSWAP8 |
214 | Xi[0] = BSWAP8(Z.hi); | |
215 | Xi[1] = BSWAP8(Z.lo); | |
216 | # else | |
217 | u8 *p = (u8 *)Xi; | |
218 | u32 v; | |
219 | v = (u32)(Z.hi >> 32); | |
220 | PUTU32(p, v); | |
221 | v = (u32)(Z.hi); | |
222 | PUTU32(p + 4, v); | |
223 | v = (u32)(Z.lo >> 32); | |
224 | PUTU32(p + 8, v); | |
225 | v = (u32)(Z.lo); | |
226 | PUTU32(p + 12, v); | |
227 | # endif | |
228 | } else { | |
229 | Xi[0] = Z.hi; | |
230 | Xi[1] = Z.lo; | |
231 | } | |
2262beef AP |
232 | } |
233 | ||
999376dc HB |
234 | # endif |
235 | ||
f3fed0d5 | 236 | # if !defined(GHASH_ASM) || defined(INCLUDE_C_GHASH_4BIT) |
0f113f3e | 237 | # if !defined(OPENSSL_SMALL_FOOTPRINT) |
2262beef AP |
238 | /* |
239 | * Streamed gcm_mult_4bit, see CRYPTO_gcm128_[en|de]crypt for | |
a595baff AP |
240 | * details... Compiler-generated code doesn't seem to give any |
241 | * performance improvement, at least not on x86[_64]. It's here | |
242 | * mostly as reference and a placeholder for possible future | |
243 | * non-trivial optimization[s]... | |
2262beef | 244 | */ |
0f113f3e MC |
245 | static void gcm_ghash_4bit(u64 Xi[2], const u128 Htable[16], |
246 | const u8 *inp, size_t len) | |
2262beef AP |
247 | { |
248 | u128 Z; | |
249 | int cnt; | |
250 | size_t rem, nlo, nhi; | |
e23d850f | 251 | DECLARE_IS_ENDIAN; |
0f113f3e | 252 | |
2262beef | 253 | do { |
0f113f3e MC |
254 | cnt = 15; |
255 | nlo = ((const u8 *)Xi)[15]; | |
256 | nlo ^= inp[15]; | |
257 | nhi = nlo >> 4; | |
258 | nlo &= 0xf; | |
259 | ||
260 | Z.hi = Htable[nlo].hi; | |
261 | Z.lo = Htable[nlo].lo; | |
262 | ||
263 | while (1) { | |
264 | rem = (size_t)Z.lo & 0xf; | |
265 | Z.lo = (Z.hi << 60) | (Z.lo >> 4); | |
266 | Z.hi = (Z.hi >> 4); | |
267 | if (sizeof(size_t) == 8) | |
268 | Z.hi ^= rem_4bit[rem]; | |
269 | else | |
270 | Z.hi ^= (u64)rem_4bit[rem] << 32; | |
271 | ||
272 | Z.hi ^= Htable[nhi].hi; | |
273 | Z.lo ^= Htable[nhi].lo; | |
274 | ||
275 | if (--cnt < 0) | |
276 | break; | |
277 | ||
278 | nlo = ((const u8 *)Xi)[cnt]; | |
279 | nlo ^= inp[cnt]; | |
280 | nhi = nlo >> 4; | |
281 | nlo &= 0xf; | |
282 | ||
283 | rem = (size_t)Z.lo & 0xf; | |
284 | Z.lo = (Z.hi << 60) | (Z.lo >> 4); | |
285 | Z.hi = (Z.hi >> 4); | |
286 | if (sizeof(size_t) == 8) | |
287 | Z.hi ^= rem_4bit[rem]; | |
288 | else | |
289 | Z.hi ^= (u64)rem_4bit[rem] << 32; | |
290 | ||
291 | Z.hi ^= Htable[nlo].hi; | |
292 | Z.lo ^= Htable[nlo].lo; | |
293 | } | |
e7f5b1cd | 294 | |
e23d850f | 295 | if (IS_LITTLE_ENDIAN) { |
0f113f3e MC |
296 | # ifdef BSWAP8 |
297 | Xi[0] = BSWAP8(Z.hi); | |
298 | Xi[1] = BSWAP8(Z.lo); | |
299 | # else | |
300 | u8 *p = (u8 *)Xi; | |
301 | u32 v; | |
302 | v = (u32)(Z.hi >> 32); | |
303 | PUTU32(p, v); | |
304 | v = (u32)(Z.hi); | |
305 | PUTU32(p + 4, v); | |
306 | v = (u32)(Z.lo >> 32); | |
307 | PUTU32(p + 8, v); | |
308 | v = (u32)(Z.lo); | |
309 | PUTU32(p + 12, v); | |
310 | # endif | |
311 | } else { | |
312 | Xi[0] = Z.hi; | |
313 | Xi[1] = Z.lo; | |
314 | } | |
36c269c3 DF |
315 | |
316 | inp += 16; | |
317 | /* Block size is 128 bits so len is a multiple of 16 */ | |
318 | len -= 16; | |
319 | } while (len > 0); | |
e7f5b1cd | 320 | } |
0f113f3e MC |
321 | # endif |
322 | # else | |
323 | void gcm_gmult_4bit(u64 Xi[2], const u128 Htable[16]); | |
324 | void gcm_ghash_4bit(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
325 | size_t len); | |
326 | # endif | |
2262beef | 327 | |
d50e0934 | 328 | # define GCM_MUL(ctx) ctx->funcs.gmult(ctx->Xi.u,ctx->Htable) |
0f113f3e | 329 | # if defined(GHASH_ASM) || !defined(OPENSSL_SMALL_FOOTPRINT) |
95201ef4 | 330 | # define GHASH(ctx,in,len) ctx->funcs.ghash((ctx)->Xi.u,(ctx)->Htable,in,len) |
0f113f3e MC |
331 | /* |
332 | * GHASH_CHUNK is "stride parameter" missioned to mitigate cache trashing | |
333 | * effect. In other words idea is to hash data while it's still in L1 cache | |
334 | * after encryption pass... | |
335 | */ | |
336 | # define GHASH_CHUNK (3*1024) | |
337 | # endif | |
2262beef | 338 | |
7b6e19fc | 339 | #if (defined(GHASH_ASM) || defined(OPENSSL_CPUID_OBJ)) |
0f113f3e MC |
340 | # if !defined(I386_ONLY) && \ |
341 | (defined(__i386) || defined(__i386__) || \ | |
342 | defined(__x86_64) || defined(__x86_64__) || \ | |
343 | defined(_M_IX86) || defined(_M_AMD64) || defined(_M_X64)) | |
1e863180 | 344 | # define GHASH_ASM_X86_OR_64 |
c1f092d1 | 345 | |
0f113f3e MC |
346 | void gcm_init_clmul(u128 Htable[16], const u64 Xi[2]); |
347 | void gcm_gmult_clmul(u64 Xi[2], const u128 Htable[16]); | |
348 | void gcm_ghash_clmul(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
349 | size_t len); | |
c1f092d1 | 350 | |
0f113f3e MC |
351 | # if defined(__i386) || defined(__i386__) || defined(_M_IX86) |
352 | # define gcm_init_avx gcm_init_clmul | |
353 | # define gcm_gmult_avx gcm_gmult_clmul | |
354 | # define gcm_ghash_avx gcm_ghash_clmul | |
355 | # else | |
356 | void gcm_init_avx(u128 Htable[16], const u64 Xi[2]); | |
357 | void gcm_gmult_avx(u64 Xi[2], const u128 Htable[16]); | |
358 | void gcm_ghash_avx(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
359 | size_t len); | |
360 | # endif | |
1da5d302 | 361 | |
0f113f3e | 362 | # if defined(__i386) || defined(__i386__) || defined(_M_IX86) |
1e863180 | 363 | # define GHASH_ASM_X86 |
0f113f3e MC |
364 | void gcm_gmult_4bit_mmx(u64 Xi[2], const u128 Htable[16]); |
365 | void gcm_ghash_4bit_mmx(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
366 | size_t len); | |
c1f092d1 | 367 | |
0f113f3e MC |
368 | void gcm_gmult_4bit_x86(u64 Xi[2], const u128 Htable[16]); |
369 | void gcm_ghash_4bit_x86(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
370 | size_t len); | |
1e863180 | 371 | # endif |
82741e9c | 372 | # elif defined(__arm__) || defined(__arm) || defined(__aarch64__) |
1e863180 | 373 | # include "arm_arch.h" |
c1669e1c | 374 | # if __ARM_MAX_ARCH__>=7 |
1e863180 | 375 | # define GHASH_ASM_ARM |
0f113f3e | 376 | # define PMULL_CAPABLE (OPENSSL_armcap_P & ARMV8_PMULL) |
82741e9c | 377 | # if defined(__arm__) || defined(__arm) |
0f113f3e | 378 | # define NEON_CAPABLE (OPENSSL_armcap_P & ARMV7_NEON) |
82741e9c | 379 | # endif |
0f113f3e MC |
380 | void gcm_init_neon(u128 Htable[16], const u64 Xi[2]); |
381 | void gcm_gmult_neon(u64 Xi[2], const u128 Htable[16]); | |
382 | void gcm_ghash_neon(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
383 | size_t len); | |
384 | void gcm_init_v8(u128 Htable[16], const u64 Xi[2]); | |
385 | void gcm_gmult_v8(u64 Xi[2], const u128 Htable[16]); | |
386 | void gcm_ghash_v8(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
387 | size_t len); | |
1e863180 | 388 | # endif |
23328d4b | 389 | # elif defined(__sparc__) || defined(__sparc) |
52f7e44e | 390 | # include "crypto/sparc_arch.h" |
23328d4b | 391 | # define GHASH_ASM_SPARC |
0f113f3e MC |
392 | void gcm_init_vis3(u128 Htable[16], const u64 Xi[2]); |
393 | void gcm_gmult_vis3(u64 Xi[2], const u128 Htable[16]); | |
394 | void gcm_ghash_vis3(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
395 | size_t len); | |
396 | # elif defined(OPENSSL_CPUID_OBJ) && (defined(__powerpc__) || defined(__ppc__) || defined(_ARCH_PPC)) | |
3d178db7 | 397 | # include "crypto/ppc_arch.h" |
0e716d92 | 398 | # define GHASH_ASM_PPC |
0f113f3e MC |
399 | void gcm_init_p8(u128 Htable[16], const u64 Xi[2]); |
400 | void gcm_gmult_p8(u64 Xi[2], const u128 Htable[16]); | |
401 | void gcm_ghash_p8(u64 Xi[2], const u128 Htable[16], const u8 *inp, | |
402 | size_t len); | |
999376dc HB |
403 | # elif defined(OPENSSL_CPUID_OBJ) && defined(__riscv) && __riscv_xlen == 64 |
404 | # include "crypto/riscv_arch.h" | |
f3fed0d5 | 405 | # define GHASH_ASM_RV64I |
b2468436 CM |
406 | /* Zbc/Zbkc (scalar crypto with clmul) based routines. */ |
407 | void gcm_init_rv64i_zbc(u128 Htable[16], const u64 Xi[2]); | |
408 | void gcm_init_rv64i_zbc__zbb(u128 Htable[16], const u64 Xi[2]); | |
409 | void gcm_init_rv64i_zbc__zbkb(u128 Htable[16], const u64 Xi[2]); | |
410 | void gcm_gmult_rv64i_zbc(u64 Xi[2], const u128 Htable[16]); | |
411 | void gcm_gmult_rv64i_zbc__zbkb(u64 Xi[2], const u128 Htable[16]); | |
f3fed0d5 CM |
412 | void gcm_ghash_rv64i_zbc(u64 Xi[2], const u128 Htable[16], |
413 | const u8 *inp, size_t len); | |
414 | void gcm_ghash_rv64i_zbc__zbkb(u64 Xi[2], const u128 Htable[16], | |
415 | const u8 *inp, size_t len); | |
c1f092d1 | 416 | # endif |
c1f092d1 AP |
417 | #endif |
418 | ||
92c9086e | 419 | static void gcm_get_funcs(struct gcm_funcs_st *ctx) |
e7f5b1cd | 420 | { |
92c9086e TS |
421 | /* set defaults -- overridden below as needed */ |
422 | ctx->ginit = gcm_init_4bit; | |
f3fed0d5 | 423 | #if !defined(GHASH_ASM) |
92c9086e | 424 | ctx->gmult = gcm_gmult_4bit; |
e7f5b1cd | 425 | #else |
92c9086e | 426 | ctx->gmult = NULL; |
e7f5b1cd | 427 | #endif |
92c9086e TS |
428 | #if !defined(GHASH_ASM) && !defined(OPENSSL_SMALL_FOOTPRINT) |
429 | ctx->ghash = gcm_ghash_4bit; | |
7b6e19fc | 430 | #else |
92c9086e | 431 | ctx->ghash = NULL; |
7b6e19fc | 432 | #endif |
92c9086e TS |
433 | |
434 | #if defined(GHASH_ASM_X86_OR_64) | |
435 | # if !defined(GHASH_ASM_X86) || defined(OPENSSL_IA32_SSE2) | |
436 | /* x86_64 */ | |
6e5a853b | 437 | if (OPENSSL_ia32cap_P[1] & (1 << 1)) { /* check PCLMULQDQ bit */ |
0f113f3e | 438 | if (((OPENSSL_ia32cap_P[1] >> 22) & 0x41) == 0x41) { /* AVX+MOVBE */ |
92c9086e | 439 | ctx->ginit = gcm_init_avx; |
0f113f3e | 440 | ctx->gmult = gcm_gmult_avx; |
92c9086e | 441 | ctx->ghash = gcm_ghash_avx; |
0f113f3e | 442 | } else { |
92c9086e | 443 | ctx->ginit = gcm_init_clmul; |
0f113f3e | 444 | ctx->gmult = gcm_gmult_clmul; |
92c9086e | 445 | ctx->ghash = gcm_ghash_clmul; |
0f113f3e MC |
446 | } |
447 | return; | |
448 | } | |
7b6e19fc | 449 | # endif |
92c9086e TS |
450 | # if defined(GHASH_ASM_X86) |
451 | /* x86 only */ | |
452 | # if defined(OPENSSL_IA32_SSE2) | |
0f113f3e | 453 | if (OPENSSL_ia32cap_P[0] & (1 << 25)) { /* check SSE bit */ |
92c9086e TS |
454 | ctx->gmult = gcm_gmult_4bit_mmx; |
455 | ctx->ghash = gcm_ghash_4bit_mmx; | |
456 | return; | |
457 | } | |
7b6e19fc | 458 | # else |
0f113f3e | 459 | if (OPENSSL_ia32cap_P[0] & (1 << 23)) { /* check MMX bit */ |
0f113f3e | 460 | ctx->gmult = gcm_gmult_4bit_mmx; |
92c9086e TS |
461 | ctx->ghash = gcm_ghash_4bit_mmx; |
462 | return; | |
0f113f3e | 463 | } |
92c9086e TS |
464 | # endif |
465 | ctx->gmult = gcm_gmult_4bit_x86; | |
466 | ctx->ghash = gcm_ghash_4bit_x86; | |
467 | return; | |
be0161ff TM |
468 | # else |
469 | /* x86_64 fallback defaults */ | |
470 | ctx->gmult = gcm_gmult_4bit; | |
471 | ctx->ghash = gcm_ghash_4bit; | |
472 | return; | |
7b6e19fc | 473 | # endif |
92c9086e | 474 | #elif defined(GHASH_ASM_ARM) |
186be8ed TM |
475 | /* ARM defaults */ |
476 | ctx->gmult = gcm_gmult_4bit; | |
477 | ctx->ghash = gcm_ghash_4bit; | |
7b6e19fc | 478 | # ifdef PMULL_CAPABLE |
0f113f3e | 479 | if (PMULL_CAPABLE) { |
92c9086e | 480 | ctx->ginit = (gcm_init_fn)gcm_init_v8; |
0f113f3e | 481 | ctx->gmult = gcm_gmult_v8; |
92c9086e TS |
482 | ctx->ghash = gcm_ghash_v8; |
483 | } | |
484 | # elif defined(NEON_CAPABLE) | |
0f113f3e | 485 | if (NEON_CAPABLE) { |
92c9086e | 486 | ctx->ginit = gcm_init_neon; |
0f113f3e | 487 | ctx->gmult = gcm_gmult_neon; |
92c9086e | 488 | ctx->ghash = gcm_ghash_neon; |
0f113f3e | 489 | } |
92c9086e TS |
490 | # endif |
491 | return; | |
492 | #elif defined(GHASH_ASM_SPARC) | |
186be8ed TM |
493 | /* SPARC defaults */ |
494 | ctx->gmult = gcm_gmult_4bit; | |
495 | ctx->ghash = gcm_ghash_4bit; | |
0f113f3e | 496 | if (OPENSSL_sparcv9cap_P[0] & SPARCV9_VIS3) { |
92c9086e | 497 | ctx->ginit = gcm_init_vis3; |
0f113f3e | 498 | ctx->gmult = gcm_gmult_vis3; |
92c9086e | 499 | ctx->ghash = gcm_ghash_vis3; |
0f113f3e | 500 | } |
92c9086e TS |
501 | return; |
502 | #elif defined(GHASH_ASM_PPC) | |
186be8ed | 503 | /* PowerPC does not define GHASH_ASM; defaults set above */ |
0f113f3e | 504 | if (OPENSSL_ppccap_P & PPC_CRYPTO207) { |
92c9086e | 505 | ctx->ginit = gcm_init_p8; |
0f113f3e | 506 | ctx->gmult = gcm_gmult_p8; |
92c9086e | 507 | ctx->ghash = gcm_ghash_p8; |
0f113f3e | 508 | } |
92c9086e | 509 | return; |
f3fed0d5 CM |
510 | #elif defined(GHASH_ASM_RV64I) |
511 | /* RISCV defaults */ | |
512 | ctx->gmult = gcm_gmult_4bit; | |
513 | ctx->ghash = gcm_ghash_4bit; | |
514 | ||
b2468436 CM |
515 | if (RISCV_HAS_ZBC()) { |
516 | if (RISCV_HAS_ZBKB()) { | |
517 | ctx->ginit = gcm_init_rv64i_zbc__zbkb; | |
518 | ctx->gmult = gcm_gmult_rv64i_zbc__zbkb; | |
f3fed0d5 | 519 | ctx->ghash = gcm_ghash_rv64i_zbc__zbkb; |
b2468436 CM |
520 | } else if (RISCV_HAS_ZBB()) { |
521 | ctx->ginit = gcm_init_rv64i_zbc__zbb; | |
522 | ctx->gmult = gcm_gmult_rv64i_zbc; | |
f3fed0d5 | 523 | ctx->ghash = gcm_ghash_rv64i_zbc; |
b2468436 CM |
524 | } else { |
525 | ctx->ginit = gcm_init_rv64i_zbc; | |
526 | ctx->gmult = gcm_gmult_rv64i_zbc; | |
f3fed0d5 | 527 | ctx->ghash = gcm_ghash_rv64i_zbc; |
b2468436 | 528 | } |
999376dc | 529 | } |
92c9086e | 530 | return; |
186be8ed TM |
531 | #elif defined(GHASH_ASM) |
532 | /* all other architectures use the generic names */ | |
48e35b99 JC |
533 | ctx->gmult = gcm_gmult_4bit; |
534 | ctx->ghash = gcm_ghash_4bit; | |
535 | return; | |
536 | #endif | |
92c9086e TS |
537 | } |
538 | ||
0113ec84 TS |
539 | void ossl_gcm_init_4bit(u128 Htable[16], const u64 H[2]) |
540 | { | |
541 | struct gcm_funcs_st funcs; | |
542 | ||
543 | gcm_get_funcs(&funcs); | |
544 | funcs.ginit(Htable, H); | |
545 | } | |
546 | ||
547 | void ossl_gcm_gmult_4bit(u64 Xi[2], const u128 Htable[16]) | |
548 | { | |
549 | struct gcm_funcs_st funcs; | |
550 | ||
551 | gcm_get_funcs(&funcs); | |
552 | funcs.gmult(Xi, Htable); | |
553 | } | |
554 | ||
555 | void ossl_gcm_ghash_4bit(u64 Xi[2], const u128 Htable[16], | |
556 | const u8 *inp, size_t len) | |
557 | { | |
558 | struct gcm_funcs_st funcs; | |
559 | u64 tmp[2]; | |
560 | size_t i; | |
561 | ||
562 | gcm_get_funcs(&funcs); | |
563 | if (funcs.ghash != NULL) { | |
564 | funcs.ghash(Xi, Htable, inp, len); | |
565 | } else { | |
566 | /* Emulate ghash if needed */ | |
567 | for (i = 0; i < len; i += 16) { | |
568 | memcpy(tmp, &inp[i], sizeof(tmp)); | |
569 | Xi[0] ^= tmp[0]; | |
570 | Xi[1] ^= tmp[1]; | |
571 | funcs.gmult(Xi, Htable); | |
572 | } | |
573 | } | |
574 | } | |
575 | ||
92c9086e TS |
576 | void CRYPTO_gcm128_init(GCM128_CONTEXT *ctx, void *key, block128_f block) |
577 | { | |
578 | DECLARE_IS_ENDIAN; | |
579 | ||
580 | memset(ctx, 0, sizeof(*ctx)); | |
581 | ctx->block = block; | |
582 | ctx->key = key; | |
583 | ||
584 | (*block) (ctx->H.c, ctx->H.c, key); | |
585 | ||
586 | if (IS_LITTLE_ENDIAN) { | |
587 | /* H is stored in host byte order */ | |
588 | #ifdef BSWAP8 | |
589 | ctx->H.u[0] = BSWAP8(ctx->H.u[0]); | |
590 | ctx->H.u[1] = BSWAP8(ctx->H.u[1]); | |
7b6e19fc | 591 | #else |
92c9086e TS |
592 | u8 *p = ctx->H.c; |
593 | u64 hi, lo; | |
594 | hi = (u64)GETU32(p) << 32 | GETU32(p + 4); | |
595 | lo = (u64)GETU32(p + 8) << 32 | GETU32(p + 12); | |
596 | ctx->H.u[0] = hi; | |
597 | ctx->H.u[1] = lo; | |
a595baff | 598 | #endif |
92c9086e TS |
599 | } |
600 | ||
601 | gcm_get_funcs(&ctx->funcs); | |
602 | ctx->funcs.ginit(ctx->Htable, ctx->H.u); | |
e7f5b1cd AP |
603 | } |
604 | ||
0f113f3e MC |
605 | void CRYPTO_gcm128_setiv(GCM128_CONTEXT *ctx, const unsigned char *iv, |
606 | size_t len) | |
e7f5b1cd | 607 | { |
e23d850f | 608 | DECLARE_IS_ENDIAN; |
0f113f3e | 609 | unsigned int ctr; |
0f113f3e | 610 | |
0f113f3e MC |
611 | ctx->len.u[0] = 0; /* AAD length */ |
612 | ctx->len.u[1] = 0; /* message length */ | |
613 | ctx->ares = 0; | |
614 | ctx->mres = 0; | |
615 | ||
616 | if (len == 12) { | |
617 | memcpy(ctx->Yi.c, iv, 12); | |
f5791af3 AP |
618 | ctx->Yi.c[12] = 0; |
619 | ctx->Yi.c[13] = 0; | |
620 | ctx->Yi.c[14] = 0; | |
0f113f3e MC |
621 | ctx->Yi.c[15] = 1; |
622 | ctr = 1; | |
623 | } else { | |
624 | size_t i; | |
625 | u64 len0 = len; | |
626 | ||
f5791af3 AP |
627 | /* Borrow ctx->Xi to calculate initial Yi */ |
628 | ctx->Xi.u[0] = 0; | |
629 | ctx->Xi.u[1] = 0; | |
630 | ||
0f113f3e MC |
631 | while (len >= 16) { |
632 | for (i = 0; i < 16; ++i) | |
f5791af3 AP |
633 | ctx->Xi.c[i] ^= iv[i]; |
634 | GCM_MUL(ctx); | |
0f113f3e MC |
635 | iv += 16; |
636 | len -= 16; | |
637 | } | |
638 | if (len) { | |
639 | for (i = 0; i < len; ++i) | |
f5791af3 AP |
640 | ctx->Xi.c[i] ^= iv[i]; |
641 | GCM_MUL(ctx); | |
0f113f3e MC |
642 | } |
643 | len0 <<= 3; | |
e23d850f | 644 | if (IS_LITTLE_ENDIAN) { |
e7f5b1cd | 645 | #ifdef BSWAP8 |
f5791af3 | 646 | ctx->Xi.u[1] ^= BSWAP8(len0); |
e7f5b1cd | 647 | #else |
f5791af3 AP |
648 | ctx->Xi.c[8] ^= (u8)(len0 >> 56); |
649 | ctx->Xi.c[9] ^= (u8)(len0 >> 48); | |
650 | ctx->Xi.c[10] ^= (u8)(len0 >> 40); | |
651 | ctx->Xi.c[11] ^= (u8)(len0 >> 32); | |
652 | ctx->Xi.c[12] ^= (u8)(len0 >> 24); | |
653 | ctx->Xi.c[13] ^= (u8)(len0 >> 16); | |
654 | ctx->Xi.c[14] ^= (u8)(len0 >> 8); | |
655 | ctx->Xi.c[15] ^= (u8)(len0); | |
e7f5b1cd | 656 | #endif |
f5791af3 AP |
657 | } else { |
658 | ctx->Xi.u[1] ^= len0; | |
659 | } | |
e7f5b1cd | 660 | |
f5791af3 | 661 | GCM_MUL(ctx); |
e7f5b1cd | 662 | |
e23d850f | 663 | if (IS_LITTLE_ENDIAN) |
997d1aac | 664 | #ifdef BSWAP4 |
f5791af3 | 665 | ctr = BSWAP4(ctx->Xi.d[3]); |
997d1aac | 666 | #else |
f5791af3 | 667 | ctr = GETU32(ctx->Xi.c + 12); |
997d1aac | 668 | #endif |
0f113f3e | 669 | else |
f5791af3 AP |
670 | ctr = ctx->Xi.d[3]; |
671 | ||
672 | /* Copy borrowed Xi to Yi */ | |
673 | ctx->Yi.u[0] = ctx->Xi.u[0]; | |
674 | ctx->Yi.u[1] = ctx->Xi.u[1]; | |
0f113f3e | 675 | } |
e7f5b1cd | 676 | |
f5791af3 AP |
677 | ctx->Xi.u[0] = 0; |
678 | ctx->Xi.u[1] = 0; | |
679 | ||
0f113f3e MC |
680 | (*ctx->block) (ctx->Yi.c, ctx->EK0.c, ctx->key); |
681 | ++ctr; | |
e23d850f | 682 | if (IS_LITTLE_ENDIAN) |
997d1aac | 683 | #ifdef BSWAP4 |
0f113f3e | 684 | ctx->Yi.d[3] = BSWAP4(ctr); |
997d1aac | 685 | #else |
0f113f3e | 686 | PUTU32(ctx->Yi.c + 12, ctr); |
997d1aac | 687 | #endif |
0f113f3e MC |
688 | else |
689 | ctx->Yi.d[3] = ctr; | |
e7f5b1cd AP |
690 | } |
691 | ||
0f113f3e MC |
692 | int CRYPTO_gcm128_aad(GCM128_CONTEXT *ctx, const unsigned char *aad, |
693 | size_t len) | |
e7f5b1cd | 694 | { |
0f113f3e MC |
695 | size_t i; |
696 | unsigned int n; | |
697 | u64 alen = ctx->len.u[0]; | |
e7f5b1cd | 698 | |
0f113f3e MC |
699 | if (ctx->len.u[1]) |
700 | return -2; | |
701 | ||
702 | alen += len; | |
703 | if (alen > (U64(1) << 61) || (sizeof(len) == 8 && alen < len)) | |
704 | return -1; | |
705 | ctx->len.u[0] = alen; | |
706 | ||
707 | n = ctx->ares; | |
708 | if (n) { | |
709 | while (n && len) { | |
710 | ctx->Xi.c[n] ^= *(aad++); | |
711 | --len; | |
712 | n = (n + 1) % 16; | |
713 | } | |
714 | if (n == 0) | |
f5791af3 | 715 | GCM_MUL(ctx); |
0f113f3e MC |
716 | else { |
717 | ctx->ares = n; | |
718 | return 0; | |
719 | } | |
720 | } | |
2262beef | 721 | #ifdef GHASH |
0f113f3e MC |
722 | if ((i = (len & (size_t)-16))) { |
723 | GHASH(ctx, aad, i); | |
724 | aad += i; | |
725 | len -= i; | |
726 | } | |
2262beef | 727 | #else |
0f113f3e MC |
728 | while (len >= 16) { |
729 | for (i = 0; i < 16; ++i) | |
730 | ctx->Xi.c[i] ^= aad[i]; | |
f5791af3 | 731 | GCM_MUL(ctx); |
0f113f3e MC |
732 | aad += 16; |
733 | len -= 16; | |
734 | } | |
2262beef | 735 | #endif |
0f113f3e MC |
736 | if (len) { |
737 | n = (unsigned int)len; | |
738 | for (i = 0; i < len; ++i) | |
739 | ctx->Xi.c[i] ^= aad[i]; | |
740 | } | |
b68c1315 | 741 | |
0f113f3e MC |
742 | ctx->ares = n; |
743 | return 0; | |
e7f5b1cd AP |
744 | } |
745 | ||
1f2502eb | 746 | int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx, |
0f113f3e MC |
747 | const unsigned char *in, unsigned char *out, |
748 | size_t len) | |
e7f5b1cd | 749 | { |
e23d850f | 750 | DECLARE_IS_ENDIAN; |
c1b2569d | 751 | unsigned int n, ctr, mres; |
0f113f3e MC |
752 | size_t i; |
753 | u64 mlen = ctx->len.u[1]; | |
754 | block128_f block = ctx->block; | |
755 | void *key = ctx->key; | |
1f2502eb | 756 | |
0f113f3e MC |
757 | mlen += len; |
758 | if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len)) | |
759 | return -1; | |
760 | ctx->len.u[1] = mlen; | |
e7f5b1cd | 761 | |
c1b2569d AP |
762 | mres = ctx->mres; |
763 | ||
0f113f3e MC |
764 | if (ctx->ares) { |
765 | /* First call to encrypt finalizes GHASH(AAD) */ | |
c1b2569d AP |
766 | #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT) |
767 | if (len == 0) { | |
768 | GCM_MUL(ctx); | |
769 | ctx->ares = 0; | |
770 | return 0; | |
771 | } | |
772 | memcpy(ctx->Xn, ctx->Xi.c, sizeof(ctx->Xi)); | |
773 | ctx->Xi.u[0] = 0; | |
774 | ctx->Xi.u[1] = 0; | |
775 | mres = sizeof(ctx->Xi); | |
776 | #else | |
f5791af3 | 777 | GCM_MUL(ctx); |
c1b2569d | 778 | #endif |
0f113f3e MC |
779 | ctx->ares = 0; |
780 | } | |
96a4cf8c | 781 | |
e23d850f | 782 | if (IS_LITTLE_ENDIAN) |
997d1aac | 783 | #ifdef BSWAP4 |
0f113f3e | 784 | ctr = BSWAP4(ctx->Yi.d[3]); |
997d1aac | 785 | #else |
0f113f3e | 786 | ctr = GETU32(ctx->Yi.c + 12); |
997d1aac | 787 | #endif |
0f113f3e MC |
788 | else |
789 | ctr = ctx->Yi.d[3]; | |
96a4cf8c | 790 | |
c1b2569d | 791 | n = mres % 16; |
0f113f3e MC |
792 | #if !defined(OPENSSL_SMALL_FOOTPRINT) |
793 | if (16 % sizeof(size_t) == 0) { /* always true actually */ | |
794 | do { | |
795 | if (n) { | |
c1b2569d AP |
796 | # if defined(GHASH) |
797 | while (n && len) { | |
798 | ctx->Xn[mres++] = *(out++) = *(in++) ^ ctx->EKi.c[n]; | |
799 | --len; | |
800 | n = (n + 1) % 16; | |
801 | } | |
802 | if (n == 0) { | |
803 | GHASH(ctx, ctx->Xn, mres); | |
804 | mres = 0; | |
805 | } else { | |
806 | ctx->mres = mres; | |
807 | return 0; | |
808 | } | |
809 | # else | |
0f113f3e MC |
810 | while (n && len) { |
811 | ctx->Xi.c[n] ^= *(out++) = *(in++) ^ ctx->EKi.c[n]; | |
812 | --len; | |
813 | n = (n + 1) % 16; | |
814 | } | |
c1b2569d | 815 | if (n == 0) { |
f5791af3 | 816 | GCM_MUL(ctx); |
c1b2569d AP |
817 | mres = 0; |
818 | } else { | |
0f113f3e MC |
819 | ctx->mres = n; |
820 | return 0; | |
821 | } | |
c1b2569d | 822 | # endif |
0f113f3e MC |
823 | } |
824 | # if defined(STRICT_ALIGNMENT) | |
825 | if (((size_t)in | (size_t)out) % sizeof(size_t) != 0) | |
826 | break; | |
827 | # endif | |
2e635aa8 | 828 | # if defined(GHASH) |
c1b2569d AP |
829 | if (len >= 16 && mres) { |
830 | GHASH(ctx, ctx->Xn, mres); | |
831 | mres = 0; | |
832 | } | |
2e635aa8 | 833 | # if defined(GHASH_CHUNK) |
0f113f3e MC |
834 | while (len >= GHASH_CHUNK) { |
835 | size_t j = GHASH_CHUNK; | |
836 | ||
837 | while (j) { | |
77286fe3 BE |
838 | size_t_aX *out_t = (size_t_aX *)out; |
839 | const size_t_aX *in_t = (const size_t_aX *)in; | |
0f113f3e MC |
840 | |
841 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
842 | ++ctr; | |
e23d850f | 843 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 844 | # ifdef BSWAP4 |
0f113f3e | 845 | ctx->Yi.d[3] = BSWAP4(ctr); |
2e635aa8 | 846 | # else |
0f113f3e | 847 | PUTU32(ctx->Yi.c + 12, ctr); |
2e635aa8 | 848 | # endif |
0f113f3e MC |
849 | else |
850 | ctx->Yi.d[3] = ctr; | |
851 | for (i = 0; i < 16 / sizeof(size_t); ++i) | |
852 | out_t[i] = in_t[i] ^ ctx->EKi.t[i]; | |
853 | out += 16; | |
854 | in += 16; | |
855 | j -= 16; | |
856 | } | |
857 | GHASH(ctx, out - GHASH_CHUNK, GHASH_CHUNK); | |
858 | len -= GHASH_CHUNK; | |
859 | } | |
2e635aa8 | 860 | # endif |
0f113f3e MC |
861 | if ((i = (len & (size_t)-16))) { |
862 | size_t j = i; | |
863 | ||
864 | while (len >= 16) { | |
77286fe3 BE |
865 | size_t_aX *out_t = (size_t_aX *)out; |
866 | const size_t_aX *in_t = (const size_t_aX *)in; | |
0f113f3e MC |
867 | |
868 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
869 | ++ctr; | |
e23d850f | 870 | if (IS_LITTLE_ENDIAN) |
0f113f3e MC |
871 | # ifdef BSWAP4 |
872 | ctx->Yi.d[3] = BSWAP4(ctr); | |
873 | # else | |
874 | PUTU32(ctx->Yi.c + 12, ctr); | |
875 | # endif | |
876 | else | |
877 | ctx->Yi.d[3] = ctr; | |
878 | for (i = 0; i < 16 / sizeof(size_t); ++i) | |
879 | out_t[i] = in_t[i] ^ ctx->EKi.t[i]; | |
880 | out += 16; | |
881 | in += 16; | |
882 | len -= 16; | |
883 | } | |
884 | GHASH(ctx, out - j, j); | |
885 | } | |
886 | # else | |
887 | while (len >= 16) { | |
888 | size_t *out_t = (size_t *)out; | |
889 | const size_t *in_t = (const size_t *)in; | |
890 | ||
891 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
892 | ++ctr; | |
e23d850f | 893 | if (IS_LITTLE_ENDIAN) |
0f113f3e MC |
894 | # ifdef BSWAP4 |
895 | ctx->Yi.d[3] = BSWAP4(ctr); | |
896 | # else | |
897 | PUTU32(ctx->Yi.c + 12, ctr); | |
898 | # endif | |
899 | else | |
900 | ctx->Yi.d[3] = ctr; | |
901 | for (i = 0; i < 16 / sizeof(size_t); ++i) | |
902 | ctx->Xi.t[i] ^= out_t[i] = in_t[i] ^ ctx->EKi.t[i]; | |
f5791af3 | 903 | GCM_MUL(ctx); |
0f113f3e MC |
904 | out += 16; |
905 | in += 16; | |
906 | len -= 16; | |
907 | } | |
908 | # endif | |
909 | if (len) { | |
910 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
911 | ++ctr; | |
e23d850f | 912 | if (IS_LITTLE_ENDIAN) |
0f113f3e MC |
913 | # ifdef BSWAP4 |
914 | ctx->Yi.d[3] = BSWAP4(ctr); | |
915 | # else | |
916 | PUTU32(ctx->Yi.c + 12, ctr); | |
917 | # endif | |
918 | else | |
919 | ctx->Yi.d[3] = ctr; | |
c1b2569d AP |
920 | # if defined(GHASH) |
921 | while (len--) { | |
922 | ctx->Xn[mres++] = out[n] = in[n] ^ ctx->EKi.c[n]; | |
923 | ++n; | |
924 | } | |
925 | # else | |
0f113f3e MC |
926 | while (len--) { |
927 | ctx->Xi.c[n] ^= out[n] = in[n] ^ ctx->EKi.c[n]; | |
928 | ++n; | |
929 | } | |
c1b2569d AP |
930 | mres = n; |
931 | # endif | |
0f113f3e MC |
932 | } |
933 | ||
c1b2569d | 934 | ctx->mres = mres; |
0f113f3e MC |
935 | return 0; |
936 | } while (0); | |
937 | } | |
e7f5b1cd | 938 | #endif |
0f113f3e MC |
939 | for (i = 0; i < len; ++i) { |
940 | if (n == 0) { | |
941 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
942 | ++ctr; | |
e23d850f | 943 | if (IS_LITTLE_ENDIAN) |
997d1aac | 944 | #ifdef BSWAP4 |
0f113f3e | 945 | ctx->Yi.d[3] = BSWAP4(ctr); |
997d1aac | 946 | #else |
0f113f3e MC |
947 | PUTU32(ctx->Yi.c + 12, ctr); |
948 | #endif | |
949 | else | |
950 | ctx->Yi.d[3] = ctr; | |
951 | } | |
c1b2569d AP |
952 | #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT) |
953 | ctx->Xn[mres++] = out[i] = in[i] ^ ctx->EKi.c[n]; | |
0f113f3e | 954 | n = (n + 1) % 16; |
c1b2569d AP |
955 | if (mres == sizeof(ctx->Xn)) { |
956 | GHASH(ctx,ctx->Xn,sizeof(ctx->Xn)); | |
957 | mres = 0; | |
958 | } | |
959 | #else | |
960 | ctx->Xi.c[n] ^= out[i] = in[i] ^ ctx->EKi.c[n]; | |
961 | mres = n = (n + 1) % 16; | |
0f113f3e | 962 | if (n == 0) |
f5791af3 | 963 | GCM_MUL(ctx); |
c1b2569d | 964 | #endif |
0f113f3e MC |
965 | } |
966 | ||
c1b2569d | 967 | ctx->mres = mres; |
0f113f3e | 968 | return 0; |
e7f5b1cd AP |
969 | } |
970 | ||
1f2502eb | 971 | int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx, |
0f113f3e MC |
972 | const unsigned char *in, unsigned char *out, |
973 | size_t len) | |
e7f5b1cd | 974 | { |
e23d850f | 975 | DECLARE_IS_ENDIAN; |
c1b2569d | 976 | unsigned int n, ctr, mres; |
0f113f3e MC |
977 | size_t i; |
978 | u64 mlen = ctx->len.u[1]; | |
979 | block128_f block = ctx->block; | |
980 | void *key = ctx->key; | |
1f2502eb | 981 | |
0f113f3e MC |
982 | mlen += len; |
983 | if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len)) | |
984 | return -1; | |
985 | ctx->len.u[1] = mlen; | |
e7f5b1cd | 986 | |
c1b2569d AP |
987 | mres = ctx->mres; |
988 | ||
0f113f3e MC |
989 | if (ctx->ares) { |
990 | /* First call to decrypt finalizes GHASH(AAD) */ | |
c1b2569d AP |
991 | #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT) |
992 | if (len == 0) { | |
993 | GCM_MUL(ctx); | |
994 | ctx->ares = 0; | |
995 | return 0; | |
996 | } | |
997 | memcpy(ctx->Xn, ctx->Xi.c, sizeof(ctx->Xi)); | |
998 | ctx->Xi.u[0] = 0; | |
999 | ctx->Xi.u[1] = 0; | |
1000 | mres = sizeof(ctx->Xi); | |
1001 | #else | |
f5791af3 | 1002 | GCM_MUL(ctx); |
c1b2569d | 1003 | #endif |
0f113f3e MC |
1004 | ctx->ares = 0; |
1005 | } | |
b68c1315 | 1006 | |
e23d850f | 1007 | if (IS_LITTLE_ENDIAN) |
997d1aac | 1008 | #ifdef BSWAP4 |
0f113f3e | 1009 | ctr = BSWAP4(ctx->Yi.d[3]); |
997d1aac | 1010 | #else |
0f113f3e | 1011 | ctr = GETU32(ctx->Yi.c + 12); |
997d1aac | 1012 | #endif |
0f113f3e MC |
1013 | else |
1014 | ctr = ctx->Yi.d[3]; | |
e7f5b1cd | 1015 | |
c1b2569d | 1016 | n = mres % 16; |
e7f5b1cd | 1017 | #if !defined(OPENSSL_SMALL_FOOTPRINT) |
0f113f3e MC |
1018 | if (16 % sizeof(size_t) == 0) { /* always true actually */ |
1019 | do { | |
1020 | if (n) { | |
c1b2569d AP |
1021 | # if defined(GHASH) |
1022 | while (n && len) { | |
1023 | *(out++) = (ctx->Xn[mres++] = *(in++)) ^ ctx->EKi.c[n]; | |
1024 | --len; | |
1025 | n = (n + 1) % 16; | |
1026 | } | |
1027 | if (n == 0) { | |
1028 | GHASH(ctx, ctx->Xn, mres); | |
1029 | mres = 0; | |
1030 | } else { | |
1031 | ctx->mres = mres; | |
1032 | return 0; | |
1033 | } | |
1034 | # else | |
0f113f3e MC |
1035 | while (n && len) { |
1036 | u8 c = *(in++); | |
1037 | *(out++) = c ^ ctx->EKi.c[n]; | |
1038 | ctx->Xi.c[n] ^= c; | |
1039 | --len; | |
1040 | n = (n + 1) % 16; | |
1041 | } | |
c1b2569d | 1042 | if (n == 0) { |
f5791af3 | 1043 | GCM_MUL(ctx); |
c1b2569d AP |
1044 | mres = 0; |
1045 | } else { | |
0f113f3e MC |
1046 | ctx->mres = n; |
1047 | return 0; | |
1048 | } | |
c1b2569d | 1049 | # endif |
0f113f3e MC |
1050 | } |
1051 | # if defined(STRICT_ALIGNMENT) | |
1052 | if (((size_t)in | (size_t)out) % sizeof(size_t) != 0) | |
1053 | break; | |
1054 | # endif | |
2e635aa8 | 1055 | # if defined(GHASH) |
c1b2569d AP |
1056 | if (len >= 16 && mres) { |
1057 | GHASH(ctx, ctx->Xn, mres); | |
1058 | mres = 0; | |
1059 | } | |
2e635aa8 | 1060 | # if defined(GHASH_CHUNK) |
0f113f3e MC |
1061 | while (len >= GHASH_CHUNK) { |
1062 | size_t j = GHASH_CHUNK; | |
1063 | ||
1064 | GHASH(ctx, in, GHASH_CHUNK); | |
1065 | while (j) { | |
77286fe3 BE |
1066 | size_t_aX *out_t = (size_t_aX *)out; |
1067 | const size_t_aX *in_t = (const size_t_aX *)in; | |
0f113f3e MC |
1068 | |
1069 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
1070 | ++ctr; | |
e23d850f | 1071 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 1072 | # ifdef BSWAP4 |
0f113f3e | 1073 | ctx->Yi.d[3] = BSWAP4(ctr); |
2e635aa8 | 1074 | # else |
0f113f3e | 1075 | PUTU32(ctx->Yi.c + 12, ctr); |
2e635aa8 | 1076 | # endif |
0f113f3e MC |
1077 | else |
1078 | ctx->Yi.d[3] = ctr; | |
1079 | for (i = 0; i < 16 / sizeof(size_t); ++i) | |
1080 | out_t[i] = in_t[i] ^ ctx->EKi.t[i]; | |
1081 | out += 16; | |
1082 | in += 16; | |
1083 | j -= 16; | |
1084 | } | |
1085 | len -= GHASH_CHUNK; | |
1086 | } | |
2e635aa8 | 1087 | # endif |
0f113f3e MC |
1088 | if ((i = (len & (size_t)-16))) { |
1089 | GHASH(ctx, in, i); | |
1090 | while (len >= 16) { | |
77286fe3 BE |
1091 | size_t_aX *out_t = (size_t_aX *)out; |
1092 | const size_t_aX *in_t = (const size_t_aX *)in; | |
0f113f3e MC |
1093 | |
1094 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
1095 | ++ctr; | |
e23d850f | 1096 | if (IS_LITTLE_ENDIAN) |
0f113f3e MC |
1097 | # ifdef BSWAP4 |
1098 | ctx->Yi.d[3] = BSWAP4(ctr); | |
1099 | # else | |
1100 | PUTU32(ctx->Yi.c + 12, ctr); | |
1101 | # endif | |
1102 | else | |
1103 | ctx->Yi.d[3] = ctr; | |
1104 | for (i = 0; i < 16 / sizeof(size_t); ++i) | |
1105 | out_t[i] = in_t[i] ^ ctx->EKi.t[i]; | |
1106 | out += 16; | |
1107 | in += 16; | |
1108 | len -= 16; | |
1109 | } | |
1110 | } | |
1111 | # else | |
1112 | while (len >= 16) { | |
1113 | size_t *out_t = (size_t *)out; | |
1114 | const size_t *in_t = (const size_t *)in; | |
1115 | ||
1116 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
1117 | ++ctr; | |
e23d850f | 1118 | if (IS_LITTLE_ENDIAN) |
0f113f3e MC |
1119 | # ifdef BSWAP4 |
1120 | ctx->Yi.d[3] = BSWAP4(ctr); | |
1121 | # else | |
1122 | PUTU32(ctx->Yi.c + 12, ctr); | |
1123 | # endif | |
1124 | else | |
1125 | ctx->Yi.d[3] = ctr; | |
1126 | for (i = 0; i < 16 / sizeof(size_t); ++i) { | |
1d724b5e ZJ |
1127 | size_t c = in_t[i]; |
1128 | out_t[i] = c ^ ctx->EKi.t[i]; | |
0f113f3e MC |
1129 | ctx->Xi.t[i] ^= c; |
1130 | } | |
f5791af3 | 1131 | GCM_MUL(ctx); |
0f113f3e MC |
1132 | out += 16; |
1133 | in += 16; | |
1134 | len -= 16; | |
1135 | } | |
1136 | # endif | |
1137 | if (len) { | |
1138 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
1139 | ++ctr; | |
e23d850f | 1140 | if (IS_LITTLE_ENDIAN) |
0f113f3e MC |
1141 | # ifdef BSWAP4 |
1142 | ctx->Yi.d[3] = BSWAP4(ctr); | |
1143 | # else | |
1144 | PUTU32(ctx->Yi.c + 12, ctr); | |
1145 | # endif | |
1146 | else | |
1147 | ctx->Yi.d[3] = ctr; | |
c1b2569d AP |
1148 | # if defined(GHASH) |
1149 | while (len--) { | |
1150 | out[n] = (ctx->Xn[mres++] = in[n]) ^ ctx->EKi.c[n]; | |
1151 | ++n; | |
1152 | } | |
1153 | # else | |
0f113f3e MC |
1154 | while (len--) { |
1155 | u8 c = in[n]; | |
1156 | ctx->Xi.c[n] ^= c; | |
1157 | out[n] = c ^ ctx->EKi.c[n]; | |
1158 | ++n; | |
1159 | } | |
c1b2569d AP |
1160 | mres = n; |
1161 | # endif | |
0f113f3e MC |
1162 | } |
1163 | ||
c1b2569d | 1164 | ctx->mres = mres; |
0f113f3e MC |
1165 | return 0; |
1166 | } while (0); | |
1167 | } | |
997d1aac | 1168 | #endif |
0f113f3e MC |
1169 | for (i = 0; i < len; ++i) { |
1170 | u8 c; | |
1171 | if (n == 0) { | |
1172 | (*block) (ctx->Yi.c, ctx->EKi.c, key); | |
1173 | ++ctr; | |
e23d850f | 1174 | if (IS_LITTLE_ENDIAN) |
997d1aac | 1175 | #ifdef BSWAP4 |
0f113f3e | 1176 | ctx->Yi.d[3] = BSWAP4(ctr); |
997d1aac | 1177 | #else |
0f113f3e MC |
1178 | PUTU32(ctx->Yi.c + 12, ctr); |
1179 | #endif | |
1180 | else | |
1181 | ctx->Yi.d[3] = ctr; | |
1182 | } | |
c1b2569d AP |
1183 | #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT) |
1184 | out[i] = (ctx->Xn[mres++] = c = in[i]) ^ ctx->EKi.c[n]; | |
1185 | n = (n + 1) % 16; | |
1186 | if (mres == sizeof(ctx->Xn)) { | |
1187 | GHASH(ctx,ctx->Xn,sizeof(ctx->Xn)); | |
1188 | mres = 0; | |
1189 | } | |
1190 | #else | |
0f113f3e MC |
1191 | c = in[i]; |
1192 | out[i] = c ^ ctx->EKi.c[n]; | |
1193 | ctx->Xi.c[n] ^= c; | |
c1b2569d | 1194 | mres = n = (n + 1) % 16; |
0f113f3e | 1195 | if (n == 0) |
f5791af3 | 1196 | GCM_MUL(ctx); |
c1b2569d | 1197 | #endif |
0f113f3e | 1198 | } |
96a4cf8c | 1199 | |
c1b2569d | 1200 | ctx->mres = mres; |
0f113f3e | 1201 | return 0; |
e7f5b1cd AP |
1202 | } |
1203 | ||
1f2502eb | 1204 | int CRYPTO_gcm128_encrypt_ctr32(GCM128_CONTEXT *ctx, |
0f113f3e MC |
1205 | const unsigned char *in, unsigned char *out, |
1206 | size_t len, ctr128_f stream) | |
f71c6ace | 1207 | { |
2e635aa8 AP |
1208 | #if defined(OPENSSL_SMALL_FOOTPRINT) |
1209 | return CRYPTO_gcm128_encrypt(ctx, in, out, len); | |
1210 | #else | |
e23d850f | 1211 | DECLARE_IS_ENDIAN; |
c1b2569d | 1212 | unsigned int n, ctr, mres; |
0f113f3e MC |
1213 | size_t i; |
1214 | u64 mlen = ctx->len.u[1]; | |
1215 | void *key = ctx->key; | |
1f2502eb | 1216 | |
0f113f3e MC |
1217 | mlen += len; |
1218 | if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len)) | |
1219 | return -1; | |
1220 | ctx->len.u[1] = mlen; | |
f71c6ace | 1221 | |
c1b2569d AP |
1222 | mres = ctx->mres; |
1223 | ||
0f113f3e MC |
1224 | if (ctx->ares) { |
1225 | /* First call to encrypt finalizes GHASH(AAD) */ | |
c1b2569d AP |
1226 | #if defined(GHASH) |
1227 | if (len == 0) { | |
1228 | GCM_MUL(ctx); | |
1229 | ctx->ares = 0; | |
1230 | return 0; | |
1231 | } | |
1232 | memcpy(ctx->Xn, ctx->Xi.c, sizeof(ctx->Xi)); | |
1233 | ctx->Xi.u[0] = 0; | |
1234 | ctx->Xi.u[1] = 0; | |
1235 | mres = sizeof(ctx->Xi); | |
1236 | #else | |
f5791af3 | 1237 | GCM_MUL(ctx); |
c1b2569d | 1238 | #endif |
0f113f3e MC |
1239 | ctx->ares = 0; |
1240 | } | |
b68c1315 | 1241 | |
e23d850f | 1242 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 1243 | # ifdef BSWAP4 |
0f113f3e | 1244 | ctr = BSWAP4(ctx->Yi.d[3]); |
2e635aa8 | 1245 | # else |
0f113f3e | 1246 | ctr = GETU32(ctx->Yi.c + 12); |
2e635aa8 | 1247 | # endif |
0f113f3e MC |
1248 | else |
1249 | ctr = ctx->Yi.d[3]; | |
1250 | ||
c1b2569d | 1251 | n = mres % 16; |
0f113f3e | 1252 | if (n) { |
c1b2569d AP |
1253 | # if defined(GHASH) |
1254 | while (n && len) { | |
1255 | ctx->Xn[mres++] = *(out++) = *(in++) ^ ctx->EKi.c[n]; | |
1256 | --len; | |
1257 | n = (n + 1) % 16; | |
1258 | } | |
1259 | if (n == 0) { | |
1260 | GHASH(ctx, ctx->Xn, mres); | |
1261 | mres = 0; | |
1262 | } else { | |
1263 | ctx->mres = mres; | |
1264 | return 0; | |
1265 | } | |
1266 | # else | |
0f113f3e MC |
1267 | while (n && len) { |
1268 | ctx->Xi.c[n] ^= *(out++) = *(in++) ^ ctx->EKi.c[n]; | |
1269 | --len; | |
1270 | n = (n + 1) % 16; | |
1271 | } | |
c1b2569d | 1272 | if (n == 0) { |
f5791af3 | 1273 | GCM_MUL(ctx); |
c1b2569d AP |
1274 | mres = 0; |
1275 | } else { | |
0f113f3e MC |
1276 | ctx->mres = n; |
1277 | return 0; | |
1278 | } | |
c1b2569d | 1279 | # endif |
0f113f3e | 1280 | } |
c1b2569d AP |
1281 | # if defined(GHASH) |
1282 | if (len >= 16 && mres) { | |
1283 | GHASH(ctx, ctx->Xn, mres); | |
1284 | mres = 0; | |
1285 | } | |
1286 | # if defined(GHASH_CHUNK) | |
0f113f3e MC |
1287 | while (len >= GHASH_CHUNK) { |
1288 | (*stream) (in, out, GHASH_CHUNK / 16, key, ctx->Yi.c); | |
1289 | ctr += GHASH_CHUNK / 16; | |
e23d850f | 1290 | if (IS_LITTLE_ENDIAN) |
c1b2569d | 1291 | # ifdef BSWAP4 |
0f113f3e | 1292 | ctx->Yi.d[3] = BSWAP4(ctr); |
c1b2569d | 1293 | # else |
0f113f3e | 1294 | PUTU32(ctx->Yi.c + 12, ctr); |
c1b2569d | 1295 | # endif |
0f113f3e MC |
1296 | else |
1297 | ctx->Yi.d[3] = ctr; | |
1298 | GHASH(ctx, out, GHASH_CHUNK); | |
1299 | out += GHASH_CHUNK; | |
1300 | in += GHASH_CHUNK; | |
1301 | len -= GHASH_CHUNK; | |
1302 | } | |
c1b2569d | 1303 | # endif |
2e635aa8 | 1304 | # endif |
0f113f3e MC |
1305 | if ((i = (len & (size_t)-16))) { |
1306 | size_t j = i / 16; | |
f71c6ace | 1307 | |
0f113f3e MC |
1308 | (*stream) (in, out, j, key, ctx->Yi.c); |
1309 | ctr += (unsigned int)j; | |
e23d850f | 1310 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 1311 | # ifdef BSWAP4 |
0f113f3e | 1312 | ctx->Yi.d[3] = BSWAP4(ctr); |
2e635aa8 | 1313 | # else |
0f113f3e | 1314 | PUTU32(ctx->Yi.c + 12, ctr); |
2e635aa8 | 1315 | # endif |
0f113f3e MC |
1316 | else |
1317 | ctx->Yi.d[3] = ctr; | |
1318 | in += i; | |
1319 | len -= i; | |
2e635aa8 | 1320 | # if defined(GHASH) |
0f113f3e MC |
1321 | GHASH(ctx, out, i); |
1322 | out += i; | |
2e635aa8 | 1323 | # else |
0f113f3e MC |
1324 | while (j--) { |
1325 | for (i = 0; i < 16; ++i) | |
1326 | ctx->Xi.c[i] ^= out[i]; | |
f5791af3 | 1327 | GCM_MUL(ctx); |
0f113f3e MC |
1328 | out += 16; |
1329 | } | |
2e635aa8 | 1330 | # endif |
0f113f3e MC |
1331 | } |
1332 | if (len) { | |
1333 | (*ctx->block) (ctx->Yi.c, ctx->EKi.c, key); | |
1334 | ++ctr; | |
e23d850f | 1335 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 1336 | # ifdef BSWAP4 |
0f113f3e | 1337 | ctx->Yi.d[3] = BSWAP4(ctr); |
2e635aa8 | 1338 | # else |
0f113f3e | 1339 | PUTU32(ctx->Yi.c + 12, ctr); |
2e635aa8 | 1340 | # endif |
0f113f3e MC |
1341 | else |
1342 | ctx->Yi.d[3] = ctr; | |
1343 | while (len--) { | |
c1b2569d AP |
1344 | # if defined(GHASH) |
1345 | ctx->Xn[mres++] = out[n] = in[n] ^ ctx->EKi.c[n]; | |
1346 | # else | |
1347 | ctx->Xi.c[mres++] ^= out[n] = in[n] ^ ctx->EKi.c[n]; | |
1348 | # endif | |
0f113f3e MC |
1349 | ++n; |
1350 | } | |
1351 | } | |
1352 | ||
c1b2569d | 1353 | ctx->mres = mres; |
0f113f3e | 1354 | return 0; |
2e635aa8 | 1355 | #endif |
f71c6ace AP |
1356 | } |
1357 | ||
1f2502eb | 1358 | int CRYPTO_gcm128_decrypt_ctr32(GCM128_CONTEXT *ctx, |
0f113f3e MC |
1359 | const unsigned char *in, unsigned char *out, |
1360 | size_t len, ctr128_f stream) | |
f71c6ace | 1361 | { |
2e635aa8 AP |
1362 | #if defined(OPENSSL_SMALL_FOOTPRINT) |
1363 | return CRYPTO_gcm128_decrypt(ctx, in, out, len); | |
1364 | #else | |
e23d850f | 1365 | DECLARE_IS_ENDIAN; |
c1b2569d | 1366 | unsigned int n, ctr, mres; |
0f113f3e MC |
1367 | size_t i; |
1368 | u64 mlen = ctx->len.u[1]; | |
1369 | void *key = ctx->key; | |
1f2502eb | 1370 | |
0f113f3e MC |
1371 | mlen += len; |
1372 | if (mlen > ((U64(1) << 36) - 32) || (sizeof(len) == 8 && mlen < len)) | |
1373 | return -1; | |
1374 | ctx->len.u[1] = mlen; | |
f71c6ace | 1375 | |
c1b2569d AP |
1376 | mres = ctx->mres; |
1377 | ||
0f113f3e MC |
1378 | if (ctx->ares) { |
1379 | /* First call to decrypt finalizes GHASH(AAD) */ | |
c1b2569d AP |
1380 | # if defined(GHASH) |
1381 | if (len == 0) { | |
1382 | GCM_MUL(ctx); | |
1383 | ctx->ares = 0; | |
1384 | return 0; | |
1385 | } | |
1386 | memcpy(ctx->Xn, ctx->Xi.c, sizeof(ctx->Xi)); | |
1387 | ctx->Xi.u[0] = 0; | |
1388 | ctx->Xi.u[1] = 0; | |
1389 | mres = sizeof(ctx->Xi); | |
1390 | # else | |
f5791af3 | 1391 | GCM_MUL(ctx); |
c1b2569d | 1392 | # endif |
0f113f3e MC |
1393 | ctx->ares = 0; |
1394 | } | |
b68c1315 | 1395 | |
e23d850f | 1396 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 1397 | # ifdef BSWAP4 |
0f113f3e | 1398 | ctr = BSWAP4(ctx->Yi.d[3]); |
2e635aa8 | 1399 | # else |
0f113f3e | 1400 | ctr = GETU32(ctx->Yi.c + 12); |
2e635aa8 | 1401 | # endif |
0f113f3e MC |
1402 | else |
1403 | ctr = ctx->Yi.d[3]; | |
1404 | ||
c1b2569d | 1405 | n = mres % 16; |
0f113f3e | 1406 | if (n) { |
c1b2569d AP |
1407 | # if defined(GHASH) |
1408 | while (n && len) { | |
1409 | *(out++) = (ctx->Xn[mres++] = *(in++)) ^ ctx->EKi.c[n]; | |
1410 | --len; | |
1411 | n = (n + 1) % 16; | |
1412 | } | |
1413 | if (n == 0) { | |
1414 | GHASH(ctx, ctx->Xn, mres); | |
1415 | mres = 0; | |
1416 | } else { | |
1417 | ctx->mres = mres; | |
1418 | return 0; | |
1419 | } | |
1420 | # else | |
0f113f3e MC |
1421 | while (n && len) { |
1422 | u8 c = *(in++); | |
1423 | *(out++) = c ^ ctx->EKi.c[n]; | |
1424 | ctx->Xi.c[n] ^= c; | |
1425 | --len; | |
1426 | n = (n + 1) % 16; | |
1427 | } | |
c1b2569d | 1428 | if (n == 0) { |
f5791af3 | 1429 | GCM_MUL(ctx); |
c1b2569d AP |
1430 | mres = 0; |
1431 | } else { | |
0f113f3e MC |
1432 | ctx->mres = n; |
1433 | return 0; | |
1434 | } | |
c1b2569d | 1435 | # endif |
0f113f3e | 1436 | } |
c1b2569d AP |
1437 | # if defined(GHASH) |
1438 | if (len >= 16 && mres) { | |
1439 | GHASH(ctx, ctx->Xn, mres); | |
1440 | mres = 0; | |
1441 | } | |
1442 | # if defined(GHASH_CHUNK) | |
0f113f3e MC |
1443 | while (len >= GHASH_CHUNK) { |
1444 | GHASH(ctx, in, GHASH_CHUNK); | |
1445 | (*stream) (in, out, GHASH_CHUNK / 16, key, ctx->Yi.c); | |
1446 | ctr += GHASH_CHUNK / 16; | |
e23d850f | 1447 | if (IS_LITTLE_ENDIAN) |
c1b2569d | 1448 | # ifdef BSWAP4 |
0f113f3e | 1449 | ctx->Yi.d[3] = BSWAP4(ctr); |
c1b2569d | 1450 | # else |
0f113f3e | 1451 | PUTU32(ctx->Yi.c + 12, ctr); |
c1b2569d | 1452 | # endif |
0f113f3e MC |
1453 | else |
1454 | ctx->Yi.d[3] = ctr; | |
1455 | out += GHASH_CHUNK; | |
1456 | in += GHASH_CHUNK; | |
1457 | len -= GHASH_CHUNK; | |
1458 | } | |
c1b2569d | 1459 | # endif |
2e635aa8 | 1460 | # endif |
0f113f3e MC |
1461 | if ((i = (len & (size_t)-16))) { |
1462 | size_t j = i / 16; | |
f71c6ace | 1463 | |
2e635aa8 | 1464 | # if defined(GHASH) |
0f113f3e | 1465 | GHASH(ctx, in, i); |
2e635aa8 | 1466 | # else |
0f113f3e MC |
1467 | while (j--) { |
1468 | size_t k; | |
1469 | for (k = 0; k < 16; ++k) | |
1470 | ctx->Xi.c[k] ^= in[k]; | |
f5791af3 | 1471 | GCM_MUL(ctx); |
0f113f3e MC |
1472 | in += 16; |
1473 | } | |
1474 | j = i / 16; | |
1475 | in -= i; | |
2e635aa8 | 1476 | # endif |
0f113f3e MC |
1477 | (*stream) (in, out, j, key, ctx->Yi.c); |
1478 | ctr += (unsigned int)j; | |
e23d850f | 1479 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 1480 | # ifdef BSWAP4 |
0f113f3e | 1481 | ctx->Yi.d[3] = BSWAP4(ctr); |
2e635aa8 | 1482 | # else |
0f113f3e | 1483 | PUTU32(ctx->Yi.c + 12, ctr); |
2e635aa8 | 1484 | # endif |
0f113f3e MC |
1485 | else |
1486 | ctx->Yi.d[3] = ctr; | |
1487 | out += i; | |
1488 | in += i; | |
1489 | len -= i; | |
1490 | } | |
1491 | if (len) { | |
1492 | (*ctx->block) (ctx->Yi.c, ctx->EKi.c, key); | |
1493 | ++ctr; | |
e23d850f | 1494 | if (IS_LITTLE_ENDIAN) |
2e635aa8 | 1495 | # ifdef BSWAP4 |
0f113f3e | 1496 | ctx->Yi.d[3] = BSWAP4(ctr); |
2e635aa8 | 1497 | # else |
0f113f3e | 1498 | PUTU32(ctx->Yi.c + 12, ctr); |
2e635aa8 | 1499 | # endif |
0f113f3e MC |
1500 | else |
1501 | ctx->Yi.d[3] = ctr; | |
1502 | while (len--) { | |
c1b2569d AP |
1503 | # if defined(GHASH) |
1504 | out[n] = (ctx->Xn[mres++] = in[n]) ^ ctx->EKi.c[n]; | |
1505 | # else | |
0f113f3e | 1506 | u8 c = in[n]; |
c1b2569d | 1507 | ctx->Xi.c[mres++] ^= c; |
0f113f3e | 1508 | out[n] = c ^ ctx->EKi.c[n]; |
c1b2569d | 1509 | # endif |
0f113f3e MC |
1510 | ++n; |
1511 | } | |
1512 | } | |
1513 | ||
c1b2569d | 1514 | ctx->mres = mres; |
0f113f3e | 1515 | return 0; |
2e635aa8 | 1516 | #endif |
f71c6ace AP |
1517 | } |
1518 | ||
0f113f3e MC |
1519 | int CRYPTO_gcm128_finish(GCM128_CONTEXT *ctx, const unsigned char *tag, |
1520 | size_t len) | |
e7f5b1cd | 1521 | { |
e23d850f | 1522 | DECLARE_IS_ENDIAN; |
0f113f3e MC |
1523 | u64 alen = ctx->len.u[0] << 3; |
1524 | u64 clen = ctx->len.u[1] << 3; | |
e7f5b1cd | 1525 | |
c1b2569d AP |
1526 | #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT) |
1527 | u128 bitlen; | |
1528 | unsigned int mres = ctx->mres; | |
1529 | ||
1530 | if (mres) { | |
1531 | unsigned blocks = (mres + 15) & -16; | |
1532 | ||
1533 | memset(ctx->Xn + mres, 0, blocks - mres); | |
1534 | mres = blocks; | |
1535 | if (mres == sizeof(ctx->Xn)) { | |
1536 | GHASH(ctx, ctx->Xn, mres); | |
1537 | mres = 0; | |
1538 | } | |
1539 | } else if (ctx->ares) { | |
1540 | GCM_MUL(ctx); | |
1541 | } | |
1542 | #else | |
0f113f3e | 1543 | if (ctx->mres || ctx->ares) |
f5791af3 | 1544 | GCM_MUL(ctx); |
c1b2569d | 1545 | #endif |
e7f5b1cd | 1546 | |
e23d850f | 1547 | if (IS_LITTLE_ENDIAN) { |
e7f5b1cd | 1548 | #ifdef BSWAP8 |
0f113f3e MC |
1549 | alen = BSWAP8(alen); |
1550 | clen = BSWAP8(clen); | |
e7f5b1cd | 1551 | #else |
0f113f3e | 1552 | u8 *p = ctx->len.c; |
e7f5b1cd | 1553 | |
0f113f3e MC |
1554 | ctx->len.u[0] = alen; |
1555 | ctx->len.u[1] = clen; | |
e7f5b1cd | 1556 | |
0f113f3e MC |
1557 | alen = (u64)GETU32(p) << 32 | GETU32(p + 4); |
1558 | clen = (u64)GETU32(p + 8) << 32 | GETU32(p + 12); | |
e7f5b1cd | 1559 | #endif |
0f113f3e | 1560 | } |
e7f5b1cd | 1561 | |
c1b2569d AP |
1562 | #if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT) |
1563 | bitlen.hi = alen; | |
1564 | bitlen.lo = clen; | |
1565 | memcpy(ctx->Xn + mres, &bitlen, sizeof(bitlen)); | |
1566 | mres += sizeof(bitlen); | |
1567 | GHASH(ctx, ctx->Xn, mres); | |
1568 | #else | |
0f113f3e MC |
1569 | ctx->Xi.u[0] ^= alen; |
1570 | ctx->Xi.u[1] ^= clen; | |
f5791af3 | 1571 | GCM_MUL(ctx); |
c1b2569d | 1572 | #endif |
e7f5b1cd | 1573 | |
0f113f3e MC |
1574 | ctx->Xi.u[0] ^= ctx->EK0.u[0]; |
1575 | ctx->Xi.u[1] ^= ctx->EK0.u[1]; | |
6acb4ff3 | 1576 | |
0f113f3e | 1577 | if (tag && len <= sizeof(ctx->Xi)) |
1e4a355d | 1578 | return CRYPTO_memcmp(ctx->Xi.c, tag, len); |
0f113f3e MC |
1579 | else |
1580 | return -1; | |
6acb4ff3 AP |
1581 | } |
1582 | ||
fd3dbc1d DSH |
1583 | void CRYPTO_gcm128_tag(GCM128_CONTEXT *ctx, unsigned char *tag, size_t len) |
1584 | { | |
0f113f3e MC |
1585 | CRYPTO_gcm128_finish(ctx, NULL, 0); |
1586 | memcpy(tag, ctx->Xi.c, | |
1587 | len <= sizeof(ctx->Xi.c) ? len : sizeof(ctx->Xi.c)); | |
fd3dbc1d DSH |
1588 | } |
1589 | ||
6acb4ff3 AP |
1590 | GCM128_CONTEXT *CRYPTO_gcm128_new(void *key, block128_f block) |
1591 | { | |
0f113f3e | 1592 | GCM128_CONTEXT *ret; |
6acb4ff3 | 1593 | |
90945fa3 | 1594 | if ((ret = OPENSSL_malloc(sizeof(*ret))) != NULL) |
0f113f3e | 1595 | CRYPTO_gcm128_init(ret, key, block); |
6acb4ff3 | 1596 | |
0f113f3e | 1597 | return ret; |
6acb4ff3 AP |
1598 | } |
1599 | ||
1600 | void CRYPTO_gcm128_release(GCM128_CONTEXT *ctx) | |
1601 | { | |
4b45c6e5 | 1602 | OPENSSL_clear_free(ctx, sizeof(*ctx)); |
e7f5b1cd | 1603 | } |