gcm128.c 49.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
/* ====================================================================
 * Copyright (c) 2010 The OpenSSL Project.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer. 
 *
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in
 *    the documentation and/or other materials provided with the
 *    distribution.
 *
 * 3. All advertising materials mentioning features or use of this
 *    software must display the following acknowledgment:
 *    "This product includes software developed by the OpenSSL Project
 *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
 *
 * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
 *    endorse or promote products derived from this software without
 *    prior written permission. For written permission, please contact
 *    openssl-core@openssl.org.
 *
 * 5. Products derived from this software may not be called "OpenSSL"
 *    nor may "OpenSSL" appear in their names without prior written
 *    permission of the OpenSSL Project.
 *
 * 6. Redistributions of any form whatsoever must retain the following
 *    acknowledgment:
 *    "This product includes software developed by the OpenSSL Project
 *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
 *
 * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
 * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
 * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
 * OF THE POSSIBILITY OF SUCH DAMAGE.
 * ====================================================================
 */

50 51
#define OPENSSL_FIPSAPI

D
Dr. Stephen Henson 已提交
52
#include <openssl/crypto.h>
A
Andy Polyakov 已提交
53
#include "modes_lcl.h"
54 55 56 57 58 59 60 61 62
#include <string.h>

#ifndef MODES_DEBUG
# ifndef NDEBUG
#  define NDEBUG
# endif
#endif
#include <assert.h>

A
Andy Polyakov 已提交
63 64 65 66 67 68
#if defined(BSWAP4) && defined(STRICT_ALIGNMENT)
/* redefine, because alignment is ensured */
#undef	GETU32
#define	GETU32(p)	BSWAP4(*(const u32 *)(p))
#undef	PUTU32
#define	PUTU32(p,v)	*(u32 *)(p) = BSWAP4(v)
69 70
#endif

A
Andy Polyakov 已提交
71 72 73 74 75 76 77 78 79 80 81 82 83 84
#define	PACK(s)		((size_t)(s)<<(sizeof(size_t)*8-16))
#define REDUCE1BIT(V)	do { \
	if (sizeof(size_t)==8) { \
		u64 T = U64(0xe100000000000000) & (0-(V.lo&1)); \
		V.lo  = (V.hi<<63)|(V.lo>>1); \
		V.hi  = (V.hi>>1 )^T; \
	} \
	else { \
		u32 T = 0xe1000000U & (0-(u32)(V.lo&1)); \
		V.lo  = (V.hi<<63)|(V.lo>>1); \
		V.hi  = (V.hi>>1 )^((u64)T<<32); \
	} \
} while(0)

85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
/*
 * Even though permitted values for TABLE_BITS are 8, 4 and 1, it should
 * never be set to 8. 8 is effectively reserved for testing purposes.
 * TABLE_BITS>1 are lookup-table-driven implementations referred to as
 * "Shoup's" in GCM specification. In other words OpenSSL does not cover
 * whole spectrum of possible table driven implementations. Why? In
 * non-"Shoup's" case memory access pattern is segmented in such manner,
 * that it's trivial to see that cache timing information can reveal
 * fair portion of intermediate hash value. Given that ciphertext is
 * always available to attacker, it's possible for him to attempt to
 * deduce secret parameter H and if successful, tamper with messages
 * [which is nothing but trivial in CTR mode]. In "Shoup's" case it's
 * not as trivial, but there is no reason to believe that it's resistant
 * to cache-timing attack. And the thing about "8-bit" implementation is
 * that it consumes 16 (sixteen) times more memory, 4KB per individual
 * key + 1KB shared. Well, on pros side it should be twice as fast as
 * "4-bit" version. And for gcc-generated x86[_64] code, "8-bit" version
 * was observed to run ~75% faster, closer to 100% for commercial
 * compilers... Yet "4-bit" procedure is preferred, because it's
 * believed to provide better security-performance balance and adequate
 * all-round performance. "All-round" refers to things like:
 *
 * - shorter setup time effectively improves overall timing for
 *   handling short messages;
 * - larger table allocation can become unbearable because of VM
 *   subsystem penalties (for example on Windows large enough free
 *   results in VM working set trimming, meaning that consequent
 *   malloc would immediately incur working set expansion);
 * - larger table has larger cache footprint, which can affect
 *   performance of other code paths (not necessarily even from same
 *   thread in Hyper-Threading world);
 *
 * Value of 1 is not appropriate for performance reasons.
 */
119 120
#if	TABLE_BITS==8

121 122 123 124 125 126 127 128 129 130 131
static void gcm_init_8bit(u128 Htable[256], u64 H[2])
{
	int  i, j;
	u128 V;

	Htable[0].hi = 0;
	Htable[0].lo = 0;
	V.hi = H[0];
	V.lo = H[1];

	for (Htable[128]=V, i=64; i>0; i>>=1) {
A
Andy Polyakov 已提交
132
		REDUCE1BIT(V);
133 134 135 136 137 138 139 140 141 142 143 144
		Htable[i] = V;
	}

	for (i=2; i<256; i<<=1) {
		u128 *Hi = Htable+i, H0 = *Hi;
		for (j=1; j<i; ++j) {
			Hi[j].hi = H0.hi^Htable[j].hi;
			Hi[j].lo = H0.lo^Htable[j].lo;
		}
	}
}

145
static void gcm_gmult_8bit(u64 Xi[2], const u128 Htable[256])
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
{
	u128 Z = { 0, 0};
	const u8 *xi = (const u8 *)Xi+15;
	size_t rem, n = *xi;
	const union { long one; char little; } is_endian = {1};
	static const size_t rem_8bit[256] = {
		PACK(0x0000), PACK(0x01C2), PACK(0x0384), PACK(0x0246),
		PACK(0x0708), PACK(0x06CA), PACK(0x048C), PACK(0x054E),
		PACK(0x0E10), PACK(0x0FD2), PACK(0x0D94), PACK(0x0C56),
		PACK(0x0918), PACK(0x08DA), PACK(0x0A9C), PACK(0x0B5E),
		PACK(0x1C20), PACK(0x1DE2), PACK(0x1FA4), PACK(0x1E66),
		PACK(0x1B28), PACK(0x1AEA), PACK(0x18AC), PACK(0x196E),
		PACK(0x1230), PACK(0x13F2), PACK(0x11B4), PACK(0x1076),
		PACK(0x1538), PACK(0x14FA), PACK(0x16BC), PACK(0x177E),
		PACK(0x3840), PACK(0x3982), PACK(0x3BC4), PACK(0x3A06),
		PACK(0x3F48), PACK(0x3E8A), PACK(0x3CCC), PACK(0x3D0E),
		PACK(0x3650), PACK(0x3792), PACK(0x35D4), PACK(0x3416),
		PACK(0x3158), PACK(0x309A), PACK(0x32DC), PACK(0x331E),
		PACK(0x2460), PACK(0x25A2), PACK(0x27E4), PACK(0x2626),
		PACK(0x2368), PACK(0x22AA), PACK(0x20EC), PACK(0x212E),
		PACK(0x2A70), PACK(0x2BB2), PACK(0x29F4), PACK(0x2836),
		PACK(0x2D78), PACK(0x2CBA), PACK(0x2EFC), PACK(0x2F3E),
		PACK(0x7080), PACK(0x7142), PACK(0x7304), PACK(0x72C6),
		PACK(0x7788), PACK(0x764A), PACK(0x740C), PACK(0x75CE),
		PACK(0x7E90), PACK(0x7F52), PACK(0x7D14), PACK(0x7CD6),
		PACK(0x7998), PACK(0x785A), PACK(0x7A1C), PACK(0x7BDE),
		PACK(0x6CA0), PACK(0x6D62), PACK(0x6F24), PACK(0x6EE6),
		PACK(0x6BA8), PACK(0x6A6A), PACK(0x682C), PACK(0x69EE),
		PACK(0x62B0), PACK(0x6372), PACK(0x6134), PACK(0x60F6),
		PACK(0x65B8), PACK(0x647A), PACK(0x663C), PACK(0x67FE),
		PACK(0x48C0), PACK(0x4902), PACK(0x4B44), PACK(0x4A86),
		PACK(0x4FC8), PACK(0x4E0A), PACK(0x4C4C), PACK(0x4D8E),
		PACK(0x46D0), PACK(0x4712), PACK(0x4554), PACK(0x4496),
		PACK(0x41D8), PACK(0x401A), PACK(0x425C), PACK(0x439E),
		PACK(0x54E0), PACK(0x5522), PACK(0x5764), PACK(0x56A6),
		PACK(0x53E8), PACK(0x522A), PACK(0x506C), PACK(0x51AE),
		PACK(0x5AF0), PACK(0x5B32), PACK(0x5974), PACK(0x58B6),
		PACK(0x5DF8), PACK(0x5C3A), PACK(0x5E7C), PACK(0x5FBE),
		PACK(0xE100), PACK(0xE0C2), PACK(0xE284), PACK(0xE346),
		PACK(0xE608), PACK(0xE7CA), PACK(0xE58C), PACK(0xE44E),
		PACK(0xEF10), PACK(0xEED2), PACK(0xEC94), PACK(0xED56),
		PACK(0xE818), PACK(0xE9DA), PACK(0xEB9C), PACK(0xEA5E),
		PACK(0xFD20), PACK(0xFCE2), PACK(0xFEA4), PACK(0xFF66),
		PACK(0xFA28), PACK(0xFBEA), PACK(0xF9AC), PACK(0xF86E),
		PACK(0xF330), PACK(0xF2F2), PACK(0xF0B4), PACK(0xF176),
		PACK(0xF438), PACK(0xF5FA), PACK(0xF7BC), PACK(0xF67E),
		PACK(0xD940), PACK(0xD882), PACK(0xDAC4), PACK(0xDB06),
		PACK(0xDE48), PACK(0xDF8A), PACK(0xDDCC), PACK(0xDC0E),
		PACK(0xD750), PACK(0xD692), PACK(0xD4D4), PACK(0xD516),
		PACK(0xD058), PACK(0xD19A), PACK(0xD3DC), PACK(0xD21E),
		PACK(0xC560), PACK(0xC4A2), PACK(0xC6E4), PACK(0xC726),
		PACK(0xC268), PACK(0xC3AA), PACK(0xC1EC), PACK(0xC02E),
		PACK(0xCB70), PACK(0xCAB2), PACK(0xC8F4), PACK(0xC936),
		PACK(0xCC78), PACK(0xCDBA), PACK(0xCFFC), PACK(0xCE3E),
		PACK(0x9180), PACK(0x9042), PACK(0x9204), PACK(0x93C6),
		PACK(0x9688), PACK(0x974A), PACK(0x950C), PACK(0x94CE),
		PACK(0x9F90), PACK(0x9E52), PACK(0x9C14), PACK(0x9DD6),
		PACK(0x9898), PACK(0x995A), PACK(0x9B1C), PACK(0x9ADE),
		PACK(0x8DA0), PACK(0x8C62), PACK(0x8E24), PACK(0x8FE6),
		PACK(0x8AA8), PACK(0x8B6A), PACK(0x892C), PACK(0x88EE),
		PACK(0x83B0), PACK(0x8272), PACK(0x8034), PACK(0x81F6),
		PACK(0x84B8), PACK(0x857A), PACK(0x873C), PACK(0x86FE),
		PACK(0xA9C0), PACK(0xA802), PACK(0xAA44), PACK(0xAB86),
		PACK(0xAEC8), PACK(0xAF0A), PACK(0xAD4C), PACK(0xAC8E),
		PACK(0xA7D0), PACK(0xA612), PACK(0xA454), PACK(0xA596),
		PACK(0xA0D8), PACK(0xA11A), PACK(0xA35C), PACK(0xA29E),
		PACK(0xB5E0), PACK(0xB422), PACK(0xB664), PACK(0xB7A6),
		PACK(0xB2E8), PACK(0xB32A), PACK(0xB16C), PACK(0xB0AE),
		PACK(0xBBF0), PACK(0xBA32), PACK(0xB874), PACK(0xB9B6),
		PACK(0xBCF8), PACK(0xBD3A), PACK(0xBF7C), PACK(0xBEBE) };

	while (1) {
		Z.hi ^= Htable[n].hi;
		Z.lo ^= Htable[n].lo;

		if ((u8 *)Xi==xi)	break;

		n = *(--xi);

		rem  = (size_t)Z.lo&0xff;
		Z.lo = (Z.hi<<56)|(Z.lo>>8);
		Z.hi = (Z.hi>>8);
		if (sizeof(size_t)==8)
			Z.hi ^= rem_8bit[rem];
		else
			Z.hi ^= (u64)rem_8bit[rem]<<32;
	}

	if (is_endian.little) {
#ifdef BSWAP8
		Xi[0] = BSWAP8(Z.hi);
		Xi[1] = BSWAP8(Z.lo);
#else
		u8 *p = (u8 *)Xi;
		u32 v;
		v = (u32)(Z.hi>>32);	PUTU32(p,v);
		v = (u32)(Z.hi);	PUTU32(p+4,v);
		v = (u32)(Z.lo>>32);	PUTU32(p+8,v);
		v = (u32)(Z.lo);	PUTU32(p+12,v);
#endif
	}
	else {
		Xi[0] = Z.hi;
		Xi[1] = Z.lo;
	}
}
252
#define GCM_MUL(ctx,Xi)   gcm_gmult_8bit(ctx->Xi.u,ctx->Htable)
253

254
#elif	TABLE_BITS==4
255

256 257 258
static void gcm_init_4bit(u128 Htable[16], u64 H[2])
{
	u128 V;
A
Andy Polyakov 已提交
259 260 261
#if defined(OPENSSL_SMALL_FOOTPRINT)
	int  i;
#endif
262 263 264 265 266 267

	Htable[0].hi = 0;
	Htable[0].lo = 0;
	V.hi = H[0];
	V.lo = H[1];

A
Andy Polyakov 已提交
268
#if defined(OPENSSL_SMALL_FOOTPRINT)
269
	for (Htable[8]=V, i=4; i>0; i>>=1) {
A
Andy Polyakov 已提交
270
		REDUCE1BIT(V);
271 272 273 274
		Htable[i] = V;
	}

	for (i=2; i<16; i<<=1) {
275 276 277 278 279
		u128 *Hi = Htable+i;
		int   j;
		for (V=*Hi, j=1; j<i; ++j) {
			Hi[j].hi = V.hi^Htable[j].hi;
			Hi[j].lo = V.lo^Htable[j].lo;
280 281
		}
	}
282
#else
A
Andy Polyakov 已提交
283
	Htable[8] = V;
A
Andy Polyakov 已提交
284
	REDUCE1BIT(V);
A
Andy Polyakov 已提交
285
	Htable[4] = V;
A
Andy Polyakov 已提交
286
	REDUCE1BIT(V);
A
Andy Polyakov 已提交
287
	Htable[2] = V;
A
Andy Polyakov 已提交
288
	REDUCE1BIT(V);
A
Andy Polyakov 已提交
289
	Htable[1] = V;
290 291 292 293 294 295 296 297 298 299 300 301 302 303
	Htable[3].hi  = V.hi^Htable[2].hi, Htable[3].lo  = V.lo^Htable[2].lo;
	V=Htable[4];
	Htable[5].hi  = V.hi^Htable[1].hi, Htable[5].lo  = V.lo^Htable[1].lo;
	Htable[6].hi  = V.hi^Htable[2].hi, Htable[6].lo  = V.lo^Htable[2].lo;
	Htable[7].hi  = V.hi^Htable[3].hi, Htable[7].lo  = V.lo^Htable[3].lo;
	V=Htable[8];
	Htable[9].hi  = V.hi^Htable[1].hi, Htable[9].lo  = V.lo^Htable[1].lo;
	Htable[10].hi = V.hi^Htable[2].hi, Htable[10].lo = V.lo^Htable[2].lo;
	Htable[11].hi = V.hi^Htable[3].hi, Htable[11].lo = V.lo^Htable[3].lo;
	Htable[12].hi = V.hi^Htable[4].hi, Htable[12].lo = V.lo^Htable[4].lo;
	Htable[13].hi = V.hi^Htable[5].hi, Htable[13].lo = V.lo^Htable[5].lo;
	Htable[14].hi = V.hi^Htable[6].hi, Htable[14].lo = V.lo^Htable[6].lo;
	Htable[15].hi = V.hi^Htable[7].hi, Htable[15].lo = V.lo^Htable[7].lo;
#endif
A
Andy Polyakov 已提交
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
#if defined(GHASH_ASM) && (defined(__arm__) || defined(__arm))
	/*
	 * ARM assembler expects specific dword order in Htable.
	 */
	{
	int j;
	const union { long one; char little; } is_endian = {1};

	if (is_endian.little)
		for (j=0;j<16;++j) {
			V = Htable[j];
			Htable[j].hi = V.lo;
			Htable[j].lo = V.hi;
		}
	else
		for (j=0;j<16;++j) {
			V = Htable[j];
			Htable[j].hi = V.lo<<32|V.lo>>32;
			Htable[j].lo = V.hi<<32|V.hi>>32;
		}
	}
#endif
326 327
}

328
#ifndef GHASH_ASM
329 330 331 332 333 334
static const size_t rem_4bit[16] = {
	PACK(0x0000), PACK(0x1C20), PACK(0x3840), PACK(0x2460),
	PACK(0x7080), PACK(0x6CA0), PACK(0x48C0), PACK(0x54E0),
	PACK(0xE100), PACK(0xFD20), PACK(0xD940), PACK(0xC560),
	PACK(0x9180), PACK(0x8DA0), PACK(0xA9C0), PACK(0xB5E0) };

335
static void gcm_gmult_4bit(u64 Xi[2], const u128 Htable[16])
336
{
337 338 339
	u128 Z;
	int cnt = 15;
	size_t rem, nlo, nhi;
340
	const union { long one; char little; } is_endian = {1};
341 342 343 344 345 346 347

	nlo  = ((const u8 *)Xi)[15];
	nhi  = nlo>>4;
	nlo &= 0xf;

	Z.hi = Htable[nlo].hi;
	Z.lo = Htable[nlo].lo;
348 349

	while (1) {
350 351 352 353 354 355 356 357 358 359 360 361 362 363
		rem  = (size_t)Z.lo&0xf;
		Z.lo = (Z.hi<<60)|(Z.lo>>4);
		Z.hi = (Z.hi>>4);
		if (sizeof(size_t)==8)
			Z.hi ^= rem_4bit[rem];
		else
			Z.hi ^= (u64)rem_4bit[rem]<<32;

		Z.hi ^= Htable[nhi].hi;
		Z.lo ^= Htable[nhi].lo;

		if (--cnt<0)		break;

		nlo  = ((const u8 *)Xi)[cnt];
364 365 366
		nhi  = nlo>>4;
		nlo &= 0xf;

367 368 369 370 371 372 373 374
		rem  = (size_t)Z.lo&0xf;
		Z.lo = (Z.hi<<60)|(Z.lo>>4);
		Z.hi = (Z.hi>>4);
		if (sizeof(size_t)==8)
			Z.hi ^= rem_4bit[rem];
		else
			Z.hi ^= (u64)rem_4bit[rem]<<32;

375 376
		Z.hi ^= Htable[nlo].hi;
		Z.lo ^= Htable[nlo].lo;
377
	}
378

379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
	if (is_endian.little) {
#ifdef BSWAP8
		Xi[0] = BSWAP8(Z.hi);
		Xi[1] = BSWAP8(Z.lo);
#else
		u8 *p = (u8 *)Xi;
		u32 v;
		v = (u32)(Z.hi>>32);	PUTU32(p,v);
		v = (u32)(Z.hi);	PUTU32(p+4,v);
		v = (u32)(Z.lo>>32);	PUTU32(p+8,v);
		v = (u32)(Z.lo);	PUTU32(p+12,v);
#endif
	}
	else {
		Xi[0] = Z.hi;
		Xi[1] = Z.lo;
	}
}

#if !defined(OPENSSL_SMALL_FOOTPRINT)
/*
 * Streamed gcm_mult_4bit, see CRYPTO_gcm128_[en|de]crypt for
401 402 403 404
 * details... Compiler-generated code doesn't seem to give any
 * performance improvement, at least not on x86[_64]. It's here
 * mostly as reference and a placeholder for possible future
 * non-trivial optimization[s]...
405
 */
406 407
static void gcm_ghash_4bit(u64 Xi[2],const u128 Htable[16],
				const u8 *inp,size_t len)
408 409 410 411 412 413
{
    u128 Z;
    int cnt;
    size_t rem, nlo, nhi;
    const union { long one; char little; } is_endian = {1};

414
#if 1
415 416 417 418 419 420 421 422 423 424 425
    do {
	cnt  = 15;
	nlo  = ((const u8 *)Xi)[15];
	nlo ^= inp[15];
	nhi  = nlo>>4;
	nlo &= 0xf;

	Z.hi = Htable[nlo].hi;
	Z.lo = Htable[nlo].lo;

	while (1) {
426 427 428 429 430 431 432 433 434 435 436
		rem  = (size_t)Z.lo&0xf;
		Z.lo = (Z.hi<<60)|(Z.lo>>4);
		Z.hi = (Z.hi>>4);
		if (sizeof(size_t)==8)
			Z.hi ^= rem_4bit[rem];
		else
			Z.hi ^= (u64)rem_4bit[rem]<<32;

		Z.hi ^= Htable[nhi].hi;
		Z.lo ^= Htable[nhi].lo;

437
		if (--cnt<0)		break;
438

439 440 441 442
		nlo  = ((const u8 *)Xi)[cnt];
		nlo ^= inp[cnt];
		nhi  = nlo>>4;
		nlo &= 0xf;
443 444 445 446 447 448 449 450

		rem  = (size_t)Z.lo&0xf;
		Z.lo = (Z.hi<<60)|(Z.lo>>4);
		Z.hi = (Z.hi>>4);
		if (sizeof(size_t)==8)
			Z.hi ^= rem_4bit[rem];
		else
			Z.hi ^= (u64)rem_4bit[rem]<<32;
451 452 453

		Z.hi ^= Htable[nlo].hi;
		Z.lo ^= Htable[nlo].lo;
454
	}
455 456 457 458
#else
    /*
     * Extra 256+16 bytes per-key plus 512 bytes shared tables
     * [should] give ~50% improvement... One could have PACK()-ed
459 460
     * the rem_8bit even here, but the priority is to minimize
     * cache footprint...
461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511
     */ 
    u128 Hshr4[16];	/* Htable shifted right by 4 bits */
    u8   Hshl4[16];	/* Htable shifted left  by 4 bits */
    static const unsigned short rem_8bit[256] = {
	0x0000, 0x01C2, 0x0384, 0x0246, 0x0708, 0x06CA, 0x048C, 0x054E,
	0x0E10, 0x0FD2, 0x0D94, 0x0C56, 0x0918, 0x08DA, 0x0A9C, 0x0B5E,
	0x1C20, 0x1DE2, 0x1FA4, 0x1E66, 0x1B28, 0x1AEA, 0x18AC, 0x196E,
	0x1230, 0x13F2, 0x11B4, 0x1076, 0x1538, 0x14FA, 0x16BC, 0x177E,
	0x3840, 0x3982, 0x3BC4, 0x3A06, 0x3F48, 0x3E8A, 0x3CCC, 0x3D0E,
	0x3650, 0x3792, 0x35D4, 0x3416, 0x3158, 0x309A, 0x32DC, 0x331E,
	0x2460, 0x25A2, 0x27E4, 0x2626, 0x2368, 0x22AA, 0x20EC, 0x212E,
	0x2A70, 0x2BB2, 0x29F4, 0x2836, 0x2D78, 0x2CBA, 0x2EFC, 0x2F3E,
	0x7080, 0x7142, 0x7304, 0x72C6, 0x7788, 0x764A, 0x740C, 0x75CE,
	0x7E90, 0x7F52, 0x7D14, 0x7CD6, 0x7998, 0x785A, 0x7A1C, 0x7BDE,
	0x6CA0, 0x6D62, 0x6F24, 0x6EE6, 0x6BA8, 0x6A6A, 0x682C, 0x69EE,
	0x62B0, 0x6372, 0x6134, 0x60F6, 0x65B8, 0x647A, 0x663C, 0x67FE,
	0x48C0, 0x4902, 0x4B44, 0x4A86, 0x4FC8, 0x4E0A, 0x4C4C, 0x4D8E,
	0x46D0, 0x4712, 0x4554, 0x4496, 0x41D8, 0x401A, 0x425C, 0x439E,
	0x54E0, 0x5522, 0x5764, 0x56A6, 0x53E8, 0x522A, 0x506C, 0x51AE,
	0x5AF0, 0x5B32, 0x5974, 0x58B6, 0x5DF8, 0x5C3A, 0x5E7C, 0x5FBE,
	0xE100, 0xE0C2, 0xE284, 0xE346, 0xE608, 0xE7CA, 0xE58C, 0xE44E,
	0xEF10, 0xEED2, 0xEC94, 0xED56, 0xE818, 0xE9DA, 0xEB9C, 0xEA5E,
	0xFD20, 0xFCE2, 0xFEA4, 0xFF66, 0xFA28, 0xFBEA, 0xF9AC, 0xF86E,
	0xF330, 0xF2F2, 0xF0B4, 0xF176, 0xF438, 0xF5FA, 0xF7BC, 0xF67E,
	0xD940, 0xD882, 0xDAC4, 0xDB06, 0xDE48, 0xDF8A, 0xDDCC, 0xDC0E,
	0xD750, 0xD692, 0xD4D4, 0xD516, 0xD058, 0xD19A, 0xD3DC, 0xD21E,
	0xC560, 0xC4A2, 0xC6E4, 0xC726, 0xC268, 0xC3AA, 0xC1EC, 0xC02E,
	0xCB70, 0xCAB2, 0xC8F4, 0xC936, 0xCC78, 0xCDBA, 0xCFFC, 0xCE3E,
	0x9180, 0x9042, 0x9204, 0x93C6, 0x9688, 0x974A, 0x950C, 0x94CE,
	0x9F90, 0x9E52, 0x9C14, 0x9DD6, 0x9898, 0x995A, 0x9B1C, 0x9ADE,
	0x8DA0, 0x8C62, 0x8E24, 0x8FE6, 0x8AA8, 0x8B6A, 0x892C, 0x88EE,
	0x83B0, 0x8272, 0x8034, 0x81F6, 0x84B8, 0x857A, 0x873C, 0x86FE,
	0xA9C0, 0xA802, 0xAA44, 0xAB86, 0xAEC8, 0xAF0A, 0xAD4C, 0xAC8E,
	0xA7D0, 0xA612, 0xA454, 0xA596, 0xA0D8, 0xA11A, 0xA35C, 0xA29E,
	0xB5E0, 0xB422, 0xB664, 0xB7A6, 0xB2E8, 0xB32A, 0xB16C, 0xB0AE,
	0xBBF0, 0xBA32, 0xB874, 0xB9B6, 0xBCF8, 0xBD3A, 0xBF7C, 0xBEBE };
    /*
     * This pre-processing phase slows down procedure by approximately
     * same time as it makes each loop spin faster. In other words
     * single block performance is approximately same as straightforward
     * "4-bit" implementation, and then it goes only faster...
     */
    for (cnt=0; cnt<16; ++cnt) {
	Z.hi = Htable[cnt].hi;
	Z.lo = Htable[cnt].lo;
	Hshr4[cnt].lo = (Z.hi<<60)|(Z.lo>>4);
	Hshr4[cnt].hi = (Z.hi>>4);
	Hshl4[cnt]    = (u8)(Z.lo<<4);
    }

    do {
512
	for (Z.lo=0, Z.hi=0, cnt=15; cnt; --cnt) {
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547
		nlo  = ((const u8 *)Xi)[cnt];
		nlo ^= inp[cnt];
		nhi  = nlo>>4;
		nlo &= 0xf;

		Z.hi ^= Htable[nlo].hi;
		Z.lo ^= Htable[nlo].lo;

		rem = (size_t)Z.lo&0xff;

		Z.lo = (Z.hi<<56)|(Z.lo>>8);
		Z.hi = (Z.hi>>8);

		Z.hi ^= Hshr4[nhi].hi;
		Z.lo ^= Hshr4[nhi].lo;
		Z.hi ^= (u64)rem_8bit[rem^Hshl4[nhi]]<<48;
	}

	nlo  = ((const u8 *)Xi)[0];
	nlo ^= inp[0];
	nhi  = nlo>>4;
	nlo &= 0xf;

	Z.hi ^= Htable[nlo].hi;
	Z.lo ^= Htable[nlo].lo;

	rem = (size_t)Z.lo&0xf;

	Z.lo = (Z.hi<<60)|(Z.lo>>4);
	Z.hi = (Z.hi>>4);

	Z.hi ^= Htable[nhi].hi;
	Z.lo ^= Htable[nhi].lo;
	Z.hi ^= ((u64)rem_8bit[rem<<4])<<48;
#endif
548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565

	if (is_endian.little) {
#ifdef BSWAP8
		Xi[0] = BSWAP8(Z.hi);
		Xi[1] = BSWAP8(Z.lo);
#else
		u8 *p = (u8 *)Xi;
		u32 v;
		v = (u32)(Z.hi>>32);	PUTU32(p,v);
		v = (u32)(Z.hi);	PUTU32(p+4,v);
		v = (u32)(Z.lo>>32);	PUTU32(p+8,v);
		v = (u32)(Z.lo);	PUTU32(p+12,v);
#endif
	}
	else {
		Xi[0] = Z.hi;
		Xi[1] = Z.lo;
	}
566
    } while (inp+=16, len-=16);
567
}
568 569
#endif
#else
570 571
void gcm_gmult_4bit(u64 Xi[2],const u128 Htable[16]);
void gcm_ghash_4bit(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len);
572 573 574
#endif

#define GCM_MUL(ctx,Xi)   gcm_gmult_4bit(ctx->Xi.u,ctx->Htable)
575
#if defined(GHASH_ASM) || !defined(OPENSSL_SMALL_FOOTPRINT)
A
Andy Polyakov 已提交
576
#define GHASH(ctx,in,len) gcm_ghash_4bit((ctx)->Xi.u,(ctx)->Htable,in,len)
577 578 579
/* GHASH_CHUNK is "stride parameter" missioned to mitigate cache
 * trashing effect. In other words idea is to hash data while it's
 * still in L1 cache after encryption pass... */
580
#define GHASH_CHUNK       (3*1024)
581
#endif
582

583
#else	/* TABLE_BITS */
584

585
static void gcm_gmult_1bit(u64 Xi[2],const u64 H[2])
586 587 588 589 590 591 592
{
	u128 V,Z = { 0,0 };
	long X;
	int  i,j;
	const long *xi = (const long *)Xi;
	const union { long one; char little; } is_endian = {1};

593
	V.hi = H[0];	/* H is in host byte order, no byte swapping */
594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
	V.lo = H[1];

	for (j=0; j<16/sizeof(long); ++j) {
		if (is_endian.little) {
			if (sizeof(long)==8) {
#ifdef BSWAP8
				X = (long)(BSWAP8(xi[j]));
#else
				const u8 *p = (const u8 *)(xi+j);
				X = (long)((u64)GETU32(p)<<32|GETU32(p+4));
#endif
			}
			else {
				const u8 *p = (const u8 *)(xi+j);
				X = (long)GETU32(p);
			}
		}
		else
			X = xi[j];

		for (i=0; i<8*sizeof(long); ++i, X<<=1) {
			u64 M = (u64)(X>>(8*sizeof(long)-1));
			Z.hi ^= V.hi&M;
			Z.lo ^= V.lo&M;

A
Andy Polyakov 已提交
619
			REDUCE1BIT(V);
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640
		}
	}

	if (is_endian.little) {
#ifdef BSWAP8
		Xi[0] = BSWAP8(Z.hi);
		Xi[1] = BSWAP8(Z.lo);
#else
		u8 *p = (u8 *)Xi;
		u32 v;
		v = (u32)(Z.hi>>32);	PUTU32(p,v);
		v = (u32)(Z.hi);	PUTU32(p+4,v);
		v = (u32)(Z.lo>>32);	PUTU32(p+8,v);
		v = (u32)(Z.lo);	PUTU32(p+12,v);
#endif
	}
	else {
		Xi[0] = Z.hi;
		Xi[1] = Z.lo;
	}
}
641
#define GCM_MUL(ctx,Xi)	  gcm_gmult_1bit(ctx->Xi.u,ctx->H.u)
642

643 644
#endif

A
Andy Polyakov 已提交
645 646 647 648
#if	TABLE_BITS==4 && defined(GHASH_ASM) && !defined(I386_ONLY) && \
	(defined(__i386)	|| defined(__i386__)	|| \
	 defined(__x86_64)	|| defined(__x86_64__)	|| \
	 defined(_M_IX86)	|| defined(_M_AMD64)	|| defined(_M_X64))
649
# define GHASH_ASM_X86_OR_64
A
Andy Polyakov 已提交
650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
extern unsigned int OPENSSL_ia32cap_P[2];

void gcm_init_clmul(u128 Htable[16],const u64 Xi[2]);
void gcm_gmult_clmul(u64 Xi[2],const u128 Htable[16]);
void gcm_ghash_clmul(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len);

# if	defined(__i386) || defined(__i386__) || defined(_M_IX86)
#  define GHASH_ASM_X86
void gcm_gmult_4bit_mmx(u64 Xi[2],const u128 Htable[16]);
void gcm_ghash_4bit_mmx(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len);

void gcm_gmult_4bit_x86(u64 Xi[2],const u128 Htable[16]);
void gcm_ghash_4bit_x86(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len);
# endif

665
# define GCM_FUNCREF_4BIT
A
Andy Polyakov 已提交
666 667
#endif

668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
void CRYPTO_gcm128_init(GCM128_CONTEXT *ctx,void *key,block128_f block)
{
	const union { long one; char little; } is_endian = {1};

	memset(ctx,0,sizeof(*ctx));
	ctx->block = block;
	ctx->key   = key;

	(*block)(ctx->H.c,ctx->H.c,key);

	if (is_endian.little) {
		/* H is stored in host byte order */
#ifdef BSWAP8
		ctx->H.u[0] = BSWAP8(ctx->H.u[0]);
		ctx->H.u[1] = BSWAP8(ctx->H.u[1]);
#else
		u8 *p = ctx->H.c;
		u64 hi,lo;
		hi = (u64)GETU32(p)  <<32|GETU32(p+4);
		lo = (u64)GETU32(p+8)<<32|GETU32(p+12);
		ctx->H.u[0] = hi;
		ctx->H.u[1] = lo;
#endif
	}

693 694 695
#if	TABLE_BITS==8
	gcm_init_8bit(ctx->Htable,ctx->H.u);
#elif	TABLE_BITS==4
696
# if	defined(GHASH_ASM_X86_OR_64)
697
#  if	!defined(GHASH_ASM_X86) || defined(OPENSSL_IA32_SSE2)
A
Andy Polyakov 已提交
698 699 700 701 702 703
	if (OPENSSL_ia32cap_P[1]&(1<<1)) {
		gcm_init_clmul(ctx->Htable,ctx->H.u);
		ctx->gmult = gcm_gmult_clmul;
		ctx->ghash = gcm_ghash_clmul;
		return;
	}
704
#  endif
705
	gcm_init_4bit(ctx->Htable,ctx->H.u);
706
#  if	defined(GHASH_ASM_X86)			/* x86 only */
A
Andy Polyakov 已提交
707 708 709 710 711 712 713 714 715 716 717 718 719 720
	if (OPENSSL_ia32cap_P[0]&(1<<23)) {
		ctx->gmult = gcm_gmult_4bit_mmx;
		ctx->ghash = gcm_ghash_4bit_mmx;
	} else {
		ctx->gmult = gcm_gmult_4bit_x86;
		ctx->ghash = gcm_ghash_4bit_x86;
	}
#  else
	ctx->gmult = gcm_gmult_4bit;
	ctx->ghash = gcm_ghash_4bit;
#  endif
# else
	gcm_init_4bit(ctx->Htable,ctx->H.u);
# endif
721
#endif
722 723 724 725 726
}

void CRYPTO_gcm128_setiv(GCM128_CONTEXT *ctx,const unsigned char *iv,size_t len)
{
	const union { long one; char little; } is_endian = {1};
A
Andy Polyakov 已提交
727
	unsigned int ctr;
728 729 730
#ifdef GCM_FUNCREF_4BIT
	void (*gcm_gmult_4bit)(u64 Xi[2],const u128 Htable[16]) = ctx->gmult;
#endif
731 732 733 734 735

	ctx->Yi.u[0]  = 0;
	ctx->Yi.u[1]  = 0;
	ctx->Xi.u[0]  = 0;
	ctx->Xi.u[1]  = 0;
736 737 738 739
	ctx->len.u[0] = 0;	/* AAD length */
	ctx->len.u[1] = 0;	/* message length */
	ctx->ares = 0;
	ctx->mres = 0;
740 741 742 743

	if (len==12) {
		memcpy(ctx->Yi.c,iv,12);
		ctx->Yi.c[15]=1;
A
Andy Polyakov 已提交
744
		ctr=1;
745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780
	}
	else {
		size_t i;
		u64 len0 = len;

		while (len>=16) {
			for (i=0; i<16; ++i) ctx->Yi.c[i] ^= iv[i];
			GCM_MUL(ctx,Yi);
			iv += 16;
			len -= 16;
		}
		if (len) {
			for (i=0; i<len; ++i) ctx->Yi.c[i] ^= iv[i];
			GCM_MUL(ctx,Yi);
		}
		len0 <<= 3;
		if (is_endian.little) {
#ifdef BSWAP8
			ctx->Yi.u[1]  ^= BSWAP8(len0);
#else
			ctx->Yi.c[8]  ^= (u8)(len0>>56);
			ctx->Yi.c[9]  ^= (u8)(len0>>48);
			ctx->Yi.c[10] ^= (u8)(len0>>40);
			ctx->Yi.c[11] ^= (u8)(len0>>32);
			ctx->Yi.c[12] ^= (u8)(len0>>24);
			ctx->Yi.c[13] ^= (u8)(len0>>16);
			ctx->Yi.c[14] ^= (u8)(len0>>8);
			ctx->Yi.c[15] ^= (u8)(len0);
#endif
		}
		else
			ctx->Yi.u[1]  ^= len0;

		GCM_MUL(ctx,Yi);

		if (is_endian.little)
A
Andy Polyakov 已提交
781
			ctr = GETU32(ctx->Yi.c+12);
782
		else
A
Andy Polyakov 已提交
783
			ctr = ctx->Yi.d[3];
784 785 786
	}

	(*ctx->block)(ctx->Yi.c,ctx->EK0.c,ctx->key);
A
Andy Polyakov 已提交
787
	++ctr;
788
	if (is_endian.little)
A
Andy Polyakov 已提交
789
		PUTU32(ctx->Yi.c+12,ctr);
790
	else
A
Andy Polyakov 已提交
791
		ctx->Yi.d[3] = ctr;
792 793
}

794
int CRYPTO_gcm128_aad(GCM128_CONTEXT *ctx,const unsigned char *aad,size_t len)
795 796
{
	size_t i;
797 798
	unsigned int n;
	u64 alen = ctx->len.u[0];
799 800 801 802 803 804 805
#ifdef GCM_FUNCREF_4BIT
	void (*gcm_gmult_4bit)(u64 Xi[2],const u128 Htable[16]) = ctx->gmult;
# ifdef GHASH
	void (*gcm_ghash_4bit)(u64 Xi[2],const u128 Htable[16],
				const u8 *inp,size_t len) = ctx->ghash;
# endif
#endif
806

807 808 809 810 811 812
	if (ctx->len.u[1]) return -2;

	alen += len;
	if (alen>(U64(1)<<61) || (sizeof(len)==8 && alen<len))
		return -1;
	ctx->len.u[0] = alen;
813

814
	n = ctx->ares;
815 816 817 818 819 820 821 822 823
	if (n) {
		while (n && len) {
			ctx->Xi.c[n] ^= *(aad++);
			--len;
			n = (n+1)%16;
		}
		if (n==0) GCM_MUL(ctx,Xi);
		else {
			ctx->ares = n;
824
			return 0;
825 826
		}
	}
827

828 829
#ifdef GHASH
	if ((i = (len&(size_t)-16))) {
A
Andy Polyakov 已提交
830
		GHASH(ctx,aad,i);
831 832 833 834
		aad += i;
		len -= i;
	}
#else
835 836 837 838 839 840
	while (len>=16) {
		for (i=0; i<16; ++i) ctx->Xi.c[i] ^= aad[i];
		GCM_MUL(ctx,Xi);
		aad += 16;
		len -= 16;
	}
841
#endif
842
	if (len) {
843
		n = (unsigned int)len;
844 845
		for (i=0; i<len; ++i) ctx->Xi.c[i] ^= aad[i];
	}
846 847

	ctx->ares = n;
848
	return 0;
849 850
}

851
int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx,
852 853 854 855 856 857
		const unsigned char *in, unsigned char *out,
		size_t len)
{
	const union { long one; char little; } is_endian = {1};
	unsigned int n, ctr;
	size_t i;
858
	u64 mlen = ctx->len.u[1];
859 860 861 862 863 864 865
#ifdef GCM_FUNCREF_4BIT
	void (*gcm_gmult_4bit)(u64 Xi[2],const u128 Htable[16]) = ctx->gmult;
# ifdef GHASH
	void (*gcm_ghash_4bit)(u64 Xi[2],const u128 Htable[16],
				const u8 *inp,size_t len) = ctx->ghash;
# endif
#endif
866 867 868 869 870 871 872 873

#if 0
	n = (unsigned int)mlen%16; /* alternative to ctx->mres */
#endif
	mlen += len;
	if (mlen>((U64(1)<<36)-32) || (sizeof(len)==8 && mlen<len))
		return -1;
	ctx->len.u[1] = mlen;
874

875 876 877 878 879 880
	if (ctx->ares) {
		/* First call to encrypt finalizes GHASH(AAD) */
		GCM_MUL(ctx,Xi);
		ctx->ares = 0;
	}

A
Andy Polyakov 已提交
881 882 883 884
	if (is_endian.little)
		ctr = GETU32(ctx->Yi.c+12);
	else
		ctr = ctx->Yi.d[3];
885

886
	n = ctx->mres;
887 888 889 890 891 892 893 894 895 896
#if !defined(OPENSSL_SMALL_FOOTPRINT)
	if (16%sizeof(size_t) == 0) do {	/* always true actually */
		if (n) {
			while (n && len) {
				ctx->Xi.c[n] ^= *(out++) = *(in++)^ctx->EKi.c[n];
				--len;
				n = (n+1)%16;
			}
			if (n==0) GCM_MUL(ctx,Xi);
			else {
897
				ctx->mres = n;
898
				return 0;
899 900 901 902 903 904
			}
		}
#if defined(STRICT_ALIGNMENT)
		if (((size_t)in|(size_t)out)%sizeof(size_t) != 0)
			break;
#endif
905
#if defined(GHASH) && defined(GHASH_CHUNK)
906 907 908 909 910
		while (len>=GHASH_CHUNK) {
		    size_t j=GHASH_CHUNK;

		    while (j) {
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
911 912 913 914 915
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
916 917 918 919 920 921 922
			for (i=0; i<16; i+=sizeof(size_t))
				*(size_t *)(out+i) =
				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
			out += 16;
			in  += 16;
			j   -= 16;
		    }
A
Andy Polyakov 已提交
923
		    GHASH(ctx,out-GHASH_CHUNK,GHASH_CHUNK);
924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942
		    len -= GHASH_CHUNK;
		}
		if ((i = (len&(size_t)-16))) {
		    size_t j=i;

		    while (len>=16) {
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
			for (i=0; i<16; i+=sizeof(size_t))
				*(size_t *)(out+i) =
				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
			out += 16;
			in  += 16;
			len -= 16;
		    }
A
Andy Polyakov 已提交
943
		    GHASH(ctx,out-j,j);
944 945 946
		}
#else
		while (len>=16) {
947
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
948 949 950 951 952
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
953 954 955 956 957 958 959 960 961
			for (i=0; i<16; i+=sizeof(size_t))
				*(size_t *)(ctx->Xi.c+i) ^=
				*(size_t *)(out+i) =
				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
			GCM_MUL(ctx,Xi);
			out += 16;
			in  += 16;
			len -= 16;
		}
962
#endif
963
		if (len) {
964
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
965 966 967 968 969 970 971 972 973 974 975
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
			while (len--) {
				ctx->Xi.c[n] ^= out[n] = in[n]^ctx->EKi.c[n];
				++n;
			}
		}

976
		ctx->mres = n;
977
		return 0;
978 979 980 981
	} while(0);
#endif
	for (i=0;i<len;++i) {
		if (n==0) {
982
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
983 984 985 986 987 988 989 990 991 992 993 994
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
		}
		ctx->Xi.c[n] ^= out[i] = in[i]^ctx->EKi.c[n];
		n = (n+1)%16;
		if (n==0)
			GCM_MUL(ctx,Xi);
	}

995
	ctx->mres = n;
996
	return 0;
997 998
}

999
int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx,
1000 1001 1002 1003 1004 1005
		const unsigned char *in, unsigned char *out,
		size_t len)
{
	const union { long one; char little; } is_endian = {1};
	unsigned int n, ctr;
	size_t i;
1006
	u64 mlen = ctx->len.u[1];
1007 1008 1009 1010 1011 1012 1013
#ifdef GCM_FUNCREF_4BIT
	void (*gcm_gmult_4bit)(u64 Xi[2],const u128 Htable[16]) = ctx->gmult;
# ifdef GHASH
	void (*gcm_ghash_4bit)(u64 Xi[2],const u128 Htable[16],
				const u8 *inp,size_t len) = ctx->ghash;
# endif
#endif
1014 1015 1016 1017 1018

	mlen += len;
	if (mlen>((U64(1)<<36)-32) || (sizeof(len)==8 && mlen<len))
		return -1;
	ctx->len.u[1] = mlen;
1019

1020 1021 1022 1023 1024 1025
	if (ctx->ares) {
		/* First call to decrypt finalizes GHASH(AAD) */
		GCM_MUL(ctx,Xi);
		ctx->ares = 0;
	}

A
Andy Polyakov 已提交
1026 1027 1028 1029
	if (is_endian.little)
		ctr = GETU32(ctx->Yi.c+12);
	else
		ctr = ctx->Yi.d[3];
1030

1031
	n = ctx->mres;
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
#if !defined(OPENSSL_SMALL_FOOTPRINT)
	if (16%sizeof(size_t) == 0) do {	/* always true actually */
		if (n) {
			while (n && len) {
				u8 c = *(in++);
				*(out++) = c^ctx->EKi.c[n];
				ctx->Xi.c[n] ^= c;
				--len;
				n = (n+1)%16;
			}
			if (n==0) GCM_MUL (ctx,Xi);
			else {
1044
				ctx->mres = n;
1045
				return 0;
1046 1047 1048 1049 1050 1051
			}
		}
#if defined(STRICT_ALIGNMENT)
		if (((size_t)in|(size_t)out)%sizeof(size_t) != 0)
			break;
#endif
1052
#if defined(GHASH) && defined(GHASH_CHUNK)
1053 1054 1055
		while (len>=GHASH_CHUNK) {
		    size_t j=GHASH_CHUNK;

A
Andy Polyakov 已提交
1056
		    GHASH(ctx,in,GHASH_CHUNK);
1057 1058
		    while (j) {
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
1059 1060 1061 1062 1063
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
			for (i=0; i<16; i+=sizeof(size_t))
				*(size_t *)(out+i) =
				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
			out += 16;
			in  += 16;
			j   -= 16;
		    }
		    len -= GHASH_CHUNK;
		}
		if ((i = (len&(size_t)-16))) {
A
Andy Polyakov 已提交
1074
		    GHASH(ctx,in,i);
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
		    while (len>=16) {
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
			for (i=0; i<16; i+=sizeof(size_t))
				*(size_t *)(out+i) =
				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
			out += 16;
			in  += 16;
			len -= 16;
		    }
		}
#else
		while (len>=16) {
1092
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
1093 1094 1095 1096 1097
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
1098 1099 1100 1101 1102
			for (i=0; i<16; i+=sizeof(size_t)) {
				size_t c = *(size_t *)(in+i);
				*(size_t *)(out+i) = c^*(size_t *)(ctx->EKi.c+i);
				*(size_t *)(ctx->Xi.c+i) ^= c;
			}
1103
			GCM_MUL(ctx,Xi);
1104 1105 1106 1107
			out += 16;
			in  += 16;
			len -= 16;
		}
1108
#endif
1109
		if (len) {
1110
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
			while (len--) {
				u8 c = in[n];
				ctx->Xi.c[n] ^= c;
				out[n] = c^ctx->EKi.c[n];
				++n;
			}
		}

1124
		ctx->mres = n;
1125
		return 0;
1126 1127 1128 1129 1130
	} while(0);
#endif
	for (i=0;i<len;++i) {
		u8 c;
		if (n==0) {
1131
			(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
1132 1133 1134 1135 1136 1137 1138
			++ctr;
			if (is_endian.little)
				PUTU32(ctx->Yi.c+12,ctr);
			else
				ctx->Yi.d[3] = ctr;
		}
		c = in[i];
1139
		out[i] = c^ctx->EKi.c[n];
1140 1141 1142 1143 1144 1145
		ctx->Xi.c[n] ^= c;
		n = (n+1)%16;
		if (n==0)
			GCM_MUL(ctx,Xi);
	}

1146
	ctx->mres = n;
1147
	return 0;
1148 1149
}

1150
int CRYPTO_gcm128_encrypt_ctr32(GCM128_CONTEXT *ctx,
1151 1152 1153 1154 1155 1156
		const unsigned char *in, unsigned char *out,
		size_t len, ctr128_f stream)
{
	const union { long one; char little; } is_endian = {1};
	unsigned int n, ctr;
	size_t i;
1157
	u64 mlen = ctx->len.u[1];
1158 1159 1160 1161 1162 1163 1164
#ifdef GCM_FUNCREF_4BIT
	void (*gcm_gmult_4bit)(u64 Xi[2],const u128 Htable[16]) = ctx->gmult;
# ifdef GHASH
	void (*gcm_ghash_4bit)(u64 Xi[2],const u128 Htable[16],
				const u8 *inp,size_t len) = ctx->ghash;
# endif
#endif
1165 1166 1167 1168 1169

	mlen += len;
	if (mlen>((U64(1)<<36)-32) || (sizeof(len)==8 && mlen<len))
		return -1;
	ctx->len.u[1] = mlen;
1170

1171 1172 1173 1174 1175 1176
	if (ctx->ares) {
		/* First call to encrypt finalizes GHASH(AAD) */
		GCM_MUL(ctx,Xi);
		ctx->ares = 0;
	}

1177 1178 1179 1180 1181
	if (is_endian.little)
		ctr = GETU32(ctx->Yi.c+12);
	else
		ctr = ctx->Yi.d[3];

1182
	n = ctx->mres;
1183 1184 1185 1186 1187 1188 1189 1190
	if (n) {
		while (n && len) {
			ctx->Xi.c[n] ^= *(out++) = *(in++)^ctx->EKi.c[n];
			--len;
			n = (n+1)%16;
		}
		if (n==0) GCM_MUL(ctx,Xi);
		else {
1191
			ctx->mres = n;
1192
			return 0;
1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
		}
	}
#if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT)
	while (len>=GHASH_CHUNK) {
		(*stream)(in,out,GHASH_CHUNK/16,ctx->key,ctx->Yi.c);
		ctr += GHASH_CHUNK/16;
		if (is_endian.little)
			PUTU32(ctx->Yi.c+12,ctr);
		else
			ctx->Yi.d[3] = ctr;
		GHASH(ctx,out,GHASH_CHUNK);
		out += GHASH_CHUNK;
		in  += GHASH_CHUNK;
		len -= GHASH_CHUNK;
	}
#endif
	if ((i = (len&(size_t)-16))) {
		size_t j=i/16;

		(*stream)(in,out,j,ctx->key,ctx->Yi.c);
1213
		ctr += (unsigned int)j;
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243
		if (is_endian.little)
			PUTU32(ctx->Yi.c+12,ctr);
		else
			ctx->Yi.d[3] = ctr;
		in  += i;
		len -= i;
#if defined(GHASH)
		GHASH(ctx,out,i);
		out += i;
#else
		while (j--) {
			for (i=0;i<16;++i) ctx->Xi.c[i] ^= out[i];
			GCM_MUL(ctx,Xi);
			out += 16;
		}
#endif
	}
	if (len) {
		(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
		++ctr;
		if (is_endian.little)
			PUTU32(ctx->Yi.c+12,ctr);
		else
			ctx->Yi.d[3] = ctr;
		while (len--) {
			ctx->Xi.c[n] ^= out[n] = in[n]^ctx->EKi.c[n];
			++n;
		}
	}

1244
	ctx->mres = n;
1245
	return 0;
1246 1247
}

1248
int CRYPTO_gcm128_decrypt_ctr32(GCM128_CONTEXT *ctx,
1249 1250 1251 1252 1253 1254
		const unsigned char *in, unsigned char *out,
		size_t len,ctr128_f stream)
{
	const union { long one; char little; } is_endian = {1};
	unsigned int n, ctr;
	size_t i;
1255
	u64 mlen = ctx->len.u[1];
1256 1257 1258 1259 1260 1261 1262
#ifdef GCM_FUNCREF_4BIT
	void (*gcm_gmult_4bit)(u64 Xi[2],const u128 Htable[16]) = ctx->gmult;
# ifdef GHASH
	void (*gcm_ghash_4bit)(u64 Xi[2],const u128 Htable[16],
				const u8 *inp,size_t len) = ctx->ghash;
# endif
#endif
1263 1264 1265 1266 1267

	mlen += len;
	if (mlen>((U64(1)<<36)-32) || (sizeof(len)==8 && mlen<len))
		return -1;
	ctx->len.u[1] = mlen;
1268

1269 1270 1271 1272 1273 1274
	if (ctx->ares) {
		/* First call to decrypt finalizes GHASH(AAD) */
		GCM_MUL(ctx,Xi);
		ctx->ares = 0;
	}

1275 1276 1277 1278 1279
	if (is_endian.little)
		ctr = GETU32(ctx->Yi.c+12);
	else
		ctr = ctx->Yi.d[3];

1280
	n = ctx->mres;
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
	if (n) {
		while (n && len) {
			u8 c = *(in++);
			*(out++) = c^ctx->EKi.c[n];
			ctx->Xi.c[n] ^= c;
			--len;
			n = (n+1)%16;
		}
		if (n==0) GCM_MUL (ctx,Xi);
		else {
1291
			ctx->mres = n;
1292
			return 0;
1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
		}
	}
#if defined(GHASH) && !defined(OPENSSL_SMALL_FOOTPRINT)
	while (len>=GHASH_CHUNK) {
		GHASH(ctx,in,GHASH_CHUNK);
		(*stream)(in,out,GHASH_CHUNK/16,ctx->key,ctx->Yi.c);
		ctr += GHASH_CHUNK/16;
		if (is_endian.little)
			PUTU32(ctx->Yi.c+12,ctr);
		else
			ctx->Yi.d[3] = ctr;
		out += GHASH_CHUNK;
		in  += GHASH_CHUNK;
		len -= GHASH_CHUNK;
	}
#endif
	if ((i = (len&(size_t)-16))) {
		size_t j=i/16;

#if defined(GHASH)
		GHASH(ctx,in,i);
#else
		while (j--) {
			size_t k;
			for (k=0;k<16;++k) ctx->Xi.c[k] ^= in[k];
			GCM_MUL(ctx,Xi);
			in += 16;
		}
		j   = i/16;
		in -= i;
#endif
		(*stream)(in,out,j,ctx->key,ctx->Yi.c);
1325
		ctr += (unsigned int)j;
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
		if (is_endian.little)
			PUTU32(ctx->Yi.c+12,ctr);
		else
			ctx->Yi.d[3] = ctr;
		out += i;
		in  += i;
		len -= i;
	}
	if (len) {
		(*ctx->block)(ctx->Yi.c,ctx->EKi.c,ctx->key);
		++ctr;
		if (is_endian.little)
			PUTU32(ctx->Yi.c+12,ctr);
		else
			ctx->Yi.d[3] = ctr;
		while (len--) {
			u8 c = in[n];
			ctx->Xi.c[n] ^= c;
			out[n] = c^ctx->EKi.c[n];
			++n;
		}
	}

1349
	ctx->mres = n;
1350
	return 0;
1351 1352
}

1353 1354
int CRYPTO_gcm128_finish(GCM128_CONTEXT *ctx,const unsigned char *tag,
			size_t len)
1355 1356 1357 1358
{
	const union { long one; char little; } is_endian = {1};
	u64 alen = ctx->len.u[0]<<3;
	u64 clen = ctx->len.u[1]<<3;
1359 1360 1361
#ifdef GCM_FUNCREF_4BIT
	void (*gcm_gmult_4bit)(u64 Xi[2],const u128 Htable[16]) = ctx->gmult;
#endif
1362

1363
	if (ctx->mres)
1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
		GCM_MUL(ctx,Xi);

	if (is_endian.little) {
#ifdef BSWAP8
		alen = BSWAP8(alen);
		clen = BSWAP8(clen);
#else
		u8 *p = ctx->len.c;

		ctx->len.u[0] = alen;
		ctx->len.u[1] = clen;

		alen = (u64)GETU32(p)  <<32|GETU32(p+4);
		clen = (u64)GETU32(p+8)<<32|GETU32(p+12);
#endif
	}

	ctx->Xi.u[0] ^= alen;
	ctx->Xi.u[1] ^= clen;
	GCM_MUL(ctx,Xi);

	ctx->Xi.u[0] ^= ctx->EK0.u[0];
	ctx->Xi.u[1] ^= ctx->EK0.u[1];
1387 1388 1389 1390 1391 1392 1393

	if (tag && len<=sizeof(ctx->Xi))
		return memcmp(ctx->Xi.c,tag,len);
	else
		return -1;
}

1394 1395 1396
void CRYPTO_gcm128_tag(GCM128_CONTEXT *ctx, unsigned char *tag, size_t len)
{
	CRYPTO_gcm128_finish(ctx, NULL, 0);
1397
	memcpy(tag, ctx->Xi.c, len<=sizeof(ctx->Xi.c)?len:sizeof(ctx->Xi.c));
1398 1399
}

1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415
GCM128_CONTEXT *CRYPTO_gcm128_new(void *key, block128_f block)
{
	GCM128_CONTEXT *ret;

	if ((ret = (GCM128_CONTEXT *)OPENSSL_malloc(sizeof(GCM128_CONTEXT))))
		CRYPTO_gcm128_init(ret,key,block);

	return ret;
}

void CRYPTO_gcm128_release(GCM128_CONTEXT *ctx)
{
	if (ctx) {
		OPENSSL_cleanse(ctx,sizeof(*ctx));
		OPENSSL_free(ctx);
	}
1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
}

#if defined(SELFTEST)
#include <stdio.h>
#include <openssl/aes.h>

/* Test Case 1 */
static const u8	K1[16],
		*P1=NULL,
		*A1=NULL,
		IV1[12],
		*C1=NULL,
		T1[]=  {0x58,0xe2,0xfc,0xce,0xfa,0x7e,0x30,0x61,0x36,0x7f,0x1d,0x57,0xa4,0xe7,0x45,0x5a};
1429

1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449
/* Test Case 2 */
#define K2 K1
#define A2 A1
#define IV2 IV1
static const u8	P2[16],
		C2[]=  {0x03,0x88,0xda,0xce,0x60,0xb6,0xa3,0x92,0xf3,0x28,0xc2,0xb9,0x71,0xb2,0xfe,0x78},
		T2[]=  {0xab,0x6e,0x47,0xd4,0x2c,0xec,0x13,0xbd,0xf5,0x3a,0x67,0xb2,0x12,0x57,0xbd,0xdf};

/* Test Case 3 */
#define A3 A2
static const u8	K3[]=  {0xfe,0xff,0xe9,0x92,0x86,0x65,0x73,0x1c,0x6d,0x6a,0x8f,0x94,0x67,0x30,0x83,0x08},
		P3[]=  {0xd9,0x31,0x32,0x25,0xf8,0x84,0x06,0xe5,0xa5,0x59,0x09,0xc5,0xaf,0xf5,0x26,0x9a,
			0x86,0xa7,0xa9,0x53,0x15,0x34,0xf7,0xda,0x2e,0x4c,0x30,0x3d,0x8a,0x31,0x8a,0x72,
			0x1c,0x3c,0x0c,0x95,0x95,0x68,0x09,0x53,0x2f,0xcf,0x0e,0x24,0x49,0xa6,0xb5,0x25,
			0xb1,0x6a,0xed,0xf5,0xaa,0x0d,0xe6,0x57,0xba,0x63,0x7b,0x39,0x1a,0xaf,0xd2,0x55},
		IV3[]= {0xca,0xfe,0xba,0xbe,0xfa,0xce,0xdb,0xad,0xde,0xca,0xf8,0x88},
		C3[]=  {0x42,0x83,0x1e,0xc2,0x21,0x77,0x74,0x24,0x4b,0x72,0x21,0xb7,0x84,0xd0,0xd4,0x9c,
			0xe3,0xaa,0x21,0x2f,0x2c,0x02,0xa4,0xe0,0x35,0xc1,0x7e,0x23,0x29,0xac,0xa1,0x2e,
			0x21,0xd5,0x14,0xb2,0x54,0x66,0x93,0x1c,0x7d,0x8f,0x6a,0x5a,0xac,0x84,0xaa,0x05,
			0x1b,0xa3,0x0b,0x39,0x6a,0x0a,0xac,0x97,0x3d,0x58,0xe0,0x91,0x47,0x3f,0x59,0x85},
A
Andy Polyakov 已提交
1450
		T3[]=  {0x4d,0x5c,0x2a,0xf3,0x27,0xcd,0x64,0xa6,0x2c,0xf3,0x5a,0xbd,0x2b,0xa6,0xfa,0xb4};
1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469

/* Test Case 4 */
#define K4 K3
#define IV4 IV3
static const u8	P4[]=  {0xd9,0x31,0x32,0x25,0xf8,0x84,0x06,0xe5,0xa5,0x59,0x09,0xc5,0xaf,0xf5,0x26,0x9a,
			0x86,0xa7,0xa9,0x53,0x15,0x34,0xf7,0xda,0x2e,0x4c,0x30,0x3d,0x8a,0x31,0x8a,0x72,
			0x1c,0x3c,0x0c,0x95,0x95,0x68,0x09,0x53,0x2f,0xcf,0x0e,0x24,0x49,0xa6,0xb5,0x25,
			0xb1,0x6a,0xed,0xf5,0xaa,0x0d,0xe6,0x57,0xba,0x63,0x7b,0x39},
		A4[]=  {0xfe,0xed,0xfa,0xce,0xde,0xad,0xbe,0xef,0xfe,0xed,0xfa,0xce,0xde,0xad,0xbe,0xef,
			0xab,0xad,0xda,0xd2},
		C4[]=  {0x42,0x83,0x1e,0xc2,0x21,0x77,0x74,0x24,0x4b,0x72,0x21,0xb7,0x84,0xd0,0xd4,0x9c,
			0xe3,0xaa,0x21,0x2f,0x2c,0x02,0xa4,0xe0,0x35,0xc1,0x7e,0x23,0x29,0xac,0xa1,0x2e,
			0x21,0xd5,0x14,0xb2,0x54,0x66,0x93,0x1c,0x7d,0x8f,0x6a,0x5a,0xac,0x84,0xaa,0x05,
			0x1b,0xa3,0x0b,0x39,0x6a,0x0a,0xac,0x97,0x3d,0x58,0xe0,0x91},
		T4[]=  {0x5b,0xc9,0x4f,0xbc,0x32,0x21,0xa5,0xdb,0x94,0xfa,0xe9,0x5a,0xe7,0x12,0x1a,0x47};

/* Test Case 5 */
#define K5 K4
#define P5 P4
1470 1471
#define A5 A4
static const u8	IV5[]= {0xca,0xfe,0xba,0xbe,0xfa,0xce,0xdb,0xad},
1472 1473 1474 1475 1476
		C5[]=  {0x61,0x35,0x3b,0x4c,0x28,0x06,0x93,0x4a,0x77,0x7f,0xf5,0x1f,0xa2,0x2a,0x47,0x55,
			0x69,0x9b,0x2a,0x71,0x4f,0xcd,0xc6,0xf8,0x37,0x66,0xe5,0xf9,0x7b,0x6c,0x74,0x23,
			0x73,0x80,0x69,0x00,0xe4,0x9f,0x24,0xb2,0x2b,0x09,0x75,0x44,0xd4,0x89,0x6b,0x42,
			0x49,0x89,0xb5,0xe1,0xeb,0xac,0x0f,0x07,0xc2,0x3f,0x45,0x98},
		T5[]=  {0x36,0x12,0xd2,0xe7,0x9e,0x3b,0x07,0x85,0x56,0x1b,0xe1,0x4a,0xac,0xa2,0xfc,0xcb};
1477

1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
/* Test Case 6 */
#define K6 K5
#define P6 P5
#define A6 A5
static const u8	IV6[]= {0x93,0x13,0x22,0x5d,0xf8,0x84,0x06,0xe5,0x55,0x90,0x9c,0x5a,0xff,0x52,0x69,0xaa,
			0x6a,0x7a,0x95,0x38,0x53,0x4f,0x7d,0xa1,0xe4,0xc3,0x03,0xd2,0xa3,0x18,0xa7,0x28,
			0xc3,0xc0,0xc9,0x51,0x56,0x80,0x95,0x39,0xfc,0xf0,0xe2,0x42,0x9a,0x6b,0x52,0x54,
			0x16,0xae,0xdb,0xf5,0xa0,0xde,0x6a,0x57,0xa6,0x37,0xb3,0x9b},
		C6[]=  {0x8c,0xe2,0x49,0x98,0x62,0x56,0x15,0xb6,0x03,0xa0,0x33,0xac,0xa1,0x3f,0xb8,0x94,
			0xbe,0x91,0x12,0xa5,0xc3,0xa2,0x11,0xa8,0xba,0x26,0x2a,0x3c,0xca,0x7e,0x2c,0xa7,
			0x01,0xe4,0xa9,0xa4,0xfb,0xa4,0x3c,0x90,0xcc,0xdc,0xb2,0x81,0xd4,0x8c,0x7c,0x6f,
			0xd6,0x28,0x75,0xd2,0xac,0xa4,0x17,0x03,0x4c,0x34,0xae,0xe5},
		T6[]=  {0x61,0x9c,0xc5,0xae,0xff,0xfe,0x0b,0xfa,0x46,0x2a,0xf4,0x3c,0x16,0x99,0xd0,0x50};

/* Test Case 7 */
static const u8 K7[24],
		*P7=NULL,
		*A7=NULL,
		IV7[12],
		*C7=NULL,
		T7[]=  {0xcd,0x33,0xb2,0x8a,0xc7,0x73,0xf7,0x4b,0xa0,0x0e,0xd1,0xf3,0x12,0x57,0x24,0x35};

/* Test Case 8 */
#define K8 K7
#define IV8 IV7
#define A8 A7
static const u8	P8[16],
		C8[]=  {0x98,0xe7,0x24,0x7c,0x07,0xf0,0xfe,0x41,0x1c,0x26,0x7e,0x43,0x84,0xb0,0xf6,0x00},
		T8[]=  {0x2f,0xf5,0x8d,0x80,0x03,0x39,0x27,0xab,0x8e,0xf4,0xd4,0x58,0x75,0x14,0xf0,0xfb};

/* Test Case 9 */
#define A9 A8
static const u8	K9[]=  {0xfe,0xff,0xe9,0x92,0x86,0x65,0x73,0x1c,0x6d,0x6a,0x8f,0x94,0x67,0x30,0x83,0x08,
			0xfe,0xff,0xe9,0x92,0x86,0x65,0x73,0x1c},
		P9[]=  {0xd9,0x31,0x32,0x25,0xf8,0x84,0x06,0xe5,0xa5,0x59,0x09,0xc5,0xaf,0xf5,0x26,0x9a,
			0x86,0xa7,0xa9,0x53,0x15,0x34,0xf7,0xda,0x2e,0x4c,0x30,0x3d,0x8a,0x31,0x8a,0x72,
			0x1c,0x3c,0x0c,0x95,0x95,0x68,0x09,0x53,0x2f,0xcf,0x0e,0x24,0x49,0xa6,0xb5,0x25,
			0xb1,0x6a,0xed,0xf5,0xaa,0x0d,0xe6,0x57,0xba,0x63,0x7b,0x39,0x1a,0xaf,0xd2,0x55},
		IV9[]= {0xca,0xfe,0xba,0xbe,0xfa,0xce,0xdb,0xad,0xde,0xca,0xf8,0x88},
		C9[]=  {0x39,0x80,0xca,0x0b,0x3c,0x00,0xe8,0x41,0xeb,0x06,0xfa,0xc4,0x87,0x2a,0x27,0x57,
			0x85,0x9e,0x1c,0xea,0xa6,0xef,0xd9,0x84,0x62,0x85,0x93,0xb4,0x0c,0xa1,0xe1,0x9c,
			0x7d,0x77,0x3d,0x00,0xc1,0x44,0xc5,0x25,0xac,0x61,0x9d,0x18,0xc8,0x4a,0x3f,0x47,
			0x18,0xe2,0x44,0x8b,0x2f,0xe3,0x24,0xd9,0xcc,0xda,0x27,0x10,0xac,0xad,0xe2,0x56},
		T9[]=  {0x99,0x24,0xa7,0xc8,0x58,0x73,0x36,0xbf,0xb1,0x18,0x02,0x4d,0xb8,0x67,0x4a,0x14};

/* Test Case 10 */
#define K10 K9
#define IV10 IV9
static const u8	P10[]= {0xd9,0x31,0x32,0x25,0xf8,0x84,0x06,0xe5,0xa5,0x59,0x09,0xc5,0xaf,0xf5,0x26,0x9a,
			0x86,0xa7,0xa9,0x53,0x15,0x34,0xf7,0xda,0x2e,0x4c,0x30,0x3d,0x8a,0x31,0x8a,0x72,
			0x1c,0x3c,0x0c,0x95,0x95,0x68,0x09,0x53,0x2f,0xcf,0x0e,0x24,0x49,0xa6,0xb5,0x25,
			0xb1,0x6a,0xed,0xf5,0xaa,0x0d,0xe6,0x57,0xba,0x63,0x7b,0x39},
		A10[]= {0xfe,0xed,0xfa,0xce,0xde,0xad,0xbe,0xef,0xfe,0xed,0xfa,0xce,0xde,0xad,0xbe,0xef,
			0xab,0xad,0xda,0xd2},
		C10[]= {0x39,0x80,0xca,0x0b,0x3c,0x00,0xe8,0x41,0xeb,0x06,0xfa,0xc4,0x87,0x2a,0x27,0x57,
			0x85,0x9e,0x1c,0xea,0xa6,0xef,0xd9,0x84,0x62,0x85,0x93,0xb4,0x0c,0xa1,0xe1,0x9c,
			0x7d,0x77,0x3d,0x00,0xc1,0x44,0xc5,0x25,0xac,0x61,0x9d,0x18,0xc8,0x4a,0x3f,0x47,
			0x18,0xe2,0x44,0x8b,0x2f,0xe3,0x24,0xd9,0xcc,0xda,0x27,0x10},
		T10[]= {0x25,0x19,0x49,0x8e,0x80,0xf1,0x47,0x8f,0x37,0xba,0x55,0xbd,0x6d,0x27,0x61,0x8c};

/* Test Case 11 */
#define K11 K10
#define P11 P10
#define A11 A10
static const u8	IV11[]={0xca,0xfe,0xba,0xbe,0xfa,0xce,0xdb,0xad},
		C11[]= {0x0f,0x10,0xf5,0x99,0xae,0x14,0xa1,0x54,0xed,0x24,0xb3,0x6e,0x25,0x32,0x4d,0xb8,
			0xc5,0x66,0x63,0x2e,0xf2,0xbb,0xb3,0x4f,0x83,0x47,0x28,0x0f,0xc4,0x50,0x70,0x57,
			0xfd,0xdc,0x29,0xdf,0x9a,0x47,0x1f,0x75,0xc6,0x65,0x41,0xd4,0xd4,0xda,0xd1,0xc9,
			0xe9,0x3a,0x19,0xa5,0x8e,0x8b,0x47,0x3f,0xa0,0xf0,0x62,0xf7},
		T11[]= {0x65,0xdc,0xc5,0x7f,0xcf,0x62,0x3a,0x24,0x09,0x4f,0xcc,0xa4,0x0d,0x35,0x33,0xf8};

/* Test Case 12 */
#define K12 K11
#define P12 P11
#define A12 A11
static const u8	IV12[]={0x93,0x13,0x22,0x5d,0xf8,0x84,0x06,0xe5,0x55,0x90,0x9c,0x5a,0xff,0x52,0x69,0xaa,
			0x6a,0x7a,0x95,0x38,0x53,0x4f,0x7d,0xa1,0xe4,0xc3,0x03,0xd2,0xa3,0x18,0xa7,0x28,
			0xc3,0xc0,0xc9,0x51,0x56,0x80,0x95,0x39,0xfc,0xf0,0xe2,0x42,0x9a,0x6b,0x52,0x54,
			0x16,0xae,0xdb,0xf5,0xa0,0xde,0x6a,0x57,0xa6,0x37,0xb3,0x9b},
		C12[]= {0xd2,0x7e,0x88,0x68,0x1c,0xe3,0x24,0x3c,0x48,0x30,0x16,0x5a,0x8f,0xdc,0xf9,0xff,
			0x1d,0xe9,0xa1,0xd8,0xe6,0xb4,0x47,0xef,0x6e,0xf7,0xb7,0x98,0x28,0x66,0x6e,0x45,
			0x81,0xe7,0x90,0x12,0xaf,0x34,0xdd,0xd9,0xe2,0xf0,0x37,0x58,0x9b,0x29,0x2d,0xb3,
			0xe6,0x7c,0x03,0x67,0x45,0xfa,0x22,0xe7,0xe9,0xb7,0x37,0x3b},
		T12[]= {0xdc,0xf5,0x66,0xff,0x29,0x1c,0x25,0xbb,0xb8,0x56,0x8f,0xc3,0xd3,0x76,0xa6,0xd9};

/* Test Case 13 */
static const u8	K13[32],
		*P13=NULL,
		*A13=NULL,
		IV13[12],
		*C13=NULL,
		T13[]={0x53,0x0f,0x8a,0xfb,0xc7,0x45,0x36,0xb9,0xa9,0x63,0xb4,0xf1,0xc4,0xcb,0x73,0x8b};

/* Test Case 14 */
#define K14 K13
#define A14 A13
static const u8	P14[16],
		IV14[12],
		C14[]= {0xce,0xa7,0x40,0x3d,0x4d,0x60,0x6b,0x6e,0x07,0x4e,0xc5,0xd3,0xba,0xf3,0x9d,0x18},
		T14[]= {0xd0,0xd1,0xc8,0xa7,0x99,0x99,0x6b,0xf0,0x26,0x5b,0x98,0xb5,0xd4,0x8a,0xb9,0x19};

/* Test Case 15 */
#define A15 A14
static const u8	K15[]= {0xfe,0xff,0xe9,0x92,0x86,0x65,0x73,0x1c,0x6d,0x6a,0x8f,0x94,0x67,0x30,0x83,0x08,
			0xfe,0xff,0xe9,0x92,0x86,0x65,0x73,0x1c,0x6d,0x6a,0x8f,0x94,0x67,0x30,0x83,0x08},
		P15[]= {0xd9,0x31,0x32,0x25,0xf8,0x84,0x06,0xe5,0xa5,0x59,0x09,0xc5,0xaf,0xf5,0x26,0x9a,
			0x86,0xa7,0xa9,0x53,0x15,0x34,0xf7,0xda,0x2e,0x4c,0x30,0x3d,0x8a,0x31,0x8a,0x72,
			0x1c,0x3c,0x0c,0x95,0x95,0x68,0x09,0x53,0x2f,0xcf,0x0e,0x24,0x49,0xa6,0xb5,0x25,
			0xb1,0x6a,0xed,0xf5,0xaa,0x0d,0xe6,0x57,0xba,0x63,0x7b,0x39,0x1a,0xaf,0xd2,0x55},
		IV15[]={0xca,0xfe,0xba,0xbe,0xfa,0xce,0xdb,0xad,0xde,0xca,0xf8,0x88},
		C15[]= {0x52,0x2d,0xc1,0xf0,0x99,0x56,0x7d,0x07,0xf4,0x7f,0x37,0xa3,0x2a,0x84,0x42,0x7d,
			0x64,0x3a,0x8c,0xdc,0xbf,0xe5,0xc0,0xc9,0x75,0x98,0xa2,0xbd,0x25,0x55,0xd1,0xaa,
			0x8c,0xb0,0x8e,0x48,0x59,0x0d,0xbb,0x3d,0xa7,0xb0,0x8b,0x10,0x56,0x82,0x88,0x38,
			0xc5,0xf6,0x1e,0x63,0x93,0xba,0x7a,0x0a,0xbc,0xc9,0xf6,0x62,0x89,0x80,0x15,0xad},
		T15[]= {0xb0,0x94,0xda,0xc5,0xd9,0x34,0x71,0xbd,0xec,0x1a,0x50,0x22,0x70,0xe3,0xcc,0x6c};

/* Test Case 16 */
#define K16 K15
#define IV16 IV15
static const u8	P16[]= {0xd9,0x31,0x32,0x25,0xf8,0x84,0x06,0xe5,0xa5,0x59,0x09,0xc5,0xaf,0xf5,0x26,0x9a,
			0x86,0xa7,0xa9,0x53,0x15,0x34,0xf7,0xda,0x2e,0x4c,0x30,0x3d,0x8a,0x31,0x8a,0x72,
			0x1c,0x3c,0x0c,0x95,0x95,0x68,0x09,0x53,0x2f,0xcf,0x0e,0x24,0x49,0xa6,0xb5,0x25,
			0xb1,0x6a,0xed,0xf5,0xaa,0x0d,0xe6,0x57,0xba,0x63,0x7b,0x39},
		A16[]= {0xfe,0xed,0xfa,0xce,0xde,0xad,0xbe,0xef,0xfe,0xed,0xfa,0xce,0xde,0xad,0xbe,0xef,
			0xab,0xad,0xda,0xd2},
		C16[]= {0x52,0x2d,0xc1,0xf0,0x99,0x56,0x7d,0x07,0xf4,0x7f,0x37,0xa3,0x2a,0x84,0x42,0x7d,
			0x64,0x3a,0x8c,0xdc,0xbf,0xe5,0xc0,0xc9,0x75,0x98,0xa2,0xbd,0x25,0x55,0xd1,0xaa,
			0x8c,0xb0,0x8e,0x48,0x59,0x0d,0xbb,0x3d,0xa7,0xb0,0x8b,0x10,0x56,0x82,0x88,0x38,
			0xc5,0xf6,0x1e,0x63,0x93,0xba,0x7a,0x0a,0xbc,0xc9,0xf6,0x62},
		T16[]= {0x76,0xfc,0x6e,0xce,0x0f,0x4e,0x17,0x68,0xcd,0xdf,0x88,0x53,0xbb,0x2d,0x55,0x1b};

/* Test Case 17 */
#define K17 K16
#define P17 P16
#define A17 A16
static const u8	IV17[]={0xca,0xfe,0xba,0xbe,0xfa,0xce,0xdb,0xad},
		C17[]= {0xc3,0x76,0x2d,0xf1,0xca,0x78,0x7d,0x32,0xae,0x47,0xc1,0x3b,0xf1,0x98,0x44,0xcb,
			0xaf,0x1a,0xe1,0x4d,0x0b,0x97,0x6a,0xfa,0xc5,0x2f,0xf7,0xd7,0x9b,0xba,0x9d,0xe0,
			0xfe,0xb5,0x82,0xd3,0x39,0x34,0xa4,0xf0,0x95,0x4c,0xc2,0x36,0x3b,0xc7,0x3f,0x78,
			0x62,0xac,0x43,0x0e,0x64,0xab,0xe4,0x99,0xf4,0x7c,0x9b,0x1f},
		T17[]= {0x3a,0x33,0x7d,0xbf,0x46,0xa7,0x92,0xc4,0x5e,0x45,0x49,0x13,0xfe,0x2e,0xa8,0xf2};

/* Test Case 18 */
#define K18 K17
#define P18 P17
#define A18 A17
static const u8	IV18[]={0x93,0x13,0x22,0x5d,0xf8,0x84,0x06,0xe5,0x55,0x90,0x9c,0x5a,0xff,0x52,0x69,0xaa,
			0x6a,0x7a,0x95,0x38,0x53,0x4f,0x7d,0xa1,0xe4,0xc3,0x03,0xd2,0xa3,0x18,0xa7,0x28,
			0xc3,0xc0,0xc9,0x51,0x56,0x80,0x95,0x39,0xfc,0xf0,0xe2,0x42,0x9a,0x6b,0x52,0x54,
			0x16,0xae,0xdb,0xf5,0xa0,0xde,0x6a,0x57,0xa6,0x37,0xb3,0x9b},
		C18[]= {0x5a,0x8d,0xef,0x2f,0x0c,0x9e,0x53,0xf1,0xf7,0x5d,0x78,0x53,0x65,0x9e,0x2a,0x20,
			0xee,0xb2,0xb2,0x2a,0xaf,0xde,0x64,0x19,0xa0,0x58,0xab,0x4f,0x6f,0x74,0x6b,0xf4,
			0x0f,0xc0,0xc3,0xb7,0x80,0xf2,0x44,0x45,0x2d,0xa3,0xeb,0xf1,0xc5,0xd8,0x2c,0xde,
			0xa2,0x41,0x89,0x97,0x20,0x0e,0xf8,0x2e,0x44,0xae,0x7e,0x3f},
		T18[]= {0xa4,0x4a,0x82,0x66,0xee,0x1c,0x8e,0xb0,0xc8,0xb5,0xd4,0xcf,0x5a,0xe9,0xf1,0x9a};

#define TEST_CASE(n)	do {					\
	u8 out[sizeof(P##n)];					\
	AES_set_encrypt_key(K##n,sizeof(K##n)*8,&key);		\
	CRYPTO_gcm128_init(&ctx,&key,(block128_f)AES_encrypt);	\
	CRYPTO_gcm128_setiv(&ctx,IV##n,sizeof(IV##n));		\
1639
	memset(out,0,sizeof(out));				\
1640 1641
	if (A##n) CRYPTO_gcm128_aad(&ctx,A##n,sizeof(A##n));	\
	if (P##n) CRYPTO_gcm128_encrypt(&ctx,P##n,out,sizeof(out));	\
1642 1643
	if (CRYPTO_gcm128_finish(&ctx,T##n,16) ||		\
	    (C##n && memcmp(out,C##n,sizeof(out))))		\
1644
		ret++, printf ("encrypt test#%d failed.\n",n);	\
1645
	CRYPTO_gcm128_setiv(&ctx,IV##n,sizeof(IV##n));		\
1646
	memset(out,0,sizeof(out));				\
1647 1648
	if (A##n) CRYPTO_gcm128_aad(&ctx,A##n,sizeof(A##n));	\
	if (C##n) CRYPTO_gcm128_decrypt(&ctx,C##n,out,sizeof(out));	\
1649 1650 1651
	if (CRYPTO_gcm128_finish(&ctx,T##n,16) ||		\
	    (P##n && memcmp(out,P##n,sizeof(out))))		\
		ret++, printf ("decrypt test#%d failed.\n",n);	\
1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
	} while(0)

int main()
{
	GCM128_CONTEXT ctx;
	AES_KEY key;
	int ret=0;

	TEST_CASE(1);
	TEST_CASE(2);
	TEST_CASE(3);
	TEST_CASE(4);
	TEST_CASE(5);
	TEST_CASE(6);
	TEST_CASE(7);
	TEST_CASE(8);
	TEST_CASE(9);
	TEST_CASE(10);
	TEST_CASE(11);
	TEST_CASE(12);
	TEST_CASE(13);
	TEST_CASE(14);
	TEST_CASE(15);
	TEST_CASE(16);
	TEST_CASE(17);
	TEST_CASE(18);

1679
#ifdef OPENSSL_CPUID_OBJ
1680 1681 1682
	{
	size_t start,stop,gcm_t,ctr_t,OPENSSL_rdtsc();
	union { u64 u; u8 c[1024]; } buf;
A
Andy Polyakov 已提交
1683
	int i;
1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694

	AES_set_encrypt_key(K1,sizeof(K1)*8,&key);
	CRYPTO_gcm128_init(&ctx,&key,(block128_f)AES_encrypt);
	CRYPTO_gcm128_setiv(&ctx,IV1,sizeof(IV1));

	CRYPTO_gcm128_encrypt(&ctx,buf.c,buf.c,sizeof(buf));
	start = OPENSSL_rdtsc();
	CRYPTO_gcm128_encrypt(&ctx,buf.c,buf.c,sizeof(buf));
	gcm_t = OPENSSL_rdtsc() - start;

	CRYPTO_ctr128_encrypt(buf.c,buf.c,sizeof(buf),
1695
			&key,ctx.Yi.c,ctx.EKi.c,&ctx.mres,
1696 1697 1698
			(block128_f)AES_encrypt);
	start = OPENSSL_rdtsc();
	CRYPTO_ctr128_encrypt(buf.c,buf.c,sizeof(buf),
1699
			&key,ctx.Yi.c,ctx.EKi.c,&ctx.mres,
1700
			(block128_f)AES_encrypt);
1701 1702 1703 1704 1705 1706
	ctr_t = OPENSSL_rdtsc() - start;

	printf("%.2f-%.2f=%.2f\n",
			gcm_t/(double)sizeof(buf),
			ctr_t/(double)sizeof(buf),
			(gcm_t-ctr_t)/(double)sizeof(buf));
1707
#ifdef GHASH
A
Andy Polyakov 已提交
1708
	GHASH(&ctx,buf.c,sizeof(buf));
1709
	start = OPENSSL_rdtsc();
A
Andy Polyakov 已提交
1710
	for (i=0;i<100;++i) GHASH(&ctx,buf.c,sizeof(buf));
1711
	gcm_t = OPENSSL_rdtsc() - start;
A
Andy Polyakov 已提交
1712
	printf("%.2f\n",gcm_t/(double)sizeof(buf)/(double)i);
1713
#endif
1714
	}
1715
#endif
1716

1717 1718 1719
	return ret;
}
#endif