Nitin Shukla / LED_Demo

nkjnm

Dependencies:   MAX44000 nexpaq_mdk

Fork of LED_Demo by Maxim nexpaq

mbd_os/features/mbedtls/src/aesni.c@1:55a6170b404f, 2016-09-17 (annotated)

Committer:
nexpaq
Date:
Sat Sep 17 16:32:05 2016 +0000
Revision:
1:55a6170b404f
checking in for sharing

Who changed what in which revision?

UserRevisionLine numberNew contents of line
nexpaq 1:55a6170b404f 1 /*
nexpaq 1:55a6170b404f 2 * AES-NI support functions
nexpaq 1:55a6170b404f 3 *
nexpaq 1:55a6170b404f 4 * Copyright (C) 2006-2015, ARM Limited, All Rights Reserved
nexpaq 1:55a6170b404f 5 * SPDX-License-Identifier: Apache-2.0
nexpaq 1:55a6170b404f 6 *
nexpaq 1:55a6170b404f 7 * Licensed under the Apache License, Version 2.0 (the "License"); you may
nexpaq 1:55a6170b404f 8 * not use this file except in compliance with the License.
nexpaq 1:55a6170b404f 9 * You may obtain a copy of the License at
nexpaq 1:55a6170b404f 10 *
nexpaq 1:55a6170b404f 11 * http://www.apache.org/licenses/LICENSE-2.0
nexpaq 1:55a6170b404f 12 *
nexpaq 1:55a6170b404f 13 * Unless required by applicable law or agreed to in writing, software
nexpaq 1:55a6170b404f 14 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
nexpaq 1:55a6170b404f 15 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
nexpaq 1:55a6170b404f 16 * See the License for the specific language governing permissions and
nexpaq 1:55a6170b404f 17 * limitations under the License.
nexpaq 1:55a6170b404f 18 *
nexpaq 1:55a6170b404f 19 * This file is part of mbed TLS (https://tls.mbed.org)
nexpaq 1:55a6170b404f 20 */
nexpaq 1:55a6170b404f 21
nexpaq 1:55a6170b404f 22 /*
nexpaq 1:55a6170b404f 23 * [AES-WP] http://software.intel.com/en-us/articles/intel-advanced-encryption-standard-aes-instructions-set
nexpaq 1:55a6170b404f 24 * [CLMUL-WP] http://software.intel.com/en-us/articles/intel-carry-less-multiplication-instruction-and-its-usage-for-computing-the-gcm-mode/
nexpaq 1:55a6170b404f 25 */
nexpaq 1:55a6170b404f 26
nexpaq 1:55a6170b404f 27 #if !defined(MBEDTLS_CONFIG_FILE)
nexpaq 1:55a6170b404f 28 #include "mbedtls/config.h"
nexpaq 1:55a6170b404f 29 #else
nexpaq 1:55a6170b404f 30 #include MBEDTLS_CONFIG_FILE
nexpaq 1:55a6170b404f 31 #endif
nexpaq 1:55a6170b404f 32
nexpaq 1:55a6170b404f 33 #if defined(MBEDTLS_AESNI_C)
nexpaq 1:55a6170b404f 34
nexpaq 1:55a6170b404f 35 #include "mbedtls/aesni.h"
nexpaq 1:55a6170b404f 36
nexpaq 1:55a6170b404f 37 #include <string.h>
nexpaq 1:55a6170b404f 38
nexpaq 1:55a6170b404f 39 #ifndef asm
nexpaq 1:55a6170b404f 40 #define asm __asm
nexpaq 1:55a6170b404f 41 #endif
nexpaq 1:55a6170b404f 42
nexpaq 1:55a6170b404f 43 #if defined(MBEDTLS_HAVE_X86_64)
nexpaq 1:55a6170b404f 44
nexpaq 1:55a6170b404f 45 /*
nexpaq 1:55a6170b404f 46 * AES-NI support detection routine
nexpaq 1:55a6170b404f 47 */
nexpaq 1:55a6170b404f 48 int mbedtls_aesni_has_support( unsigned int what )
nexpaq 1:55a6170b404f 49 {
nexpaq 1:55a6170b404f 50 static int done = 0;
nexpaq 1:55a6170b404f 51 static unsigned int c = 0;
nexpaq 1:55a6170b404f 52
nexpaq 1:55a6170b404f 53 if( ! done )
nexpaq 1:55a6170b404f 54 {
nexpaq 1:55a6170b404f 55 asm( "movl $1, %%eax \n\t"
nexpaq 1:55a6170b404f 56 "cpuid \n\t"
nexpaq 1:55a6170b404f 57 : "=c" (c)
nexpaq 1:55a6170b404f 58 :
nexpaq 1:55a6170b404f 59 : "eax", "ebx", "edx" );
nexpaq 1:55a6170b404f 60 done = 1;
nexpaq 1:55a6170b404f 61 }
nexpaq 1:55a6170b404f 62
nexpaq 1:55a6170b404f 63 return( ( c & what ) != 0 );
nexpaq 1:55a6170b404f 64 }
nexpaq 1:55a6170b404f 65
nexpaq 1:55a6170b404f 66 /*
nexpaq 1:55a6170b404f 67 * Binutils needs to be at least 2.19 to support AES-NI instructions.
nexpaq 1:55a6170b404f 68 * Unfortunately, a lot of users have a lower version now (2014-04).
nexpaq 1:55a6170b404f 69 * Emit bytecode directly in order to support "old" version of gas.
nexpaq 1:55a6170b404f 70 *
nexpaq 1:55a6170b404f 71 * Opcodes from the Intel architecture reference manual, vol. 3.
nexpaq 1:55a6170b404f 72 * We always use registers, so we don't need prefixes for memory operands.
nexpaq 1:55a6170b404f 73 * Operand macros are in gas order (src, dst) as opposed to Intel order
nexpaq 1:55a6170b404f 74 * (dst, src) in order to blend better into the surrounding assembly code.
nexpaq 1:55a6170b404f 75 */
nexpaq 1:55a6170b404f 76 #define AESDEC ".byte 0x66,0x0F,0x38,0xDE,"
nexpaq 1:55a6170b404f 77 #define AESDECLAST ".byte 0x66,0x0F,0x38,0xDF,"
nexpaq 1:55a6170b404f 78 #define AESENC ".byte 0x66,0x0F,0x38,0xDC,"
nexpaq 1:55a6170b404f 79 #define AESENCLAST ".byte 0x66,0x0F,0x38,0xDD,"
nexpaq 1:55a6170b404f 80 #define AESIMC ".byte 0x66,0x0F,0x38,0xDB,"
nexpaq 1:55a6170b404f 81 #define AESKEYGENA ".byte 0x66,0x0F,0x3A,0xDF,"
nexpaq 1:55a6170b404f 82 #define PCLMULQDQ ".byte 0x66,0x0F,0x3A,0x44,"
nexpaq 1:55a6170b404f 83
nexpaq 1:55a6170b404f 84 #define xmm0_xmm0 "0xC0"
nexpaq 1:55a6170b404f 85 #define xmm0_xmm1 "0xC8"
nexpaq 1:55a6170b404f 86 #define xmm0_xmm2 "0xD0"
nexpaq 1:55a6170b404f 87 #define xmm0_xmm3 "0xD8"
nexpaq 1:55a6170b404f 88 #define xmm0_xmm4 "0xE0"
nexpaq 1:55a6170b404f 89 #define xmm1_xmm0 "0xC1"
nexpaq 1:55a6170b404f 90 #define xmm1_xmm2 "0xD1"
nexpaq 1:55a6170b404f 91
nexpaq 1:55a6170b404f 92 /*
nexpaq 1:55a6170b404f 93 * AES-NI AES-ECB block en(de)cryption
nexpaq 1:55a6170b404f 94 */
nexpaq 1:55a6170b404f 95 int mbedtls_aesni_crypt_ecb( mbedtls_aes_context *ctx,
nexpaq 1:55a6170b404f 96 int mode,
nexpaq 1:55a6170b404f 97 const unsigned char input[16],
nexpaq 1:55a6170b404f 98 unsigned char output[16] )
nexpaq 1:55a6170b404f 99 {
nexpaq 1:55a6170b404f 100 asm( "movdqu (%3), %%xmm0 \n\t" // load input
nexpaq 1:55a6170b404f 101 "movdqu (%1), %%xmm1 \n\t" // load round key 0
nexpaq 1:55a6170b404f 102 "pxor %%xmm1, %%xmm0 \n\t" // round 0
nexpaq 1:55a6170b404f 103 "add $16, %1 \n\t" // point to next round key
nexpaq 1:55a6170b404f 104 "subl $1, %0 \n\t" // normal rounds = nr - 1
nexpaq 1:55a6170b404f 105 "test %2, %2 \n\t" // mode?
nexpaq 1:55a6170b404f 106 "jz 2f \n\t" // 0 = decrypt
nexpaq 1:55a6170b404f 107
nexpaq 1:55a6170b404f 108 "1: \n\t" // encryption loop
nexpaq 1:55a6170b404f 109 "movdqu (%1), %%xmm1 \n\t" // load round key
nexpaq 1:55a6170b404f 110 AESENC xmm1_xmm0 "\n\t" // do round
nexpaq 1:55a6170b404f 111 "add $16, %1 \n\t" // point to next round key
nexpaq 1:55a6170b404f 112 "subl $1, %0 \n\t" // loop
nexpaq 1:55a6170b404f 113 "jnz 1b \n\t"
nexpaq 1:55a6170b404f 114 "movdqu (%1), %%xmm1 \n\t" // load round key
nexpaq 1:55a6170b404f 115 AESENCLAST xmm1_xmm0 "\n\t" // last round
nexpaq 1:55a6170b404f 116 "jmp 3f \n\t"
nexpaq 1:55a6170b404f 117
nexpaq 1:55a6170b404f 118 "2: \n\t" // decryption loop
nexpaq 1:55a6170b404f 119 "movdqu (%1), %%xmm1 \n\t"
nexpaq 1:55a6170b404f 120 AESDEC xmm1_xmm0 "\n\t" // do round
nexpaq 1:55a6170b404f 121 "add $16, %1 \n\t"
nexpaq 1:55a6170b404f 122 "subl $1, %0 \n\t"
nexpaq 1:55a6170b404f 123 "jnz 2b \n\t"
nexpaq 1:55a6170b404f 124 "movdqu (%1), %%xmm1 \n\t" // load round key
nexpaq 1:55a6170b404f 125 AESDECLAST xmm1_xmm0 "\n\t" // last round
nexpaq 1:55a6170b404f 126
nexpaq 1:55a6170b404f 127 "3: \n\t"
nexpaq 1:55a6170b404f 128 "movdqu %%xmm0, (%4) \n\t" // export output
nexpaq 1:55a6170b404f 129 :
nexpaq 1:55a6170b404f 130 : "r" (ctx->nr), "r" (ctx->rk), "r" (mode), "r" (input), "r" (output)
nexpaq 1:55a6170b404f 131 : "memory", "cc", "xmm0", "xmm1" );
nexpaq 1:55a6170b404f 132
nexpaq 1:55a6170b404f 133
nexpaq 1:55a6170b404f 134 return( 0 );
nexpaq 1:55a6170b404f 135 }
nexpaq 1:55a6170b404f 136
nexpaq 1:55a6170b404f 137 /*
nexpaq 1:55a6170b404f 138 * GCM multiplication: c = a times b in GF(2^128)
nexpaq 1:55a6170b404f 139 * Based on [CLMUL-WP] algorithms 1 (with equation 27) and 5.
nexpaq 1:55a6170b404f 140 */
nexpaq 1:55a6170b404f 141 void mbedtls_aesni_gcm_mult( unsigned char c[16],
nexpaq 1:55a6170b404f 142 const unsigned char a[16],
nexpaq 1:55a6170b404f 143 const unsigned char b[16] )
nexpaq 1:55a6170b404f 144 {
nexpaq 1:55a6170b404f 145 unsigned char aa[16], bb[16], cc[16];
nexpaq 1:55a6170b404f 146 size_t i;
nexpaq 1:55a6170b404f 147
nexpaq 1:55a6170b404f 148 /* The inputs are in big-endian order, so byte-reverse them */
nexpaq 1:55a6170b404f 149 for( i = 0; i < 16; i++ )
nexpaq 1:55a6170b404f 150 {
nexpaq 1:55a6170b404f 151 aa[i] = a[15 - i];
nexpaq 1:55a6170b404f 152 bb[i] = b[15 - i];
nexpaq 1:55a6170b404f 153 }
nexpaq 1:55a6170b404f 154
nexpaq 1:55a6170b404f 155 asm( "movdqu (%0), %%xmm0 \n\t" // a1:a0
nexpaq 1:55a6170b404f 156 "movdqu (%1), %%xmm1 \n\t" // b1:b0
nexpaq 1:55a6170b404f 157
nexpaq 1:55a6170b404f 158 /*
nexpaq 1:55a6170b404f 159 * Caryless multiplication xmm2:xmm1 = xmm0 * xmm1
nexpaq 1:55a6170b404f 160 * using [CLMUL-WP] algorithm 1 (p. 13).
nexpaq 1:55a6170b404f 161 */
nexpaq 1:55a6170b404f 162 "movdqa %%xmm1, %%xmm2 \n\t" // copy of b1:b0
nexpaq 1:55a6170b404f 163 "movdqa %%xmm1, %%xmm3 \n\t" // same
nexpaq 1:55a6170b404f 164 "movdqa %%xmm1, %%xmm4 \n\t" // same
nexpaq 1:55a6170b404f 165 PCLMULQDQ xmm0_xmm1 ",0x00 \n\t" // a0*b0 = c1:c0
nexpaq 1:55a6170b404f 166 PCLMULQDQ xmm0_xmm2 ",0x11 \n\t" // a1*b1 = d1:d0
nexpaq 1:55a6170b404f 167 PCLMULQDQ xmm0_xmm3 ",0x10 \n\t" // a0*b1 = e1:e0
nexpaq 1:55a6170b404f 168 PCLMULQDQ xmm0_xmm4 ",0x01 \n\t" // a1*b0 = f1:f0
nexpaq 1:55a6170b404f 169 "pxor %%xmm3, %%xmm4 \n\t" // e1+f1:e0+f0
nexpaq 1:55a6170b404f 170 "movdqa %%xmm4, %%xmm3 \n\t" // same
nexpaq 1:55a6170b404f 171 "psrldq $8, %%xmm4 \n\t" // 0:e1+f1
nexpaq 1:55a6170b404f 172 "pslldq $8, %%xmm3 \n\t" // e0+f0:0
nexpaq 1:55a6170b404f 173 "pxor %%xmm4, %%xmm2 \n\t" // d1:d0+e1+f1
nexpaq 1:55a6170b404f 174 "pxor %%xmm3, %%xmm1 \n\t" // c1+e0+f1:c0
nexpaq 1:55a6170b404f 175
nexpaq 1:55a6170b404f 176 /*
nexpaq 1:55a6170b404f 177 * Now shift the result one bit to the left,
nexpaq 1:55a6170b404f 178 * taking advantage of [CLMUL-WP] eq 27 (p. 20)
nexpaq 1:55a6170b404f 179 */
nexpaq 1:55a6170b404f 180 "movdqa %%xmm1, %%xmm3 \n\t" // r1:r0
nexpaq 1:55a6170b404f 181 "movdqa %%xmm2, %%xmm4 \n\t" // r3:r2
nexpaq 1:55a6170b404f 182 "psllq $1, %%xmm1 \n\t" // r1<<1:r0<<1
nexpaq 1:55a6170b404f 183 "psllq $1, %%xmm2 \n\t" // r3<<1:r2<<1
nexpaq 1:55a6170b404f 184 "psrlq $63, %%xmm3 \n\t" // r1>>63:r0>>63
nexpaq 1:55a6170b404f 185 "psrlq $63, %%xmm4 \n\t" // r3>>63:r2>>63
nexpaq 1:55a6170b404f 186 "movdqa %%xmm3, %%xmm5 \n\t" // r1>>63:r0>>63
nexpaq 1:55a6170b404f 187 "pslldq $8, %%xmm3 \n\t" // r0>>63:0
nexpaq 1:55a6170b404f 188 "pslldq $8, %%xmm4 \n\t" // r2>>63:0
nexpaq 1:55a6170b404f 189 "psrldq $8, %%xmm5 \n\t" // 0:r1>>63
nexpaq 1:55a6170b404f 190 "por %%xmm3, %%xmm1 \n\t" // r1<<1|r0>>63:r0<<1
nexpaq 1:55a6170b404f 191 "por %%xmm4, %%xmm2 \n\t" // r3<<1|r2>>62:r2<<1
nexpaq 1:55a6170b404f 192 "por %%xmm5, %%xmm2 \n\t" // r3<<1|r2>>62:r2<<1|r1>>63
nexpaq 1:55a6170b404f 193
nexpaq 1:55a6170b404f 194 /*
nexpaq 1:55a6170b404f 195 * Now reduce modulo the GCM polynomial x^128 + x^7 + x^2 + x + 1
nexpaq 1:55a6170b404f 196 * using [CLMUL-WP] algorithm 5 (p. 20).
nexpaq 1:55a6170b404f 197 * Currently xmm2:xmm1 holds x3:x2:x1:x0 (already shifted).
nexpaq 1:55a6170b404f 198 */
nexpaq 1:55a6170b404f 199 /* Step 2 (1) */
nexpaq 1:55a6170b404f 200 "movdqa %%xmm1, %%xmm3 \n\t" // x1:x0
nexpaq 1:55a6170b404f 201 "movdqa %%xmm1, %%xmm4 \n\t" // same
nexpaq 1:55a6170b404f 202 "movdqa %%xmm1, %%xmm5 \n\t" // same
nexpaq 1:55a6170b404f 203 "psllq $63, %%xmm3 \n\t" // x1<<63:x0<<63 = stuff:a
nexpaq 1:55a6170b404f 204 "psllq $62, %%xmm4 \n\t" // x1<<62:x0<<62 = stuff:b
nexpaq 1:55a6170b404f 205 "psllq $57, %%xmm5 \n\t" // x1<<57:x0<<57 = stuff:c
nexpaq 1:55a6170b404f 206
nexpaq 1:55a6170b404f 207 /* Step 2 (2) */
nexpaq 1:55a6170b404f 208 "pxor %%xmm4, %%xmm3 \n\t" // stuff:a+b
nexpaq 1:55a6170b404f 209 "pxor %%xmm5, %%xmm3 \n\t" // stuff:a+b+c
nexpaq 1:55a6170b404f 210 "pslldq $8, %%xmm3 \n\t" // a+b+c:0
nexpaq 1:55a6170b404f 211 "pxor %%xmm3, %%xmm1 \n\t" // x1+a+b+c:x0 = d:x0
nexpaq 1:55a6170b404f 212
nexpaq 1:55a6170b404f 213 /* Steps 3 and 4 */
nexpaq 1:55a6170b404f 214 "movdqa %%xmm1,%%xmm0 \n\t" // d:x0
nexpaq 1:55a6170b404f 215 "movdqa %%xmm1,%%xmm4 \n\t" // same
nexpaq 1:55a6170b404f 216 "movdqa %%xmm1,%%xmm5 \n\t" // same
nexpaq 1:55a6170b404f 217 "psrlq $1, %%xmm0 \n\t" // e1:x0>>1 = e1:e0'
nexpaq 1:55a6170b404f 218 "psrlq $2, %%xmm4 \n\t" // f1:x0>>2 = f1:f0'
nexpaq 1:55a6170b404f 219 "psrlq $7, %%xmm5 \n\t" // g1:x0>>7 = g1:g0'
nexpaq 1:55a6170b404f 220 "pxor %%xmm4, %%xmm0 \n\t" // e1+f1:e0'+f0'
nexpaq 1:55a6170b404f 221 "pxor %%xmm5, %%xmm0 \n\t" // e1+f1+g1:e0'+f0'+g0'
nexpaq 1:55a6170b404f 222 // e0'+f0'+g0' is almost e0+f0+g0, ex\tcept for some missing
nexpaq 1:55a6170b404f 223 // bits carried from d. Now get those\t bits back in.
nexpaq 1:55a6170b404f 224 "movdqa %%xmm1,%%xmm3 \n\t" // d:x0
nexpaq 1:55a6170b404f 225 "movdqa %%xmm1,%%xmm4 \n\t" // same
nexpaq 1:55a6170b404f 226 "movdqa %%xmm1,%%xmm5 \n\t" // same
nexpaq 1:55a6170b404f 227 "psllq $63, %%xmm3 \n\t" // d<<63:stuff
nexpaq 1:55a6170b404f 228 "psllq $62, %%xmm4 \n\t" // d<<62:stuff
nexpaq 1:55a6170b404f 229 "psllq $57, %%xmm5 \n\t" // d<<57:stuff
nexpaq 1:55a6170b404f 230 "pxor %%xmm4, %%xmm3 \n\t" // d<<63+d<<62:stuff
nexpaq 1:55a6170b404f 231 "pxor %%xmm5, %%xmm3 \n\t" // missing bits of d:stuff
nexpaq 1:55a6170b404f 232 "psrldq $8, %%xmm3 \n\t" // 0:missing bits of d
nexpaq 1:55a6170b404f 233 "pxor %%xmm3, %%xmm0 \n\t" // e1+f1+g1:e0+f0+g0
nexpaq 1:55a6170b404f 234 "pxor %%xmm1, %%xmm0 \n\t" // h1:h0
nexpaq 1:55a6170b404f 235 "pxor %%xmm2, %%xmm0 \n\t" // x3+h1:x2+h0
nexpaq 1:55a6170b404f 236
nexpaq 1:55a6170b404f 237 "movdqu %%xmm0, (%2) \n\t" // done
nexpaq 1:55a6170b404f 238 :
nexpaq 1:55a6170b404f 239 : "r" (aa), "r" (bb), "r" (cc)
nexpaq 1:55a6170b404f 240 : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5" );
nexpaq 1:55a6170b404f 241
nexpaq 1:55a6170b404f 242 /* Now byte-reverse the outputs */
nexpaq 1:55a6170b404f 243 for( i = 0; i < 16; i++ )
nexpaq 1:55a6170b404f 244 c[i] = cc[15 - i];
nexpaq 1:55a6170b404f 245
nexpaq 1:55a6170b404f 246 return;
nexpaq 1:55a6170b404f 247 }
nexpaq 1:55a6170b404f 248
nexpaq 1:55a6170b404f 249 /*
nexpaq 1:55a6170b404f 250 * Compute decryption round keys from encryption round keys
nexpaq 1:55a6170b404f 251 */
nexpaq 1:55a6170b404f 252 void mbedtls_aesni_inverse_key( unsigned char *invkey,
nexpaq 1:55a6170b404f 253 const unsigned char *fwdkey, int nr )
nexpaq 1:55a6170b404f 254 {
nexpaq 1:55a6170b404f 255 unsigned char *ik = invkey;
nexpaq 1:55a6170b404f 256 const unsigned char *fk = fwdkey + 16 * nr;
nexpaq 1:55a6170b404f 257
nexpaq 1:55a6170b404f 258 memcpy( ik, fk, 16 );
nexpaq 1:55a6170b404f 259
nexpaq 1:55a6170b404f 260 for( fk -= 16, ik += 16; fk > fwdkey; fk -= 16, ik += 16 )
nexpaq 1:55a6170b404f 261 asm( "movdqu (%0), %%xmm0 \n\t"
nexpaq 1:55a6170b404f 262 AESIMC xmm0_xmm0 "\n\t"
nexpaq 1:55a6170b404f 263 "movdqu %%xmm0, (%1) \n\t"
nexpaq 1:55a6170b404f 264 :
nexpaq 1:55a6170b404f 265 : "r" (fk), "r" (ik)
nexpaq 1:55a6170b404f 266 : "memory", "xmm0" );
nexpaq 1:55a6170b404f 267
nexpaq 1:55a6170b404f 268 memcpy( ik, fk, 16 );
nexpaq 1:55a6170b404f 269 }
nexpaq 1:55a6170b404f 270
nexpaq 1:55a6170b404f 271 /*
nexpaq 1:55a6170b404f 272 * Key expansion, 128-bit case
nexpaq 1:55a6170b404f 273 */
nexpaq 1:55a6170b404f 274 static void aesni_setkey_enc_128( unsigned char *rk,
nexpaq 1:55a6170b404f 275 const unsigned char *key )
nexpaq 1:55a6170b404f 276 {
nexpaq 1:55a6170b404f 277 asm( "movdqu (%1), %%xmm0 \n\t" // copy the original key
nexpaq 1:55a6170b404f 278 "movdqu %%xmm0, (%0) \n\t" // as round key 0
nexpaq 1:55a6170b404f 279 "jmp 2f \n\t" // skip auxiliary routine
nexpaq 1:55a6170b404f 280
nexpaq 1:55a6170b404f 281 /*
nexpaq 1:55a6170b404f 282 * Finish generating the next round key.
nexpaq 1:55a6170b404f 283 *
nexpaq 1:55a6170b404f 284 * On entry xmm0 is r3:r2:r1:r0 and xmm1 is X:stuff:stuff:stuff
nexpaq 1:55a6170b404f 285 * with X = rot( sub( r3 ) ) ^ RCON.
nexpaq 1:55a6170b404f 286 *
nexpaq 1:55a6170b404f 287 * On exit, xmm0 is r7:r6:r5:r4
nexpaq 1:55a6170b404f 288 * with r4 = X + r0, r5 = r4 + r1, r6 = r5 + r2, r7 = r6 + r3
nexpaq 1:55a6170b404f 289 * and those are written to the round key buffer.
nexpaq 1:55a6170b404f 290 */
nexpaq 1:55a6170b404f 291 "1: \n\t"
nexpaq 1:55a6170b404f 292 "pshufd $0xff, %%xmm1, %%xmm1 \n\t" // X:X:X:X
nexpaq 1:55a6170b404f 293 "pxor %%xmm0, %%xmm1 \n\t" // X+r3:X+r2:X+r1:r4
nexpaq 1:55a6170b404f 294 "pslldq $4, %%xmm0 \n\t" // r2:r1:r0:0
nexpaq 1:55a6170b404f 295 "pxor %%xmm0, %%xmm1 \n\t" // X+r3+r2:X+r2+r1:r5:r4
nexpaq 1:55a6170b404f 296 "pslldq $4, %%xmm0 \n\t" // etc
nexpaq 1:55a6170b404f 297 "pxor %%xmm0, %%xmm1 \n\t"
nexpaq 1:55a6170b404f 298 "pslldq $4, %%xmm0 \n\t"
nexpaq 1:55a6170b404f 299 "pxor %%xmm1, %%xmm0 \n\t" // update xmm0 for next time!
nexpaq 1:55a6170b404f 300 "add $16, %0 \n\t" // point to next round key
nexpaq 1:55a6170b404f 301 "movdqu %%xmm0, (%0) \n\t" // write it
nexpaq 1:55a6170b404f 302 "ret \n\t"
nexpaq 1:55a6170b404f 303
nexpaq 1:55a6170b404f 304 /* Main "loop" */
nexpaq 1:55a6170b404f 305 "2: \n\t"
nexpaq 1:55a6170b404f 306 AESKEYGENA xmm0_xmm1 ",0x01 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 307 AESKEYGENA xmm0_xmm1 ",0x02 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 308 AESKEYGENA xmm0_xmm1 ",0x04 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 309 AESKEYGENA xmm0_xmm1 ",0x08 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 310 AESKEYGENA xmm0_xmm1 ",0x10 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 311 AESKEYGENA xmm0_xmm1 ",0x20 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 312 AESKEYGENA xmm0_xmm1 ",0x40 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 313 AESKEYGENA xmm0_xmm1 ",0x80 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 314 AESKEYGENA xmm0_xmm1 ",0x1B \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 315 AESKEYGENA xmm0_xmm1 ",0x36 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 316 :
nexpaq 1:55a6170b404f 317 : "r" (rk), "r" (key)
nexpaq 1:55a6170b404f 318 : "memory", "cc", "0" );
nexpaq 1:55a6170b404f 319 }
nexpaq 1:55a6170b404f 320
nexpaq 1:55a6170b404f 321 /*
nexpaq 1:55a6170b404f 322 * Key expansion, 192-bit case
nexpaq 1:55a6170b404f 323 */
nexpaq 1:55a6170b404f 324 static void aesni_setkey_enc_192( unsigned char *rk,
nexpaq 1:55a6170b404f 325 const unsigned char *key )
nexpaq 1:55a6170b404f 326 {
nexpaq 1:55a6170b404f 327 asm( "movdqu (%1), %%xmm0 \n\t" // copy original round key
nexpaq 1:55a6170b404f 328 "movdqu %%xmm0, (%0) \n\t"
nexpaq 1:55a6170b404f 329 "add $16, %0 \n\t"
nexpaq 1:55a6170b404f 330 "movq 16(%1), %%xmm1 \n\t"
nexpaq 1:55a6170b404f 331 "movq %%xmm1, (%0) \n\t"
nexpaq 1:55a6170b404f 332 "add $8, %0 \n\t"
nexpaq 1:55a6170b404f 333 "jmp 2f \n\t" // skip auxiliary routine
nexpaq 1:55a6170b404f 334
nexpaq 1:55a6170b404f 335 /*
nexpaq 1:55a6170b404f 336 * Finish generating the next 6 quarter-keys.
nexpaq 1:55a6170b404f 337 *
nexpaq 1:55a6170b404f 338 * On entry xmm0 is r3:r2:r1:r0, xmm1 is stuff:stuff:r5:r4
nexpaq 1:55a6170b404f 339 * and xmm2 is stuff:stuff:X:stuff with X = rot( sub( r3 ) ) ^ RCON.
nexpaq 1:55a6170b404f 340 *
nexpaq 1:55a6170b404f 341 * On exit, xmm0 is r9:r8:r7:r6 and xmm1 is stuff:stuff:r11:r10
nexpaq 1:55a6170b404f 342 * and those are written to the round key buffer.
nexpaq 1:55a6170b404f 343 */
nexpaq 1:55a6170b404f 344 "1: \n\t"
nexpaq 1:55a6170b404f 345 "pshufd $0x55, %%xmm2, %%xmm2 \n\t" // X:X:X:X
nexpaq 1:55a6170b404f 346 "pxor %%xmm0, %%xmm2 \n\t" // X+r3:X+r2:X+r1:r4
nexpaq 1:55a6170b404f 347 "pslldq $4, %%xmm0 \n\t" // etc
nexpaq 1:55a6170b404f 348 "pxor %%xmm0, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 349 "pslldq $4, %%xmm0 \n\t"
nexpaq 1:55a6170b404f 350 "pxor %%xmm0, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 351 "pslldq $4, %%xmm0 \n\t"
nexpaq 1:55a6170b404f 352 "pxor %%xmm2, %%xmm0 \n\t" // update xmm0 = r9:r8:r7:r6
nexpaq 1:55a6170b404f 353 "movdqu %%xmm0, (%0) \n\t"
nexpaq 1:55a6170b404f 354 "add $16, %0 \n\t"
nexpaq 1:55a6170b404f 355 "pshufd $0xff, %%xmm0, %%xmm2 \n\t" // r9:r9:r9:r9
nexpaq 1:55a6170b404f 356 "pxor %%xmm1, %%xmm2 \n\t" // stuff:stuff:r9+r5:r10
nexpaq 1:55a6170b404f 357 "pslldq $4, %%xmm1 \n\t" // r2:r1:r0:0
nexpaq 1:55a6170b404f 358 "pxor %%xmm2, %%xmm1 \n\t" // xmm1 = stuff:stuff:r11:r10
nexpaq 1:55a6170b404f 359 "movq %%xmm1, (%0) \n\t"
nexpaq 1:55a6170b404f 360 "add $8, %0 \n\t"
nexpaq 1:55a6170b404f 361 "ret \n\t"
nexpaq 1:55a6170b404f 362
nexpaq 1:55a6170b404f 363 "2: \n\t"
nexpaq 1:55a6170b404f 364 AESKEYGENA xmm1_xmm2 ",0x01 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 365 AESKEYGENA xmm1_xmm2 ",0x02 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 366 AESKEYGENA xmm1_xmm2 ",0x04 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 367 AESKEYGENA xmm1_xmm2 ",0x08 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 368 AESKEYGENA xmm1_xmm2 ",0x10 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 369 AESKEYGENA xmm1_xmm2 ",0x20 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 370 AESKEYGENA xmm1_xmm2 ",0x40 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 371 AESKEYGENA xmm1_xmm2 ",0x80 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 372
nexpaq 1:55a6170b404f 373 :
nexpaq 1:55a6170b404f 374 : "r" (rk), "r" (key)
nexpaq 1:55a6170b404f 375 : "memory", "cc", "0" );
nexpaq 1:55a6170b404f 376 }
nexpaq 1:55a6170b404f 377
nexpaq 1:55a6170b404f 378 /*
nexpaq 1:55a6170b404f 379 * Key expansion, 256-bit case
nexpaq 1:55a6170b404f 380 */
nexpaq 1:55a6170b404f 381 static void aesni_setkey_enc_256( unsigned char *rk,
nexpaq 1:55a6170b404f 382 const unsigned char *key )
nexpaq 1:55a6170b404f 383 {
nexpaq 1:55a6170b404f 384 asm( "movdqu (%1), %%xmm0 \n\t"
nexpaq 1:55a6170b404f 385 "movdqu %%xmm0, (%0) \n\t"
nexpaq 1:55a6170b404f 386 "add $16, %0 \n\t"
nexpaq 1:55a6170b404f 387 "movdqu 16(%1), %%xmm1 \n\t"
nexpaq 1:55a6170b404f 388 "movdqu %%xmm1, (%0) \n\t"
nexpaq 1:55a6170b404f 389 "jmp 2f \n\t" // skip auxiliary routine
nexpaq 1:55a6170b404f 390
nexpaq 1:55a6170b404f 391 /*
nexpaq 1:55a6170b404f 392 * Finish generating the next two round keys.
nexpaq 1:55a6170b404f 393 *
nexpaq 1:55a6170b404f 394 * On entry xmm0 is r3:r2:r1:r0, xmm1 is r7:r6:r5:r4 and
nexpaq 1:55a6170b404f 395 * xmm2 is X:stuff:stuff:stuff with X = rot( sub( r7 )) ^ RCON
nexpaq 1:55a6170b404f 396 *
nexpaq 1:55a6170b404f 397 * On exit, xmm0 is r11:r10:r9:r8 and xmm1 is r15:r14:r13:r12
nexpaq 1:55a6170b404f 398 * and those have been written to the output buffer.
nexpaq 1:55a6170b404f 399 */
nexpaq 1:55a6170b404f 400 "1: \n\t"
nexpaq 1:55a6170b404f 401 "pshufd $0xff, %%xmm2, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 402 "pxor %%xmm0, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 403 "pslldq $4, %%xmm0 \n\t"
nexpaq 1:55a6170b404f 404 "pxor %%xmm0, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 405 "pslldq $4, %%xmm0 \n\t"
nexpaq 1:55a6170b404f 406 "pxor %%xmm0, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 407 "pslldq $4, %%xmm0 \n\t"
nexpaq 1:55a6170b404f 408 "pxor %%xmm2, %%xmm0 \n\t"
nexpaq 1:55a6170b404f 409 "add $16, %0 \n\t"
nexpaq 1:55a6170b404f 410 "movdqu %%xmm0, (%0) \n\t"
nexpaq 1:55a6170b404f 411
nexpaq 1:55a6170b404f 412 /* Set xmm2 to stuff:Y:stuff:stuff with Y = subword( r11 )
nexpaq 1:55a6170b404f 413 * and proceed to generate next round key from there */
nexpaq 1:55a6170b404f 414 AESKEYGENA xmm0_xmm2 ",0x00 \n\t"
nexpaq 1:55a6170b404f 415 "pshufd $0xaa, %%xmm2, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 416 "pxor %%xmm1, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 417 "pslldq $4, %%xmm1 \n\t"
nexpaq 1:55a6170b404f 418 "pxor %%xmm1, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 419 "pslldq $4, %%xmm1 \n\t"
nexpaq 1:55a6170b404f 420 "pxor %%xmm1, %%xmm2 \n\t"
nexpaq 1:55a6170b404f 421 "pslldq $4, %%xmm1 \n\t"
nexpaq 1:55a6170b404f 422 "pxor %%xmm2, %%xmm1 \n\t"
nexpaq 1:55a6170b404f 423 "add $16, %0 \n\t"
nexpaq 1:55a6170b404f 424 "movdqu %%xmm1, (%0) \n\t"
nexpaq 1:55a6170b404f 425 "ret \n\t"
nexpaq 1:55a6170b404f 426
nexpaq 1:55a6170b404f 427 /*
nexpaq 1:55a6170b404f 428 * Main "loop" - Generating one more key than necessary,
nexpaq 1:55a6170b404f 429 * see definition of mbedtls_aes_context.buf
nexpaq 1:55a6170b404f 430 */
nexpaq 1:55a6170b404f 431 "2: \n\t"
nexpaq 1:55a6170b404f 432 AESKEYGENA xmm1_xmm2 ",0x01 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 433 AESKEYGENA xmm1_xmm2 ",0x02 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 434 AESKEYGENA xmm1_xmm2 ",0x04 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 435 AESKEYGENA xmm1_xmm2 ",0x08 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 436 AESKEYGENA xmm1_xmm2 ",0x10 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 437 AESKEYGENA xmm1_xmm2 ",0x20 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 438 AESKEYGENA xmm1_xmm2 ",0x40 \n\tcall 1b \n\t"
nexpaq 1:55a6170b404f 439 :
nexpaq 1:55a6170b404f 440 : "r" (rk), "r" (key)
nexpaq 1:55a6170b404f 441 : "memory", "cc", "0" );
nexpaq 1:55a6170b404f 442 }
nexpaq 1:55a6170b404f 443
nexpaq 1:55a6170b404f 444 /*
nexpaq 1:55a6170b404f 445 * Key expansion, wrapper
nexpaq 1:55a6170b404f 446 */
nexpaq 1:55a6170b404f 447 int mbedtls_aesni_setkey_enc( unsigned char *rk,
nexpaq 1:55a6170b404f 448 const unsigned char *key,
nexpaq 1:55a6170b404f 449 size_t bits )
nexpaq 1:55a6170b404f 450 {
nexpaq 1:55a6170b404f 451 switch( bits )
nexpaq 1:55a6170b404f 452 {
nexpaq 1:55a6170b404f 453 case 128: aesni_setkey_enc_128( rk, key ); break;
nexpaq 1:55a6170b404f 454 case 192: aesni_setkey_enc_192( rk, key ); break;
nexpaq 1:55a6170b404f 455 case 256: aesni_setkey_enc_256( rk, key ); break;
nexpaq 1:55a6170b404f 456 default : return( MBEDTLS_ERR_AES_INVALID_KEY_LENGTH );
nexpaq 1:55a6170b404f 457 }
nexpaq 1:55a6170b404f 458
nexpaq 1:55a6170b404f 459 return( 0 );
nexpaq 1:55a6170b404f 460 }
nexpaq 1:55a6170b404f 461
nexpaq 1:55a6170b404f 462 #endif /* MBEDTLS_HAVE_X86_64 */
nexpaq 1:55a6170b404f 463
nexpaq 1:55a6170b404f 464 #endif /* MBEDTLS_AESNI_C */