_ Git - BearSSL/blob - src/symcipher/aes_x86ni_ctr.c

   1 /*
   2  * Copyright (c) 2017 Thomas Pornin <pornin@bolet.org>
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining
   5  * a copy of this software and associated documentation files (the
   6  * "Software"), to deal in the Software without restriction, including
   7  * without limitation the rights to use, copy, modify, merge, publish,
   8  * distribute, sublicense, and/or sell copies of the Software, and to
   9  * permit persons to whom the Software is furnished to do so, subject to
  10  * the following conditions:
  11  *
  12  * The above copyright notice and this permission notice shall be
  13  * included in all copies or substantial portions of the Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  16  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  18  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  19  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  20  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  21  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  22  * SOFTWARE.
  23  */
  24
  25 #include "inner.h"
  26
  27 #if BR_AES_X86NI
  28
  29 #if BR_AES_X86NI_GCC
  30 #include <smmintrin.h>
  31 #include <wmmintrin.h>
  32 #define bswap32   __builtin_bswap32
  33 #endif
  34
  35 #if BR_AES_X86NI_MSC
  36 #include <stdlib.h>
  37 #include <intrin.h>
  38 #define bswap32   _byteswap_ulong
  39 #endif
  40
  41 /* see bearssl_block.h */
  42 void
  43 br_aes_x86ni_ctr_init(br_aes_x86ni_ctr_keys *ctx,
  44         const void *key, size_t len)
  45 {
  46         ctx->vtable = &br_aes_x86ni_ctr_vtable;
  47         ctx->num_rounds = br_aes_x86ni_keysched_enc(ctx->skey.skni, key, len);
  48 }
  49
  50 /* see bearssl_block.h */
  51 BR_TARGET("sse2,sse4.1,aes")
  52 uint32_t
  53 br_aes_x86ni_ctr_run(const br_aes_x86ni_ctr_keys *ctx,
  54         const void *iv, uint32_t cc, void *data, size_t len)
  55 {
  56         unsigned char *buf;
  57         unsigned char ivbuf[16];
  58         unsigned num_rounds;
  59         __m128i sk[15];
  60         __m128i ivx;
  61         unsigned u;
  62
  63         buf = data;
  64         memcpy(ivbuf, iv, 12);
  65         num_rounds = ctx->num_rounds;
  66         for (u = 0; u <= num_rounds; u ++) {
  67                 sk[u] = _mm_loadu_si128((void *)(ctx->skey.skni + (u << 4)));
  68         }
  69         ivx = _mm_loadu_si128((void *)ivbuf);
  70         while (len > 0) {
  71                 __m128i x0, x1, x2, x3;
  72
  73                 x0 = _mm_insert_epi32(ivx, bswap32(cc + 0), 3);
  74                 x1 = _mm_insert_epi32(ivx, bswap32(cc + 1), 3);
  75                 x2 = _mm_insert_epi32(ivx, bswap32(cc + 2), 3);
  76                 x3 = _mm_insert_epi32(ivx, bswap32(cc + 3), 3);
  77                 x0 = _mm_xor_si128(x0, sk[0]);
  78                 x1 = _mm_xor_si128(x1, sk[0]);
  79                 x2 = _mm_xor_si128(x2, sk[0]);
  80                 x3 = _mm_xor_si128(x3, sk[0]);
  81                 x0 = _mm_aesenc_si128(x0, sk[1]);
  82                 x1 = _mm_aesenc_si128(x1, sk[1]);
  83                 x2 = _mm_aesenc_si128(x2, sk[1]);
  84                 x3 = _mm_aesenc_si128(x3, sk[1]);
  85                 x0 = _mm_aesenc_si128(x0, sk[2]);
  86                 x1 = _mm_aesenc_si128(x1, sk[2]);
  87                 x2 = _mm_aesenc_si128(x2, sk[2]);
  88                 x3 = _mm_aesenc_si128(x3, sk[2]);
  89                 x0 = _mm_aesenc_si128(x0, sk[3]);
  90                 x1 = _mm_aesenc_si128(x1, sk[3]);
  91                 x2 = _mm_aesenc_si128(x2, sk[3]);
  92                 x3 = _mm_aesenc_si128(x3, sk[3]);
  93                 x0 = _mm_aesenc_si128(x0, sk[4]);
  94                 x1 = _mm_aesenc_si128(x1, sk[4]);
  95                 x2 = _mm_aesenc_si128(x2, sk[4]);
  96                 x3 = _mm_aesenc_si128(x3, sk[4]);
  97                 x0 = _mm_aesenc_si128(x0, sk[5]);
  98                 x1 = _mm_aesenc_si128(x1, sk[5]);
  99                 x2 = _mm_aesenc_si128(x2, sk[5]);
 100                 x3 = _mm_aesenc_si128(x3, sk[5]);
 101                 x0 = _mm_aesenc_si128(x0, sk[6]);
 102                 x1 = _mm_aesenc_si128(x1, sk[6]);
 103                 x2 = _mm_aesenc_si128(x2, sk[6]);
 104                 x3 = _mm_aesenc_si128(x3, sk[6]);
 105                 x0 = _mm_aesenc_si128(x0, sk[7]);
 106                 x1 = _mm_aesenc_si128(x1, sk[7]);
 107                 x2 = _mm_aesenc_si128(x2, sk[7]);
 108                 x3 = _mm_aesenc_si128(x3, sk[7]);
 109                 x0 = _mm_aesenc_si128(x0, sk[8]);
 110                 x1 = _mm_aesenc_si128(x1, sk[8]);
 111                 x2 = _mm_aesenc_si128(x2, sk[8]);
 112                 x3 = _mm_aesenc_si128(x3, sk[8]);
 113                 x0 = _mm_aesenc_si128(x0, sk[9]);
 114                 x1 = _mm_aesenc_si128(x1, sk[9]);
 115                 x2 = _mm_aesenc_si128(x2, sk[9]);
 116                 x3 = _mm_aesenc_si128(x3, sk[9]);
 117                 if (num_rounds == 10) {
 118                         x0 = _mm_aesenclast_si128(x0, sk[10]);
 119                         x1 = _mm_aesenclast_si128(x1, sk[10]);
 120                         x2 = _mm_aesenclast_si128(x2, sk[10]);
 121                         x3 = _mm_aesenclast_si128(x3, sk[10]);
 122                 } else if (num_rounds == 12) {
 123                         x0 = _mm_aesenc_si128(x0, sk[10]);
 124                         x1 = _mm_aesenc_si128(x1, sk[10]);
 125                         x2 = _mm_aesenc_si128(x2, sk[10]);
 126                         x3 = _mm_aesenc_si128(x3, sk[10]);
 127                         x0 = _mm_aesenc_si128(x0, sk[11]);
 128                         x1 = _mm_aesenc_si128(x1, sk[11]);
 129                         x2 = _mm_aesenc_si128(x2, sk[11]);
 130                         x3 = _mm_aesenc_si128(x3, sk[11]);
 131                         x0 = _mm_aesenclast_si128(x0, sk[12]);
 132                         x1 = _mm_aesenclast_si128(x1, sk[12]);
 133                         x2 = _mm_aesenclast_si128(x2, sk[12]);
 134                         x3 = _mm_aesenclast_si128(x3, sk[12]);
 135                 } else {
 136                         x0 = _mm_aesenc_si128(x0, sk[10]);
 137                         x1 = _mm_aesenc_si128(x1, sk[10]);
 138                         x2 = _mm_aesenc_si128(x2, sk[10]);
 139                         x3 = _mm_aesenc_si128(x3, sk[10]);
 140                         x0 = _mm_aesenc_si128(x0, sk[11]);
 141                         x1 = _mm_aesenc_si128(x1, sk[11]);
 142                         x2 = _mm_aesenc_si128(x2, sk[11]);
 143                         x3 = _mm_aesenc_si128(x3, sk[11]);
 144                         x0 = _mm_aesenc_si128(x0, sk[12]);
 145                         x1 = _mm_aesenc_si128(x1, sk[12]);
 146                         x2 = _mm_aesenc_si128(x2, sk[12]);
 147                         x3 = _mm_aesenc_si128(x3, sk[12]);
 148                         x0 = _mm_aesenc_si128(x0, sk[13]);
 149                         x1 = _mm_aesenc_si128(x1, sk[13]);
 150                         x2 = _mm_aesenc_si128(x2, sk[13]);
 151                         x3 = _mm_aesenc_si128(x3, sk[13]);
 152                         x0 = _mm_aesenclast_si128(x0, sk[14]);
 153                         x1 = _mm_aesenclast_si128(x1, sk[14]);
 154                         x2 = _mm_aesenclast_si128(x2, sk[14]);
 155                         x3 = _mm_aesenclast_si128(x3, sk[14]);
 156                 }
 157                 if (len >= 64) {
 158                         x0 = _mm_xor_si128(x0,
 159                                 _mm_loadu_si128((void *)(buf +  0)));
 160                         x1 = _mm_xor_si128(x1,
 161                                 _mm_loadu_si128((void *)(buf + 16)));
 162                         x2 = _mm_xor_si128(x2,
 163                                 _mm_loadu_si128((void *)(buf + 32)));
 164                         x3 = _mm_xor_si128(x3,
 165                                 _mm_loadu_si128((void *)(buf + 48)));
 166                         _mm_storeu_si128((void *)(buf +  0), x0);
 167                         _mm_storeu_si128((void *)(buf + 16), x1);
 168                         _mm_storeu_si128((void *)(buf + 32), x2);
 169                         _mm_storeu_si128((void *)(buf + 48), x3);
 170                         buf += 64;
 171                         len -= 64;
 172                         cc += 4;
 173                 } else {
 174                         unsigned char tmp[64];
 175
 176                         _mm_storeu_si128((void *)(tmp +  0), x0);
 177                         _mm_storeu_si128((void *)(tmp + 16), x1);
 178                         _mm_storeu_si128((void *)(tmp + 32), x2);
 179                         _mm_storeu_si128((void *)(tmp + 48), x3);
 180                         for (u = 0; u < len; u ++) {
 181                                 buf[u] ^= tmp[u];
 182                         }
 183                         cc += (uint32_t)len >> 4;
 184                         break;
 185                 }
 186         }
 187         return cc;
 188 }
 189
 190 /* see bearssl_block.h */
 191 const br_block_ctr_class br_aes_x86ni_ctr_vtable = {
 192         sizeof(br_aes_x86ni_ctr_keys),
 193         16,
 194         4,
 195         (void (*)(const br_block_ctr_class **, const void *, size_t))
 196                 &br_aes_x86ni_ctr_init,
 197         (uint32_t (*)(const br_block_ctr_class *const *,
 198                 const void *, uint32_t, void *, size_t))
 199                 &br_aes_x86ni_ctr_run
 200 };
 201
 202 /* see bearssl_block.h */
 203 const br_block_ctr_class *
 204 br_aes_x86ni_ctr_get_vtable(void)
 205 {
 206         return br_aes_x86ni_supported() ? &br_aes_x86ni_ctr_vtable : NULL;
 207 }
 208
 209 #else
 210
 211 /* see bearssl_block.h */
 212 const br_block_ctr_class *
 213 br_aes_x86ni_ctr_get_vtable(void)
 214 {
 215         return NULL;
 216 }
 217
 218 #endif