release/src/router/nettle/aes-encrypt-internal.c

   1 /* aes-encrypt-internal.c
   2  *
   3  * Encryption function for the aes/rijndael block cipher.
   4  */
   5
   6 /* nettle, low-level cryptographics library
   7  *
   8  * Copyright (C) 2002 Niels Möller
   9  *
  10  * The nettle library is free software; you can redistribute it and/or modify
  11  * it under the terms of the GNU Lesser General Public License as published by
  12  * the Free Software Foundation; either version 2.1 of the License, or (at your
  13  * option) any later version.
  14  *
  15  * The nettle library is distributed in the hope that it will be useful, but
  16  * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
  17  * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
  18  * License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public License
  21  * along with the nettle library; see the file COPYING.LIB.  If not, write to
  22  * the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
  23  * MA 02111-1301, USA.
  24  */
  25
  26 #if HAVE_CONFIG_H
  27 # include "config.h"
  28 #endif
  29
  30 #include <assert.h>
  31
  32 #include "aes-internal.h"
  33 #include "macros.h"
  34
  35 void
  36 _nettle_aes_encrypt(const struct aes_ctx *ctx,
  37                     const struct aes_table *T,
  38                     unsigned length, uint8_t *dst,
  39                     const uint8_t *src)
  40 {
  41   FOR_BLOCKS(length, dst, src, AES_BLOCK_SIZE)
  42     {
  43       uint32_t w0, w1, w2, w3;          /* working ciphertext */
  44       uint32_t t0, t1, t2, t3;
  45       unsigned round;
  46
  47       /* Get clear text, using little-endian byte order.
  48        * Also XOR with the first subkey. */
  49
  50       w0 = LE_READ_UINT32(src)      ^ ctx->keys[0];
  51       w1 = LE_READ_UINT32(src + 4)  ^ ctx->keys[1];
  52       w2 = LE_READ_UINT32(src + 8)  ^ ctx->keys[2];
  53       w3 = LE_READ_UINT32(src + 12) ^ ctx->keys[3];
  54
  55       for (round = 1; round < ctx->nrounds; round++)
  56         {
  57           t0 = AES_ROUND(T, w0, w1, w2, w3, ctx->keys[4*round]);
  58           t1 = AES_ROUND(T, w1, w2, w3, w0, ctx->keys[4*round + 1]);
  59           t2 = AES_ROUND(T, w2, w3, w0, w1, ctx->keys[4*round + 2]);
  60           t3 = AES_ROUND(T, w3, w0, w1, w2, ctx->keys[4*round + 3]);
  61
  62           /* We could unroll the loop twice, to avoid these
  63              assignments. If all eight variables fit in registers,
  64              that should give a slight speedup. */
  65           w0 = t0;
  66           w1 = t1;
  67           w2 = t2;
  68           w3 = t3;
  69         }
  70
  71       /* Final round */
  72
  73       t0 = AES_FINAL_ROUND(T, w0, w1, w2, w3, ctx->keys[4*round]);
  74       t1 = AES_FINAL_ROUND(T, w1, w2, w3, w0, ctx->keys[4*round + 1]);
  75       t2 = AES_FINAL_ROUND(T, w2, w3, w0, w1, ctx->keys[4*round + 2]);
  76       t3 = AES_FINAL_ROUND(T, w3, w0, w1, w2, ctx->keys[4*round + 3]);
  77
  78       LE_WRITE_UINT32(dst, t0);
  79       LE_WRITE_UINT32(dst + 8, t2);
  80       LE_WRITE_UINT32(dst + 4, t1);
  81       LE_WRITE_UINT32(dst + 12, t3);
  82     }
  83 }
  84
  85 /* Some stats, all for AES 128:
  86
  87    A. Table-driven indexing (the approach of the old unified
  88       _aes_crypt function).
  89    B. Unrolling the j-loop.
  90
  91    C. Eliminated the use of IDXk(j) in the main loop.
  92
  93    D. Put wtxt in four scalar variables.
  94
  95    E. Also put t in four scalar variables.
  96
  97        P4 2.2 GHz         AMD Duron 1.4GHz
  98
  99        MB/s  code size
 100    A   35.9  0x202        17 MB/s
 101    B   37.3  0x334
 102    C   33.0  0x2a7
 103    D   40.7  0x3f9
 104    E   42.9  0x44a        26 MB/s
 105  */