Home | History | Annotate | Download | only in chacha
      1 /* Copyright (c) 2014, Google Inc.
      2  *
      3  * Permission to use, copy, modify, and/or distribute this software for any
      4  * purpose with or without fee is hereby granted, provided that the above
      5  * copyright notice and this permission notice appear in all copies.
      6  *
      7  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
      8  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
      9  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
     10  * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
     11  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
     12  * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
     13  * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
     14 
     15 // Adapted from the public domain, estream code by D. Bernstein.
     16 
     17 #include <openssl/chacha.h>
     18 
     19 #include <assert.h>
     20 #include <string.h>
     21 
     22 #include <openssl/cpu.h>
     23 
     24 #include "../internal.h"
     25 
     26 
     27 #define U8TO32_LITTLE(p)                              \
     28   (((uint32_t)((p)[0])) | ((uint32_t)((p)[1]) << 8) | \
     29    ((uint32_t)((p)[2]) << 16) | ((uint32_t)((p)[3]) << 24))
     30 
     31 #if !defined(OPENSSL_NO_ASM) &&                         \
     32     (defined(OPENSSL_X86) || defined(OPENSSL_X86_64) || \
     33      defined(OPENSSL_ARM) || defined(OPENSSL_AARCH64))
     34 
     35 // ChaCha20_ctr32 is defined in asm/chacha-*.pl.
     36 void ChaCha20_ctr32(uint8_t *out, const uint8_t *in, size_t in_len,
     37                     const uint32_t key[8], const uint32_t counter[4]);
     38 
     39 void CRYPTO_chacha_20(uint8_t *out, const uint8_t *in, size_t in_len,
     40                       const uint8_t key[32], const uint8_t nonce[12],
     41                       uint32_t counter) {
     42   assert(!buffers_alias(out, in_len, in, in_len) || in == out);
     43 
     44   uint32_t counter_nonce[4];  counter_nonce[0] = counter;
     45   counter_nonce[1] = U8TO32_LITTLE(nonce + 0);
     46   counter_nonce[2] = U8TO32_LITTLE(nonce + 4);
     47   counter_nonce[3] = U8TO32_LITTLE(nonce + 8);
     48 
     49   const uint32_t *key_ptr = (const uint32_t *)key;
     50 #if !defined(OPENSSL_X86) && !defined(OPENSSL_X86_64)
     51   // The assembly expects the key to be four-byte aligned.
     52   uint32_t key_u32[8];
     53   if ((((uintptr_t)key) & 3) != 0) {
     54     key_u32[0] = U8TO32_LITTLE(key + 0);
     55     key_u32[1] = U8TO32_LITTLE(key + 4);
     56     key_u32[2] = U8TO32_LITTLE(key + 8);
     57     key_u32[3] = U8TO32_LITTLE(key + 12);
     58     key_u32[4] = U8TO32_LITTLE(key + 16);
     59     key_u32[5] = U8TO32_LITTLE(key + 20);
     60     key_u32[6] = U8TO32_LITTLE(key + 24);
     61     key_u32[7] = U8TO32_LITTLE(key + 28);
     62 
     63     key_ptr = key_u32;
     64   }
     65 #endif
     66 
     67   ChaCha20_ctr32(out, in, in_len, key_ptr, counter_nonce);
     68 }
     69 
     70 #else
     71 
     72 // sigma contains the ChaCha constants, which happen to be an ASCII string.
     73 static const uint8_t sigma[16] = { 'e', 'x', 'p', 'a', 'n', 'd', ' ', '3',
     74                                    '2', '-', 'b', 'y', 't', 'e', ' ', 'k' };
     75 
     76 #define ROTATE(v, n) (((v) << (n)) | ((v) >> (32 - (n))))
     77 
     78 #define U32TO8_LITTLE(p, v)    \
     79   {                            \
     80     (p)[0] = (v >> 0) & 0xff;  \
     81     (p)[1] = (v >> 8) & 0xff;  \
     82     (p)[2] = (v >> 16) & 0xff; \
     83     (p)[3] = (v >> 24) & 0xff; \
     84   }
     85 
     86 // QUARTERROUND updates a, b, c, d with a ChaCha "quarter" round.
     87 #define QUARTERROUND(a, b, c, d)                \
     88   x[a] += x[b]; x[d] = ROTATE(x[d] ^ x[a], 16); \
     89   x[c] += x[d]; x[b] = ROTATE(x[b] ^ x[c], 12); \
     90   x[a] += x[b]; x[d] = ROTATE(x[d] ^ x[a],  8); \
     91   x[c] += x[d]; x[b] = ROTATE(x[b] ^ x[c],  7);
     92 
     93 // chacha_core performs 20 rounds of ChaCha on the input words in
     94 // |input| and writes the 64 output bytes to |output|.
     95 static void chacha_core(uint8_t output[64], const uint32_t input[16]) {
     96   uint32_t x[16];
     97   int i;
     98 
     99   OPENSSL_memcpy(x, input, sizeof(uint32_t) * 16);
    100   for (i = 20; i > 0; i -= 2) {
    101     QUARTERROUND(0, 4, 8, 12)
    102     QUARTERROUND(1, 5, 9, 13)
    103     QUARTERROUND(2, 6, 10, 14)
    104     QUARTERROUND(3, 7, 11, 15)
    105     QUARTERROUND(0, 5, 10, 15)
    106     QUARTERROUND(1, 6, 11, 12)
    107     QUARTERROUND(2, 7, 8, 13)
    108     QUARTERROUND(3, 4, 9, 14)
    109   }
    110 
    111   for (i = 0; i < 16; ++i) {
    112     x[i] += input[i];
    113   }
    114   for (i = 0; i < 16; ++i) {
    115     U32TO8_LITTLE(output + 4 * i, x[i]);
    116   }
    117 }
    118 
    119 void CRYPTO_chacha_20(uint8_t *out, const uint8_t *in, size_t in_len,
    120                       const uint8_t key[32], const uint8_t nonce[12],
    121                       uint32_t counter) {
    122   assert(!buffers_alias(out, in_len, in, in_len) || in == out);
    123 
    124   uint32_t input[16];
    125   uint8_t buf[64];
    126   size_t todo, i;
    127 
    128   input[0] = U8TO32_LITTLE(sigma + 0);
    129   input[1] = U8TO32_LITTLE(sigma + 4);
    130   input[2] = U8TO32_LITTLE(sigma + 8);
    131   input[3] = U8TO32_LITTLE(sigma + 12);
    132 
    133   input[4] = U8TO32_LITTLE(key + 0);
    134   input[5] = U8TO32_LITTLE(key + 4);
    135   input[6] = U8TO32_LITTLE(key + 8);
    136   input[7] = U8TO32_LITTLE(key + 12);
    137 
    138   input[8] = U8TO32_LITTLE(key + 16);
    139   input[9] = U8TO32_LITTLE(key + 20);
    140   input[10] = U8TO32_LITTLE(key + 24);
    141   input[11] = U8TO32_LITTLE(key + 28);
    142 
    143   input[12] = counter;
    144   input[13] = U8TO32_LITTLE(nonce + 0);
    145   input[14] = U8TO32_LITTLE(nonce + 4);
    146   input[15] = U8TO32_LITTLE(nonce + 8);
    147 
    148   while (in_len > 0) {
    149     todo = sizeof(buf);
    150     if (in_len < todo) {
    151       todo = in_len;
    152     }
    153 
    154     chacha_core(buf, input);
    155     for (i = 0; i < todo; i++) {
    156       out[i] = in[i] ^ buf[i];
    157     }
    158 
    159     out += todo;
    160     in += todo;
    161     in_len -= todo;
    162 
    163     input[12]++;
    164   }
    165 }
    166 
    167 #endif
    168