diff options
Diffstat (limited to 'src/stream/salsa20/salsa20.cpp')
-rw-r--r-- | src/stream/salsa20/salsa20.cpp | 154 |
1 files changed, 89 insertions, 65 deletions
diff --git a/src/stream/salsa20/salsa20.cpp b/src/stream/salsa20/salsa20.cpp index 1b97f4421..a38e6e305 100644 --- a/src/stream/salsa20/salsa20.cpp +++ b/src/stream/salsa20/salsa20.cpp @@ -1,6 +1,6 @@ /* -* Salsa20 -* (C) 1999-2008 Jack Lloyd +* Salsa20 / XSalsa20 +* (C) 1999-2010 Jack Lloyd * * Distributed under the terms of the Botan license */ @@ -14,63 +14,68 @@ namespace Botan { namespace { +#define SALSA20_QUARTER_ROUND(x1, x2, x3, x4) \ + do { \ + x2 ^= rotate_left(x1 + x4, 7); \ + x3 ^= rotate_left(x2 + x1, 9); \ + x4 ^= rotate_left(x3 + x2, 13); \ + x1 ^= rotate_left(x4 + x3, 18); \ + } while(0) + +/* +* Generate HSalsa20 cipher stream (for XSalsa20 IV setup) +*/ +void hsalsa20(u32bit output[8], const u32bit input[16]) + { + u32bit x00 = input[ 0], x01 = input[ 1], x02 = input[ 2], x03 = input[ 3], + x04 = input[ 4], x05 = input[ 5], x06 = input[ 6], x07 = input[ 7], + x08 = input[ 8], x09 = input[ 9], x10 = input[10], x11 = input[11], + x12 = input[12], x13 = input[13], x14 = input[14], x15 = input[15]; + + for(u32bit i = 0; i != 10; ++i) + { + SALSA20_QUARTER_ROUND(x00, x04, x08, x12); + SALSA20_QUARTER_ROUND(x05, x09, x13, x01); + SALSA20_QUARTER_ROUND(x10, x14, x02, x06); + SALSA20_QUARTER_ROUND(x15, x03, x07, x11); + + SALSA20_QUARTER_ROUND(x00, x01, x02, x03); + SALSA20_QUARTER_ROUND(x05, x06, x07, x04); + SALSA20_QUARTER_ROUND(x10, x11, x08, x09); + SALSA20_QUARTER_ROUND(x15, x12, x13, x14); + } + + output[0] = x00; + output[1] = x05; + output[2] = x10; + output[3] = x15; + output[4] = x06; + output[5] = x07; + output[6] = x08; + output[7] = x09; + } + /* * Generate Salsa20 cipher stream */ void salsa20(byte output[64], const u32bit input[16]) { - u32bit x00 = input[0]; - u32bit x01 = input[1]; - u32bit x02 = input[2]; - u32bit x03 = input[3]; - u32bit x04 = input[4]; - u32bit x05 = input[5]; - u32bit x06 = input[6]; - u32bit x07 = input[7]; - u32bit x08 = input[8]; - u32bit x09 = input[9]; - u32bit x10 = input[10]; - u32bit x11 = input[11]; - u32bit x12 = input[12]; - u32bit x13 = input[13]; - u32bit x14 = input[14]; - u32bit x15 = input[15]; + u32bit x00 = input[ 0], x01 = input[ 1], x02 = input[ 2], x03 = input[ 3], + x04 = input[ 4], x05 = input[ 5], x06 = input[ 6], x07 = input[ 7], + x08 = input[ 8], x09 = input[ 9], x10 = input[10], x11 = input[11], + x12 = input[12], x13 = input[13], x14 = input[14], x15 = input[15]; for(u32bit i = 0; i != 10; ++i) { - x04 ^= rotate_left(x00 + x12, 7); - x08 ^= rotate_left(x04 + x00, 9); - x12 ^= rotate_left(x08 + x04, 13); - x00 ^= rotate_left(x12 + x08, 18); - x09 ^= rotate_left(x05 + x01, 7); - x13 ^= rotate_left(x09 + x05, 9); - x01 ^= rotate_left(x13 + x09, 13); - x05 ^= rotate_left(x01 + x13, 18); - x14 ^= rotate_left(x10 + x06, 7); - x02 ^= rotate_left(x14 + x10, 9); - x06 ^= rotate_left(x02 + x14, 13); - x10 ^= rotate_left(x06 + x02, 18); - x03 ^= rotate_left(x15 + x11, 7); - x07 ^= rotate_left(x03 + x15, 9); - x11 ^= rotate_left(x07 + x03, 13); - x15 ^= rotate_left(x11 + x07, 18); - - x01 ^= rotate_left(x00 + x03, 7); - x02 ^= rotate_left(x01 + x00, 9); - x03 ^= rotate_left(x02 + x01, 13); - x00 ^= rotate_left(x03 + x02, 18); - x06 ^= rotate_left(x05 + x04, 7); - x07 ^= rotate_left(x06 + x05, 9); - x04 ^= rotate_left(x07 + x06, 13); - x05 ^= rotate_left(x04 + x07, 18); - x11 ^= rotate_left(x10 + x09, 7); - x08 ^= rotate_left(x11 + x10, 9); - x09 ^= rotate_left(x08 + x11, 13); - x10 ^= rotate_left(x09 + x08, 18); - x12 ^= rotate_left(x15 + x14, 7); - x13 ^= rotate_left(x12 + x15, 9); - x14 ^= rotate_left(x13 + x12, 13); - x15 ^= rotate_left(x14 + x13, 18); + SALSA20_QUARTER_ROUND(x00, x04, x08, x12); + SALSA20_QUARTER_ROUND(x05, x09, x13, x01); + SALSA20_QUARTER_ROUND(x10, x14, x02, x06); + SALSA20_QUARTER_ROUND(x15, x03, x07, x11); + + SALSA20_QUARTER_ROUND(x00, x01, x02, x03); + SALSA20_QUARTER_ROUND(x05, x06, x07, x04); + SALSA20_QUARTER_ROUND(x10, x11, x08, x09); + SALSA20_QUARTER_ROUND(x15, x12, x13, x14); } store_le(x00 + input[ 0], output + 4 * 0); @@ -100,11 +105,11 @@ void Salsa20::cipher(const byte in[], byte out[], u32bit length) { while(length >= buffer.size() - position) { - xor_buf(out, in, buffer.begin() + position, buffer.size() - position); + xor_buf(out, in, &buffer[position], buffer.size() - position); length -= (buffer.size() - position); in += (buffer.size() - position); out += (buffer.size() - position); - salsa20(buffer.begin(), state); + salsa20(&buffer[0], state); ++state[8]; if(!state[8]) // if overflow in state[8] @@ -113,7 +118,7 @@ void Salsa20::cipher(const byte in[], byte out[], u32bit length) position = 0; } - xor_buf(out, in, buffer.begin() + position, length); + xor_buf(out, in, &buffer[position], length); position += length; } @@ -174,12 +179,39 @@ void Salsa20::set_iv(const byte iv[], u32bit length) if(!valid_iv_length(length)) throw Invalid_IV_Length(name(), length); - state[6] = load_le<u32bit>(iv, 0); - state[7] = load_le<u32bit>(iv, 1); + if(length == 8) + { + // Salsa20 + state[6] = load_le<u32bit>(iv, 0); + state[7] = load_le<u32bit>(iv, 1); + } + else + { + // XSalsa20 + state[6] = load_le<u32bit>(iv, 0); + state[7] = load_le<u32bit>(iv, 1); + state[8] = load_le<u32bit>(iv, 2); + state[9] = load_le<u32bit>(iv, 3); + + SecureVector<u32bit> hsalsa(8); + hsalsa20(hsalsa, state); + + state[ 1] = hsalsa[0]; + state[ 2] = hsalsa[1]; + state[ 3] = hsalsa[2]; + state[ 4] = hsalsa[3]; + state[ 6] = load_le<u32bit>(iv, 4); + state[ 7] = load_le<u32bit>(iv, 5); + state[11] = hsalsa[4]; + state[12] = hsalsa[5]; + state[13] = hsalsa[6]; + state[14] = hsalsa[7]; + } + state[8] = 0; state[9] = 0; - salsa20(buffer.begin(), state); + salsa20(&buffer[0], state); ++state[8]; if(!state[8]) // if overflow in state[8] ++state[9]; // carry to state[9] @@ -205,12 +237,4 @@ void Salsa20::clear() position = 0; } -/* -* Salsa20 Constructor -*/ -Salsa20::Salsa20() : StreamCipher(16, 32, 16) - { - clear(); - } - } |