From 546d6cff4447a56bbf7c0e1a8b6f89dba5d3183b Mon Sep 17 00:00:00 2001 From: Mounir IDRASSI Date: Fri, 23 Jun 2017 02:07:32 +0200 Subject: Crypto: Add optimized SHA-512 and SHA-256 assembly implementations for x86_64 and x86. This improves speed by 30%. --- src/Crypto/Sha2.c | 1401 ++++++++++++++++++++++++++++------------------------- 1 file changed, 747 insertions(+), 654 deletions(-) (limited to 'src/Crypto/Sha2.c') diff --git a/src/Crypto/Sha2.c b/src/Crypto/Sha2.c index 9dbb529f..05da532e 100644 --- a/src/Crypto/Sha2.c +++ b/src/Crypto/Sha2.c @@ -1,767 +1,860 @@ /* - --------------------------------------------------------------------------- - Copyright (c) 2002, Dr Brian Gladman, Worcester, UK. All rights reserved. - - LICENSE TERMS - - The free distribution and use of this software is allowed (with or without - changes) provided that: - - 1. source code distributions include the above copyright notice, this - list of conditions and the following disclaimer; - - 2. binary distributions include the above copyright notice, this list - of conditions and the following disclaimer in their documentation; - - 3. the name of the copyright holder is not used to endorse products - built using this software without specific written permission. - - DISCLAIMER - - This software is provided 'as is' with no explicit or implied warranties - in respect of its properties, including, but not limited to, correctness - and/or fitness for purpose. - --------------------------------------------------------------------------- - Issue Date: 01/08/2005 - - This is a byte oriented version of SHA2 that operates on arrays of bytes - stored in memory. This code implements sha256, sha384 and sha512 but the - latter two functions rely on efficient 64-bit integer operations that - may not be very efficient on 32-bit machines - - The sha256 functions use a type 'sha256_ctx' to hold details of the - current hash state and uses the following three calls: - - void sha256_begin(sha256_ctx ctx[1]) - void sha256_hash(const unsigned char data[], - unsigned long len, sha256_ctx ctx[1]) - void sha_end1(unsigned char hval[], sha256_ctx ctx[1]) - - The first subroutine initialises a hash computation by setting up the - context in the sha256_ctx context. The second subroutine hashes 8-bit - bytes from array data[] into the hash state withinh sha256_ctx context, - the number of bytes to be hashed being given by the the unsigned long - integer len. The third subroutine completes the hash calculation and - places the resulting digest value in the array of 8-bit bytes hval[]. - - The sha384 and sha512 functions are similar and use the interfaces: - - void sha384_begin(sha384_ctx ctx[1]); - void sha384_hash(const unsigned char data[], - unsigned long len, sha384_ctx ctx[1]); - void sha384_end(unsigned char hval[], sha384_ctx ctx[1]); - - void sha512_begin(sha512_ctx ctx[1]); - void sha512_hash(const unsigned char data[], - unsigned long len, sha512_ctx ctx[1]); - void sha512_end(unsigned char hval[], sha512_ctx ctx[1]); - - In addition there is a function sha2 that can be used to call all these - functions using a call with a hash length parameter as follows: - - int sha2_begin(unsigned long len, sha2_ctx ctx[1]); - void sha2_hash(const unsigned char data[], - unsigned long len, sha2_ctx ctx[1]); - void sha2_end(unsigned char hval[], sha2_ctx ctx[1]); - - My thanks to Erik Andersen for testing this code - on big-endian systems and for his assistance with corrections +This code is written by kerukuro for cppcrypto library (http://cppcrypto.sourceforge.net/) +and released into public domain. */ +/* Modified for VeraCrypt with speed optimization for C implementation */ + +#include "Sha2.h" #include "Common/Endian.h" -#include "Common/Tcdefs.h" +#include "Crypto/cpu.h" #include "Crypto/misc.h" -#define PLATFORM_BYTE_ORDER BYTE_ORDER -#define IS_LITTLE_ENDIAN LITTLE_ENDIAN - -#if 0 -#define UNROLL_SHA2 /* for SHA2 loop unroll */ +#ifdef _UEFI +#define NO_OPTIMIZED_VERSIONS #endif -#if !defined(_UEFI) -#include /* for memcpy() etc. */ -#endif // !defined(_UEFI) - -#include "Sha2.h" +#ifndef NO_OPTIMIZED_VERSIONS #if defined(__cplusplus) extern "C" { #endif - -#if defined( _MSC_VER ) && ( _MSC_VER > 800 ) && !defined(_UEFI) -#pragma intrinsic(memcpy) +#if CRYPTOPP_BOOL_X64 + void sha512_rorx(const void* M, void* D, uint_64t l); + void sha512_sse4(const void* M, uint_64t D[8], uint_64t l); + void sha512_avx(const void* M, void* D, uint_64t l); #endif - -#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN) -#define SWAP_BYTES -#else -#undef SWAP_BYTES + +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 || CRYPTOPP_BOOL_X64 + void sha512_compress_nayuki(uint_64t state[8], const uint_8t block[128]); #endif - -#if 0 - -#define ch(x,y,z) (((x) & (y)) ^ (~(x) & (z))) -#define maj(x,y,z) (((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z))) - -#else /* Thanks to Rich Schroeppel and Colin Plumb for the following */ - -#define ch(x,y,z) ((z) ^ ((x) & ((y) ^ (z)))) -#define maj(x,y,z) (((x) & (y)) | ((z) & ((x) ^ (y)))) - +#if defined(__cplusplus) +} #endif -/* round transforms for SHA256 and SHA512 compression functions */ - -#define vf(n,i) v[(n - i) & 7] - -#define hf(i) (p[i & 15] += \ - g_1(p[(i + 14) & 15]) + p[(i + 9) & 15] + g_0(p[(i + 1) & 15])) - -#define v_cycle(i,j) \ - vf(7,i) += (j ? hf(i) : p[i]) + k_0[i+j] \ - + s_1(vf(4,i)) + ch(vf(4,i),vf(5,i),vf(6,i)); \ - vf(3,i) += vf(7,i); \ - vf(7,i) += s_0(vf(0,i))+ maj(vf(0,i),vf(1,i),vf(2,i)) - -#if defined(SHA_224) || defined(SHA_256) - -#define SHA256_MASK (SHA256_BLOCK_SIZE - 1) +#endif -#if defined(SWAP_BYTES) -#define bsw_32(p,n) \ - { int _i = (n); while(_i--) ((uint_32t*)p)[_i] = bswap_32(((uint_32t*)p)[_i]); } -#else -#define bsw_32(p,n) -#endif - -#define s_0(x) (rotr32((x), 2) ^ rotr32((x), 13) ^ rotr32((x), 22)) -#define s_1(x) (rotr32((x), 6) ^ rotr32((x), 11) ^ rotr32((x), 25)) -#define g_0(x) (rotr32((x), 7) ^ rotr32((x), 18) ^ ((x) >> 3)) -#define g_1(x) (rotr32((x), 17) ^ rotr32((x), 19) ^ ((x) >> 10)) -#define k_0 k256 - -/* rotated SHA256 round definition. Rather than swapping variables as in */ -/* FIPS-180, different variables are 'rotated' on each round, returning */ -/* to their starting positions every eight rounds */ - -#define q(n) v##n - -#define one_cycle(a,b,c,d,e,f,g,h,k,w) \ - q(h) += s_1(q(e)) + ch(q(e), q(f), q(g)) + k + w; \ - q(d) += q(h); q(h) += s_0(q(a)) + maj(q(a), q(b), q(c)) - -/* SHA256 mixing data */ - -const uint_32t k256[64] = -{ 0x428a2f98ul, 0x71374491ul, 0xb5c0fbcful, 0xe9b5dba5ul, - 0x3956c25bul, 0x59f111f1ul, 0x923f82a4ul, 0xab1c5ed5ul, - 0xd807aa98ul, 0x12835b01ul, 0x243185beul, 0x550c7dc3ul, - 0x72be5d74ul, 0x80deb1feul, 0x9bdc06a7ul, 0xc19bf174ul, - 0xe49b69c1ul, 0xefbe4786ul, 0x0fc19dc6ul, 0x240ca1ccul, - 0x2de92c6ful, 0x4a7484aaul, 0x5cb0a9dcul, 0x76f988daul, - 0x983e5152ul, 0xa831c66dul, 0xb00327c8ul, 0xbf597fc7ul, - 0xc6e00bf3ul, 0xd5a79147ul, 0x06ca6351ul, 0x14292967ul, - 0x27b70a85ul, 0x2e1b2138ul, 0x4d2c6dfcul, 0x53380d13ul, - 0x650a7354ul, 0x766a0abbul, 0x81c2c92eul, 0x92722c85ul, - 0xa2bfe8a1ul, 0xa81a664bul, 0xc24b8b70ul, 0xc76c51a3ul, - 0xd192e819ul, 0xd6990624ul, 0xf40e3585ul, 0x106aa070ul, - 0x19a4c116ul, 0x1e376c08ul, 0x2748774cul, 0x34b0bcb5ul, - 0x391c0cb3ul, 0x4ed8aa4aul, 0x5b9cca4ful, 0x682e6ff3ul, - 0x748f82eeul, 0x78a5636ful, 0x84c87814ul, 0x8cc70208ul, - 0x90befffaul, 0xa4506cebul, 0xbef9a3f7ul, 0xc67178f2ul, +typedef void (*transformFn)(sha512_ctx* ctx, void* m, uint_64t num_blks); + +transformFn transfunc = NULL; + +static const uint_64t K[80] = { + 0x428a2f98d728ae22, 0x7137449123ef65cd, 0xb5c0fbcfec4d3b2f, 0xe9b5dba58189dbbc, + 0x3956c25bf348b538, 0x59f111f1b605d019, 0x923f82a4af194f9b, 0xab1c5ed5da6d8118, + 0xd807aa98a3030242, 0x12835b0145706fbe, 0x243185be4ee4b28c, 0x550c7dc3d5ffb4e2, + 0x72be5d74f27b896f, 0x80deb1fe3b1696b1, 0x9bdc06a725c71235, 0xc19bf174cf692694, + 0xe49b69c19ef14ad2, 0xefbe4786384f25e3, 0x0fc19dc68b8cd5b5, 0x240ca1cc77ac9c65, + 0x2de92c6f592b0275, 0x4a7484aa6ea6e483, 0x5cb0a9dcbd41fbd4, 0x76f988da831153b5, + 0x983e5152ee66dfab, 0xa831c66d2db43210, 0xb00327c898fb213f, 0xbf597fc7beef0ee4, + 0xc6e00bf33da88fc2, 0xd5a79147930aa725, 0x06ca6351e003826f, 0x142929670a0e6e70, + 0x27b70a8546d22ffc, 0x2e1b21385c26c926, 0x4d2c6dfc5ac42aed, 0x53380d139d95b3df, + 0x650a73548baf63de, 0x766a0abb3c77b2a8, 0x81c2c92e47edaee6, 0x92722c851482353b, + 0xa2bfe8a14cf10364, 0xa81a664bbc423001, 0xc24b8b70d0f89791, 0xc76c51a30654be30, + 0xd192e819d6ef5218, 0xd69906245565a910, 0xf40e35855771202a, 0x106aa07032bbd1b8, + 0x19a4c116b8d2d0c8, 0x1e376c085141ab53, 0x2748774cdf8eeb99, 0x34b0bcb5e19b48a8, + 0x391c0cb3c5c95a63, 0x4ed8aa4ae3418acb, 0x5b9cca4f7763e373, 0x682e6ff3d6b2b8a3, + 0x748f82ee5defb2fc, 0x78a5636f43172f60, 0x84c87814a1f0ab72, 0x8cc702081a6439ec, + 0x90befffa23631e28, 0xa4506cebde82bde9, 0xbef9a3f7b2c67915, 0xc67178f2e372532b, + 0xca273eceea26619c, 0xd186b8c721c0c207, 0xeada7dd6cde0eb1e, 0xf57d4f7fee6ed178, + 0x06f067aa72176fba, 0x0a637dc5a2c898a6, 0x113f9804bef90dae, 0x1b710b35131c471b, + 0x28db77f523047d84, 0x32caab7b40c72493, 0x3c9ebe0a15c9bebc, 0x431d67c49c100d4c, + 0x4cc5d4becb3e42b6, 0x597f299cfc657e2a, 0x5fcb6fab3ad6faec, 0x6c44198c4a475817 }; -/* Compile 64 bytes of hash data into SHA256 digest value */ -/* NOTE: this routine assumes that the byte order in the */ -/* ctx->wbuf[] at this point is such that low address bytes */ -/* in the ORIGINAL byte stream will go into the high end of */ -/* words on BOTH big and little endian systems */ -VOID_RETURN sha256_compile(sha256_ctx ctx[1]) +#define Ch(x,y,z) ((z) ^ ((x) & ((y) ^ (z)))) +#define Maj(x,y,z) (((x) & (y)) | ((z) & ((x) ^ (y)))) +#define sum0(x) (rotr64((x), 28) ^ rotr64((x), 34) ^ rotr64((x), 39)) +#define sum1(x) (rotr64((x), 14) ^ rotr64((x), 18) ^ rotr64((x), 41)) +#define sigma0(x) (rotr64((x), 1) ^ rotr64((x), 8) ^ ((x) >> 7)) +#define sigma1(x) (rotr64((x), 19) ^ rotr64((x), 61) ^ ((x) >> 6)) + +#define WU(j) (W[j & 15] += sigma1(W[(j + 14) & 15]) + W[(j + 9) & 15] + sigma0(W[(j + 1) & 15])) + +#define COMPRESS_ROUND(i, j, K) \ + T1 = h + sum1(e) + Ch(e, f, g) + K[i + j] + (i? WU(j): W[j]); \ + T2 = sum0(a) + Maj(a, b, c); \ + h = g; \ + g = f; \ + f = e; \ + e = d + T1; \ + d = c; \ + c = b; \ + b = a; \ + a = T1 + T2; + +void StdTransform(sha512_ctx* ctx, void* mp, uint_64t num_blks) { -#if !defined(UNROLL_SHA2) - - uint_32t j, *p = ctx->wbuf, v[8]; - - memcpy(v, ctx->hash, 8 * sizeof(uint_32t)); - - for(j = 0; j < 64; j += 16) - { - v_cycle( 0, j); v_cycle( 1, j); - v_cycle( 2, j); v_cycle( 3, j); - v_cycle( 4, j); v_cycle( 5, j); - v_cycle( 6, j); v_cycle( 7, j); - v_cycle( 8, j); v_cycle( 9, j); - v_cycle(10, j); v_cycle(11, j); - v_cycle(12, j); v_cycle(13, j); - v_cycle(14, j); v_cycle(15, j); - } - - ctx->hash[0] += v[0]; ctx->hash[1] += v[1]; - ctx->hash[2] += v[2]; ctx->hash[3] += v[3]; - ctx->hash[4] += v[4]; ctx->hash[5] += v[5]; - ctx->hash[6] += v[6]; ctx->hash[7] += v[7]; + uint_64t blk; + for (blk = 0; blk < num_blks; blk++) + { + uint_64t W[16]; + uint_64t a,b,c,d,e,f,g,h; + uint_64t T1, T2; + int i; +#if defined (TC_WINDOWS_DRIVER) && defined (DEBUG) + int j; +#endif + for (i = 0; i < 128 / 8; i++) + { + W[i] = bswap_64((((const uint_64t*)(mp))[blk * 16 + i])); + } + + a = ctx->hash[0]; + b = ctx->hash[1]; + c = ctx->hash[2]; + d = ctx->hash[3]; + e = ctx->hash[4]; + f = ctx->hash[5]; + g = ctx->hash[6]; + h = ctx->hash[7]; + + for (i = 0; i <= 79; i+=16) + { +#if defined (TC_WINDOWS_DRIVER) && defined (DEBUG) + for (j = 0; j < 16; j++) + { + COMPRESS_ROUND(i, j, K); + } #else - - uint_32t *p = ctx->wbuf,v0,v1,v2,v3,v4,v5,v6,v7; - - v0 = ctx->hash[0]; v1 = ctx->hash[1]; - v2 = ctx->hash[2]; v3 = ctx->hash[3]; - v4 = ctx->hash[4]; v5 = ctx->hash[5]; - v6 = ctx->hash[6]; v7 = ctx->hash[7]; - - one_cycle(0,1,2,3,4,5,6,7,k256[ 0],p[ 0]); - one_cycle(7,0,1,2,3,4,5,6,k256[ 1],p[ 1]); - one_cycle(6,7,0,1,2,3,4,5,k256[ 2],p[ 2]); - one_cycle(5,6,7,0,1,2,3,4,k256[ 3],p[ 3]); - one_cycle(4,5,6,7,0,1,2,3,k256[ 4],p[ 4]); - one_cycle(3,4,5,6,7,0,1,2,k256[ 5],p[ 5]); - one_cycle(2,3,4,5,6,7,0,1,k256[ 6],p[ 6]); - one_cycle(1,2,3,4,5,6,7,0,k256[ 7],p[ 7]); - one_cycle(0,1,2,3,4,5,6,7,k256[ 8],p[ 8]); - one_cycle(7,0,1,2,3,4,5,6,k256[ 9],p[ 9]); - one_cycle(6,7,0,1,2,3,4,5,k256[10],p[10]); - one_cycle(5,6,7,0,1,2,3,4,k256[11],p[11]); - one_cycle(4,5,6,7,0,1,2,3,k256[12],p[12]); - one_cycle(3,4,5,6,7,0,1,2,k256[13],p[13]); - one_cycle(2,3,4,5,6,7,0,1,k256[14],p[14]); - one_cycle(1,2,3,4,5,6,7,0,k256[15],p[15]); - - one_cycle(0,1,2,3,4,5,6,7,k256[16],hf( 0)); - one_cycle(7,0,1,2,3,4,5,6,k256[17],hf( 1)); - one_cycle(6,7,0,1,2,3,4,5,k256[18],hf( 2)); - one_cycle(5,6,7,0,1,2,3,4,k256[19],hf( 3)); - one_cycle(4,5,6,7,0,1,2,3,k256[20],hf( 4)); - one_cycle(3,4,5,6,7,0,1,2,k256[21],hf( 5)); - one_cycle(2,3,4,5,6,7,0,1,k256[22],hf( 6)); - one_cycle(1,2,3,4,5,6,7,0,k256[23],hf( 7)); - one_cycle(0,1,2,3,4,5,6,7,k256[24],hf( 8)); - one_cycle(7,0,1,2,3,4,5,6,k256[25],hf( 9)); - one_cycle(6,7,0,1,2,3,4,5,k256[26],hf(10)); - one_cycle(5,6,7,0,1,2,3,4,k256[27],hf(11)); - one_cycle(4,5,6,7,0,1,2,3,k256[28],hf(12)); - one_cycle(3,4,5,6,7,0,1,2,k256[29],hf(13)); - one_cycle(2,3,4,5,6,7,0,1,k256[30],hf(14)); - one_cycle(1,2,3,4,5,6,7,0,k256[31],hf(15)); - - one_cycle(0,1,2,3,4,5,6,7,k256[32],hf( 0)); - one_cycle(7,0,1,2,3,4,5,6,k256[33],hf( 1)); - one_cycle(6,7,0,1,2,3,4,5,k256[34],hf( 2)); - one_cycle(5,6,7,0,1,2,3,4,k256[35],hf( 3)); - one_cycle(4,5,6,7,0,1,2,3,k256[36],hf( 4)); - one_cycle(3,4,5,6,7,0,1,2,k256[37],hf( 5)); - one_cycle(2,3,4,5,6,7,0,1,k256[38],hf( 6)); - one_cycle(1,2,3,4,5,6,7,0,k256[39],hf( 7)); - one_cycle(0,1,2,3,4,5,6,7,k256[40],hf( 8)); - one_cycle(7,0,1,2,3,4,5,6,k256[41],hf( 9)); - one_cycle(6,7,0,1,2,3,4,5,k256[42],hf(10)); - one_cycle(5,6,7,0,1,2,3,4,k256[43],hf(11)); - one_cycle(4,5,6,7,0,1,2,3,k256[44],hf(12)); - one_cycle(3,4,5,6,7,0,1,2,k256[45],hf(13)); - one_cycle(2,3,4,5,6,7,0,1,k256[46],hf(14)); - one_cycle(1,2,3,4,5,6,7,0,k256[47],hf(15)); - - one_cycle(0,1,2,3,4,5,6,7,k256[48],hf( 0)); - one_cycle(7,0,1,2,3,4,5,6,k256[49],hf( 1)); - one_cycle(6,7,0,1,2,3,4,5,k256[50],hf( 2)); - one_cycle(5,6,7,0,1,2,3,4,k256[51],hf( 3)); - one_cycle(4,5,6,7,0,1,2,3,k256[52],hf( 4)); - one_cycle(3,4,5,6,7,0,1,2,k256[53],hf( 5)); - one_cycle(2,3,4,5,6,7,0,1,k256[54],hf( 6)); - one_cycle(1,2,3,4,5,6,7,0,k256[55],hf( 7)); - one_cycle(0,1,2,3,4,5,6,7,k256[56],hf( 8)); - one_cycle(7,0,1,2,3,4,5,6,k256[57],hf( 9)); - one_cycle(6,7,0,1,2,3,4,5,k256[58],hf(10)); - one_cycle(5,6,7,0,1,2,3,4,k256[59],hf(11)); - one_cycle(4,5,6,7,0,1,2,3,k256[60],hf(12)); - one_cycle(3,4,5,6,7,0,1,2,k256[61],hf(13)); - one_cycle(2,3,4,5,6,7,0,1,k256[62],hf(14)); - one_cycle(1,2,3,4,5,6,7,0,k256[63],hf(15)); - - ctx->hash[0] += v0; ctx->hash[1] += v1; - ctx->hash[2] += v2; ctx->hash[3] += v3; - ctx->hash[4] += v4; ctx->hash[5] += v5; - ctx->hash[6] += v6; ctx->hash[7] += v7; + COMPRESS_ROUND(i, 0, K); + COMPRESS_ROUND(i, 1, K); + COMPRESS_ROUND(i , 2, K); + COMPRESS_ROUND(i, 3, K); + COMPRESS_ROUND(i, 4, K); + COMPRESS_ROUND(i, 5, K); + COMPRESS_ROUND(i, 6, K); + COMPRESS_ROUND(i, 7, K); + COMPRESS_ROUND(i, 8, K); + COMPRESS_ROUND(i, 9, K); + COMPRESS_ROUND(i, 10, K); + COMPRESS_ROUND(i, 11, K); + COMPRESS_ROUND(i, 12, K); + COMPRESS_ROUND(i, 13, K); + COMPRESS_ROUND(i, 14, K); + COMPRESS_ROUND(i, 15, K); #endif + } + ctx->hash[0] += a; + ctx->hash[1] += b; + ctx->hash[2] += c; + ctx->hash[3] += d; + ctx->hash[4] += e; + ctx->hash[5] += f; + ctx->hash[6] += g; + ctx->hash[7] += h; + } } -/* SHA256 hash data in an array of bytes into hash buffer */ -/* and call the hash_compile function as required. */ - -VOID_RETURN sha256_hash(const unsigned char data[], unsigned long len, sha256_ctx ctx[1]) -{ uint_32t pos = (uint_32t)(ctx->count[0] & SHA256_MASK), - space = SHA256_BLOCK_SIZE - pos; - const unsigned char *sp = data; - - if((ctx->count[0] += len) < len) - ++(ctx->count[1]); - - while(len >= space) /* tranfer whole blocks while possible */ - { - memcpy(((unsigned char*)ctx->wbuf) + pos, sp, space); - sp += space; len -= space; space = SHA256_BLOCK_SIZE; pos = 0; - bsw_32(ctx->wbuf, SHA256_BLOCK_SIZE >> 2) - sha256_compile(ctx); - } - - memcpy(((unsigned char*)ctx->wbuf) + pos, sp, len); -} - -/* SHA256 Final padding and digest calculation */ - -static void sha_end1(unsigned char hval[], sha256_ctx ctx[1], const unsigned int hlen) -{ uint_32t i = (uint_32t)(ctx->count[0] & SHA256_MASK); - - /* put bytes in the buffer in an order in which references to */ - /* 32-bit words will put bytes with lower addresses into the */ - /* top of 32 bit words on BOTH big and little endian machines */ - bsw_32(ctx->wbuf, (i + 3) >> 2) - - /* we now need to mask valid bytes and add the padding which is */ - /* a single 1 bit and as many zero bits as necessary. Note that */ - /* we can always add the first padding byte here because the */ - /* buffer always has at least one empty slot */ - ctx->wbuf[i >> 2] &= 0xffffff80 << 8 * (~i & 3); - ctx->wbuf[i >> 2] |= 0x00000080 << 8 * (~i & 3); - - /* we need 9 or more empty positions, one for the padding byte */ - /* (above) and eight for the length count. If there is not */ - /* enough space pad and empty the buffer */ - if(i > SHA256_BLOCK_SIZE - 9) - { - if(i < 60) ctx->wbuf[15] = 0; - sha256_compile(ctx); - i = 0; - } - else /* compute a word index for the empty buffer positions */ - i = (i >> 2) + 1; - - while(i < 14) /* and zero pad all but last two positions */ - ctx->wbuf[i++] = 0; - - /* the following 32-bit length fields are assembled in the */ - /* wrong byte order on little endian machines but this is */ - /* corrected later since they are only ever used as 32-bit */ - /* word values. */ - ctx->wbuf[14] = (ctx->count[1] << 3) | (ctx->count[0] >> 29); - ctx->wbuf[15] = ctx->count[0] << 3; - sha256_compile(ctx); - - /* extract the hash value as bytes in case the hash buffer is */ - /* mislaigned for 32-bit words */ - for(i = 0; i < hlen; ++i) - hval[i] = (unsigned char)(ctx->hash[i >> 2] >> (8 * (~i & 3))); -} - -#endif - -#if defined(SHA_224) +#ifndef NO_OPTIMIZED_VERSIONS -const uint_32t i224[8] = +#if CRYPTOPP_BOOL_X64 +void Avx2Transform(sha512_ctx* ctx, void* mp, uint_64t num_blks) { - 0xc1059ed8ul, 0x367cd507ul, 0x3070dd17ul, 0xf70e5939ul, - 0xffc00b31ul, 0x68581511ul, 0x64f98fa7ul, 0xbefa4fa4ul -}; + if (num_blks > 1) + sha512_rorx(mp, ctx->hash, num_blks); + else + sha512_sse4(mp, ctx->hash, num_blks); +} -VOID_RETURN sha224_begin(sha224_ctx ctx[1]) +void AvxTransform(sha512_ctx* ctx, void* mp, uint_64t num_blks) { - ctx->count[0] = ctx->count[1] = 0; - memcpy(ctx->hash, i224, 8 * sizeof(uint_32t)); + if (num_blks > 1) + sha512_avx(mp, ctx->hash, num_blks); + else + sha512_sse4(mp, ctx->hash, num_blks); } -VOID_RETURN sha224_end(unsigned char hval[], sha224_ctx ctx[1]) +void SSE4Transform(sha512_ctx* ctx, void* mp, uint_64t num_blks) { - sha_end1(hval, ctx, SHA224_DIGEST_SIZE); + sha512_sse4(mp, ctx->hash, num_blks); } +#endif -VOID_RETURN sha224(unsigned char hval[], const unsigned char data[], unsigned long len) -{ sha224_ctx cx[1]; +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 || CRYPTOPP_BOOL_X64 - sha224_begin(cx); - sha224_hash(data, len, cx); - sha_end1(hval, cx, SHA224_DIGEST_SIZE); +void SSE2Transform(sha512_ctx* ctx, void* mp, uint_64t num_blks) +{ + uint_64t i; + for (i = 0; i < num_blks; i++) + sha512_compress_nayuki(ctx->hash, (uint_8t*)mp + i * 128); } #endif -#if defined(SHA_256) +#endif // NO_OPTIMIZED_VERSIONS -const uint_32t i256[8] = +void sha512_begin(sha512_ctx* ctx) { - 0x6a09e667ul, 0xbb67ae85ul, 0x3c6ef372ul, 0xa54ff53aul, - 0x510e527ful, 0x9b05688cul, 0x1f83d9abul, 0x5be0cd19ul -}; + ctx->hash[0] = 0x6a09e667f3bcc908; + ctx->hash[1] = 0xbb67ae8584caa73b; + ctx->hash[2] = 0x3c6ef372fe94f82b; + ctx->hash[3] = 0xa54ff53a5f1d36f1; + ctx->hash[4] = 0x510e527fade682d1; + ctx->hash[5] = 0x9b05688c2b3e6c1f; + ctx->hash[6] = 0x1f83d9abfb41bd6b; + ctx->hash[7] = 0x5be0cd19137e2179; + ctx->count[0] = 0; + ctx->count[1] = 0; + + if (!transfunc) + { +#ifndef NO_OPTIMIZED_VERSIONS +#if CRYPTOPP_BOOL_X64 + if (g_isIntel&& HasSAVX2() && HasSBMI2()) + transfunc = Avx2Transform; + else if (g_isIntel && HasSAVX()) + { + transfunc = AvxTransform; + } + else if (HasSSE41()) + { + transfunc = SSE4Transform; + } + else +#endif -VOID_RETURN sha256_begin(sha256_ctx ctx[1]) +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 || CRYPTOPP_BOOL_X64 + if (HasSSE2() && HasMMX()) + transfunc = SSE2Transform; + else +#endif + +#endif + transfunc = StdTransform; + } +} + +void sha512_end(unsigned char * result, sha512_ctx* ctx) { - ctx->count[0] = ctx->count[1] = 0; - memcpy(ctx->hash, i256, 8 * sizeof(uint_32t)); + int i; + uint_64t mlen, pos = ctx->count[0]; + uint_8t* m = (uint_8t*) ctx->wbuf; + m[pos++] = 0x80; + if (pos > 112) + { + memset(m + pos, 0, (size_t) (128 - pos)); + transfunc(ctx, m, 1); + pos = 0; + } + memset(m + pos, 0, (size_t) (128 - pos)); + mlen = bswap_64(ctx->count[1]); + memcpy(m + (128 - 8), &mlen, 64 / 8); + transfunc(ctx, m, 1); + for (i = 0; i < 8; i++) + { + ctx->hash[i] = bswap_64(ctx->hash[i]); + } + memcpy(result, ctx->hash, 64); } -VOID_RETURN sha256_end(unsigned char hval[], sha256_ctx ctx[1]) +void sha512_hash(const unsigned char * data, uint_64t len, sha512_ctx *ctx) { - sha_end1(hval, ctx, SHA256_DIGEST_SIZE); + uint_64t pos = ctx->count[0]; + uint_64t total = ctx->count[1]; + uint_8t* m = (uint_8t*) ctx->wbuf; + if (pos && pos + len >= 128) + { + memcpy(m + pos, data, (size_t) (128 - pos)); + transfunc(ctx, m, 1); + len -= 128 - pos; + total += (128 - pos) * 8; + data += 128 - pos; + pos = 0; + } + if (len >= 128) + { + uint_64t blocks = len / 128; + uint_64t bytes = blocks * 128; + transfunc(ctx, (void*)data, blocks); + len -= bytes; + total += (bytes)* 8; + data += bytes; + } + memcpy(m+pos, data, (size_t) (len)); + pos += len; + total += len * 8; + ctx->count[0] = pos; + ctx->count[1] = total; } -VOID_RETURN sha256(unsigned char hval[], const unsigned char data[], unsigned long len) -{ sha256_ctx cx[1]; +void sha512(unsigned char * result, const unsigned char* source, uint_64t sourceLen) +{ + sha512_ctx ctx; - sha256_begin(cx); - sha256_hash(data, len, cx); - sha_end1(hval, cx, SHA256_DIGEST_SIZE); + sha512_begin(&ctx); + sha512_hash(source, sourceLen, &ctx); + sha512_end(result, &ctx); } -#endif +///////////////////////////// -#if defined(SHA_384) || defined(SHA_512) +#ifndef NO_OPTIMIZED_VERSIONS -#define SHA512_MASK (SHA512_BLOCK_SIZE - 1) +#if defined(__cplusplus) +extern "C" +{ +#endif -#if defined(SWAP_BYTES) -#define bsw_64(p,n) \ - { int _i = (n); while(_i--) ((uint_64t*)p)[_i] = bswap_64(((uint_64t*)p)[_i]); } -#else -#define bsw_64(p,n) +#if CRYPTOPP_BOOL_X64 + void sha256_sse4(void *input_data, uint_32t digest[8], uint_64t num_blks); + void sha256_rorx(void *input_data, uint_32t digest[8], uint_64t num_blks); + void sha256_avx(void *input_data, uint_32t digest[8], uint_64t num_blks); #endif -/* SHA512 mixing function definitions */ +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 + void sha256_compress_nayuki(uint_32t state[8], const uint_8t block[64]); +#endif -#ifdef s_0 -# undef s_0 -# undef s_1 -# undef g_0 -# undef g_1 -# undef k_0 +#if defined(__cplusplus) +} #endif -#define s_0(x) (rotr64((x), 28) ^ rotr64((x), 34) ^ rotr64((x), 39)) -#define s_1(x) (rotr64((x), 14) ^ rotr64((x), 18) ^ rotr64((x), 41)) -#define g_0(x) (rotr64((x), 1) ^ rotr64((x), 8) ^ ((x) >> 7)) -#define g_1(x) (rotr64((x), 19) ^ rotr64((x), 61) ^ ((x) >> 6)) -#define k_0 k512 +#endif -/* SHA384/SHA512 mixing data */ +CRYPTOPP_ALIGN_DATA(16) uint_32t SHA256_K[64] CRYPTOPP_SECTION_ALIGN16 = { + 0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5, 0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5, + 0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3, 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174, + 0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc, 0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da, + 0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7, 0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967, + 0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13, 0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85, + 0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3, 0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070, + 0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5, 0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3, + 0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208, 0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2 + }; + +#if (defined(CRYPTOPP_X86_ASM_AVAILABLE) || defined(CRYPTOPP_X32_ASM_AVAILABLE)) + +#ifdef _MSC_VER +# pragma warning(disable: 4100 4731) +#endif -const uint_64t k512[80] = +static void CRYPTOPP_FASTCALL X86_SHA256_HashBlocks(uint_32t *state, const uint_32t *data, size_t len) { - li_64(428a2f98d728ae22), li_64(7137449123ef65cd), - li_64(b5c0fbcfec4d3b2f), li_64(e9b5dba58189dbbc), - li_64(3956c25bf348b538), li_64(59f111f1b605d019), - li_64(923f82a4af194f9b), li_64(ab1c5ed5da6d8118), - li_64(d807aa98a3030242), li_64(12835b0145706fbe), - li_64(243185be4ee4b28c), li_64(550c7dc3d5ffb4e2), - li_64(72be5d74f27b896f), li_64(80deb1fe3b1696b1), - li_64(9bdc06a725c71235), li_64(c19bf174cf692694), - li_64(e49b69c19ef14ad2), li_64(efbe4786384f25e3), - li_64(0fc19dc68b8cd5b5), li_64(240ca1cc77ac9c65), - li_64(2de92c6f592b0275), li_64(4a7484aa6ea6e483), - li_64(5cb0a9dcbd41fbd4), li_64(76f988da831153b5), - li_64(983e5152ee66dfab), li_64(a831c66d2db43210), - li_64(b00327c898fb213f), li_64(bf597fc7beef0ee4), - li_64(c6e00bf33da88fc2), li_64(d5a79147930aa725), - li_64(06ca6351e003826f), li_64(142929670a0e6e70), - li_64(27b70a8546d22ffc), li_64(2e1b21385c26c926), - li_64(4d2c6dfc5ac42aed), li_64(53380d139d95b3df), - li_64(650a73548baf63de), li_64(766a0abb3c77b2a8), - li_64(81c2c92e47edaee6), li_64(92722c851482353b), - li_64(a2bfe8a14cf10364), li_64(a81a664bbc423001), - li_64(c24b8b70d0f89791), li_64(c76c51a30654be30), - li_64(d192e819d6ef5218), li_64(d69906245565a910), - li_64(f40e35855771202a), li_64(106aa07032bbd1b8), - li_64(19a4c116b8d2d0c8), li_64(1e376c085141ab53), - li_64(2748774cdf8eeb99), li_64(34b0bcb5e19b48a8), - li_64(391c0cb3c5c95a63), li_64(4ed8aa4ae3418acb), - li_64(5b9cca4f7763e373), li_64(682e6ff3d6b2b8a3), - li_64(748f82ee5defb2fc), li_64(78a5636f43172f60), - li_64(84c87814a1f0ab72), li_64(8cc702081a6439ec), - li_64(90befffa23631e28), li_64(a4506cebde82bde9), - li_64(bef9a3f7b2c67915), li_64(c67178f2e372532b), - li_64(ca273eceea26619c), li_64(d186b8c721c0c207), - li_64(eada7dd6cde0eb1e), li_64(f57d4f7fee6ed178), - li_64(06f067aa72176fba), li_64(0a637dc5a2c898a6), - li_64(113f9804bef90dae), li_64(1b710b35131c471b), - li_64(28db77f523047d84), li_64(32caab7b40c72493), - li_64(3c9ebe0a15c9bebc), li_64(431d67c49c100d4c), - li_64(4cc5d4becb3e42b6), li_64(597f299cfc657e2a), - li_64(5fcb6fab3ad6faec), li_64(6c44198c4a475817) -}; + #define LOCALS_SIZE 8*4 + 16*4 + 4*WORD_SZ + #define H(i) [BASE+ASM_MOD(1024+7-(i),8)*4] + #define G(i) H(i+1) + #define F(i) H(i+2) + #define E(i) H(i+3) + #define D(i) H(i+4) + #define C(i) H(i+5) + #define B(i) H(i+6) + #define A(i) H(i+7) + #define Wt(i) BASE+8*4+ASM_MOD(1024+15-(i),16)*4 + #define Wt_2(i) Wt((i)-2) + #define Wt_15(i) Wt((i)-15) + #define Wt_7(i) Wt((i)-7) + #define K_END [BASE+8*4+16*4+0*WORD_SZ] + #define STATE_SAVE [BASE+8*4+16*4+1*WORD_SZ] + #define DATA_SAVE [BASE+8*4+16*4+2*WORD_SZ] + #define DATA_END [BASE+8*4+16*4+3*WORD_SZ] + #define Kt(i) WORD_REG(si)+(i)*4 +#if CRYPTOPP_BOOL_X32 + #define BASE esp+8 +#elif CRYPTOPP_BOOL_X86 + #define BASE esp+4 +#elif defined(__GNUC__) + #define BASE r8 +#else + #define BASE rsp +#endif -/* Compile 128 bytes of hash data into SHA384/512 digest */ -/* NOTE: this routine assumes that the byte order in the */ -/* ctx->wbuf[] at this point is such that low address bytes */ -/* in the ORIGINAL byte stream will go into the high end of */ -/* words on BOTH big and little endian systems */ +#define RA0(i, edx, edi) \ + AS2( add edx, [Kt(i)] )\ + AS2( add edx, [Wt(i)] )\ + AS2( add edx, H(i) )\ + +#define RA1(i, edx, edi) + +#define RB0(i, edx, edi) + +#define RB1(i, edx, edi) \ + AS2( mov AS_REG_7d, [Wt_2(i)] )\ + AS2( mov edi, [Wt_15(i)])\ + AS2( mov ebx, AS_REG_7d )\ + AS2( shr AS_REG_7d, 10 )\ + AS2( ror ebx, 17 )\ + AS2( xor AS_REG_7d, ebx )\ + AS2( ror ebx, 2 )\ + AS2( xor ebx, AS_REG_7d )/* s1(W_t-2) */\ + AS2( add ebx, [Wt_7(i)])\ + AS2( mov AS_REG_7d, edi )\ + AS2( shr AS_REG_7d, 3 )\ + AS2( ror edi, 7 )\ + AS2( add ebx, [Wt(i)])/* s1(W_t-2) + W_t-7 + W_t-16 */\ + AS2( xor AS_REG_7d, edi )\ + AS2( add edx, [Kt(i)])\ + AS2( ror edi, 11 )\ + AS2( add edx, H(i) )\ + AS2( xor AS_REG_7d, edi )/* s0(W_t-15) */\ + AS2( add AS_REG_7d, ebx )/* W_t = s1(W_t-2) + W_t-7 + s0(W_t-15) W_t-16*/\ + AS2( mov [Wt(i)], AS_REG_7d)\ + AS2( add edx, AS_REG_7d )\ + +#define ROUND(i, r, eax, ecx, edi, edx)\ + /* in: edi = E */\ + /* unused: eax, ecx, temp: ebx, AS_REG_7d, out: edx = T1 */\ + AS2( mov edx, F(i) )\ + AS2( xor edx, G(i) )\ + AS2( and edx, edi )\ + AS2( xor edx, G(i) )/* Ch(E,F,G) = (G^(E&(F^G))) */\ + AS2( mov AS_REG_7d, edi )\ + AS2( ror edi, 6 )\ + AS2( ror AS_REG_7d, 25 )\ + RA##r(i, edx, edi )/* H + Wt + Kt + Ch(E,F,G) */\ + AS2( xor AS_REG_7d, edi )\ + AS2( ror edi, 5 )\ + AS2( xor AS_REG_7d, edi )/* S1(E) */\ + AS2( add edx, AS_REG_7d )/* T1 = S1(E) + Ch(E,F,G) + H + Wt + Kt */\ + RB##r(i, edx, edi )/* H + Wt + Kt + Ch(E,F,G) */\ + /* in: ecx = A, eax = B^C, edx = T1 */\ + /* unused: edx, temp: ebx, AS_REG_7d, out: eax = A, ecx = B^C, edx = E */\ + AS2( mov ebx, ecx )\ + AS2( xor ecx, B(i) )/* A^B */\ + AS2( and eax, ecx )\ + AS2( xor eax, B(i) )/* Maj(A,B,C) = B^((A^B)&(B^C) */\ + AS2( mov AS_REG_7d, ebx )\ + AS2( ror ebx, 2 )\ + AS2( add eax, edx )/* T1 + Maj(A,B,C) */\ + AS2( add edx, D(i) )\ + AS2( mov D(i), edx )\ + AS2( ror AS_REG_7d, 22 )\ + AS2( xor AS_REG_7d, ebx )\ + AS2( ror ebx, 11 )\ + AS2( xor AS_REG_7d, ebx )\ + AS2( add eax, AS_REG_7d )/* T1 + S0(A) + Maj(A,B,C) */\ + AS2( mov H(i), eax )\ + +// Unroll the use of CRYPTOPP_BOOL_X64 in assembler math. The GAS assembler on X32 (version 2.25) +// complains "Error: invalid operands (*ABS* and *UND* sections) for `*` and `-`" +#if CRYPTOPP_BOOL_X64 +#define SWAP_COPY(i) \ + AS2( mov WORD_REG(bx), [WORD_REG(dx)+i*WORD_SZ])\ + AS1( bswap WORD_REG(bx))\ + AS2( mov [Wt(i*2+1)], WORD_REG(bx)) +#else // X86 and X32 +#define SWAP_COPY(i) \ + AS2( mov WORD_REG(bx), [WORD_REG(dx)+i*WORD_SZ])\ + AS1( bswap WORD_REG(bx))\ + AS2( mov [Wt(i)], WORD_REG(bx)) +#endif -VOID_RETURN sha512_compile(sha512_ctx ctx[1]) -{ uint_64t v[8], *p = ctx->wbuf; - uint_32t j; -#if defined (TC_WINDOWS_DRIVER) && defined (DEBUG) - uint_32t i; +#if defined(__GNUC__) + #if CRYPTOPP_BOOL_X64 + CRYPTOPP_ALIGN_DATA(16) byte workspace[LOCALS_SIZE] ; + #endif + __asm__ __volatile__ + ( + #if CRYPTOPP_BOOL_X64 + "lea %4, %%r8;" + #endif + INTEL_NOPREFIX #endif - memcpy(v, ctx->hash, 8 * sizeof(uint_64t)); +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 + #ifndef __GNUC__ + AS2( mov edi, [len]) + AS2( lea WORD_REG(si), [SHA256_K+48*4]) + #endif + #if !defined(_MSC_VER) || (_MSC_VER < 1400) + AS_PUSH_IF86(bx) + #endif + + AS_PUSH_IF86(bp) + AS2( mov ebx, esp) + AS2( and esp, -16) + AS2( sub WORD_REG(sp), LOCALS_SIZE) + AS_PUSH_IF86(bx) +#endif + AS2( mov STATE_SAVE, WORD_REG(cx)) + AS2( mov DATA_SAVE, WORD_REG(dx)) + AS2( lea WORD_REG(ax), [WORD_REG(di) + WORD_REG(dx)]) + AS2( mov DATA_END, WORD_REG(ax)) + AS2( mov K_END, WORD_REG(si)) + +#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 + AS2( test edi, 1) + ASJ( jnz, 2, f) + AS1( dec DWORD PTR K_END) +#endif + AS2( movdqa xmm0, XMMWORD_PTR [WORD_REG(cx)+0*16]) + AS2( movdqa xmm1, XMMWORD_PTR [WORD_REG(cx)+1*16]) +#endif - for(j = 0; j < 80; j += 16) - { -#if defined (TC_WINDOWS_DRIVER) && defined (DEBUG) - for (i = 0; i < 16; i++) - { - v_cycle( i, j); - } -#else - v_cycle( 0, j); v_cycle( 1, j); - v_cycle( 2, j); v_cycle( 3, j); - v_cycle( 4, j); v_cycle( 5, j); - v_cycle( 6, j); v_cycle( 7, j); - v_cycle( 8, j); v_cycle( 9, j); - v_cycle(10, j); v_cycle(11, j); - v_cycle(12, j); v_cycle(13, j); - v_cycle(14, j); v_cycle(15, j); -#endif - } - - ctx->hash[0] += v[0]; ctx->hash[1] += v[1]; - ctx->hash[2] += v[2]; ctx->hash[3] += v[3]; - ctx->hash[4] += v[4]; ctx->hash[5] += v[5]; - ctx->hash[6] += v[6]; ctx->hash[7] += v[7]; -} +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 +#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE + ASJ( jmp, 0, f) +#endif + ASL(2) // non-SSE2 + AS2( mov esi, ecx) + AS2( lea edi, A(0)) + AS2( mov ecx, 8) +ATT_NOPREFIX + AS1( rep movsd) +INTEL_NOPREFIX + AS2( mov esi, K_END) + ASJ( jmp, 3, f) +#endif -/* Compile 128 bytes of hash data into SHA256 digest value */ -/* NOTE: this routine assumes that the byte order in the */ -/* ctx->wbuf[] at this point is in such an order that low */ -/* address bytes in the ORIGINAL byte stream placed in this */ -/* buffer will now go to the high end of words on BOTH big */ -/* and little endian systems */ - -VOID_RETURN sha512_hash(const unsigned char data[], unsigned long len, sha512_ctx ctx[1]) -{ uint_32t pos = (uint_32t)(ctx->count[0] & SHA512_MASK), - space = SHA512_BLOCK_SIZE - pos; - const unsigned char *sp = data; - - if((ctx->count[0] += len) < len) - ++(ctx->count[1]); - - while(len >= space) /* tranfer whole blocks while possible */ - { - memcpy(((unsigned char*)ctx->wbuf) + pos, sp, space); - sp += space; len -= space; space = SHA512_BLOCK_SIZE; pos = 0; - bsw_64(ctx->wbuf, SHA512_BLOCK_SIZE >> 3); - sha512_compile(ctx); - } - - memcpy(((unsigned char*)ctx->wbuf) + pos, sp, len); -} +#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE + ASL(0) + AS2( movdqa E(0), xmm1) + AS2( movdqa A(0), xmm0) +#endif +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 + ASL(3) +#endif + AS2( sub WORD_REG(si), 48*4) + SWAP_COPY(0) SWAP_COPY(1) SWAP_COPY(2) SWAP_COPY(3) + SWAP_COPY(4) SWAP_COPY(5) SWAP_COPY(6) SWAP_COPY(7) +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 + SWAP_COPY(8) SWAP_COPY(9) SWAP_COPY(10) SWAP_COPY(11) + SWAP_COPY(12) SWAP_COPY(13) SWAP_COPY(14) SWAP_COPY(15) +#endif + AS2( mov edi, E(0)) // E + AS2( mov eax, B(0)) // B + AS2( xor eax, C(0)) // B^C + AS2( mov ecx, A(0)) // A + + ROUND(0, 0, eax, ecx, edi, edx) + ROUND(1, 0, ecx, eax, edx, edi) + ROUND(2, 0, eax, ecx, edi, edx) + ROUND(3, 0, ecx, eax, edx, edi) + ROUND(4, 0, eax, ecx, edi, edx) + ROUND(5, 0, ecx, eax, edx, edi) + ROUND(6, 0, eax, ecx, edi, edx) + ROUND(7, 0, ecx, eax, edx, edi) + ROUND(8, 0, eax, ecx, edi, edx) + ROUND(9, 0, ecx, eax, edx, edi) + ROUND(10, 0, eax, ecx, edi, edx) + ROUND(11, 0, ecx, eax, edx, edi) + ROUND(12, 0, eax, ecx, edi, edx) + ROUND(13, 0, ecx, eax, edx, edi) + ROUND(14, 0, eax, ecx, edi, edx) + ROUND(15, 0, ecx, eax, edx, edi) + + ASL(1) + AS2(add WORD_REG(si), 4*16) + ROUND(0, 1, eax, ecx, edi, edx) + ROUND(1, 1, ecx, eax, edx, edi) + ROUND(2, 1, eax, ecx, edi, edx) + ROUND(3, 1, ecx, eax, edx, edi) + ROUND(4, 1, eax, ecx, edi, edx) + ROUND(5, 1, ecx, eax, edx, edi) + ROUND(6, 1, eax, ecx, edi, edx) + ROUND(7, 1, ecx, eax, edx, edi) + ROUND(8, 1, eax, ecx, edi, edx) + ROUND(9, 1, ecx, eax, edx, edi) + ROUND(10, 1, eax, ecx, edi, edx) + ROUND(11, 1, ecx, eax, edx, edi) + ROUND(12, 1, eax, ecx, edi, edx) + ROUND(13, 1, ecx, eax, edx, edi) + ROUND(14, 1, eax, ecx, edi, edx) + ROUND(15, 1, ecx, eax, edx, edi) + AS2( cmp WORD_REG(si), K_END) + ATT_NOPREFIX + ASJ( jb, 1, b) + INTEL_NOPREFIX + + AS2( mov WORD_REG(dx), DATA_SAVE) + AS2( add WORD_REG(dx), 64) + AS2( mov AS_REG_7, STATE_SAVE) + AS2( mov DATA_SAVE, WORD_REG(dx)) + +#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 + AS2( test DWORD PTR K_END, 1) + ASJ( jz, 4, f) +#endif + AS2( movdqa xmm1, XMMWORD_PTR [AS_REG_7+1*16]) + AS2( movdqa xmm0, XMMWORD_PTR [AS_REG_7+0*16]) + AS2( paddd xmm1, E(0)) + AS2( paddd xmm0, A(0)) + AS2( movdqa [AS_REG_7+1*16], xmm1) + AS2( movdqa [AS_REG_7+0*16], xmm0) + AS2( cmp WORD_REG(dx), DATA_END) + ATT_NOPREFIX + ASJ( jb, 0, b) + INTEL_NOPREFIX +#endif -/* SHA384/512 Final padding and digest calculation */ - -static void sha_end2(unsigned char hval[], sha512_ctx ctx[1], const unsigned int hlen) -{ uint_32t i = (uint_32t)(ctx->count[0] & SHA512_MASK); - - /* put bytes in the buffer in an order in which references to */ - /* 32-bit words will put bytes with lower addresses into the */ - /* top of 32 bit words on BOTH big and little endian machines */ - bsw_64(ctx->wbuf, (i + 7) >> 3); - - /* we now need to mask valid bytes and add the padding which is */ - /* a single 1 bit and as many zero bits as necessary. Note that */ - /* we can always add the first padding byte here because the */ - /* buffer always has at least one empty slot */ - ctx->wbuf[i >> 3] &= li_64(ffffffffffffff00) << 8 * (~i & 7); - ctx->wbuf[i >> 3] |= li_64(0000000000000080) << 8 * (~i & 7); - - /* we need 17 or more empty byte positions, one for the padding */ - /* byte (above) and sixteen for the length count. If there is */ - /* not enough space pad and empty the buffer */ - if(i > SHA512_BLOCK_SIZE - 17) - { - if(i < 120) ctx->wbuf[15] = 0; - sha512_compile(ctx); - i = 0; - } - else - i = (i >> 3) + 1; - - while(i < 14) - ctx->wbuf[i++] = 0; - - /* the following 64-bit length fields are assembled in the */ - /* wrong byte order on little endian machines but this is */ - /* corrected later since they are only ever used as 64-bit */ - /* word values. */ - ctx->wbuf[14] = (ctx->count[1] << 3) | (ctx->count[0] >> 61); - ctx->wbuf[15] = ctx->count[0] << 3; - sha512_compile(ctx); - - /* extract the hash value as bytes in case the hash buffer is */ - /* misaligned for 32-bit words */ - for(i = 0; i < hlen; ++i) - hval[i] = (unsigned char)(ctx->hash[i >> 3] >> (8 * (~i & 7))); -} +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 +#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE + ASJ( jmp, 5, f) + ASL(4) // non-SSE2 +#endif + AS2( add [AS_REG_7+0*4], ecx) // A + AS2( add [AS_REG_7+4*4], edi) // E + AS2( mov eax, B(0)) + AS2( mov ebx, C(0)) + AS2( mov ecx, D(0)) + AS2( add [AS_REG_7+1*4], eax) + AS2( add [AS_REG_7+2*4], ebx) + AS2( add [AS_REG_7+3*4], ecx) + AS2( mov eax, F(0)) + AS2( mov ebx, G(0)) + AS2( mov ecx, H(0)) + AS2( add [AS_REG_7+5*4], eax) + AS2( add [AS_REG_7+6*4], ebx) + AS2( add [AS_REG_7+7*4], ecx) + AS2( mov ecx, AS_REG_7d) + AS2( cmp WORD_REG(dx), DATA_END) + ASJ( jb, 2, b) +#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE + ASL(5) +#endif +#endif + AS_POP_IF86(sp) + AS_POP_IF86(bp) + #if !defined(_MSC_VER) || (_MSC_VER < 1400) + AS_POP_IF86(bx) + #endif + +#ifdef __GNUC__ + ATT_PREFIX + : + : "c" (state), "d" (data), "S" (SHA256_K+48), "D" (len) + #if CRYPTOPP_BOOL_X64 + , "m" (workspace[0]) + #endif + : "memory", "cc", "%eax" + #if CRYPTOPP_BOOL_X64 + , "%rbx", "%r8", "%r10" + #endif + ); #endif +} -#if defined(SHA_384) +#endif // (defined(CRYPTOPP_X86_ASM_AVAILABLE)) -/* SHA384 initialisation data */ +#undef sum0 +#undef sum1 +#undef sigma0 +#undef sigma1 -const uint_64t i384[80] = -{ - li_64(cbbb9d5dc1059ed8), li_64(629a292a367cd507), - li_64(9159015a3070dd17), li_64(152fecd8f70e5939), - li_64(67332667ffc00b31), li_64(8eb44a8768581511), - li_64(db0c2e0d64f98fa7), li_64(47b5481dbefa4fa4) -}; +#define sum0(x) (rotr32((x), 2) ^ rotr32((x), 13) ^ rotr32((x), 22)) +#define sum1(x) (rotr32((x), 6) ^ rotr32((x), 11) ^ rotr32((x), 25)) +#define sigma0(x) (rotr32((x), 7) ^ rotr32((x), 18) ^ ((x) >> 3)) +#define sigma1(x) (rotr32((x), 17) ^ rotr32((x), 19) ^ ((x) >> 10)) -VOID_RETURN sha384_begin(sha384_ctx ctx[1]) -{ - ctx->count[0] = ctx->count[1] = 0; - memcpy(ctx->hash, i384, 8 * sizeof(uint_64t)); -} -VOID_RETURN sha384_end(unsigned char hval[], sha384_ctx ctx[1]) -{ - sha_end2(hval, ctx, SHA384_DIGEST_SIZE); -} +typedef void (*sha256transformFn)(sha256_ctx* ctx, void* m, uint_64t num_blks); -VOID_RETURN sha384(unsigned char hval[], const unsigned char data[], unsigned long len) -{ sha384_ctx cx[1]; - - sha384_begin(cx); - sha384_hash(data, len, cx); - sha_end2(hval, cx, SHA384_DIGEST_SIZE); -} +sha256transformFn sha256transfunc = NULL; +void StdSha256Transform(sha256_ctx* ctx, void* mp, uint_64t num_blks) +{ + uint_64t blk; + for (blk = 0; blk < num_blks; blk++) + { + uint_32t W[16]; + uint_32t a,b,c,d,e,f,g,h; + uint_32t T1, T2; + int i; +#if defined (TC_WINDOWS_DRIVER) && defined (DEBUG) + int j; #endif -#if defined(SHA_512) - -/* SHA512 initialisation data */ + for (i = 0; i < 64 / 4; i++) + { + W[i] = bswap_32((((const uint_32t*)(mp))[blk * 16 + i])); + } + + a = ctx->hash[0]; + b = ctx->hash[1]; + c = ctx->hash[2]; + d = ctx->hash[3]; + e = ctx->hash[4]; + f = ctx->hash[5]; + g = ctx->hash[6]; + h = ctx->hash[7]; + + for (i = 0; i <= 63; i+=16) + { +#if defined (TC_WINDOWS_DRIVER) && defined (DEBUG) + for (j = 0; j < 16; j++) + { + COMPRESS_ROUND(i, j, SHA256_K); + } +#else + COMPRESS_ROUND(i, 0, SHA256_K); + COMPRESS_ROUND(i, 1, SHA256_K); + COMPRESS_ROUND(i , 2, SHA256_K); + COMPRESS_ROUND(i, 3, SHA256_K); + COMPRESS_ROUND(i, 4, SHA256_K); + COMPRESS_ROUND(i, 5, SHA256_K); + COMPRESS_ROUND(i, 6, SHA256_K); + COMPRESS_ROUND(i, 7, SHA256_K); + COMPRESS_ROUND(i, 8, SHA256_K); + COMPRESS_ROUND(i, 9, SHA256_K); + COMPRESS_ROUND(i, 10, SHA256_K); + COMPRESS_ROUND(i, 11, SHA256_K); + COMPRESS_ROUND(i, 12, SHA256_K); + COMPRESS_ROUND(i, 13, SHA256_K); + COMPRESS_ROUND(i, 14, SHA256_K); + COMPRESS_ROUND(i, 15, SHA256_K); +#endif + } + ctx->hash[0] += a; + ctx->hash[1] += b; + ctx->hash[2] += c; + ctx->hash[3] += d; + ctx->hash[4] += e; + ctx->hash[5] += f; + ctx->hash[6] += g; + ctx->hash[7] += h; + } +} -const uint_64t i512[80] = -{ - li_64(6a09e667f3bcc908), li_64(bb67ae8584caa73b), - li_64(3c6ef372fe94f82b), li_64(a54ff53a5f1d36f1), - li_64(510e527fade682d1), li_64(9b05688c2b3e6c1f), - li_64(1f83d9abfb41bd6b), li_64(5be0cd19137e2179) -}; +#ifndef NO_OPTIMIZED_VERSIONS -VOID_RETURN sha512_begin(sha512_ctx ctx[1]) +#if CRYPTOPP_BOOL_X64 +void Avx2Sha256Transform(sha256_ctx* ctx, void* mp, uint_64t num_blks) { - ctx->count[0] = ctx->count[1] = 0; - memcpy(ctx->hash, i512, 8 * sizeof(uint_64t)); + if (num_blks > 1) + sha256_rorx(mp, ctx->hash, num_blks); + else + sha256_sse4(mp, ctx->hash, num_blks); } -VOID_RETURN sha512_end(unsigned char hval[], sha512_ctx ctx[1]) +void AvxSha256Transform(sha256_ctx* ctx, void* mp, uint_64t num_blks) { - sha_end2(hval, ctx, SHA512_DIGEST_SIZE); + if (num_blks > 1) + sha256_avx(mp, ctx->hash, num_blks); + else + sha256_sse4(mp, ctx->hash, num_blks); } -VOID_RETURN sha512(unsigned char hval[], const unsigned char data[], unsigned long len) -{ sha512_ctx cx[1]; - - sha512_begin(cx); - sha512_hash(data, len, cx); - sha_end2(hval, cx, SHA512_DIGEST_SIZE); +void SSE4Sha256Transform(sha256_ctx* ctx, void* mp, uint_64t num_blks) +{ + sha256_sse4(mp, ctx->hash, num_blks); } #endif -#if defined(SHA_2) - -#define CTX_224(x) ((x)->uu->ctx256) -#define CTX_256(x) ((x)->uu->ctx256) -#define CTX_384(x) ((x)->uu->ctx512) -#define CTX_512(x) ((x)->uu->ctx512) - -/* SHA2 initialisation */ - -INT_RETURN sha2_begin(unsigned long len, sha2_ctx ctx[1]) +#if (defined(CRYPTOPP_X86_ASM_AVAILABLE) || defined(CRYPTOPP_X32_ASM_AVAILABLE)) +void SSE2Sha256Transform(sha256_ctx* ctx, void* mp, uint_64t num_blks) { - switch(len) - { -#if defined(SHA_224) - case 224: - case 28: CTX_256(ctx)->count[0] = CTX_256(ctx)->count[1] = 0; - memcpy(CTX_256(ctx)->hash, i224, 32); - ctx->sha2_len = 28; return EXIT_SUCCESS; -#endif -#if defined(SHA_256) - case 256: - case 32: CTX_256(ctx)->count[0] = CTX_256(ctx)->count[1] = 0; - memcpy(CTX_256(ctx)->hash, i256, 32); - ctx->sha2_len = 32; return EXIT_SUCCESS; -#endif -#if defined(SHA_384) - case 384: - case 48: CTX_384(ctx)->count[0] = CTX_384(ctx)->count[1] = 0; - memcpy(CTX_384(ctx)->hash, i384, 64); - ctx->sha2_len = 48; return EXIT_SUCCESS; -#endif -#if defined(SHA_512) - case 512: - case 64: CTX_512(ctx)->count[0] = CTX_512(ctx)->count[1] = 0; - memcpy(CTX_512(ctx)->hash, i512, 64); - ctx->sha2_len = 64; return EXIT_SUCCESS; -#endif - default: return EXIT_FAILURE; - } + X86_SHA256_HashBlocks(ctx->hash, (const uint_32t*)mp, (size_t)(num_blks * 64)); } +#endif -VOID_RETURN sha2_hash(const unsigned char data[], unsigned long len, sha2_ctx ctx[1]) +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 +void Sha256AsmTransform(sha256_ctx* ctx, void* mp, uint_64t num_blks) { - switch(ctx->sha2_len) - { -#if defined(SHA_224) - case 28: sha224_hash(data, len, CTX_224(ctx)); return; -#endif -#if defined(SHA_256) - case 32: sha256_hash(data, len, CTX_256(ctx)); return; -#endif -#if defined(SHA_384) - case 48: sha384_hash(data, len, CTX_384(ctx)); return; + uint_64t i; + for (i = 0; i < num_blks; i++) + sha256_compress_nayuki(ctx->hash, (uint_8t*)mp + i * 64); +} #endif -#if defined(SHA_512) - case 64: sha512_hash(data, len, CTX_512(ctx)); return; + #endif - } -} -VOID_RETURN sha2_end(unsigned char hval[], sha2_ctx ctx[1]) +void sha256_begin(sha256_ctx* ctx) { - switch(ctx->sha2_len) - { -#if defined(SHA_224) - case 28: sha_end1(hval, CTX_224(ctx), SHA224_DIGEST_SIZE); return; + ctx->hash[0] = 0x6a09e667; + ctx->hash[1] = 0xbb67ae85; + ctx->hash[2] = 0x3c6ef372; + ctx->hash[3] = 0xa54ff53a; + ctx->hash[4] = 0x510e527f; + ctx->hash[5] = 0x9b05688c; + ctx->hash[6] = 0x1f83d9ab; + ctx->hash[7] = 0x5be0cd19; + ctx->count[0] = 0; + ctx->count[1] = 0; + + if (!sha256transfunc) + { +#ifndef NO_OPTIMIZED_VERSIONS +#ifdef _M_X64 + if (g_isIntel && HasSAVX2() && HasSBMI2()) + sha256transfunc = Avx2Sha256Transform; + else if (g_isIntel && HasSAVX()) + sha256transfunc = AvxSha256Transform; + else if (HasSSE41()) + sha256transfunc = SSE4Sha256Transform; + else #endif -#if defined(SHA_256) - case 32: sha_end1(hval, CTX_256(ctx), SHA256_DIGEST_SIZE); return; + +#if (defined(CRYPTOPP_X86_ASM_AVAILABLE) || defined(CRYPTOPP_X32_ASM_AVAILABLE)) + if (HasSSE2 ()) + sha256transfunc = SSE2Sha256Transform; + else #endif -#if defined(SHA_384) - case 48: sha_end2(hval, CTX_384(ctx), SHA384_DIGEST_SIZE); return; + +#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X32 + sha256transfunc = Sha256AsmTransform; +#else + sha256transfunc = StdSha256Transform; #endif -#if defined(SHA_512) - case 64: sha_end2(hval, CTX_512(ctx), SHA512_DIGEST_SIZE); return; +#else + sha256transfunc = StdSha256Transform; #endif - } + } } -INT_RETURN sha2(unsigned char hval[], unsigned long size, - const unsigned char data[], unsigned long len) -{ sha2_ctx cx[1]; +void sha256_end(unsigned char * result, sha256_ctx* ctx) +{ + int i; + uint_64t mlen, pos = ctx->count[0]; + uint_8t* m = (uint_8t*) ctx->wbuf; + m[pos++] = 0x80; + if (pos > 56) + { + memset(m + pos, 0, (size_t) (64 - pos)); + sha256transfunc(ctx, m, 1); + pos = 0; + } + memset(m + pos, 0, (size_t) (56 - pos)); + mlen = bswap_64((uint_64t) ctx->count[1]); + memcpy(m + (64 - 8), &mlen, 64 / 8); + sha256transfunc(ctx, m, 1); + for (i = 0; i < 8; i++) + { + ctx->hash[i] = bswap_32(ctx->hash[i]); + } + memcpy(result, ctx->hash, 32); +} - if(sha2_begin(size, cx) == EXIT_SUCCESS) - { - sha2_hash(data, len, cx); sha2_end(hval, cx); return EXIT_SUCCESS; - } - else - return EXIT_FAILURE; +void sha256_hash(const unsigned char * data, uint_32t len, sha256_ctx *ctx) +{ + uint_32t pos = ctx->count[0]; + uint_32t total = ctx->count[1]; + uint_8t* m = (uint_8t*) ctx->wbuf; + if (pos && pos + len >= 64) + { + memcpy(m + pos, data, 64 - pos); + sha256transfunc(ctx, m, 1); + len -= 64 - pos; + total += (64 - pos) * 8; + data += 64 - pos; + pos = 0; + } + if (len >= 64) + { + uint_32t blocks = len / 64; + uint_32t bytes = blocks * 64; + sha256transfunc(ctx, (void*)data, blocks); + len -= bytes; + total += (bytes)* 8; + data += bytes; + } + memcpy(m+pos, data, len); + pos += len; + total += len * 8; + ctx->count[0] = pos; + ctx->count[1] = total; } -#endif +void sha256(unsigned char * result, const unsigned char* source, uint_32t sourceLen) +{ + sha256_ctx ctx; -#if defined(__cplusplus) + sha256_begin(&ctx); + sha256_hash(source, sourceLen, &ctx); + sha256_end(result, &ctx); } -#endif -- cgit v1.2.3