freebsd-skq/sys/crypto/sha2/sha256c.c
cperciva b67598a549 Retune SHA2 code for improved performance on CPUs with more ILP and
a preference for memory load instructions over large code footprints
with embedded immediate variables.

On amd64 CPUs from 2007-2008 there is not a significant change, but
amd64 CPUs from 2009-2010 get roughly 10% more throughput with this
code; amd64 CPUs from 2011-2012 get roughly 15% more throughput; and
AMD64 CPUs from 2013-2015 get 20-25% more throughput.  The Raspberry
Pi 2 increases its throughput by 6-8%.

Sponsored by:	Tarsnap Backup Inc.
Performance tested by:	allanjude
MFC after:	3 weeks
2016-05-29 17:26:40 +00:00

317 lines
8.5 KiB
C

/*-
* Copyright 2005 Colin Percival
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/endian.h>
#include <sys/types.h>
#ifdef _KERNEL
#include <sys/systm.h>
#else
#include <string.h>
#endif
#include "sha256.h"
#if BYTE_ORDER == BIG_ENDIAN
/* Copy a vector of big-endian uint32_t into a vector of bytes */
#define be32enc_vect(dst, src, len) \
memcpy((void *)dst, (const void *)src, (size_t)len)
/* Copy a vector of bytes into a vector of big-endian uint32_t */
#define be32dec_vect(dst, src, len) \
memcpy((void *)dst, (const void *)src, (size_t)len)
#else /* BYTE_ORDER != BIG_ENDIAN */
/*
* Encode a length len/4 vector of (uint32_t) into a length len vector of
* (unsigned char) in big-endian form. Assumes len is a multiple of 4.
*/
static void
be32enc_vect(unsigned char *dst, const uint32_t *src, size_t len)
{
size_t i;
for (i = 0; i < len / 4; i++)
be32enc(dst + i * 4, src[i]);
}
/*
* Decode a big-endian length len vector of (unsigned char) into a length
* len/4 vector of (uint32_t). Assumes len is a multiple of 4.
*/
static void
be32dec_vect(uint32_t *dst, const unsigned char *src, size_t len)
{
size_t i;
for (i = 0; i < len / 4; i++)
dst[i] = be32dec(src + i * 4);
}
#endif /* BYTE_ORDER != BIG_ENDIAN */
/* SHA256 round constants. */
static const uint32_t K[64] = {
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
};
/* Elementary functions used by SHA256 */
#define Ch(x, y, z) ((x & (y ^ z)) ^ z)
#define Maj(x, y, z) ((x & (y | z)) | (y & z))
#define SHR(x, n) (x >> n)
#define ROTR(x, n) ((x >> n) | (x << (32 - n)))
#define S0(x) (ROTR(x, 2) ^ ROTR(x, 13) ^ ROTR(x, 22))
#define S1(x) (ROTR(x, 6) ^ ROTR(x, 11) ^ ROTR(x, 25))
#define s0(x) (ROTR(x, 7) ^ ROTR(x, 18) ^ SHR(x, 3))
#define s1(x) (ROTR(x, 17) ^ ROTR(x, 19) ^ SHR(x, 10))
/* SHA256 round function */
#define RND(a, b, c, d, e, f, g, h, k) \
h += S1(e) + Ch(e, f, g) + k; \
d += h; \
h += S0(a) + Maj(a, b, c);
/* Adjusted round function for rotating state */
#define RNDr(S, W, i, ii) \
RND(S[(64 - i) % 8], S[(65 - i) % 8], \
S[(66 - i) % 8], S[(67 - i) % 8], \
S[(68 - i) % 8], S[(69 - i) % 8], \
S[(70 - i) % 8], S[(71 - i) % 8], \
W[i + ii] + K[i + ii])
/* Message schedule computation */
#define MSCH(W, ii, i) \
W[i + ii + 16] = s1(W[i + ii + 14]) + W[i + ii + 9] + s0(W[i + ii + 1]) + W[i + ii]
/*
* SHA256 block compression function. The 256-bit state is transformed via
* the 512-bit input block to produce a new state.
*/
static void
SHA256_Transform(uint32_t * state, const unsigned char block[64])
{
uint32_t W[64];
uint32_t S[8];
int i;
/* 1. Prepare the first part of the message schedule W. */
be32dec_vect(W, block, 64);
/* 2. Initialize working variables. */
memcpy(S, state, 32);
/* 3. Mix. */
for (i = 0; i < 64; i += 16) {
RNDr(S, W, 0, i);
RNDr(S, W, 1, i);
RNDr(S, W, 2, i);
RNDr(S, W, 3, i);
RNDr(S, W, 4, i);
RNDr(S, W, 5, i);
RNDr(S, W, 6, i);
RNDr(S, W, 7, i);
RNDr(S, W, 8, i);
RNDr(S, W, 9, i);
RNDr(S, W, 10, i);
RNDr(S, W, 11, i);
RNDr(S, W, 12, i);
RNDr(S, W, 13, i);
RNDr(S, W, 14, i);
RNDr(S, W, 15, i);
if (i == 48)
break;
MSCH(W, 0, i);
MSCH(W, 1, i);
MSCH(W, 2, i);
MSCH(W, 3, i);
MSCH(W, 4, i);
MSCH(W, 5, i);
MSCH(W, 6, i);
MSCH(W, 7, i);
MSCH(W, 8, i);
MSCH(W, 9, i);
MSCH(W, 10, i);
MSCH(W, 11, i);
MSCH(W, 12, i);
MSCH(W, 13, i);
MSCH(W, 14, i);
MSCH(W, 15, i);
}
/* 4. Mix local working variables into global state */
for (i = 0; i < 8; i++)
state[i] += S[i];
}
static unsigned char PAD[64] = {
0x80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
/* Add padding and terminating bit-count. */
static void
SHA256_Pad(SHA256_CTX * ctx)
{
size_t r;
/* Figure out how many bytes we have buffered. */
r = (ctx->count >> 3) & 0x3f;
/* Pad to 56 mod 64, transforming if we finish a block en route. */
if (r < 56) {
/* Pad to 56 mod 64. */
memcpy(&ctx->buf[r], PAD, 56 - r);
} else {
/* Finish the current block and mix. */
memcpy(&ctx->buf[r], PAD, 64 - r);
SHA256_Transform(ctx->state, ctx->buf);
/* The start of the final block is all zeroes. */
memset(&ctx->buf[0], 0, 56);
}
/* Add the terminating bit-count. */
be64enc(&ctx->buf[56], ctx->count);
/* Mix in the final block. */
SHA256_Transform(ctx->state, ctx->buf);
}
/* SHA-256 initialization. Begins a SHA-256 operation. */
void
SHA256_Init(SHA256_CTX * ctx)
{
/* Zero bits processed so far */
ctx->count = 0;
/* Magic initialization constants */
ctx->state[0] = 0x6A09E667;
ctx->state[1] = 0xBB67AE85;
ctx->state[2] = 0x3C6EF372;
ctx->state[3] = 0xA54FF53A;
ctx->state[4] = 0x510E527F;
ctx->state[5] = 0x9B05688C;
ctx->state[6] = 0x1F83D9AB;
ctx->state[7] = 0x5BE0CD19;
}
/* Add bytes into the hash */
void
SHA256_Update(SHA256_CTX * ctx, const void *in, size_t len)
{
uint64_t bitlen;
uint32_t r;
const unsigned char *src = in;
/* Number of bytes left in the buffer from previous updates */
r = (ctx->count >> 3) & 0x3f;
/* Convert the length into a number of bits */
bitlen = len << 3;
/* Update number of bits */
ctx->count += bitlen;
/* Handle the case where we don't need to perform any transforms */
if (len < 64 - r) {
memcpy(&ctx->buf[r], src, len);
return;
}
/* Finish the current block */
memcpy(&ctx->buf[r], src, 64 - r);
SHA256_Transform(ctx->state, ctx->buf);
src += 64 - r;
len -= 64 - r;
/* Perform complete blocks */
while (len >= 64) {
SHA256_Transform(ctx->state, src);
src += 64;
len -= 64;
}
/* Copy left over data into buffer */
memcpy(ctx->buf, src, len);
}
/*
* SHA-256 finalization. Pads the input data, exports the hash value,
* and clears the context state.
*/
void
SHA256_Final(unsigned char digest[static SHA256_DIGEST_LENGTH], SHA256_CTX *ctx)
{
/* Add padding */
SHA256_Pad(ctx);
/* Write the hash */
be32enc_vect(digest, ctx->state, SHA256_DIGEST_LENGTH);
/* Clear the context state */
memset(ctx, 0, sizeof(*ctx));
}
#ifdef WEAK_REFS
/* When building libmd, provide weak references. Note: this is not
activated in the context of compiling these sources for internal
use in libcrypt.
*/
#undef SHA256_Init
__weak_reference(_libmd_SHA256_Init, SHA256_Init);
#undef SHA256_Update
__weak_reference(_libmd_SHA256_Update, SHA256_Update);
#undef SHA256_Final
__weak_reference(_libmd_SHA256_Final, SHA256_Final);
#undef SHA256_Transform
__weak_reference(_libmd_SHA256_Transform, SHA256_Transform);
#endif