Print this page
6665607 Need a SHA256/SHA384/SHA512 implementation optimized for 64-bit x86
@@ -1,13 +1,12 @@
/*
- * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
-#pragma ident "@(#)sha2.c 1.7 07/04/10 SMI"
+#pragma ident "@(#)sha2.c 1.8 08/03/20 SMI"
-
/*
* The basic framework for this code came from the reference
* implementation for MD5. That implementation is Copyright (C)
* 1991-2, RSA Data Security, Inc. Created 1991. All rights reserved.
*
@@ -41,12 +40,14 @@
#include <sys/sysmacros.h>
#define _SHA2_IMPL
#include <sys/sha2.h>
#include <sys/sha2_consts.h>
-#ifndef _KERNEL
+#ifdef _KERNEL
+#include <sys/cmn_err.h>
+#else
#include <strings.h>
#include <stdlib.h>
#include <errno.h>
#pragma weak SHA256Update = SHA2Update
@@ -55,20 +56,26 @@
#pragma weak SHA256Final = SHA2Final
#pragma weak SHA384Final = SHA2Final
#pragma weak SHA512Final = SHA2Final
-#endif /* !_KERNEL */
-
-#ifdef _KERNEL
-#include <sys/cmn_err.h>
#endif /* _KERNEL */
static void Encode(uint8_t *, uint32_t *, size_t);
static void Encode64(uint8_t *, uint64_t *, size_t);
+
+#if defined(__amd64)
+#define SHA512Transform(ctx, in) SHA512TransformBlocks((ctx), (in), 1)
+#define SHA256Transform(ctx, in) SHA256TransformBlocks((ctx), (in), 1)
+
+void SHA512TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num);
+void SHA256TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num);
+
+#else
static void SHA256Transform(SHA2_CTX *, const uint8_t *);
static void SHA512Transform(SHA2_CTX *, const uint8_t *);
+#endif /* __amd64 */
static uint8_t PADDING[128] = { 0x80, /* all zeros */ };
/* Ch and Maj are the basic SHA2 functions. */
#define Ch(b, c, d) (((b) & (c)) ^ ((~b) & (d)))
@@ -133,20 +140,19 @@
#define LOAD_BIG_64(addr) \
(((uint64_t)(addr)[0] << 56) | ((uint64_t)(addr)[1] << 48) | \
((uint64_t)(addr)[2] << 40) | ((uint64_t)(addr)[3] << 32) | \
((uint64_t)(addr)[4] << 24) | ((uint64_t)(addr)[5] << 16) | \
((uint64_t)(addr)[6] << 8) | (uint64_t)(addr)[7])
-
#endif
+#if !defined(__amd64)
/* SHA256 Transform */
static void
SHA256Transform(SHA2_CTX *ctx, const uint8_t *blk)
{
-
uint32_t a = ctx->state.s32[0];
uint32_t b = ctx->state.s32[1];
uint32_t c = ctx->state.s32[2];
uint32_t d = ctx->state.s32[3];
uint32_t e = ctx->state.s32[4];
@@ -181,11 +187,11 @@
SHA256_CONST_54, SHA256_CONST_55, SHA256_CONST_56,
SHA256_CONST_57, SHA256_CONST_58, SHA256_CONST_59,
SHA256_CONST_60, SHA256_CONST_61, SHA256_CONST_62,
SHA256_CONST_63
};
-#endif
+#endif /* __sparc */
if ((uintptr_t)blk & 0x3) { /* not 4-byte aligned? */
bcopy(blk, ctx->buf_un.buf32, sizeof (ctx->buf_un.buf32));
blk = (uint8_t *)ctx->buf_un.buf32;
}
@@ -396,11 +402,11 @@
SHA512_CONST_69, SHA512_CONST_70, SHA512_CONST_71,
SHA512_CONST_72, SHA512_CONST_73, SHA512_CONST_74,
SHA512_CONST_75, SHA512_CONST_76, SHA512_CONST_77,
SHA512_CONST_78, SHA512_CONST_79
};
-#endif
+#endif /* __sparc */
if ((uintptr_t)blk & 0x7) { /* not 8-byte aligned? */
bcopy(blk, ctx->buf_un.buf64, sizeof (ctx->buf_un.buf64));
blk = (uint8_t *)ctx->buf_un.buf64;
@@ -595,10 +601,11 @@
ctx->state.s64[5] += f;
ctx->state.s64[6] += g;
ctx->state.s64[7] += h;
}
+#endif /* !__amd64 */
/*
* Encode()
*
@@ -747,25 +754,30 @@
*
* purpose: continues an sha2 digest operation, using the message block
* to update the context.
* input: SHA2_CTX * : the context to update
* void * : the message block
- * size_t : the length of the message block in bytes
+ * size_t : the length of the message block, in bytes
* output: void
*/
void
SHA2Update(SHA2_CTX *ctx, const void *inptr, size_t input_len)
{
uint32_t i, buf_index, buf_len, buf_limit;
const uint8_t *input = inptr;
+ uint32_t algotype = ctx->algotype;
+#if defined(__amd64)
+ uint32_t block_count;
+#endif /* !__amd64 */
+
/* check for noop */
if (input_len == 0)
return;
- if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
+ if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
buf_limit = 64;
/* compute number of bytes mod 64 */
buf_index = (ctx->count.c32[1] >> 3) & 0x3F;
@@ -803,32 +815,53 @@
* wasn't any data left over from a previous call to
* SHA2Update().
*/
if (buf_index) {
bcopy(input, &ctx->buf_un.buf8[buf_index], buf_len);
- if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
+ if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
SHA256Transform(ctx, ctx->buf_un.buf8);
else
SHA512Transform(ctx, ctx->buf_un.buf8);
i = buf_len;
}
-
+#if !defined(__amd64)
+ if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
for (; i + buf_limit - 1 < input_len; i += buf_limit) {
- if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
SHA256Transform(ctx, &input[i]);
- else
+ }
+ } else {
+ for (; i + buf_limit - 1 < input_len; i += buf_limit) {
SHA512Transform(ctx, &input[i]);
}
+ }
+#else
+ if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
+ block_count = (input_len - i) >> 6;
+ if (block_count > 0) {
+ SHA256TransformBlocks(ctx, &input[i],
+ block_count);
+ i += block_count << 6;
+ }
+ } else {
+ block_count = (input_len - i) >> 7;
+ if (block_count > 0) {
+ SHA512TransformBlocks(ctx, &input[i],
+ block_count);
+ i += block_count << 7;
+ }
+ }
+#endif /* !__amd64 */
+
/*
* general optimization:
*
* if i and input_len are the same, return now instead
* of calling bcopy(), since the bcopy() in this case
- * will be an expensive nop.
+ * will be an expensive noop.
*/
if (input_len == i)
return;
@@ -843,11 +876,11 @@
/*
* SHA2Final()
*
* purpose: ends an sha2 digest operation, finalizing the message digest and
* zeroing the context.
- * input: uchar_t * : a buffer to store the digest in
+ * input: uchar_t * : a buffer to store the digest
* : The function actually uses void* because many
* : callers pass things other than uchar_t here.
* SHA2_CTX * : the context to finalize, save, and zero
* output: void
*/
@@ -856,13 +889,13 @@
SHA2Final(void *digest, SHA2_CTX *ctx)
{
uint8_t bitcount_be[sizeof (ctx->count.c32)];
uint8_t bitcount_be64[sizeof (ctx->count.c64)];
uint32_t index;
+ uint32_t algotype = ctx->algotype;
-
- if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
+ if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
index = (ctx->count.c32[1] >> 3) & 0x3f;
Encode(bitcount_be, ctx->count.c32, sizeof (bitcount_be));
SHA2Update(ctx, PADDING, ((index < 56) ? 56 : 120) - index);
SHA2Update(ctx, bitcount_be, sizeof (bitcount_be));
Encode(digest, ctx->state.s32, sizeof (ctx->state.s32));
@@ -871,11 +904,11 @@
index = (ctx->count.c64[1] >> 3) & 0x7f;
Encode64(bitcount_be64, ctx->count.c64,
sizeof (bitcount_be64));
SHA2Update(ctx, PADDING, ((index < 112) ? 112 : 240) - index);
SHA2Update(ctx, bitcount_be64, sizeof (bitcount_be64));
- if (ctx->algotype <= SHA384_HMAC_GEN_MECH_INFO_TYPE) {
+ if (algotype <= SHA384_HMAC_GEN_MECH_INFO_TYPE) {
ctx->state.s64[6] = ctx->state.s64[7] = 0;
Encode64(digest, ctx->state.s64,
sizeof (uint64_t) * 6);
} else
Encode64(digest, ctx->state.s64,