switch python and oops hashing to sha1

Update sha1 code to a smaller version.
Verified that ccpp hashing produces the same results as the old code.

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>

5 files changed, 128 insertions, 607 deletions

diff --git a/src/lib/Makefile.am b/src/lib/Makefile.am
index 36fe7b4b..d2dc9cae 100644
--- a/src/lib/Makefile.am
+++ b/src/lib/Makefile.am
@@ -15,7 +15,6 @@ libreport_la_SOURCES = \
     encbase64.c \
     binhex.c \
     stdio_helpers.c \
-    hash_md5.c hash_md5.h \
     hash_sha1.c hash_sha1.h \
     read_write.c read_write.h \
     logging.c logging.h \
diff --git a/src/lib/hash_md5.c b/src/lib/hash_md5.c
deleted file mode 100644
index 8bfddd54..00000000
--- a/src/lib/hash_md5.c
+++ /dev/null
@@ -1,444 +0,0 @@
-/* vi: set sw=4 ts=4: */
-/*
- *  md5.c - Compute MD5 checksum of strings according to the
- *          definition of MD5 in RFC 1321 from April 1992.
- *
- *  Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995.
- *
- *  Copyright (C) 1995-1999 Free Software Foundation, Inc.
- *  Copyright (C) 2001 Manuel Novoa III
- *  Copyright (C) 2003 Glenn L. McGrath
- *  Copyright (C) 2003 Erik Andersen
- *
- *  Licensed under the GPL v2 or later, see the file LICENSE in this tarball.
- */
-#include "abrtlib.h"
-#include <byteswap.h>
-
-#if defined(__BIG_ENDIAN__) && __BIG_ENDIAN__
-# define MD5_BIG_ENDIAN 1
-# define MD5_LITTLE_ENDIAN 0
-#elif __BYTE_ORDER == __BIG_ENDIAN
-# define MD5_BIG_ENDIAN 1
-# define MD5_LITTLE_ENDIAN 0
-#elif __BYTE_ORDER == __LITTLE_ENDIAN
-# define MD5_BIG_ENDIAN 0
-# define MD5_LITTLE_ENDIAN 1
-#else
-# error "Can't determine endianness"
-#endif
-
-/* SWAP_LEnn means "convert CPU<->little_endian if needed (by swapping bytes)" */
-#if MD5_BIG_ENDIAN
-# define SWAP_BE32(x) (x)
-# define SWAP_LE32(x) bswap_32(x)
-#else
-# define SWAP_BE32(x) bswap_32(x)
-# define SWAP_LE32(x) (x)
-#endif
-
-
-/* 0: fastest, 3: smallest */
-#define MD5_SIZE_VS_SPEED 3
-
-/* Initialize structure containing state of computation.
- * (RFC 1321, 3.3: Step 3)
- */
-void md5_begin(md5_ctx_t *ctx)
-{
-	ctx->A = 0x67452301;
-	ctx->B = 0xefcdab89;
-	ctx->C = 0x98badcfe;
-	ctx->D = 0x10325476;
-	ctx->total = 0;
-	ctx->buflen = 0;
-}
-
-/* These are the four functions used in the four steps of the MD5 algorithm
- * and defined in the RFC 1321.  The first function is a little bit optimized
- * (as found in Colin Plumbs public domain implementation).
- * #define FF(b, c, d) ((b & c) | (~b & d))
- */
-#define FF(b, c, d) (d ^ (b & (c ^ d)))
-#define FG(b, c, d) FF(d, b, c)
-#define FH(b, c, d) (b ^ c ^ d)
-#define FI(b, c, d) (c ^ (b | ~d))
-
-#define rotl32(w, s) (((w) << (s)) | ((w) >> (32 - (s))))
-
-/* Hash a single block, 64 bytes long and 4-byte aligned. */
-static void md5_hash_block(const void *buffer, md5_ctx_t *ctx)
-{
-	uint32_t correct_words[16];
-	const uint32_t *words = (const uint32_t *)buffer;
-
-#if MD5_SIZE_VS_SPEED > 0
-	static const uint32_t C_array[] = {
-		/* round 1 */
-		0xd76aa478, 0xe8c7b756, 0x242070db, 0xc1bdceee,
-		0xf57c0faf, 0x4787c62a, 0xa8304613, 0xfd469501,
-		0x698098d8, 0x8b44f7af, 0xffff5bb1, 0x895cd7be,
-		0x6b901122, 0xfd987193, 0xa679438e, 0x49b40821,
-		/* round 2 */
-		0xf61e2562, 0xc040b340, 0x265e5a51, 0xe9b6c7aa,
-		0xd62f105d, 0x2441453, 0xd8a1e681, 0xe7d3fbc8,
-		0x21e1cde6, 0xc33707d6, 0xf4d50d87, 0x455a14ed,
-		0xa9e3e905, 0xfcefa3f8, 0x676f02d9, 0x8d2a4c8a,
-		/* round 3 */
-		0xfffa3942, 0x8771f681, 0x6d9d6122, 0xfde5380c,
-		0xa4beea44, 0x4bdecfa9, 0xf6bb4b60, 0xbebfbc70,
-		0x289b7ec6, 0xeaa127fa, 0xd4ef3085, 0x4881d05,
-		0xd9d4d039, 0xe6db99e5, 0x1fa27cf8, 0xc4ac5665,
-		/* round 4 */
-		0xf4292244, 0x432aff97, 0xab9423a7, 0xfc93a039,
-		0x655b59c3, 0x8f0ccc92, 0xffeff47d, 0x85845dd1,
-		0x6fa87e4f, 0xfe2ce6e0, 0xa3014314, 0x4e0811a1,
-		0xf7537e82, 0xbd3af235, 0x2ad7d2bb, 0xeb86d391
-	};
-	static const char P_array[] = {
-# if MD5_SIZE_VS_SPEED > 1
-		0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,	/* 1 */
-# endif
-		1, 6, 11, 0, 5, 10, 15, 4, 9, 14, 3, 8, 13, 2, 7, 12,	/* 2 */
-		5, 8, 11, 14, 1, 4, 7, 10, 13, 0, 3, 6, 9, 12, 15, 2,	/* 3 */
-		0, 7, 14, 5, 12, 3, 10, 1, 8, 15, 6, 13, 4, 11, 2, 9	/* 4 */
-	};
-# if MD5_SIZE_VS_SPEED > 1
-	static const char S_array[] = {
-		7, 12, 17, 22,
-		5, 9, 14, 20,
-		4, 11, 16, 23,
-		6, 10, 15, 21
-	};
-# endif	/* MD5_SIZE_VS_SPEED > 1 */
-#endif
-	uint32_t A = ctx->A;
-	uint32_t B = ctx->B;
-	uint32_t C = ctx->C;
-	uint32_t D = ctx->D;
-
-	/* Process all bytes in the buffer with 64 bytes in each round of
-	   the loop.  */
-	uint32_t *cwp = correct_words;
-	uint32_t A_save = A;
-	uint32_t B_save = B;
-	uint32_t C_save = C;
-	uint32_t D_save = D;
-
-#if MD5_SIZE_VS_SPEED > 1
-	const uint32_t *pc;
-	const char *pp;
-	const char *ps;
-	int i;
-	uint32_t temp;
-
-	for (i = 0; i < 16; i++) {
-		cwp[i] = SWAP_LE32(words[i]);
-	}
-	words += 16;
-
-# if MD5_SIZE_VS_SPEED > 2
-	pc = C_array;
-	pp = P_array;
-	ps = S_array - 4;
-
-	for (i = 0; i < 64; i++) {
-		if ((i & 0x0f) == 0)
-			ps += 4;
-		temp = A;
-		switch (i >> 4) {
-		case 0:
-			temp += FF(B, C, D);
-			break;
-		case 1:
-			temp += FG(B, C, D);
-			break;
-		case 2:
-			temp += FH(B, C, D);
-			break;
-		case 3:
-			temp += FI(B, C, D);
-		}
-		temp += cwp[(int) (*pp++)] + *pc++;
-		temp = rotl32(temp, ps[i & 3]);
-		temp += B;
-		A = D;
-		D = C;
-		C = B;
-		B = temp;
-	}
-# else
-	pc = C_array;
-	pp = P_array;
-	ps = S_array;
-
-	for (i = 0; i < 16; i++) {
-		temp = A + FF(B, C, D) + cwp[(int) (*pp++)] + *pc++;
-		temp = rotl32(temp, ps[i & 3]);
-		temp += B;
-		A = D;
-		D = C;
-		C = B;
-		B = temp;
-	}
-	ps += 4;
-	for (i = 0; i < 16; i++) {
-		temp = A + FG(B, C, D) + cwp[(int) (*pp++)] + *pc++;
-		temp = rotl32(temp, ps[i & 3]);
-		temp += B;
-		A = D;
-		D = C;
-		C = B;
-		B = temp;
-	}
-	ps += 4;
-	for (i = 0; i < 16; i++) {
-		temp = A + FH(B, C, D) + cwp[(int) (*pp++)] + *pc++;
-		temp = rotl32(temp, ps[i & 3]);
-		temp += B;
-		A = D;
-		D = C;
-		C = B;
-		B = temp;
-	}
-	ps += 4;
-	for (i = 0; i < 16; i++) {
-		temp = A + FI(B, C, D) + cwp[(int) (*pp++)] + *pc++;
-		temp = rotl32(temp, ps[i & 3]);
-		temp += B;
-		A = D;
-		D = C;
-		C = B;
-		B = temp;
-	}
-
-# endif /* MD5_SIZE_VS_SPEED > 2 */
-#else
-	/* First round: using the given function, the context and a constant
-	   the next context is computed.  Because the algorithms processing
-	   unit is a 32-bit word and it is determined to work on words in
-	   little endian byte order we perhaps have to change the byte order
-	   before the computation.  To reduce the work for the next steps
-	   we store the swapped words in the array CORRECT_WORDS.  */
-# define OP(a, b, c, d, s, T) \
-	do { \
-		a += FF(b, c, d) + (*cwp++ = SWAP_LE32(*words)) + T; \
-		++words; \
-		a = rotl32(a, s); \
-		a += b; \
-	} while (0)
-
-	/* Before we start, one word to the strange constants.
-	   They are defined in RFC 1321 as
-	   T[i] = (int)(4294967296.0 * fabs(sin(i))), i=1..64
-	 */
-
-# if MD5_SIZE_VS_SPEED == 1
-	const uint32_t *pc;
-	const char *pp;
-	int i;
-# endif	/* MD5_SIZE_VS_SPEED */
-
-	/* Round 1.  */
-# if MD5_SIZE_VS_SPEED == 1
-	pc = C_array;
-	for (i = 0; i < 4; i++) {
-		OP(A, B, C, D, 7, *pc++);
-		OP(D, A, B, C, 12, *pc++);
-		OP(C, D, A, B, 17, *pc++);
-		OP(B, C, D, A, 22, *pc++);
-	}
-# else
-	OP(A, B, C, D, 7, 0xd76aa478);
-	OP(D, A, B, C, 12, 0xe8c7b756);
-	OP(C, D, A, B, 17, 0x242070db);
-	OP(B, C, D, A, 22, 0xc1bdceee);
-	OP(A, B, C, D, 7, 0xf57c0faf);
-	OP(D, A, B, C, 12, 0x4787c62a);
-	OP(C, D, A, B, 17, 0xa8304613);
-	OP(B, C, D, A, 22, 0xfd469501);
-	OP(A, B, C, D, 7, 0x698098d8);
-	OP(D, A, B, C, 12, 0x8b44f7af);
-	OP(C, D, A, B, 17, 0xffff5bb1);
-	OP(B, C, D, A, 22, 0x895cd7be);
-	OP(A, B, C, D, 7, 0x6b901122);
-	OP(D, A, B, C, 12, 0xfd987193);
-	OP(C, D, A, B, 17, 0xa679438e);
-	OP(B, C, D, A, 22, 0x49b40821);
-# endif /* MD5_SIZE_VS_SPEED == 1 */
-
-	/* For the second to fourth round we have the possibly swapped words
-	   in CORRECT_WORDS.  Redefine the macro to take an additional first
-	   argument specifying the function to use.  */
-# undef OP
-# define OP(f, a, b, c, d, k, s, T) \
-	do { \
-		a += f(b, c, d) + correct_words[k] + T; \
-		a = rotl32(a, s); \
-		a += b; \
-	} while (0)
-
-	/* Round 2.  */
-# if MD5_SIZE_VS_SPEED == 1
-	pp = P_array;
-	for (i = 0; i < 4; i++) {
-		OP(FG, A, B, C, D, (int) (*pp++), 5, *pc++);
-		OP(FG, D, A, B, C, (int) (*pp++), 9, *pc++);
-		OP(FG, C, D, A, B, (int) (*pp++), 14, *pc++);
-		OP(FG, B, C, D, A, (int) (*pp++), 20, *pc++);
-	}
-# else
-	OP(FG, A, B, C, D, 1, 5, 0xf61e2562);
-	OP(FG, D, A, B, C, 6, 9, 0xc040b340);
-	OP(FG, C, D, A, B, 11, 14, 0x265e5a51);
-	OP(FG, B, C, D, A, 0, 20, 0xe9b6c7aa);
-	OP(FG, A, B, C, D, 5, 5, 0xd62f105d);
-	OP(FG, D, A, B, C, 10, 9, 0x02441453);
-	OP(FG, C, D, A, B, 15, 14, 0xd8a1e681);
-	OP(FG, B, C, D, A, 4, 20, 0xe7d3fbc8);
-	OP(FG, A, B, C, D, 9, 5, 0x21e1cde6);
-	OP(FG, D, A, B, C, 14, 9, 0xc33707d6);
-	OP(FG, C, D, A, B, 3, 14, 0xf4d50d87);
-	OP(FG, B, C, D, A, 8, 20, 0x455a14ed);
-	OP(FG, A, B, C, D, 13, 5, 0xa9e3e905);
-	OP(FG, D, A, B, C, 2, 9, 0xfcefa3f8);
-	OP(FG, C, D, A, B, 7, 14, 0x676f02d9);
-	OP(FG, B, C, D, A, 12, 20, 0x8d2a4c8a);
-# endif /* MD5_SIZE_VS_SPEED == 1 */
-
-	/* Round 3.  */
-# if MD5_SIZE_VS_SPEED == 1
-	for (i = 0; i < 4; i++) {
-		OP(FH, A, B, C, D, (int) (*pp++), 4, *pc++);
-		OP(FH, D, A, B, C, (int) (*pp++), 11, *pc++);
-		OP(FH, C, D, A, B, (int) (*pp++), 16, *pc++);
-		OP(FH, B, C, D, A, (int) (*pp++), 23, *pc++);
-	}
-# else
-	OP(FH, A, B, C, D, 5, 4, 0xfffa3942);
-	OP(FH, D, A, B, C, 8, 11, 0x8771f681);
-	OP(FH, C, D, A, B, 11, 16, 0x6d9d6122);
-	OP(FH, B, C, D, A, 14, 23, 0xfde5380c);
-	OP(FH, A, B, C, D, 1, 4, 0xa4beea44);
-	OP(FH, D, A, B, C, 4, 11, 0x4bdecfa9);
-	OP(FH, C, D, A, B, 7, 16, 0xf6bb4b60);
-	OP(FH, B, C, D, A, 10, 23, 0xbebfbc70);
-	OP(FH, A, B, C, D, 13, 4, 0x289b7ec6);
-	OP(FH, D, A, B, C, 0, 11, 0xeaa127fa);
-	OP(FH, C, D, A, B, 3, 16, 0xd4ef3085);
-	OP(FH, B, C, D, A, 6, 23, 0x04881d05);
-	OP(FH, A, B, C, D, 9, 4, 0xd9d4d039);
-	OP(FH, D, A, B, C, 12, 11, 0xe6db99e5);
-	OP(FH, C, D, A, B, 15, 16, 0x1fa27cf8);
-	OP(FH, B, C, D, A, 2, 23, 0xc4ac5665);
-# endif /* MD5_SIZE_VS_SPEED == 1 */
-
-	/* Round 4.  */
-# if MD5_SIZE_VS_SPEED == 1
-	for (i = 0; i < 4; i++) {
-		OP(FI, A, B, C, D, (int) (*pp++), 6, *pc++);
-		OP(FI, D, A, B, C, (int) (*pp++), 10, *pc++);
-		OP(FI, C, D, A, B, (int) (*pp++), 15, *pc++);
-		OP(FI, B, C, D, A, (int) (*pp++), 21, *pc++);
-	}
-# else
-	OP(FI, A, B, C, D, 0, 6, 0xf4292244);
-	OP(FI, D, A, B, C, 7, 10, 0x432aff97);
-	OP(FI, C, D, A, B, 14, 15, 0xab9423a7);
-	OP(FI, B, C, D, A, 5, 21, 0xfc93a039);
-	OP(FI, A, B, C, D, 12, 6, 0x655b59c3);
-	OP(FI, D, A, B, C, 3, 10, 0x8f0ccc92);
-	OP(FI, C, D, A, B, 10, 15, 0xffeff47d);
-	OP(FI, B, C, D, A, 1, 21, 0x85845dd1);
-	OP(FI, A, B, C, D, 8, 6, 0x6fa87e4f);
-	OP(FI, D, A, B, C, 15, 10, 0xfe2ce6e0);
-	OP(FI, C, D, A, B, 6, 15, 0xa3014314);
-	OP(FI, B, C, D, A, 13, 21, 0x4e0811a1);
-	OP(FI, A, B, C, D, 4, 6, 0xf7537e82);
-	OP(FI, D, A, B, C, 11, 10, 0xbd3af235);
-	OP(FI, C, D, A, B, 2, 15, 0x2ad7d2bb);
-	OP(FI, B, C, D, A, 9, 21, 0xeb86d391);
-# endif	/* MD5_SIZE_VS_SPEED == 1 */
-#endif	/* MD5_SIZE_VS_SPEED > 1 */
-
-	/* Add the starting values of the context.  */
-	A += A_save;
-	B += B_save;
-	C += C_save;
-	D += D_save;
-
-	/* Put checksum in context given as argument.  */
-	ctx->A = A;
-	ctx->B = B;
-	ctx->C = C;
-	ctx->D = D;
-}
-
-/* Feed data through a temporary buffer to call md5_hash_aligned_block()
- * with chunks of data that are 4-byte aligned and a multiple of 64 bytes.
- * This function's internal buffer remembers previous data until it has 64
- * bytes worth to pass on.  Call md5_end() to flush this buffer. */
-void md5_hash(const void *buffer, size_t len, md5_ctx_t *ctx)
-{
-	char *buf = (char *)buffer;
-
-	/* RFC 1321 specifies the possible length of the file up to 2^64 bits,
-	 * Here we only track the number of bytes.  */
-	ctx->total += len;
-
-	/* Process all input. */
-	while (len) {
-		unsigned i = 64 - ctx->buflen;
-
-		/* Copy data into aligned buffer. */
-		if (i > len) i = len;
-		memcpy(ctx->buffer + ctx->buflen, buf, i);
-		len -= i;
-		ctx->buflen += i;
-		buf += i;
-
-		/* When buffer fills up, process it. */
-		if (ctx->buflen == 64) {
-			md5_hash_block(ctx->buffer, ctx);
-			ctx->buflen = 0;
-		}
-	}
-}
-
-/* Process the remaining bytes in the buffer and put result from CTX
- * in first 16 bytes following RESBUF.  The result is always in little
- * endian byte order, so that a byte-wise output yields to the wanted
- * ASCII representation of the message digest.
- */
-void md5_end(void *resbuf, md5_ctx_t *ctx)
-{
-	char *buf = ctx->buffer;
-	int i;
-
-	/* Pad data to block size.  */
-	buf[ctx->buflen++] = 0x80;
-	memset(buf + ctx->buflen, 0, 128 - ctx->buflen);
-
-	/* Put the 64-bit file length in *bits* at the end of the buffer.  */
-	ctx->total <<= 3;
-	if (ctx->buflen > 56)
-		buf += 64;
-	for (i = 0; i < 8; i++)
-		buf[56 + i] = ctx->total >> (i*8);
-
-	/* Process last bytes.  */
-	if (buf != ctx->buffer)
-		md5_hash_block(ctx->buffer, ctx);
-	md5_hash_block(buf, ctx);
-
-	/* The MD5 result is in little endian byte order.
-	 * We (ab)use the fact that A-D are consecutive in memory.
-	 */
-#if MD5_BIG_ENDIAN
-	ctx->A = SWAP_LE32(ctx->A);
-	ctx->B = SWAP_LE32(ctx->B);
-	ctx->C = SWAP_LE32(ctx->C);
-	ctx->D = SWAP_LE32(ctx->D);
-#endif
-	memcpy(resbuf, &ctx->A, sizeof(ctx->A) * 4);
-}
diff --git a/src/lib/hash_md5.h b/src/lib/hash_md5.h
deleted file mode 100644
index f7e9f398..00000000
--- a/src/lib/hash_md5.h
+++ /dev/null
@@ -1,32 +0,0 @@
-/* vi: set sw=4 ts=4: */
-/*
- *  md5.c - Compute MD5 checksum of strings according to the
- *          definition of MD5 in RFC 1321 from April 1992.
- *
- *  Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995.
- *
- *  Copyright (C) 1995-1999 Free Software Foundation, Inc.
- *  Copyright (C) 2001 Manuel Novoa III
- *  Copyright (C) 2003 Glenn L. McGrath
- *  Copyright (C) 2003 Erik Andersen
- *
- *  Licensed under the GPL v2 or later, see the file LICENSE in this tarball.
- */
-
-#define MD5_RESULT_LEN 16
-
-typedef struct md5_ctx_t {
-        uint32_t A;
-        uint32_t B;
-        uint32_t C;
-        uint32_t D;
-        uint64_t total;
-        uint32_t buflen;
-        char buffer[128];
-} md5_ctx_t;
-#define md5_begin abrt_md5_begin
-void md5_begin(md5_ctx_t *ctx);
-#define md5_hash abrt_md5_hash
-void md5_hash(const void *data, size_t length, md5_ctx_t *ctx);
-#define md5_end abrt_md5_end
-void md5_end(void *resbuf, md5_ctx_t *ctx);
diff --git a/src/lib/hash_sha1.c b/src/lib/hash_sha1.c
index 954e7bc8..f1a9e0a6 100644
--- a/src/lib/hash_sha1.c
+++ b/src/lib/hash_sha1.c
@@ -18,6 +18,7 @@
  * ---------------------------------------------------------------------------
  */
 #include "abrtlib.h"
+#include <byteswap.h>
 
 #if defined(__BIG_ENDIAN__) && __BIG_ENDIAN__
 # define SHA1_BIG_ENDIAN 1
@@ -32,51 +33,37 @@
 # error "Can't determine endianness"
 #endif
 
-
 #define rotl32(x,n) (((x) << (n)) | ((x) >> (32 - (n))))
+/* for sha256: */
 #define rotr32(x,n) (((x) >> (n)) | ((x) << (32 - (n))))
 /* for sha512: */
 #define rotr64(x,n) (((x) >> (n)) | ((x) << (64 - (n))))
-#if SHA1_LITTLE_ENDIAN
-static inline uint64_t hton64(uint64_t v)
-{
-	return (((uint64_t)htonl(v)) << 32) | htonl(v >> 32);
-}
-#else
-#define hton64(v) (v)
-#endif
-#define ntoh64(v) hton64(v)
 
-/* To check alignment gcc has an appropriate operator.  Other
-   compilers don't.  */
-#if defined(__GNUC__) && __GNUC__ >= 2
-# define UNALIGNED_P(p,type) (((uintptr_t) p) % __alignof__(type) != 0)
-#else
-# define UNALIGNED_P(p,type) (((uintptr_t) p) % sizeof(type) != 0)
-#endif
+
+/* Generic 64-byte helpers for 64-byte block hashes */
+static void common64_hash(sha1_ctx_t *ctx, const void *buffer, size_t len);
+static void common64_end(sha1_ctx_t *ctx, int swap_needed);
 
 
-/* Some arch headers have conflicting defines */
-#undef ch
-#undef parity
-#undef maj
-#undef rnd
+/* sha1 specific code */
 
 static void sha1_process_block64(sha1_ctx_t *ctx)
 {
-	unsigned t;
-	uint32_t W[80], a, b, c, d, e;
-	const uint32_t *words = (uint32_t*) ctx->wbuffer;
-
-	for (t = 0; t < 16; ++t) {
-		W[t] = ntohl(*words);
-		words++;
-	}
-
-	for (/*t = 16*/; t < 80; ++t) {
-		uint32_t T = W[t - 3] ^ W[t - 8] ^ W[t - 14] ^ W[t - 16];
-		W[t] = rotl32(T, 1);
-	}
+	static const uint32_t rconsts[] = {
+		0x5A827999, 0x6ED9EBA1, 0x8F1BBCDC, 0xCA62C1D6
+	};
+	int i, j;
+	int cnt;
+	uint32_t W[16+16];
+	uint32_t a, b, c, d, e;
+
+	/* On-stack work buffer frees up one register in the main loop
+	 * which otherwise will be needed to hold ctx pointer */
+	for (i = 0; i < 16; i++)
+		if (SHA1_BIG_ENDIAN)
+			W[i] = W[i+16] = ((uint32_t*)ctx->wbuffer)[i];
+		else
+			W[i] = W[i+16] = bswap_32(((uint32_t*)ctx->wbuffer)[i]);
 
 	a = ctx->hash[0];
 	b = ctx->hash[1];
@@ -84,37 +71,41 @@ static void sha1_process_block64(sha1_ctx_t *ctx)
 	d = ctx->hash[3];
 	e = ctx->hash[4];
 
-/* Reverse byte order in 32-bit words   */
-#define ch(x,y,z)        ((z) ^ ((x) & ((y) ^ (z))))
-#define parity(x,y,z)    ((x) ^ (y) ^ (z))
-#define maj(x,y,z)       (((x) & (y)) | ((z) & ((x) | (y))))
-/* A normal version as set out in the FIPS. This version uses   */
-/* partial loop unrolling and is optimised for the Pentium 4    */
-#define rnd(f,k) \
-	do { \
-		uint32_t T = a; \
-		a = rotl32(a, 5) + f(b, c, d) + e + k + W[t]; \
-		e = d; \
-		d = c; \
-		c = rotl32(b, 30); \
-		b = T; \
-	} while (0)
-
-	for (t = 0; t < 20; ++t)
-		rnd(ch, 0x5a827999);
-
-	for (/*t = 20*/; t < 40; ++t)
-		rnd(parity, 0x6ed9eba1);
-
-	for (/*t = 40*/; t < 60; ++t)
-		rnd(maj, 0x8f1bbcdc);
-
-	for (/*t = 60*/; t < 80; ++t)
-		rnd(parity, 0xca62c1d6);
-#undef ch
-#undef parity
-#undef maj
-#undef rnd
+	/* 4 rounds of 20 operations each */
+	cnt = 0;
+	for (i = 0; i < 4; i++) {
+		j = 19;
+		do {
+			uint32_t work;
+
+			work = c ^ d;
+			if (i == 0) {
+				work = (work & b) ^ d;
+				if (j <= 3)
+					goto ge16;
+				/* Used to do bswap_32 here, but this
+				 * requires ctx (see comment above) */
+				work += W[cnt];
+			} else {
+				if (i == 2)
+					work = ((b | c) & d) | (b & c);
+				else /* i = 1 or 3 */
+					work ^= b;
+ ge16:
+				W[cnt] = W[cnt+16] = rotl32(W[cnt+13] ^ W[cnt+8] ^ W[cnt+2] ^ W[cnt], 1);
+				work += W[cnt];
+			}
+			work += e + rotl32(a, 5) + rconsts[i];
+
+			/* Rotate by one for next time */
+			e = d;
+			d = c;
+			c = /* b = */ rotl32(b, 30);
+			b = a;
+			a = work;
+			cnt = (cnt + 1) & 15;
+		} while (--j >= 0);
+	}
 
 	ctx->hash[0] += a;
 	ctx->hash[1] += b;
@@ -131,83 +122,90 @@ void sha1_begin(sha1_ctx_t *ctx)
 	ctx->hash[3] = 0x10325476;
 	ctx->hash[4] = 0xc3d2e1f0;
 	ctx->total64 = 0;
-	ctx->process_block = sha1_process_block64;
+	/* for sha256: ctx->process_block = sha1_process_block64; */
 }
 
-static const uint32_t init256[] = {
-	0x6a09e667,
-	0xbb67ae85,
-	0x3c6ef372,
-	0xa54ff53a,
-	0x510e527f,
-	0x9b05688c,
-	0x1f83d9ab,
-	0x5be0cd19
-};
-static const uint32_t init512_lo[] = {
-	0xf3bcc908,
-	0x84caa73b,
-	0xfe94f82b,
-	0x5f1d36f1,
-	0xade682d1,
-	0x2b3e6c1f,
-	0xfb41bd6b,
-	0x137e2179
-};
-
-/* Used also for sha256 */
-void sha1_hash(const void *buffer, size_t len, sha1_ctx_t *ctx)
+void sha1_hash(sha1_ctx_t *ctx, const void *buffer, size_t len)
 {
-	unsigned in_buf = ctx->total64 & 63;
-	unsigned add = 64 - in_buf;
+        common64_hash(ctx, buffer, len);
+}
 
-	ctx->total64 += len;
+/* May be used also for sha256 */
+void sha1_end(sha1_ctx_t *ctx, void *resbuf)
+{
+	unsigned hash_size;
 
-	while (len >= add) {	/* transfer whole blocks while possible  */
-		memcpy(ctx->wbuffer + in_buf, buffer, add);
-		buffer = (const char *)buffer + add;
-		len -= add;
-		add = 64;
-		in_buf = 0;
-		ctx->process_block(ctx);
-	}
+	/* SHA stores total in BE, need to swap on LE arches: */
+	common64_end(ctx, /*swap_needed:*/ SHA1_LITTLE_ENDIAN);
 
-	memcpy(ctx->wbuffer + in_buf, buffer, len);
+	hash_size = 5; /* (ctx->process_block == sha1_process_block64) ? 5 : 8; */
+	/* This way we do not impose alignment constraints on resbuf: */
+	if (SHA1_LITTLE_ENDIAN) {
+		unsigned i;
+		for (i = 0; i < hash_size; ++i)
+			ctx->hash[i] = bswap_32(ctx->hash[i]);
+	}
+	memcpy(resbuf, ctx->hash, sizeof(ctx->hash[0]) * hash_size);
 }
 
-/* Used also for sha256 */
-void sha1_end(void *resbuf, sha1_ctx_t *ctx)
+
+/* Generic 64-byte helpers for 64-byte block hashes */
+
+/*#define PROCESS_BLOCK(ctx) ctx->process_block(ctx)*/
+#define PROCESS_BLOCK(ctx) sha1_process_block64(ctx)
+
+/* Feed data through a temporary buffer.
+ * The internal buffer remembers previous data until it has 64
+ * bytes worth to pass on.
+ */
+static void common64_hash(sha1_ctx_t *ctx, const void *buffer, size_t len)
 {
-	unsigned pad, in_buf;
+	unsigned bufpos = ctx->total64 & 63;
+
+	ctx->total64 += len;
 
-	in_buf = ctx->total64 & 63;
+	while (1) {
+		unsigned remaining = 64 - bufpos;
+		if (remaining > len)
+			remaining = len;
+		/* Copy data into aligned buffer */
+		memcpy(ctx->wbuffer + bufpos, buffer, remaining);
+		len -= remaining;
+		buffer = (const char *)buffer + remaining;
+		bufpos += remaining;
+		/* clever way to do "if (bufpos != 64) break; ... ; bufpos = 0;" */
+		bufpos -= 64;
+		if (bufpos != 0)
+			break;
+		/* Buffer is filled up, process it */
+		PROCESS_BLOCK(ctx);
+		/*bufpos = 0; - already is */
+	}
+}
+
+/* Process the remaining bytes in the buffer */
+static void common64_end(sha1_ctx_t *ctx, int swap_needed)
+{
+	unsigned bufpos = ctx->total64 & 63;
 	/* Pad the buffer to the next 64-byte boundary with 0x80,0,0,0... */
-	ctx->wbuffer[in_buf++] = 0x80;
+	ctx->wbuffer[bufpos++] = 0x80;
 
 	/* This loop iterates either once or twice, no more, no less */
 	while (1) {
-		pad = 64 - in_buf;
-		memset(ctx->wbuffer + in_buf, 0, pad);
-		in_buf = 0;
+		unsigned remaining = 64 - bufpos;
+		memset(ctx->wbuffer + bufpos, 0, remaining);
 		/* Do we have enough space for the length count? */
-		if (pad >= 8) {
-			/* Store the 64-bit counter of bits in the buffer in BE format */
+		if (remaining >= 8) {
+			/* Store the 64-bit counter of bits in the buffer */
 			uint64_t t = ctx->total64 << 3;
-			t = hton64(t);
+			if (swap_needed)
+				t = bswap_64(t);
 			/* wbuffer is suitably aligned for this */
 			*(uint64_t *) (&ctx->wbuffer[64 - 8]) = t;
 		}
-		ctx->process_block(ctx);
-		if (pad >= 8)
+		PROCESS_BLOCK(ctx);
+		if (remaining >= 8)
 			break;
+		bufpos = 0;
 	}
-
-	in_buf = (ctx->process_block == sha1_process_block64) ? 5 : 8;
-	/* This way we do not impose alignment constraints on resbuf: */
-	if (SHA1_LITTLE_ENDIAN) {
-		unsigned i;
-		for (i = 0; i < in_buf; ++i)
-			ctx->hash[i] = htonl(ctx->hash[i]);
-	}
-	memcpy(resbuf, ctx->hash, sizeof(ctx->hash[0]) * in_buf);
 }
diff --git a/src/lib/hash_sha1.h b/src/lib/hash_sha1.h
index 09f50d12..2b6b7606 100644
--- a/src/lib/hash_sha1.h
+++ b/src/lib/hash_sha1.h
@@ -25,18 +25,18 @@ extern "C" {
 #define SHA1_RESULT_LEN (5 * 4)
 
 typedef struct sha1_ctx_t {
-	uint32_t hash[8];    /* 5, +3 elements for sha256 */
-	uint64_t total64;
-	uint8_t wbuffer[64]; /* NB: always correctly aligned for uint64_t */
-	void (*process_block)(struct sha1_ctx_t*);
+	uint8_t wbuffer[64]; /* always correctly aligned for uint64_t */
+	/* for sha256: void (*process_block)(struct md5_ctx_t*); */
+	uint64_t total64;    /* must be directly before hash[] */
+	uint32_t hash[8];    /* 4 elements for md5, 5 for sha1, 8 for sha256 */
 } sha1_ctx_t;
 
 #define sha1_begin abrt_sha1_begin
 void sha1_begin(sha1_ctx_t *ctx);
 #define sha1_hash abrt_sha1_hash
-void sha1_hash(const void *buffer, size_t len, sha1_ctx_t *ctx);
+void sha1_hash(sha1_ctx_t *ctx, const void *buffer, size_t len);
 #define sha1_end abrt_sha1_end
-void sha1_end(void *resbuf, sha1_ctx_t *ctx);
+void sha1_end(sha1_ctx_t *ctx, void *resbuf);
 
 #ifdef __cplusplus
 }