Unmangle scrypt.c (revert to cd0b57640c0de03ac23bc965649d4085b1a2556a)

The mangled scrypt.c from Art Forz is too much broken on big endian
systems. Revert it back to something that is more maintainable.

1 files changed, 472 insertions, 201 deletions

diff --git a/scrypt.c b/scrypt.c
index 803a45a..3b80a81 100644
--- a/scrypt.c
+++ b/scrypt.c
@@ -34,24 +34,83 @@
 #include <stdint.h>
 #include <string.h>
 
-#define byteswap(x) ((((x) << 24) & 0xff000000u) | (((x) << 8) & 0x00ff0000u) | (((x) >> 8) & 0x0000ff00u) | (((x) >> 24) & 0x000000ffu))
+
+static inline uint32_t
+be32dec(const void *pp)
+{
+	const uint8_t *p = (uint8_t const *)pp;
+
+	return ((uint32_t)(p[3]) + ((uint32_t)(p[2]) << 8) +
+	    ((uint32_t)(p[1]) << 16) + ((uint32_t)(p[0]) << 24));
+}
+
+static inline void
+be32enc(void *pp, uint32_t x)
+{
+	uint8_t * p = (uint8_t *)pp;
+
+	p[3] = x & 0xff;
+	p[2] = (x >> 8) & 0xff;
+	p[1] = (x >> 16) & 0xff;
+	p[0] = (x >> 24) & 0xff;
+}
+
+static inline uint32_t
+le32dec(const void *pp)
+{
+	const uint8_t *p = (uint8_t const *)pp;
+
+	return ((uint32_t)(p[0]) + ((uint32_t)(p[1]) << 8) +
+	    ((uint32_t)(p[2]) << 16) + ((uint32_t)(p[3]) << 24));
+}
+
+static inline void
+le32enc(void *pp, uint32_t x)
+{
+	uint8_t * p = (uint8_t *)pp;
+
+	p[0] = x & 0xff;
+	p[1] = (x >> 8) & 0xff;
+	p[2] = (x >> 16) & 0xff;
+	p[3] = (x >> 24) & 0xff;
+}
+
 
 typedef struct SHA256Context {
 	uint32_t state[8];
-	uint32_t buf[16];
+	uint32_t count[2];
+	unsigned char buf[64];
 } SHA256_CTX;
 
+typedef struct HMAC_SHA256Context {
+	SHA256_CTX ictx;
+	SHA256_CTX octx;
+} HMAC_SHA256_CTX;
+
 /*
  * 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 inline void
-be32enc_vect(uint32_t *dst, const uint32_t *src, uint32_t len)
+static void
+be32enc_vect(unsigned char *dst, const uint32_t *src, size_t len)
 {
-	uint32_t i;
+	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; i++)
-		dst[i] = byteswap(src[i]);
+	for (i = 0; i < len / 4; i++)
+		dst[i] = be32dec(src + i * 4);
 }
 
 /* Elementary functions used by SHA256 */
@@ -84,7 +143,7 @@ be32enc_vect(uint32_t *dst, const uint32_t *src, uint32_t len)
  * the 512-bit input block to produce a new state.
  */
 static void
-SHA256_Transform(uint32_t * state, const uint32_t block[16], int swap)
+SHA256_Transform(uint32_t * state, const unsigned char block[64])
 {
 	uint32_t W[64];
 	uint32_t S[8];
@@ -92,15 +151,9 @@ SHA256_Transform(uint32_t * state, const uint32_t block[16], int swap)
 	int i;
 
 	/* 1. Prepare message schedule W. */
-	if(swap)
-		for (i = 0; i < 16; i++)
-			W[i] = byteswap(block[i]);
-	else
-		memcpy(W, block, 64);
-	for (i = 16; i < 64; i += 2) {
+	be32dec_vect(W, block, 64);
+	for (i = 16; i < 64; i++)
 		W[i] = s1(W[i - 2]) + W[i - 7] + s0(W[i - 15]) + W[i - 16];
-		W[i+1] = s1(W[i - 1]) + W[i - 6] + s0(W[i - 14]) + W[i - 15];
-	}
 
 	/* 2. Initialize working variables. */
 	memcpy(S, state, 32);
@@ -174,260 +227,478 @@ SHA256_Transform(uint32_t * state, const uint32_t block[16], int swap)
 	/* 4. Mix local working variables into global state */
 	for (i = 0; i < 8; i++)
 		state[i] += S[i];
+
+	/* Clean the stack. */
+	memset(W, 0, 256);
+	memset(S, 0, 32);
+	t0 = t1 = 0;
 }
 
-static inline void
-SHA256_InitState(uint32_t * state)
+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
+};
+
+/* SHA-256 initialization.  Begins a SHA-256 operation. */
+static void
+SHA256_Init(SHA256_CTX * ctx)
 {
+
+	/* Zero bits processed so far */
+	ctx->count[0] = ctx->count[1] = 0;
+
 	/* Magic initialization constants */
-	state[0] = 0x6A09E667;
-	state[1] = 0xBB67AE85;
-	state[2] = 0x3C6EF372;
-	state[3] = 0xA54FF53A;
-	state[4] = 0x510E527F;
-	state[5] = 0x9B05688C;
-	state[6] = 0x1F83D9AB;
-	state[7] = 0x5BE0CD19;
+	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 */
+static void
+SHA256_Update(SHA256_CTX * ctx, const void *in, size_t len)
+{
+	uint32_t bitlen[2];
+	uint32_t r;
+	const unsigned char *src = in;
+
+	/* Number of bytes left in the buffer from previous updates */
+	r = (ctx->count[1] >> 3) & 0x3f;
+
+	/* Convert the length into a number of bits */
+	bitlen[1] = ((uint32_t)len) << 3;
+	bitlen[0] = (uint32_t)(len >> 29);
+
+	/* Update number of bits */
+	if ((ctx->count[1] += bitlen[1]) < bitlen[1])
+		ctx->count[0]++;
+	ctx->count[0] += bitlen[0];
+
+	/* 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);
+}
+
+/* Add padding and terminating bit-count. */
+static void
+SHA256_Pad(SHA256_CTX * ctx)
+{
+	unsigned char len[8];
+	uint32_t r, plen;
+
+	/*
+	 * Convert length to a vector of bytes -- we do this now rather
+	 * than later because the length will change after we pad.
+	 */
+	be32enc_vect(len, ctx->count, 8);
+
+	/* Add 1--64 bytes so that the resulting length is 56 mod 64 */
+	r = (ctx->count[1] >> 3) & 0x3f;
+	plen = (r < 56) ? (56 - r) : (120 - r);
+	SHA256_Update(ctx, PAD, (size_t)plen);
+
+	/* Add the terminating bit-count */
+	SHA256_Update(ctx, len, 8);
 }
 
-static const uint32_t passwdpad[12] = {0x00000080, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0x80020000};
-static const uint32_t outerpad[8] = {0x80000000, 0, 0, 0, 0, 0, 0, 0x00000300};
+/*
+ * SHA-256 finalization.  Pads the input data, exports the hash value,
+ * and clears the context state.
+ */
+static void
+SHA256_Final(unsigned char digest[32], SHA256_CTX * ctx)
+{
+
+	/* Add padding */
+	SHA256_Pad(ctx);
+
+	/* Write the hash */
+	be32enc_vect(digest, ctx->state, 32);
+
+	/* Clear the context state */
+	memset((void *)ctx, 0, sizeof(*ctx));
+}
+
+/* Initialize an HMAC-SHA256 operation with the given key. */
+static void
+HMAC_SHA256_Init(HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen)
+{
+	unsigned char pad[64];
+	unsigned char khash[32];
+	const unsigned char * K = _K;
+	size_t i;
+
+	/* If Klen > 64, the key is really SHA256(K). */
+	if (Klen > 64) {
+		SHA256_Init(&ctx->ictx);
+		SHA256_Update(&ctx->ictx, K, Klen);
+		SHA256_Final(khash, &ctx->ictx);
+		K = khash;
+		Klen = 32;
+	}
+
+	/* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
+	SHA256_Init(&ctx->ictx);
+	memset(pad, 0x36, 64);
+	for (i = 0; i < Klen; i++)
+		pad[i] ^= K[i];
+	SHA256_Update(&ctx->ictx, pad, 64);
+
+	/* Outer SHA256 operation is SHA256(K xor [block of 0x5c] || hash). */
+	SHA256_Init(&ctx->octx);
+	memset(pad, 0x5c, 64);
+	for (i = 0; i < Klen; i++)
+		pad[i] ^= K[i];
+	SHA256_Update(&ctx->octx, pad, 64);
+
+	/* Clean the stack. */
+	memset(khash, 0, 32);
+}
+
+/* Add bytes to the HMAC-SHA256 operation. */
+static void
+HMAC_SHA256_Update(HMAC_SHA256_CTX * ctx, const void *in, size_t len)
+{
+
+	/* Feed data to the inner SHA256 operation. */
+	SHA256_Update(&ctx->ictx, in, len);
+}
+
+/* Finish an HMAC-SHA256 operation. */
+static void
+HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx)
+{
+	unsigned char ihash[32];
+
+	/* Finish the inner SHA256 operation. */
+	SHA256_Final(ihash, &ctx->ictx);
+
+	/* Feed the inner hash to the outer SHA256 operation. */
+	SHA256_Update(&ctx->octx, ihash, 32);
+
+	/* Finish the outer SHA256 operation. */
+	SHA256_Final(digest, &ctx->octx);
+
+	/* Clean the stack. */
+	memset(ihash, 0, 32);
+}
 
 /**
  * PBKDF2_SHA256(passwd, passwdlen, salt, saltlen, c, buf, dkLen):
  * Compute PBKDF2(passwd, salt, c, dkLen) using HMAC-SHA256 as the PRF, and
  * write the output to buf.  The value dkLen must be at most 32 * (2^32 - 1).
  */
-static inline void
-PBKDF2_SHA256_80_128(const uint32_t * passwd, uint32_t * buf)
+static void
+PBKDF2_SHA256(const uint8_t * passwd, size_t passwdlen, const uint8_t * salt,
+    size_t saltlen, uint64_t c, uint8_t * buf, size_t dkLen)
 {
-	SHA256_CTX PShictx, PShoctx;
-	uint32_t tstate[8];
-	uint32_t ihash[8];
-	uint32_t i;
-	uint32_t pad[16];
-	
-	static const uint32_t innerpad[11] = {0x00000080, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xa0040000};
+	HMAC_SHA256_CTX PShctx, hctx;
+	size_t i;
+	uint8_t ivec[4];
+	uint8_t U[32];
+	uint8_t T[32];
+	uint64_t j;
+	int k;
+	size_t clen;
 
-	/* If Klen > 64, the key is really SHA256(K). */
-	SHA256_InitState(tstate);
-	SHA256_Transform(tstate, passwd, 1);
-	memcpy(pad, passwd+16, 16);
-	memcpy(pad+4, passwdpad, 48);
-	SHA256_Transform(tstate, pad, 1);
-	memcpy(ihash, tstate, 32);
-
-	SHA256_InitState(PShictx.state);
-	for (i = 0; i < 8; i++)
-		pad[i] = ihash[i] ^ 0x36363636;
-	for (; i < 16; i++)
-		pad[i] = 0x36363636;
-	SHA256_Transform(PShictx.state, pad, 0);
-	SHA256_Transform(PShictx.state, passwd, 1);
-	be32enc_vect(PShictx.buf, passwd+16, 4);
-	be32enc_vect(PShictx.buf+5, innerpad, 11);
-
-	SHA256_InitState(PShoctx.state);
-	for (i = 0; i < 8; i++)
-		pad[i] = ihash[i] ^ 0x5c5c5c5c;
-	for (; i < 16; i++)
-		pad[i] = 0x5c5c5c5c;
-	SHA256_Transform(PShoctx.state, pad, 0);
-	memcpy(PShoctx.buf+8, outerpad, 32);
+	/* Compute HMAC state after processing P and S. */
+	HMAC_SHA256_Init(&PShctx, passwd, passwdlen);
+	HMAC_SHA256_Update(&PShctx, salt, saltlen);
 
 	/* Iterate through the blocks. */
-	for (i = 0; i < 4; i++) {
-		uint32_t istate[8];
-		uint32_t ostate[8];
-		
-		memcpy(istate, PShictx.state, 32);
-		PShictx.buf[4] = i + 1;
-		SHA256_Transform(istate, PShictx.buf, 0);
-		memcpy(PShoctx.buf, istate, 32);
-
-		memcpy(ostate, PShoctx.state, 32);
-		SHA256_Transform(ostate, PShoctx.buf, 0);
-		be32enc_vect(buf+i*8, ostate, 8);
+	for (i = 0; i * 32 < dkLen; i++) {
+		/* Generate INT(i + 1). */
+		be32enc(ivec, (uint32_t)(i + 1));
+
+		/* Compute U_1 = PRF(P, S || INT(i)). */
+		memcpy(&hctx, &PShctx, sizeof(HMAC_SHA256_CTX));
+		HMAC_SHA256_Update(&hctx, ivec, 4);
+		HMAC_SHA256_Final(U, &hctx);
+
+		/* T_i = U_1 ... */
+		memcpy(T, U, 32);
+
+		for (j = 2; j <= c; j++) {
+			/* Compute U_j. */
+			HMAC_SHA256_Init(&hctx, passwd, passwdlen);
+			HMAC_SHA256_Update(&hctx, U, 32);
+			HMAC_SHA256_Final(U, &hctx);
+
+			/* ... xor U_j ... */
+			for (k = 0; k < 32; k++)
+				T[k] ^= U[k];
+		}
+
+		/* Copy as many bytes as necessary into buf. */
+		clen = dkLen - i * 32;
+		if (clen > 32)
+			clen = 32;
+		memcpy(&buf[i * 32], T, clen);
 	}
+
+	/* Clean PShctx, since we never called _Final on it. */
+	memset(&PShctx, 0, sizeof(HMAC_SHA256_CTX));
 }
 
 
-static inline uint32_t
-PBKDF2_SHA256_80_128_32(const uint32_t * passwd, const uint32_t * salt)
-{
-	uint32_t tstate[8];
-	uint32_t ostate[8];
-	uint32_t ihash[8];
-	uint32_t i;
+static void blkcpy(void *, void *, size_t);
+static void blkxor(void *, void *, size_t);
+static void salsa20_8(uint32_t[16]);
+static void blockmix_salsa8(uint32_t *, uint32_t *, uint32_t *, size_t);
+static uint64_t integerify(void *, size_t);
+static void smix(uint8_t *, size_t, uint64_t, uint32_t *, uint32_t *);
 
-	/* Compute HMAC state after processing P and S. */
-	uint32_t pad[16];
-	
-	static const uint32_t ihash_finalblk[16] = {0x00000001,0x80000000,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x00000620};
+static void
+blkcpy(void * dest, void * src, size_t len)
+{
+	size_t * D = dest;
+	size_t * S = src;
+	size_t L = len / sizeof(size_t);
+	size_t i;
 
-	/* If Klen > 64, the key is really SHA256(K). */
-	SHA256_InitState(tstate);
-	SHA256_Transform(tstate, passwd, 1);
-	memcpy(pad, passwd+16, 16);
-	memcpy(pad+4, passwdpad, 48);
-	SHA256_Transform(tstate, pad, 1);
-	memcpy(ihash, tstate, 32);
-
-	SHA256_InitState(ostate);
-	for (i = 0; i < 8; i++)
-		pad[i] = ihash[i] ^ 0x5c5c5c5c;
-	for (; i < 16; i++)
-		pad[i] = 0x5c5c5c5c;
-	SHA256_Transform(ostate, pad, 0);
+	for (i = 0; i < L; i++)
+		D[i] = S[i];
+}
 
-	SHA256_InitState(tstate);
-	for (i = 0; i < 8; i++)
-		pad[i] = ihash[i] ^ 0x36363636;
-	for (; i < 16; i++)
-		pad[i] = 0x36363636;
-	SHA256_Transform(tstate, pad, 0);
-	SHA256_Transform(tstate, salt, 1);
-	SHA256_Transform(tstate, salt+16, 1);
-	SHA256_Transform(tstate, ihash_finalblk, 0);
-	memcpy(pad, tstate, 32);
-	memcpy(pad+8, outerpad, 32);
+static void
+blkxor(void * dest, void * src, size_t len)
+{
+	size_t * D = dest;
+	size_t * S = src;
+	size_t L = len / sizeof(size_t);
+	size_t i;
 
-	/* Feed the inner hash to the outer SHA256 operation. */
-	SHA256_Transform(ostate, pad, 0);
-	/* Finish the outer SHA256 operation. */
-	return byteswap(ostate[7]);
+	for (i = 0; i < L; i++)
+		D[i] ^= S[i];
 }
 
-
 /**
  * salsa20_8(B):
  * Apply the salsa20/8 core to the provided block.
  */
-static inline void
-salsa20_8(uint32_t B[16], const uint32_t Bx[16])
+static void
+salsa20_8(uint32_t B[16])
 {
-	uint32_t x00,x01,x02,x03,x04,x05,x06,x07,x08,x09,x10,x11,x12,x13,x14,x15;
+	uint32_t x[16];
 	size_t i;
 
-	x00 = (B[ 0] ^= Bx[ 0]);
-	x01 = (B[ 1] ^= Bx[ 1]);
-	x02 = (B[ 2] ^= Bx[ 2]);
-	x03 = (B[ 3] ^= Bx[ 3]);
-	x04 = (B[ 4] ^= Bx[ 4]);
-	x05 = (B[ 5] ^= Bx[ 5]);
-	x06 = (B[ 6] ^= Bx[ 6]);
-	x07 = (B[ 7] ^= Bx[ 7]);
-	x08 = (B[ 8] ^= Bx[ 8]);
-	x09 = (B[ 9] ^= Bx[ 9]);
-	x10 = (B[10] ^= Bx[10]);
-	x11 = (B[11] ^= Bx[11]);
-	x12 = (B[12] ^= Bx[12]);
-	x13 = (B[13] ^= Bx[13]);
-	x14 = (B[14] ^= Bx[14]);
-	x15 = (B[15] ^= Bx[15]);
+	blkcpy(x, B, 64);
 	for (i = 0; i < 8; i += 2) {
 #define R(a,b) (((a) << (b)) | ((a) >> (32 - (b))))
 		/* Operate on columns. */
-		x04 ^= R(x00+x12, 7);	x09 ^= R(x05+x01, 7);	x14 ^= R(x10+x06, 7);	x03 ^= R(x15+x11, 7);
-		x08 ^= R(x04+x00, 9);	x13 ^= R(x09+x05, 9);	x02 ^= R(x14+x10, 9);	x07 ^= R(x03+x15, 9);
-		x12 ^= R(x08+x04,13);	x01 ^= R(x13+x09,13);	x06 ^= R(x02+x14,13);	x11 ^= R(x07+x03,13);
-		x00 ^= R(x12+x08,18);	x05 ^= R(x01+x13,18);	x10 ^= R(x06+x02,18);	x15 ^= R(x11+x07,18);
+		x[ 4] ^= R(x[ 0]+x[12], 7);  x[ 8] ^= R(x[ 4]+x[ 0], 9);
+		x[12] ^= R(x[ 8]+x[ 4],13);  x[ 0] ^= R(x[12]+x[ 8],18);
+
+		x[ 9] ^= R(x[ 5]+x[ 1], 7);  x[13] ^= R(x[ 9]+x[ 5], 9);
+		x[ 1] ^= R(x[13]+x[ 9],13);  x[ 5] ^= R(x[ 1]+x[13],18);
+
+		x[14] ^= R(x[10]+x[ 6], 7);  x[ 2] ^= R(x[14]+x[10], 9);
+		x[ 6] ^= R(x[ 2]+x[14],13);  x[10] ^= R(x[ 6]+x[ 2],18);
+
+		x[ 3] ^= R(x[15]+x[11], 7);  x[ 7] ^= R(x[ 3]+x[15], 9);
+		x[11] ^= R(x[ 7]+x[ 3],13);  x[15] ^= R(x[11]+x[ 7],18);
 
 		/* Operate on rows. */
-		x01 ^= R(x00+x03, 7);	x06 ^= R(x05+x04, 7);	x11 ^= R(x10+x09, 7);	x12 ^= R(x15+x14, 7);
-		x02 ^= R(x01+x00, 9);	x07 ^= R(x06+x05, 9);	x08 ^= R(x11+x10, 9);	x13 ^= R(x12+x15, 9);
-		x03 ^= R(x02+x01,13);	x04 ^= R(x07+x06,13);	x09 ^= R(x08+x11,13);	x14 ^= R(x13+x12,13);
-		x00 ^= R(x03+x02,18);	x05 ^= R(x04+x07,18);	x10 ^= R(x09+x08,18);	x15 ^= R(x14+x13,18);
+		x[ 1] ^= R(x[ 0]+x[ 3], 7);  x[ 2] ^= R(x[ 1]+x[ 0], 9);
+		x[ 3] ^= R(x[ 2]+x[ 1],13);  x[ 0] ^= R(x[ 3]+x[ 2],18);
+
+		x[ 6] ^= R(x[ 5]+x[ 4], 7);  x[ 7] ^= R(x[ 6]+x[ 5], 9);
+		x[ 4] ^= R(x[ 7]+x[ 6],13);  x[ 5] ^= R(x[ 4]+x[ 7],18);
+
+		x[11] ^= R(x[10]+x[ 9], 7);  x[ 8] ^= R(x[11]+x[10], 9);
+		x[ 9] ^= R(x[ 8]+x[11],13);  x[10] ^= R(x[ 9]+x[ 8],18);
+
+		x[12] ^= R(x[15]+x[14], 7);  x[13] ^= R(x[12]+x[15], 9);
+		x[14] ^= R(x[13]+x[12],13);  x[15] ^= R(x[14]+x[13],18);
 #undef R
 	}
-	B[ 0] += x00;
-	B[ 1] += x01;
-	B[ 2] += x02;
-	B[ 3] += x03;
-	B[ 4] += x04;
-	B[ 5] += x05;
-	B[ 6] += x06;
-	B[ 7] += x07;
-	B[ 8] += x08;
-	B[ 9] += x09;
-	B[10] += x10;
-	B[11] += x11;
-	B[12] += x12;
-	B[13] += x13;
-	B[14] += x14;
-	B[15] += x15;
+	for (i = 0; i < 16; i++)
+		B[i] += x[i];
+}
+
+/**
+ * blockmix_salsa8(Bin, Bout, X, r):
+ * Compute Bout = BlockMix_{salsa20/8, r}(Bin).  The input Bin must be 128r
+ * bytes in length; the output Bout must also be the same size.  The
+ * temporary space X must be 64 bytes.
+ */
+static void
+blockmix_salsa8(uint32_t * Bin, uint32_t * Bout, uint32_t * X, size_t r)
+{
+	size_t i;
+
+	/* 1: X <-- B_{2r - 1} */
+	blkcpy(X, &Bin[(2 * r - 1) * 16], 64);
+
+	/* 2: for i = 0 to 2r - 1 do */
+	for (i = 0; i < 2 * r; i += 2) {
+		/* 3: X <-- H(X \xor B_i) */
+		blkxor(X, &Bin[i * 16], 64);
+		salsa20_8(X);
+
+		/* 4: Y_i <-- X */
+		/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
+		blkcpy(&Bout[i * 8], X, 64);
+
+		/* 3: X <-- H(X \xor B_i) */
+		blkxor(X, &Bin[i * 16 + 16], 64);
+		salsa20_8(X);
+
+		/* 4: Y_i <-- X */
+		/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
+		blkcpy(&Bout[i * 8 + r * 16], X, 64);
+	}
+}
+
+/**
+ * integerify(B, r):
+ * Return the result of parsing B_{2r-1} as a little-endian integer.
+ */
+static uint64_t
+integerify(void * B, size_t r)
+{
+	uint32_t * X = (void *)((uintptr_t)(B) + (2 * r - 1) * 64);
+
+	return (((uint64_t)(X[1]) << 32) + X[0]);
+}
+
+/**
+ * smix(B, r, N, V, XY):
+ * Compute B = SMix_r(B, N).  The input B must be 128r bytes in length;
+ * the temporary storage V must be 128rN bytes in length; the temporary
+ * storage XY must be 256r + 64 bytes in length.  The value N must be a
+ * power of 2 greater than 1.  The arrays B, V, and XY must be aligned to a
+ * multiple of 64 bytes.
+ */
+static void
+smix(uint8_t * B, size_t r, uint64_t N, uint32_t * V, uint32_t * XY)
+{
+	uint32_t * X = XY;
+	uint32_t * Y = &XY[32 * r];
+	uint32_t * Z = &XY[64 * r];
+	uint64_t i;
+	uint64_t j;
+	size_t k;
+
+	/* 1: X <-- B */
+	for (k = 0; k < 32 * r; k++)
+		X[k] = le32dec(&B[4 * k]);
+
+	/* 2: for i = 0 to N - 1 do */
+	for (i = 0; i < N; i += 2) {
+		/* 3: V_i <-- X */
+		blkcpy(&V[i * (32 * r)], X, 128 * r);
+
+		/* 4: X <-- H(X) */
+		blockmix_salsa8(X, Y, Z, r);
+
+		/* 3: V_i <-- X */
+		blkcpy(&V[(i + 1) * (32 * r)], Y, 128 * r);
+
+		/* 4: X <-- H(X) */
+		blockmix_salsa8(Y, X, Z, r);
+	}
+
+	/* 6: for i = 0 to N - 1 do */
+	for (i = 0; i < N; i += 2) {
+		/* 7: j <-- Integerify(X) mod N */
+		j = integerify(X, r) & (N - 1);
+
+		/* 8: X <-- H(X \xor V_j) */
+		blkxor(X, &V[j * (32 * r)], 128 * r);
+		blockmix_salsa8(X, Y, Z, r);
+
+		/* 7: j <-- Integerify(X) mod N */
+		j = integerify(Y, r) & (N - 1);
+
+		/* 8: X <-- H(X \xor V_j) */
+		blkxor(Y, &V[j * (32 * r)], 128 * r);
+		blockmix_salsa8(Y, X, Z, r);
+	}
+
+	/* 10: B' <-- X */
+	for (k = 0; k < 32 * r; k++)
+		le32enc(&B[4 * k], X[k]);
 }
 
 /* cpu and memory intensive function to transform a 80 byte buffer into a 32 byte output
    scratchpad size needs to be at least 63 + (128 * r * p) + (256 * r + 64) + (128 * r * N) bytes
  */
-static uint32_t scrypt_1024_1_1_256_sp(const uint32_t* input, char* scratchpad)
+static void scrypt_1024_1_1_256_sp(const char* input, char* output, char* scratchpad)
 {
+	uint8_t * B;
 	uint32_t * V;
-	uint32_t X[32];
+	uint32_t * XY;
 	uint32_t i;
-	uint32_t j;
-	uint32_t k;
-	uint64_t *p1, *p2;
-
-	p1 = (uint64_t *)X;
-	V = (uint32_t *)(((uintptr_t)(scratchpad) + 63) & ~ (uintptr_t)(63));
 
-	PBKDF2_SHA256_80_128(input, X);
+	const uint32_t N = 1024;
+	const uint32_t r = 1;
+	const uint32_t p = 1;
 
-	for (i = 0; i < 1024; i += 2) {
-		memcpy(&V[i * 32], X, 128);
+	B = (uint8_t *)(((uintptr_t)(scratchpad) + 63) & ~ (uintptr_t)(63));
+	XY = (uint32_t *)(B + (128 * r * p));
+	V = (uint32_t *)(B + (128 * r * p) + (256 * r + 64));
 
-		salsa20_8(&X[0], &X[16]);
-		salsa20_8(&X[16], &X[0]);
+	/* 1: (B_0 ... B_{p-1}) <-- PBKDF2(P, S, 1, p * MFLen) */
+	PBKDF2_SHA256((const uint8_t*)input, 80, (const uint8_t*)input, 80, 1, B, p * 128 * r);
 
-		memcpy(&V[(i + 1) * 32], X, 128);
-
-		salsa20_8(&X[0], &X[16]);
-		salsa20_8(&X[16], &X[0]);
-	}
-	for (i = 0; i < 1024; i += 2) {
-		j = X[16] & 1023;
-		p2 = (uint64_t *)(&V[j * 32]);
-		for(k = 0; k < 16; k++)
-			p1[k] ^= p2[k];
-
-		salsa20_8(&X[0], &X[16]);
-		salsa20_8(&X[16], &X[0]);
-
-		j = X[16] & 1023;
-		p2 = (uint64_t *)(&V[j * 32]);
-		for(k = 0; k < 16; k++)
-			p1[k] ^= p2[k];
-
-		salsa20_8(&X[0], &X[16]);
-		salsa20_8(&X[16], &X[0]);
+	/* 2: for i = 0 to p - 1 do */
+	for (i = 0; i < p; i++) {
+		/* 3: B_i <-- MF(B_i, N) */
+		smix(&B[i * 128 * r], r, N, V, XY);
 	}
 
-	return PBKDF2_SHA256_80_128_32(input, X);
+	/* 5: DK <-- PBKDF2(P, B, 1, dkLen) */
+	PBKDF2_SHA256((const uint8_t*)input, 80, B, p * 128 * r, 1, (uint8_t*)output, 32);
 }
 
 int scanhash_scrypt(int thr_id, unsigned char *pdata, unsigned char *scratchbuf,
 	const unsigned char *ptarget,
 	uint32_t max_nonce, unsigned long *hashes_done)
 {
-	uint32_t data[20];
-	uint32_t tmp_hash7;
+	unsigned char data[80];
+	unsigned char tmp_hash[32];
+	uint32_t *nonce = (uint32_t *)(data + 64 + 12);
 	uint32_t n = 0;
-	uint32_t Htarg = ((const uint32_t *)ptarget)[7];
+	uint32_t Htarg = *(uint32_t *)(ptarget + 28);
 	int i;
 
 	work_restart[thr_id].restart = 0;
 	
-	be32enc_vect(data, (const uint32_t *)pdata, 19);
+	for (i = 0; i < 80/4; i++)
+		((uint32_t *)data)[i] = swab32(((uint32_t *)pdata)[i]);
 	
 	while(1) {
 		n++;
-		data[19] = n;
-		tmp_hash7 = scrypt_1024_1_1_256_sp(data, scratchbuf);
+		*nonce = n;
+		scrypt_1024_1_1_256_sp(data, tmp_hash, scratchbuf);
 
-		if (tmp_hash7 <= Htarg) {
-			((uint32_t *)pdata)[19] = byteswap(n);
+		if (*(uint32_t *)(tmp_hash+28) <= Htarg) {
+			*(uint32_t *)(pdata + 64 + 12) = swab32(n);
 			*hashes_done = n;
 			return true;
 		}