OpenZFS 4185 - add new cryptographic checksums to ZFS: SHA-512, Skein, Edon-R

Reviewed by: George Wilson <[email protected]>
Reviewed by: Prakash Surya <[email protected]>
Reviewed by: Saso Kiselkov <[email protected]>
Reviewed by: Richard Lowe <[email protected]>
Approved by: Garrett D'Amore <[email protected]>
Ported by: Tony Hutter <[email protected]>

OpenZFS-issue: https://www.illumos.org/issues/4185
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/45818ee

Porting Notes:
This code is ported on top of the Illumos Crypto Framework code:

    https://github.com/zfsonlinux/zfs/pull/4329/commits/b5e030c8dbb9cd393d313571dee4756fbba8c22d

The list of porting changes includes:

- Copied module/icp/include/sha2/sha2.h directly from illumos

- Removed from module/icp/algs/sha2/sha2.c:
	#pragma inline(SHA256Init, SHA384Init, SHA512Init)

- Added 'ctx' to lib/libzfs/libzfs_sendrecv.c:zio_checksum_SHA256() since
  it now takes in an extra parameter.

- Added CTASSERT() to assert.h from for module/zfs/edonr_zfs.c

- Added skein & edonr to libicp/Makefile.am

- Added sha512.S.  It was generated from sha512-x86_64.pl in Illumos.

- Updated ztest.c with new fletcher_4_*() args; used NULL for new CTX argument.

- In icp/algs/edonr/edonr_byteorder.h, Removed the #if defined(__linux) section
  to not #include the non-existant endian.h.

- In skein_test.c, renane NULL to 0 in "no test vector" array entries to get
  around a compiler warning.

- Fixup test files:
	- Rename <sys/varargs.h> -> <varargs.h>, <strings.h> -> <string.h>,
	- Remove <note.h> and define NOTE() as NOP.
	- Define u_longlong_t
	- Rename "#!/usr/bin/ksh" -> "#!/bin/ksh -p"
	- Rename NULL to 0 in "no test vector" array entries to get around a
	  compiler warning.
	- Remove "for isa in $($ISAINFO); do" stuff
	- Add/update Makefiles
	- Add some userspace headers like stdio.h/stdlib.h in places of
	  sys/types.h.

- EXPORT_SYMBOL *_Init/*_Update/*_Final... routines in ICP modules.

- Update scripts/zfs2zol-patch.sed

- include <sys/sha2.h> in sha2_impl.h

- Add sha2.h to include/sys/Makefile.am

- Add skein and edonr dirs to icp Makefile

- Add new checksums to zpool_get.cfg

- Move checksum switch block from zfs_secpolicy_setprop() to
  zfs_check_settable()

- Fix -Wuninitialized error in edonr_byteorder.h on PPC

- Fix stack frame size errors on ARM32
  	- Don't unroll loops in Skein on 32-bit to save stack space
  	- Add memory barriers in sha2.c on 32-bit to save stack space

- Add filetest_001_pos.ksh checksum sanity test

- Add option to write psudorandom data in file_write utility

7 files changed, 2320 insertions, 0 deletions

diff --git a/module/icp/algs/skein/THIRDPARTYLICENSE b/module/icp/algs/skein/THIRDPARTYLICENSE
new file mode 100644
index 000000000..b7434fd17
--- /dev/null
+++ b/module/icp/algs/skein/THIRDPARTYLICENSE
@@ -0,0 +1,3 @@
+Implementation of the Skein hash function.
+Source code author: Doug Whiting, 2008.
+This algorithm and source code is released to the public domain.
diff --git a/module/icp/algs/skein/THIRDPARTYLICENSE.descrip b/module/icp/algs/skein/THIRDPARTYLICENSE.descrip
new file mode 100644
index 000000000..0ae89cfdf
--- /dev/null
+++ b/module/icp/algs/skein/THIRDPARTYLICENSE.descrip
@@ -0,0 +1 @@
+LICENSE TERMS OF SKEIN HASH ALGORITHM IMPLEMENTATION
diff --git a/module/icp/algs/skein/skein.c b/module/icp/algs/skein/skein.c
new file mode 100644
index 000000000..0981eee08
--- /dev/null
+++ b/module/icp/algs/skein/skein.c
@@ -0,0 +1,921 @@
+/*
+ * Implementation of the Skein hash function.
+ * Source code author: Doug Whiting, 2008.
+ * This algorithm and source code is released to the public domain.
+ */
+/* Copyright 2013 Doug Whiting. This code is released to the public domain. */
+
+#define	SKEIN_PORT_CODE		/* instantiate any code in skein_port.h */
+
+#include <sys/types.h>
+#include <sys/note.h>
+#include <sys/skein.h>		/* get the Skein API definitions   */
+#include "skein_impl.h"		/* get internal definitions */
+
+/* External function to process blkCnt (nonzero) full block(s) of data. */
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx, const uint8_t *blkPtr,
+    size_t blkCnt, size_t byteCntAdd);
+void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx, const uint8_t *blkPtr,
+    size_t blkCnt, size_t byteCntAdd);
+void Skein1024_Process_Block(Skein1024_Ctxt_t *ctx, const uint8_t *blkPtr,
+    size_t blkCnt, size_t byteCntAdd);
+
+/* 256-bit Skein */
+/* init the context for a straight hashing operation  */
+int
+Skein_256_Init(Skein_256_Ctxt_t *ctx, size_t hashBitLen)
+{
+	union {
+		uint8_t b[SKEIN_256_STATE_BYTES];
+		uint64_t w[SKEIN_256_STATE_WORDS];
+	} cfg;			/* config block */
+
+	Skein_Assert(hashBitLen > 0, SKEIN_BAD_HASHLEN);
+	ctx->h.hashBitLen = hashBitLen;	/* output hash bit count */
+
+	switch (hashBitLen) {	/* use pre-computed values, where available */
+#ifndef	SKEIN_NO_PRECOMP
+	case 256:
+		bcopy(SKEIN_256_IV_256, ctx->X, sizeof (ctx->X));
+		break;
+	case 224:
+		bcopy(SKEIN_256_IV_224, ctx->X, sizeof (ctx->X));
+		break;
+	case 160:
+		bcopy(SKEIN_256_IV_160, ctx->X, sizeof (ctx->X));
+		break;
+	case 128:
+		bcopy(SKEIN_256_IV_128, ctx->X, sizeof (ctx->X));
+		break;
+#endif
+	default:
+		/* here if there is no precomputed IV value available */
+		/*
+		 * build/process the config block, type == CONFIG (could be
+		 * precomputed)
+		 */
+		/* set tweaks: T0=0; T1=CFG | FINAL */
+		Skein_Start_New_Type(ctx, CFG_FINAL);
+
+		/* set the schema, version */
+		cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+		/* hash result length in bits */
+		cfg.w[1] = Skein_Swap64(hashBitLen);
+		cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+		/* zero pad config block */
+		bzero(&cfg.w[3], sizeof (cfg) - 3 * sizeof (cfg.w[0]));
+
+		/* compute the initial chaining values from config block */
+		/* zero the chaining variables */
+		bzero(ctx->X, sizeof (ctx->X));
+		Skein_256_Process_Block(ctx, cfg.b, 1, SKEIN_CFG_STR_LEN);
+		break;
+	}
+	/*
+	 * The chaining vars ctx->X are now initialized for the given
+	 * hashBitLen.
+	 * Set up to process the data message portion of the hash (default)
+	 */
+	Skein_Start_New_Type(ctx, MSG);	/* T0=0, T1= MSG type */
+
+	return (SKEIN_SUCCESS);
+}
+
+/* init the context for a MAC and/or tree hash operation */
+/*
+ * [identical to Skein_256_Init() when keyBytes == 0 &&
+ * treeInfo == SKEIN_CFG_TREE_INFO_SEQUENTIAL]
+ */
+int
+Skein_256_InitExt(Skein_256_Ctxt_t *ctx, size_t hashBitLen, uint64_t treeInfo,
+    const uint8_t *key, size_t keyBytes)
+{
+	union {
+		uint8_t b[SKEIN_256_STATE_BYTES];
+		uint64_t w[SKEIN_256_STATE_WORDS];
+	} cfg;			/* config block */
+
+	Skein_Assert(hashBitLen > 0, SKEIN_BAD_HASHLEN);
+	Skein_Assert(keyBytes == 0 || key != NULL, SKEIN_FAIL);
+
+	/* compute the initial chaining values ctx->X[], based on key */
+	if (keyBytes == 0) {	/* is there a key? */
+		/* no key: use all zeroes as key for config block */
+		bzero(ctx->X, sizeof (ctx->X));
+	} else {		/* here to pre-process a key */
+
+		Skein_assert(sizeof (cfg.b) >= sizeof (ctx->X));
+		/* do a mini-Init right here */
+		/* set output hash bit count = state size */
+		ctx->h.hashBitLen = 8 * sizeof (ctx->X);
+		/* set tweaks: T0 = 0; T1 = KEY type */
+		Skein_Start_New_Type(ctx, KEY);
+		/* zero the initial chaining variables */
+		bzero(ctx->X, sizeof (ctx->X));
+		/* hash the key */
+		(void) Skein_256_Update(ctx, key, keyBytes);
+		/* put result into cfg.b[] */
+		(void) Skein_256_Final_Pad(ctx, cfg.b);
+		/* copy over into ctx->X[] */
+		bcopy(cfg.b, ctx->X, sizeof (cfg.b));
+#if	SKEIN_NEED_SWAP
+		{
+			uint_t i;
+			/* convert key bytes to context words */
+			for (i = 0; i < SKEIN_256_STATE_WORDS; i++)
+				ctx->X[i] = Skein_Swap64(ctx->X[i]);
+		}
+#endif
+	}
+	/*
+	 * build/process the config block, type == CONFIG (could be
+	 * precomputed for each key)
+	 */
+	ctx->h.hashBitLen = hashBitLen;	/* output hash bit count */
+	Skein_Start_New_Type(ctx, CFG_FINAL);
+
+	bzero(&cfg.w, sizeof (cfg.w));	/* pre-pad cfg.w[] with zeroes */
+	cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+	cfg.w[1] = Skein_Swap64(hashBitLen);	/* hash result length in bits */
+	/* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+	cfg.w[2] = Skein_Swap64(treeInfo);
+
+	Skein_Show_Key(256, &ctx->h, key, keyBytes);
+
+	/* compute the initial chaining values from config block */
+	Skein_256_Process_Block(ctx, cfg.b, 1, SKEIN_CFG_STR_LEN);
+
+	/* The chaining vars ctx->X are now initialized */
+	/* Set up to process the data message portion of the hash (default) */
+	ctx->h.bCnt = 0;	/* buffer b[] starts out empty */
+	Skein_Start_New_Type(ctx, MSG);
+
+	return (SKEIN_SUCCESS);
+}
+
+/* process the input bytes */
+int
+Skein_256_Update(Skein_256_Ctxt_t *ctx, const uint8_t *msg, size_t msgByteCnt)
+{
+	size_t n;
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES, SKEIN_FAIL);
+
+	/* process full blocks, if any */
+	if (msgByteCnt + ctx->h.bCnt > SKEIN_256_BLOCK_BYTES) {
+		/* finish up any buffered message data */
+		if (ctx->h.bCnt) {
+			/* # bytes free in buffer b[] */
+			n = SKEIN_256_BLOCK_BYTES - ctx->h.bCnt;
+			if (n) {
+				/* check on our logic here */
+				Skein_assert(n < msgByteCnt);
+				bcopy(msg, &ctx->b[ctx->h.bCnt], n);
+				msgByteCnt -= n;
+				msg += n;
+				ctx->h.bCnt += n;
+			}
+			Skein_assert(ctx->h.bCnt == SKEIN_256_BLOCK_BYTES);
+			Skein_256_Process_Block(ctx, ctx->b, 1,
+			    SKEIN_256_BLOCK_BYTES);
+			ctx->h.bCnt = 0;
+		}
+		/*
+		 * now process any remaining full blocks, directly from input
+		 * message data
+		 */
+		if (msgByteCnt > SKEIN_256_BLOCK_BYTES) {
+			/* number of full blocks to process */
+			n = (msgByteCnt - 1) / SKEIN_256_BLOCK_BYTES;
+			Skein_256_Process_Block(ctx, msg, n,
+			    SKEIN_256_BLOCK_BYTES);
+			msgByteCnt -= n * SKEIN_256_BLOCK_BYTES;
+			msg += n * SKEIN_256_BLOCK_BYTES;
+		}
+		Skein_assert(ctx->h.bCnt == 0);
+	}
+
+	/* copy any remaining source message data bytes into b[] */
+	if (msgByteCnt) {
+		Skein_assert(msgByteCnt + ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES);
+		bcopy(msg, &ctx->b[ctx->h.bCnt], msgByteCnt);
+		ctx->h.bCnt += msgByteCnt;
+	}
+
+	return (SKEIN_SUCCESS);
+}
+
+/* finalize the hash computation and output the result */
+int
+Skein_256_Final(Skein_256_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	size_t i, n, byteCnt;
+	uint64_t X[SKEIN_256_STATE_WORDS];
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES, SKEIN_FAIL);
+
+	ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;	/* tag as the final block */
+	/* zero pad b[] if necessary */
+	if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES)
+		bzero(&ctx->b[ctx->h.bCnt],
+		    SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+
+	/* process the final block */
+	Skein_256_Process_Block(ctx, ctx->b, 1, ctx->h.bCnt);
+
+	/* now output the result */
+	/* total number of output bytes */
+	byteCnt = (ctx->h.hashBitLen + 7) >> 3;
+
+	/* run Threefish in "counter mode" to generate output */
+	/* zero out b[], so it can hold the counter */
+	bzero(ctx->b, sizeof (ctx->b));
+	/* keep a local copy of counter mode "key" */
+	bcopy(ctx->X, X, sizeof (X));
+	for (i = 0; i * SKEIN_256_BLOCK_BYTES < byteCnt; i++) {
+		/* build the counter block */
+		uint64_t tmp = Skein_Swap64((uint64_t)i);
+		bcopy(&tmp, ctx->b, sizeof (tmp));
+		Skein_Start_New_Type(ctx, OUT_FINAL);
+		/* run "counter mode" */
+		Skein_256_Process_Block(ctx, ctx->b, 1, sizeof (uint64_t));
+		/* number of output bytes left to go */
+		n = byteCnt - i * SKEIN_256_BLOCK_BYTES;
+		if (n >= SKEIN_256_BLOCK_BYTES)
+			n = SKEIN_256_BLOCK_BYTES;
+		Skein_Put64_LSB_First(hashVal + i * SKEIN_256_BLOCK_BYTES,
+		    ctx->X, n);	/* "output" the ctr mode bytes */
+		Skein_Show_Final(256, &ctx->h, n,
+		    hashVal + i * SKEIN_256_BLOCK_BYTES);
+		/* restore the counter mode key for next time */
+		bcopy(X, ctx->X, sizeof (X));
+	}
+	return (SKEIN_SUCCESS);
+}
+
+/* 512-bit Skein */
+
+/* init the context for a straight hashing operation  */
+int
+Skein_512_Init(Skein_512_Ctxt_t *ctx, size_t hashBitLen)
+{
+	union {
+		uint8_t b[SKEIN_512_STATE_BYTES];
+		uint64_t w[SKEIN_512_STATE_WORDS];
+	} cfg;			/* config block */
+
+	Skein_Assert(hashBitLen > 0, SKEIN_BAD_HASHLEN);
+	ctx->h.hashBitLen = hashBitLen;	/* output hash bit count */
+
+	switch (hashBitLen) {	/* use pre-computed values, where available */
+#ifndef	SKEIN_NO_PRECOMP
+	case 512:
+		bcopy(SKEIN_512_IV_512, ctx->X, sizeof (ctx->X));
+		break;
+	case 384:
+		bcopy(SKEIN_512_IV_384, ctx->X, sizeof (ctx->X));
+		break;
+	case 256:
+		bcopy(SKEIN_512_IV_256, ctx->X, sizeof (ctx->X));
+		break;
+	case 224:
+		bcopy(SKEIN_512_IV_224, ctx->X, sizeof (ctx->X));
+		break;
+#endif
+	default:
+		/*
+		 * here if there is no precomputed IV value available
+		 * build/process the config block, type == CONFIG (could be
+		 * precomputed)
+		 */
+		/* set tweaks: T0=0; T1=CFG | FINAL */
+		Skein_Start_New_Type(ctx, CFG_FINAL);
+
+		/* set the schema, version */
+		cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+		/* hash result length in bits */
+		cfg.w[1] = Skein_Swap64(hashBitLen);
+		cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+		/* zero pad config block */
+		bzero(&cfg.w[3], sizeof (cfg) - 3 * sizeof (cfg.w[0]));
+
+		/* compute the initial chaining values from config block */
+		/* zero the chaining variables */
+		bzero(ctx->X, sizeof (ctx->X));
+		Skein_512_Process_Block(ctx, cfg.b, 1, SKEIN_CFG_STR_LEN);
+		break;
+	}
+
+	/*
+	 * The chaining vars ctx->X are now initialized for the given
+	 * hashBitLen. Set up to process the data message portion of the
+	 * hash (default)
+	 */
+	Skein_Start_New_Type(ctx, MSG);	/* T0=0, T1= MSG type */
+
+	return (SKEIN_SUCCESS);
+}
+
+/* init the context for a MAC and/or tree hash operation */
+/*
+ * [identical to Skein_512_Init() when keyBytes == 0 &&
+ * treeInfo == SKEIN_CFG_TREE_INFO_SEQUENTIAL]
+ */
+int
+Skein_512_InitExt(Skein_512_Ctxt_t *ctx, size_t hashBitLen, uint64_t treeInfo,
+    const uint8_t *key, size_t keyBytes)
+{
+	union {
+		uint8_t b[SKEIN_512_STATE_BYTES];
+		uint64_t w[SKEIN_512_STATE_WORDS];
+	} cfg;			/* config block */
+
+	Skein_Assert(hashBitLen > 0, SKEIN_BAD_HASHLEN);
+	Skein_Assert(keyBytes == 0 || key != NULL, SKEIN_FAIL);
+
+	/* compute the initial chaining values ctx->X[], based on key */
+	if (keyBytes == 0) {	/* is there a key? */
+		/* no key: use all zeroes as key for config block */
+		bzero(ctx->X, sizeof (ctx->X));
+	} else {		/* here to pre-process a key */
+
+		Skein_assert(sizeof (cfg.b) >= sizeof (ctx->X));
+		/* do a mini-Init right here */
+		/* set output hash bit count = state size */
+		ctx->h.hashBitLen = 8 * sizeof (ctx->X);
+		/* set tweaks: T0 = 0; T1 = KEY type */
+		Skein_Start_New_Type(ctx, KEY);
+		/* zero the initial chaining variables */
+		bzero(ctx->X, sizeof (ctx->X));
+		(void) Skein_512_Update(ctx, key, keyBytes); /* hash the key */
+		/* put result into cfg.b[] */
+		(void) Skein_512_Final_Pad(ctx, cfg.b);
+		/* copy over into ctx->X[] */
+		bcopy(cfg.b, ctx->X, sizeof (cfg.b));
+#if	SKEIN_NEED_SWAP
+		{
+			uint_t i;
+			/* convert key bytes to context words */
+			for (i = 0; i < SKEIN_512_STATE_WORDS; i++)
+				ctx->X[i] = Skein_Swap64(ctx->X[i]);
+		}
+#endif
+	}
+	/*
+	 * build/process the config block, type == CONFIG (could be
+	 * precomputed for each key)
+	 */
+	ctx->h.hashBitLen = hashBitLen;	/* output hash bit count */
+	Skein_Start_New_Type(ctx, CFG_FINAL);
+
+	bzero(&cfg.w, sizeof (cfg.w));	/* pre-pad cfg.w[] with zeroes */
+	cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+	cfg.w[1] = Skein_Swap64(hashBitLen);	/* hash result length in bits */
+	/* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+	cfg.w[2] = Skein_Swap64(treeInfo);
+
+	Skein_Show_Key(512, &ctx->h, key, keyBytes);
+
+	/* compute the initial chaining values from config block */
+	Skein_512_Process_Block(ctx, cfg.b, 1, SKEIN_CFG_STR_LEN);
+
+	/* The chaining vars ctx->X are now initialized */
+	/* Set up to process the data message portion of the hash (default) */
+	ctx->h.bCnt = 0;	/* buffer b[] starts out empty */
+	Skein_Start_New_Type(ctx, MSG);
+
+	return (SKEIN_SUCCESS);
+}
+
+/* process the input bytes */
+int
+Skein_512_Update(Skein_512_Ctxt_t *ctx, const uint8_t *msg, size_t msgByteCnt)
+{
+	size_t n;
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES, SKEIN_FAIL);
+
+	/* process full blocks, if any */
+	if (msgByteCnt + ctx->h.bCnt > SKEIN_512_BLOCK_BYTES) {
+		/* finish up any buffered message data */
+		if (ctx->h.bCnt) {
+			/* # bytes free in buffer b[] */
+			n = SKEIN_512_BLOCK_BYTES - ctx->h.bCnt;
+			if (n) {
+				/* check on our logic here */
+				Skein_assert(n < msgByteCnt);
+				bcopy(msg, &ctx->b[ctx->h.bCnt], n);
+				msgByteCnt -= n;
+				msg += n;
+				ctx->h.bCnt += n;
+			}
+			Skein_assert(ctx->h.bCnt == SKEIN_512_BLOCK_BYTES);
+			Skein_512_Process_Block(ctx, ctx->b, 1,
+			    SKEIN_512_BLOCK_BYTES);
+			ctx->h.bCnt = 0;
+		}
+		/*
+		 * now process any remaining full blocks, directly from input
+		 * message data
+		 */
+		if (msgByteCnt > SKEIN_512_BLOCK_BYTES) {
+			/* number of full blocks to process */
+			n = (msgByteCnt - 1) / SKEIN_512_BLOCK_BYTES;
+			Skein_512_Process_Block(ctx, msg, n,
+			    SKEIN_512_BLOCK_BYTES);
+			msgByteCnt -= n * SKEIN_512_BLOCK_BYTES;
+			msg += n * SKEIN_512_BLOCK_BYTES;
+		}
+		Skein_assert(ctx->h.bCnt == 0);
+	}
+
+	/* copy any remaining source message data bytes into b[] */
+	if (msgByteCnt) {
+		Skein_assert(msgByteCnt + ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES);
+		bcopy(msg, &ctx->b[ctx->h.bCnt], msgByteCnt);
+		ctx->h.bCnt += msgByteCnt;
+	}
+
+	return (SKEIN_SUCCESS);
+}
+
+/* finalize the hash computation and output the result */
+int
+Skein_512_Final(Skein_512_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	size_t i, n, byteCnt;
+	uint64_t X[SKEIN_512_STATE_WORDS];
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES, SKEIN_FAIL);
+
+	ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;	/* tag as the final block */
+	/* zero pad b[] if necessary */
+	if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES)
+		bzero(&ctx->b[ctx->h.bCnt],
+		    SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+
+	/* process the final block */
+	Skein_512_Process_Block(ctx, ctx->b, 1, ctx->h.bCnt);
+
+	/* now output the result */
+	/* total number of output bytes */
+	byteCnt = (ctx->h.hashBitLen + 7) >> 3;
+
+	/* run Threefish in "counter mode" to generate output */
+	/* zero out b[], so it can hold the counter */
+	bzero(ctx->b, sizeof (ctx->b));
+	/* keep a local copy of counter mode "key" */
+	bcopy(ctx->X, X, sizeof (X));
+	for (i = 0; i * SKEIN_512_BLOCK_BYTES < byteCnt; i++) {
+		/* build the counter block */
+		uint64_t tmp = Skein_Swap64((uint64_t)i);
+		bcopy(&tmp, ctx->b, sizeof (tmp));
+		Skein_Start_New_Type(ctx, OUT_FINAL);
+		/* run "counter mode" */
+		Skein_512_Process_Block(ctx, ctx->b, 1, sizeof (uint64_t));
+		/* number of output bytes left to go */
+		n = byteCnt - i * SKEIN_512_BLOCK_BYTES;
+		if (n >= SKEIN_512_BLOCK_BYTES)
+			n = SKEIN_512_BLOCK_BYTES;
+		Skein_Put64_LSB_First(hashVal + i * SKEIN_512_BLOCK_BYTES,
+		    ctx->X, n);	/* "output" the ctr mode bytes */
+		Skein_Show_Final(512, &ctx->h, n,
+		    hashVal + i * SKEIN_512_BLOCK_BYTES);
+		/* restore the counter mode key for next time */
+		bcopy(X, ctx->X, sizeof (X));
+	}
+	return (SKEIN_SUCCESS);
+}
+
+/* 1024-bit Skein */
+
+/* init the context for a straight hashing operation  */
+int
+Skein1024_Init(Skein1024_Ctxt_t *ctx, size_t hashBitLen)
+{
+	union {
+		uint8_t b[SKEIN1024_STATE_BYTES];
+		uint64_t w[SKEIN1024_STATE_WORDS];
+	} cfg;			/* config block */
+
+	Skein_Assert(hashBitLen > 0, SKEIN_BAD_HASHLEN);
+	ctx->h.hashBitLen = hashBitLen;	/* output hash bit count */
+
+	switch (hashBitLen) {	/* use pre-computed values, where available */
+#ifndef	SKEIN_NO_PRECOMP
+	case 512:
+		bcopy(SKEIN1024_IV_512, ctx->X, sizeof (ctx->X));
+		break;
+	case 384:
+		bcopy(SKEIN1024_IV_384, ctx->X, sizeof (ctx->X));
+		break;
+	case 1024:
+		bcopy(SKEIN1024_IV_1024, ctx->X, sizeof (ctx->X));
+		break;
+#endif
+	default:
+		/* here if there is no precomputed IV value available */
+		/*
+		 * build/process the config block, type == CONFIG (could be
+		 * precomputed)
+		 */
+		/* set tweaks: T0=0; T1=CFG | FINAL */
+		Skein_Start_New_Type(ctx, CFG_FINAL);
+
+		/* set the schema, version */
+		cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+		/* hash result length in bits */
+		cfg.w[1] = Skein_Swap64(hashBitLen);
+		cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+		/* zero pad config block */
+		bzero(&cfg.w[3], sizeof (cfg) - 3 * sizeof (cfg.w[0]));
+
+		/* compute the initial chaining values from config block */
+		/* zero the chaining variables */
+		bzero(ctx->X, sizeof (ctx->X));
+		Skein1024_Process_Block(ctx, cfg.b, 1, SKEIN_CFG_STR_LEN);
+		break;
+	}
+
+	/*
+	 * The chaining vars ctx->X are now initialized for the given
+	 * hashBitLen. Set up to process the data message portion of the hash
+	 * (default)
+	 */
+	Skein_Start_New_Type(ctx, MSG);	/* T0=0, T1= MSG type */
+
+	return (SKEIN_SUCCESS);
+}
+
+/* init the context for a MAC and/or tree hash operation */
+/*
+ * [identical to Skein1024_Init() when keyBytes == 0 &&
+ * treeInfo == SKEIN_CFG_TREE_INFO_SEQUENTIAL]
+ */
+int
+Skein1024_InitExt(Skein1024_Ctxt_t *ctx, size_t hashBitLen, uint64_t treeInfo,
+    const uint8_t *key, size_t keyBytes)
+{
+	union {
+		uint8_t b[SKEIN1024_STATE_BYTES];
+		uint64_t w[SKEIN1024_STATE_WORDS];
+	} cfg;			/* config block */
+
+	Skein_Assert(hashBitLen > 0, SKEIN_BAD_HASHLEN);
+	Skein_Assert(keyBytes == 0 || key != NULL, SKEIN_FAIL);
+
+	/* compute the initial chaining values ctx->X[], based on key */
+	if (keyBytes == 0) {	/* is there a key? */
+		/* no key: use all zeroes as key for config block */
+		bzero(ctx->X, sizeof (ctx->X));
+	} else {		/* here to pre-process a key */
+		Skein_assert(sizeof (cfg.b) >= sizeof (ctx->X));
+		/* do a mini-Init right here */
+		/* set output hash bit count = state size */
+		ctx->h.hashBitLen = 8 * sizeof (ctx->X);
+		/* set tweaks: T0 = 0; T1 = KEY type */
+		Skein_Start_New_Type(ctx, KEY);
+		/* zero the initial chaining variables */
+		bzero(ctx->X, sizeof (ctx->X));
+		(void) Skein1024_Update(ctx, key, keyBytes); /* hash the key */
+		/* put result into cfg.b[] */
+		(void) Skein1024_Final_Pad(ctx, cfg.b);
+		/* copy over into ctx->X[] */
+		bcopy(cfg.b, ctx->X, sizeof (cfg.b));
+#if	SKEIN_NEED_SWAP
+		{
+			uint_t i;
+			/* convert key bytes to context words */
+			for (i = 0; i < SKEIN1024_STATE_WORDS; i++)
+				ctx->X[i] = Skein_Swap64(ctx->X[i]);
+		}
+#endif
+	}
+	/*
+	 * build/process the config block, type == CONFIG (could be
+	 * precomputed for each key)
+	 */
+	ctx->h.hashBitLen = hashBitLen;	/* output hash bit count */
+	Skein_Start_New_Type(ctx, CFG_FINAL);
+
+	bzero(&cfg.w, sizeof (cfg.w));	/* pre-pad cfg.w[] with zeroes */
+	cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER);
+	/* hash result length in bits */
+	cfg.w[1] = Skein_Swap64(hashBitLen);
+	/* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+	cfg.w[2] = Skein_Swap64(treeInfo);
+
+	Skein_Show_Key(1024, &ctx->h, key, keyBytes);
+
+	/* compute the initial chaining values from config block */
+	Skein1024_Process_Block(ctx, cfg.b, 1, SKEIN_CFG_STR_LEN);
+
+	/* The chaining vars ctx->X are now initialized */
+	/* Set up to process the data message portion of the hash (default) */
+	ctx->h.bCnt = 0;	/* buffer b[] starts out empty */
+	Skein_Start_New_Type(ctx, MSG);
+
+	return (SKEIN_SUCCESS);
+}
+
+/* process the input bytes */
+int
+Skein1024_Update(Skein1024_Ctxt_t *ctx, const uint8_t *msg, size_t msgByteCnt)
+{
+	size_t n;
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES, SKEIN_FAIL);
+
+	/* process full blocks, if any */
+	if (msgByteCnt + ctx->h.bCnt > SKEIN1024_BLOCK_BYTES) {
+		/* finish up any buffered message data */
+		if (ctx->h.bCnt) {
+			/* # bytes free in buffer b[] */
+			n = SKEIN1024_BLOCK_BYTES - ctx->h.bCnt;
+			if (n) {
+				/* check on our logic here */
+				Skein_assert(n < msgByteCnt);
+				bcopy(msg, &ctx->b[ctx->h.bCnt], n);
+				msgByteCnt -= n;
+				msg += n;
+				ctx->h.bCnt += n;
+			}
+			Skein_assert(ctx->h.bCnt == SKEIN1024_BLOCK_BYTES);
+			Skein1024_Process_Block(ctx, ctx->b, 1,
+			    SKEIN1024_BLOCK_BYTES);
+			ctx->h.bCnt = 0;
+		}
+		/*
+		 * now process any remaining full blocks, directly from
+		 * input message data
+		 */
+		if (msgByteCnt > SKEIN1024_BLOCK_BYTES) {
+			/* number of full blocks to process */
+			n = (msgByteCnt - 1) / SKEIN1024_BLOCK_BYTES;
+			Skein1024_Process_Block(ctx, msg, n,
+			    SKEIN1024_BLOCK_BYTES);
+			msgByteCnt -= n * SKEIN1024_BLOCK_BYTES;
+			msg += n * SKEIN1024_BLOCK_BYTES;
+		}
+		Skein_assert(ctx->h.bCnt == 0);
+	}
+
+	/* copy any remaining source message data bytes into b[] */
+	if (msgByteCnt) {
+		Skein_assert(msgByteCnt + ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES);
+		bcopy(msg, &ctx->b[ctx->h.bCnt], msgByteCnt);
+		ctx->h.bCnt += msgByteCnt;
+	}
+
+	return (SKEIN_SUCCESS);
+}
+
+/* finalize the hash computation and output the result */
+int
+Skein1024_Final(Skein1024_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	size_t i, n, byteCnt;
+	uint64_t X[SKEIN1024_STATE_WORDS];
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES, SKEIN_FAIL);
+
+	ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;	/* tag as the final block */
+	/* zero pad b[] if necessary */
+	if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES)
+		bzero(&ctx->b[ctx->h.bCnt],
+		    SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+
+	/* process the final block */
+	Skein1024_Process_Block(ctx, ctx->b, 1, ctx->h.bCnt);
+
+	/* now output the result */
+	/* total number of output bytes */
+	byteCnt = (ctx->h.hashBitLen + 7) >> 3;
+
+	/* run Threefish in "counter mode" to generate output */
+	/* zero out b[], so it can hold the counter */
+	bzero(ctx->b, sizeof (ctx->b));
+	/* keep a local copy of counter mode "key" */
+	bcopy(ctx->X, X, sizeof (X));
+	for (i = 0; i * SKEIN1024_BLOCK_BYTES < byteCnt; i++) {
+		/* build the counter block */
+		uint64_t tmp = Skein_Swap64((uint64_t)i);
+		bcopy(&tmp, ctx->b, sizeof (tmp));
+		Skein_Start_New_Type(ctx, OUT_FINAL);
+		/* run "counter mode" */
+		Skein1024_Process_Block(ctx, ctx->b, 1, sizeof (uint64_t));
+		/* number of output bytes left to go */
+		n = byteCnt - i * SKEIN1024_BLOCK_BYTES;
+		if (n >= SKEIN1024_BLOCK_BYTES)
+			n = SKEIN1024_BLOCK_BYTES;
+		Skein_Put64_LSB_First(hashVal + i * SKEIN1024_BLOCK_BYTES,
+		    ctx->X, n);	/* "output" the ctr mode bytes */
+		Skein_Show_Final(1024, &ctx->h, n,
+		    hashVal + i * SKEIN1024_BLOCK_BYTES);
+		/* restore the counter mode key for next time */
+		bcopy(X, ctx->X, sizeof (X));
+	}
+	return (SKEIN_SUCCESS);
+}
+
+/* Functions to support MAC/tree hashing */
+/* (this code is identical for Optimized and Reference versions) */
+
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int
+Skein_256_Final_Pad(Skein_256_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES, SKEIN_FAIL);
+
+	ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;	/* tag as the final block */
+	/* zero pad b[] if necessary */
+	if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES)
+		bzero(&ctx->b[ctx->h.bCnt],
+		    SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+	/* process the final block */
+	Skein_256_Process_Block(ctx, ctx->b, 1, ctx->h.bCnt);
+
+	/* "output" the state bytes */
+	Skein_Put64_LSB_First(hashVal, ctx->X, SKEIN_256_BLOCK_BYTES);
+
+	return (SKEIN_SUCCESS);
+}
+
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int
+Skein_512_Final_Pad(Skein_512_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES, SKEIN_FAIL);
+
+	ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;	/* tag as the final block */
+	/* zero pad b[] if necessary */
+	if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES)
+		bzero(&ctx->b[ctx->h.bCnt],
+		    SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+	/* process the final block */
+	Skein_512_Process_Block(ctx, ctx->b, 1, ctx->h.bCnt);
+
+	/* "output" the state bytes */
+	Skein_Put64_LSB_First(hashVal, ctx->X, SKEIN_512_BLOCK_BYTES);
+
+	return (SKEIN_SUCCESS);
+}
+
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int
+Skein1024_Final_Pad(Skein1024_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES, SKEIN_FAIL);
+
+	/* tag as the final block */
+	ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;
+	/* zero pad b[] if necessary */
+	if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES)
+		bzero(&ctx->b[ctx->h.bCnt],
+		    SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+	/* process the final block */
+	Skein1024_Process_Block(ctx, ctx->b, 1, ctx->h.bCnt);
+
+	/* "output" the state bytes */
+	Skein_Put64_LSB_First(hashVal, ctx->X, SKEIN1024_BLOCK_BYTES);
+
+	return (SKEIN_SUCCESS);
+}
+
+#if	SKEIN_TREE_HASH
+/* just do the OUTPUT stage */
+int
+Skein_256_Output(Skein_256_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	size_t i, n, byteCnt;
+	uint64_t X[SKEIN_256_STATE_WORDS];
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES, SKEIN_FAIL);
+
+	/* now output the result */
+	/* total number of output bytes */
+	byteCnt = (ctx->h.hashBitLen + 7) >> 3;
+
+	/* run Threefish in "counter mode" to generate output */
+	/* zero out b[], so it can hold the counter */
+	bzero(ctx->b, sizeof (ctx->b));
+	/* keep a local copy of counter mode "key" */
+	bcopy(ctx->X, X, sizeof (X));
+	for (i = 0; i * SKEIN_256_BLOCK_BYTES < byteCnt; i++) {
+		/* build the counter block */
+		uint64_t tmp = Skein_Swap64((uint64_t)i);
+		bcopy(&tmp, ctx->b, sizeof (tmp));
+		Skein_Start_New_Type(ctx, OUT_FINAL);
+		/* run "counter mode" */
+		Skein_256_Process_Block(ctx, ctx->b, 1, sizeof (uint64_t));
+		/* number of output bytes left to go */
+		n = byteCnt - i * SKEIN_256_BLOCK_BYTES;
+		if (n >= SKEIN_256_BLOCK_BYTES)
+			n = SKEIN_256_BLOCK_BYTES;
+		Skein_Put64_LSB_First(hashVal + i * SKEIN_256_BLOCK_BYTES,
+		    ctx->X, n);	/* "output" the ctr mode bytes */
+		Skein_Show_Final(256, &ctx->h, n,
+		    hashVal + i * SKEIN_256_BLOCK_BYTES);
+		/* restore the counter mode key for next time */
+		bcopy(X, ctx->X, sizeof (X));
+	}
+	return (SKEIN_SUCCESS);
+}
+
+/* just do the OUTPUT stage */
+int
+Skein_512_Output(Skein_512_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	size_t i, n, byteCnt;
+	uint64_t X[SKEIN_512_STATE_WORDS];
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES, SKEIN_FAIL);
+
+	/* now output the result */
+	/* total number of output bytes */
+	byteCnt = (ctx->h.hashBitLen + 7) >> 3;
+
+	/* run Threefish in "counter mode" to generate output */
+	/* zero out b[], so it can hold the counter */
+	bzero(ctx->b, sizeof (ctx->b));
+	/* keep a local copy of counter mode "key" */
+	bcopy(ctx->X, X, sizeof (X));
+	for (i = 0; i * SKEIN_512_BLOCK_BYTES < byteCnt; i++) {
+		/* build the counter block */
+		uint64_t tmp = Skein_Swap64((uint64_t)i);
+		bcopy(&tmp, ctx->b, sizeof (tmp));
+		Skein_Start_New_Type(ctx, OUT_FINAL);
+		/* run "counter mode" */
+		Skein_512_Process_Block(ctx, ctx->b, 1, sizeof (uint64_t));
+		/* number of output bytes left to go */
+		n = byteCnt - i * SKEIN_512_BLOCK_BYTES;
+		if (n >= SKEIN_512_BLOCK_BYTES)
+			n = SKEIN_512_BLOCK_BYTES;
+		Skein_Put64_LSB_First(hashVal + i * SKEIN_512_BLOCK_BYTES,
+		    ctx->X, n);	/* "output" the ctr mode bytes */
+		Skein_Show_Final(256, &ctx->h, n,
+		    hashVal + i * SKEIN_512_BLOCK_BYTES);
+		/* restore the counter mode key for next time */
+		bcopy(X, ctx->X, sizeof (X));
+	}
+	return (SKEIN_SUCCESS);
+}
+
+/* just do the OUTPUT stage */
+int
+Skein1024_Output(Skein1024_Ctxt_t *ctx, uint8_t *hashVal)
+{
+	size_t i, n, byteCnt;
+	uint64_t X[SKEIN1024_STATE_WORDS];
+
+	/* catch uninitialized context */
+	Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES, SKEIN_FAIL);
+
+	/* now output the result */
+	/* total number of output bytes */
+	byteCnt = (ctx->h.hashBitLen + 7) >> 3;
+
+	/* run Threefish in "counter mode" to generate output */
+	/* zero out b[], so it can hold the counter */
+	bzero(ctx->b, sizeof (ctx->b));
+	/* keep a local copy of counter mode "key" */
+	bcopy(ctx->X, X, sizeof (X));
+	for (i = 0; i * SKEIN1024_BLOCK_BYTES < byteCnt; i++) {
+		/* build the counter block */
+		uint64_t tmp = Skein_Swap64((uint64_t)i);
+		bcopy(&tmp, ctx->b, sizeof (tmp));
+		Skein_Start_New_Type(ctx, OUT_FINAL);
+		/* run "counter mode" */
+		Skein1024_Process_Block(ctx, ctx->b, 1, sizeof (uint64_t));
+		/* number of output bytes left to go */
+		n = byteCnt - i * SKEIN1024_BLOCK_BYTES;
+		if (n >= SKEIN1024_BLOCK_BYTES)
+			n = SKEIN1024_BLOCK_BYTES;
+		Skein_Put64_LSB_First(hashVal + i * SKEIN1024_BLOCK_BYTES,
+		    ctx->X, n);	/* "output" the ctr mode bytes */
+		Skein_Show_Final(256, &ctx->h, n,
+		    hashVal + i * SKEIN1024_BLOCK_BYTES);
+		/* restore the counter mode key for next time */
+		bcopy(X, ctx->X, sizeof (X));
+	}
+	return (SKEIN_SUCCESS);
+}
+#endif
+
+#ifdef _KERNEL
+EXPORT_SYMBOL(Skein_512_Init);
+EXPORT_SYMBOL(Skein_512_InitExt);
+EXPORT_SYMBOL(Skein_512_Update);
+EXPORT_SYMBOL(Skein_512_Final);
+#endif
diff --git a/module/icp/algs/skein/skein_block.c b/module/icp/algs/skein/skein_block.c
new file mode 100644
index 000000000..d2e811963
--- /dev/null
+++ b/module/icp/algs/skein/skein_block.c
@@ -0,0 +1,793 @@
+/*
+ * Implementation of the Skein block functions.
+ * Source code author: Doug Whiting, 2008.
+ * This algorithm and source code is released to the public domain.
+ * Compile-time switches:
+ *  SKEIN_USE_ASM  -- set bits (256/512/1024) to select which
+ *                    versions use ASM code for block processing
+ *                    [default: use C for all block sizes]
+ */
+/* Copyright 2013 Doug Whiting. This code is released to the public domain. */
+
+#include <sys/skein.h>
+#include "skein_impl.h"
+#include <sys/isa_defs.h>	/* for _ILP32 */
+
+#ifndef	SKEIN_USE_ASM
+#define	SKEIN_USE_ASM	(0)	/* default is all C code (no ASM) */
+#endif
+
+#ifndef	SKEIN_LOOP
+/*
+ * The low-level checksum routines use a lot of stack space. On systems where
+ * small stacks frame are enforced (like 32-bit kernel builds), do not unroll
+ * checksum calculations to save stack space.
+ *
+ * Even with no loops unrolled, we still can exceed the 1k stack frame limit
+ * in Skein1024_Process_Block() (it hits 1272 bytes on ARM32).  We can
+ * safely ignore it though, since that the checksum functions will be called
+ * from a worker thread that won't be using much stack.  That's why we have
+ * the #pragma here to ignore the warning.
+ */
+#if defined(_ILP32) || defined(__powerpc)	/* Assume small stack */
+#pragma GCC diagnostic ignored "-Wframe-larger-than="
+/*
+ * We're running on 32-bit, don't unroll loops to save stack frame space
+ *
+ * Due to the ways the calculations on SKEIN_LOOP are done in
+ * Skein_*_Process_Block(), a value of 111 disables unrolling loops
+ * in any of those functions.
+ */
+#define	SKEIN_LOOP 111
+#else
+/* We're compiling with large stacks */
+#define	SKEIN_LOOP 001		/* default: unroll 256 and 512, but not 1024 */
+#endif
+#endif
+
+/* some useful definitions for code here */
+#define	BLK_BITS	(WCNT*64)
+#define	KW_TWK_BASE	(0)
+#define	KW_KEY_BASE	(3)
+#define	ks		(kw + KW_KEY_BASE)
+#define	ts		(kw + KW_TWK_BASE)
+
+/* no debugging in Illumos version */
+#define	DebugSaveTweak(ctx)
+
+/* Skein_256 */
+#if	!(SKEIN_USE_ASM & 256)
+
+void
+Skein_256_Process_Block(Skein_256_Ctxt_t *ctx, const uint8_t *blkPtr,
+    size_t blkCnt, size_t byteCntAdd)
+{				/* do it in C */
+	enum {
+		WCNT = SKEIN_256_STATE_WORDS
+	};
+#undef  RCNT
+#define	RCNT  (SKEIN_256_ROUNDS_TOTAL / 8)
+
+#ifdef	SKEIN_LOOP		/* configure how much to unroll the loop */
+#define	SKEIN_UNROLL_256 (((SKEIN_LOOP) / 100) % 10)
+#else
+#define	SKEIN_UNROLL_256 (0)
+#endif
+
+#if	SKEIN_UNROLL_256
+#if	(RCNT % SKEIN_UNROLL_256)
+#error "Invalid SKEIN_UNROLL_256"	/* sanity check on unroll count */
+#endif
+	size_t r;
+	/* key schedule words : chaining vars + tweak + "rotation" */
+	uint64_t kw[WCNT + 4 + RCNT * 2];
+#else
+	uint64_t kw[WCNT + 4];	/* key schedule words : chaining vars + tweak */
+#endif
+	/* local copy of context vars, for speed */
+	uint64_t X0, X1, X2, X3;
+	uint64_t w[WCNT];		/* local copy of input block */
+#ifdef	SKEIN_DEBUG
+	/* use for debugging (help compiler put Xn in registers) */
+	const uint64_t *Xptr[4];
+	Xptr[0] = &X0;
+	Xptr[1] = &X1;
+	Xptr[2] = &X2;
+	Xptr[3] = &X3;
+#endif
+	Skein_assert(blkCnt != 0);	/* never call with blkCnt == 0! */
+	ts[0] = ctx->h.T[0];
+	ts[1] = ctx->h.T[1];
+	do {
+		/*
+		 * this implementation only supports 2**64 input bytes
+		 * (no carry out here)
+		 */
+		ts[0] += byteCntAdd;	/* update processed length */
+
+		/* precompute the key schedule for this block */
+		ks[0] = ctx->X[0];
+		ks[1] = ctx->X[1];
+		ks[2] = ctx->X[2];
+		ks[3] = ctx->X[3];
+		ks[4] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^ SKEIN_KS_PARITY;
+
+		ts[2] = ts[0] ^ ts[1];
+
+		/* get input block in little-endian format */
+		Skein_Get64_LSB_First(w, blkPtr, WCNT);
+		DebugSaveTweak(ctx);
+		Skein_Show_Block(BLK_BITS, &ctx->h, ctx->X, blkPtr, w, ks, ts);
+
+		X0 = w[0] + ks[0];	/* do the first full key injection */
+		X1 = w[1] + ks[1] + ts[0];
+		X2 = w[2] + ks[2] + ts[1];
+		X3 = w[3] + ks[3];
+
+		Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INITIAL,
+		    Xptr);	/* show starting state values */
+
+		blkPtr += SKEIN_256_BLOCK_BYTES;
+
+		/* run the rounds */
+
+#define	Round256(p0, p1, p2, p3, ROT, rNum)                          \
+    X##p0 += X##p1; X##p1 = RotL_64(X##p1, ROT##_0); X##p1 ^= X##p0; \
+    X##p2 += X##p3; X##p3 = RotL_64(X##p3, ROT##_1); X##p3 ^= X##p2; \
+
+#if	SKEIN_UNROLL_256 == 0
+#define	R256(p0, p1, p2, p3, ROT, rNum)		/* fully unrolled */	\
+    Round256(p0, p1, p2, p3, ROT, rNum)					\
+    Skein_Show_R_Ptr(BLK_BITS, &ctx->h, rNum, Xptr);
+
+#define	I256(R)								\
+    X0 += ks[((R) + 1) % 5];	/* inject the key schedule value */	\
+    X1 += ks[((R) + 2) % 5] + ts[((R) + 1) % 3];			\
+    X2 += ks[((R) + 3) % 5] + ts[((R) + 2) % 3];			\
+    X3 += ks[((R) + 4) % 5] + (R) + 1;					\
+    Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, Xptr);
+#else				/* looping version */
+#define	R256(p0, p1, p2, p3, ROT, rNum)                             \
+    Round256(p0, p1, p2, p3, ROT, rNum)                             \
+    Skein_Show_R_Ptr(BLK_BITS, &ctx->h, 4 * (r - 1) + rNum, Xptr);
+
+#define	I256(R)								\
+	X0 += ks[r + (R) + 0];	/* inject the key schedule value */	\
+	X1 += ks[r + (R) + 1] + ts[r + (R) + 0];			\
+	X2 += ks[r + (R) + 2] + ts[r + (R) + 1];			\
+	X3 += ks[r + (R) + 3] + r + (R);				\
+	ks[r + (R) + 4] = ks[r + (R) - 1];   /* rotate key schedule */	\
+    ts[r + (R) + 2] = ts[r + (R) - 1];					\
+    Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, Xptr);
+
+		/* loop thru it */
+		for (r = 1; r < 2 * RCNT; r += 2 * SKEIN_UNROLL_256)
+#endif
+		{
+#define	R256_8_rounds(R)                         \
+	R256(0, 1, 2, 3, R_256_0, 8 * (R) + 1);  \
+	R256(0, 3, 2, 1, R_256_1, 8 * (R) + 2);  \
+	R256(0, 1, 2, 3, R_256_2, 8 * (R) + 3);  \
+	R256(0, 3, 2, 1, R_256_3, 8 * (R) + 4);  \
+	I256(2 * (R));                           \
+	R256(0, 1, 2, 3, R_256_4, 8 * (R) + 5);  \
+	R256(0, 3, 2, 1, R_256_5, 8 * (R) + 6);  \
+	R256(0, 1, 2, 3, R_256_6, 8 * (R) + 7);  \
+	R256(0, 3, 2, 1, R_256_7, 8 * (R) + 8);  \
+	I256(2 * (R) + 1);
+
+			R256_8_rounds(0);
+
+#define	R256_Unroll_R(NN) \
+	((SKEIN_UNROLL_256 == 0 && SKEIN_256_ROUNDS_TOTAL / 8 > (NN)) || \
+	(SKEIN_UNROLL_256 > (NN)))
+
+#if	R256_Unroll_R(1)
+			R256_8_rounds(1);
+#endif
+#if	R256_Unroll_R(2)
+			R256_8_rounds(2);
+#endif
+#if	R256_Unroll_R(3)
+			R256_8_rounds(3);
+#endif
+#if	R256_Unroll_R(4)
+			R256_8_rounds(4);
+#endif
+#if	R256_Unroll_R(5)
+			R256_8_rounds(5);
+#endif
+#if	R256_Unroll_R(6)
+			R256_8_rounds(6);
+#endif
+#if	R256_Unroll_R(7)
+			R256_8_rounds(7);
+#endif
+#if	R256_Unroll_R(8)
+			R256_8_rounds(8);
+#endif
+#if	R256_Unroll_R(9)
+			R256_8_rounds(9);
+#endif
+#if	R256_Unroll_R(10)
+			R256_8_rounds(10);
+#endif
+#if	R256_Unroll_R(11)
+			R256_8_rounds(11);
+#endif
+#if	R256_Unroll_R(12)
+			R256_8_rounds(12);
+#endif
+#if	R256_Unroll_R(13)
+			R256_8_rounds(13);
+#endif
+#if	R256_Unroll_R(14)
+			R256_8_rounds(14);
+#endif
+#if	(SKEIN_UNROLL_256 > 14)
+#error  "need more unrolling in Skein_256_Process_Block"
+#endif
+		}
+		/*
+		 * do the final "feedforward" xor, update context chaining vars
+		 */
+		ctx->X[0] = X0 ^ w[0];
+		ctx->X[1] = X1 ^ w[1];
+		ctx->X[2] = X2 ^ w[2];
+		ctx->X[3] = X3 ^ w[3];
+
+		Skein_Show_Round(BLK_BITS, &ctx->h, SKEIN_RND_FEED_FWD, ctx->X);
+
+		ts[1] &= ~SKEIN_T1_FLAG_FIRST;
+	}
+	while (--blkCnt);
+	ctx->h.T[0] = ts[0];
+	ctx->h.T[1] = ts[1];
+}
+
+#if	defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t
+Skein_256_Process_Block_CodeSize(void)
+{
+	return ((uint8_t *)Skein_256_Process_Block_CodeSize) -
+	    ((uint8_t *)Skein_256_Process_Block);
+}
+
+uint_t
+Skein_256_Unroll_Cnt(void)
+{
+	return (SKEIN_UNROLL_256);
+}
+#endif
+#endif
+
+/* Skein_512 */
+#if	!(SKEIN_USE_ASM & 512)
+void
+Skein_512_Process_Block(Skein_512_Ctxt_t *ctx, const uint8_t *blkPtr,
+    size_t blkCnt, size_t byteCntAdd)
+{				/* do it in C */
+	enum {
+		WCNT = SKEIN_512_STATE_WORDS
+	};
+#undef  RCNT
+#define	RCNT  (SKEIN_512_ROUNDS_TOTAL / 8)
+
+#ifdef	SKEIN_LOOP		/* configure how much to unroll the loop */
+#define	SKEIN_UNROLL_512 (((SKEIN_LOOP) / 10) % 10)
+#else
+#define	SKEIN_UNROLL_512 (0)
+#endif
+
+#if	SKEIN_UNROLL_512
+#if	(RCNT % SKEIN_UNROLL_512)
+#error "Invalid SKEIN_UNROLL_512"	/* sanity check on unroll count */
+#endif
+	size_t r;
+	/* key schedule words : chaining vars + tweak + "rotation" */
+	uint64_t kw[WCNT + 4 + RCNT * 2];
+#else
+	uint64_t kw[WCNT + 4];	/* key schedule words : chaining vars + tweak */
+#endif
+	/* local copy of vars, for speed */
+	uint64_t X0, X1, X2, X3, X4, X5, X6, X7;
+	uint64_t w[WCNT];		/* local copy of input block */
+#ifdef	SKEIN_DEBUG
+	/* use for debugging (help compiler put Xn in registers) */
+	const uint64_t *Xptr[8];
+	Xptr[0] = &X0;
+	Xptr[1] = &X1;
+	Xptr[2] = &X2;
+	Xptr[3] = &X3;
+	Xptr[4] = &X4;
+	Xptr[5] = &X5;
+	Xptr[6] = &X6;
+	Xptr[7] = &X7;
+#endif
+
+	Skein_assert(blkCnt != 0);	/* never call with blkCnt == 0! */
+	ts[0] = ctx->h.T[0];
+	ts[1] = ctx->h.T[1];
+	do {
+		/*
+		 * this implementation only supports 2**64 input bytes
+		 * (no carry out here)
+		 */
+		ts[0] += byteCntAdd;	/* update processed length */
+
+		/* precompute the key schedule for this block */
+		ks[0] = ctx->X[0];
+		ks[1] = ctx->X[1];
+		ks[2] = ctx->X[2];
+		ks[3] = ctx->X[3];
+		ks[4] = ctx->X[4];
+		ks[5] = ctx->X[5];
+		ks[6] = ctx->X[6];
+		ks[7] = ctx->X[7];
+		ks[8] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
+		    ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^ SKEIN_KS_PARITY;
+
+		ts[2] = ts[0] ^ ts[1];
+
+		/* get input block in little-endian format */
+		Skein_Get64_LSB_First(w, blkPtr, WCNT);
+		DebugSaveTweak(ctx);
+		Skein_Show_Block(BLK_BITS, &ctx->h, ctx->X, blkPtr, w, ks, ts);
+
+		X0 = w[0] + ks[0];	/* do the first full key injection */
+		X1 = w[1] + ks[1];
+		X2 = w[2] + ks[2];
+		X3 = w[3] + ks[3];
+		X4 = w[4] + ks[4];
+		X5 = w[5] + ks[5] + ts[0];
+		X6 = w[6] + ks[6] + ts[1];
+		X7 = w[7] + ks[7];
+
+		blkPtr += SKEIN_512_BLOCK_BYTES;
+
+		Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INITIAL,
+		    Xptr);
+		/* run the rounds */
+#define	Round512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, rNum)		\
+	X##p0 += X##p1; X##p1 = RotL_64(X##p1, ROT##_0); X##p1 ^= X##p0;\
+	X##p2 += X##p3; X##p3 = RotL_64(X##p3, ROT##_1); X##p3 ^= X##p2;\
+	X##p4 += X##p5; X##p5 = RotL_64(X##p5, ROT##_2); X##p5 ^= X##p4;\
+	X##p6 += X##p7; X##p7 = RotL_64(X##p7, ROT##_3); X##p7 ^= X##p6;
+
+#if	SKEIN_UNROLL_512 == 0
+#define	R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, rNum)	/* unrolled */	\
+	Round512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, rNum)		\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, rNum, Xptr);
+
+#define	I512(R)								\
+	X0 += ks[((R) + 1) % 9];	/* inject the key schedule value */\
+	X1 += ks[((R) + 2) % 9];					\
+	X2 += ks[((R) + 3) % 9];					\
+	X3 += ks[((R) + 4) % 9];					\
+	X4 += ks[((R) + 5) % 9];					\
+	X5 += ks[((R) + 6) % 9] + ts[((R) + 1) % 3];			\
+	X6 += ks[((R) + 7) % 9] + ts[((R) + 2) % 3];			\
+	X7 += ks[((R) + 8) % 9] + (R) + 1;				\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, Xptr);
+#else				/* looping version */
+#define	R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, rNum)			\
+	Round512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, rNum)		\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, 4 * (r - 1) + rNum, Xptr);
+
+#define	I512(R)								\
+	X0 += ks[r + (R) + 0];	/* inject the key schedule value */	\
+	X1 += ks[r + (R) + 1];						\
+	X2 += ks[r + (R) + 2];						\
+	X3 += ks[r + (R) + 3];						\
+	X4 += ks[r + (R) + 4];						\
+	X5 += ks[r + (R) + 5] + ts[r + (R) + 0];			\
+	X6 += ks[r + (R) + 6] + ts[r + (R) + 1];			\
+	X7 += ks[r + (R) + 7] + r + (R);				\
+	ks[r + (R)+8] = ks[r + (R) - 1];	/* rotate key schedule */\
+	ts[r + (R)+2] = ts[r + (R) - 1];				\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, Xptr);
+
+		/* loop thru it */
+		for (r = 1; r < 2 * RCNT; r += 2 * SKEIN_UNROLL_512)
+#endif				/* end of looped code definitions */
+		{
+#define	R512_8_rounds(R)	/* do 8 full rounds */			\
+	R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_0, 8 * (R) + 1);		\
+	R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_1, 8 * (R) + 2);		\
+	R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_2, 8 * (R) + 3);		\
+	R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_3, 8 * (R) + 4);		\
+	I512(2 * (R));							\
+	R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_4, 8 * (R) + 5);		\
+	R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_5, 8 * (R) + 6);		\
+	R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_6, 8 * (R) + 7);		\
+	R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_7, 8 * (R) + 8);		\
+	I512(2*(R) + 1);		/* and key injection */
+
+			R512_8_rounds(0);
+
+#define	R512_Unroll_R(NN) \
+	((SKEIN_UNROLL_512 == 0 && SKEIN_512_ROUNDS_TOTAL / 8 > (NN)) || \
+	(SKEIN_UNROLL_512 > (NN)))
+
+#if	R512_Unroll_R(1)
+			R512_8_rounds(1);
+#endif
+#if	R512_Unroll_R(2)
+			R512_8_rounds(2);
+#endif
+#if	R512_Unroll_R(3)
+			R512_8_rounds(3);
+#endif
+#if	R512_Unroll_R(4)
+			R512_8_rounds(4);
+#endif
+#if	R512_Unroll_R(5)
+			R512_8_rounds(5);
+#endif
+#if	R512_Unroll_R(6)
+			R512_8_rounds(6);
+#endif
+#if	R512_Unroll_R(7)
+			R512_8_rounds(7);
+#endif
+#if	R512_Unroll_R(8)
+			R512_8_rounds(8);
+#endif
+#if	R512_Unroll_R(9)
+			R512_8_rounds(9);
+#endif
+#if	R512_Unroll_R(10)
+			R512_8_rounds(10);
+#endif
+#if	R512_Unroll_R(11)
+			R512_8_rounds(11);
+#endif
+#if	R512_Unroll_R(12)
+			R512_8_rounds(12);
+#endif
+#if	R512_Unroll_R(13)
+			R512_8_rounds(13);
+#endif
+#if	R512_Unroll_R(14)
+			R512_8_rounds(14);
+#endif
+#if	(SKEIN_UNROLL_512 > 14)
+#error "need more unrolling in Skein_512_Process_Block"
+#endif
+		}
+
+		/*
+		 * do the final "feedforward" xor, update context chaining vars
+		 */
+		ctx->X[0] = X0 ^ w[0];
+		ctx->X[1] = X1 ^ w[1];
+		ctx->X[2] = X2 ^ w[2];
+		ctx->X[3] = X3 ^ w[3];
+		ctx->X[4] = X4 ^ w[4];
+		ctx->X[5] = X5 ^ w[5];
+		ctx->X[6] = X6 ^ w[6];
+		ctx->X[7] = X7 ^ w[7];
+		Skein_Show_Round(BLK_BITS, &ctx->h, SKEIN_RND_FEED_FWD, ctx->X);
+
+		ts[1] &= ~SKEIN_T1_FLAG_FIRST;
+	}
+	while (--blkCnt);
+	ctx->h.T[0] = ts[0];
+	ctx->h.T[1] = ts[1];
+}
+
+#if	defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t
+Skein_512_Process_Block_CodeSize(void)
+{
+	return ((uint8_t *)Skein_512_Process_Block_CodeSize) -
+	    ((uint8_t *)Skein_512_Process_Block);
+}
+
+uint_t
+Skein_512_Unroll_Cnt(void)
+{
+	return (SKEIN_UNROLL_512);
+}
+#endif
+#endif
+
+/*  Skein1024 */
+#if	!(SKEIN_USE_ASM & 1024)
+void
+Skein1024_Process_Block(Skein1024_Ctxt_t *ctx, const uint8_t *blkPtr,
+    size_t blkCnt, size_t byteCntAdd)
+{
+	/* do it in C, always looping (unrolled is bigger AND slower!) */
+	enum {
+		WCNT = SKEIN1024_STATE_WORDS
+	};
+#undef  RCNT
+#define	RCNT  (SKEIN1024_ROUNDS_TOTAL/8)
+
+#ifdef	SKEIN_LOOP		/* configure how much to unroll the loop */
+#define	SKEIN_UNROLL_1024 ((SKEIN_LOOP)%10)
+#else
+#define	SKEIN_UNROLL_1024 (0)
+#endif
+
+#if	(SKEIN_UNROLL_1024 != 0)
+#if	(RCNT % SKEIN_UNROLL_1024)
+#error "Invalid SKEIN_UNROLL_1024"	/* sanity check on unroll count */
+#endif
+	size_t r;
+	/* key schedule words : chaining vars + tweak + "rotation" */
+	uint64_t kw[WCNT + 4 + RCNT * 2];
+#else
+	uint64_t kw[WCNT + 4];	/* key schedule words : chaining vars + tweak */
+#endif
+
+	/* local copy of vars, for speed */
+	uint64_t X00, X01, X02, X03, X04, X05, X06, X07, X08, X09, X10, X11,
+	    X12, X13, X14, X15;
+	uint64_t w[WCNT];		/* local copy of input block */
+#ifdef	SKEIN_DEBUG
+	/* use for debugging (help compiler put Xn in registers) */
+	const uint64_t *Xptr[16];
+	Xptr[0] = &X00;
+	Xptr[1] = &X01;
+	Xptr[2] = &X02;
+	Xptr[3] = &X03;
+	Xptr[4] = &X04;
+	Xptr[5] = &X05;
+	Xptr[6] = &X06;
+	Xptr[7] = &X07;
+	Xptr[8] = &X08;
+	Xptr[9] = &X09;
+	Xptr[10] = &X10;
+	Xptr[11] = &X11;
+	Xptr[12] = &X12;
+	Xptr[13] = &X13;
+	Xptr[14] = &X14;
+	Xptr[15] = &X15;
+#endif
+
+	Skein_assert(blkCnt != 0);	/* never call with blkCnt == 0! */
+	ts[0] = ctx->h.T[0];
+	ts[1] = ctx->h.T[1];
+	do {
+		/*
+		 * this implementation only supports 2**64 input bytes
+		 * (no carry out here)
+		 */
+		ts[0] += byteCntAdd;	/* update processed length */
+
+		/* precompute the key schedule for this block */
+		ks[0] = ctx->X[0];
+		ks[1] = ctx->X[1];
+		ks[2] = ctx->X[2];
+		ks[3] = ctx->X[3];
+		ks[4] = ctx->X[4];
+		ks[5] = ctx->X[5];
+		ks[6] = ctx->X[6];
+		ks[7] = ctx->X[7];
+		ks[8] = ctx->X[8];
+		ks[9] = ctx->X[9];
+		ks[10] = ctx->X[10];
+		ks[11] = ctx->X[11];
+		ks[12] = ctx->X[12];
+		ks[13] = ctx->X[13];
+		ks[14] = ctx->X[14];
+		ks[15] = ctx->X[15];
+		ks[16] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
+		    ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^
+		    ks[8] ^ ks[9] ^ ks[10] ^ ks[11] ^
+		    ks[12] ^ ks[13] ^ ks[14] ^ ks[15] ^ SKEIN_KS_PARITY;
+
+		ts[2] = ts[0] ^ ts[1];
+
+		/* get input block in little-endian format */
+		Skein_Get64_LSB_First(w, blkPtr, WCNT);
+		DebugSaveTweak(ctx);
+		Skein_Show_Block(BLK_BITS, &ctx->h, ctx->X, blkPtr, w, ks, ts);
+
+		X00 = w[0] + ks[0];	/* do the first full key injection */
+		X01 = w[1] + ks[1];
+		X02 = w[2] + ks[2];
+		X03 = w[3] + ks[3];
+		X04 = w[4] + ks[4];
+		X05 = w[5] + ks[5];
+		X06 = w[6] + ks[6];
+		X07 = w[7] + ks[7];
+		X08 = w[8] + ks[8];
+		X09 = w[9] + ks[9];
+		X10 = w[10] + ks[10];
+		X11 = w[11] + ks[11];
+		X12 = w[12] + ks[12];
+		X13 = w[13] + ks[13] + ts[0];
+		X14 = w[14] + ks[14] + ts[1];
+		X15 = w[15] + ks[15];
+
+		Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INITIAL,
+		    Xptr);
+
+#define	Round1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC,	\
+	pD, pE, pF, ROT, rNum)						\
+	X##p0 += X##p1; X##p1 = RotL_64(X##p1, ROT##_0); X##p1 ^= X##p0;\
+	X##p2 += X##p3; X##p3 = RotL_64(X##p3, ROT##_1); X##p3 ^= X##p2;\
+	X##p4 += X##p5; X##p5 = RotL_64(X##p5, ROT##_2); X##p5 ^= X##p4;\
+	X##p6 += X##p7; X##p7 = RotL_64(X##p7, ROT##_3); X##p7 ^= X##p6;\
+	X##p8 += X##p9; X##p9 = RotL_64(X##p9, ROT##_4); X##p9 ^= X##p8;\
+	X##pA += X##pB; X##pB = RotL_64(X##pB, ROT##_5); X##pB ^= X##pA;\
+	X##pC += X##pD; X##pD = RotL_64(X##pD, ROT##_6); X##pD ^= X##pC;\
+	X##pE += X##pF; X##pF = RotL_64(X##pF, ROT##_7); X##pF ^= X##pE;
+
+#if	SKEIN_UNROLL_1024 == 0
+#define	R1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD,	\
+	pE, pF, ROT, rn)						\
+	Round1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC,	\
+	pD, pE, pF, ROT, rn)						\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, rn, Xptr);
+
+#define	I1024(R)							\
+	X00 += ks[((R) + 1) % 17];	/* inject the key schedule value */\
+	X01 += ks[((R) + 2) % 17];					\
+	X02 += ks[((R) + 3) % 17];					\
+	X03 += ks[((R) + 4) % 17];					\
+	X04 += ks[((R) + 5) % 17];					\
+	X05 += ks[((R) + 6) % 17];					\
+	X06 += ks[((R) + 7) % 17];					\
+	X07 += ks[((R) + 8) % 17];					\
+	X08 += ks[((R) + 9) % 17];					\
+	X09 += ks[((R) + 10) % 17];					\
+	X10 += ks[((R) + 11) % 17];					\
+	X11 += ks[((R) + 12) % 17];					\
+	X12 += ks[((R) + 13) % 17];					\
+	X13 += ks[((R) + 14) % 17] + ts[((R) + 1) % 3];			\
+	X14 += ks[((R) + 15) % 17] + ts[((R) + 2) % 3];			\
+	X15 += ks[((R) + 16) % 17] + (R) +1;				\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, Xptr);
+#else				/* looping version */
+#define	R1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD,	\
+	pE, pF, ROT, rn)						\
+	Round1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC,	\
+	pD, pE, pF, ROT, rn)						\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, 4 * (r - 1) + rn, Xptr);
+
+#define	I1024(R)							\
+	X00 += ks[r + (R) + 0];	/* inject the key schedule value */	\
+	X01 += ks[r + (R) + 1];						\
+	X02 += ks[r + (R) + 2];						\
+	X03 += ks[r + (R) + 3];						\
+	X04 += ks[r + (R) + 4];						\
+	X05 += ks[r + (R) + 5];						\
+	X06 += ks[r + (R) + 6];						\
+	X07 += ks[r + (R) + 7];						\
+	X08 += ks[r + (R) + 8];						\
+	X09 += ks[r + (R) + 9];						\
+	X10 += ks[r + (R) + 10];					\
+	X11 += ks[r + (R) + 11];					\
+	X12 += ks[r + (R) + 12];					\
+	X13 += ks[r + (R) + 13] + ts[r + (R) + 0];			\
+	X14 += ks[r + (R) + 14] + ts[r + (R) + 1];			\
+	X15 += ks[r + (R) + 15] +  r + (R);				\
+	ks[r + (R) + 16] = ks[r + (R) - 1];	/* rotate key schedule */\
+	ts[r + (R) + 2] = ts[r + (R) - 1];				\
+	Skein_Show_R_Ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, Xptr);
+
+		/* loop thru it */
+		for (r = 1; r <= 2 * RCNT; r += 2 * SKEIN_UNROLL_1024)
+#endif
+		{
+#define	R1024_8_rounds(R)	/* do 8 full rounds */			\
+	R1024(00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13,	\
+	    14, 15, R1024_0, 8 * (R) + 1);				\
+	R1024(00, 09, 02, 13, 06, 11, 04, 15, 10, 07, 12, 03, 14, 05,	\
+	    08, 01, R1024_1, 8 * (R) + 2);				\
+	R1024(00, 07, 02, 05, 04, 03, 06, 01, 12, 15, 14, 13, 08, 11,	\
+	    10, 09, R1024_2, 8 * (R) + 3);				\
+	R1024(00, 15, 02, 11, 06, 13, 04, 09, 14, 01, 08, 05, 10, 03,	\
+	    12, 07, R1024_3, 8 * (R) + 4);				\
+	I1024(2 * (R));							\
+	R1024(00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13,	\
+	    14, 15, R1024_4, 8 * (R) + 5);				\
+	R1024(00, 09, 02, 13, 06, 11, 04, 15, 10, 07, 12, 03, 14, 05,	\
+	    08, 01, R1024_5, 8 * (R) + 6);				\
+	R1024(00, 07, 02, 05, 04, 03, 06, 01, 12, 15, 14, 13, 08, 11,	\
+	    10, 09, R1024_6, 8 * (R) + 7);				\
+	R1024(00, 15, 02, 11, 06, 13, 04, 09, 14, 01, 08, 05, 10, 03,	\
+	    12, 07, R1024_7, 8 * (R) + 8);				\
+	I1024(2 * (R) + 1);
+
+			R1024_8_rounds(0);
+
+#define	R1024_Unroll_R(NN)						\
+	((SKEIN_UNROLL_1024 == 0 && SKEIN1024_ROUNDS_TOTAL/8 > (NN)) ||	\
+	(SKEIN_UNROLL_1024 > (NN)))
+
+#if	R1024_Unroll_R(1)
+			R1024_8_rounds(1);
+#endif
+#if	R1024_Unroll_R(2)
+			R1024_8_rounds(2);
+#endif
+#if	R1024_Unroll_R(3)
+			R1024_8_rounds(3);
+#endif
+#if	R1024_Unroll_R(4)
+			R1024_8_rounds(4);
+#endif
+#if	R1024_Unroll_R(5)
+			R1024_8_rounds(5);
+#endif
+#if	R1024_Unroll_R(6)
+			R1024_8_rounds(6);
+#endif
+#if	R1024_Unroll_R(7)
+			R1024_8_rounds(7);
+#endif
+#if	R1024_Unroll_R(8)
+			R1024_8_rounds(8);
+#endif
+#if	R1024_Unroll_R(9)
+			R1024_8_rounds(9);
+#endif
+#if	R1024_Unroll_R(10)
+			R1024_8_rounds(10);
+#endif
+#if	R1024_Unroll_R(11)
+			R1024_8_rounds(11);
+#endif
+#if	R1024_Unroll_R(12)
+			R1024_8_rounds(12);
+#endif
+#if	R1024_Unroll_R(13)
+			R1024_8_rounds(13);
+#endif
+#if	R1024_Unroll_R(14)
+			R1024_8_rounds(14);
+#endif
+#if	(SKEIN_UNROLL_1024 > 14)
+#error  "need more unrolling in Skein_1024_Process_Block"
+#endif
+		}
+		/*
+		 * do the final "feedforward" xor, update context chaining vars
+		 */
+
+		ctx->X[0] = X00 ^ w[0];
+		ctx->X[1] = X01 ^ w[1];
+		ctx->X[2] = X02 ^ w[2];
+		ctx->X[3] = X03 ^ w[3];
+		ctx->X[4] = X04 ^ w[4];
+		ctx->X[5] = X05 ^ w[5];
+		ctx->X[6] = X06 ^ w[6];
+		ctx->X[7] = X07 ^ w[7];
+		ctx->X[8] = X08 ^ w[8];
+		ctx->X[9] = X09 ^ w[9];
+		ctx->X[10] = X10 ^ w[10];
+		ctx->X[11] = X11 ^ w[11];
+		ctx->X[12] = X12 ^ w[12];
+		ctx->X[13] = X13 ^ w[13];
+		ctx->X[14] = X14 ^ w[14];
+		ctx->X[15] = X15 ^ w[15];
+
+		Skein_Show_Round(BLK_BITS, &ctx->h, SKEIN_RND_FEED_FWD, ctx->X);
+
+		ts[1] &= ~SKEIN_T1_FLAG_FIRST;
+		blkPtr += SKEIN1024_BLOCK_BYTES;
+	} while (--blkCnt);
+	ctx->h.T[0] = ts[0];
+	ctx->h.T[1] = ts[1];
+}
+
+#if	defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t
+Skein1024_Process_Block_CodeSize(void)
+{
+	return ((uint8_t *)Skein1024_Process_Block_CodeSize) -
+	    ((uint8_t *)Skein1024_Process_Block);
+}
+
+uint_t
+Skein1024_Unroll_Cnt(void)
+{
+	return (SKEIN_UNROLL_1024);
+}
+#endif
+#endif
diff --git a/module/icp/algs/skein/skein_impl.h b/module/icp/algs/skein/skein_impl.h
new file mode 100644
index 000000000..e83a06971
--- /dev/null
+++ b/module/icp/algs/skein/skein_impl.h
@@ -0,0 +1,289 @@
+/*
+ * Internal definitions for Skein hashing.
+ * Source code author: Doug Whiting, 2008.
+ * This algorithm and source code is released to the public domain.
+ *
+ * The following compile-time switches may be defined to control some
+ * tradeoffs between speed, code size, error checking, and security.
+ *
+ * The "default" note explains what happens when the switch is not defined.
+ *
+ *  SKEIN_DEBUG            -- make callouts from inside Skein code
+ *                            to examine/display intermediate values.
+ *                            [default: no callouts (no overhead)]
+ *
+ *  SKEIN_ERR_CHECK        -- how error checking is handled inside Skein
+ *                            code. If not defined, most error checking
+ *                            is disabled (for performance). Otherwise,
+ *                            the switch value is interpreted as:
+ *                                0: use assert()      to flag errors
+ *                                1: return SKEIN_FAIL to flag errors
+ */
+/* Copyright 2013 Doug Whiting. This code is released to the public domain. */
+
+#ifndef	_SKEIN_IMPL_H_
+#define	_SKEIN_IMPL_H_
+
+#include <sys/skein.h>
+#include "skein_impl.h"
+#include "skein_port.h"
+
+/* determine where we can get bcopy/bzero declarations */
+#ifdef	_KERNEL
+#include <sys/systm.h>
+#else
+#include <strings.h>
+#endif
+
+/*
+ * "Internal" Skein definitions
+ *    -- not needed for sequential hashing API, but will be
+ *           helpful for other uses of Skein (e.g., tree hash mode).
+ *    -- included here so that they can be shared between
+ *           reference and optimized code.
+ */
+
+/* tweak word T[1]: bit field starting positions */
+/* offset 64 because it's the second word  */
+#define	SKEIN_T1_BIT(BIT)	((BIT) - 64)
+
+/* bits 112..118: level in hash tree */
+#define	SKEIN_T1_POS_TREE_LVL	SKEIN_T1_BIT(112)
+/* bit  119: partial final input byte */
+#define	SKEIN_T1_POS_BIT_PAD	SKEIN_T1_BIT(119)
+/* bits 120..125: type field */
+#define	SKEIN_T1_POS_BLK_TYPE	SKEIN_T1_BIT(120)
+/* bits 126: first block flag */
+#define	SKEIN_T1_POS_FIRST	SKEIN_T1_BIT(126)
+/* bit  127: final block flag */
+#define	SKEIN_T1_POS_FINAL	SKEIN_T1_BIT(127)
+
+/* tweak word T[1]: flag bit definition(s) */
+#define	SKEIN_T1_FLAG_FIRST	(((uint64_t)1) << SKEIN_T1_POS_FIRST)
+#define	SKEIN_T1_FLAG_FINAL	(((uint64_t)1) << SKEIN_T1_POS_FINAL)
+#define	SKEIN_T1_FLAG_BIT_PAD	(((uint64_t)1) << SKEIN_T1_POS_BIT_PAD)
+
+/* tweak word T[1]: tree level bit field mask */
+#define	SKEIN_T1_TREE_LVL_MASK	(((uint64_t)0x7F) << SKEIN_T1_POS_TREE_LVL)
+#define	SKEIN_T1_TREE_LEVEL(n)	(((uint64_t)(n)) << SKEIN_T1_POS_TREE_LVL)
+
+/* tweak word T[1]: block type field */
+#define	SKEIN_BLK_TYPE_KEY	(0)	/* key, for MAC and KDF */
+#define	SKEIN_BLK_TYPE_CFG	(4)	/* configuration block */
+#define	SKEIN_BLK_TYPE_PERS	(8)	/* personalization string */
+#define	SKEIN_BLK_TYPE_PK	(12)	/* public key (for signature hashing) */
+#define	SKEIN_BLK_TYPE_KDF	(16)	/* key identifier for KDF */
+#define	SKEIN_BLK_TYPE_NONCE	(20)	/* nonce for PRNG */
+#define	SKEIN_BLK_TYPE_MSG	(48)	/* message processing */
+#define	SKEIN_BLK_TYPE_OUT	(63)	/* output stage */
+#define	SKEIN_BLK_TYPE_MASK	(63)	/* bit field mask */
+
+#define	SKEIN_T1_BLK_TYPE(T)	\
+	(((uint64_t)(SKEIN_BLK_TYPE_##T)) << SKEIN_T1_POS_BLK_TYPE)
+/* key, for MAC and KDF */
+#define	SKEIN_T1_BLK_TYPE_KEY	SKEIN_T1_BLK_TYPE(KEY)
+/* configuration block */
+#define	SKEIN_T1_BLK_TYPE_CFG	SKEIN_T1_BLK_TYPE(CFG)
+/* personalization string */
+#define	SKEIN_T1_BLK_TYPE_PERS	SKEIN_T1_BLK_TYPE(PERS)
+/* public key (for digital signature hashing) */
+#define	SKEIN_T1_BLK_TYPE_PK	SKEIN_T1_BLK_TYPE(PK)
+/* key identifier for KDF */
+#define	SKEIN_T1_BLK_TYPE_KDF	SKEIN_T1_BLK_TYPE(KDF)
+/* nonce for PRNG */
+#define	SKEIN_T1_BLK_TYPE_NONCE	SKEIN_T1_BLK_TYPE(NONCE)
+/* message processing */
+#define	SKEIN_T1_BLK_TYPE_MSG	SKEIN_T1_BLK_TYPE(MSG)
+/* output stage */
+#define	SKEIN_T1_BLK_TYPE_OUT	SKEIN_T1_BLK_TYPE(OUT)
+/* field bit mask */
+#define	SKEIN_T1_BLK_TYPE_MASK	SKEIN_T1_BLK_TYPE(MASK)
+
+#define	SKEIN_T1_BLK_TYPE_CFG_FINAL	\
+	(SKEIN_T1_BLK_TYPE_CFG | SKEIN_T1_FLAG_FINAL)
+#define	SKEIN_T1_BLK_TYPE_OUT_FINAL	\
+	(SKEIN_T1_BLK_TYPE_OUT | SKEIN_T1_FLAG_FINAL)
+
+#define	SKEIN_VERSION		(1)
+
+#ifndef	SKEIN_ID_STRING_LE	/* allow compile-time personalization */
+#define	SKEIN_ID_STRING_LE	(0x33414853)	/* "SHA3" (little-endian) */
+#endif
+
+#define	SKEIN_MK_64(hi32, lo32)	((lo32) + (((uint64_t)(hi32)) << 32))
+#define	SKEIN_SCHEMA_VER	SKEIN_MK_64(SKEIN_VERSION, SKEIN_ID_STRING_LE)
+#define	SKEIN_KS_PARITY		SKEIN_MK_64(0x1BD11BDA, 0xA9FC1A22)
+
+#define	SKEIN_CFG_STR_LEN	(4*8)
+
+/* bit field definitions in config block treeInfo word */
+#define	SKEIN_CFG_TREE_LEAF_SIZE_POS	(0)
+#define	SKEIN_CFG_TREE_NODE_SIZE_POS	(8)
+#define	SKEIN_CFG_TREE_MAX_LEVEL_POS	(16)
+
+#define	SKEIN_CFG_TREE_LEAF_SIZE_MSK	\
+	(((uint64_t)0xFF) << SKEIN_CFG_TREE_LEAF_SIZE_POS)
+#define	SKEIN_CFG_TREE_NODE_SIZE_MSK	\
+	(((uint64_t)0xFF) << SKEIN_CFG_TREE_NODE_SIZE_POS)
+#define	SKEIN_CFG_TREE_MAX_LEVEL_MSK	\
+	(((uint64_t)0xFF) << SKEIN_CFG_TREE_MAX_LEVEL_POS)
+
+#define	SKEIN_CFG_TREE_INFO(leaf, node, maxLvl)			\
+	((((uint64_t)(leaf)) << SKEIN_CFG_TREE_LEAF_SIZE_POS) |	\
+	(((uint64_t)(node)) << SKEIN_CFG_TREE_NODE_SIZE_POS) |	\
+	(((uint64_t)(maxLvl)) << SKEIN_CFG_TREE_MAX_LEVEL_POS))
+
+/* use as treeInfo in InitExt() call for sequential processing */
+#define	SKEIN_CFG_TREE_INFO_SEQUENTIAL	SKEIN_CFG_TREE_INFO(0, 0, 0)
+
+/*
+ * Skein macros for getting/setting tweak words, etc.
+ * These are useful for partial input bytes, hash tree init/update, etc.
+ */
+#define	Skein_Get_Tweak(ctxPtr, TWK_NUM)	((ctxPtr)->h.T[TWK_NUM])
+#define	Skein_Set_Tweak(ctxPtr, TWK_NUM, tVal)		\
+	do {						\
+		(ctxPtr)->h.T[TWK_NUM] = (tVal);	\
+		_NOTE(CONSTCOND)			\
+	} while (0)
+
+#define	Skein_Get_T0(ctxPtr)		Skein_Get_Tweak(ctxPtr, 0)
+#define	Skein_Get_T1(ctxPtr)		Skein_Get_Tweak(ctxPtr, 1)
+#define	Skein_Set_T0(ctxPtr, T0)	Skein_Set_Tweak(ctxPtr, 0, T0)
+#define	Skein_Set_T1(ctxPtr, T1)	Skein_Set_Tweak(ctxPtr, 1, T1)
+
+/* set both tweak words at once */
+#define	Skein_Set_T0_T1(ctxPtr, T0, T1)		\
+	do {					\
+		Skein_Set_T0(ctxPtr, (T0));	\
+		Skein_Set_T1(ctxPtr, (T1));	\
+		_NOTE(CONSTCOND)		\
+	} while (0)
+
+#define	Skein_Set_Type(ctxPtr, BLK_TYPE)	\
+	Skein_Set_T1(ctxPtr, SKEIN_T1_BLK_TYPE_##BLK_TYPE)
+
+/*
+ * set up for starting with a new type: h.T[0]=0; h.T[1] = NEW_TYPE; h.bCnt=0;
+ */
+#define	Skein_Start_New_Type(ctxPtr, BLK_TYPE)				\
+	do {								\
+		Skein_Set_T0_T1(ctxPtr, 0, SKEIN_T1_FLAG_FIRST |	\
+		    SKEIN_T1_BLK_TYPE_ ## BLK_TYPE);			\
+		(ctxPtr)->h.bCnt = 0;	\
+		_NOTE(CONSTCOND)					\
+	} while (0)
+
+#define	Skein_Clear_First_Flag(hdr)					\
+	do {								\
+		(hdr).T[1] &= ~SKEIN_T1_FLAG_FIRST;			\
+		_NOTE(CONSTCOND)					\
+	} while (0)
+#define	Skein_Set_Bit_Pad_Flag(hdr)					\
+	do {								\
+		(hdr).T[1] |=  SKEIN_T1_FLAG_BIT_PAD;			\
+		_NOTE(CONSTCOND)					\
+	} while (0)
+
+#define	Skein_Set_Tree_Level(hdr, height)				\
+	do {								\
+		(hdr).T[1] |= SKEIN_T1_TREE_LEVEL(height);		\
+		_NOTE(CONSTCOND)					\
+	} while (0)
+
+/*
+ * "Internal" Skein definitions for debugging and error checking
+ * Note: in Illumos we always disable debugging features.
+ */
+#define	Skein_Show_Block(bits, ctx, X, blkPtr, wPtr, ksEvenPtr, ksOddPtr)
+#define	Skein_Show_Round(bits, ctx, r, X)
+#define	Skein_Show_R_Ptr(bits, ctx, r, X_ptr)
+#define	Skein_Show_Final(bits, ctx, cnt, outPtr)
+#define	Skein_Show_Key(bits, ctx, key, keyBytes)
+
+/* run-time checks (e.g., bad params, uninitialized context)? */
+#ifndef	SKEIN_ERR_CHECK
+/* default: ignore all Asserts, for performance */
+#define	Skein_Assert(x, retCode)
+#define	Skein_assert(x)
+#elif	defined(SKEIN_ASSERT)
+#include <sys/debug.h>
+#define	Skein_Assert(x, retCode)	ASSERT(x)
+#define	Skein_assert(x)			ASSERT(x)
+#else
+#include <sys/debug.h>
+/*  caller error */
+#define	Skein_Assert(x, retCode)		\
+	do {					\
+		if (!(x))			\
+			return (retCode);	\
+		_NOTE(CONSTCOND)		\
+	} while (0)
+/* internal error */
+#define	Skein_assert(x)	ASSERT(x)
+#endif
+
+/*
+ * Skein block function constants (shared across Ref and Opt code)
+ */
+enum {
+	/* Skein_256 round rotation constants */
+	R_256_0_0 = 14, R_256_0_1 = 16,
+	R_256_1_0 = 52, R_256_1_1 = 57,
+	R_256_2_0 = 23, R_256_2_1 = 40,
+	R_256_3_0 = 5, R_256_3_1 = 37,
+	R_256_4_0 = 25, R_256_4_1 = 33,
+	R_256_5_0 = 46, R_256_5_1 = 12,
+	R_256_6_0 = 58, R_256_6_1 = 22,
+	R_256_7_0 = 32, R_256_7_1 = 32,
+
+	/* Skein_512 round rotation constants */
+	R_512_0_0 = 46, R_512_0_1 = 36, R_512_0_2 = 19, R_512_0_3 = 37,
+	R_512_1_0 = 33, R_512_1_1 = 27, R_512_1_2 = 14, R_512_1_3 = 42,
+	R_512_2_0 = 17, R_512_2_1 = 49, R_512_2_2 = 36, R_512_2_3 = 39,
+	R_512_3_0 = 44, R_512_3_1 = 9, R_512_3_2 = 54, R_512_3_3 = 56,
+	R_512_4_0 = 39, R_512_4_1 = 30, R_512_4_2 = 34, R_512_4_3 = 24,
+	R_512_5_0 = 13, R_512_5_1 = 50, R_512_5_2 = 10, R_512_5_3 = 17,
+	R_512_6_0 = 25, R_512_6_1 = 29, R_512_6_2 = 39, R_512_6_3 = 43,
+	R_512_7_0 = 8, R_512_7_1 = 35, R_512_7_2 = 56, R_512_7_3 = 22,
+
+	/* Skein1024 round rotation constants */
+	R1024_0_0 = 24, R1024_0_1 = 13, R1024_0_2 = 8, R1024_0_3 =
+	    47, R1024_0_4 = 8, R1024_0_5 = 17, R1024_0_6 = 22, R1024_0_7 = 37,
+	R1024_1_0 = 38, R1024_1_1 = 19, R1024_1_2 = 10, R1024_1_3 =
+	    55, R1024_1_4 = 49, R1024_1_5 = 18, R1024_1_6 = 23, R1024_1_7 = 52,
+	R1024_2_0 = 33, R1024_2_1 = 4, R1024_2_2 = 51, R1024_2_3 =
+	    13, R1024_2_4 = 34, R1024_2_5 = 41, R1024_2_6 = 59, R1024_2_7 = 17,
+	R1024_3_0 = 5, R1024_3_1 = 20, R1024_3_2 = 48, R1024_3_3 =
+	    41, R1024_3_4 = 47, R1024_3_5 = 28, R1024_3_6 = 16, R1024_3_7 = 25,
+	R1024_4_0 = 41, R1024_4_1 = 9, R1024_4_2 = 37, R1024_4_3 =
+	    31, R1024_4_4 = 12, R1024_4_5 = 47, R1024_4_6 = 44, R1024_4_7 = 30,
+	R1024_5_0 = 16, R1024_5_1 = 34, R1024_5_2 = 56, R1024_5_3 =
+	    51, R1024_5_4 = 4, R1024_5_5 = 53, R1024_5_6 = 42, R1024_5_7 = 41,
+	R1024_6_0 = 31, R1024_6_1 = 44, R1024_6_2 = 47, R1024_6_3 =
+	    46, R1024_6_4 = 19, R1024_6_5 = 42, R1024_6_6 = 44, R1024_6_7 = 25,
+	R1024_7_0 = 9, R1024_7_1 = 48, R1024_7_2 = 35, R1024_7_3 =
+	    52, R1024_7_4 = 23, R1024_7_5 = 31, R1024_7_6 = 37, R1024_7_7 = 20
+};
+
+/* number of rounds for the different block sizes */
+#define	SKEIN_256_ROUNDS_TOTAL	(72)
+#define	SKEIN_512_ROUNDS_TOTAL	(72)
+#define	SKEIN1024_ROUNDS_TOTAL	(80)
+
+
+extern const uint64_t SKEIN_256_IV_128[];
+extern const uint64_t SKEIN_256_IV_160[];
+extern const uint64_t SKEIN_256_IV_224[];
+extern const uint64_t SKEIN_256_IV_256[];
+extern const uint64_t SKEIN_512_IV_128[];
+extern const uint64_t SKEIN_512_IV_160[];
+extern const uint64_t SKEIN_512_IV_224[];
+extern const uint64_t SKEIN_512_IV_256[];
+extern const uint64_t SKEIN_512_IV_384[];
+extern const uint64_t SKEIN_512_IV_512[];
+extern const uint64_t SKEIN1024_IV_384[];
+extern const uint64_t SKEIN1024_IV_512[];
+extern const uint64_t SKEIN1024_IV_1024[];
+
+#endif	/* _SKEIN_IMPL_H_ */
diff --git a/module/icp/algs/skein/skein_iv.c b/module/icp/algs/skein/skein_iv.c
new file mode 100644
index 000000000..140d38f76
--- /dev/null
+++ b/module/icp/algs/skein/skein_iv.c
@@ -0,0 +1,185 @@
+/*
+ * Pre-computed Skein IVs
+ *
+ * NOTE: these values are not "magic" constants, but
+ * are generated using the Threefish block function.
+ * They are pre-computed here only for speed; i.e., to
+ * avoid the need for a Threefish call during Init().
+ *
+ * The IV for any fixed hash length may be pre-computed.
+ * Only the most common values are included here.
+ */
+/* Copyright 2013 Doug Whiting. This code is released to the public domain. */
+/*
+ * Illumos implementation note: these constants are for Skein v1.3 as per:
+ * http://www.skein-hash.info/sites/default/files/skein1.3.pdf
+ */
+
+#include <sys/skein.h>		/* get Skein macros and types */
+#include "skein_impl.h"		/* get internal definitions */
+
+#define	MK_64 SKEIN_MK_64
+
+/* blkSize =  256 bits. hashSize =  128 bits */
+const uint64_t SKEIN_256_IV_128[] = {
+	MK_64(0xE1111906, 0x964D7260),
+	MK_64(0x883DAAA7, 0x7C8D811C),
+	MK_64(0x10080DF4, 0x91960F7A),
+	MK_64(0xCCF7DDE5, 0xB45BC1C2)
+};
+
+/* blkSize =  256 bits. hashSize =  160 bits */
+const uint64_t SKEIN_256_IV_160[] = {
+	MK_64(0x14202314, 0x72825E98),
+	MK_64(0x2AC4E9A2, 0x5A77E590),
+	MK_64(0xD47A5856, 0x8838D63E),
+	MK_64(0x2DD2E496, 0x8586AB7D)
+};
+
+/* blkSize =  256 bits. hashSize =  224 bits */
+const uint64_t SKEIN_256_IV_224[] = {
+	MK_64(0xC6098A8C, 0x9AE5EA0B),
+	MK_64(0x876D5686, 0x08C5191C),
+	MK_64(0x99CB88D7, 0xD7F53884),
+	MK_64(0x384BDDB1, 0xAEDDB5DE)
+};
+
+/* blkSize =  256 bits. hashSize =  256 bits */
+const uint64_t SKEIN_256_IV_256[] = {
+	MK_64(0xFC9DA860, 0xD048B449),
+	MK_64(0x2FCA6647, 0x9FA7D833),
+	MK_64(0xB33BC389, 0x6656840F),
+	MK_64(0x6A54E920, 0xFDE8DA69)
+};
+
+/* blkSize =  512 bits. hashSize =  128 bits */
+const uint64_t SKEIN_512_IV_128[] = {
+	MK_64(0xA8BC7BF3, 0x6FBF9F52),
+	MK_64(0x1E9872CE, 0xBD1AF0AA),
+	MK_64(0x309B1790, 0xB32190D3),
+	MK_64(0xBCFBB854, 0x3F94805C),
+	MK_64(0x0DA61BCD, 0x6E31B11B),
+	MK_64(0x1A18EBEA, 0xD46A32E3),
+	MK_64(0xA2CC5B18, 0xCE84AA82),
+	MK_64(0x6982AB28, 0x9D46982D)
+};
+
+/* blkSize =  512 bits. hashSize =  160 bits */
+const uint64_t SKEIN_512_IV_160[] = {
+	MK_64(0x28B81A2A, 0xE013BD91),
+	MK_64(0xC2F11668, 0xB5BDF78F),
+	MK_64(0x1760D8F3, 0xF6A56F12),
+	MK_64(0x4FB74758, 0x8239904F),
+	MK_64(0x21EDE07F, 0x7EAF5056),
+	MK_64(0xD908922E, 0x63ED70B8),
+	MK_64(0xB8EC76FF, 0xECCB52FA),
+	MK_64(0x01A47BB8, 0xA3F27A6E)
+};
+
+/* blkSize =  512 bits. hashSize =  224 bits */
+const uint64_t SKEIN_512_IV_224[] = {
+	MK_64(0xCCD06162, 0x48677224),
+	MK_64(0xCBA65CF3, 0xA92339EF),
+	MK_64(0x8CCD69D6, 0x52FF4B64),
+	MK_64(0x398AED7B, 0x3AB890B4),
+	MK_64(0x0F59D1B1, 0x457D2BD0),
+	MK_64(0x6776FE65, 0x75D4EB3D),
+	MK_64(0x99FBC70E, 0x997413E9),
+	MK_64(0x9E2CFCCF, 0xE1C41EF7)
+};
+
+/* blkSize =  512 bits. hashSize =  256 bits */
+const uint64_t SKEIN_512_IV_256[] = {
+	MK_64(0xCCD044A1, 0x2FDB3E13),
+	MK_64(0xE8359030, 0x1A79A9EB),
+	MK_64(0x55AEA061, 0x4F816E6F),
+	MK_64(0x2A2767A4, 0xAE9B94DB),
+	MK_64(0xEC06025E, 0x74DD7683),
+	MK_64(0xE7A436CD, 0xC4746251),
+	MK_64(0xC36FBAF9, 0x393AD185),
+	MK_64(0x3EEDBA18, 0x33EDFC13)
+};
+
+/* blkSize =  512 bits. hashSize =  384 bits */
+const uint64_t SKEIN_512_IV_384[] = {
+	MK_64(0xA3F6C6BF, 0x3A75EF5F),
+	MK_64(0xB0FEF9CC, 0xFD84FAA4),
+	MK_64(0x9D77DD66, 0x3D770CFE),
+	MK_64(0xD798CBF3, 0xB468FDDA),
+	MK_64(0x1BC4A666, 0x8A0E4465),
+	MK_64(0x7ED7D434, 0xE5807407),
+	MK_64(0x548FC1AC, 0xD4EC44D6),
+	MK_64(0x266E1754, 0x6AA18FF8)
+};
+
+/* blkSize =  512 bits. hashSize =  512 bits */
+const uint64_t SKEIN_512_IV_512[] = {
+	MK_64(0x4903ADFF, 0x749C51CE),
+	MK_64(0x0D95DE39, 0x9746DF03),
+	MK_64(0x8FD19341, 0x27C79BCE),
+	MK_64(0x9A255629, 0xFF352CB1),
+	MK_64(0x5DB62599, 0xDF6CA7B0),
+	MK_64(0xEABE394C, 0xA9D5C3F4),
+	MK_64(0x991112C7, 0x1A75B523),
+	MK_64(0xAE18A40B, 0x660FCC33)
+};
+
+/* blkSize = 1024 bits. hashSize =  384 bits */
+const uint64_t SKEIN1024_IV_384[] = {
+	MK_64(0x5102B6B8, 0xC1894A35),
+	MK_64(0xFEEBC9E3, 0xFE8AF11A),
+	MK_64(0x0C807F06, 0xE32BED71),
+	MK_64(0x60C13A52, 0xB41A91F6),
+	MK_64(0x9716D35D, 0xD4917C38),
+	MK_64(0xE780DF12, 0x6FD31D3A),
+	MK_64(0x797846B6, 0xC898303A),
+	MK_64(0xB172C2A8, 0xB3572A3B),
+	MK_64(0xC9BC8203, 0xA6104A6C),
+	MK_64(0x65909338, 0xD75624F4),
+	MK_64(0x94BCC568, 0x4B3F81A0),
+	MK_64(0x3EBBF51E, 0x10ECFD46),
+	MK_64(0x2DF50F0B, 0xEEB08542),
+	MK_64(0x3B5A6530, 0x0DBC6516),
+	MK_64(0x484B9CD2, 0x167BBCE1),
+	MK_64(0x2D136947, 0xD4CBAFEA)
+};
+
+/* blkSize = 1024 bits. hashSize =  512 bits */
+const uint64_t SKEIN1024_IV_512[] = {
+	MK_64(0xCAEC0E5D, 0x7C1B1B18),
+	MK_64(0xA01B0E04, 0x5F03E802),
+	MK_64(0x33840451, 0xED912885),
+	MK_64(0x374AFB04, 0xEAEC2E1C),
+	MK_64(0xDF25A0E2, 0x813581F7),
+	MK_64(0xE4004093, 0x8B12F9D2),
+	MK_64(0xA662D539, 0xC2ED39B6),
+	MK_64(0xFA8B85CF, 0x45D8C75A),
+	MK_64(0x8316ED8E, 0x29EDE796),
+	MK_64(0x053289C0, 0x2E9F91B8),
+	MK_64(0xC3F8EF1D, 0x6D518B73),
+	MK_64(0xBDCEC3C4, 0xD5EF332E),
+	MK_64(0x549A7E52, 0x22974487),
+	MK_64(0x67070872, 0x5B749816),
+	MK_64(0xB9CD28FB, 0xF0581BD1),
+	MK_64(0x0E2940B8, 0x15804974)
+};
+
+/* blkSize = 1024 bits. hashSize = 1024 bits */
+const uint64_t SKEIN1024_IV_1024[] = {
+	MK_64(0xD593DA07, 0x41E72355),
+	MK_64(0x15B5E511, 0xAC73E00C),
+	MK_64(0x5180E5AE, 0xBAF2C4F0),
+	MK_64(0x03BD41D3, 0xFCBCAFAF),
+	MK_64(0x1CAEC6FD, 0x1983A898),
+	MK_64(0x6E510B8B, 0xCDD0589F),
+	MK_64(0x77E2BDFD, 0xC6394ADA),
+	MK_64(0xC11E1DB5, 0x24DCB0A3),
+	MK_64(0xD6D14AF9, 0xC6329AB5),
+	MK_64(0x6A9B0BFC, 0x6EB67E0D),
+	MK_64(0x9243C60D, 0xCCFF1332),
+	MK_64(0x1A1F1DDE, 0x743F02D4),
+	MK_64(0x0996753C, 0x10ED0BB8),
+	MK_64(0x6572DD22, 0xF2B4969A),
+	MK_64(0x61FD3062, 0xD00A579A),
+	MK_64(0x1DE0536E, 0x8682E539)
+};
diff --git a/module/icp/algs/skein/skein_port.h b/module/icp/algs/skein/skein_port.h
new file mode 100644
index 000000000..1b0225236
--- /dev/null
+++ b/module/icp/algs/skein/skein_port.h
@@ -0,0 +1,128 @@
+/*
+ * Platform-specific definitions for Skein hash function.
+ *
+ * Source code author: Doug Whiting, 2008.
+ *
+ * This algorithm and source code is released to the public domain.
+ *
+ * Many thanks to Brian Gladman for his portable header files.
+ *
+ * To port Skein to an "unsupported" platform, change the definitions
+ * in this file appropriately.
+ */
+/* Copyright 2013 Doug Whiting. This code is released to the public domain. */
+
+#ifndef	_SKEIN_PORT_H_
+#define	_SKEIN_PORT_H_
+
+#include <sys/types.h>	/* get integer type definitions */
+#include <sys/systm.h>	/* for bcopy() */
+
+#ifndef	RotL_64
+#define	RotL_64(x, N)	(((x) << (N)) | ((x) >> (64 - (N))))
+#endif
+
+/*
+ * Skein is "natively" little-endian (unlike SHA-xxx), for optimal
+ * performance on x86 CPUs. The Skein code requires the following
+ * definitions for dealing with endianness:
+ *
+ *    SKEIN_NEED_SWAP:  0 for little-endian, 1 for big-endian
+ *    Skein_Put64_LSB_First
+ *    Skein_Get64_LSB_First
+ *    Skein_Swap64
+ *
+ * If SKEIN_NEED_SWAP is defined at compile time, it is used here
+ * along with the portable versions of Put64/Get64/Swap64, which
+ * are slow in general.
+ *
+ * Otherwise, an "auto-detect" of endianness is attempted below.
+ * If the default handling doesn't work well, the user may insert
+ * platform-specific code instead (e.g., for big-endian CPUs).
+ *
+ */
+#ifndef	SKEIN_NEED_SWAP		/* compile-time "override" for endianness? */
+
+#include <sys/isa_defs.h>	/* get endianness selection */
+
+#define	PLATFORM_MUST_ALIGN	_ALIGNMENT_REQUIRED
+#if	defined(_BIG_ENDIAN)
+/* here for big-endian CPUs */
+#define	SKEIN_NEED_SWAP   (1)
+#else
+/* here for x86 and x86-64 CPUs (and other detected little-endian CPUs) */
+#define	SKEIN_NEED_SWAP   (0)
+#if	PLATFORM_MUST_ALIGN == 0	/* ok to use "fast" versions? */
+#define	Skein_Put64_LSB_First(dst08, src64, bCnt) bcopy(src64, dst08, bCnt)
+#define	Skein_Get64_LSB_First(dst64, src08, wCnt) \
+	bcopy(src08, dst64, 8 * (wCnt))
+#endif
+#endif
+
+#endif				/* ifndef SKEIN_NEED_SWAP */
+
+/*
+ * Provide any definitions still needed.
+ */
+#ifndef	Skein_Swap64	/* swap for big-endian, nop for little-endian */
+#if	SKEIN_NEED_SWAP
+#define	Skein_Swap64(w64)				\
+	(((((uint64_t)(w64)) & 0xFF) << 56) |		\
+	(((((uint64_t)(w64)) >> 8) & 0xFF) << 48) |	\
+	(((((uint64_t)(w64)) >> 16) & 0xFF) << 40) |	\
+	(((((uint64_t)(w64)) >> 24) & 0xFF) << 32) |	\
+	(((((uint64_t)(w64)) >> 32) & 0xFF) << 24) |	\
+	(((((uint64_t)(w64)) >> 40) & 0xFF) << 16) |	\
+	(((((uint64_t)(w64)) >> 48) & 0xFF) << 8) |	\
+	(((((uint64_t)(w64)) >> 56) & 0xFF)))
+#else
+#define	Skein_Swap64(w64)  (w64)
+#endif
+#endif				/* ifndef Skein_Swap64 */
+
+#ifndef	Skein_Put64_LSB_First
+void
+Skein_Put64_LSB_First(uint8_t *dst, const uint64_t *src, size_t bCnt)
+#ifdef	SKEIN_PORT_CODE		/* instantiate the function code here? */
+{
+	/*
+	 * this version is fully portable (big-endian or little-endian),
+	 * but slow
+	 */
+	size_t n;
+
+	for (n = 0; n < bCnt; n++)
+		dst[n] = (uint8_t)(src[n >> 3] >> (8 * (n & 7)));
+}
+#else
+;				/* output only the function prototype */
+#endif
+#endif				/* ifndef Skein_Put64_LSB_First */
+
+#ifndef	Skein_Get64_LSB_First
+void
+Skein_Get64_LSB_First(uint64_t *dst, const uint8_t *src, size_t wCnt)
+#ifdef	SKEIN_PORT_CODE		/* instantiate the function code here? */
+{
+	/*
+	 * this version is fully portable (big-endian or little-endian),
+	 * but slow
+	 */
+	size_t n;
+
+	for (n = 0; n < 8 * wCnt; n += 8)
+		dst[n / 8] = (((uint64_t)src[n])) +
+		    (((uint64_t)src[n + 1]) << 8) +
+		    (((uint64_t)src[n + 2]) << 16) +
+		    (((uint64_t)src[n + 3]) << 24) +
+		    (((uint64_t)src[n + 4]) << 32) +
+		    (((uint64_t)src[n + 5]) << 40) +
+		    (((uint64_t)src[n + 6]) << 48) +
+		    (((uint64_t)src[n + 7]) << 56);
+}
+#else
+;				/* output only the function prototype */
+#endif
+#endif				/* ifndef Skein_Get64_LSB_First */
+
+#endif	/* _SKEIN_PORT_H_ */