Page MenuHomeFreeBSD
Files F174551186

D6166.id15779.diff
No OneTemporary
Actions

D6166.id15779.diff
View Options

This file is larger than 256 KB, so syntax highlighting was skipped.
Index: lib/libcrypt/Makefile
===================================================================
--- lib/libcrypt/Makefile
+++ lib/libcrypt/Makefile
@@ -10,16 +10,25 @@
SHLIB_MAJOR= 5
LIB= crypt
-.PATH: ${.CURDIR}/../libmd ${.CURDIR}/../../sys/crypto/sha2
+.PATH: ${.CURDIR}/../libmd ${.CURDIR}/../../sys/crypto/sha2 ${.CURDIR}/../../sys/crypto/skein
SRCS= crypt.c misc.c \
crypt-md5.c md5c.c \
crypt-nthash.c md4c.c \
crypt-sha256.c sha256c.c \
- crypt-sha512.c sha512c.c
+ crypt-sha512.c sha512c.c \
+ skein.c skein_block.c
MAN= crypt.3
MLINKS= crypt.3 crypt_get_format.3 crypt.3 crypt_set_format.3
CFLAGS+= -I${.CURDIR}/../libmd -I${.CURDIR}/../libutil \
- -I${.CURDIR}/../../sys/crypto/sha2
+ -I${.CURDIR}/../../sys/crypto/sha2 -I${.CURDIR}/../../sys/crypto/skein
+
+# Use assembly optimized skein if available
+.if exists(${MACHINE_ARCH}/skein_block_asm.s)
+.PATH: ${.CURDIR}/../../sys/crypto/skein/${MACHINE_ARCH}
+SRCS += skein_block_asm.s
+CFLAGS += -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792
+ACFLAGS += -DELF -Wa,--noexecstack
+.endif
# Pull in the strong crypto, if it is present.
.if exists(${.CURDIR}/../../secure/lib/libcrypt) && ${MK_CRYPT} != "no"
@@ -32,7 +41,10 @@
MD5Init MD5Final MD5Update MD5Pad \
SHA256_Init SHA256_Final SHA256_Update \
SHA384_Init SHA384_Final SHA384_Update \
- SHA512_Init SHA512_Final SHA512_Update
+ SHA512_Init SHA512_Final SHA512_Update \
+ SKEIN256_Init SKEIN256_Final SKEIN256_Update \
+ SKEIN512_Init SKEIN512_Final SKEIN512_Update \
+ SKEIN1024_Init SKEIN1024_Final SKEIN1024_Update
CFLAGS+= -D${sym}=__${sym}
.endfor
Index: lib/libmd/Makefile
===================================================================
--- lib/libmd/Makefile
+++ lib/libmd/Makefile
@@ -9,12 +9,16 @@
sha0c.c sha0hl.c sha1c.c sha1hl.c \
sha256c.c sha256hl.c \
sha384hl.c \
- sha512c.c sha512hl.c
-INCS= md4.h md5.h ripemd.h sha.h sha256.h sha384.h sha512.h
+ sha512c.c sha512hl.c \
+ skein.c skein_block.c \
+ skein256hl.c skein512hl.c skein1024hl.c
+INCS= md4.h md5.h ripemd.h sha.h sha256.h sha384.h sha512.h \
+ skein.h skein_port.h skein_freebsd.h skein_iv.h \
+ brg_types.h brg_endian.h
WARNS?= 0
-MAN+= md4.3 md5.3 ripemd.3 sha.3 sha256.3 sha512.3
+MAN+= md4.3 md5.3 ripemd.3 sha.3 sha256.3 sha512.3 skein.3
MLINKS+=md4.3 MD4Init.3 md4.3 MD4Update.3 md4.3 MD4Final.3
MLINKS+=md4.3 MD4End.3 md4.3 MD4File.3 md4.3 MD4FileChunk.3
MLINKS+=md4.3 MD4Data.3
@@ -43,11 +47,27 @@
MLINKS+=sha512.3 SHA512_Final.3 sha512.3 SHA512_End.3
MLINKS+=sha512.3 SHA512_File.3 sha512.3 SHA512_FileChunk.3
MLINKS+=sha512.3 SHA512_Data.3
+MLINKS+=skein.3 SKEIN256_Init.3 skein.3 SKEIN256_Update.3
+MLINKS+=skein.3 SKEIN256_Final.3 skein.3 SKEIN256_End.3
+MLINKS+=skein.3 SKEIN256_File.3 skein.3 SKEIN256_FileChunk.3
+MLINKS+=skein.3 SKEIN256_Data.3 skein.3 skein256.3
+MLINKS+=skein.3 SKEIN512_Init.3 skein.3 SKEIN512_Update.3
+MLINKS+=skein.3 SKEIN512_Final.3 skein.3 SKEIN512_End.3
+MLINKS+=skein.3 SKEIN512_File.3 skein.3 SKEIN512_FileChunk.3
+MLINKS+=skein.3 SKEIN512_Data.3 skein.3 skein512.3
+MLINKS+=skein.3 SKEIN1024_Init.3 skein.3 SKEIN1024_Update.3
+MLINKS+=skein.3 SKEIN1024_Final.3 skein.3 SKEIN1024_End.3
+MLINKS+=skein.3 SKEIN1024_File.3 skein.3 SKEIN1024_FileChunk.3
+MLINKS+=skein.3 SKEIN1024_Data.3 skein.3 skein1024.3
+
CLEANFILES+= md[245]hl.c md[245].ref md[245].3 mddriver \
rmd160.ref rmd160hl.c rmddriver \
sha0.ref sha0hl.c sha1.ref sha1hl.c shadriver \
sha256.ref sha256hl.c sha384hl.c sha384.ref \
- sha512.ref sha512hl.c
+ sha512.ref sha512hl.c \
+ skein256hl.c skein512hl.c skein1024hl.c \
+ skein256.ref skein512.ref skein1024.ref \
+ skeindriver
# Define WEAK_REFS to provide weak aliases for libmd symbols
#
@@ -56,8 +76,10 @@
# * macros are used to rename symbols to libcrypt internal names
# * no weak aliases are generated
CFLAGS+= -I${.CURDIR} -I${.CURDIR}/../../sys/crypto/sha2
+CFLAGS+= -I${.CURDIR}/../../sys/crypto/skein
CFLAGS+= -DWEAK_REFS
.PATH: ${.CURDIR}/${MACHINE_ARCH} ${.CURDIR}/../../sys/crypto/sha2
+.PATH: ${.CURDIR}/../../sys/crypto/skein ${.CURDIR}/../../sys/crypto/skein/${MACHINE_ARCH}
.if exists(${MACHINE_ARCH}/sha.S)
SRCS+= sha.S
@@ -67,7 +89,11 @@
SRCS+= rmd160.S
CFLAGS+= -DRMD160_ASM
.endif
-.if exists(${MACHINE_ARCH}/sha.S) || exists(${MACHINE_ARCH}/rmd160.S)
+.if exists(${MACHINE_ARCH}/skein_block_asm.s)
+SRCS+= skein_block_asm.s
+CFLAGS+= -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792
+.endif
+.if exists(${MACHINE_ARCH}/sha.S) || exists(${MACHINE_ARCH}/rmd160.S) || exists(${MACHINE_ARCH}/skein_block_asm.s)
ACFLAGS+= -DELF -Wa,--noexecstack
.endif
@@ -113,6 +139,25 @@
-e 's/RIPEMD160__/RIPEMD160_/g' \
${.ALLSRC}) > ${.TARGET}
+skein256hl.c: mdXhl.c
+ (echo '#define LENGTH 32'; \
+ sed -e 's/mdX/skein/g' -e 's/MDX/SKEIN256_/g' \
+ -e 's/SKEIN256__/SKEIN256_/g' \
+ ${.ALLSRC}) > ${.TARGET}
+
+skein512hl.c: mdXhl.c
+ (echo '#define LENGTH 64'; \
+ sed -e 's/mdX/skein/g' -e 's/MDX/SKEIN512_/g' \
+ -e 's/SKEIN512__/SKEIN512_/g' \
+ ${.ALLSRC}) > ${.TARGET}
+
+skein1024hl.c: mdXhl.c
+ (echo '#define LENGTH 128'; \
+ sed -e 's/mdX/skein/g' -e 's/MDX/SKEIN1024_/g' \
+ -e 's/SKEIN1024__/SKEIN1024_/g' \
+ ${.ALLSRC}) > ${.TARGET}
+
+
.for i in 2 4 5
md${i}.3: ${.CURDIR}/mdX.3
sed -e "s/mdX/md${i}/g" -e "s/MDX/MD${i}/g" ${.ALLSRC} > ${.TARGET}
@@ -224,8 +269,51 @@
@echo 'RIPEMD160 ("12345678901234567890123456789012345678901234567890123456789012345678901234567890") =' \
'9b752e45573d4b39f4dbd3323cab82bf63326bfb' >> ${.TARGET}
+skein256.ref:
+ echo 'SKEIN256 test suite:' > ${.TARGET}
+ @echo 'SKEIN256 ("") = c8877087da56e072870daa843f176e9453115929094c3a40c463a196c29bf7ba' >> ${.TARGET}
+ @echo 'SKEIN256 ("abc") = 258bdec343b9fde1639221a5ae0144a96e552e5288753c5fec76c05fc2fc1870' >> ${.TARGET}
+ @echo 'SKEIN256 ("message digest") =' \
+ '4d2ce0062b5eb3a4db95bc1117dd8aa014f6cd50fdc8e64f31f7d41f9231e488' >> ${.TARGET}
+ @echo 'SKEIN256 ("abcdefghijklmnopqrstuvwxyz") =' \
+ '46d8440685461b00e3ddb891b2ecc6855287d2bd8834a95fb1c1708b00ea5e82' >> ${.TARGET}
+ @echo 'SKEIN256 ("ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789") =' \
+ '7c5eb606389556b33d34eb2536459528dc0af97adbcd0ce273aeb650f598d4b2' >> ${.TARGET}
+ @echo 'SKEIN256 ("12345678901234567890123456789012345678901234567890123456789012345678901234567890") =' \
+ '4def7a7e5464a140ae9c3a80279fbebce4bd00f9faad819ab7e001512f67a10d' >> ${.TARGET}
+
+skein512.ref:
+ echo 'SKEIN512 test suite:' > ${.TARGET}
+ @echo 'SKEIN512 ("") =' \
+ 'bc5b4c50925519c290cc634277ae3d6257212395cba733bbad37a4af0fa06af41fca7903d06564fea7a2d3730dbdb80c1f85562dfcc070334ea4d1d9e72cba7a' >> ${.TARGET}
+ @echo 'SKEIN512 ("abc") =' \
+ '8f5dd9ec798152668e35129496b029a960c9a9b88662f7f9482f110b31f9f93893ecfb25c009baad9e46737197d5630379816a886aa05526d3a70df272d96e75' >> ${.TARGET}
+ @echo 'SKEIN512 ("message digest") =' \
+ '15b73c158ffb875fed4d72801ded0794c720b121c0c78edf45f900937e6933d9e21a3a984206933d504b5dbb2368000411477ee1b204c986068df77886542fcc' >> ${.TARGET}
+ @echo 'SKEIN512 ("abcdefghijklmnopqrstuvwxyz") =' \
+ '23793ad900ef12f9165c8080da6fdfd2c8354a2929b8aadf83aa82a3c6470342f57cf8c035ec0d97429b626c4d94f28632c8f5134fd367dca5cf293d2ec13f8c' >> ${.TARGET}
+ @echo 'SKEIN512 ("ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789") =' \
+ '0c6bed927e022f5ddcf81877d42e5f75798a9f8fd3ede3d83baac0a2f364b082e036c11af35fe478745459dd8f5c0b73efe3c56ba5bb2009208d5a29cc6e469c' >> ${.TARGET}
+ @echo 'SKEIN512 ("12345678901234567890123456789012345678901234567890123456789012345678901234567890") =' \
+ '2ca9fcffb3456f297d1b5f407014ecb856f0baac8eb540f534b1f187196f21e88f31103128c2f03fcc9857d7a58eb66f9525e2302d88833ee069295537a434ce' >> ${.TARGET}
+
+skein1024.ref:
+ echo 'SKEIN1024 test suite:' > ${.TARGET}
+ @echo 'SKEIN1024 ("") =' \
+ '0fff9563bb3279289227ac77d319b6fff8d7e9f09da1247b72a0a265cd6d2a62645ad547ed8193db48cff847c06494a03f55666d3b47eb4c20456c9373c86297d630d5578ebd34cb40991578f9f52b18003efa35d3da6553ff35db91b81ab890bec1b189b7f52cb2a783ebb7d823d725b0b4a71f6824e88f68f982eefc6d19c6' >> ${.TARGET}
+ @echo 'SKEIN1024 ("abc") =' \
+ '35a599a0f91abcdb4cb73c19b8cb8d947742d82c309137a7caed29e8e0a2ca7a9ff9a90c34c1908cc7e7fd99bb15032fb86e76df21b72628399b5f7c3cc209d7bb31c99cd4e19465622a049afbb87c03b5ce3888d17e6e667279ec0aa9b3e2712624c01b5f5bbe1a564220bdcf6990af0c2539019f313fdd7406cca3892a1f1f' >> ${.TARGET}
+ @echo 'SKEIN1024 ("message digest") =' \
+ 'ea891f5268acd0fac97467fc1aa89d1ce8681a9992a42540e53babee861483110c2d16f49e73bac27653ff173003e40cfb08516cd34262e6af95a5d8645c9c1abb3e813604d508b8511b30f9a5c1b352aa0791c7d2f27b2706dccea54bc7de6555b5202351751c3299f97c09cf89c40f67187e2521c0fad82b30edbb224f0458' >> ${.TARGET}
+ @echo 'SKEIN1024 ("abcdefghijklmnopqrstuvwxyz") =' \
+ 'f23d95c2a25fbcd0e797cd058fec39d3c52d2b5afd7a9af1df934e63257d1d3dcf3246e7329c0f1104c1e51e3d22e300507b0c3b9f985bb1f645ef49835080536becf83788e17fed09c9982ba65c3cb7ffe6a5f745b911c506962adf226e435c42f6f6bc08d288f9c810e807e3216ef444f3db22744441deefa4900982a1371f' >> ${.TARGET}
+ @echo 'SKEIN1024 ("ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789") =' \
+ 'cf3889e8a8d11bfd3938055d7d061437962bc5eac8ae83b1b71c94be201b8cf657fdbfc38674997a008c0c903f56a23feb3ae30e012377f1cfa080a9ca7fe8b96138662653fb3335c7d06595bf8baf65e215307532094cfdfa056bd8052ab792a3944a2adaa47b30335b8badb8fe9eb94fe329cdca04e58bbc530f0af709f469' >> ${.TARGET}
+ @echo 'SKEIN1024 ("12345678901234567890123456789012345678901234567890123456789012345678901234567890") =' \
+ 'cf21a613620e6c119eca31fdfaad449a8e02f95ca256c21d2a105f8e4157048f9fe1e897893ea18b64e0e37cb07d5ac947f27ba544caf7cbc1ad094e675aed77a366270f7eb7f46543bccfa61c526fd628408058ed00ed566ac35a9761d002e629c4fb0d430b2f4ad016fcc49c44d2981c4002da0eecc42144160e2eaea4855a' >> ${.TARGET}
+
test: md4.ref md5.ref sha0.ref rmd160.ref sha1.ref sha256.ref sha384.ref \
- sha512.ref
+ sha512.ref skein256.ref skein512.ref skein1024.ref
@${ECHO} if any of these test fail, the code produces wrong results
@${ECHO} and should NOT be used.
${CC} ${CFLAGS} ${LDFLAGS} -DMD=4 -o mddriver ${.CURDIR}/mddriver.c libmd.a
@@ -255,5 +343,15 @@
./shadriver | cmp sha512.ref -
@${ECHO} SHA-512 passed test
-rm -f shadriver
+ ${CC} ${CFLAGS} ${LDFLAGS} -DSKEIN=256 -o skeindriver ${.CURDIR}/skeindriver.c libmd.a
+ ./skeindriver | cmp skein256.ref -
+ @${ECHO} SKEIN256 passed test
+ ${CC} ${CFLAGS} ${LDFLAGS} -DSKEIN=512 -o skeindriver ${.CURDIR}/skeindriver.c libmd.a
+ ./skeindriver | cmp skein512.ref -
+ @${ECHO} SKEIN512 passed test
+ ${CC} ${CFLAGS} ${LDFLAGS} -DSKEIN=1024 -o skeindriver ${.CURDIR}/skeindriver.c libmd.a
+ ./skeindriver | cmp skein1024.ref -
+ @${ECHO} SKEIN1024 passed test
+ -rm -f skeindriver
.include <bsd.lib.mk>
Index: lib/libmd/mdX.3
===================================================================
--- lib/libmd/mdX.3
+++ lib/libmd/mdX.3
@@ -8,7 +8,7 @@
.\"
.\" $FreeBSD$
.\"
-.Dd February 11, 1999
+.Dd April 26, 2016
.Dt MDX 3
.Os
.Sh NAME
@@ -145,7 +145,11 @@
.Sh SEE ALSO
.Xr md4 3 ,
.Xr md5 3 ,
-.Xr sha 3
+.Xr ripemd 3 ,
+.Xr sha 3 ,
+.Xr sha256 3 ,
+.Xr sha512 3 ,
+.Xr skein 3
.Rs
.%A R. Rivest
.%T The MD4 Message-Digest Algorithm
Index: lib/libmd/ripemd.3
===================================================================
--- lib/libmd/ripemd.3
+++ lib/libmd/ripemd.3
@@ -9,7 +9,7 @@
.\" From: Id: mdX.3,v 1.14 1999/02/11 20:31:49 wollman Exp
.\" $FreeBSD$
.\"
-.Dd March 28, 2014
+.Dd April 26, 2016
.Dt RIPEMD 3
.Os
.Sh NAME
@@ -125,7 +125,10 @@
.Sh SEE ALSO
.Xr md4 3 ,
.Xr md5 3 ,
-.Xr sha 3
+.Xr sha 3 ,
+.Xr sha256 3 ,
+.Xr sha512 3 ,
+.Xr skein 3
.Sh HISTORY
These functions appeared in
.Fx 4.0 .
Index: lib/libmd/sha.3
===================================================================
--- lib/libmd/sha.3
+++ lib/libmd/sha.3
@@ -9,7 +9,7 @@
.\" From: Id: mdX.3,v 1.14 1999/02/11 20:31:49 wollman Exp
.\" $FreeBSD$
.\"
-.Dd March 28, 2014
+.Dd April 26, 2016
.Dt SHA 3
.Os
.Sh NAME
@@ -157,7 +157,9 @@
.Xr md4 3 ,
.Xr md5 3 ,
.Xr ripemd 3 ,
-.Xr sha256 3
+.Xr sha256 3 ,
+.Xr sha512 3 ,
+.Xr skein 3
.Sh HISTORY
These functions appeared in
.Fx 4.0 .
Index: lib/libmd/sha256.3
===================================================================
--- lib/libmd/sha256.3
+++ lib/libmd/sha256.3
@@ -9,7 +9,7 @@
.\" From: Id: mdX.3,v 1.14 1999/02/11 20:31:49 wollman Exp
.\" $FreeBSD$
.\"
-.Dd March 28, 2014
+.Dd April 26, 2016
.Dt SHA256 3
.Os
.Sh NAME
@@ -123,7 +123,10 @@
.Xr md4 3 ,
.Xr md5 3 ,
.Xr ripemd 3 ,
-.Xr sha 3
+.Xr sha 3 ,
+.Xr sha256 3 ,
+.Xr sha512 3 ,
+.Xr skein 3
.Sh HISTORY
These functions appeared in
.Fx 6.0 .
Index: lib/libmd/sha512.3
===================================================================
--- lib/libmd/sha512.3
+++ lib/libmd/sha512.3
@@ -159,7 +159,10 @@
.Xr md4 3 ,
.Xr md5 3 ,
.Xr ripemd 3 ,
-.Xr sha 3
+.Xr sha 3 ,
+.Xr sha256 3 ,
+.Xr sha512 3 ,
+.Xr skein 3
.Sh HISTORY
These functions appeared in
.Fx 9.0 .
Index: lib/libmd/skein.3
===================================================================
--- /dev/null
+++ lib/libmd/skein.3
@@ -0,0 +1,202 @@
+.\"
+.\" ----------------------------------------------------------------------------
+.\" "THE BEER-WARE LICENSE" (Revision 42):
+.\" <phk@FreeBSD.org> wrote this file. As long as you retain this notice you
+.\" can do whatever you want with this stuff. If we meet some day, and you think
+.\" this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
+.\" ----------------------------------------------------------------------------
+.\"
+.\" From: Id: sha512.3 292782 2015-12-27 17:33:59Z allanjude
+.\" $FreeBSD$
+.\"
+.Dd April 26, 2016
+.Dt SKEIN 3
+.Os
+.Sh NAME
+.Nm SKEIN256_Init ,
+.Nm SKEIN256_Update ,
+.Nm SKEIN256_Final ,
+.Nm SKEIN256_End ,
+.Nm SKEIN256_File ,
+.Nm SKEIN256_FileChunk ,
+.Nm SKEIN256_Data ,
+.Nm SKEIN512_Init ,
+.Nm SKEIN512_Update ,
+.Nm SKEIN512_Final ,
+.Nm SKEIN512_End ,
+.Nm SKEIN512_File ,
+.Nm SKEIN512_FileChunk ,
+.Nm SKEIN512_Data ,
+.Nm SKEIN1024_Init ,
+.Nm SKEIN1024_Update ,
+.Nm SKEIN1024_Final ,
+.Nm SKEIN1024_End ,
+.Nm SKEIN1024_File ,
+.Nm SKEIN1024_FileChunk ,
+.Nm SKEIN1024_Data
+.Nd calculate the ``SKEIN'' family of message digests
+.Sh LIBRARY
+.Lb libmd
+.Sh SYNOPSIS
+.In sys/types.h
+.In skein.h
+.Ft void
+.Fn SKEIN256_Init "SKEIN256_CTX *context"
+.Ft void
+.Fn SKEIN256_Update "SKEIN256_CTX *context" "const unsigned char *data" "size_t len"
+.Ft void
+.Fn SKEIN256_Final "unsigned char digest[32]" "SKEIN256_CTX *context"
+.Ft "char *"
+.Fn SKEIN256_End "SKEIN256_CTX *context" "char *buf"
+.Ft "char *"
+.Fn SKEIN256_File "const char *filename" "char *buf"
+.Ft "char *"
+.Fn SKEIN256_FileChunk "const char *filename" "char *buf" "off_t offset" "off_t length"
+.Ft "char *"
+.Fn SKEIN256_Data "const unsigned char *data" "unsigned int len" "char *buf"
+.Ft void
+.Fn SKEIN512_Init "SKEIN512_CTX *context"
+.Ft void
+.Fn SKEIN512_Update "SKEIN512_CTX *context" "const unsigned char *data" "size_t len"
+.Ft void
+.Fn SKEIN512_Final "unsigned char digest[64]" "SKEIN512_CTX *context"
+.Ft "char *"
+.Fn SKEIN512_End "SKEIN512_CTX *context" "char *buf"
+.Ft "char *"
+.Fn SKEIN512_File "const char *filename" "char *buf"
+.Ft "char *"
+.Fn SKEIN512_FileChunk "const char *filename" "char *buf" "off_t offset" "off_t length"
+.Ft "char *"
+.Fn SKEIN512_Data "const unsigned char *data" "unsigned int len" "char *buf"
+.Ft void
+.Fn SKEIN1024_Init "SKEIN1024_CTX *context"
+.Ft void
+.Fn SKEIN1024_Update "SKEIN1024_CTX *context" "const unsigned char *data" "size_t len"
+.Ft void
+.Fn SKEIN1024_Final "unsigned char digest[128]" "SKEIN1024_CTX *context"
+.Ft "char *"
+.Fn SKEIN1024_End "SKEIN1024_CTX *context" "char *buf"
+.Ft "char *"
+.Fn SKEIN1024_File "const char *filename" "char *buf"
+.Ft "char *"
+.Fn SKEIN1024_FileChunk "const char *filename" "char *buf" "off_t offset" "off_t length"
+.Ft "char *"
+.Fn SKEIN1024_Data "const unsigned char *data" "unsigned int len" "char *buf"
+.Sh DESCRIPTION
+The
+.Li SKEIN
+functions calculate a 256, 512, or 1024-bit cryptographic checksum (digest)
+for any number of input bytes.
+A cryptographic checksum is a one-way
+hash function; that is, it is computationally impractical to find
+the input corresponding to a particular output.
+This net result is
+a
+.Dq fingerprint
+of the input-data, which does not disclose the actual input.
+.Pp
+The
+.Fn SKEIN256_Init ,
+.Fn SKEIN256_Update ,
+and
+.Fn SKEIN256_Final
+functions are the core functions.
+Allocate an
+.Vt SKEIN256_CTX ,
+initialize it with
+.Fn SKEIN256_Init ,
+run over the data with
+.Fn SKEIN256_Update ,
+and finally extract the result using
+.Fn SKEIN256_Final .
+.Pp
+.Fn SKEIN256_End
+is a wrapper for
+.Fn SKEIN256_Final
+which converts the return value to a 33-character
+(including the terminating '\e0')
+.Tn ASCII
+string which represents the 256 bits in hexadecimal.
+.Pp
+.Fn SKEIN256_File
+calculates the digest of a file, and uses
+.Fn SKEIN256_End
+to return the result.
+If the file cannot be opened, a null pointer is returned.
+.Fn SKEIN256_FileChunk
+is similar to
+.Fn SKEIN256_File ,
+but it only calculates the digest over a byte-range of the file specified,
+starting at
+.Fa offset
+and spanning
+.Fa length
+bytes.
+If the
+.Fa length
+parameter is specified as 0, or more than the length of the remaining part
+of the file,
+.Fn SKEIN256_FileChunk
+calculates the digest from
+.Fa offset
+to the end of file.
+.Fn SKEIN256_Data
+calculates the digest of a chunk of data in memory, and uses
+.Fn SKEIN256_End
+to return the result.
+.Pp
+When using
+.Fn SKEIN256_End ,
+.Fn SKEIN256_File ,
+or
+.Fn SKEIN256_Data ,
+the
+.Fa buf
+argument can be a null pointer, in which case the returned string
+is allocated with
+.Xr malloc 3
+and subsequently must be explicitly deallocated using
+.Xr free 3
+after use.
+If the
+.Fa buf
+argument is non-null it must point to at least 33 characters of buffer space.
+.Pp
+The
+.Li SKEIN512_
+and
+.Li SKEIN1024_
+functions are similar to the
+.Li SKEIN256_
+functions except they produce a 512-bit, 65 character,
+or 1024-bit, 129 character, output.
+.Sh SEE ALSO
+.Xr md4 3 ,
+.Xr md5 3 ,
+.Xr ripemd 3 ,
+.Xr sha 3 ,
+.Xr sha256 3 ,
+.Xr sha512 3
+.Sh HISTORY
+These functions appeared in
+.Fx 11.0 .
+.Sh AUTHORS
+.An -nosplit
+The core hash routines were imported from version 1.3 of the optimized
+reference implementation written by
+.An Doug Whiting
+as submitted to the NSA SHA-3 contest.
+The algorithms were developed by
+.An Niels Ferguson ,
+.An Stefan Lucks ,
+.An Bruce Schneier ,
+.An Doug Whiting ,
+.An Mihir Bellare ,
+.An Tadayoshi Kohno ,
+.An Jon Callas,
+and
+.An Jesse Walker .
+.Sh BUGS
+No method is known to exist which finds two files having the same hash value,
+nor to find a file with a specific hash value.
+There is on the other hand no guarantee that such a method does not exist.
Index: lib/libmd/skeindriver.c
===================================================================
--- /dev/null
+++ lib/libmd/skeindriver.c
@@ -0,0 +1,68 @@
+/* SKEINDRIVER.C - test driver for SKEIN */
+
+/* Copyright (C) 1990-2, RSA Data Security, Inc. Created 1990. All rights
+ * reserved.
+ *
+ * RSA Data Security, Inc. makes no representations concerning either the
+ * merchantability of this software or the suitability of this software for
+ * any particular purpose. It is provided "as is" without express or implied
+ * warranty of any kind.
+ *
+ * These notices must be retained in any copies of any part of this
+ * documentation and/or software. */
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/types.h>
+
+#include <stdio.h>
+#include <time.h>
+#include <string.h>
+
+#include "skein.h"
+
+/* The following makes SKEIN default to SKEIN512 if it has not already been
+ * defined with C compiler flags. */
+#ifndef SKEIN
+#define SKEIN 512
+#endif
+
+#if SKEIN == 256
+#undef SKEIN_Data
+#define SKEIN_Data SKEIN256_Data
+#elif SKEIN == 512
+#undef SKEIN_Data
+#define SKEIN_Data SKEIN512_Data
+#elif SKEIN == 1024
+#undef SKEIN_Data
+#define SKEIN_Data SKEIN1024_Data
+#endif
+
+/* Digests a string and prints the result. */
+static void
+SKEINString(char *string)
+{
+ char buf[2*128 + 1];
+
+ printf("SKEIN%d (\"%s\") = %s\n",
+ SKEIN, string, SKEIN_Data(string, strlen(string), buf));
+}
+
+/* Digests a reference suite of strings and prints the results. */
+int
+main(void)
+{
+ printf("SKEIN%d test suite:\n", SKEIN);
+
+ SKEINString("");
+ SKEINString("abc");
+ SKEINString("message digest");
+ SKEINString("abcdefghijklmnopqrstuvwxyz");
+ SKEINString("ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+ "abcdefghijklmnopqrstuvwxyz0123456789");
+ SKEINString("1234567890123456789012345678901234567890"
+ "1234567890123456789012345678901234567890");
+
+ return 0;
+}
Index: sbin/md5/Makefile
===================================================================
--- sbin/md5/Makefile
+++ sbin/md5/Makefile
@@ -8,7 +8,10 @@
${BINDIR}/md5 ${BINDIR}/sha1 \
${BINDIR}/md5 ${BINDIR}/sha256 \
${BINDIR}/md5 ${BINDIR}/sha384 \
- ${BINDIR}/md5 ${BINDIR}/sha512
+ ${BINDIR}/md5 ${BINDIR}/sha512 \
+ ${BINDIR}/md5 ${BINDIR}/skein256 \
+ ${BINDIR}/md5 ${BINDIR}/skein512 \
+ ${BINDIR}/md5 ${BINDIR}/skein1024
MLINKS= md5.1 rmd160.1 \
md5.1 sha1.1 \
Index: sbin/md5/md5.c
===================================================================
--- sbin/md5/md5.c
+++ sbin/md5/md5.c
@@ -30,6 +30,7 @@
#include <sha256.h>
#include <sha384.h>
#include <sha512.h>
+#include <skein.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
@@ -59,6 +60,9 @@
extern const char *SHA384_TestOutput[MDTESTCOUNT];
extern const char *SHA512_TestOutput[MDTESTCOUNT];
extern const char *RIPEMD160_TestOutput[MDTESTCOUNT];
+extern const char *SKEIN256_TestOutput[MDTESTCOUNT];
+extern const char *SKEIN512_TestOutput[MDTESTCOUNT];
+extern const char *SKEIN1024_TestOutput[MDTESTCOUNT];
typedef struct Algorithm_t {
const char *progname;
@@ -85,11 +89,14 @@
SHA384_CTX sha384;
SHA512_CTX sha512;
RIPEMD160_CTX ripemd160;
+ SKEIN256_CTX skein256;
+ SKEIN512_CTX skein512;
+ SKEIN1024_CTX skein1024;
} DIGEST_CTX;
/* max(MD5_DIGEST_LENGTH, SHA_DIGEST_LENGTH,
SHA256_DIGEST_LENGTH, SHA512_DIGEST_LENGTH,
- RIPEMD160_DIGEST_LENGTH)*2+1 */
+ RIPEMD160_DIGEST_LENGTH, SKEIN1024_DIGEST_LENGTH)*2+1 */
#define HEX_DIGEST_LENGTH 129
/* algorithm function table */
@@ -112,7 +119,19 @@
&SHA512_Data, &SHA512_File },
{ "rmd160", "RMD160", &RIPEMD160_TestOutput,
(DIGEST_Init*)&RIPEMD160_Init, (DIGEST_Update*)&RIPEMD160_Update,
- (DIGEST_End*)&RIPEMD160_End, &RIPEMD160_Data, &RIPEMD160_File }
+ (DIGEST_End*)&RIPEMD160_End, &RIPEMD160_Data, &RIPEMD160_File },
+ { "skein256", "Skein256", &SKEIN256_TestOutput,
+ (DIGEST_Init*)&SKEIN256_Init, (DIGEST_Update*)&SKEIN256_Update,
+ (DIGEST_End*)&SKEIN256_End, &SKEIN256_Data, &SKEIN256_File,
+ (DIGEST_InitExt*)&Skein_256_InitExt, 256 },
+ { "skein512", "Skein512", &SKEIN512_TestOutput,
+ (DIGEST_Init*)&SKEIN512_Init, (DIGEST_Update*)&SKEIN512_Update,
+ (DIGEST_End*)&SKEIN512_End, &SKEIN512_Data, &SKEIN512_File,
+ (DIGEST_InitExt*)&Skein_512_InitExt, 512 },
+ { "skein1024", "Skein1024", &SKEIN1024_TestOutput,
+ (DIGEST_Init*)&SKEIN1024_Init, (DIGEST_Update*)&SKEIN1024_Update,
+ (DIGEST_End*)&SKEIN1024_End, &SKEIN1024_Data, &SKEIN1024_File,
+ (DIGEST_InitExt*)&Skein1024_InitExt, 1024 }
};
static void
@@ -281,8 +300,8 @@
printf(" done\n");
printf("Digest = %s", p);
printf("\nTime = %f seconds\n", seconds);
- printf("Speed = %f bytes/second\n",
- (float) TEST_BLOCK_LEN * (float) TEST_BLOCK_COUNT / seconds);
+ printf("Speed = %f MiB/second\n", (float) TEST_BLOCK_LEN *
+ (float) TEST_BLOCK_COUNT / seconds / (1 << 20));
}
/*
* Digests a reference suite of strings and prints the results.
@@ -366,6 +385,39 @@
"5feb69c6bf7c29d95715ad55f57d8ac5b2b7dd32"
};
+const char *SKEIN256_TestOutput[MDTESTCOUNT] = {
+ "c8877087da56e072870daa843f176e9453115929094c3a40c463a196c29bf7ba",
+ "7fba44ff1a31d71a0c1f82e6e82fb5e9ac6c92a39c9185b9951fed82d82fe635",
+ "258bdec343b9fde1639221a5ae0144a96e552e5288753c5fec76c05fc2fc1870",
+ "4d2ce0062b5eb3a4db95bc1117dd8aa014f6cd50fdc8e64f31f7d41f9231e488",
+ "46d8440685461b00e3ddb891b2ecc6855287d2bd8834a95fb1c1708b00ea5e82",
+ "7c5eb606389556b33d34eb2536459528dc0af97adbcd0ce273aeb650f598d4b2",
+ "4def7a7e5464a140ae9c3a80279fbebce4bd00f9faad819ab7e001512f67a10d",
+ "d9c017dbe355f318d036469eb9b5fbe129fc2b5786a9dc6746a516eab6fe0126"
+};
+
+const char *SKEIN512_TestOutput[MDTESTCOUNT] = {
+ "bc5b4c50925519c290cc634277ae3d6257212395cba733bbad37a4af0fa06af41fca7903d06564fea7a2d3730dbdb80c1f85562dfcc070334ea4d1d9e72cba7a",
+ "b1cd8d33f61b3737adfd59bb13ad82f4a9548e92f22956a8976cca3fdb7fee4fe91698146c4197cec85d38b83c5d93bdba92c01fd9a53870d0c7f967bc62bdce",
+ "8f5dd9ec798152668e35129496b029a960c9a9b88662f7f9482f110b31f9f93893ecfb25c009baad9e46737197d5630379816a886aa05526d3a70df272d96e75",
+ "15b73c158ffb875fed4d72801ded0794c720b121c0c78edf45f900937e6933d9e21a3a984206933d504b5dbb2368000411477ee1b204c986068df77886542fcc",
+ "23793ad900ef12f9165c8080da6fdfd2c8354a2929b8aadf83aa82a3c6470342f57cf8c035ec0d97429b626c4d94f28632c8f5134fd367dca5cf293d2ec13f8c",
+ "0c6bed927e022f5ddcf81877d42e5f75798a9f8fd3ede3d83baac0a2f364b082e036c11af35fe478745459dd8f5c0b73efe3c56ba5bb2009208d5a29cc6e469c",
+ "2ca9fcffb3456f297d1b5f407014ecb856f0baac8eb540f534b1f187196f21e88f31103128c2f03fcc9857d7a58eb66f9525e2302d88833ee069295537a434ce",
+ "1131f2aaa0e97126c9314f9f968cc827259bbfabced2943bb8c9274448998fb3b78738b4580dd500c76105fd3c03e465e1414f2c29664286b1f79d3e51128125"
+};
+
+const char *SKEIN1024_TestOutput[MDTESTCOUNT] = {
+ "0fff9563bb3279289227ac77d319b6fff8d7e9f09da1247b72a0a265cd6d2a62645ad547ed8193db48cff847c06494a03f55666d3b47eb4c20456c9373c86297d630d5578ebd34cb40991578f9f52b18003efa35d3da6553ff35db91b81ab890bec1b189b7f52cb2a783ebb7d823d725b0b4a71f6824e88f68f982eefc6d19c6",
+ "6ab4c4ba9814a3d976ec8bffa7fcc638ceba0544a97b3c98411323ffd2dc936315d13dc93c13c4e88cda6f5bac6f2558b2d8694d3b6143e40d644ae43ca940685cb37f809d3d0550c56cba8036dee729a4f8fb960732e59e64d57f7f7710f8670963cdcdc95b41daab4855fcf8b6762a64b173ee61343a2c7689af1d293eba97",
+ "35a599a0f91abcdb4cb73c19b8cb8d947742d82c309137a7caed29e8e0a2ca7a9ff9a90c34c1908cc7e7fd99bb15032fb86e76df21b72628399b5f7c3cc209d7bb31c99cd4e19465622a049afbb87c03b5ce3888d17e6e667279ec0aa9b3e2712624c01b5f5bbe1a564220bdcf6990af0c2539019f313fdd7406cca3892a1f1f",
+ "ea891f5268acd0fac97467fc1aa89d1ce8681a9992a42540e53babee861483110c2d16f49e73bac27653ff173003e40cfb08516cd34262e6af95a5d8645c9c1abb3e813604d508b8511b30f9a5c1b352aa0791c7d2f27b2706dccea54bc7de6555b5202351751c3299f97c09cf89c40f67187e2521c0fad82b30edbb224f0458",
+ "f23d95c2a25fbcd0e797cd058fec39d3c52d2b5afd7a9af1df934e63257d1d3dcf3246e7329c0f1104c1e51e3d22e300507b0c3b9f985bb1f645ef49835080536becf83788e17fed09c9982ba65c3cb7ffe6a5f745b911c506962adf226e435c42f6f6bc08d288f9c810e807e3216ef444f3db22744441deefa4900982a1371f",
+ "cf3889e8a8d11bfd3938055d7d061437962bc5eac8ae83b1b71c94be201b8cf657fdbfc38674997a008c0c903f56a23feb3ae30e012377f1cfa080a9ca7fe8b96138662653fb3335c7d06595bf8baf65e215307532094cfdfa056bd8052ab792a3944a2adaa47b30335b8badb8fe9eb94fe329cdca04e58bbc530f0af709f469",
+ "cf21a613620e6c119eca31fdfaad449a8e02f95ca256c21d2a105f8e4157048f9fe1e897893ea18b64e0e37cb07d5ac947f27ba544caf7cbc1ad094e675aed77a366270f7eb7f46543bccfa61c526fd628408058ed00ed566ac35a9761d002e629c4fb0d430b2f4ad016fcc49c44d2981c4002da0eecc42144160e2eaea4855a",
+ "e6799b78db54085a2be7ff4c8007f147fa88d326abab30be0560b953396d8802feee9a15419b48a467574e9283be15685ca8a079ee52b27166b64dd70b124b1d4e4f6aca37224c3f2685e67e67baef9f94b905698adc794a09672aba977a61b20966912acdb08c21a2c37001785355dc884751a21f848ab36e590331ff938138"
+};
+
static void
MDTestSuite(const Algorithm_t *alg)
{
Index: sys/contrib/skein/SHA3api_ref.h
===================================================================
--- /dev/null
+++ sys/contrib/skein/SHA3api_ref.h
@@ -0,0 +1,66 @@
+#ifndef _AHS_API_H_
+#define _AHS_API_H_
+
+/***********************************************************************
+**
+** Interface declarations of the AHS API using the Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+
+#include "skein.h"
+
+typedef enum
+ {
+ SUCCESS = SKEIN_SUCCESS,
+ FAIL = SKEIN_FAIL,
+ BAD_HASHLEN = SKEIN_BAD_HASHLEN
+ }
+ HashReturn;
+
+typedef size_t DataLength; /* bit count type */
+typedef u08b_t BitSequence; /* bit stream type */
+
+typedef struct
+ {
+ uint_t statebits; /* 256, 512, or 1024 */
+ union
+ {
+ Skein_Ctxt_Hdr_t h; /* common header "overlay" */
+ Skein_256_Ctxt_t ctx_256;
+ Skein_512_Ctxt_t ctx_512;
+ Skein1024_Ctxt_t ctx1024;
+ } u;
+ }
+ hashState;
+
+/* "incremental" hashing API */
+HashReturn Init (hashState *state, int hashbitlen);
+HashReturn Update(hashState *state, const BitSequence *data, DataLength databitlen);
+HashReturn Final (hashState *state, BitSequence *hashval);
+
+/* "all-in-one" call */
+HashReturn Hash (int hashbitlen, const BitSequence *data,
+ DataLength databitlen, BitSequence *hashval);
+
+
+/*
+** Re-define the compile-time constants below to change the selection
+** of the Skein state size in the Init() function in SHA3api_ref.c.
+**
+** That is, the NIST API does not allow for explicit selection of the
+** Skein block size, so it must be done implicitly in the Init() function.
+** The selection is controlled by these constants.
+*/
+#ifndef SKEIN_256_NIST_MAX_HASHBITS
+#define SKEIN_256_NIST_MAX_HASHBITS (0)
+#endif
+
+#ifndef SKEIN_512_NIST_MAX_HASHBITS
+#define SKEIN_512_NIST_MAX_HASHBITS (512)
+#endif
+
+#endif /* ifdef _AHS_API_H_ */
Index: sys/contrib/skein/SHA3api_ref.c
===================================================================
--- /dev/null
+++ sys/contrib/skein/SHA3api_ref.c
@@ -0,0 +1,115 @@
+/***********************************************************************
+**
+** Implementation of the AHS API using the Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+
+#include <string.h> /* get the memcpy/memset functions */
+#include "skein.h" /* get the Skein API definitions */
+#include "SHA3api_ref.h"/* get the AHS API definitions */
+
+/******************************************************************/
+/* AHS API code */
+/******************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* select the context size and init the context */
+HashReturn Init(hashState *state, int hashbitlen)
+ {
+#if SKEIN_256_NIST_MAX_HASH_BITS
+ if (hashbitlen <= SKEIN_256_NIST_MAX_HASHBITS)
+ {
+ Skein_Assert(hashbitlen > 0,BAD_HASHLEN);
+ state->statebits = 64*SKEIN_256_STATE_WORDS;
+ return Skein_256_Init(&state->u.ctx_256,(size_t) hashbitlen);
+ }
+#endif
+ if (hashbitlen <= SKEIN_512_NIST_MAX_HASHBITS)
+ {
+ state->statebits = 64*SKEIN_512_STATE_WORDS;
+ return Skein_512_Init(&state->u.ctx_512,(size_t) hashbitlen);
+ }
+ else
+ {
+ state->statebits = 64*SKEIN1024_STATE_WORDS;
+ return Skein1024_Init(&state->u.ctx1024,(size_t) hashbitlen);
+ }
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* process data to be hashed */
+HashReturn Update(hashState *state, const BitSequence *data, DataLength databitlen)
+ {
+ /* only the final Update() call is allowed do partial bytes, else assert an error */
+ Skein_Assert((state->u.h.T[1] & SKEIN_T1_FLAG_BIT_PAD) == 0 || databitlen == 0, FAIL);
+
+ Skein_Assert(state->statebits % 256 == 0 && (state->statebits-256) < 1024,FAIL);
+ if ((databitlen & 7) == 0) /* partial bytes? */
+ {
+ switch ((state->statebits >> 8) & 3)
+ {
+ case 2: return Skein_512_Update(&state->u.ctx_512,data,databitlen >> 3);
+ case 1: return Skein_256_Update(&state->u.ctx_256,data,databitlen >> 3);
+ case 0: return Skein1024_Update(&state->u.ctx1024,data,databitlen >> 3);
+ default: return FAIL;
+ }
+ }
+ else
+ { /* handle partial final byte */
+ size_t bCnt = (databitlen >> 3) + 1; /* number of bytes to handle (nonzero here!) */
+ u08b_t b,mask;
+
+ mask = (u08b_t) (1u << (7 - (databitlen & 7))); /* partial byte bit mask */
+ b = (u08b_t) ((data[bCnt-1] & (0-mask)) | mask); /* apply bit padding on final byte */
+
+ switch ((state->statebits >> 8) & 3)
+ {
+ case 2: Skein_512_Update(&state->u.ctx_512,data,bCnt-1); /* process all but the final byte */
+ Skein_512_Update(&state->u.ctx_512,&b , 1 ); /* process the (masked) partial byte */
+ break;
+ case 1: Skein_256_Update(&state->u.ctx_256,data,bCnt-1); /* process all but the final byte */
+ Skein_256_Update(&state->u.ctx_256,&b , 1 ); /* process the (masked) partial byte */
+ break;
+ case 0: Skein1024_Update(&state->u.ctx1024,data,bCnt-1); /* process all but the final byte */
+ Skein1024_Update(&state->u.ctx1024,&b , 1 ); /* process the (masked) partial byte */
+ break;
+ default: return FAIL;
+ }
+ Skein_Set_Bit_Pad_Flag(state->u.h); /* set tweak flag for the final call */
+
+ return SUCCESS;
+ }
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize hash computation and output the result (hashbitlen bits) */
+HashReturn Final(hashState *state, BitSequence *hashval)
+ {
+ Skein_Assert(state->statebits % 256 == 0 && (state->statebits-256) < 1024,FAIL);
+ switch ((state->statebits >> 8) & 3)
+ {
+ case 2: return Skein_512_Final(&state->u.ctx_512,hashval);
+ case 1: return Skein_256_Final(&state->u.ctx_256,hashval);
+ case 0: return Skein1024_Final(&state->u.ctx1024,hashval);
+ default: return FAIL;
+ }
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* all-in-one hash function */
+HashReturn Hash(int hashbitlen, const BitSequence *data, /* all-in-one call */
+ DataLength databitlen,BitSequence *hashval)
+ {
+ hashState state;
+ HashReturn r = Init(&state,hashbitlen);
+ if (r == SUCCESS)
+ { /* these calls do not fail when called properly */
+ r = Update(&state,data,databitlen);
+ Final(&state,hashval);
+ }
+ return r;
+ }
Index: sys/contrib/skein/asm/skein_block_x64.asm
===================================================================
--- /dev/null
+++ sys/contrib/skein/asm/skein_block_x64.asm
@@ -0,0 +1,1335 @@
+;
+;----------------------------------------------------------------
+; 64-bit x86 assembler code (Microsoft ML64) for Skein block functions
+;
+; Author: Doug Whiting, Hifn
+;
+; This code is released to the public domain.
+;----------------------------------------------------------------
+;
+ .code
+;
+_MASK_ALL_ equ (256+512+1024) ;all three algorithm bits
+_MAX_FRAME_ equ 240
+;
+;;;;;;;;;;;;;;;;;
+ifndef SKEIN_USE_ASM
+_USE_ASM_ = _MASK_ALL_
+elseif SKEIN_USE_ASM and _MASK_ALL_
+_USE_ASM_ = SKEIN_USE_ASM
+else
+_USE_ASM_ = _MASK_ALL_
+endif
+;;;;;;;;;;;;;;;;;
+ifndef SKEIN_LOOP ;configure loop unrolling
+_SKEIN_LOOP = 0 ;default is all fully unrolled
+else
+_SKEIN_LOOP = SKEIN_LOOP
+endif
+; the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256 = (_SKEIN_LOOP / 100) mod 10
+SKEIN_UNROLL_512 = (_SKEIN_LOOP / 10) mod 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP ) mod 10
+;
+SKEIN_ASM_UNROLL = 0
+ irp _NN_,<256,512,1024>
+ if (SKEIN_UNROLL_&_NN_) eq 0
+SKEIN_ASM_UNROLL = SKEIN_ASM_UNROLL + _NN_
+ endif
+ endm
+;;;;;;;;;;;;;;;;;
+;
+ifndef SKEIN_ROUNDS
+ROUNDS_256 = 72
+ROUNDS_512 = 72
+ROUNDS_1024 = 80
+else
+ROUNDS_256 = 8*((((SKEIN_ROUNDS / 100) + 5) mod 10) + 5)
+ROUNDS_512 = 8*((((SKEIN_ROUNDS / 10) + 5) mod 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS ) + 5) mod 10) + 5)
+endif
+;
+irp _NN_,<256,512,1024>
+ if _USE_ASM_ and _NN_
+ irp _RR_,<%(ROUNDS_&_NN_)>
+ if _NN_ eq 1024
+%out +++ SKEIN_ROUNDS_&_NN_ = _RR_
+ else
+%out +++ SKEIN_ROUNDS_&_NN_ = _RR_
+ endif
+ endm
+ endif
+endm
+;;;;;;;;;;;;;;;;;
+;
+ifndef SKEIN_CODE_SIZE
+ifdef SKEIN_PERF
+SKEIN_CODE_SIZE equ (1)
+endif
+endif
+;
+;;;;;;;;;;;;;;;;;
+;
+ifndef SKEIN_DEBUG
+_SKEIN_DEBUG = 0
+else
+_SKEIN_DEBUG = 1
+endif
+;;;;;;;;;;;;;;;;;
+;
+; define offsets of fields in hash context structure
+;
+HASH_BITS = 0 ;# bits of hash output
+BCNT = 8 + HASH_BITS ;number of bytes in BUFFER[]
+TWEAK = 8 + BCNT ;tweak values[0..1]
+X_VARS = 16 + TWEAK ;chaining vars
+;
+;(Note: buffer[] in context structure is NOT needed here :-)
+;
+r08 equ <r8>
+r09 equ <r9>
+;
+KW_PARITY = 01BD11BDAA9FC1A22h ;overall parity of key schedule words
+FIRST_MASK = NOT (1 SHL 62)
+;
+; rotation constants for Skein
+;
+RC_256_0_0 = 14
+RC_256_0_1 = 16
+
+RC_256_1_0 = 52
+RC_256_1_1 = 57
+
+RC_256_2_0 = 23
+RC_256_2_1 = 40
+
+RC_256_3_0 = 5
+RC_256_3_1 = 37
+
+RC_256_4_0 = 25
+RC_256_4_1 = 33
+
+RC_256_5_0 = 46
+RC_256_5_1 = 12
+
+RC_256_6_0 = 58
+RC_256_6_1 = 22
+
+RC_256_7_0 = 32
+RC_256_7_1 = 32
+
+RC_512_0_0 = 46
+RC_512_0_1 = 36
+RC_512_0_2 = 19
+RC_512_0_3 = 37
+
+RC_512_1_0 = 33
+RC_512_1_1 = 27
+RC_512_1_2 = 14
+RC_512_1_3 = 42
+
+RC_512_2_0 = 17
+RC_512_2_1 = 49
+RC_512_2_2 = 36
+RC_512_2_3 = 39
+
+RC_512_3_0 = 44
+RC_512_3_1 = 9
+RC_512_3_2 = 54
+RC_512_3_3 = 56
+
+RC_512_4_0 = 39
+RC_512_4_1 = 30
+RC_512_4_2 = 34
+RC_512_4_3 = 24
+
+RC_512_5_0 = 13
+RC_512_5_1 = 50
+RC_512_5_2 = 10
+RC_512_5_3 = 17
+
+RC_512_6_0 = 25
+RC_512_6_1 = 29
+RC_512_6_2 = 39
+RC_512_6_3 = 43
+
+RC_512_7_0 = 8
+RC_512_7_1 = 35
+RC_512_7_2 = 56
+RC_512_7_3 = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 = 8
+RC_1024_0_3 = 47
+RC_1024_0_4 = 8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 = 4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 = 5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 = 9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 = 4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 = 9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+;
+; Input: reg
+; Output: <reg> <<< RC_BlkSize_roundNum_mixNum, BlkSize=256/512/1024
+;
+RotL64 macro reg,BLK_SIZE,ROUND_NUM,MIX_NUM
+_RCNT_ = ( RC_&BLK_SIZE&_&ROUND_NUM&_&MIX_NUM AND 63 )
+ if _RCNT_ ;is there anything to do?
+ rol reg,_RCNT_
+ endif
+endm
+;
+;----------------------------------------------------------------
+;
+; MACROS: define local vars and configure stack
+;
+;----------------------------------------------------------------
+; declare allocated space on the stack
+StackVar macro localName,localSize
+localName = _STK_OFFS_
+_STK_OFFS_ = _STK_OFFS_+(localSize)
+endm ;StackVar
+;
+;----------------------------------------------------------------
+;
+; MACRO: Configure stack frame, allocate local vars
+;
+Setup_Stack macro BLK_BITS,KS_CNT,NO_FRAME,debugCnt
+ WCNT = (BLK_BITS)/64
+;
+_PushCnt_ = 0 ;save nonvolatile regs on stack
+ irp _reg_,<rbp,rsi,rdi,rbx,r12,r13,r14,r15>
+ push _reg_
+ .pushreg _reg_ ;pseudo-op push for exception handling
+_PushCnt_ = _PushCnt_ + 1 ;track count to keep alignment
+ endm
+;
+_STK_OFFS_ = 0 ;starting offset from rsp
+ ;---- local variables ;<-- rsp
+ StackVar X_stk ,8*(WCNT) ;local context vars
+ StackVar ksTwk ,8*3 ;key schedule: tweak words
+ StackVar ksKey ,8*(WCNT)+8 ;key schedule: key words
+ if (SKEIN_ASM_UNROLL and (BLK_BITS)) eq 0
+ StackVar ksRot ,16*(KS_CNT+0);leave space for "rotation" to happen
+ endif
+ StackVar Wcopy ,8*(WCNT) ;copy of input block
+ if _SKEIN_DEBUG
+ ifnb <debugCnt> ;temp location for debug X[] info
+ StackVar xDebug_&BLK_BITS ,8*(debugCnt)
+ endif
+ endif
+ if ((8*_PushCnt_ + _STK_OFFS_) and 8) eq 0
+ StackVar align16,8 ;keep 16-byte aligned (adjust for retAddr?)
+tmpStk_&BLK_BITS = align16 ;use this
+ endif
+LOCAL_SIZE = _STK_OFFS_ ;size of local vars
+ ;----
+ StackVar savRegs,8*_PushCnt_ ;saved registers
+ StackVar retAddr,8 ;return address
+ ;---- caller parameters
+ StackVar ctxPtr ,8 ;context ptr
+ StackVar blkPtr ,8 ;pointer to block data
+ StackVar blkCnt ,8 ;number of full blocks to process
+ StackVar bitAdd ,8 ;bit count to add to tweak
+ ;---- caller's stack frame
+;
+; set up the stack frame pointer (rbp)
+;
+FRAME_OFFS = ksTwk + 128 ;allow short (negative) offset to ksTwk, kwKey
+ if FRAME_OFFS gt _STK_OFFS_ ;keep rbp in the "locals" range
+FRAME_OFFS = _STK_OFFS_
+ endif
+ if FRAME_OFFS gt _MAX_FRAME_ ;keep Microsoft .setframe happy
+FRAME_OFFS = _MAX_FRAME_
+ endif
+;
+ifdef SKEIN_ASM_INFO
+ if FRAME_OFFS+128 lt savRegs
+%out +++ SKEIN_&BLK_BITS: Unable to reach all of Wcopy with short offset from rbp.
+ elseif FRAME_OFFS+128 lt Wcopy
+%out +++ SKEIN_&BLK_BITS: Unable to reach end of Wcopy with short offset from rbp.
+ elseif FRAME_OFFS+128 lt _STK_OFFS_
+%out +++ SKEIN_&BLK_BITS: Unable to reach caller parms with short offset from rbp
+ endif
+endif
+ ;put some useful defines in the .lst file (for grep)
+__STK_LCL_SIZE_&BLK_BITS = LOCAL_SIZE
+__STK_TOT_SIZE_&BLK_BITS = _STK_OFFS_
+__STK_FRM_OFFS_&BLK_BITS = FRAME_OFFS
+;
+; Notes on stack frame setup:
+; * the most frequently used variable is X_stk[], based at [rsp+0]
+; * the next most used is the key schedule arrays, ksKey and ksTwk
+; so rbp is "centered" there, allowing short offsets to the key
+; schedule even in 1024-bit Skein case
+; * the Wcopy variables are infrequently accessed, but they have long
+; offsets from both rsp and rbp only in the 1024-bit case.
+; * all other local vars and calling parameters can be accessed
+; with short offsets, except in the 1024-bit case
+;
+ sub rsp,LOCAL_SIZE ;make room for the locals
+ .allocstack LOCAL_SIZE ;pseudo op for exception handling
+ lea rbp,[rsp+FRAME_OFFS] ;maximize use of short offsets
+ ifb <NO_FRAME>
+ .setframe rbp, FRAME_OFFS ;pseudo op for exception handling
+ endif
+ mov [FP_+ctxPtr],rcx ;save caller's parameters on the stack
+ mov [FP_+blkPtr],rdx
+ mov [FP_+blkCnt],r08
+ mov [FP_+bitAdd],r09
+ .endprolog ;pseudo op to support exception handling
+
+ mov rdi,[FP_+ctxPtr ] ;rdi --> context
+;
+endm ;Setup_Stack
+;
+FP_ equ <rbp-FRAME_OFFS> ;keep as many short offsets as possible
+;
+;----------------------------------------------------------------
+;
+Reset_Stack macro procStart
+ add rsp,LOCAL_SIZE ;get rid of locals (wipe??)
+ irp _reg_,<r15,r14,r13,r12,rbx,rdi,rsi,rbp>
+ pop _reg_
+_PushCnt_ = _PushCnt_ - 1
+ endm
+ if _PushCnt_
+ .err "Mismatched push/pops?"
+ endif
+
+ ;display code size in bytes to stdout
+ irp _BCNT_,<%($+1-procStart)> ;account for return opcode
+_ProcBytes_ = _BCNT_
+if _BCNT_ ge 10000
+%out procStart code size = _BCNT_ bytes
+elseif _BCNT_ ge 1000
+%out procStart code size = _BCNT_ bytes
+else
+%out procStart code size = _BCNT_ bytes
+endif
+ endm ;irp _BCNT_
+endm ; Reset_Stack
+;
+;----------------------------------------------------------------
+; macros to help debug internals
+;
+if _SKEIN_DEBUG
+ extrn Skein_Show_Block:proc ;calls to C routines
+ extrn Skein_Show_Round:proc
+;
+SKEIN_RND_SPECIAL = 1000
+SKEIN_RND_KEY_INITIAL = SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT = SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD = SKEIN_RND_SPECIAL+2
+;
+Skein_Debug_Block macro BLK_BITS
+;
+;void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+; const u08b_t *blkPtr, const u64b_t *wPtr,
+; const u64b_t *ksPtr,const u64b_t *tsPtr);
+;
+ irp _reg_,<rax,rcx,rdx,r08,r09,r10,r11>
+ push _reg_ ;save all volatile regs on tack before the call
+ endm
+ ; get and push call parameters
+ lea rax,[FP_+ksTwk] ;tweak pointer
+ push rax
+ lea rax,[FP_+ksKey] ;key pointer
+ push rax
+ lea rax,[FP_+Wcopy] ;wPtr
+ push rax
+ mov r09,[FP_+blkPtr] ;blkPtr
+ push r09 ;(push register parameters anyway to make room on stack)
+ mov rdx,[FP_+ctxPtr]
+ lea r08,[rdx+X_VARS] ;X (pointer)
+ push r08
+ push rdx ;h (pointer)
+ mov rcx, BLK_BITS ;bits
+ push rdx
+ call Skein_Show_Block ;call external debug handler
+ add rsp,7*8 ;discard parameters on stack
+ irp _reg_,<r11,r10,r09,r08,rdx,rcx,rax>
+ pop _reg_ ;restore regs
+ endm
+endm ; Skein_Debug_Block
+;
+;
+; the macro to "call" to debug a round
+;
+Skein_Debug_Round macro BLK_BITS,R,RDI_OFFS,afterOp
+ ; call the appropriate (local) debug function
+ push r08
+ if (SKEIN_ASM_UNROLL and BLK_BITS) or (R ge SKEIN_RND_SPECIAL)
+ mov r08, R
+ else ;compute round number using edi
+_rOffs_ = RDI_OFFS + 0
+ if BLK_BITS eq 1024
+ mov r08,[rsp+8+rIdx_offs] ;get rIdx off the stack (adjust for push r08)
+ lea r08,[4*r08+1+(((R)-1) and 3)+_rOffs_]
+ else
+ lea r08,[4*rdi+1+(((R)-1) and 3)+_rOffs_]
+ endif
+ endif
+ call Skein_Debug_Round_&BLK_BITS
+ pop r08
+;
+ afterOp
+endm ; Skein_Debug_Round
+else ;------- _SKEIN_DEBUG (dummy macros if debug not enabled)
+Skein_Debug_Block macro BLK_BITS,afterOp
+endm
+;
+Skein_Debug_Round macro BLK_BITS,R,RDI_OFFS,afterOp
+endm
+;
+endif ; _SKEIN_DEBUG
+;
+;----------------------------------------------------------------
+;
+addReg macro dstReg,srcReg_A,srcReg_B,useAddOp,immOffs
+ ifnb <immOffs>
+ lea dstReg,[srcReg_A&&srcReg_B + dstReg + immOffs]
+ elseif ((useAddOp + 0) eq 0)
+ ifndef ASM_NO_LEA
+ ;lea seems to be faster on Core 2 Duo CPUs!
+ lea dstReg,[srcReg_A&&srcReg_B + dstReg]
+ else
+ add dstReg, srcReg_A&&srcReg_B
+ endif
+ else
+ add dstReg, srcReg_A&&srcReg_B
+ endif
+endm
+;
+;=================================== Skein_256 =============================================
+;
+if _USE_ASM_ and 256
+ public Skein_256_Process_Block
+;
+; void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+;;;;;;;;;;;;;;;;;
+;
+; code
+;
+Skein_256_Process_Block proc frame
+ Setup_Stack 256,((ROUNDS_256/8)+1)
+ mov r14,[rdi+TWEAK+8]
+ jmp short Skein_256_block_loop
+ align 16
+ ; main hash loop for Skein_256
+Skein_256_block_loop:
+ ;
+ ; general register usage:
+ ; RAX..RDX = X0..X3
+ ; R08..R12 = ks[0..4]
+ ; R13..R15 = ts[0..2]
+ ; RSP, RBP = stack/frame pointers
+ ; RDI = round counter or context pointer
+ ; RSI = temp
+ ;
+ mov r13,[rdi+TWEAK+0]
+ add r13,[FP_+bitAdd] ;computed updated tweak value T0
+ mov r15,r14
+ xor r15,r13 ;now r13.r15 is set as the tweak
+
+ mov r12,KW_PARITY
+ mov r08,[rdi+X_VARS+ 0]
+ mov r09,[rdi+X_VARS+ 8]
+ mov r10,[rdi+X_VARS+16]
+ mov r11,[rdi+X_VARS+24]
+ mov [rdi+TWEAK+0],r13 ;save updated tweak value ctx->h.T[0]
+ xor r12,r08 ;start accumulating overall parity
+
+ mov rsi,[FP_+blkPtr ] ;esi --> input block
+ xor r12,r09
+ mov rax,[rsi+ 0] ;get X[0..3]
+ xor r12,r10
+ mov rbx,[rsi+ 8]
+ xor r12,r11
+ mov rcx,[rsi+16]
+ mov rdx,[rsi+24]
+
+ mov [FP_+Wcopy+ 0],rax ;save copy of input block
+ mov [FP_+Wcopy+ 8],rbx
+ mov [FP_+Wcopy+16],rcx
+ mov [FP_+Wcopy+24],rdx
+
+ add rax, r08 ;initial key injection
+ add rbx, r09
+ add rcx, r10
+ add rdx, r11
+ add rbx, r13
+ add rcx, r14
+
+if _SKEIN_DEBUG
+ mov [rdi+TWEAK+ 8],r14 ;save updated tweak T[1] (start bit cleared?)
+ mov [FP_+ksKey+ 0],r08 ;save key schedule on stack for Skein_Debug_Block
+ mov [FP_+ksKey+ 8],r09
+ mov [FP_+ksKey+16],r10
+ mov [FP_+ksKey+24],r11
+ mov [FP_+ksKey+32],r12
+
+ mov [FP_+ksTwk+ 0],r13
+ mov [FP_+ksTwk+ 8],r14
+ mov [FP_+ksTwk+16],r15
+
+ mov [rsp+X_stk + 0],rax ;save X[] on stack for Skein_Debug_Block
+ mov [rsp+X_stk + 8],rbx
+ mov [rsp+X_stk +16],rcx
+ mov [rsp+X_stk +24],rdx
+
+ Skein_Debug_Block 256 ;debug dump
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL
+endif
+;
+if ((SKEIN_ASM_UNROLL and 256) eq 0)
+ mov [FP_+ksKey+40],r08 ;save key schedule on stack for looping code
+ mov [FP_+ksKey+ 8],r09
+ mov [FP_+ksKey+16],r10
+ mov [FP_+ksKey+24],r11
+ mov [FP_+ksKey+32],r12
+
+ mov [FP_+ksTwk+24],r13
+ mov [FP_+ksTwk+ 8],r14
+ mov [FP_+ksTwk+16],r15
+endif
+ add rsi, WCNT*8 ;skip the block
+ mov [FP_+blkPtr ],rsi ;update block pointer
+;
+opLoop macro op1,op2
+ if (SKEIN_ASM_UNROLL and 256) eq 0
+ op1
+ else
+ op2
+ endif
+endm
+;
+ ;
+ ; now the key schedule is computed. Start the rounds
+ ;
+if SKEIN_ASM_UNROLL and 256
+_UNROLL_CNT = ROUNDS_256/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_256
+ if ((ROUNDS_256/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_256"
+ endif
+ xor rdi,rdi ;rdi = iteration count
+Skein_256_round_loop:
+endif
+_Rbase_ = 0
+rept _UNROLL_CNT*2
+ ; all X and ks vars in regs ; (ops to "rotate" ks vars, via mem, if not unrolled)
+ ; round 4*_RBase_ + 0
+ addReg rax, rbx
+ RotL64 rbx, 256,%((4*_RBase_+0) and 7),0
+ addReg rcx, rdx
+ opLoop <mov r08,[FP_+ksKey+8*rdi+8*1]>
+ xor rbx, rax
+ RotL64 rdx, 256,%((4*_RBase_+0) and 7),1
+ xor rdx, rcx
+ if SKEIN_ASM_UNROLL and 256
+ irp _r0_,<%(08+(_Rbase_+3) mod 5)>
+ irp _r1_,<%(13+(_Rbase_+2) mod 3)>
+ lea rdi,[r&_r0_+r&_r1_] ;precompute key injection value for rcx
+ endm
+ endm
+ endif
+ opLoop <mov r13,[FP_+ksTwk+8*rdi+8*1]>
+ Skein_Debug_Round 256,%(4*_RBase_+1)
+
+ ; round 4*_RBase_ + 1
+ addReg rax, rdx
+ RotL64 rdx, 256,%((4*_RBase_+1) and 7),0
+ xor rdx, rax
+ opLoop <mov r09,[FP_+ksKey+8*rdi+8*2]>
+ addReg rcx, rbx
+ RotL64 rbx, 256,%((4*_RBase_+1) and 7),1
+ xor rbx, rcx
+ opLoop <mov r11,[FP_+ksKey+8*rdi+8*4]>
+ Skein_Debug_Round 256,%(4*_RBase_+2)
+ if SKEIN_ASM_UNROLL and 256
+ irp _r0_,<%(08+(_Rbase_+2) mod 5)>
+ irp _r1_,<%(13+(_Rbase_+1) mod 3)>
+ lea rsi,[r&_r0_+r&_r1_] ;precompute key injection value for rbx
+ endm
+ endm
+ endif
+ ; round 4*_RBase_ + 2
+ addReg rax, rbx
+ RotL64 rbx, 256,%((4*_RBase_+2) and 7),0
+ addReg rcx, rdx
+ opLoop <mov r10,[FP_+ksKey+8*rdi+8*3]>
+ xor rbx, rax
+ RotL64 rdx, 256,%((4*_RBase_+2) and 7),1
+ xor rdx, rcx
+ opLoop <mov [FP_+ksKey+8*rdi+8*6],r08> ;"rotate" the key
+ opLoop <lea r11,[r11+rdi+1]> ;precompute key + tweak
+ Skein_Debug_Round 256,%(4*_RBase_+3)
+ ; round 4*_RBase_ + 3
+ addReg rax, rdx
+ RotL64 rdx, 256,%((4*_RBase_+3) and 7),0
+ addReg rcx, rbx
+ opLoop <add r10,[FP_+ksTwk+8*rdi+8*2]> ;precompute key + tweak
+ opLoop <mov [FP_+ksTwk+8*rdi+8*4],r13> ;"rotate" the tweak
+ xor rdx, rax
+ RotL64 rbx, 256,%((4*_RBase_+3) and 7),1
+ xor rbx, rcx
+ Skein_Debug_Round 256,%(4*_RBase_+4)
+ opLoop <addReg r09,r13> ;precompute key+tweak
+ ;inject key schedule words
+_Rbase_ = _Rbase_+1
+ if SKEIN_ASM_UNROLL and 256
+ addReg rax,r,%(08+((_Rbase_+0) mod 5))
+ addReg rbx,rsi
+ addReg rcx,rdi
+ addReg rdx,r,%(08+((_Rbase_+3) mod 5)),,_Rbase_
+ else
+ inc rdi
+ addReg rax,r08
+ addReg rcx,r10
+ addReg rbx,r09
+ addReg rdx,r11
+ endif
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT
+endm ;rept _UNROLL_CNT
+
+;
+if (SKEIN_ASM_UNROLL and 256) eq 0
+ cmp rdi,2*(ROUNDS_256/8)
+ jb Skein_256_round_loop
+endif ; (SKEIN_ASM_UNROLL and 256) eq 0
+ mov rdi,[FP_+ctxPtr ] ;restore edi --> context
+
+ ;----------------------------
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..3}
+ xor rax,[FP_+Wcopy + 0]
+ mov r14,FIRST_MASK
+ xor rbx,[FP_+Wcopy + 8]
+ xor rcx,[FP_+Wcopy +16]
+ xor rdx,[FP_+Wcopy +24]
+ mov [rdi+X_VARS+ 0],rax ;store final result
+ and r14,[rdi+TWEAK + 8]
+ dec qword ptr [FP_+blkCnt] ;set zero flag
+ mov [rdi+X_VARS+ 8],rbx
+ mov [rdi+X_VARS+16],rcx
+ mov [rdi+X_VARS+24],rdx
+
+ Skein_Debug_Round 256,SKEIN_RND_FEED_FWD,,<cmp qword ptr [FP_+blkCnt],0>
+
+ ; go back for more blocks, if needed
+ jnz Skein_256_block_loop
+ mov [rdi+TWEAK + 8],r14
+ Reset_Stack Skein_256_Process_Block
+ ret
+
+ if _SKEIN_DEBUG
+Skein_Debug_Round_256:
+ mov [FP_+X_stk+ 0],rax ;first, save X[] state on stack so debug routines can access it
+ mov [FP_+X_stk+ 8],rbx ;(use FP_ since rsp has changed!)
+ mov [FP_+X_stk+16],rcx
+ mov [FP_+X_stk+24],rdx
+ push rdx ;save two regs for BLK_BITS-specific parms
+ push rcx
+ mov rdx,[FP_+ctxPtr] ;ctx_hdr_ptr
+ mov rcx, 256
+ jmp Skein_Debug_Round_Common
+ endif
+
+Skein_256_Process_Block endp
+;
+ifdef SKEIN_CODE_SIZE
+ public Skein_256_Process_Block_CodeSize
+Skein_256_Process_Block_CodeSize proc
+ mov rax,_ProcBytes_
+ ret
+Skein_256_Process_Block_CodeSize endp
+;
+ public Skein_256_Unroll_Cnt
+Skein_256_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_256/8
+ mov rax,_UNROLL_CNT
+ else
+ xor rax,rax
+ endif
+ ret
+Skein_256_Unroll_Cnt endp
+endif
+;
+endif ;_USE_ASM_ and 256
+;
+;=================================== Skein_512 =============================================
+;
+if _USE_ASM_ and 512
+ public Skein_512_Process_Block
+;
+; void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+rX_512_0 equ r08 ;register assignments for X[] values during rounds
+rX_512_1 equ r09
+rX_512_2 equ r10
+rX_512_3 equ r11
+rX_512_4 equ r12
+rX_512_5 equ r13
+rX_512_6 equ r14
+rX_512_7 equ r15
+;
+;;;;;;;;;;;;;;;;;
+; MACRO: one round for 512-bit blocks
+;
+R_512_OneRound macro r0,r1,r2,r3,r4,r5,r6,r7,_Rn_,op1,op2,op3,op4
+;
+ addReg rX_512_&r0, rX_512_&r1
+ RotL64 rX_512_&r1, 512,%((_Rn_) and 7),0
+ xor rX_512_&r1, rX_512_&r0
+ op1
+ addReg rX_512_&r2, rX_512_&r3
+ RotL64 rX_512_&r3, 512,%((_Rn_) and 7),1
+ xor rX_512_&r3, rX_512_&r2
+ op2
+ addReg rX_512_&r4, rX_512_&r5
+ RotL64 rX_512_&r5, 512,%((_Rn_) and 7),2
+ xor rX_512_&r5, rX_512_&r4
+ op3
+ addReg rX_512_&r6, rX_512_&r7
+ RotL64 rX_512_&r7, 512,%((_Rn_) and 7),3
+ xor rX_512_&r7, rX_512_&r6
+ op4
+ Skein_Debug_Round 512,%(_Rn_+1),-4
+;
+endm ;R_512_OneRound
+;
+;;;;;;;;;;;;;;;;;
+; MACRO: eight rounds for 512-bit blocks
+;
+R_512_FourRounds macro _RR_ ;RR = base round number (0 mod 8)
+ if SKEIN_ASM_UNROLL and 512
+ ; here for fully unrolled case.
+ _II_ = ((_RR_)/4) + 1 ;key injection counter
+ R_512_OneRound 0,1,2,3,4,5,6,7,%((_RR_)+0),<mov rax,[FP_+ksKey+8*(((_II_)+3) mod 9)]>,,<mov rbx,[FP_+ksKey+8*(((_II_)+4) mod 9)]>
+ R_512_OneRound 2,1,4,7,6,5,0,3,%((_RR_)+1),<mov rcx,[FP_+ksKey+8*(((_II_)+5) mod 9)]>,,<mov rdx,[FP_+ksKey+8*(((_II_)+6) mod 9)]>
+ R_512_OneRound 4,1,6,3,0,5,2,7,%((_RR_)+2),<mov rsi,[FP_+ksKey+8*(((_II_)+7) mod 9)]>,,<add rcx,[FP_+ksTwk+8*(((_II_)+0) mod 3)]>
+ R_512_OneRound 6,1,0,7,2,5,4,3,%((_RR_)+3),<add rdx,[FP_+ksTwk+8*(((_II_)+1) mod 3)]>,
+ ; inject the key schedule
+ add r08,[FP_+ksKey+8*(((_II_)+0) mod 9)]
+ addReg r11,rax
+ add r09,[FP_+ksKey+8*(((_II_)+1) mod 9)]
+ addReg r12,rbx
+ add r10,[FP_+ksKey+8*(((_II_)+2) mod 9)]
+ addReg r13,rcx
+ addReg r14,rdx
+ addReg r15,rsi,,,(_II_)
+ else
+ ; here for looping case ;"rotate" key/tweak schedule (move up on stack)
+ inc rdi ;bump key injection counter
+ R_512_OneRound 0,1,2,3,4,5,6,7,%((_RR_)+0),<mov rdx,[FP_+ksKey+8*rdi+8*6]>,<mov rax,[FP_+ksTwk+8*rdi-8*1]> ,<mov rsi,[FP_+ksKey+8*rdi-8*1]>
+ R_512_OneRound 2,1,4,7,6,5,0,3,%((_RR_)+1),<mov rcx,[FP_+ksKey+8*rdi+8*5]>,<mov [FP_+ksTwk+8*rdi+8*2],rax>,<mov [FP_+ksKey+8*rdi+8*8],rsi>
+ R_512_OneRound 4,1,6,3,0,5,2,7,%((_RR_)+2),<mov rbx,[FP_+ksKey+8*rdi+8*4]>,<add rdx,[FP_+ksTwk+8*rdi+8*1]> ,<mov rsi,[FP_+ksKey+8*rdi+8*7]>
+ R_512_OneRound 6,1,0,7,2,5,4,3,%((_RR_)+3),<mov rax,[FP_+ksKey+8*rdi+8*3]>,<add rcx,[FP_+ksTwk+8*rdi+8*0]>
+ ; inject the key schedule
+ add r08,[FP_+ksKey+8*rdi+8*0]
+ addReg r11,rax
+ addReg r12,rbx
+ add r09,[FP_+ksKey+8*rdi+8*1]
+ addReg r13,rcx
+ addReg r14,rdx
+ add r10,[FP_+ksKey+8*rdi+8*2]
+ addReg r15,rsi
+ addReg r15,rdi ;inject the round number
+ endif
+ ;show the result of the key injection
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT
+endm ;R_512_EightRounds
+;
+;;;;;;;;;;;;;;;;;
+; instantiated code
+;
+Skein_512_Process_Block proc frame
+ Setup_Stack 512,ROUNDS_512/8
+ mov rbx,[rdi+TWEAK+ 8]
+ jmp short Skein_512_block_loop
+ align 16
+ ; main hash loop for Skein_512
+Skein_512_block_loop:
+ ; general register usage:
+ ; RAX..RDX = temps for key schedule pre-loads
+ ; R08..R15 = X0..X7
+ ; RSP, RBP = stack/frame pointers
+ ; RDI = round counter or context pointer
+ ; RSI = temp
+ ;
+ mov rax,[rdi+TWEAK+ 0]
+ add rax,[FP_+bitAdd] ;computed updated tweak value T0
+ mov rcx,rbx
+ xor rcx,rax ;rax/rbx/rcx = tweak schedule
+ mov [rdi+TWEAK+ 0],rax ;save updated tweak value ctx->h.T[0]
+ mov [FP_+ksTwk+ 0],rax
+ mov rdx,KW_PARITY
+ mov rsi,[FP_+blkPtr ] ;rsi --> input block
+ mov [FP_+ksTwk+ 8],rbx
+ mov [FP_+ksTwk+16],rcx
+
+ irp _Rn_,<0,1,2,3,4,5,6,7>
+ mov rX_512_&_Rn_,[rdi+X_VARS+8*(_Rn_)]
+ xor rdx,rX_512_&_Rn_ ;compute overall parity
+ mov [FP_+ksKey+8*(_Rn_)],rX_512_&_Rn_
+ endm ;load state into r08..r15, compute parity
+ mov [FP_+ksKey+8*(8)],rdx ;save key schedule parity
+
+ addReg rX_512_5,rax ;precompute key injection for tweak
+ addReg rX_512_6,rbx
+if _SKEIN_DEBUG
+ mov [rdi+TWEAK+ 8],rbx ;save updated tweak value ctx->h.T[1] for Skein_Debug_Block below
+endif
+ mov rax,[rsi+ 0] ;load input block
+ mov rbx,[rsi+ 8]
+ mov rcx,[rsi+16]
+ mov rdx,[rsi+24]
+ addReg r08,rax ;do initial key injection
+ addReg r09,rbx
+ mov [FP_+Wcopy+ 0],rax ;keep local copy for feedforward
+ mov [FP_+Wcopy+ 8],rbx
+ addReg r10,rcx
+ addReg r11,rdx
+ mov [FP_+Wcopy+16],rcx
+ mov [FP_+Wcopy+24],rdx
+
+ mov rax,[rsi+32]
+ mov rbx,[rsi+40]
+ mov rcx,[rsi+48]
+ mov rdx,[rsi+56]
+ addReg r12,rax
+ addReg r13,rbx
+ addReg r14,rcx
+ addReg r15,rdx
+ mov [FP_+Wcopy+32],rax
+ mov [FP_+Wcopy+40],rbx
+ mov [FP_+Wcopy+48],rcx
+ mov [FP_+Wcopy+56],rdx
+
+if _SKEIN_DEBUG
+ irp _Rn_,<0,1,2,3,4,5,6,7> ;save values on stack for debug output
+ mov [rsp+X_stk+8*(_Rn_)],rX_512_&_Rn_
+ endm
+
+ Skein_Debug_Block 512 ;debug dump
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL
+endif
+ add rsi, 8*WCNT ;skip the block
+ mov [FP_+blkPtr ],rsi ;update block pointer
+ ;
+ ;;;;;;;;;;;;;;;;;
+ ; now the key schedule is computed. Start the rounds
+ ;
+if SKEIN_ASM_UNROLL and 512
+_UNROLL_CNT = ROUNDS_512/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_512
+ if ((ROUNDS_512/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_512"
+ endif
+ xor rdi,rdi ;rdi = round counter
+Skein_512_round_loop:
+endif
+;
+_Rbase_ = 0
+rept _UNROLL_CNT*2
+ R_512_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+endm ;rept _UNROLL_CNT
+;
+if (SKEIN_ASM_UNROLL and 512) eq 0
+ cmp rdi,2*(ROUNDS_512/8)
+ jb Skein_512_round_loop
+ mov rdi,[FP_+ctxPtr ] ;restore rdi --> context
+endif
+ ; end of rounds
+ ;;;;;;;;;;;;;;;;;
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..7}
+ irp _Rn_,<0,1,2,3,4,5,6,7>
+ if (_Rn_ eq 0)
+ mov rbx,FIRST_MASK
+ endif
+ xor rX_512_&_Rn_,[FP_+Wcopy+8*(_Rn_)] ;feedforward XOR
+ mov [rdi+X_VARS+8*(_Rn_)],rX_512_&_Rn_ ;and store result
+ if (_Rn_ eq 6)
+ and rbx,[rdi+TWEAK+ 8]
+ endif
+ endm
+ Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+
+ ; go back for more blocks, if needed
+ dec qword ptr [FP_+blkCnt]
+ jnz Skein_512_block_loop
+ mov [rdi+TWEAK + 8],rbx
+
+ Reset_Stack Skein_512_Process_Block
+ ret
+;
+ if _SKEIN_DEBUG
+; call here with r08 = "round number"
+Skein_Debug_Round_512:
+ push rdx ;save two regs for BLK_BITS-specific parms
+ push rcx
+ mov rcx,[rsp+24] ;get back original r08 (pushed on stack in macro call)
+ mov [FP_+X_stk],rcx ;and save it in X_stk
+ irp _Rn_,<1,2,3,4,5,6,7> ;save rest of X[] state on stack so debug routines can access it
+ mov [FP_+X_stk+8*(_Rn_)],rX_512_&_Rn_
+ endm
+ mov rdx,[FP_+ctxPtr] ;ctx_hdr_ptr
+ mov rcx, 512 ;block size
+ jmp Skein_Debug_Round_Common
+ endif
+;
+Skein_512_Process_Block endp
+;
+ifdef SKEIN_CODE_SIZE
+ public Skein_512_Process_Block_CodeSize
+Skein_512_Process_Block_CodeSize proc
+ mov rax,_ProcBytes_
+ ret
+Skein_512_Process_Block_CodeSize endp
+;
+ public Skein_512_Unroll_Cnt
+Skein_512_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_512/8
+ mov rax,_UNROLL_CNT
+ else
+ xor rax,rax
+ endif
+ ret
+Skein_512_Unroll_Cnt endp
+endif
+;
+endif ; _USE_ASM_ and 512
+;
+;=================================== Skein1024 =============================================
+if _USE_ASM_ and 1024
+ public Skein1024_Process_Block
+;
+; void Skein1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+;;;;;;;;;;;;;;;;;
+; use details of permutation to make register assignments
+;
+r1K_x0 equ rdi
+r1K_x1 equ rsi
+r1K_x2 equ rbp
+r1K_x3 equ rax
+r1K_x4 equ rcx ;"shared" with X6, since X4/X6 alternate
+r1K_x5 equ rbx
+r1K_x6 equ rcx
+r1K_x7 equ rdx
+r1K_x8 equ r08
+r1K_x9 equ r09
+r1K_xA equ r10
+r1K_xB equ r11
+r1K_xC equ r12
+r1K_xD equ r13
+r1K_xE equ r14
+r1K_xF equ r15
+;
+rIdx equ r1K_x0 ;index register for looping versions
+rIdx_offs equ tmpStk_1024
+;
+R1024_Mix macro w0,w1,_RN0_,_Rn1_,op1
+_w0 = 0&w0&h ;handle the hex conversion
+_w1 = 0&w1&h
+_II_ = ((_RN0_)/4)+1 ;injection count
+ ;
+ addReg r1K_x&w0 , r1K_x&w1 ;perform the MIX
+ RotL64 r1K_x&w1 , 1024,%((_RN0_) and 7),_Rn1_
+ xor r1K_x&w1 , r1K_x&w0
+ if ((_RN0_) and 3) eq 3 ;time to do key injection?
+ if _SKEIN_DEBUG
+ mov [rsp+xDebug_1024+8*_w0],r1K_x&w0 ;save intermediate values for Debug_Round
+ mov [rsp+xDebug_1024+8*_w1],r1K_x&w1 ; (before inline key injection)
+ endif
+ if SKEIN_ASM_UNROLL and 1024 ;here to do fully unrolled key injection
+ add r1K_x&w0, [rsp+ksKey+ 8*((_II_+_w0) mod 17)]
+ add r1K_x&w1, [rsp+ksKey+ 8*((_II_+_w1) mod 17)]
+ if _w1 eq 13 ;tweak injection
+ add r1K_x&w1, [rsp+ksTwk+ 8*((_II_+0 ) mod 3)]
+ elseif _w0 eq 14
+ add r1K_x&w0, [rsp+ksTwk+ 8*((_II_+1 ) mod 3)]
+ elseif _w1 eq 15
+ add r1K_x&w1, _II_ ;(injection counter)
+ endif
+ else ;here to do looping key injection
+ if (_w0 eq 0)
+ mov [rsp+X_stk+8*_w0],r1K_x0 ;if so, store N0 so we can use reg as index
+ mov rIdx, [rsp+rIdx_offs] ;get the injection counter index into rIdx (N0)
+ else
+ add r1K_x&w0, [rsp+ksKey+8+8*rIdx+8*_w0] ;even key injection
+ endif
+ if _w1 eq 13 ;tweak injection
+ add r1K_x&w1, [rsp+ksTwk+8+8*rIdx+8*0 ]
+ elseif _w0 eq 14
+ add r1K_x&w0, [rsp+ksTwk+8+8*rIdx+8*1 ]
+ elseif _w1 eq 15
+ addReg r1K_x&w1, rIdx,,,1 ;(injection counter)
+ endif
+ add r1K_x&w1, [rsp+ksKey+8+8*rIdx+8*_w1] ;odd key injection
+ endif
+ endif
+ ; insert the op provided, if any
+ op1
+endm
+;;;;;;;;;;;;;;;;;
+; MACRO: one round for 1024-bit blocks
+;
+R1024_OneRound macro x0,x1,x2,x3,x4,x5,x6,x7,x8,x9,xA,xB,xC,xD,xE,xF,_Rn_
+ if (x0 ne 0) or ((x4 ne 4) and (x4 ne 6)) or (x4 ne (x6 xor 2))
+ .err "faulty register assignment!"
+ endif
+ R1024_Mix x0,x1,_Rn_,0
+ R1024_Mix x2,x3,_Rn_,1
+ R1024_Mix x4,x5,_Rn_,2, <mov [rsp+X_stk+8*0&x4&h],r1K_x4> ;save x4 on stack (x4/x6 alternate)
+ R1024_Mix x8,x9,_Rn_,4, <mov r1K_x6,[rsp+X_stk+8*0&x6&h]> ;load x6 from stack
+ R1024_Mix xA,xB,_Rn_,5
+ R1024_Mix xC,xD,_Rn_,6
+ R1024_Mix x6,x7,_Rn_,3
+ R1024_Mix xE,xF,_Rn_,7
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ endif
+endm ;R1024_OneRound
+;;;;;;;;;;;;;;;;;
+; MACRO: four rounds for 1024-bit blocks
+;
+R1024_FourRounds macro _RR_ ;RR = base round number (0 mod 4)
+ ; should be here with r1K_x4 set properly, x6 stored on stack
+ R1024_OneRound 0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F,%((_RR_)+0)
+ R1024_OneRound 0,9,2,D,6,B,4,F,A,7,C,3,E,5,8,1,%((_RR_)+1)
+ R1024_Oneround 0,7,2,5,4,3,6,1,C,F,E,D,8,B,A,9,%((_RR_)+2)
+ R1024_Oneround 0,F,2,B,6,D,4,9,E,1,8,5,A,3,C,7,%((_RR_)+3)
+ if (SKEIN_ASM_UNROLL and 1024) eq 0 ;here with r1K_x0 == rIdx, X0 on stack
+ ;rotate the key schedule on the stack
+ mov [rsp+X_stk+ 8* 8],r1K_x8;free up a reg
+ mov r1K_x8,[rsp+ksKey+8*rIdx+8* 0] ;get key
+ mov [rsp+ksKey+8*rIdx+8*17],r1K_x8 ;rotate it (must do key first or tweak clobbers it!)
+ mov r1K_x8,[rsp+ksTwk+8*rIdx+8* 0] ;get tweak
+ mov [rsp+ksTwk+8*rIdx+8* 3],r1K_x8 ;rotate it
+ mov r1K_x8,[rsp+X_stk+ 8* 8] ;get the reg back
+ inc rIdx ;bump the index
+ mov [rsp+rIdx_offs],rIdx ;save it
+ mov r1K_x0,[rsp+ksKey+8*rIdx] ;get the key schedule word for X0
+ add r1K_x0,[rsp+X_stk+8*0] ;perform the X0 key injection
+ endif
+ ;show the result of the key injection
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT
+endm ;R1024_FourRounds
+;
+;;;;;;;;;;;;;;;;
+; code
+;
+Skein1024_Process_Block proc frame
+;
+ Setup_Stack 1024,ROUNDS_1024/8,NO_FRAME,<WCNT>
+ mov r09,[rdi+TWEAK+ 8]
+ jmp short Skein1024_block_loop
+ align 16
+ ; main hash loop for Skein1024
+Skein1024_block_loop:
+ ; general register usage:
+ ; RSP = stack pointer
+ ; RAX..RDX,RSI,RDI= X1, X3..X7 (state words)
+ ; R08..R15 = X8..X15 (state words)
+ ; RBP = temp (used for X0 and X2)
+ ;
+ if (SKEIN_ASM_UNROLL and 1024) eq 0
+ xor rax,rax ;init loop index on the stack
+ mov [rsp+rIdx_offs],rax
+ endif
+ mov r08,[rdi+TWEAK+ 0]
+ add r08,[FP_+bitAdd] ;computed updated tweak value T0
+ mov r10,r09
+ xor r10,r08 ;rax/rbx/rcx = tweak schedule
+ mov [rdi+TWEAK+ 0],r08 ;save updated tweak value ctx->h.T[0]
+ mov [FP_+ksTwk+ 0],r08
+ mov [FP_+ksTwk+ 8],r09 ;keep values in r08,r09 for initial tweak injection below
+ mov [FP_+ksTwk+16],r10
+ if _SKEIN_DEBUG
+ mov [rdi+TWEAK+ 8],r09 ;save updated tweak value ctx->h.T[1] for Skein_Debug_Block
+ endif
+ mov rsi ,[FP_+blkPtr ] ;r1K_x2 --> input block
+ mov rax , KW_PARITY ;overall key schedule parity
+
+ ; logic here assumes the set {rdi,rsi,rbp,rax} = r1K_x{0,1,2,3}
+
+ irp _rN_,<0,1,2,3,4,6> ;process the "initial" words, using r14,r15 as temps
+ mov r14,[rdi+X_VARS+8*_rN_] ;get state word
+ mov r15,[rsi+ 8*_rN_] ;get msg word
+ xor rax,r14 ;update key schedule parity
+ mov [FP_+ksKey +8*_rN_],r14 ;save key schedule word on stack
+ mov [FP_+Wcopy +8*_rN_],r15 ;save local msg Wcopy
+ add r14,r15 ;do the initial key injection
+ mov [rsp+X_stk +8*_rN_],r14 ;save initial state var on stack
+ endm
+ ; now process the rest, using the "real" registers
+ ; (MUST do it in reverse order to inject tweaks r08/r09 first)
+ irp _rN_,<F,E,D,C,B,A,9,8,7,5>
+_rr_ = 0&_rN_&h
+ mov r1K_x&_rN_,[rdi+X_VARS+8*_rr_] ;get key schedule word from context
+ mov r1K_x4 ,[rsi+ 8*_rr_] ;get next input msg word
+ mov [rsp+ksKey +8*_rr_],r1K_x&_rN_ ;save key schedule on stack
+ xor rax , r1K_x&_rN_ ;accumulate key schedule parity
+ mov [FP_+Wcopy +8*_rr_],r1K_x4 ;save copy of msg word for feedforward
+ add r1K_x&_rN_, r1K_x4 ;do the initial key injection
+ if _rr_ eq 13 ;do the initial tweak injection
+ addReg r1K_x&_rN_,r08 ; (only in words 13/14)
+ elseif _rr_ eq 14
+ addReg r1K_x&_rN_,r09
+ endif
+ endm
+ mov [FP_+ksKey+8*WCNT],rax ;save key schedule parity
+if _SKEIN_DEBUG
+ Skein_Debug_Block 1024 ;debug dump
+endif
+ addReg rsi,8*WCNT ;bump the msg ptr
+ mov [FP_+blkPtr],rsi ;save bumped msg ptr
+ ; re-load words 0..4 [rbp,rsi,rdi,rax,rbx] from stack, enter the main loop
+ irp _rN_,<0,1,2,3,4> ;(no need to re-load x6)
+ mov r1K_x&_rN_,[rsp+X_stk+8*_rN_] ;re-load state and get ready to go!
+ endm
+if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL ;show state after initial key injection
+endif
+ ;
+ ;;;;;;;;;;;;;;;;;
+ ; now the key schedule is computed. Start the rounds
+ ;
+if SKEIN_ASM_UNROLL and 1024
+_UNROLL_CNT = ROUNDS_1024/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_1024
+ if ((ROUNDS_1024/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL1024"
+ endif
+Skein1024_round_loop:
+endif
+;
+_Rbase_ = 0
+rept _UNROLL_CNT*2 ;implement the rounds, 4 at a time
+ R1024_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+endm ;rept _UNROLL_CNT
+;
+if (SKEIN_ASM_UNROLL and 1024) eq 0
+ cmp qword ptr [rsp+tmpStk_1024],2*(ROUNDS_1024/8) ;see if we are done
+ jb Skein1024_round_loop
+endif
+ ; end of rounds
+ ;;;;;;;;;;;;;;;;;
+ ;
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..15}
+ mov [rsp+X_stk+8*7],r1K_x7 ;we need a register. x6 already on stack
+ mov r1K_x7,[rsp+ctxPtr]
+
+ irp _rN_,<0,1,2,3,4,5,8,9,A,B,C,D,E,F> ;do all but x6,x7
+ xor r1K_x&_rN_,[rsp +Wcopy +8*(0&_rN_&h)] ;feedforward XOR
+ mov [r1K_x7+X_VARS+8*(0&_rN_&h)],r1K_x&_rN_ ;save result into context
+ if (0&_rN_&h eq 9)
+ mov r09,FIRST_MASK
+ endif
+ if (0&_rN_&h eq 0eh)
+ and r09,[r1K_x7+TWEAK+ 8]
+ endif
+ endm
+ ;
+ mov rax,[rsp+X_stk +8*6] ;now process x6,x7
+ mov rbx,[rsp+X_stk +8*7]
+ xor rax,[rsp+Wcopy +8*6]
+ xor rbx,[rsp+Wcopy +8*7]
+ mov [r1K_x7+X_VARS+8*6],rax
+ dec qword ptr [rsp+blkCnt] ;set zero flag iff done
+ mov [r1K_x7+X_VARS+8*7],rbx
+
+ Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD,,<cmp qword ptr [rsp+blkCnt],0>
+ ; go back for more blocks, if needed
+ mov rdi,[rsp+ctxPtr] ;don't muck with the flags here!
+ lea rbp,[rsp+FRAME_OFFS]
+ jnz Skein1024_block_loop
+ mov [r1K_x7+TWEAK+ 8],r09
+ Reset_Stack Skein1024_Process_Block
+ ret
+;
+if _SKEIN_DEBUG
+; call here with r08 = "round number"
+Skein_Debug_Round_1024:
+_SP_OFFS_ = 8*2 ;stack "offset" here: r08, return addr
+ SP_ equ <rsp + _SP_OFFS_> ;useful shorthand below
+;
+ irp _wN_,<1,2,3,5,7,9,A,B,C,D,E,F> ;save rest of X[] state on stack so debug routines can access it
+ mov [SP_+X_stk+8*(0&_wN_&h)],r1K_x&_wN_
+ endm
+ ;figure out what to do with x0. On rounds R where R==0 mod 4, it's already on the stack
+ cmp r08,SKEIN_RND_SPECIAL ;special rounds always save
+ jae save_x0
+ test r08,3
+ jz save_x0_not
+save_x0:
+ mov [SP_+X_stk+8*0],r1K_x0
+save_x0_not:
+ ;figure out the x4/x6 swapping state and save the correct one!
+ cmp r08,SKEIN_RND_SPECIAL ;special rounds always do x4
+ jae save_x4
+ test r08,1 ;and even ones have r4 as well
+ jz save_x4
+ mov [SP_+X_stk+8*6],r1K_x6
+ jmp short debug_1024_go
+save_x4:
+ mov [SP_+X_stk+8*4],r1K_x4
+debug_1024_go:
+ ;now all is saved in Xstk[] except for X8
+ push rdx ;save two regs for BLK_BITS-specific parms
+ push rcx
+_SP_OFFS_ = _SP_OFFS_ + 16 ;adjust stack offset accordingly
+ ; now stack offset is 32 to X_stk
+ mov rcx,[SP_ - 8] ;get back original r08 (pushed on stack in macro call)
+ mov [SP_+X_stk+8*8],rcx ;and save it in its rightful place in X_stk[8]
+ mov rdx,[SP_+ctxPtr] ;ctx_hdr_ptr
+ mov rcx, 1024 ;block size
+ jmp Skein_Debug_Round_Common
+endif
+;
+Skein1024_Process_Block endp
+;
+ifdef SKEIN_CODE_SIZE
+ public Skein1024_Process_Block_CodeSize
+Skein1024_Process_Block_CodeSize proc
+ mov rax,_ProcBytes_
+ ret
+Skein1024_Process_Block_CodeSize endp
+;
+ public Skein1024_Unroll_Cnt
+Skein1024_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_1024/8
+ mov rax,_UNROLL_CNT
+ else
+ xor rax,rax
+ endif
+ ret
+Skein1024_Unroll_Cnt endp
+endif
+;
+endif ; _USE_ASM_ and 1024
+;
+if _SKEIN_DEBUG
+;----------------------------------------------------------------
+;local debug routine to set up for calls to:
+; void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X);
+;
+; here with r08 = round number
+; rdx = ctx_hdr_ptr
+; rcx = block size (256/512/1024)
+;
+Skein_Debug_Round_Common:
+_SP_OFFS_ = 32 ;current stack "offset": r08, retAddr, rcx, rdx
+ irp _rr_,<rax,rbx,rsi,rdi,rbp,r09,r10,r11,r12,r13,r14,r15> ;save the rest of the regs
+ push _rr_
+_SP_OFFS_ = _SP_OFFS_+8
+ endm
+ if (_SP_OFFS_ and 0Fh) ; make sure stack is still 16-byte aligned here
+ .err "Debug_Round_Common: stack alignment"
+ endif
+ ; compute r09 = ptr to the X[] array on the stack
+ lea r09,[SP_+X_stk] ;adjust for reg pushes, return address
+ cmp r08,SKEIN_RND_FEED_FWD ;special handling for feedforward "round"?
+ jnz _got_r09a
+ lea r09,[rdx+X_VARS]
+_got_r09a:
+ if _USE_ASM_ and 1024
+ ; special handling for 1024-bit case
+ ; (for rounds right before with key injection:
+ ; use xDebug_1024[] instead of X_stk[])
+ cmp r08,SKEIN_RND_SPECIAL
+ jae _got_r09b ;must be a normal round
+ or r08,r08
+ jz _got_r09b ;just before key injection
+ test r08,3
+ jne _got_r09b
+ cmp rcx,1024 ;only 1024-bit(s) for now
+ jne _got_r09b
+ lea r09,[SP_+xDebug_1024]
+_got_r09b:
+ endif
+ sub rsp, 8*4 ;make room for parms on stack
+ call Skein_Show_Round ;call external debug handler
+ add rsp, 8*4 ;discard parm space on the stack
+
+ irp _rr_,<r15,r14,r13,r12,r11,r10,r09,rbp,rdi,rsi,rbx,rax> ;restore regs
+ pop _rr_
+_SP_OFFS_ = _SP_OFFS_-8
+ endm
+ if _SP_OFFS_ - 32
+ .err "Debug_Round_Common: push/pop misalignment!"
+ endif
+ pop rcx
+ pop rdx
+ ret
+endif
+;----------------------------------------------------------------
+ end
Index: sys/contrib/skein/asm/skein_block_x64.s
===================================================================
--- /dev/null
+++ sys/contrib/skein/asm/skein_block_x64.s
@@ -0,0 +1,1328 @@
+#
+#----------------------------------------------------------------
+# 64-bit x86 assembler code (gnu as) for Skein block functions
+#
+# Author: Doug Whiting, Hifn/Exar
+#
+# This code is released to the public domain.
+#----------------------------------------------------------------
+#
+ .text
+ .altmacro
+ .psize 0,128 #list file has no page boundaries
+#
+_MASK_ALL_ = (256+512+1024) #all three algorithm bits
+_MAX_FRAME_ = 240
+#
+#################
+.ifndef SKEIN_USE_ASM
+_USE_ASM_ = _MASK_ALL_
+.else
+_USE_ASM_ = SKEIN_USE_ASM
+.endif
+#################
+.ifndef SKEIN_LOOP #configure loop unrolling
+_SKEIN_LOOP = 2 #default is fully unrolled for 256/512, twice for 1024
+.else
+_SKEIN_LOOP = SKEIN_LOOP
+ .irp _NN_,%_SKEIN_LOOP #only display loop unrolling if default changed on command line
+.print "+++ SKEIN_LOOP = \_NN_"
+ .endr
+.endif
+# the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256 = (_SKEIN_LOOP / 100) % 10
+SKEIN_UNROLL_512 = (_SKEIN_LOOP / 10) % 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP ) % 10
+#
+SKEIN_ASM_UNROLL = 0
+ .irp _NN_,256,512,1024
+ .if (SKEIN_UNROLL_\_NN_) == 0
+SKEIN_ASM_UNROLL = SKEIN_ASM_UNROLL + \_NN_
+ .endif
+ .endr
+#################
+#
+.ifndef SKEIN_ROUNDS
+ROUNDS_256 = 72
+ROUNDS_512 = 72
+ROUNDS_1024 = 80
+.else
+ROUNDS_256 = 8*((((SKEIN_ROUNDS / 100) + 5) % 10) + 5)
+ROUNDS_512 = 8*((((SKEIN_ROUNDS / 10) + 5) % 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS ) + 5) % 10) + 5)
+# only display rounds if default size is changed on command line
+.irp _NN_,256,512,1024
+ .if _USE_ASM_ && \_NN_
+ .irp _RR_,%(ROUNDS_\_NN_)
+ .if _NN_ < 1024
+.print "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+ .else
+.print "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+ .endif
+ .endr
+ .endif
+.endr
+.endif
+#################
+#
+.ifdef SKEIN_CODE_SIZE
+_SKEIN_CODE_SIZE = (1)
+.else
+.ifdef SKEIN_PERF #use code size if SKEIN_PERF is defined
+_SKEIN_CODE_SIZE = (1)
+.else
+_SKEIN_CODE_SIZE = (0)
+.endif
+.endif
+#
+#################
+#
+.ifndef SKEIN_DEBUG
+_SKEIN_DEBUG = 0
+.else
+_SKEIN_DEBUG = 1
+.endif
+#################
+#
+# define offsets of fields in hash context structure
+#
+HASH_BITS = 0 #bits of hash output
+BCNT = 8 + HASH_BITS #number of bytes in BUFFER[]
+TWEAK = 8 + BCNT #tweak values[0..1]
+X_VARS = 16 + TWEAK #chaining vars
+#
+#(Note: buffer[] in context structure is NOT needed here :-)
+#
+KW_PARITY = 0x1BD11BDAA9FC1A22 #overall parity of key schedule words
+FIRST_MASK = ~ (1 << 6)
+FIRST_MASK64= ~ (1 << 62)
+#
+# rotation constants for Skein
+#
+RC_256_0_0 = 14
+RC_256_0_1 = 16
+
+RC_256_1_0 = 52
+RC_256_1_1 = 57
+
+RC_256_2_0 = 23
+RC_256_2_1 = 40
+
+RC_256_3_0 = 5
+RC_256_3_1 = 37
+
+RC_256_4_0 = 25
+RC_256_4_1 = 33
+
+RC_256_5_0 = 46
+RC_256_5_1 = 12
+
+RC_256_6_0 = 58
+RC_256_6_1 = 22
+
+RC_256_7_0 = 32
+RC_256_7_1 = 32
+
+RC_512_0_0 = 46
+RC_512_0_1 = 36
+RC_512_0_2 = 19
+RC_512_0_3 = 37
+
+RC_512_1_0 = 33
+RC_512_1_1 = 27
+RC_512_1_2 = 14
+RC_512_1_3 = 42
+
+RC_512_2_0 = 17
+RC_512_2_1 = 49
+RC_512_2_2 = 36
+RC_512_2_3 = 39
+
+RC_512_3_0 = 44
+RC_512_3_1 = 9
+RC_512_3_2 = 54
+RC_512_3_3 = 56
+
+RC_512_4_0 = 39
+RC_512_4_1 = 30
+RC_512_4_2 = 34
+RC_512_4_3 = 24
+
+RC_512_5_0 = 13
+RC_512_5_1 = 50
+RC_512_5_2 = 10
+RC_512_5_3 = 17
+
+RC_512_6_0 = 25
+RC_512_6_1 = 29
+RC_512_6_2 = 39
+RC_512_6_3 = 43
+
+RC_512_7_0 = 8
+RC_512_7_1 = 35
+RC_512_7_2 = 56
+RC_512_7_3 = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 = 8
+RC_1024_0_3 = 47
+RC_1024_0_4 = 8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 = 4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 = 5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 = 9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 = 4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 = 9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+#
+# Input: reg
+# Output: <reg> <<< RC_BlkSize_roundNum_mixNum, BlkSize=256/512/1024
+#
+.macro RotL64 reg,BLK_SIZE,ROUND_NUM,MIX_NUM
+_RCNT_ = RC_\BLK_SIZE&_\ROUND_NUM&_\MIX_NUM
+ .if _RCNT_ #is there anything to do?
+ rolq $_RCNT_,%\reg
+ .endif
+.endm
+#
+#----------------------------------------------------------------
+#
+# MACROS: define local vars and configure stack
+#
+#----------------------------------------------------------------
+# declare allocated space on the stack
+.macro StackVar localName,localSize
+\localName = _STK_OFFS_
+_STK_OFFS_ = _STK_OFFS_+(\localSize)
+.endm #StackVar
+#
+#----------------------------------------------------------------
+#
+# MACRO: Configure stack frame, allocate local vars
+#
+.macro Setup_Stack BLK_BITS,KS_CNT,debugCnt
+ WCNT = (\BLK_BITS)/64
+#
+_PushCnt_ = 0 #save nonvolatile regs on stack
+ .irp _reg_,rbp,rbx,r12,r13,r14,r15
+ pushq %\_reg_
+_PushCnt_ = _PushCnt_ + 1 #track count to keep alignment
+ .endr
+#
+_STK_OFFS_ = 0 #starting offset from rsp
+ #---- local variables #<-- rsp
+ StackVar X_stk ,8*(WCNT) #local context vars
+ StackVar ksTwk ,8*3 #key schedule: tweak words
+ StackVar ksKey ,8*(WCNT)+8 #key schedule: key words
+ .if (SKEIN_ASM_UNROLL && (\BLK_BITS)) == 0
+ StackVar ksRot ,16*(\KS_CNT) #leave space for "rotation" to happen
+ .endif
+ StackVar Wcopy ,8*(WCNT) #copy of input block
+ .if _SKEIN_DEBUG
+ .if \debugCnt + 0 #temp location for debug X[] info
+ StackVar xDebug_\BLK_BITS ,8*(\debugCnt)
+ .endif
+ .endif
+ .if ((8*_PushCnt_ + _STK_OFFS_) % 8) == 0
+ StackVar align16,8 #keep 16-byte aligned (adjust for retAddr?)
+tmpStk_\BLK_BITS = align16 #use this
+ .endif
+ #---- saved caller parameters (from regs rdi, rsi, rdx, rcx)
+ StackVar ctxPtr ,8 #context ptr
+ StackVar blkPtr ,8 #pointer to block data
+ StackVar blkCnt ,8 #number of full blocks to process
+ StackVar bitAdd ,8 #bit count to add to tweak
+LOCAL_SIZE = _STK_OFFS_ #size of "local" vars
+ #----
+ StackVar savRegs,8*_PushCnt_ #saved registers
+ StackVar retAddr,8 #return address
+ #---- caller's stack frame (aligned mod 16)
+#
+# set up the stack frame pointer (rbp)
+#
+FRAME_OFFS = ksTwk + 128 #allow short (negative) offset to ksTwk, kwKey
+ .if FRAME_OFFS > _STK_OFFS_ #keep rbp in the "locals" range
+FRAME_OFFS = _STK_OFFS_
+ .endif
+F_O = -FRAME_OFFS
+#
+ #put some useful defines in the .lst file (for grep)
+__STK_LCL_SIZE_\BLK_BITS = LOCAL_SIZE
+__STK_TOT_SIZE_\BLK_BITS = _STK_OFFS_
+__STK_FRM_OFFS_\BLK_BITS = FRAME_OFFS
+#
+# Notes on stack frame setup:
+# * the most frequently used variable is X_stk[], based at [rsp+0]
+# * the next most used is the key schedule arrays, ksKey and ksTwk
+# so rbp is "centered" there, allowing short offsets to the key
+# schedule even in 1024-bit Skein case
+# * the Wcopy variables are infrequently accessed, but they have long
+# offsets from both rsp and rbp only in the 1024-bit case.
+# * all other local vars and calling parameters can be accessed
+# with short offsets, except in the 1024-bit case
+#
+ subq $LOCAL_SIZE,%rsp #make room for the locals
+ leaq FRAME_OFFS(%rsp),%rbp #maximize use of short offsets
+ movq %rdi, ctxPtr+F_O(%rbp) #save caller's parameters on the stack
+ movq %rsi, blkPtr+F_O(%rbp)
+ movq %rdx, blkCnt+F_O(%rbp)
+ movq %rcx, bitAdd+F_O(%rbp)
+#
+.endm #Setup_Stack
+#
+#----------------------------------------------------------------
+#
+.macro Reset_Stack
+ addq $LOCAL_SIZE,%rsp #get rid of locals (wipe??)
+ .irp _reg_,r15,r14,r13,r12,rbx,rbp
+ popq %\_reg_ #restore caller's regs
+_PushCnt_ = _PushCnt_ - 1
+ .endr
+ .if _PushCnt_
+ .error "Mismatched push/pops?"
+ .endif
+.endm # Reset_Stack
+#
+#----------------------------------------------------------------
+# macros to help debug internals
+#
+.if _SKEIN_DEBUG
+ .extern Skein_Show_Block #calls to C routines
+ .extern Skein_Show_Round
+#
+SKEIN_RND_SPECIAL = 1000
+SKEIN_RND_KEY_INITIAL = SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT = SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD = SKEIN_RND_SPECIAL+2
+#
+.macro Skein_Debug_Block BLK_BITS
+#
+#void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+# const u08b_t *blkPtr, const u64b_t *wPtr,
+# const u64b_t *ksPtr,const u64b_t *tsPtr)
+#
+_NN_ = 0
+ .irp _reg_,rax,rcx,rdx,rsi,rdi,r8,r9,r10,r11
+ pushq %\_reg_ #save all volatile regs on tack before the call
+_NN_ = _NN_ + 1
+ .endr
+ # get and push call parameters
+ movq $\BLK_BITS ,%rdi #bits
+ movq ctxPtr+F_O(%rbp),%rsi #h (pointer)
+ leaq X_VARS (%rsi),%rdx #X (pointer)
+ movq blkPtr+F_O(%rbp),%rcx #blkPtr
+ leaq Wcopy +F_O(%rbp),%r8 #wPtr
+ leaq ksKey +F_O(%rbp),%r9 #key pointer
+ leaq ksTwk +F_O(%rbp),%rax #tweak pointer
+ pushq %rax # (pass on the stack)
+ call Skein_Show_Block #call external debug handler
+ addq $8*1,%rsp #discard parameters on stack
+ .if (_NN_ % 2 ) == 0 #check stack alignment
+ .error "Stack misalignment problem in Skein_Debug_Block_\_BLK_BITS"
+ .endif
+ .irp _reg_,r11,r10,r9,r8,rdi,rsi,rdx,rcx,rax
+ popq %\_reg_ #restore regs
+_NN_ = _NN_ - 1
+ .endr
+ .if _NN_
+ .error "Push/pop mismatch problem in Skein_Debug_Block_\_BLK_BITS"
+ .endif
+.endm # Skein_Debug_Block
+#
+# the macro to "call" to debug a round
+#
+.macro Skein_Debug_Round BLK_BITS,R,RDI_OFFS,afterOp
+ # call the appropriate (local) debug "function"
+ pushq %rdx #save rdx, so we can use it for round "number"
+ .if (SKEIN_ASM_UNROLL && \BLK_BITS) || (\R >= SKEIN_RND_SPECIAL)
+ movq $\R,%rdx
+ .else #compute round number using edi
+_rOffs_ = \RDI_OFFS + 0
+ .if \BLK_BITS == 1024
+ movq rIdx_offs+8(%rsp),%rdx #get rIdx off the stack (adjust for pushq rdx above)
+ leaq 1+(((\R)-1) && 3)+_rOffs_(,%rdx,4),%rdx
+ .else
+ leaq 1+(((\R)-1) && 3)+_rOffs_(,%rdi,4),%rdx
+ .endif
+ .endif
+ call Skein_Debug_Round_\BLK_BITS
+ popq %rdx #restore origianl rdx value
+#
+ afterOp
+.endm # Skein_Debug_Round
+.else #------- _SKEIN_DEBUG (dummy macros if debug not enabled)
+.macro Skein_Debug_Block BLK_BITS
+.endm
+#
+.macro Skein_Debug_Round BLK_BITS,R,RDI_OFFS,afterOp
+.endm
+#
+.endif # _SKEIN_DEBUG
+#
+#----------------------------------------------------------------
+#
+.macro addReg dstReg,srcReg_A,srcReg_B,useAddOp,immOffs
+ .if \immOffs + 0
+ leaq \immOffs(%\srcReg_A\srcReg_B,%\dstReg),%\dstReg
+ .elseif ((\useAddOp + 0) == 0)
+ .ifndef ASM_NO_LEA #lea seems to be faster on Core 2 Duo CPUs!
+ leaq (%\srcReg_A\srcReg_B,%\dstReg),%\dstReg
+ .else
+ addq %\srcReg_A\srcReg_B,%\dstReg
+ .endif
+ .else
+ addq %\srcReg_A\srcReg_B,%\dstReg
+ .endif
+.endm
+
+# keep Intel-style ordering here, to match addReg
+.macro xorReg dstReg,srcReg_A,srcReg_B
+ xorq %\srcReg_A\srcReg_B,%\dstReg
+.endm
+#
+#----------------------------------------------------------------
+#
+.macro C_label lName
+ \lName: #use both "genders" to work across linkage conventions
+_\lName:
+ .global \lName
+ .global _\lName
+.endm
+#
+#=================================== Skein_256 =============================================
+#
+.if _USE_ASM_ & 256
+#
+# void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+#
+# code
+#
+C_label Skein_256_Process_Block
+ Setup_Stack 256,((ROUNDS_256/8)+1)
+ movq TWEAK+8(%rdi),%r14
+ jmp Skein_256_block_loop
+ .p2align 4
+ # main hash loop for Skein_256
+Skein_256_block_loop:
+ #
+ # general register usage:
+ # RAX..RDX = X0..X3
+ # R08..R12 = ks[0..4]
+ # R13..R15 = ts[0..2]
+ # RSP, RBP = stack/frame pointers
+ # RDI = round counter or context pointer
+ # RSI = temp
+ #
+ movq TWEAK+0(%rdi) ,%r13
+ addq bitAdd+F_O(%rbp) ,%r13 #computed updated tweak value T0
+ movq %r14 ,%r15
+ xorq %r13 ,%r15 #now %r13.%r15 is set as the tweak
+
+ movq $KW_PARITY ,%r12
+ movq X_VARS+ 0(%rdi),%r8
+ movq X_VARS+ 8(%rdi),%r9
+ movq X_VARS+16(%rdi),%r10
+ movq X_VARS+24(%rdi),%r11
+ movq %r13,TWEAK+0(%rdi) #save updated tweak value ctx->h.T[0]
+ xorq %r8 ,%r12 #start accumulating overall parity
+
+ movq blkPtr +F_O(%rbp) ,%rsi #esi --> input block
+ xorq %r9 ,%r12
+ movq 0(%rsi) ,%rax #get X[0..3]
+ xorq %r10 ,%r12
+ movq 8(%rsi) ,%rbx
+ xorq %r11 ,%r12
+ movq 16(%rsi) ,%rcx
+ movq 24(%rsi) ,%rdx
+
+ movq %rax,Wcopy+ 0+F_O(%rbp) #save copy of input block
+ movq %rbx,Wcopy+ 8+F_O(%rbp)
+ movq %rcx,Wcopy+16+F_O(%rbp)
+ movq %rdx,Wcopy+24+F_O(%rbp)
+
+ addq %r8 ,%rax #initial key injection
+ addq %r9 ,%rbx
+ addq %r10,%rcx
+ addq %r11,%rdx
+ addq %r13,%rbx
+ addq %r14,%rcx
+
+.if _SKEIN_DEBUG
+ movq %r14,TWEAK+ 8(%rdi) #save updated tweak T[1] (start bit cleared?)
+ movq %r8 ,ksKey+ 0+F_O(%rbp) #save key schedule on stack for Skein_Debug_Block
+ movq %r9 ,ksKey+ 8+F_O(%rbp)
+ movq %r10,ksKey+16+F_O(%rbp)
+ movq %r11,ksKey+24+F_O(%rbp)
+ movq %r12,ksKey+32+F_O(%rbp)
+
+ movq %r13,ksTwk+ 0+F_O(%rbp)
+ movq %r14,ksTwk+ 8+F_O(%rbp)
+ movq %r15,ksTwk+16+F_O(%rbp)
+
+ movq %rax,X_stk + 0(%rsp) #save X[] on stack for Skein_Debug_Block
+ movq %rbx,X_stk + 8(%rsp)
+ movq %rcx,X_stk +16(%rsp)
+ movq %rdx,X_stk +24(%rsp)
+
+ Skein_Debug_Block 256 #debug dump
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL
+.endif
+#
+.if ((SKEIN_ASM_UNROLL & 256) == 0)
+ movq %r8 ,ksKey+40+F_O(%rbp) #save key schedule on stack for looping code
+ movq %r9 ,ksKey+ 8+F_O(%rbp)
+ movq %r10,ksKey+16+F_O(%rbp)
+ movq %r11,ksKey+24+F_O(%rbp)
+ movq %r12,ksKey+32+F_O(%rbp)
+
+ movq %r13,ksTwk+24+F_O(%rbp)
+ movq %r14,ksTwk+ 8+F_O(%rbp)
+ movq %r15,ksTwk+16+F_O(%rbp)
+.endif
+ addq $WCNT*8,%rsi #skip the block
+ movq %rsi,blkPtr +F_O(%rbp) #update block pointer
+ #
+ # now the key schedule is computed. Start the rounds
+ #
+.if SKEIN_ASM_UNROLL & 256
+_UNROLL_CNT = ROUNDS_256/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_256
+ .if ((ROUNDS_256/8) % _UNROLL_CNT)
+ .error "Invalid SKEIN_UNROLL_256"
+ .endif
+ xorq %rdi,%rdi #rdi = iteration count
+Skein_256_round_loop:
+.endif
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+ # all X and ks vars in regs # (ops to "rotate" ks vars, via mem, if not unrolled)
+ # round 4*_RBase_ + 0
+ addReg rax, rbx
+ RotL64 rbx, 256,%((4*_Rbase_+0) % 8),0
+ addReg rcx, rdx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*1+F_O(%rbp,%rdi,8),%r8
+ .endif
+ xorReg rbx, rax
+ RotL64 rdx, 256,%((4*_Rbase_+0) % 8),1
+ xorReg rdx, rcx
+ .if SKEIN_ASM_UNROLL & 256
+ .irp _r0_,%( 8+(_Rbase_+3) % 5)
+ .irp _r1_,%(13+(_Rbase_+2) % 3)
+ leaq (%r\_r0_,%r\_r1_),%rdi #precompute key injection value for %rcx
+ .endr
+ .endr
+ .endif
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksTwk+8*1+F_O(%rbp,%rdi,8),%r13
+ .endif
+ Skein_Debug_Round 256,%(4*_Rbase_+1)
+
+ # round 4*_Rbase_ + 1
+ addReg rax, rdx
+ RotL64 rdx, 256,%((4*_Rbase_+1) % 8),0
+ xorReg rdx, rax
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*2+F_O(%rbp,%rdi,8),%r9
+ .endif
+ addReg rcx, rbx
+ RotL64 rbx, 256,%((4*_Rbase_+1) % 8),1
+ xorReg rbx, rcx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*4+F_O(%rbp,%rdi,8),%r11
+ .endif
+ Skein_Debug_Round 256,%(4*_Rbase_+2)
+ .if SKEIN_ASM_UNROLL & 256
+ .irp _r0_,%( 8+(_Rbase_+2) % 5)
+ .irp _r1_,%(13+(_Rbase_+1) % 3)
+ leaq (%r\_r0_,%r\_r1_),%rsi #precompute key injection value for %rbx
+ .endr
+ .endr
+ .endif
+ # round 4*_Rbase_ + 2
+ addReg rax, rbx
+ RotL64 rbx, 256,%((4*_Rbase_+2) % 8),0
+ addReg rcx, rdx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*3+F_O(%rbp,%rdi,8),%r10
+ .endif
+ xorReg rbx, rax
+ RotL64 rdx, 256,%((4*_Rbase_+2) % 8),1
+ xorReg rdx, rcx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq %r8,ksKey+8*6+F_O(%rbp,%rdi,8) #"rotate" the key
+ leaq 1(%r11,%rdi),%r11 #precompute key + tweak
+ .endif
+ Skein_Debug_Round 256,%(4*_Rbase_+3)
+ # round 4*_Rbase_ + 3
+ addReg rax, rdx
+ RotL64 rdx, 256,%((4*_Rbase_+3) % 8),0
+ addReg rcx, rbx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ addq ksTwk+8*2+F_O(%rbp,%rdi,8),%r10 #precompute key + tweak
+ movq %r13,ksTwk+8*4+F_O(%rbp,%rdi,8) #"rotate" the tweak
+ .endif
+ xorReg rdx, rax
+ RotL64 rbx, 256,%((4*_Rbase_+3) % 8),1
+ xorReg rbx, rcx
+ Skein_Debug_Round 256,%(4*_Rbase_+4)
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ addReg r9 ,r13 #precompute key+tweak
+ .endif
+ #inject key schedule words
+_Rbase_ = _Rbase_+1
+ .if SKEIN_ASM_UNROLL & 256
+ addReg rax,r,%(8+((_Rbase_+0) % 5))
+ addReg rbx,rsi
+ addReg rcx,rdi
+ addReg rdx,r,%(8+((_Rbase_+3) % 5)),,_Rbase_
+ .else
+ incq %rdi
+ addReg rax,r8
+ addReg rcx,r10
+ addReg rbx,r9
+ addReg rdx,r11
+ .endif
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 256) == 0
+ cmpq $2*(ROUNDS_256/8),%rdi
+ jb Skein_256_round_loop
+.endif # (SKEIN_ASM_UNROLL & 256) == 0
+ movq ctxPtr +F_O(%rbp),%rdi #restore rdi --> context
+
+ #----------------------------
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..3}
+ movq $FIRST_MASK64 ,%r14
+ xorq Wcopy + 0+F_O (%rbp),%rax
+ xorq Wcopy + 8+F_O (%rbp),%rbx
+ xorq Wcopy +16+F_O (%rbp),%rcx
+ xorq Wcopy +24+F_O (%rbp),%rdx
+ andq TWEAK + 8 (%rdi),%r14
+ movq %rax,X_VARS+ 0(%rdi) #store final result
+ movq %rbx,X_VARS+ 8(%rdi)
+ movq %rcx,X_VARS+16(%rdi)
+ movq %rdx,X_VARS+24(%rdi)
+
+ Skein_Debug_Round 256,SKEIN_RND_FEED_FWD
+
+ # go back for more blocks, if needed
+ decq blkCnt+F_O(%rbp)
+ jnz Skein_256_block_loop
+ movq %r14,TWEAK + 8(%rdi)
+ Reset_Stack
+ ret
+Skein_256_Process_Block_End:
+
+ .if _SKEIN_DEBUG
+Skein_Debug_Round_256: #here with rdx == round "number" from macro
+ pushq %rsi #save two regs for BLK_BITS-specific parms
+ pushq %rdi
+ movq 24(%rsp),%rdi #get back original rdx (pushed on stack in macro call) to rdi
+ movq %rax,X_stk+ 0+F_O(%rbp) #save X[] state on stack so debug routines can access it
+ movq %rbx,X_stk+ 8+F_O(%rbp) #(use FP_ since rsp has changed!)
+ movq %rcx,X_stk+16+F_O(%rbp)
+ movq %rdi,X_stk+24+F_O(%rbp)
+
+ movq ctxPtr+F_O(%rbp),%rsi #ctx_hdr_ptr
+ movq $256,%rdi #now <rdi,rsi,rdx> are set for the call
+ jmp Skein_Debug_Round_Common
+ .endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein_256_Process_Block_CodeSize
+ movq $(Skein_256_Process_Block_End-Skein_256_Process_Block),%rax
+ ret
+#
+C_label Skein_256_Unroll_Cnt
+ .if _UNROLL_CNT <> ROUNDS_256/8
+ movq $_UNROLL_CNT,%rax
+ .else
+ xorq %rax,%rax
+ .endif
+ ret
+.endif
+#
+.endif #_USE_ASM_ & 256
+#
+#=================================== Skein_512 =============================================
+#
+.if _USE_ASM_ & 512
+#
+# void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)
+#
+# X[i] == %r[8+i] #register assignments for X[] values during rounds (i=0..7)
+#
+#################
+# MACRO: one round for 512-bit blocks
+#
+.macro R_512_OneRound rn0,rn1,rn2,rn3,rn4,rn5,rn6,rn7,_Rn_,op1,op2,op3,op4
+#
+ addReg r\rn0, r\rn1
+ RotL64 r\rn1, 512,%((_Rn_) % 8),0
+ xorReg r\rn1, r\rn0
+ op1
+ addReg r\rn2, r\rn3
+ RotL64 r\rn3, 512,%((_Rn_) % 8),1
+ xorReg r\rn3, r\rn2
+ op2
+ addReg r\rn4, r\rn5
+ RotL64 r\rn5, 512,%((_Rn_) % 8),2
+ xorReg r\rn5, r\rn4
+ op3
+ addReg r\rn6, r\rn7
+ RotL64 r\rn7, 512,%((_Rn_) % 8),3
+ xorReg r\rn7, r\rn6
+ op4
+ Skein_Debug_Round 512,%(_Rn_+1),-4
+#
+.endm #R_512_OneRound
+#
+#################
+# MACRO: eight rounds for 512-bit blocks
+#
+.macro R_512_FourRounds _RR_ #RR = base round number (0 % 8)
+ .if (SKEIN_ASM_UNROLL && 512)
+ # here for fully unrolled case.
+ _II_ = ((_RR_)/4) + 1 #key injection counter
+ R_512_OneRound 8, 9,10,11,12,13,14,15,%((_RR_)+0),<movq ksKey+8*(((_II_)+3) % 9)+F_O(%rbp),%rax>,,<movq ksKey+8*(((_II_)+4) % 9)+F_O(%rbp),%rbx>
+ R_512_OneRound 10, 9,12,15,14,13, 8,11,%((_RR_)+1),<movq ksKey+8*(((_II_)+5) % 9)+F_O(%rbp),%rcx>,,<movq ksKey+8*(((_II_)+6) % 9)+F_O(%rbp),%rdx>
+ R_512_OneRound 12, 9,14,11, 8,13,10,15,%((_RR_)+2),<movq ksKey+8*(((_II_)+7) % 9)+F_O(%rbp),%rsi>,,<addq ksTwk+8*(((_II_)+0) % 3)+F_O(%rbp),%rcx>
+ R_512_OneRound 14, 9, 8,15,10,13,12,11,%((_RR_)+3),<addq ksTwk+8*(((_II_)+1) % 3)+F_O(%rbp),%rdx>,
+ # inject the key schedule
+ addq ksKey+8*(((_II_)+0)%9)+F_O(%rbp),%r8
+ addReg r11, rax
+ addq ksKey+8*(((_II_)+1)%9)+F_O(%rbp),%r9
+ addReg r12, rbx
+ addq ksKey+8*(((_II_)+2)%9)+F_O(%rbp),%r10
+ addReg r13, rcx
+ addReg r14, rdx
+ addReg r15, rsi,,,(_II_)
+ .else
+ # here for looping case #"rotate" key/tweak schedule (move up on stack)
+ incq %rdi #bump key injection counter
+ R_512_OneRound 8, 9,10,11,12,13,14,15,%((_RR_)+0),<movq ksKey+8*6+F_O(%rbp,%rdi,8),%rdx>,<movq ksTwk-8*1+F_O(%rbp,%rdi,8),%rax>,<movq ksKey-8*1+F_O(%rbp,%rdi,8),%rsi>
+ R_512_OneRound 10, 9,12,15,14,13, 8,11,%((_RR_)+1),<movq ksKey+8*5+F_O(%rbp,%rdi,8),%rcx>,<movq %rax,ksTwk+8*2+F_O(%rbp,%rdi,8) >,<movq %rsi,ksKey+8*8+F_O(%rbp,%rdi,8)>
+ R_512_OneRound 12, 9,14,11, 8,13,10,15,%((_RR_)+2),<movq ksKey+8*4+F_O(%rbp,%rdi,8),%rbx>,<addq ksTwk+8*1+F_O(%rbp,%rdi,8),%rdx>,<movq ksKey+8*7+F_O(%rbp,%rdi,8),%rsi>
+ R_512_OneRound 14, 9, 8,15,10,13,12,11,%((_RR_)+3),<movq ksKey+8*3+F_O(%rbp,%rdi,8),%rax>,<addq ksTwk+8*0+F_O(%rbp,%rdi,8),%rcx>
+ # inject the key schedule
+ addq ksKey+8*0+F_O(%rbp,%rdi,8),%r8
+ addReg r11, rax
+ addReg r12, rbx
+ addq ksKey+8*1+F_O(%rbp,%rdi,8),%r9
+ addReg r13, rcx
+ addReg r14, rdx
+ addq ksKey+8*2+F_O(%rbp,%rdi,8),%r10
+ addReg r15, rsi
+ addReg r15, rdi #inject the round number
+ .endif
+
+ #show the result of the key injection
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT
+.endm #R_512_EightRounds
+#
+#################
+# instantiated code
+#
+C_label Skein_512_Process_Block
+ Setup_Stack 512,ROUNDS_512/8
+ movq TWEAK+ 8(%rdi),%rbx
+ jmp Skein_512_block_loop
+ .p2align 4
+ # main hash loop for Skein_512
+Skein_512_block_loop:
+ # general register usage:
+ # RAX..RDX = temps for key schedule pre-loads
+ # R8 ..R15 = X0..X7
+ # RSP, RBP = stack/frame pointers
+ # RDI = round counter or context pointer
+ # RSI = temp
+ #
+ movq TWEAK + 0(%rdi),%rax
+ addq bitAdd+F_O(%rbp),%rax #computed updated tweak value T0
+ movq %rbx,%rcx
+ xorq %rax,%rcx #%rax/%rbx/%rcx = tweak schedule
+ movq %rax,TWEAK+ 0 (%rdi) #save updated tweak value ctx->h.T[0]
+ movq %rax,ksTwk+ 0+F_O(%rbp)
+ movq $KW_PARITY,%rdx
+ movq blkPtr +F_O(%rbp),%rsi #%rsi --> input block
+ movq %rbx,ksTwk+ 8+F_O(%rbp)
+ movq %rcx,ksTwk+16+F_O(%rbp)
+ .irp _Rn_,8,9,10,11,12,13,14,15
+ movq X_VARS+8*(_Rn_-8)(%rdi),%r\_Rn_
+ xorq %r\_Rn_,%rdx #compute overall parity
+ movq %r\_Rn_,ksKey+8*(_Rn_-8)+F_O(%rbp)
+ .endr #load state into %r8 ..%r15, compute parity
+ movq %rdx,ksKey+8*(8)+F_O(%rbp)#save key schedule parity
+
+ addReg r13,rax #precompute key injection for tweak
+ addReg r14, rbx
+.if _SKEIN_DEBUG
+ movq %rbx,TWEAK+ 8(%rdi) #save updated tweak value ctx->h.T[1] for Skein_Debug_Block below
+.endif
+ movq 0(%rsi),%rax #load input block
+ movq 8(%rsi),%rbx
+ movq 16(%rsi),%rcx
+ movq 24(%rsi),%rdx
+ addReg r8 , rax #do initial key injection
+ addReg r9 , rbx
+ movq %rax,Wcopy+ 0+F_O(%rbp) #keep local copy for feedforward
+ movq %rbx,Wcopy+ 8+F_O(%rbp)
+ addReg r10, rcx
+ addReg r11, rdx
+ movq %rcx,Wcopy+16+F_O(%rbp)
+ movq %rdx,Wcopy+24+F_O(%rbp)
+
+ movq 32(%rsi),%rax
+ movq 40(%rsi),%rbx
+ movq 48(%rsi),%rcx
+ movq 56(%rsi),%rdx
+ addReg r12, rax
+ addReg r13, rbx
+ addReg r14, rcx
+ addReg r15, rdx
+ movq %rax,Wcopy+32+F_O(%rbp)
+ movq %rbx,Wcopy+40+F_O(%rbp)
+ movq %rcx,Wcopy+48+F_O(%rbp)
+ movq %rdx,Wcopy+56+F_O(%rbp)
+
+.if _SKEIN_DEBUG
+ .irp _Rn_,8,9,10,11,12,13,14,15 #save values on stack for debug output
+ movq %r\_Rn_,X_stk+8*(_Rn_-8)(%rsp)
+ .endr
+
+ Skein_Debug_Block 512 #debug dump
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL
+.endif
+ addq $8*WCNT,%rsi #skip the block
+ movq %rsi,blkPtr+F_O(%rbp) #update block pointer
+ #
+ #################
+ # now the key schedule is computed. Start the rounds
+ #
+.if SKEIN_ASM_UNROLL & 512
+_UNROLL_CNT = ROUNDS_512/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_512
+ .if ((ROUNDS_512/8) % _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_512"
+ .endif
+ xorq %rdi,%rdi #rdi = round counter
+Skein_512_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+ R_512_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 512) == 0
+ cmpq $2*(ROUNDS_512/8),%rdi
+ jb Skein_512_round_loop
+ movq ctxPtr +F_O(%rbp),%rdi #restore rdi --> context
+.endif
+ # end of rounds
+ #################
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..7}
+ .irp _Rn_,8,9,10,11,12,13,14,15
+ .if (_Rn_ == 8)
+ movq $FIRST_MASK64,%rbx
+ .endif
+ xorq Wcopy+8*(_Rn_-8)+F_O(%rbp),%r\_Rn_ #feedforward XOR
+ movq %r\_Rn_,X_VARS+8*(_Rn_-8)(%rdi) #and store result
+ .if (_Rn_ == 14)
+ andq TWEAK+ 8(%rdi),%rbx
+ .endif
+ .endr
+ Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+
+ # go back for more blocks, if needed
+ decq blkCnt+F_O(%rbp)
+ jnz Skein_512_block_loop
+ movq %rbx,TWEAK + 8(%rdi)
+
+ Reset_Stack
+ ret
+Skein_512_Process_Block_End:
+#
+ .if _SKEIN_DEBUG
+# call here with rdx = "round number"
+Skein_Debug_Round_512:
+ pushq %rsi #save two regs for BLK_BITS-specific parms
+ pushq %rdi
+ .irp _Rn_,8,9,10,11,12,13,14,15 #save X[] state on stack so debug routines can access it
+ movq %r\_Rn_,X_stk+8*(_Rn_-8)+F_O(%rbp)
+ .endr
+ movq ctxPtr+F_O(%rbp),%rsi #ctx_hdr_ptr
+ movq $512,%rdi #now <rdi,rsi,rdx> are set for the call
+ jmp Skein_Debug_Round_Common
+ .endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein_512_Process_Block_CodeSize
+ movq $(Skein_512_Process_Block_End-Skein_512_Process_Block),%rax
+ ret
+#
+C_label Skein_512_Unroll_Cnt
+ .if _UNROLL_CNT <> (ROUNDS_512/8)
+ movq $_UNROLL_CNT,%rax
+ .else
+ xorq %rax,%rax
+ .endif
+ ret
+.endif
+#
+.endif # _USE_ASM_ & 512
+#
+#=================================== Skein1024 =============================================
+.if _USE_ASM_ & 1024
+#
+# void Skein1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+# use details of permutation to make register assignments
+#
+o1K_rdi = 0 #offsets in X[] associated with each register
+o1K_rsi = 1
+o1K_rbp = 2
+o1K_rax = 3
+o1K_rcx = 4 #rcx is "shared" with X6, since X4/X6 alternate
+o1K_rbx = 5
+o1K_rdx = 7
+o1K_r8 = 8
+o1K_r9 = 9
+o1K_r10 = 10
+o1K_r11 = 11
+o1K_r12 = 12
+o1K_r13 = 13
+o1K_r14 = 14
+o1K_r15 = 15
+#
+rIdx_offs = tmpStk_1024
+#
+.macro r1024_Mix w0,w1,reg0,reg1,_RN0_,_Rn1_,op1
+ addReg \reg0 , \reg1 #perform the MIX
+ RotL64 \reg1 , 1024,%((_RN0_) % 8),_Rn1_
+ xorReg \reg1 , \reg0
+.if ((_RN0_) && 3) == 3 #time to do key injection?
+ .if _SKEIN_DEBUG
+ movq %\reg0 , xDebug_1024+8*w0(%rsp) #save intermediate values for Debug_Round
+ movq %\reg1 , xDebug_1024+8*w1(%rsp) # (before inline key injection)
+ .endif
+_II_ = ((_RN0_)/4)+1 #injection count
+ .if SKEIN_ASM_UNROLL && 1024 #here to do fully unrolled key injection
+ addq ksKey+ 8*((_II_+w0) % 17)(%rsp),%\reg0
+ addq ksKey+ 8*((_II_+w1) % 17)(%rsp),%\reg1
+ .if w1 == 13 #tweak injection
+ addq ksTwk+ 8*((_II_+ 0) % 3)(%rsp),%\reg1
+ .elseif w0 == 14
+ addq ksTwk+ 8*((_II_+ 1) % 3)(%rsp),%\reg0
+ .elseif w1 == 15
+ addq $_II_, %\reg1 #(injection counter)
+ .endif
+ .else #here to do looping key injection
+ .if (w0 == 0)
+ movq %rdi, X_stk+8*w0(%rsp) #if so, store N0 so we can use reg as index
+ movq rIdx_offs(%rsp),%rdi #get the injection counter index into rdi
+ .else
+ addq ksKey+8+8*w0(%rsp,%rdi,8),%\reg0 #even key injection
+ .endif
+ .if w1 == 13 #tweak injection
+ addq ksTwk+8+8* 0(%rsp,%rdi,8),%\reg1
+ .elseif w0 == 14
+ addq ksTwk+8+8* 1(%rsp,%rdi,8),%\reg0
+ .elseif w1 == 15
+ addReg \reg1,rdi,,,1 #(injection counter)
+ .endif
+ addq ksKey+8+8*w1(%rsp,%rdi,8),%\reg1 #odd key injection
+ .endif
+.endif
+ # insert the op provided, .if any
+ op1
+.endm
+#################
+# MACRO: four rounds for 1024-bit blocks
+#
+.macro r1024_FourRounds _RR_ #RR = base round number (0 mod 4)
+ # should be here with X4 set properly, X6 stored on stack
+_Rn_ = (_RR_) + 0
+ r1024_Mix 0, 1,rdi,rsi,_Rn_,0
+ r1024_Mix 2, 3,rbp,rax,_Rn_,1
+ r1024_Mix 4, 5,rcx,rbx,_Rn_,2,<movq %rcx,X_stk+8*4(%rsp)> #save X4 on stack (x4/x6 alternate)
+ r1024_Mix 8, 9,r8 ,r9 ,_Rn_,4,<movq X_stk+8*6(%rsp),%rcx> #load X6 from stack
+ r1024_Mix 10,11,r10,r11,_Rn_,5
+ r1024_Mix 12,13,r12,r13,_Rn_,6
+ r1024_Mix 6, 7,rcx,rdx,_Rn_,3
+ r1024_Mix 14,15,r14,r15,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+_Rn_ = (_RR_) + 1
+ r1024_Mix 0, 9,rdi,r9 ,_Rn_,0
+ r1024_Mix 2,13,rbp,r13,_Rn_,1
+ r1024_Mix 6,11,rcx,r11,_Rn_,2,<movq %rcx,X_stk+8*6(%rsp)> #save X6 on stack (x4/x6 alternate)
+ r1024_Mix 10, 7,r10,rdx,_Rn_,4,<movq X_stk+8*4(%rsp),%rcx> #load X4 from stack
+ r1024_Mix 12, 3,r12,rax,_Rn_,5
+ r1024_Mix 14, 5,r14,rbx,_Rn_,6
+ r1024_Mix 4,15,rcx,r15,_Rn_,3
+ r1024_Mix 8, 1,r8 ,rsi,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+_Rn_ = (_RR_) + 2
+ r1024_Mix 0, 7,rdi,rdx,_Rn_,0
+ r1024_Mix 2, 5,rbp,rbx,_Rn_,1
+ r1024_Mix 4, 3,rcx,rax,_Rn_,2,<movq %rcx,X_stk+8*4(%rsp)> #save X4 on stack (x4/x6 alternate)
+ r1024_Mix 12,15,r12,r15,_Rn_,4,<movq X_stk+8*6(%rsp),%rcx> #load X6 from stack
+ r1024_Mix 14,13,r14,r13,_Rn_,5
+ r1024_Mix 8,11,r8 ,r11,_Rn_,6
+ r1024_Mix 6, 1,rcx,rsi,_Rn_,3
+ r1024_Mix 10, 9,r10,r9 ,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+_Rn_ = (_RR_) + 3
+ r1024_Mix 0,15,rdi,r15,_Rn_,0
+ r1024_Mix 2,11,rbp,r11,_Rn_,1
+ r1024_Mix 6,13,rcx,r13,_Rn_,2,<movq %rcx,X_stk+8*6(%rsp)> #save X6 on stack (x4/x6 alternate)
+ r1024_Mix 14, 1,r14,rsi,_Rn_,4,<movq X_stk+8*4(%rsp),%rcx> #load X4 from stack
+ r1024_Mix 8, 5,r8 ,rbx,_Rn_,5
+ r1024_Mix 10, 3,r10,rax,_Rn_,6
+ r1024_Mix 4, 9,rcx,r9 ,_Rn_,3
+ r1024_Mix 12, 7,r12,rdx,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+
+ .if (SKEIN_ASM_UNROLL && 1024) == 0 #here with rdi == rIdx, X0 on stack
+ #"rotate" the key schedule on the stack
+i8 = o1K_r8
+i0 = o1K_rdi
+ movq %r8 , X_stk+8*i8(%rsp) #free up a register (save it on the stack)
+ movq ksKey+8* 0(%rsp,%rdi,8),%r8 #get key word
+ movq %r8 , ksKey+8*17(%rsp,%rdi,8) #rotate key (must do key first or tweak clobbers it!)
+ movq ksTwk+8* 0(%rsp,%rdi,8),%r8 #get tweak word
+ movq %r8 , ksTwk+8* 3(%rsp,%rdi,8) #rotate tweak (onto the stack)
+ movq X_stk+8*i8(%rsp) ,%r8 #get the reg back
+ incq %rdi #bump the index
+ movq %rdi, rIdx_offs (%rsp) #save rdi again
+ movq ksKey+8*i0(%rsp,%rdi,8),%rdi #get the key schedule word for X0 back
+ addq X_stk+8*i0(%rsp) ,%rdi #perform the X0 key injection
+ .endif
+ #show the result of the key injection
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT
+.endm #r1024_FourRounds
+#
+################
+# code
+#
+C_label Skein1024_Process_Block
+#
+ Setup_Stack 1024,ROUNDS_1024/8,WCNT
+ movq TWEAK+ 8(%rdi),%r9
+ jmp Skein1024_block_loop
+ # main hash loop for Skein1024
+ .p2align 4
+Skein1024_block_loop:
+ # general register usage:
+ # RSP = stack pointer
+ # RAX..RDX,RSI,RDI = X1, X3..X7 (state words)
+ # R8 ..R15 = X8..X15 (state words)
+ # RBP = temp (used for X0 and X2)
+ #
+ .if (SKEIN_ASM_UNROLL & 1024) == 0
+ xorq %rax,%rax #init loop index on the stack
+ movq %rax,rIdx_offs(%rsp)
+ .endif
+ movq TWEAK+ 0(%rdi),%r8
+ addq bitAdd+ F_O(%rbp),%r8 #computed updated tweak value T0
+ movq %r9 ,%r10
+ xorq %r8 ,%r10 #%rax/%rbx/%rcx = tweak schedule
+ movq %r8 ,TWEAK+ 0(%rdi) #save updated tweak value ctx->h.T[0]
+ movq %r8 ,ksTwk+ 0+F_O(%rbp)
+ movq %r9 ,ksTwk+ 8+F_O(%rbp) #keep values in %r8 ,%r9 for initial tweak injection below
+ movq %r10,ksTwk+16+F_O(%rbp)
+ .if _SKEIN_DEBUG
+ movq %r9 ,TWEAK+ 8(%rdi) #save updated tweak value ctx->h.T[1] for Skein_Debug_Block
+ .endif
+ movq blkPtr +F_O(%rbp),%rsi # rsi --> input block
+ movq $KW_PARITY ,%rax #overall key schedule parity
+
+ # the logic here assumes the set {rdi,rsi,rbp,rax} = X[0,1,2,3]
+ .irp _rN_,0,1,2,3,4,6 #process the "initial" words, using r14/r15 as temps
+ movq X_VARS+8*_rN_(%rdi),%r14 #get state word
+ movq 8*_rN_(%rsi),%r15 #get msg word
+ xorq %r14,%rax #update key schedule overall parity
+ movq %r14,ksKey +8*_rN_+F_O(%rbp) #save key schedule word on stack
+ movq %r15,Wcopy +8*_rN_+F_O(%rbp) #save local msg Wcopy
+ addq %r15,%r14 #do the initial key injection
+ movq %r14,X_stk +8*_rN_ (%rsp) #save initial state var on stack
+ .endr
+ # now process the rest, using the "real" registers
+ # (MUST do it in reverse order to inject tweaks r8/r9 first)
+ .irp _rr_,r15,r14,r13,r12,r11,r10,r9,r8,rdx,rbx
+_oo_ = o1K_\_rr_ #offset assocated with the register
+ movq X_VARS+8*_oo_(%rdi),%\_rr_ #get key schedule word from context
+ movq 8*_oo_(%rsi),%rcx #get next input msg word
+ movq %\_rr_, ksKey +8*_oo_(%rsp) #save key schedule on stack
+ xorq %\_rr_, %rax #accumulate key schedule parity
+ movq %rcx,Wcopy+8*_oo_+F_O(%rbp) #save copy of msg word for feedforward
+ addq %rcx,%\_rr_ #do the initial key injection
+ .if _oo_ == 13 #do the initial tweak injection
+ addReg _rr_,r8 # (only in words 13/14)
+ .elseif _oo_ == 14
+ addReg _rr_,r9
+ .endif
+ .endr
+ movq %rax,ksKey+8*WCNT+F_O(%rbp) #save key schedule parity
+.if _SKEIN_DEBUG
+ Skein_Debug_Block 1024 #initial debug dump
+.endif
+ addq $8*WCNT,%rsi #bump the msg ptr
+ movq %rsi,blkPtr+F_O(%rbp) #save bumped msg ptr
+ # re-load words 0..4 from stack, enter the main loop
+ .irp _rr_,rdi,rsi,rbp,rax,rcx #(no need to re-load x6, already on stack)
+ movq X_stk+8*o1K_\_rr_(%rsp),%\_rr_ #re-load state and get ready to go!
+ .endr
+.if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL #show state after initial key injection
+.endif
+ #
+ #################
+ # now the key schedule is computed. Start the rounds
+ #
+.if SKEIN_ASM_UNROLL & 1024
+_UNROLL_CNT = ROUNDS_1024/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_1024
+ .if ((ROUNDS_1024/8) % _UNROLL_CNT)
+ .error "Invalid SKEIN_UNROLL_1024"
+ .endif
+Skein1024_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2 #implement the rounds, 4 at a time
+ r1024_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 1024) == 0
+ cmpq $2*(ROUNDS_1024/8),tmpStk_1024(%rsp) #see .if we are done
+ jb Skein1024_round_loop
+.endif
+ # end of rounds
+ #################
+ #
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..15}
+ movq %rdx,X_stk+8*o1K_rdx(%rsp) #we need a register. x6 already on stack
+ movq ctxPtr(%rsp),%rdx
+
+ .irp _rr_,rdi,rsi,rbp,rax,rcx,rbx,r8,r9,r10,r11,r12,r13,r14,r15 #do all but x6,x7
+_oo_ = o1K_\_rr_
+ xorq Wcopy +8*_oo_(%rsp),%\_rr_ #feedforward XOR
+ movq %\_rr_,X_VARS+8*_oo_(%rdx) #save result into context
+ .if (_oo_ == 9)
+ movq $FIRST_MASK64 ,%r9
+ .endif
+ .if (_oo_ == 14)
+ andq TWEAK+ 8(%rdx),%r9
+ .endif
+ .endr
+ #
+ movq X_stk +8*6(%rsp),%rax #now process x6,x7 (skipped in .irp above)
+ movq X_stk +8*7(%rsp),%rbx
+ xorq Wcopy +8*6(%rsp),%rax
+ xorq Wcopy +8*7(%rsp),%rbx
+ movq %rax,X_VARS+8*6(%rdx)
+ decq blkCnt(%rsp) #set zero flag iff done
+ movq %rbx,X_VARS+8*7(%rdx)
+
+ Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD,,<cmpq $0,blkCnt(%rsp)>
+ # go back for more blocks, if needed
+ movq ctxPtr(%rsp),%rdi #don't muck with the flags here!
+ lea FRAME_OFFS(%rsp),%rbp
+ jnz Skein1024_block_loop
+ movq %r9 ,TWEAK+ 8(%rdx)
+ Reset_Stack
+ ret
+#
+Skein1024_Process_Block_End:
+#
+.if _SKEIN_DEBUG
+Skein_Debug_Round_1024:
+ # call here with rdx = "round number",
+_SP_OFFS_ = 8*2 #stack "offset" here: rdx, return addr
+ #
+ #save rest of X[] state on stack so debug routines can access it
+ .irp _rr_,rsi,rbp,rax,rbx,r8,r9,r10,r11,r12,r13,r14,r15
+ movq %\_rr_,X_stk+8*o1K_\_rr_+_SP_OFFS_(%rsp)
+ .endr
+ # Figure out what to do with x0 (rdi). When rdx == 0 mod 4, it's already on stack
+ cmpq $SKEIN_RND_SPECIAL,%rdx #special rounds always save
+ jae save_x0
+ testq $3,%rdx #otherwise only if rdx != 0 mod 4
+ jz save_x0_not
+save_x0:
+ movq %rdi,X_stk+8*o1K_rdi+_SP_OFFS_(%rsp)
+save_x0_not:
+ #figure out the x4/x6 swapping state and save the correct one!
+ cmpq $SKEIN_RND_SPECIAL,%rdx #special rounds always do x4
+ jae save_x4
+ testq $1,%rdx #and even ones have r4 as well
+ jz save_x4
+ movq %rcx,X_stk+8*6+_SP_OFFS_(%rsp)
+ jmp debug_1024_go
+save_x4:
+ movq %rcx,X_stk+8*4+_SP_OFFS_(%rsp)
+debug_1024_go:
+ #now all is saved in Xstk[] except for rdx
+ push %rsi #save two regs for BLK_BITS-specific parms
+ push %rdi
+_SP_OFFS_ = _SP_OFFS_ + 16 #adjust stack offset accordingly (now 32)
+
+ movq _SP_OFFS_-8(%rsp),%rsi #get back original %rdx (pushed on stack in macro call)
+ movq %rsi,X_stk+8*o1K_rdx+_SP_OFFS_(%rsp) #and save it in its rightful place in X_stk[]
+
+ movq ctxPtr+_SP_OFFS_(%rsp),%rsi #rsi = ctx_hdr_ptr
+ movq $1024,%rdi #rdi = block size
+ jmp Skein_Debug_Round_Common
+.endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein1024_Process_Block_CodeSize
+ movq $(Skein1024_Process_Block_End-Skein1024_Process_Block),%rax
+ ret
+#
+C_label Skein1024_Unroll_Cnt
+ .if _UNROLL_CNT <> (ROUNDS_1024/8)
+ movq $_UNROLL_CNT,%rax
+ .else
+ xorq %rax,%rax
+ .endif
+ ret
+.endif
+#
+.endif # _USE_ASM_ and 1024
+#
+.if _SKEIN_DEBUG
+#----------------------------------------------------------------
+#local debug routine to set up for calls to:
+# void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X)
+# [ rdi rsi rdx rcx]
+#
+# here with %rdx = round number
+# %rsi = ctx_hdr_ptr
+# %rdi = block size (256/512/1024)
+# on stack: saved rdi, saved rsi, retAddr, saved rdx
+#
+Skein_Debug_Round_Common:
+_SP_OFFS_ = 32 #account for four words on stack already
+ .irp _rr_,rax,rbx,rcx,rbp,r8,r9,r10,r11,r12,r13,r14,r15 #save the rest of the regs
+ pushq %\_rr_
+_SP_OFFS_ = _SP_OFFS_+8
+ .endr
+ .if (_SP_OFFS_ % 16) # make sure stack is still 16-byte aligned here
+ .error "Debug_Round_Common: stack alignment"
+ .endif
+ # compute %rcx = ptr to the X[] array on the stack (final parameter to call)
+ leaq X_stk+_SP_OFFS_(%rsp),%rcx #adjust for reg pushes, return address
+ cmpq $SKEIN_RND_FEED_FWD,%rdx #special handling for feedforward "round"?
+ jnz _got_rcxA
+ leaq X_VARS(%rsi),%rcx
+_got_rcxA:
+ .if _USE_ASM_ & 1024
+ # special handling for 1024-bit case
+ # (for rounds right before with key injection:
+ # use xDebug_1024[] instead of X_stk[])
+ cmpq $SKEIN_RND_SPECIAL,%rdx
+ jae _got_rcxB #must be a normal round
+ orq %rdx,%rdx
+ jz _got_rcxB #just before key injection
+ test $3,%rdx
+ jne _got_rcxB
+ cmp $1024,%rdi #only 1024-bit(s) for now
+ jne _got_rcxB
+ leaq xDebug_1024+_SP_OFFS_(%rsp),%rcx
+_got_rcxB:
+ .endif
+ call Skein_Show_Round #call external debug handler
+
+ .irp _rr_,r15,r14,r13,r12,r11,r10,r9,r8,rbp,rcx,rbx,rax #restore regs
+ popq %\_rr_
+_SP_OFFS_ = _SP_OFFS_-8
+ .endr
+ .if _SP_OFFS_ - 32
+ .error "Debug_Round_Common: push/pop misalignment!"
+ .endif
+ popq %rdi
+ popq %rsi
+ ret
+.endif
+#----------------------------------------------------------------
+ .end
Index: sys/contrib/skein/asm/skein_block_x86.asm
===================================================================
--- /dev/null
+++ sys/contrib/skein/asm/skein_block_x86.asm
@@ -0,0 +1,1180 @@
+;
+;----------------------------------------------------------------
+; 32-bit x86 assembler code for Skein block functions
+;
+; Author: Doug Whiting, Hifn
+;
+; This code is released to the public domain.
+;----------------------------------------------------------------
+;
+ .386p
+ .model flat
+ .code
+;
+_MASK_ALL_ equ (256+512+1024) ;all three algorithm bits
+;
+;;;;;;;;;;;;;;;;;
+ifndef SKEIN_USE_ASM
+_USE_ASM_ = _MASK_ALL_
+elseif SKEIN_USE_ASM and _MASK_ALL_
+_USE_ASM_ = SKEIN_USE_ASM
+else
+_USE_ASM_ = _MASK_ALL_
+endif
+;;;;;;;;;;;;;;;;;
+ifndef SKEIN_LOOP
+_SKEIN_LOOP = 0 ;default is all fully unrolled
+else
+_SKEIN_LOOP = SKEIN_LOOP
+endif
+; the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256 = (_SKEIN_LOOP / 100) mod 10
+SKEIN_UNROLL_512 = (_SKEIN_LOOP / 10) mod 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP ) mod 10
+;
+SKEIN_ASM_UNROLL = 0
+ irp _NN_,<256,512,1024>
+ if (SKEIN_UNROLL_&_NN_) eq 0
+SKEIN_ASM_UNROLL = SKEIN_ASM_UNROLL + _NN_
+ endif
+ endm
+;;;;;;;;;;;;;;;;;
+;
+ifndef SKEIN_ROUNDS
+ROUNDS_256 = 72
+ROUNDS_512 = 72
+ROUNDS_1024 = 80
+else
+ROUNDS_256 = 8*((((SKEIN_ROUNDS / 100) + 5) mod 10) + 5)
+ROUNDS_512 = 8*((((SKEIN_ROUNDS / 10) + 5) mod 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS ) + 5) mod 10) + 5)
+endif
+irp _NN_,<256,512,1024>
+ if _USE_ASM_ and _NN_
+ irp _RR_,<%(ROUNDS_&_NN_)>
+ if _NN_ eq 1024
+%out +++ SKEIN_ROUNDS_&_NN_ = _RR_
+ else
+%out +++ SKEIN_ROUNDS_&_NN_ = _RR_
+ endif
+ endm
+ endif
+endm
+;;;;;;;;;;;;;;;;;
+;
+ifdef SKEIN_CODE_SIZE
+_SKEIN_CODE_SIZE equ (1)
+else
+ifdef SKEIN_PERF ;use code size if SKEIN_PERF is defined
+_SKEIN_CODE_SIZE equ (1)
+endif
+endif
+;
+;;;;;;;;;;;;;;;;;
+;
+ifndef SKEIN_DEBUG
+_SKEIN_DEBUG = 0
+else
+_SKEIN_DEBUG = 1
+endif
+;;;;;;;;;;;;;;;;;
+;
+; define offsets of fields in hash context structure
+;
+HASH_BITS = 0 ;# bits of hash output
+BCNT = 4 + HASH_BITS ;number of bytes in BUFFER[]
+TWEAK = 4 + BCNT ;tweak values[0..1]
+X_VARS = 16 + TWEAK ;chaining vars
+;
+;(Note: buffer[] in context structure is NOT needed here :-)
+;
+KW_PARITY_LO= 0A9FC1A22h ;overall parity of key schedule words (hi32/lo32)
+KW_PARITY_HI= 01BD11BDAh ;overall parity of key schedule words (hi32/lo32)
+FIRST_MASK = NOT (1 SHL 30) ;FIRST block flag bit
+;
+; rotation constants for Skein
+;
+RC_256_0_0 = 14
+RC_256_0_1 = 16
+
+RC_256_1_0 = 52
+RC_256_1_1 = 57
+
+RC_256_2_0 = 23
+RC_256_2_1 = 40
+
+RC_256_3_0 = 5
+RC_256_3_1 = 37
+
+RC_256_4_0 = 25
+RC_256_4_1 = 33
+
+RC_256_5_0 = 46
+RC_256_5_1 = 12
+
+RC_256_6_0 = 58
+RC_256_6_1 = 22
+
+RC_256_7_0 = 32
+RC_256_7_1 = 32
+
+RC_512_0_0 = 46
+RC_512_0_1 = 36
+RC_512_0_2 = 19
+RC_512_0_3 = 37
+
+RC_512_1_0 = 33
+RC_512_1_1 = 27
+RC_512_1_2 = 14
+RC_512_1_3 = 42
+
+RC_512_2_0 = 17
+RC_512_2_1 = 49
+RC_512_2_2 = 36
+RC_512_2_3 = 39
+
+RC_512_3_0 = 44
+RC_512_3_1 = 9
+RC_512_3_2 = 54
+RC_512_3_3 = 56
+
+RC_512_4_0 = 39
+RC_512_4_1 = 30
+RC_512_4_2 = 34
+RC_512_4_3 = 24
+
+RC_512_5_0 = 13
+RC_512_5_1 = 50
+RC_512_5_2 = 10
+RC_512_5_3 = 17
+
+RC_512_6_0 = 25
+RC_512_6_1 = 29
+RC_512_6_2 = 39
+RC_512_6_3 = 43
+
+RC_512_7_0 = 8
+RC_512_7_1 = 35
+RC_512_7_2 = 56
+RC_512_7_3 = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 = 8
+RC_1024_0_3 = 47
+RC_1024_0_4 = 8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 = 4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 = 5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 = 9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 = 4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 = 9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+;
+; Input: rHi,rLo
+; Output: <rHi,rLo> <<< _RCNT_
+Rol64 macro rHi,rLo,tmp,_RCNT_
+ if _RCNT_ ;is there anything to do?
+ if _RCNT_ lt 32
+ mov tmp,rLo
+ shld rLo,rHi,_RCNT_
+ shld rHi,tmp,_RCNT_
+ elseif _RCNT_ gt 32
+ mov tmp,rLo
+ shrd rLo,rHi,((64-_RCNT_) AND 63)
+ shrd rHi,tmp,((64-_RCNT_) AND 63)
+ else
+ xchg rHi,rLo ;special case for _RCNT_ == 32
+ endif
+ endif
+endm
+;
+; Input: rHi,rLo
+; Output: <rHi,rLo> <<< rName&&rNum, and tmp trashed;
+RotL64 macro rHi,rLo,tmp,BLK_SIZE,ROUND_NUM,MIX_NUM
+_RCNT_ = ( RC_&BLK_SIZE&_&ROUND_NUM&_&MIX_NUM AND 63 )
+ Rol64 rHi,rLo,tmp,_RCNT_
+endm
+;
+;----------------------------------------------------------------
+; declare allocated space on the stack
+StackVar macro localName,localSize
+localName = _STK_OFFS_
+_STK_OFFS_ = _STK_OFFS_+(localSize)
+endm ;StackVar
+;
+;----------------------------------------------------------------
+;
+; MACRO: Configure stack frame, allocate local vars
+;
+Setup_Stack macro WCNT,KS_CNT
+_STK_OFFS_ = 0 ;starting offset from esp
+ ;----- local variables ;<-- esp
+ StackVar X_stk ,8*(WCNT) ;local context vars
+ StackVar Wcopy ,8*(WCNT) ;copy of input block
+ StackVar ksTwk ,8*3 ;key schedule: tweak words
+ StackVar ksKey ,8*(WCNT)+8 ;key schedule: key words
+ if WCNT le 8
+FRAME_OFFS = _STK_OFFS_ ;<-- ebp
+ else
+FRAME_OFFS = _STK_OFFS_-8*4 ;<-- ebp
+ endif
+ if (SKEIN_ASM_UNROLL and (WCNT*64)) eq 0
+ StackVar ksRot ,16*(KS_CNT+0);leave space for "rotation" to happen
+ endif
+LOCAL_SIZE = _STK_OFFS_ ;size of local vars
+ ;-----
+ StackVar savRegs,8*4 ;pushad data
+ StackVar retAddr,4 ;return address
+ ;----- caller parameters
+ StackVar ctxPtr ,4 ;context ptr
+ StackVar blkPtr ,4 ;pointer to block data
+ StackVar blkCnt ,4 ;number of full blocks to process
+ StackVar bitAdd ,4 ;bit count to add to tweak
+ ;----- caller's stack frame
+;
+; Notes on stack frame setup:
+; * the most frequently used variable is X_stk[], based at [esp+0]
+; * the next most used is the key schedule words
+; so ebp is "centered" there, allowing short offsets to the key/tweak
+; schedule even in 1024-bit Skein case
+; * the Wcopy variables are infrequently accessed, but they have long
+; offsets from both esp and ebp only in the 1024-bit case.
+; * all other local vars and calling parameters can be accessed
+; with short offsets, except in the 1024-bit case
+;
+ pushad ;save all regs
+ sub esp,LOCAL_SIZE ;make room for the locals
+ lea ebp,[esp+FRAME_OFFS] ;maximize use of short offsets
+ mov edi,[FP_+ctxPtr ] ;edi --> context
+;
+endm ;Setup_Stack
+;
+FP_ equ <ebp-FRAME_OFFS> ;keep as many short offsets as possible
+;
+;----------------------------------------------------------------
+;
+Reset_Stack macro procStart
+ add esp,LOCAL_SIZE ;get rid of locals (wipe??)
+ popad ;restore all regs
+
+ ;display code size in bytes to stdout
+ irp _BCNT_,<%($+1-procStart)> ;account for return opcode
+if _BCNT_ ge 10000 ;(align it all pretty)
+%out procStart code size = _BCNT_ bytes
+elseif _BCNT_ ge 1000
+%out procStart code size = _BCNT_ bytes
+else
+%out procStart code size = _BCNT_ bytes
+endif
+ endm ;irp _BCNT_
+
+endm ; Reset_Stack
+;
+;----------------------------------------------------------------
+; macros to help debug internals
+;
+if _SKEIN_DEBUG
+ extrn _Skein_Show_Block:near ;calls to C routines
+ extrn _Skein_Show_Round:near
+;
+SKEIN_RND_SPECIAL = 1000
+SKEIN_RND_KEY_INITIAL = SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT = SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD = SKEIN_RND_SPECIAL+2
+;
+Skein_Debug_Block macro BLK_BITS
+;
+;void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+; const u08b_t *blkPtr, const u64b_t *wPtr,
+; const u64b_t *ksPtr,const u64b_t *tsPtr);
+;
+ pushad ;save all regs
+ lea eax,[FP_+ksTwk]
+ lea ebx,[FP_+ksKey]
+ lea ecx,[esp+32+Wcopy]
+ mov edx,[FP_+ctxPtr] ;ctx_hdr_ptr
+ lea edx,[edx+X_VARS] ;edx ==> cxt->X[]
+ push eax ;tsPtr
+ push ebx ;ksPtr
+ push ecx ;wPtr
+ push dword ptr [FP_+blkPtr] ;blkPtr
+ push edx ;ctx->Xptr
+ push dword ptr [FP_+ctxPtr] ;ctx_hdr_ptr
+ mov eax,BLK_BITS
+ push eax ;bits
+ ifdef _MINGW_
+ call _Skein_Show_Block-4 ;strange linkage??
+ else
+ call _Skein_Show_Block
+ endif
+ add esp,7*4 ;discard parameter space on stack
+ popad ;restore regs
+endm ;Skein_Debug_Block
+
+;
+Skein_Debug_Round macro BLK_SIZE,R,saveRegs
+;
+;void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X);
+;
+ ifnb <saveRegs>
+ mov [esp+X_stk+ 0],eax ;save internal vars for debug dump
+ mov [esp+X_stk+ 4],ebx
+ mov [esp+X_stk+ 8],ecx
+ mov [esp+X_stk+12],edx
+ endif
+ pushad ;save all regs
+ if R ne SKEIN_RND_FEED_FWD
+ lea eax,[esp+32+X_stk]
+ else
+ mov eax,[FP_+ctxPtr]
+ add eax,X_VARS
+ endif
+ push eax ;Xptr
+ if (SKEIN_ASM_UNROLL and BLK_SIZE) or (R ge SKEIN_RND_SPECIAL)
+ mov eax,R
+ else
+ lea eax,[4*edi+1+(((R)-1) and 3)] ;compute round number using edi
+ endif
+ push eax ;round number
+ push dword ptr [FP_+ctxPtr] ;ctx_hdr_ptr
+ mov eax,BLK_SIZE
+ push eax ;bits
+ ifdef _MINGW_
+ call _Skein_Show_Round-4 ;strange linkage??
+ else
+ call _Skein_Show_Round
+ endif
+ add esp,4*4 ;discard parameter space on stack
+ popad ;restore regs
+endm ;Skein_Debug_Round
+endif ;ifdef SKEIN_DEBUG
+;
+;----------------------------------------------------------------
+;
+; MACRO: a mix step
+;
+MixStep macro BLK_SIZE,ld_A,ld_C,st_A,st_C,RotNum0,RotNum1,_debug_
+ ifnb <ld_A>
+ mov eax,[esp+X_stk+8*(ld_A)+0]
+ mov ebx,[esp+X_stk+8*(ld_A)+4]
+ endif
+ ifnb <ld_C>
+ mov ecx,[esp+X_stk+8*(ld_C)+0]
+ mov edx,[esp+X_stk+8*(ld_C)+4]
+ endif
+ add eax, ecx ;X[A] += X[C]
+ adc ebx, edx
+ ifnb <st_A>
+ mov [esp+X_stk+8*(st_A)+0],eax
+ mov [esp+X_stk+8*(st_A)+4],ebx
+ endif
+__rNum0 = (RotNum0) AND 7
+ RotL64 ecx, edx, esi,%(BLK_SIZE),%(__rNum0),%(RotNum1) ;X[C] <<<= RC_<BLK_BITS,RotNum0,RotNum1>
+ xor ecx, eax ;X[C] ^= X[A]
+ xor edx, ebx
+ if _SKEIN_DEBUG or (0 eq (_debug_ + 0))
+ ifb <st_C>
+ mov [esp+X_stk+8*(ld_C)+0],ecx
+ mov [esp+X_stk+8*(ld_C)+4],edx
+ else
+ mov [esp+X_stk+8*(st_C)+0],ecx
+ mov [esp+X_stk+8*(st_C)+4],edx
+ endif
+ endif
+ if _SKEIN_DEBUG and (0 ne (_debug_ + 0))
+ Skein_Debug_Round BLK_SIZE,%(RotNum0+1)
+ endif
+endm ;MixStep
+;
+;;;;;;;;;;;;;;;;;
+;
+; MACRO: key schedule injection
+;
+ks_Inject macro BLK_SIZE,X_load,X_stor,rLo,rHi,rndBase,keyIdx,twkIdx,ROUND_ADD
+ ;are rLo,rHi values already loaded? if not, load them now
+ ifnb <X_load>
+ mov rLo,[esp+X_stk +8*(X_load) ]
+ mov rHi,[esp+X_stk +8*(X_load)+4]
+ endif
+
+ ;inject the 64-bit key schedule value (and maybe the tweak as well)
+if SKEIN_ASM_UNROLL and BLK_SIZE
+_kOffs_ = ((rndBase)+(keyIdx)) mod ((BLK_SIZE/64)+1)
+ add rLo,[FP_+ksKey+8*_kOffs_+ 0]
+ adc rHi,[FP_+ksKey+8*_kOffs_+ 4]
+ ifnb <twkIdx>
+_tOffs_ = ((rndBase)+(twkIdx)) mod 3
+ add rLo,[FP_+ksTwk+8*_tOffs_+ 0]
+ adc rHi,[FP_+ksTwk+8*_tOffs_+ 4]
+ endif
+ ifnb <ROUND_ADD>
+ add rLo,(ROUND_ADD)
+ adc rHi,0
+ endif
+else
+ add rLo,[FP_+ksKey+8*(keyIdx)+8*edi ]
+ adc rHi,[FP_+ksKey+8*(keyIdx)+8*edi+4]
+ ifnb <twkIdx>
+ add rLo,[FP_+ksTwk+8*(twkIdx)+8*edi ]
+ adc rHi,[FP_+ksTwk+8*(twkIdx)+8*edi+4]
+ endif
+ ifnb <ROUND_ADD>
+ add rLo,edi ;edi is the round number
+ adc rHi,0
+ endif
+endif
+
+ ;do we need to store updated rLo,rHi values? if so, do it now
+ ifnb <X_stor>
+ mov [esp+X_stk +8*(X_stor) ],rLo
+ mov [esp+X_stk +8*(X_stor)+4],rHi
+ endif
+endm ;ks_Inject
+;
+;----------------------------------------------------------------
+; MACRO: key schedule rotation
+;
+ks_Rotate macro rLo,rHi,WCNT
+ mov rLo,[FP_+ksKey+8*edi+ 0] ;"rotate" the key schedule in memory
+ mov rHi,[FP_+ksKey+8*edi+ 4]
+ mov [FP_+ksKey+8*edi+8*(WCNT+1)+ 0],rLo
+ mov [FP_+ksKey+8*edi+8*(WCNT+1)+ 4],rHi
+ mov rLo,[FP_+ksTwk+8*edi+ 0]
+ mov rHi,[FP_+ksTwk+8*edi+ 4]
+ mov [FP_+ksTwk+8*edi+8*3+ 0],rLo
+ mov [FP_+ksTwk+8*edi+8*3+ 4],rHi
+endm
+;
+;----------------------------------------------------------------
+;
+if _USE_ASM_ and 256
+ public _Skein_256_Process_Block
+;
+; void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+;;;;;;;;;;;;;;;;;
+;
+; MACRO: two rounds
+;
+R_256_TwoRounds macro _RR_,ld_0
+ ; here with edx:ecx = X[1]
+ ;--------- round _RR_
+ MixStep 256,ld_0, ,0,1,((_RR_)+0),0
+ MixStep 256, 2,3,2,3,((_RR_)+0),1,1
+
+ ; here with edx:ecx = X[3]
+ ;--------- round _RR_ + 1
+ MixStep 256, 0, ,0,3,((_RR_)+1),0
+ MixStep 256, 2,1,2,1,((_RR_)+1),1,1
+
+ ; here with edx:ecx = X[1]
+endm ;R_256_TwoRounds
+;
+;;;;;;;;;;;;;;;;;
+;
+; code
+;
+_Skein_256_Process_Block proc near
+ WCNT = 4 ;WCNT=4 for Skein-256
+ Setup_Stack WCNT,(ROUNDS_256/8)
+
+ ; main hash loop for Skein_256
+Skein_256_block_loop:
+ mov eax,[edi+TWEAK+ 0] ;ebx:eax = tweak word T0
+ mov ebx,[edi+TWEAK+ 4]
+ mov ecx,[edi+TWEAK+ 8] ;edx:ecx = tweak word T1
+ mov edx,[edi+TWEAK+12]
+
+ add eax,[FP_+bitAdd ] ;bump T0 by the bitAdd parameter
+ adc ebx, 0
+ mov [edi+TWEAK ],eax ;save updated tweak value T0
+ mov [edi+TWEAK+ 4],ebx
+
+ mov [FP_+ksTwk ],eax ;build the tweak schedule on the stack
+ mov [FP_+ksTwk+ 4],ebx
+ xor eax,ecx ;ebx:eax = T0 ^ T1
+ xor ebx,edx
+ mov [FP_+ksTwk+ 8],ecx
+ mov [FP_+ksTwk+12],edx
+ mov [FP_+ksTwk+16],eax
+ mov [FP_+ksTwk+20],ebx
+
+ mov eax,KW_PARITY_LO ;init parity accumulator
+ mov ebx,KW_PARITY_HI
+;
+_NN_ = 0
+ rept WCNT ;copy in the chaining vars
+ mov ecx,[edi+X_VARS+_NN_ ]
+ mov edx,[edi+X_VARS+_NN_+ 4]
+ xor eax,ecx ;compute overall parity along the way
+ xor ebx,edx
+ mov [FP_+ksKey +_NN_ ],ecx
+ mov [FP_+ksKey +_NN_+ 4],edx
+_NN_ = _NN_+8
+ endm
+;
+ mov [FP_+ksKey +_NN_ ],eax ;save overall parity at the end of the array
+ mov [FP_+ksKey +_NN_+ 4],ebx
+
+ mov esi,[FP_+blkPtr ] ;esi --> input block
+;
+_NN_ = WCNT*8-16 ;work down from the end
+ rept WCNT/2 ;perform initial key injection
+ mov eax,[esi+_NN_ + 0]
+ mov ebx,[esi+_NN_ + 4]
+ mov ecx,[esi+_NN_ + 8]
+ mov edx,[esi+_NN_ +12]
+ mov [esp+_NN_+Wcopy + 0],eax
+ mov [esp+_NN_+Wcopy + 4],ebx
+ mov [esp+_NN_+Wcopy + 8],ecx
+ mov [esp+_NN_+Wcopy +12],edx
+ add eax,[FP_+_NN_+ksKey + 0]
+ adc ebx,[FP_+_NN_+ksKey + 4]
+ add ecx,[FP_+_NN_+ksKey + 8]
+ adc edx,[FP_+_NN_+ksKey +12]
+ if _NN_ eq (WCNT*8-16) ;inject the tweak words
+ add eax,[FP_+ ksTwk + 8]; (at the appropriate points)
+ adc ebx,[FP_+ ksTwk +12]
+ elseif _NN_ eq (WCNT*8-32)
+ add ecx,[FP_+ ksTwk + 0]
+ adc edx,[FP_+ ksTwk + 4]
+ endif
+ if _NN_ or _SKEIN_DEBUG
+ mov [esp+_NN_+X_stk + 0],eax
+ mov [esp+_NN_+X_stk + 4],ebx
+ mov [esp+_NN_+X_stk + 8],ecx
+ mov [esp+_NN_+X_stk +12],edx
+ endif
+_NN_ = _NN_ - 16 ;end at X[0], so regs are already loaded for first MIX!
+ endm
+;
+if _SKEIN_DEBUG ;debug dump of state at this point
+ Skein_Debug_Block WCNT*64
+ Skein_Debug_Round WCNT*64,SKEIN_RND_KEY_INITIAL
+endif
+ add esi, WCNT*8 ;skip the block
+ mov [FP_+blkPtr ],esi ;update block pointer
+ ;
+ ; now the key schedule is computed. Start the rounds
+ ;
+if SKEIN_ASM_UNROLL and 256
+_UNROLL_CNT = ROUNDS_256/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_256 ;unroll count
+ if ((ROUNDS_256/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_256"
+ endif
+ xor edi,edi ;edi = iteration count
+Skein_256_round_loop:
+endif
+_Rbase_ = 0
+rept _UNROLL_CNT*2
+ ; here with X[0], X[1] already loaded into eax..edx
+ R_256_TwoRounds %(4*_Rbase_+00),
+ R_256_TwoRounds %(4*_Rbase_+02),0
+
+ ;inject key schedule
+ if _UNROLL_CNT ne (ROUNDS_256/8)
+ ks_Rotate eax,ebx,WCNT
+ inc edi ;edi = round number
+ endif
+_Rbase_ = _Rbase_+1
+ ks_Inject 256,3,3,eax,ebx,_Rbase_,3, ,_Rbase_
+ ks_Inject 256,2,2,eax,ebx,_Rbase_,2,1
+ ks_Inject 256, , ,ecx,edx,_Rbase_,1,0
+ ks_Inject 256,0, ,eax,ebx,_Rbase_,0
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT,saveRegs
+ endif
+endm ;rept _UNROLL_CNT
+;
+ if _UNROLL_CNT ne (ROUNDS_256/8)
+ cmp edi,2*(ROUNDS_256/8)
+ jb Skein_256_round_loop
+ mov edi,[FP_+ctxPtr ] ;restore edi --> context
+ endif
+ ;----------------------------
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..3}
+_NN_ = 0
+ rept WCNT/2
+ if _NN_ ;eax..edx already loaded the first time
+ mov eax,[esp+X_stk + _NN_ + 0]
+ mov ebx,[esp+X_stk + _NN_ + 4]
+ mov ecx,[esp+X_stk + _NN_ + 8]
+ mov edx,[esp+X_stk + _NN_ +12]
+ endif
+ if _NN_ eq 0
+ and dword ptr [edi +TWEAK +12],FIRST_MASK
+ endif
+ xor eax,[esp+Wcopy + _NN_ + 0]
+ xor ebx,[esp+Wcopy + _NN_ + 4]
+ xor ecx,[esp+Wcopy + _NN_ + 8]
+ xor edx,[esp+Wcopy + _NN_ +12]
+ mov [edi+X_VARS+ _NN_ + 0],eax
+ mov [edi+X_VARS+ _NN_ + 4],ebx
+ mov [edi+X_VARS+ _NN_ + 8],ecx
+ mov [edi+X_VARS+ _NN_ +12],edx
+_NN_ = _NN_+16
+ endm
+if _SKEIN_DEBUG
+ Skein_Debug_Round 256,SKEIN_RND_FEED_FWD
+endif
+ ; go back for more blocks, if needed
+ dec dword ptr [FP_+blkCnt]
+ jnz Skein_256_block_loop
+
+ Reset_Stack _Skein_256_Process_Block
+ ret
+_Skein_256_Process_Block endp
+;
+ifdef _SKEIN_CODE_SIZE
+ public _Skein_256_Process_Block_CodeSize
+_Skein_256_Process_Block_CodeSize proc
+ mov eax,_Skein_256_Process_Block_CodeSize - _Skein_256_Process_Block
+ ret
+_Skein_256_Process_Block_CodeSize endp
+;
+ public _Skein_256_Unroll_Cnt
+_Skein_256_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_256/8
+ mov eax,_UNROLL_CNT
+ else
+ xor eax,eax
+ endif
+ ret
+_Skein_256_Unroll_Cnt endp
+endif
+endif ;_USE_ASM_ and 256
+;
+;----------------------------------------------------------------
+;
+if _USE_ASM_ and 512
+ public _Skein_512_Process_Block
+;
+; void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+;;;;;;;;;;;;;;;;;
+; MACRO: four rounds
+;
+R_512_FourRounds macro _RR_,ld_0
+ ; here with edx:ecx = X[1]
+ ;--------- round _RR_
+ ; R512(0,1,2,3,4,5,6,7,R_0, 1);
+ MixStep 512, ld_0, ,0,1,((_RR_)+0),0
+ MixStep 512, 2,3,2,3,((_RR_)+0),1
+ MixStep 512, 4,5,4,5,((_RR_)+0),2
+ MixStep 512, 6,7,6, ,((_RR_)+0),3,1
+
+ ; here with edx:ecx = X[7]
+ ; R512(2,1,4,7,6,5,0,3,R_1, 2);
+ MixStep 512, 4, ,4,7,((_RR_)+1),1
+ MixStep 512, 6,5,6,5,((_RR_)+1),2
+ MixStep 512, 0,3,0,3,((_RR_)+1),3
+ MixStep 512, 2,1,2, ,((_RR_)+1),0,1
+
+ ; here with edx:ecx = X[1]
+ ; R512(4,1,6,3,0,5,2,7,R_2, 3);
+ MixStep 512, 4, ,4,1,((_RR_)+2),0
+ MixStep 512, 6,3,6,3,((_RR_)+2),1
+ MixStep 512, 0,5,0,5,((_RR_)+2),2
+ MixStep 512, 2,7,2, ,((_RR_)+2),3,1
+
+ ; here with edx:ecx = X[7]
+ ; R512(6,1,0,7,2,5,4,3,R_3, 4);
+ MixStep 512, 0, ,0,7,((_RR_)+3),1
+ MixStep 512, 2,5,2,5,((_RR_)+3),2
+ MixStep 512, 4,3,4,3,((_RR_)+3),3
+ MixStep 512, 6,1,6, ,((_RR_)+3),0,1
+
+endm ;R_512_FourRounds
+;
+;;;;;;;;;;;;;;;;;
+; code
+;
+_Skein_512_Process_Block proc near
+ WCNT = 8 ;WCNT=8 for Skein-512
+ Setup_Stack WCNT,(ROUNDS_512/8)
+
+ ; main hash loop for Skein_512
+Skein_512_block_loop:
+ mov eax,[edi+TWEAK+ 0] ;ebx:eax = tweak word T0
+ mov ebx,[edi+TWEAK+ 4]
+ mov ecx,[edi+TWEAK+ 8] ;edx:ecx = tweak word T1
+ mov edx,[edi+TWEAK+12]
+
+ add eax,[FP_+bitAdd ] ;bump T0 by the bitAdd parameter
+ adc ebx, 0
+ mov [edi+TWEAK ],eax ;save updated tweak value T0
+ mov [edi+TWEAK+ 4],ebx
+
+ mov [FP_+ksTwk ],eax ;build the tweak schedule on the stack
+ mov [FP_+ksTwk+ 4],ebx
+ xor eax,ecx ;ebx:eax = T0 ^ T1
+ xor ebx,edx
+ mov [FP_+ksTwk+ 8],ecx
+ mov [FP_+ksTwk+12],edx
+ mov [FP_+ksTwk+16],eax
+ mov [FP_+ksTwk+20],ebx
+
+ mov eax,KW_PARITY_LO ;init parity accumulator
+ mov ebx,KW_PARITY_HI
+;
+_NN_ = 0
+ rept WCNT ;copy in the chaining vars
+ mov ecx,[edi+X_VARS+_NN_ ]
+ mov edx,[edi+X_VARS+_NN_+ 4]
+ xor eax,ecx ;compute overall parity along the way
+ xor ebx,edx
+ mov [FP_+ksKey +_NN_ ],ecx
+ mov [FP_+ksKey +_NN_+ 4],edx
+_NN_ = _NN_+8
+ endm
+;
+ mov [FP_+ksKey +_NN_ ],eax ;save overall parity at the end of the array
+ mov [FP_+ksKey +_NN_+ 4],ebx
+
+ mov esi,[FP_+blkPtr ] ;esi --> input block
+;
+_NN_ = WCNT*8-16 ;work down from the end
+ rept WCNT/2 ;perform initial key injection
+ mov eax,[esi+_NN_ + 0]
+ mov ebx,[esi+_NN_ + 4]
+ mov ecx,[esi+_NN_ + 8]
+ mov edx,[esi+_NN_ +12]
+ mov [esp+_NN_+Wcopy + 0],eax
+ mov [esp+_NN_+Wcopy + 4],ebx
+ mov [esp+_NN_+Wcopy + 8],ecx
+ mov [esp+_NN_+Wcopy +12],edx
+ add eax,[FP_+_NN_+ksKey + 0]
+ adc ebx,[FP_+_NN_+ksKey + 4]
+ add ecx,[FP_+_NN_+ksKey + 8]
+ adc edx,[FP_+_NN_+ksKey +12]
+ if _NN_ eq (WCNT*8-16) ;inject the tweak words
+ add eax,[FP_+ ksTwk + 8]; (at the appropriate points)
+ adc ebx,[FP_+ ksTwk +12]
+ elseif _NN_ eq (WCNT*8-32)
+ add ecx,[FP_+ ksTwk + 0]
+ adc edx,[FP_+ ksTwk + 4]
+ endif
+ if _NN_ or _SKEIN_DEBUG
+ mov [esp+_NN_+X_stk + 0],eax
+ mov [esp+_NN_+X_stk + 4],ebx
+ mov [esp+_NN_+X_stk + 8],ecx
+ mov [esp+_NN_+X_stk +12],edx
+ endif
+_NN_ = _NN_ - 16 ;end at X[0], so regs are already loaded for first MIX!
+ endm
+;
+if _SKEIN_DEBUG ;debug dump of state at this point
+ Skein_Debug_Block WCNT*64
+ Skein_Debug_Round WCNT*64,SKEIN_RND_KEY_INITIAL
+endif
+ add esi, WCNT*8 ;skip the block
+ mov [FP_+blkPtr ],esi ;update block pointer
+ ;
+ ; now the key schedule is computed. Start the rounds
+ ;
+if SKEIN_ASM_UNROLL and 512
+_UNROLL_CNT = ROUNDS_512/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_512
+ if ((ROUNDS_512/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_512"
+ endif
+ xor edi,edi ;edi = round counter
+Skein_512_round_loop:
+endif
+_Rbase_ = 0
+rept _UNROLL_CNT*2
+ ; here with X[0], X[1] already loaded into eax..edx
+ R_512_FourRounds %(4*_Rbase_+00),
+
+ ;inject odd key schedule words
+ if _UNROLL_CNT ne (ROUNDS_512/8)
+ ks_Rotate eax,ebx,WCNT
+ inc edi ;edi = round number
+ endif
+_Rbase_ = _Rbase_+1
+ ks_Inject 512,7,7,eax,ebx,_Rbase_,7, ,_Rbase_
+ ks_Inject 512,6,6,eax,ebx,_Rbase_,6,1
+ ks_Inject 512,5,5,eax,ebx,_Rbase_,5,0
+ ks_Inject 512,4,4,eax,ebx,_Rbase_,4
+ ks_Inject 512,3,3,eax,ebx,_Rbase_,3
+ ks_Inject 512,2,2,eax,ebx,_Rbase_,2
+ ks_Inject 512, , ,ecx,edx,_Rbase_,1
+ ks_Inject 512,0, ,eax,ebx,_Rbase_,0
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT ,saveRegs
+ endif
+endm ;rept _UNROLL_CNT
+;
+if (SKEIN_ASM_UNROLL and 512) eq 0
+ cmp edi,2*(ROUNDS_512/8)
+ jb Skein_512_round_loop
+ mov edi,[FP_+ctxPtr ] ;restore edi --> context
+endif
+ ;----------------------------
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..7}
+_NN_ = 0
+ rept WCNT/2
+ if _NN_ ;eax..edx already loaded the first time
+ mov eax,[esp+X_stk + _NN_ + 0]
+ mov ebx,[esp+X_stk + _NN_ + 4]
+ mov ecx,[esp+X_stk + _NN_ + 8]
+ mov edx,[esp+X_stk + _NN_ +12]
+ endif
+ if _NN_ eq 0
+ and dword ptr [edi + TWEAK+12],FIRST_MASK
+ endif
+ xor eax,[esp+Wcopy + _NN_ + 0]
+ xor ebx,[esp+Wcopy + _NN_ + 4]
+ xor ecx,[esp+Wcopy + _NN_ + 8]
+ xor edx,[esp+Wcopy + _NN_ +12]
+ mov [edi+X_VARS+ _NN_ + 0],eax
+ mov [edi+X_VARS+ _NN_ + 4],ebx
+ mov [edi+X_VARS+ _NN_ + 8],ecx
+ mov [edi+X_VARS+ _NN_ +12],edx
+_NN_ = _NN_+16
+ endm
+if _SKEIN_DEBUG
+ Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+endif
+ ; go back for more blocks, if needed
+ dec dword ptr [FP_+blkCnt]
+ jnz Skein_512_block_loop
+
+ Reset_Stack _Skein_512_Process_Block
+ ret
+_Skein_512_Process_Block endp
+;
+ifdef _SKEIN_CODE_SIZE
+ public _Skein_512_Process_Block_CodeSize
+_Skein_512_Process_Block_CodeSize proc
+ mov eax,_Skein_512_Process_Block_CodeSize - _Skein_512_Process_Block
+ ret
+_Skein_512_Process_Block_CodeSize endp
+;
+ public _Skein_512_Unroll_Cnt
+_Skein_512_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_512/8
+ mov eax,_UNROLL_CNT
+ else
+ xor eax,eax
+ endif
+ ret
+_Skein_512_Unroll_Cnt endp
+endif
+;
+endif ; _USE_ASM_ and 512
+;
+;----------------------------------------------------------------
+;
+if _USE_ASM_ and 1024
+ public _Skein1024_Process_Block
+;
+; void Skein_1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+;;;;;;;;;;;;;;;;;
+; MACRO: four rounds
+;
+R_1024_FourRounds macro _RR_,ld_0
+ ; here with edx:ecx = X[1]
+
+ ;--------- round _RR_
+ MixStep 1024, ld_0, , 0, 1,((_RR_)+0),0
+ MixStep 1024, 2, 3, 2, 3,((_RR_)+0),1
+ MixStep 1024, 4, 5, 4, 5,((_RR_)+0),2
+ MixStep 1024, 6, 7, 6, 7,((_RR_)+0),3
+ MixStep 1024, 8, 9, 8, 9,((_RR_)+0),4
+ MixStep 1024, 10,11,10,11,((_RR_)+0),5
+ MixStep 1024, 12,13,12,13,((_RR_)+0),6
+ MixStep 1024, 14,15,14, ,((_RR_)+0),7,1
+ ; here with edx:ecx = X[15]
+
+ ;--------- round _RR_+1
+ MixStep 1024, 4, , 4,15,((_RR_)+1),3
+ MixStep 1024, 0, 9, 0, 9,((_RR_)+1),0
+ MixStep 1024, 2,13, 2,13,((_RR_)+1),1
+ MixStep 1024, 6,11, 6,11,((_RR_)+1),2
+ MixStep 1024, 10, 7,10, 7,((_RR_)+1),4
+ MixStep 1024, 12, 3,12, 3,((_RR_)+1),5
+ MixStep 1024, 14, 5,14, 5,((_RR_)+1),6
+ MixStep 1024, 8, 1, 8, ,((_RR_)+1),7,1
+ ; here with edx:ecx = X[1]
+
+ ;--------- round _RR_+2
+ MixStep 1024, 6, , 6, 1,((_RR_)+2),3
+ MixStep 1024, 0, 7, 0, 7,((_RR_)+2),0
+ MixStep 1024, 2, 5, 2, 5,((_RR_)+2),1
+ MixStep 1024, 4, 3, 4, 3,((_RR_)+2),2
+ MixStep 1024, 12,15,12,15,((_RR_)+2),4
+ MixStep 1024, 14,13,14,13,((_RR_)+2),5
+ MixStep 1024, 8,11, 8,11,((_RR_)+2),6
+ MixStep 1024, 10, 9,10, ,((_RR_)+2),7,1
+ ; here with edx:ecx = X[9]
+
+ ;--------- round _RR_+3
+ MixStep 1024, 4, , 4, 9,((_RR_)+3),3
+ MixStep 1024, 0,15, 0,15,((_RR_)+3),0
+ MixStep 1024, 2,11, 2,11,((_RR_)+3),1
+ MixStep 1024, 6,13, 6,13,((_RR_)+3),2
+ MixStep 1024, 8, 5, 8, 5,((_RR_)+3),5
+ MixStep 1024, 10, 3,10, 3,((_RR_)+3),6
+ MixStep 1024, 12, 7,12, 7,((_RR_)+3),7
+ MixStep 1024, 14, 1,14, ,((_RR_)+3),4,1
+
+ ; here with edx:ecx = X[1]
+endm ;R_1024_FourRounds
+;
+;;;;;;;;;;;;;;;;;
+; code
+;
+_Skein1024_Process_Block proc near
+;
+ WCNT = 16 ;WCNT=16 for Skein-1024
+ Setup_Stack WCNT,(ROUNDS_1024/8)
+
+ ; main hash loop for Skein1024
+Skein1024_block_loop:
+ mov eax,[edi+TWEAK+ 0] ;ebx:eax = tweak word T0
+ mov ebx,[edi+TWEAK+ 4]
+ mov ecx,[edi+TWEAK+ 8] ;edx:ecx = tweak word T1
+ mov edx,[edi+TWEAK+12]
+
+ add eax,[FP_+bitAdd ] ;bump T0 by the bitAdd parameter
+ adc ebx, 0
+ mov [edi+TWEAK ],eax ;save updated tweak value T0
+ mov [edi+TWEAK+ 4],ebx
+
+ mov [FP_+ksTwk ],eax ;build the tweak schedule on the stack
+ mov [FP_+ksTwk+ 4],ebx
+ xor eax,ecx ;ebx:eax = T0 ^ T1
+ xor ebx,edx
+ mov [FP_+ksTwk+ 8],ecx
+ mov [FP_+ksTwk+12],edx
+ mov [FP_+ksTwk+16],eax
+ mov [FP_+ksTwk+20],ebx
+
+ mov eax,KW_PARITY_LO ;init parity accumulator
+ mov ebx,KW_PARITY_HI
+EDI_BIAS equ 70h ;bias the edi offsets to make them short!
+ add edi, EDI_BIAS
+CT_ equ <edi-EDI_BIAS>
+;
+_NN_ = 0
+ rept WCNT ;copy in the chaining vars
+ mov ecx,[CT_+X_VARS+_NN_ ]
+ mov edx,[CT_+X_VARS+_NN_+ 4]
+ xor eax,ecx ;compute overall parity along the way
+ xor ebx,edx
+ mov [FP_+ksKey +_NN_ ],ecx
+ mov [FP_+ksKey +_NN_+ 4],edx
+_NN_ = _NN_+8
+ endm
+;
+ mov [FP_+ksKey +_NN_ ],eax ;save overall parity at the end of the array
+ mov [FP_+ksKey +_NN_+ 4],ebx
+
+ mov esi,[FP_+blkPtr ] ;esi --> input block
+ lea edi,[esp+Wcopy]
+;
+_NN_ = WCNT*8-16 ;work down from the end
+ rept WCNT/2 ;perform initial key injection
+ mov eax,[esi+_NN_ + 0]
+ mov ebx,[esi+_NN_ + 4]
+ mov ecx,[esi+_NN_ + 8]
+ mov edx,[esi+_NN_ +12]
+ mov [edi+_NN_+ + 0],eax
+ mov [edi+_NN_+ + 4],ebx
+ mov [edi+_NN_+ + 8],ecx
+ mov [edi+_NN_+ +12],edx
+ add eax,[FP_+_NN_+ksKey + 0]
+ adc ebx,[FP_+_NN_+ksKey + 4]
+ add ecx,[FP_+_NN_+ksKey + 8]
+ adc edx,[FP_+_NN_+ksKey +12]
+ if _NN_ eq (WCNT*8-16) ;inject the tweak words
+ add eax,[FP_+ ksTwk + 8]; (at the appropriate points)
+ adc ebx,[FP_+ ksTwk +12]
+ elseif _NN_ eq (WCNT*8-32)
+ add ecx,[FP_+ ksTwk + 0]
+ adc edx,[FP_+ ksTwk + 4]
+ endif
+ if _NN_ or _SKEIN_DEBUG
+ mov [esp+_NN_+X_stk + 0],eax
+ mov [esp+_NN_+X_stk + 4],ebx
+ mov [esp+_NN_+X_stk + 8],ecx
+ mov [esp+_NN_+X_stk +12],edx
+ endif
+_NN_ = _NN_ - 16 ;end at X[0], so regs are already loaded for first MIX!
+ endm
+;
+if _SKEIN_DEBUG ;debug dump of state at this point
+ Skein_Debug_Block WCNT*64
+ Skein_Debug_Round WCNT*64,SKEIN_RND_KEY_INITIAL
+endif
+ sub esi,-WCNT*8 ;skip the block (short immediate)
+ mov [FP_+blkPtr ],esi ;update block pointer
+ ;
+ ; now the key schedule is computed. Start the rounds
+ ;
+if SKEIN_ASM_UNROLL and 1024
+_UNROLL_CNT = ROUNDS_1024/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_1024
+ if ((ROUNDS_1024/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_1024"
+ endif
+ xor edi,edi ;edi = round counter
+Skein_1024_round_loop:
+endif
+
+_Rbase_ = 0
+rept _UNROLL_CNT*2
+ ; here with X[0], X[1] already loaded into eax..edx
+ R_1024_FourRounds %(4*_Rbase_+00),
+
+ ;inject odd key schedule words
+ ;inject odd key schedule words
+ if _UNROLL_CNT ne (ROUNDS_1024/8)
+ ks_Rotate eax,ebx,WCNT
+ inc edi ;edi = round number
+ endif
+_Rbase_ = _Rbase_+1
+ ks_Inject 1024,15,15,eax,ebx,_Rbase_,15, ,_Rbase_
+ ks_Inject 1024,14,14,eax,ebx,_Rbase_,14,1
+ ks_Inject 1024,13,13,eax,ebx,_Rbase_,13,0
+ irp _w,<12,11,10,9,8,7,6,5,4,3,2>
+ ks_Inject 1024,_w,_w,eax,ebx,_Rbase_,_w
+ endm
+ ks_Inject 1024, , ,ecx,edx,_Rbase_,1
+ ks_Inject 1024, 0, ,eax,ebx,_Rbase_,0
+
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT ,saveRegs
+ endif
+endm ;rept _UNROLL_CNT
+;
+if (SKEIN_ASM_UNROLL and 1024) eq 0
+ cmp edi,2*(ROUNDS_1024/8)
+ jb Skein_1024_round_loop
+endif
+ mov edi,[FP_+ctxPtr ] ;restore edi --> context
+ add edi,EDI_BIAS ;and bias it for short offsets below
+ ;----------------------------
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..15}
+ lea esi,[esp+Wcopy] ;use short offsets below
+_NN_ = 0
+ rept WCNT/2
+ if _NN_ ;eax..edx already loaded the first time
+ mov eax,[esp+X_stk + _NN_ + 0]
+ mov ebx,[esp+X_stk + _NN_ + 4]
+ mov ecx,[esp+X_stk + _NN_ + 8]
+ mov edx,[esp+X_stk + _NN_ +12]
+ endif
+ if _NN_ eq 0
+ and dword ptr [CT_ + TWEAK+12],FIRST_MASK
+ endif
+ xor eax,[esi + _NN_ + 0]
+ xor ebx,[esi + _NN_ + 4]
+ xor ecx,[esi + _NN_ + 8]
+ xor edx,[esi + _NN_ +12]
+ mov [CT_+X_VARS+ _NN_ + 0],eax
+ mov [CT_+X_VARS+ _NN_ + 4],ebx
+ mov [CT_+X_VARS+ _NN_ + 8],ecx
+ mov [CT_+X_VARS+ _NN_ +12],edx
+_NN_ = _NN_+16
+ endm
+ sub edi,EDI_BIAS ;undo the bias for return
+
+if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD
+endif
+ ; go back for more blocks, if needed
+ dec dword ptr [FP_+blkCnt]
+ jnz Skein1024_block_loop
+
+ Reset_Stack _Skein1024_Process_Block
+ ret
+_Skein1024_Process_Block endp
+;
+ifdef _SKEIN_CODE_SIZE
+ public _Skein1024_Process_Block_CodeSize
+_Skein1024_Process_Block_CodeSize proc
+ mov eax,_Skein1024_Process_Block_CodeSize - _Skein1024_Process_Block
+ ret
+_Skein1024_Process_Block_CodeSize endp
+;
+ public _Skein1024_Unroll_Cnt
+_Skein1024_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_1024/8
+ mov eax,_UNROLL_CNT
+ else
+ xor eax,eax
+ endif
+ ret
+_Skein1024_Unroll_Cnt endp
+endif
+;
+endif ; _USE_ASM_ and 1024
+;----------------------------------------------------------------
+ end
Index: sys/contrib/skein/asm/skein_block_xmm32.asm
===================================================================
--- /dev/null
+++ sys/contrib/skein/asm/skein_block_xmm32.asm
@@ -0,0 +1,1167 @@
+;
+;----------------------------------------------------------------
+; 32-bit x86 assembler code for Skein block functions using XMM registers
+;
+; Author: Doug Whiting, Hifn
+;
+; This code is released to the public domain.
+;----------------------------------------------------------------
+;
+ .386p
+ .model flat
+ .code
+ .xmm ;enable XMM instructions
+;
+_MASK_ALL_ equ (256+512+1024) ;all three algorithm bits
+;
+;;;;;;;;;;;;;;;;;
+ifndef SKEIN_USE_ASM
+_USE_ASM_ = _MASK_ALL_
+elseif SKEIN_USE_ASM and _MASK_ALL_
+_USE_ASM_ = SKEIN_USE_ASM
+else
+_USE_ASM_ = _MASK_ALL_
+endif
+;
+;;;;;;;;;;;;;;;;;
+ifndef SKEIN_LOOP
+_SKEIN_LOOP = 0 ;default is all fully unrolled
+else
+_SKEIN_LOOP = SKEIN_LOOP
+endif
+;--------------
+; the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256 = (_SKEIN_LOOP / 100) mod 10
+SKEIN_UNROLL_512 = (_SKEIN_LOOP / 10) mod 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP ) mod 10
+;
+SKEIN_ASM_UNROLL = 0
+ irp _NN_,<256,512,1024>
+ if (SKEIN_UNROLL_&_NN_) eq 0
+SKEIN_ASM_UNROLL = SKEIN_ASM_UNROLL + _NN_
+ endif
+ endm
+;
+;;;;;;;;;;;;;;;;;
+;
+ifndef SKEIN_ROUNDS
+ROUNDS_256 = 72
+ROUNDS_512 = 72
+ROUNDS_1024 = 80
+else
+ROUNDS_256 = 8*((((SKEIN_ROUNDS / 100) + 5) mod 10) + 5)
+ROUNDS_512 = 8*((((SKEIN_ROUNDS / 10) + 5) mod 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS ) + 5) mod 10) + 5)
+endif
+irp _NN_,<256,512,1024>
+ if _USE_ASM_ and _NN_
+ irp _RR_,<%(ROUNDS_&_NN_)>
+ if _NN_ eq 1024
+%out +++ SKEIN_ROUNDS_&_NN_ = _RR_
+ else
+%out +++ SKEIN_ROUNDS_&_NN_ = _RR_
+ endif
+ endm
+ endif
+endm
+;;;;;;;;;;;;;;;;;
+;
+ifdef SKEIN_CODE_SIZE
+_SKEIN_CODE_SIZE equ (1)
+else
+ifdef SKEIN_PERF ;use code size if SKEIN_PERF is defined
+_SKEIN_CODE_SIZE equ (1)
+endif
+endif
+;
+;;;;;;;;;;;;;;;;;
+;
+ifndef SKEIN_DEBUG
+_SKEIN_DEBUG = 0
+else
+_SKEIN_DEBUG = 1
+endif
+;;;;;;;;;;;;;;;;;
+;
+; define offsets of fields in hash context structure
+;
+HASH_BITS = 0 ;# bits of hash output
+BCNT = 4 + HASH_BITS ;number of bytes in BUFFER[]
+TWEAK = 4 + BCNT ;tweak values[0..1]
+X_VARS = 16 + TWEAK ;chaining vars
+;
+;(Note: buffer[] in context structure is NOT needed here :-)
+;
+KW_PARITY_LO= 0A9FC1A22h ;overall parity of key schedule words (hi32/lo32)
+KW_PARITY_HI= 01BD11BDAh
+FIRST_MASK8 = NOT (1 SHL 6) ;FIRST block flag bit
+;
+; rotation constants for Skein
+;
+RC_256_0_0 = 14
+RC_256_0_1 = 16
+
+RC_256_1_0 = 52
+RC_256_1_1 = 57
+
+RC_256_2_0 = 23
+RC_256_2_1 = 40
+
+RC_256_3_0 = 5
+RC_256_3_1 = 37
+
+RC_256_4_0 = 25
+RC_256_4_1 = 33
+
+RC_256_5_0 = 46
+RC_256_5_1 = 12
+
+RC_256_6_0 = 58
+RC_256_6_1 = 22
+
+RC_256_7_0 = 32
+RC_256_7_1 = 32
+
+RC_512_0_0 = 46
+RC_512_0_1 = 36
+RC_512_0_2 = 19
+RC_512_0_3 = 37
+
+RC_512_1_0 = 33
+RC_512_1_1 = 27
+RC_512_1_2 = 14
+RC_512_1_3 = 42
+
+RC_512_2_0 = 17
+RC_512_2_1 = 49
+RC_512_2_2 = 36
+RC_512_2_3 = 39
+
+RC_512_3_0 = 44
+RC_512_3_1 = 9
+RC_512_3_2 = 54
+RC_512_3_3 = 56
+
+RC_512_4_0 = 39
+RC_512_4_1 = 30
+RC_512_4_2 = 34
+RC_512_4_3 = 24
+
+RC_512_5_0 = 13
+RC_512_5_1 = 50
+RC_512_5_2 = 10
+RC_512_5_3 = 17
+
+RC_512_6_0 = 25
+RC_512_6_1 = 29
+RC_512_6_2 = 39
+RC_512_6_3 = 43
+
+RC_512_7_0 = 8
+RC_512_7_1 = 35
+RC_512_7_2 = 56
+RC_512_7_3 = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 = 8
+RC_1024_0_3 = 47
+RC_1024_0_4 = 8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 = 4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 = 5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 = 9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 = 4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 = 9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+;
+mov64 macro x0,x1
+ movq x0,x1
+endm
+;
+;----------------------------------------------------------------
+; declare allocated space on the stack
+StackVar macro localName,localSize
+localName = _STK_OFFS_
+_STK_OFFS_ = _STK_OFFS_+(localSize)
+endm ;StackVar
+;
+;----------------------------------------------------------------
+;
+; MACRO: Configure stack frame, allocate local vars
+;
+Setup_Stack macro WCNT,RND_CNT
+_STK_OFFS_ = 0 ;starting offset from esp, forced on 16-byte alignment
+ ;----- local variables ;<-- esp
+ StackVar X_stk , 8*(WCNT) ;local context vars
+ StackVar Wcopy , 8*(WCNT) ;copy of input block
+ StackVar ksTwk ,16*3 ;key schedule: tweak words
+ StackVar ksKey ,16*(WCNT)+16;key schedule: key words
+FRAME_OFFS = ksTwk+128 ;<-- ebp
+ if (SKEIN_ASM_UNROLL and (WCNT*64)) eq 0
+ StackVar ksRot,16*(RND_CNT/4);leave space for ks "rotation" to happen
+ endif
+LOCAL_SIZE = _STK_OFFS_ ;size of local vars
+ ;
+ ;"restart" the stack defns, because we relocate esp to guarantee alignment
+ ; (i.e., these vars are NOT at fixed offsets from esp)
+_STK_OFFS_ = 0
+ ;-----
+ StackVar savRegs,8*4 ;pushad data
+ StackVar retAddr,4 ;return address
+ ;----- caller parameters
+ StackVar ctxPtr ,4 ;context ptr
+ StackVar blkPtr ,4 ;pointer to block data
+ StackVar blkCnt ,4 ;number of full blocks to process
+ StackVar bitAdd ,4 ;bit count to add to tweak
+ ;----- caller's stack frame
+;
+; Notes on stack frame setup:
+; * the most used variable (except for Skein-256) is X_stk[], based at [esp+0]
+; * the next most used is the key schedule words
+; so ebp is "centered" there, allowing short offsets to the key/tweak
+; schedule in 256/512-bit Skein cases, but not posible for Skein-1024 :-(
+; * the Wcopy variables are infrequently accessed, and they have long
+; offsets from both esp and ebp only in the 1024-bit case.
+; * all other local vars and calling parameters can be accessed
+; with short offsets, except in the 1024-bit case
+;
+ pushad ;save all regs
+ mov ebx,esp ;keep ebx as pointer to caller parms
+ sub esp,LOCAL_SIZE ;make room for the locals
+ and esp,not 15 ;force alignment
+ mov edi,[ebx+ctxPtr ] ;edi --> Skein context
+ lea ebp,[esp+FRAME_OFFS] ;maximize use of short offsets from ebp
+ mov ecx,ptr32 [ebx+blkCnt] ;keep block cnt in ecx
+;
+endm ;Setup_Stack
+;
+FP_ equ <ebp-FRAME_OFFS> ;keep as many short offsets as possible
+SI_ equ <esi-FRAME_OFFS> ;keep as many short offsets as possible
+ptr64 equ <qword ptr> ;useful abbreviations
+ptr32 equ <dword ptr>
+ptr08 equ <byte ptr>
+;
+;----------------------------------------------------------------
+;
+Reset_Stack macro procStart
+ mov esp,ebx ;get rid of locals (wipe??)
+ popad ;restore all regs
+
+ ;display code size in bytes to stdout
+ irp _BCNT_,<%($+1-procStart)> ;account for return opcode
+if _BCNT_ ge 10000 ;(align it all pretty)
+%out procStart code size = _BCNT_ bytes
+elseif _BCNT_ ge 1000
+%out procStart code size = _BCNT_ bytes
+else
+%out procStart code size = _BCNT_ bytes
+endif
+ endm ;irp _BCNT_
+
+endm ; Reset_Stack
+;
+;----------------------------------------------------------------
+; macros to help debug internals
+;
+if _SKEIN_DEBUG
+ extrn _Skein_Show_Block:near ;calls to C routines
+ extrn _Skein_Show_Round:near
+;
+SKEIN_RND_SPECIAL = 1000
+SKEIN_RND_KEY_INITIAL = SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT = SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD = SKEIN_RND_SPECIAL+2
+;
+Skein_Debug_Block macro BLK_BITS
+;
+;void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+; const u08b_t *blkPtr, const u64b_t *wPtr,
+; const u64b_t *ksPtr,const u64b_t *tsPtr);
+;
+ Put_XMM_&BLK_BITS
+ pushad ;save all regs
+ lea eax,[FP_+ksTwk+1] ;+1 = flag: "stride" size = 2 qwords
+ lea esi,[FP_+ksKey+1]
+ lea ecx,[esp+32+Wcopy] ;adjust offset by 32 for pushad
+ mov edx,[ebx+ctxPtr] ;ctx_hdr_ptr
+ lea edx,[edx+X_VARS] ;edx ==> cxt->X[]
+ push eax ;tsPtr
+ push esi ;ksPtr
+ push ecx ;wPtr
+ push ptr32 [ebx+blkPtr] ;blkPtr
+ push edx ;ctx->Xptr
+ push ptr32 [ebx+ctxPtr] ;ctx_hdr_ptr
+ mov eax,BLK_BITS
+ push eax ;bits
+ ifdef _MINGW_
+ call _Skein_Show_Block-4 ;strange linkage??
+ else
+ call _Skein_Show_Block
+ endif
+ add esp,7*4 ;discard parameter space on stack
+ popad ;restore regs
+;
+ Get_XMM_&BLK_BITS
+endm ;Skein_Debug_Block
+
+;
+Skein_Debug_Round macro BLK_BITS,R,saveRegs
+;
+;void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X);
+;
+ ifnb <saveRegs>
+ Put_XMM_&BLK_BITS
+ endif
+ pushad ;save all regs
+ if R ne SKEIN_RND_FEED_FWD
+ lea eax,[esp+32+X_stk] ;adjust offset by 32 for pushad
+ else
+ mov eax,[ebx+ctxPtr]
+ add eax,X_VARS
+ endif
+ push eax ;Xptr
+ if (SKEIN_ASM_UNROLL and BLK_BITS) or (R ge SKEIN_RND_SPECIAL)
+ mov eax,R
+ else
+ lea eax,[4*edx+1+(((R)-1) and 3)] ;compute round number using edx
+ endif
+ push eax ;round number
+ push ptr32 [ebx+ctxPtr] ;ctx_hdr_ptr
+ mov eax,BLK_BITS
+ push eax ;bits
+ ifdef _MINGW_
+ call _Skein_Show_Round-4 ;strange linkage??
+ else
+ call _Skein_Show_Round
+ endif
+ add esp,4*4 ;discard parameter space on stack
+ popad ;restore regs
+
+ ifnb <saveRegs>
+ Get_XMM_&BLK_BITS ;save internal vars for debug dump
+ endif
+endm ;Skein_Debug_Round
+endif ;ifdef SKEIN_DEBUG
+;
+;----------------------------------------------------------------
+; useful macros
+_ldX macro xn
+ ifnb <xn>
+ mov64 xmm&xn,ptr64 [esp+X_stk+8*xn]
+ endif
+endm
+
+_stX macro xn
+ ifnb <xn>
+ mov64 ptr64 [esp+X_stk+8*xn],xmm&xn
+ endif
+endm
+;
+;----------------------------------------------------------------
+;
+if _USE_ASM_ and 256
+ public _Skein_256_Process_Block
+;
+; void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+;;;;;;;;;;;;;;;;;
+;
+; Skein-256 round macros
+;
+R_256_OneRound macro _RR_,x0,x1,x2,x3,t0,t1
+ irp _qq_,<%((_RR_) and 7)> ;figure out which rotation constants to use
+ if x0 eq 0
+_RC0_ = RC_256_&_qq_&_0
+_RC1_ = RC_256_&_qq_&_1
+ else
+_RC0_ = RC_256_&_qq_&_1
+_RC1_ = RC_256_&_qq_&_0
+ endif
+ endm
+;
+ paddq xmm&x0,xmm&x1
+ mov64 xmm&t0,xmm&x1
+ psllq xmm&x1, _RC0_
+ psrlq xmm&t0,64-_RC0_
+ xorpd xmm&x1,xmm&x0
+ xorpd xmm&x1,xmm&t0
+;
+ paddq xmm&x2,xmm&x3
+ mov64 xmm&t1,xmm&x3
+ psllq xmm&x3, _RC1_
+ psrlq xmm&t1,64-_RC1_
+ xorpd xmm&x3,xmm&x2
+ xorpd xmm&x3,xmm&t1
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 256,%(_RR_+1),saveRegs
+ endif
+endm ;R_256_OneRound
+;
+R_256_FourRounds macro _RN_
+ R_256_OneRound (_RN_+0),0,1,2,3,4,5
+ R_256_OneRound (_RN_+1),2,1,0,3,4,5
+
+ R_256_OneRound (_RN_+2),0,1,2,3,4,5
+ R_256_OneRound (_RN_+3),2,1,0,3,4,5
+
+ ;inject key schedule
+ inc edx ;bump round number
+ movd xmm4,edx
+ if _UNROLL_CNT eq (ROUNDS_256/8)
+ ;fully unrolled version
+_RK_ = ((_RN_)/4) ;key injection counter
+ paddq xmm0,[FP_+ksKey+16*((_RK_+1) mod 5)]
+ paddq xmm1,[FP_+ksKey+16*((_RK_+2) mod 5)]
+ paddq xmm2,[FP_+ksKey+16*((_RK_+3) mod 5)]
+ paddq xmm3,[FP_+ksKey+16*((_RK_+4) mod 5)]
+ paddq xmm1,[FP_+ksTwk+16*((_RK_+1) mod 3)]
+ paddq xmm2,[FP_+ksTwk+16*((_RK_+2) mod 3)]
+ paddq xmm3,xmm4
+ else ;looping version
+ paddq xmm0,[SI_+ksKey+16*1]
+ paddq xmm1,[SI_+ksKey+16*2]
+ paddq xmm2,[SI_+ksKey+16*3]
+ paddq xmm3,[SI_+ksKey+16*4]
+ paddq xmm1,[SI_+ksTwk+16*1]
+ paddq xmm2,[SI_+ksTwk+16*2]
+ paddq xmm3,xmm4
+;
+ mov64 xmm4,<ptr64 [SI_+ksKey]>;first, "rotate" key schedule on the stack
+ mov64 xmm5,<ptr64 [SI_+ksTwk]>; (for next time through)
+ mov64 <ptr64 [SI_+ksKey+16*(WCNT+1)]>,xmm4
+ mov64 <ptr64 [SI_+ksTwk+16*3]>,xmm5
+ add esi,16 ;bump rolling pointer
+ endif
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT,saveRegs
+ endif
+endm ;R256_FourRounds
+;
+if _SKEIN_DEBUG ; macros for saving/restoring X_stk for debug routines
+Put_XMM_256 equ <call _Put_XMM_256>
+Get_XMM_256 equ <call _Get_XMM_256>
+
+_Put_XMM_256:
+ irp _NN_,<0,1,2,3>
+ mov64 ptr64 [esp+X_stk+4+_NN_*8],xmm&_NN_
+ endm
+ ret
+;
+_Get_XMM_256:
+ irp _NN_,<0,1,2,3>
+ mov64 xmm&_NN_,ptr64 [esp+X_stk+4+_NN_*8]
+ endm
+ ret
+endif
+;
+;;;;;;;;;;;;;;;;;
+;
+; code
+;
+_Skein_256_Process_Block proc near
+ WCNT = 4 ;WCNT=4 for Skein-256
+ Setup_Stack WCNT,ROUNDS_256
+ ; main hash loop for Skein_256
+Skein_256_block_loop:
+ movd xmm4,ptr32 [ebx+bitAdd]
+ mov64 xmm5,ptr64 [edi+TWEAK+0]
+ mov64 xmm6,ptr64 [edi+TWEAK+8]
+ paddq xmm5,xmm4 ;bump T0 by the bitAdd parameter
+ mov64 ptr64 [edi+TWEAK],xmm5 ;save updated tweak value T0 (for next time)
+ movapd xmm7,xmm6
+ xorpd xmm7,xmm5 ;compute overall tweak parity
+ movdqa [FP_+ksTwk ],xmm5 ;save the expanded tweak schedule on the stack
+ movdqa [FP_+ksTwk+16],xmm6
+ movdqa [FP_+ksTwk+32],xmm7
+
+ mov esi,[ebx+blkPtr] ;esi --> input block
+ mov eax,KW_PARITY_LO ;init key schedule parity accumulator
+ mov edx,KW_PARITY_HI
+ movd xmm4,eax
+ movd xmm0,edx
+ unpcklps xmm4,xmm0 ;pack two 32-bit words into xmm4
+;
+ irp _NN_,<0,1,2,3> ;copy in the chaining vars
+ mov64 xmm&_NN_,ptr64 [edi+X_VARS+8*_NN_]
+ xorpd xmm4,xmm&_NN_ ;update overall parity
+ movdqa [FP_+ksKey+16*_NN_],xmm&_NN_
+ endm
+ movdqa [FP_+ksKey+16*WCNT],xmm4;save overall parity at the end of the array
+;
+ paddq xmm1,xmm5 ;inject the initial tweak words
+ paddq xmm2,xmm6
+;
+ irp _NN_,<0,1,2,3> ;perform the initial key injection
+ mov64 xmm4,ptr64 [esi+8*_NN_] ;and save a copy of the input block on stack
+ mov64 ptr64 [esp+8*_NN_+Wcopy],xmm4
+ paddq xmm&_NN_,xmm4
+ endm
+;
+if _SKEIN_DEBUG ;debug dump of state at this point
+ Skein_Debug_Block 256
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL,saveRegs
+endif
+ add esi, WCNT*8 ;skip to the next block
+ mov [ebx+blkPtr ],esi ;save the updated block pointer
+ ;
+ ; now the key schedule is computed. Start the rounds
+ ;
+ xor edx,edx ;edx = iteration count
+if SKEIN_ASM_UNROLL and 256
+_UNROLL_CNT = ROUNDS_256/8 ;fully unrolled
+else
+_UNROLL_CNT = SKEIN_UNROLL_256 ;partial unroll count
+ if ((ROUNDS_256/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_256" ;sanity check
+ endif
+ mov esi,ebp ;use this as "rolling" pointer into ksTwk/ksKey
+Skein_256_round_loop: ; (since there's no 16* scaled address mode)
+endif
+;
+_Rbase_ = 0
+rept _UNROLL_CNT*2 ; here with X[0..3] in XMM0..XMM3
+ R_256_FourRounds _Rbase_
+_Rbase_ = _Rbase_+4
+endm ;rept _UNROLL_CNT*2
+;
+ if _UNROLL_CNT ne (ROUNDS_256/8)
+ cmp edx,2*(ROUNDS_256/8)
+ jb Skein_256_round_loop
+ endif
+ ;----------------------------
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..3}
+ irp _NN_,<0,1,2,3>
+ mov64 xmm4,ptr64 [esp+Wcopy+8*_NN_]
+ xorpd xmm&_NN_,xmm4
+ mov64 ptr64 [edi+X_VARS+8*_NN_],xmm&_NN_
+ endm
+ and ptr08 [edi +TWEAK +15],FIRST_MASK8
+if _SKEIN_DEBUG
+ Skein_Debug_Round 256,SKEIN_RND_FEED_FWD,saveRegs
+endif
+ ; go back for more blocks, if needed
+ dec ecx
+ jnz Skein_256_block_loop
+
+ Reset_Stack _Skein_256_Process_Block
+ ret
+;
+_Skein_256_Process_Block endp
+;
+ifdef _SKEIN_CODE_SIZE
+ public _Skein_256_Process_Block_CodeSize
+_Skein_256_Process_Block_CodeSize proc
+ mov eax,_Skein_256_Process_Block_CodeSize - _Skein_256_Process_Block
+ ret
+_Skein_256_Process_Block_CodeSize endp
+;
+ public _Skein_256_Unroll_Cnt
+_Skein_256_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_256/8
+ mov eax,_UNROLL_CNT
+ else
+ xor eax,eax
+ endif
+ ret
+_Skein_256_Unroll_Cnt endp
+endif
+endif ;_USE_ASM_ and 256
+;
+;----------------------------------------------------------------
+;
+if _USE_ASM_ and 512
+ public _Skein_512_Process_Block
+;
+; void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+;;;;;;;;;;;;;;;;;
+; MACRO: one round
+;
+R_512_Round macro _RR_, a0,a1,Ra, b0,b1,Rb, c0,c1,Rc, d0,d1,Rd
+irp _nr_,<%((_RR_) and 7)>
+_Ra_ = RC_512_&_nr_&_&Ra
+_Rb_ = RC_512_&_nr_&_&Rb
+_Rc_ = RC_512_&_nr_&_&Rc
+_Rd_ = RC_512_&_nr_&_&Rd
+endm
+ paddq xmm&a0,xmm&a1
+ _stX c0
+ mov64 xmm&c0,xmm&a1
+ psllq xmm&a1, _Ra_
+ psrlq xmm&c0,64-_Ra_
+ xorpd xmm&a1,xmm&c0
+ xorpd xmm&a1,xmm&a0
+
+ paddq xmm&b0,xmm&b1
+ _stX a0
+ mov64 xmm&a0,xmm&b1
+ psllq xmm&b1, _Rb_
+ psrlq xmm&a0,64-_Rb_
+ xorpd xmm&b1,xmm&b0
+ _ldX c0
+ xorpd xmm&b1,xmm&a0
+
+ paddq xmm&c0,xmm&c1
+ mov64 xmm&a0,xmm&c1
+ psllq xmm&c1, _Rc_
+ psrlq xmm&a0,64-_Rc_
+ xorpd xmm&c1,xmm&c0
+ xorpd xmm&c1,xmm&a0
+
+ paddq xmm&d0,xmm&d1
+ mov64 xmm&a0,xmm&d1
+ psllq xmm&d1, _Rd_
+ psrlq xmm&a0,64-_Rd_
+ xorpd xmm&d1,xmm&a0
+ _ldX a0
+ xorpd xmm&d1,xmm&d0
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 512,%(_RR_+1),saveRegs
+ endif
+endm
+;
+; MACRO: four rounds
+R_512_FourRounds macro _RN_
+ R_512_Round (_RN_) , 0,1,0, 2,3,1, 4,5,2, 6,7,3
+ R_512_Round (_RN_)+1, 2,1,0, 4,7,1, 6,5,2, 0,3,3
+ R_512_Round (_RN_)+2, 4,1,0, 6,3,1, 0,5,2, 2,7,3
+ R_512_Round (_RN_)+3, 6,1,0, 0,7,1, 2,5,2, 4,3,3
+
+ ;inject key schedule
+ irp _NN_,<0,1,2,3,4,5,6,7>
+ if _UNROLL_CNT eq (ROUNDS_512/8)
+ paddq xmm&_NN_,[FP_+ksKey+16*((((_RN_)/4)+(_NN_)+1) mod 9)]
+ else
+ paddq xmm&_NN_,[SI_+ksKey+16*((_NN_)+1)]
+ endif
+ endm
+ _stX 0 ;free up a register
+ inc edx ;bump round counter
+ movd xmm0,edx ;inject the tweak
+ if _UNROLL_CNT eq (ROUNDS_512/8)
+ paddq xmm5,[FP_+ksTwk+16*(((_RN_)+1) mod 3)]
+ paddq xmm6,[FP_+ksTwk+16*(((_RN_)+2) mod 3)]
+ paddq xmm7,xmm0
+ else ;looping version
+ paddq xmm5,[SI_+ksTwk+16*1]
+ paddq xmm6,[SI_+ksTwk+16*2]
+ paddq xmm7,xmm0
+;
+ mov64 xmm0,<ptr64 [SI_+ksKey]>;first, "rotate" key schedule on the stack
+ mov64 <ptr64 [SI_+ksKey+16*(WCNT+1)]>,xmm0
+ mov64 xmm0,<ptr64 [SI_+ksTwk]>; (for next time through)
+ mov64 <ptr64 [SI_+ksTwk+16*3]>,xmm0
+ add esi,16 ;bump rolling pointer
+ endif
+ _ldX 0 ;restore X0
+ if _SKEIN_DEBUG
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT,saveRegs
+ endif
+endm ;R_512_FourRounds
+;;;;;;;;;;;;;;;;;
+if _SKEIN_DEBUG ; macros for saving/restoring X_stk for debug routines
+Put_XMM_512 equ <call _Put_XMM_512>
+Get_XMM_512 equ <call _Get_XMM_512>
+
+_Put_XMM_512:
+ irp _NN_,<0,1,2,3,4,5,6,7>
+ mov64 ptr64 [esp+X_stk+4+_NN_*8],xmm&_NN_
+ endm
+ ret
+;
+_Get_XMM_512:
+ irp _NN_,<0,1,2,3,4,5,6,7>
+ mov64 xmm&_NN_,ptr64 [esp+X_stk+4+_NN_*8]
+ endm
+ ret
+endif
+;
+;;;;;;;;;;;;;;;;;
+; code
+;
+_Skein_512_Process_Block proc near
+ WCNT = 8 ;WCNT=8 for Skein-512
+ Setup_Stack WCNT,ROUNDS_512
+ ; main hash loop for Skein_512
+Skein_512_block_loop:
+ movd xmm0,ptr32 [ebx+bitAdd]
+ mov64 xmm1,ptr64 [edi+TWEAK+0]
+ mov64 xmm2,ptr64 [edi+TWEAK+8]
+ paddq xmm1,xmm0 ;bump T0 by the bitAdd parameter
+ mov64 ptr64 [edi+TWEAK],xmm1 ;save updated tweak value T0 (for next time)
+ mov64 xmm0,xmm2
+ xorpd xmm0,xmm1 ;compute overall tweak parity
+ movdqa [FP_+ksTwk ],xmm1 ;save the expanded tweak schedule on the stack
+ movdqa [FP_+ksTwk+16*1],xmm2
+ movdqa [FP_+ksTwk+16*2],xmm0
+
+ mov esi,[ebx+blkPtr] ;esi --> input block
+ mov eax,KW_PARITY_LO ;init key schedule parity accumulator
+ mov edx,KW_PARITY_HI
+ movd xmm0,eax
+ movd xmm7,edx
+ unpcklps xmm0,xmm7 ;pack two 32-bit words into xmm0
+;
+ irp _NN_,<7,6,5,4,3,2,1> ;copy in the chaining vars (skip #0 for now)
+ mov64 xmm&_NN_,ptr64 [edi+X_VARS+8*_NN_]
+ xorpd xmm0,xmm&_NN_ ;update overall parity
+ movdqa [FP_+ksKey+16*_NN_],xmm&_NN_
+ if _NN_ eq 5
+ paddq xmm5,xmm1 ;inject the initial tweak words
+ paddq xmm6,xmm2 ; (before they get trashed in xmm1/2)
+ endif
+ endm
+ mov64 xmm4,ptr64 [edi+X_VARS] ;handle #0 now
+ xorpd xmm0,xmm4 ;update overall parity
+ movdqa [FP_+ksKey+16* 0 ],xmm4;save the key value in slot #0
+ movdqa [FP_+ksKey+16*WCNT],xmm0;save overall parity at the end of the array
+;
+ mov64 xmm0,xmm4
+ irp _NN_,<7,6,5, 4,3,2,1,0> ;perform the initial key injection (except #4)
+ mov64 xmm4,ptr64 [esi+ 8*_NN_];and save a copy of the input block on stack
+ mov64 ptr64 [esp+ 8*_NN_+Wcopy],xmm4
+ paddq xmm&_NN_,xmm4
+ endm
+ mov64 xmm4,ptr64 [esi+ 8*4] ;get input block word #4
+ mov64 ptr64 [esp+ 8*4+Wcopy],xmm4
+ paddq xmm4,[FP_+ksKey+16*4] ;inject the initial key
+;
+if _SKEIN_DEBUG ;debug dump of state at this point
+ Skein_Debug_Block 512
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL,saveRegs
+endif
+ add esi, WCNT*8 ;skip to the next block
+ mov [ebx+blkPtr],esi ;save the updated block pointer
+ ;
+ ; now the key schedule is computed. Start the rounds
+ ;
+ xor edx,edx ;edx = round counter
+if SKEIN_ASM_UNROLL and 512
+_UNROLL_CNT = ROUNDS_512/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_512
+ if ((ROUNDS_512/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_512"
+ endif
+ mov esi,ebp ;use this as "rolling" pointer into ksTwk/ksKey
+Skein_512_round_loop: ; (since there's no 16* scaled address mode)
+endif
+_Rbase_ = 0
+rept _UNROLL_CNT*2
+ R_512_FourRounds _Rbase_
+_Rbase_ = _Rbase_+4
+endm ;rept _UNROLL_CNT
+;
+if (SKEIN_ASM_UNROLL and 512) eq 0
+ cmp edx,2*(ROUNDS_512/8)
+ jb Skein_512_round_loop
+endif
+ ;----------------------------
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..7}
+ and ptr08 [edi +TWEAK +15],FIRST_MASK8
+irp _NN_,<0,2,4,6> ;do the aligned ones first
+ xorpd xmm&_NN_,[esp+Wcopy+8*_NN_]
+ mov64 ptr64 [edi+X_VARS+8*_NN_],xmm&_NN_
+endm
+irp _NN_,<1,3,5,7> ;now we have some register space available
+ mov64 xmm0,ptr64 [esp+Wcopy+8*_NN_]
+ xorpd xmm&_NN_,xmm0
+ mov64 ptr64 [edi+X_VARS+8*_NN_],xmm&_NN_
+endm
+if _SKEIN_DEBUG
+ Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+endif
+ ; go back for more blocks, if needed
+ dec ecx
+ jnz Skein_512_block_loop
+
+ Reset_Stack _Skein_512_Process_Block
+ ret
+_Skein_512_Process_Block endp
+;
+ifdef _SKEIN_CODE_SIZE
+ public _Skein_512_Process_Block_CodeSize
+_Skein_512_Process_Block_CodeSize proc
+ mov eax,_Skein_512_Process_Block_CodeSize - _Skein_512_Process_Block
+ ret
+_Skein_512_Process_Block_CodeSize endp
+;
+ public _Skein_512_Unroll_Cnt
+_Skein_512_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_512/8
+ mov eax,_UNROLL_CNT
+ else
+ xor eax,eax
+ endif
+ ret
+_Skein_512_Unroll_Cnt endp
+endif
+;
+endif ; _USE_ASM_ and 512
+;
+;----------------------------------------------------------------
+;
+if _USE_ASM_ and 1024
+ public _Skein1024_Process_Block
+;
+; void Skein_1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd);
+;
+R_1024_REGS equ (5) ;keep this many block variables in registers
+;
+;;;;;;;;;;;;;;;;
+if _SKEIN_DEBUG ; macros for saving/restoring X_stk for debug routines
+Put_XMM_1024 equ <call _Put_XMM_1024>
+Get_XMM_1024 equ <call _Get_XMM_1024>
+
+_Put_XMM_1024:
+_NN_ = 0
+ rept R_1024_REGS
+ irp _rr_,<%(_NN_)>
+ mov64 ptr64 [esp+X_stk+4+8*_NN_],xmm&_rr_
+ endm
+_NN_ = _NN_+1
+ endm
+ ret
+;
+_Get_XMM_1024:
+_NN_ = 0
+ rept R_1024_REGS
+ irp _rr_,<%(_NN_)>
+ mov64 xmm&_rr_,ptr64 [esp+X_stk+4+8*_NN_]
+ endm
+_NN_ = _NN_+1
+ endm
+ ret
+endif
+;
+;;;;;;;;;;;;;;;;;
+; MACRO: one mix step
+MixStep_1024 macro x0,x1,rotIdx0,rotIdx1,_debug_
+_r0_ = x0 ;default, if already loaded
+_r1_ = x1
+ ; load the regs (if necessary)
+ if (x0 ge R_1024_REGS)
+_r0_ = 5
+ mov64 xmm5,ptr64 [esp+X_stk+8*(x0)]
+ endif
+ if (x1 ge R_1024_REGS)
+_r1_ = 6
+ mov64 xmm6,ptr64 [esp+X_stk+8*(x1)]
+ endif
+ ; do the mix
+ irp _rx_,<%((rotIdx0) and 7)>
+_Rc_ = RC_1024_&_rx_&_&rotIdx1 ;rotation constant
+ endm
+ irp _x0_,<%_r0_>
+ irp _x1_,<%_r1_>
+ paddq xmm&_x0_,xmm&_x1_
+ mov64 xmm7 ,xmm&_x1_
+ psllq xmm&_x1_, _Rc_
+ psrlq xmm7 ,64-_Rc_
+ xorpd xmm&_x1_,xmm&_x0_
+ xorpd xmm&_x1_,xmm7
+ endm
+ endm
+ ; save the regs (if necessary)
+ if (x0 ge R_1024_REGS)
+ mov64 ptr64 [esp+X_stk+8*(x0)],xmm5
+ endif
+ if (x1 ge R_1024_REGS)
+ mov64 ptr64 [esp+X_stk+8*(x1)],xmm6
+ endif
+ ; debug output
+ if _SKEIN_DEBUG and (0 ne (_debug_ + 0))
+ Skein_Debug_Round 1024,%((RotIdx0)+1),saveRegs
+ endif
+endm
+;;;;;;;;;;;;;;;;;
+; MACRO: four rounds
+;
+R_1024_FourRounds macro _RR_
+ ;--------- round _RR_
+ MixStep_1024 0, 1,%((_RR_)+0),0
+ MixStep_1024 2, 3,%((_RR_)+0),1
+ MixStep_1024 4, 5,%((_RR_)+0),2
+ MixStep_1024 6, 7,%((_RR_)+0),3
+ MixStep_1024 8, 9,%((_RR_)+0),4
+ MixStep_1024 10,11,%((_RR_)+0),5
+ MixStep_1024 12,13,%((_RR_)+0),6
+ MixStep_1024 14,15,%((_RR_)+0),7,1
+ ;--------- round _RR_+1
+ MixStep_1024 0, 9,%((_RR_)+1),0
+ MixStep_1024 2,13,%((_RR_)+1),1
+ MixStep_1024 6,11,%((_RR_)+1),2
+ MixStep_1024 4,15,%((_RR_)+1),3
+ MixStep_1024 10, 7,%((_RR_)+1),4
+ MixStep_1024 12, 3,%((_RR_)+1),5
+ MixStep_1024 14, 5,%((_RR_)+1),6
+ MixStep_1024 8, 1,%((_RR_)+1),7,1
+ ;--------- round _RR_+2
+ MixStep_1024 0, 7,%((_RR_)+2),0
+ MixStep_1024 2, 5,%((_RR_)+2),1
+ MixStep_1024 4, 3,%((_RR_)+2),2
+ MixStep_1024 6, 1,%((_RR_)+2),3
+ MixStep_1024 12,15,%((_RR_)+2),4
+ MixStep_1024 14,13,%((_RR_)+2),5
+ MixStep_1024 8,11,%((_RR_)+2),6
+ MixStep_1024 10, 9,%((_RR_)+2),7,1
+ ;--------- round _RR_+3
+ MixStep_1024 0,15,%((_RR_)+3),0
+ MixStep_1024 2,11,%((_RR_)+3),1
+ MixStep_1024 6,13,%((_RR_)+3),2
+ MixStep_1024 4, 9,%((_RR_)+3),3
+ MixStep_1024 14, 1,%((_RR_)+3),4
+ MixStep_1024 8, 5,%((_RR_)+3),5
+ MixStep_1024 10, 3,%((_RR_)+3),6
+ MixStep_1024 12, 7,%((_RR_)+3),7,1
+
+ inc edx ;edx = round number
+ movd xmm7,edx
+ ;inject the key
+irp _NN_,<15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0>
+ if _UNROLL_CNT ne (ROUNDS_1024/8)
+ if _NN_ lt R_1024_REGS
+ paddq xmm&_NN_,ptr64 [SI_+ksKey+16*_NN_+16]
+ else
+ mov64 xmm6 ,ptr64 [esp+X_stk+ 8*_NN_]
+ if _NN_ eq 15
+ paddq xmm6,xmm7
+ elseif _NN_ eq 14
+ paddq xmm6,ptr64 [SI_+ksTwk+16*2]
+ elseif _NN_ eq 13
+ paddq xmm6,ptr64 [SI_+ksTwk+16*1]
+ endif
+ paddq xmm6 ,ptr64 [SI_+ksKey+16*_NN_+16]
+ mov64 ptr64 [esp+X_stk+ 8*_NN_],xmm6
+ endif
+ else
+ if _NN_ lt R_1024_REGS
+ paddq xmm&_NN_,ptr64 [FP_+ksKey+16*(((_Rbase_/4)+(_NN_)+1) mod 17)]
+ else
+ mov64 xmm6,ptr64 [esp+X_stk+ 8*_NN_]
+ paddq xmm6,ptr64 [FP_+ksKey+16*(((_Rbase_/4)+(_NN_)+1) mod 17)]
+ if _NN_ eq 15
+ paddq xmm6,xmm7
+ elseif _NN_ eq 14
+ paddq xmm6,ptr64 [FP_+ksTwk+16*(((_Rbase_/4)+2) mod 3)]
+ elseif _NN_ eq 13
+ paddq xmm6,ptr64 [FP_+ksTwk+16*(((_Rbase_/4)+1) mod 3)]
+ endif
+ mov64 ptr64 [esp+X_stk+ 8*_NN_],xmm6
+ endif
+ endif
+endm
+if _UNROLL_CNT ne (ROUNDS_1024/8) ;rotate the key schedule on the stack
+ mov64 xmm6,ptr64 [SI_+ksKey]
+ mov64 xmm7,ptr64 [SI_+ksTwk]
+ mov64 ptr64 [SI_+ksKey+16*(WCNT+1)],xmm6
+ mov64 ptr64 [SI_+ksTwk+16* 3 ],xmm7
+ add esi,16 ;bump rolling pointer
+endif
+if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT ,saveRegs
+endif
+endm ;R_1024_FourRounds
+;;;;;;;;;;;;;;;;
+; code
+;
+_Skein1024_Process_Block proc near
+;
+ WCNT = 16 ;WCNT=16 for Skein-1024
+ Setup_Stack WCNT,ROUNDS_1024
+ add edi,80h ;bias the edi ctxt offsets to keep them all short
+ctx equ <edi-80h> ;offset alias
+ ; main hash loop for Skein1024
+Skein1024_block_loop:
+ movd xmm0,ptr32 [ebx+bitAdd]
+ mov64 xmm1,ptr64 [ctx+TWEAK+0]
+ mov64 xmm2,ptr64 [ctx+TWEAK+8]
+ paddq xmm1,xmm0 ;bump T0 by the bitAdd parameter
+ mov64 ptr64 [ctx+TWEAK],xmm1 ;save updated tweak value T0 (for next time)
+ mov64 xmm0,xmm2
+ xorpd xmm0,xmm1 ;compute overall tweak parity
+ movdqa [FP_+ksTwk ],xmm1 ;save the expanded tweak schedule on the stack
+ movdqa [FP_+ksTwk+16],xmm2
+ movdqa [FP_+ksTwk+32],xmm0
+
+ mov esi,[ebx+blkPtr] ;esi --> input block
+ mov eax,KW_PARITY_LO ;init key schedule parity accumulator
+ mov edx,KW_PARITY_HI
+ movd xmm7,eax
+ movd xmm6,edx
+ unpcklps xmm7,xmm6 ;pack two 32-bit words into xmm7
+;
+ lea eax,[esp+80h] ;use short offsets for Wcopy, X_stk writes below
+SP_ equ <eax-80h> ;[eax+OFFS] mode is one byte shorter than [esp+OFFS]
+irp _NN_,<15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0>
+ mov64 xmm6,ptr64 [ctx+X_VARS+8*_NN_]
+ xorpd xmm7,xmm6 ;update overall parity
+ movdqa [FP_+ksKey+16*_NN_],xmm6;save the key schedule on the stack
+ if _NN_ lt R_1024_REGS
+ _rr_ = _NN_
+ else
+ _rr_ = R_1024_REGS
+ endif
+ irp _rn_,<%(_rr_)>
+ mov64 xmm&_rn_,ptr64 [esi+ 8*_NN_];save copy of the input block on stack
+ mov64 ptr64 [SP_+ Wcopy + 8*_NN_],xmm&_rn_ ;(for feedforward later)
+ paddq xmm&_rn_,xmm6 ;inject the key into the block
+ if _NN_ eq 13
+ paddq xmm&_rn_,xmm1 ;inject the initial tweak words
+ elseif _NN_ eq 14
+ paddq xmm&_rn_,xmm2
+ endif
+ if _NN_ ge R_1024_REGS ;only save X[5..15] on stack, leave X[0..4] in regs
+ mov64 ptr64 [SP_+X_stk+8*_NN_],xmm&_rn_
+ endif
+ endm
+endm
+ movdqa [FP_+ksKey+16*WCNT],xmm7;save overall key parity at the end of the array
+;
+if _SKEIN_DEBUG ;debug dump of state at this point
+ Skein_Debug_Block 1024
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL,saveRegs
+endif
+ add esi, WCNT*8 ;skip to the next block
+ mov [ebx+blkPtr],esi ;save the updated block pointer
+ ;
+ ; now the key schedule is computed. Start the rounds
+ ;
+ xor edx,edx ;edx = round counter
+if SKEIN_ASM_UNROLL and 1024
+_UNROLL_CNT = ROUNDS_1024/8
+else
+_UNROLL_CNT = SKEIN_UNROLL_1024
+ if ((ROUNDS_1024/8) mod _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_1024"
+ endif
+ mov esi,ebp ;use this as "rolling" pointer into ksTwk/ksKey
+Skein_1024_round_loop:
+endif
+;
+_Rbase_ = 0
+rept _UNROLL_CNT*2
+ R_1024_FourRounds %_Rbase_
+_Rbase_ = _Rbase_+4
+endm ;rept _UNROLL_CNT
+;
+if (SKEIN_ASM_UNROLL and 1024) eq 0
+ cmp edx,2*(ROUNDS_1024/8)
+ jb Skein_1024_round_loop
+endif
+ and ptr08 [ctx +TWEAK +15],FIRST_MASK8 ;clear tweak bit for next time thru
+ ;----------------------------
+ ; feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..15}
+ lea eax,[esp+80h] ;allow short offsets to X_stk and Wcopy
+irp _NN_,<0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15>
+ if _NN_ lt R_1024_REGS
+ if _NN_ and 1 ;already in regs: no load needed
+ mov64 xmm7 ,ptr64 [SP_+ Wcopy + 8*_NN_] ;unaligned
+ xorpd xmm&_NN_,xmm7
+ else
+ xorpd xmm&_NN_, [SP_+ Wcopy + 8*_NN_] ;aligned
+ endif
+ mov64 ptr64 [ctx+ X_vars+ 8*_NN_],xmm&_NN_
+ else
+ mov64 xmm7,ptr64 [SP_+ X_stk + 8*_NN_] ;load X value from stack
+ if _NN_ and 1
+ mov64 xmm6,ptr64 [SP_+ Wcopy + 8*_NN_] ;unaligned
+ xorpd xmm7,xmm6
+ else
+ xorpd xmm7, [SP_+ Wcopy + 8*_NN_] ;aligned
+ endif
+ mov64 ptr64 [ctx+ X_vars+ 8*_NN_],xmm7
+ endif
+endm
+if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD ;no need to save regs on stack here
+endif
+ ; go back for more blocks, if needed
+ dec ecx
+ jnz Skein1024_block_loop
+
+ Reset_Stack _Skein1024_Process_Block
+ ret
+_Skein1024_Process_Block endp
+;
+ifdef _SKEIN_CODE_SIZE
+ public _Skein1024_Process_Block_CodeSize
+_Skein1024_Process_Block_CodeSize proc
+ mov eax,_Skein1024_Process_Block_CodeSize - _Skein1024_Process_Block
+ ret
+_Skein1024_Process_Block_CodeSize endp
+;
+ public _Skein1024_Unroll_Cnt
+_Skein1024_Unroll_Cnt proc
+ if _UNROLL_CNT ne ROUNDS_1024/8
+ mov eax,_UNROLL_CNT
+ else
+ xor eax,eax
+ endif
+ ret
+_Skein1024_Unroll_Cnt endp
+endif
+;
+endif ; _USE_ASM_ and 1024
+;----------------------------------------------------------------
+ end
Index: sys/contrib/skein/asm/skein_block_xmm32.s
===================================================================
--- /dev/null
+++ sys/contrib/skein/asm/skein_block_xmm32.s
@@ -0,0 +1,1110 @@
+#
+#----------------------------------------------------------------
+# 32-bit x86 assembler code for Skein block functions using XMM registers
+#
+# Author: Doug Whiting, Hifn/Exar
+#
+# This code is released to the public domain.
+#----------------------------------------------------------------
+#
+ .text
+ .altmacro #use advanced macro features
+ .psize 0,128 #list file has no page boundaries
+#
+_MASK_ALL_ = (256+512+1024) #all three algorithm bits
+SAVE_REGS = 1
+#
+#################
+.ifndef SKEIN_USE_ASM
+_USE_ASM_ = _MASK_ALL_
+.elseif SKEIN_USE_ASM & _MASK_ALL_
+_USE_ASM_ = SKEIN_USE_ASM
+.else
+_USE_ASM_ = _MASK_ALL_
+.endif
+#
+#################
+.ifndef SKEIN_LOOP
+_SKEIN_LOOP = 002 #default is all fully unrolled, except Skein1024
+.else
+_SKEIN_LOOP = SKEIN_LOOP
+.endif
+#--------------
+# the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256 = (_SKEIN_LOOP / 100) % 10
+SKEIN_UNROLL_512 = (_SKEIN_LOOP / 10) % 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP ) % 10
+#
+SKEIN_ASM_UNROLL = 0
+ .irp _NN_,256,512,1024
+ .if (SKEIN_UNROLL_\_NN_) == 0
+SKEIN_ASM_UNROLL = SKEIN_ASM_UNROLL + \_NN_
+ .endif
+ .endr
+#
+#################
+#
+.ifndef SKEIN_ROUNDS
+ROUNDS_256 = 72
+ROUNDS_512 = 72
+ROUNDS_1024 = 80
+.else
+ROUNDS_256 = 8*((((SKEIN_ROUNDS / 100) + 5) % 10) + 5)
+ROUNDS_512 = 8*((((SKEIN_ROUNDS / 10) + 5) % 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS ) + 5) % 10) + 5)
+.irp _NN_,256,512,1024
+ .if _USE_ASM_ && \_NN_
+ .irp _RR_,%(ROUNDS_\_NN_)
+ .if \_NN_ < 1024
+.print "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+ .else
+.print "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+ .endif
+ .endr
+ .endif
+.endr
+.endif
+#################
+#
+.ifdef SKEIN_CODE_SIZE
+_SKEIN_CODE_SIZE = (1)
+.else
+.ifdef SKEIN_PERF #use code size if SKEIN_PERF is defined
+_SKEIN_CODE_SIZE = (1)
+.endif
+.endif
+#
+#################
+#
+.ifndef SKEIN_DEBUG
+_SKEIN_DEBUG = 0
+.else
+_SKEIN_DEBUG = 1
+.endif
+#################
+#
+# define offsets of fields in hash context structure
+#
+HASH_BITS = 0 ## bits of hash output
+BCNT = 4 + HASH_BITS #number of bytes in BUFFER[]
+TWEAK = 4 + BCNT #tweak values[0..1]
+X_VARS = 16 + TWEAK #chaining vars
+#
+#(Note: buffer[] in context structure is NOT needed here :-)
+#
+KW_PARITY_LO= 0xA9FC1A22 #overall parity of key schedule words (hi32/lo32)
+KW_PARITY_HI= 0x1BD11BDA
+FIRST_MASK8 = ~ (1 << 6) #FIRST block flag bit
+#
+# rotation constants for Skein
+#
+RC_256_0_0 = 14
+RC_256_0_1 = 16
+
+RC_256_1_0 = 52
+RC_256_1_1 = 57
+
+RC_256_2_0 = 23
+RC_256_2_1 = 40
+
+RC_256_3_0 = 5
+RC_256_3_1 = 37
+
+RC_256_4_0 = 25
+RC_256_4_1 = 33
+
+RC_256_5_0 = 46
+RC_256_5_1 = 12
+
+RC_256_6_0 = 58
+RC_256_6_1 = 22
+
+RC_256_7_0 = 32
+RC_256_7_1 = 32
+
+RC_512_0_0 = 46
+RC_512_0_1 = 36
+RC_512_0_2 = 19
+RC_512_0_3 = 37
+
+RC_512_1_0 = 33
+RC_512_1_1 = 27
+RC_512_1_2 = 14
+RC_512_1_3 = 42
+
+RC_512_2_0 = 17
+RC_512_2_1 = 49
+RC_512_2_2 = 36
+RC_512_2_3 = 39
+
+RC_512_3_0 = 44
+RC_512_3_1 = 9
+RC_512_3_2 = 54
+RC_512_3_3 = 56
+
+RC_512_4_0 = 39
+RC_512_4_1 = 30
+RC_512_4_2 = 34
+RC_512_4_3 = 24
+
+RC_512_5_0 = 13
+RC_512_5_1 = 50
+RC_512_5_2 = 10
+RC_512_5_3 = 17
+
+RC_512_6_0 = 25
+RC_512_6_1 = 29
+RC_512_6_2 = 39
+RC_512_6_3 = 43
+
+RC_512_7_0 = 8
+RC_512_7_1 = 35
+RC_512_7_2 = 56
+RC_512_7_3 = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 = 8
+RC_1024_0_3 = 47
+RC_1024_0_4 = 8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 = 4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 = 5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 = 9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 = 4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 = 9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+#
+#----------------------------------------------------------------
+# declare allocated space on the stack
+.macro StackVar localName,localSize
+\localName = _STK_OFFS_
+_STK_OFFS_ = _STK_OFFS_+(\localSize)
+.endm #StackVar
+#
+#----------------------------------------------------------------
+#
+# MACRO: Configure stack frame, allocate local vars
+#
+.macro Setup_Stack WCNT,RND_CNT
+_STK_OFFS_ = 0 #starting offset from esp, forced on 16-byte alignment
+ #----- local variables #<-- esp
+ StackVar X_stk , 8*(WCNT) #local context vars
+ StackVar Wcopy , 8*(WCNT) #copy of input block
+ StackVar ksTwk ,16*3 #key schedule: tweak words
+ StackVar ksKey ,16*(WCNT)+16#key schedule: key words
+FRAME_OFFS = ksTwk+128 #<-- ebp
+F_O = FRAME_OFFS #syntactic shorthand
+ .if (SKEIN_ASM_UNROLL && (WCNT*64)) == 0
+ StackVar ksRot,16*(RND_CNT/4)#leave space for ks "rotation" to happen
+ .endif
+LOCAL_SIZE = _STK_OFFS_ #size of local vars
+ #
+ #"restart" the stack defns, because we relocate esp to guarantee alignment
+ # (i.e., these vars are NOT at fixed offsets from esp)
+_STK_OFFS_ = 0
+ #-----
+ StackVar savRegs,8*4 #pushad data
+ StackVar retAddr,4 #return address
+ #----- caller parameters
+ StackVar ctxPtr ,4 #context ptr
+ StackVar blkPtr ,4 #pointer to block data
+ StackVar blkCnt ,4 #number of full blocks to process
+ StackVar bitAdd ,4 #bit count to add to tweak
+ #----- caller's stack frame
+#
+# Notes on stack frame setup:
+# * the most used variable (except for Skein-256) is X_stk[], based at [esp+0]
+# * the next most used is the key schedule words
+# so ebp is "centered" there, allowing short offsets to the key/tweak
+# schedule in 256/512-bit Skein cases, but not posible for Skein-1024 :-(
+# * the Wcopy variables are infrequently accessed, and they have long
+# offsets from both esp and ebp only in the 1024-bit case.
+# * all other local vars and calling parameters can be accessed
+# with short offsets, except in the 1024-bit case
+#
+ pushal #save all regs
+ movl %esp,%ebx #keep ebx as pointer to caller parms
+ subl $LOCAL_SIZE,%esp #make room for the locals
+ andl $~15,%esp #force alignment
+ movl ctxPtr(%ebx),%edi #edi --> Skein context
+ leal FRAME_OFFS(%esp),%ebp #maximize use of short offsets from ebp
+ movl blkCnt(%ebx),%ecx #keep block cnt in ecx
+.endm #Setup_Stack
+#
+#----------------------------------------------------------------
+#
+.macro Reset_Stack,procStart
+ movl %ebx,%esp #get rid of locals (wipe??)
+ popal #restore all regs
+.endm # Reset_Stack
+#
+#----------------------------------------------------------------
+# macros to help debug internals
+#
+.if _SKEIN_DEBUG
+ .extern _Skein_Show_Block #calls to C routines
+ .extern _Skein_Show_Round
+#
+SKEIN_RND_SPECIAL = 1000
+SKEIN_RND_KEY_INITIAL = SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT = SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD = SKEIN_RND_SPECIAL+2
+#
+.macro Skein_Debug_Block BLK_BITS
+#
+#void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+# const u08b_t *blkPtr, const u64b_t *wPtr,
+# const u64b_t *ksPtr,const u64b_t *tsPtr)#
+#
+ call _Put_XMM_\BLK_BITS
+ pushal #save all regs
+ leal ksTwk+1-F_O(%ebp),%eax #+1 = flag: "stride" size = 2 qwords
+ leal ksKey+1-F_O(%ebp),%esi
+ leal Wcopy+32(%esp),%ecx #adjust offset by 32 for pushad
+ movl ctxPtr(%ebx) ,%edx #ctx_hdr_ptr
+ leal X_VARS(%edx) ,%edx #edx ==> cxt->X[]
+ pushl %eax #tsPtr
+ pushl %esi #ksPtr
+ pushl %ecx #wPtr
+ pushl blkPtr(%ebx) #blkPtr
+ pushl %edx #ctx->Xptr
+ pushl ctxPtr(%ebx) #ctx_hdr_ptr
+ movl $\BLK_BITS,%eax
+ pushl %eax #bits
+ call _Skein_Show_Block
+ addl $7*4,%esp #discard parameter space on stack
+ popal #restore regs
+#
+ call _Get_XMM_\BLK_BITS
+.endm #Skein_Debug_Block
+
+#
+.macro Skein_Debug_Round BLK_BITS,R,saveRegs=0
+#
+#void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X)#
+#
+ .if \saveRegs
+ call _Put_XMM_\BLK_BITS
+ .endif
+ pushal #save all regs
+ .if R <> SKEIN_RND_FEED_FWD
+ leal 32+X_stk(%esp),%eax #adjust offset by 32 for pushal
+ .else
+ movl ctxPtr(%ebx),%eax
+ addl $X_VARS,%eax
+ .endif
+ pushl %eax #Xptr
+ .if (SKEIN_ASM_UNROLL && \BLK_BITS) || (\R >= SKEIN_RND_SPECIAL)
+ movl $\R,%eax
+ .else #compute round number from edx, R
+ leal 1+(((\R)-1) && 3)(,%edx,4),%eax
+ .endif
+ pushl %eax #round number
+ pushl ctxPtr(%ebx) #ctx_hdr_ptr
+ movl $\BLK_BITS,%eax
+ pushl %eax #bits
+ call _Skein_Show_Round
+ addl $4*4,%esp #discard parameter space on stack
+ popal #restore regs
+ .if \saveRegs
+ call _Get_XMM_\BLK_BITS #save internal vars for debug dump
+ .endif
+.endm #Skein_Debug_Round
+.endif #ifdef SKEIN_DEBUG
+#
+#----------------------------------------------------------------
+# useful macros
+.macro _ldX xn
+ movq X_stk+8*(\xn)(%esp),%xmm\xn
+.endm
+
+.macro _stX xn
+ movq %xmm\xn,X_stk+8*(\xn)(%esp)
+.endm
+#
+#----------------------------------------------------------------
+#
+.macro C_label lName
+ \lName: #use both "genders" to work across linkage conventions
+_\lName:
+ .global \lName
+ .global _\lName
+.endm
+#
+
+.if _USE_ASM_ & 256
+#
+# void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+#
+# Skein-256 round macros
+#
+.macro R_256_OneRound _RR_,x0,x1,x2,x3,t0,t1
+ .irp _qq_,%((\_RR_) && 7) #figure out which rotation constants to use
+ .if \x0 == 0
+_RC0_ = RC_256_\_qq_&&_0
+_RC1_ = RC_256_\_qq_&&_1
+ .else
+_RC0_ = RC_256_\_qq_&&_1
+_RC1_ = RC_256_\_qq_&&_0
+ .endif
+ .endr
+#
+ paddq %xmm\x1,%xmm\x0
+ movq %xmm\x1,%xmm\t0
+ psllq $ _RC0_,%xmm\x1
+ psrlq $64-_RC0_,%xmm\t0
+ xorpd %xmm\x0,%xmm\x1
+ xorpd %xmm\t0,%xmm\x1
+#
+ paddq %xmm\x3,%xmm\x2
+ movq %xmm\x3,%xmm\t1
+ psllq $ _RC1_,%xmm\x3
+ psrlq $64-_RC1_,%xmm\t1
+ xorpd %xmm\x2,%xmm\x3
+ xorpd %xmm\t1,%xmm\x3
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 256,%(\_RR_+1),SAVE_REGS
+ .endif
+.endm #R_256_OneRound
+#
+.macro R_256_FourRounds _RN_
+ R_256_OneRound %(_RN_+0),0,1,2,3,4,5
+ R_256_OneRound (_RN_+1),2,1,0,3,4,5
+
+ R_256_OneRound (_RN_+2),0,1,2,3,4,5
+ R_256_OneRound (_RN_+3),2,1,0,3,4,5
+
+ #inject key schedule
+ incl %edx #bump round number
+ movd %edx,%xmm4
+ .if _UNROLL_CNT == (ROUNDS_256/8)
+ #fully unrolled version
+_RK_ = ((_RN_)/4) #key injection counter
+ paddq ksKey+16*((_RK_+1) % 5)-F_O(%ebp),%xmm0
+ paddq ksKey+16*((_RK_+2) % 5)-F_O(%ebp),%xmm1
+ paddq ksKey+16*((_RK_+3) % 5)-F_O(%ebp),%xmm2
+ paddq ksKey+16*((_RK_+4) % 5)-F_O(%ebp),%xmm3
+ paddq ksTwk+16*((_RK_+1) % 3)-F_O(%ebp),%xmm1
+ paddq ksTwk+16*((_RK_+2) % 3)-F_O(%ebp),%xmm2
+ paddq %xmm4,%xmm3
+ .else #looping version
+ paddq ksKey+16*1-F_O(%esi),%xmm0
+ paddq ksKey+16*2-F_O(%esi),%xmm1
+ paddq ksKey+16*3-F_O(%esi),%xmm2
+ paddq ksKey+16*4-F_O(%esi),%xmm3
+ paddq ksTwk+16*1-F_O(%esi),%xmm1
+ paddq ksTwk+16*2-F_O(%esi),%xmm2
+ paddq %xmm4,%xmm3
+#
+ movq ksKey-F_O(%esi),%xmm4 #first, "rotate" key schedule on the stack
+ movq ksTwk-F_O(%esi),%xmm5 # (for next time through)
+ movq %xmm4,ksKey+16*(WCNT+1)-F_O(%esi)
+ movq %xmm5,ksTwk+16*3-F_O(%esi)
+ addl $16,%esi #bump rolling pointer
+ .endif
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT,SAVE_REGS
+ .endif
+.endm #R256_FourRounds
+#
+.if _SKEIN_DEBUG # macros for saving/restoring X_stk for debug routines
+_Put_XMM_256:
+ .irp _NN_,0,1,2,3
+ movq %xmm\_NN_,X_stk+4+\_NN_*8(%esp)
+ .endr
+ ret
+#
+_Get_XMM_256:
+ .irp _NN_,0,1,2,3
+ movq X_stk+4+_NN_*8(%esp),%xmm\_NN_
+ .endr
+ ret
+.endif
+#
+#################
+#
+# code
+#
+C_label Skein_256_Process_Block
+ WCNT = 4 #WCNT=4 for Skein-256
+ Setup_Stack WCNT,ROUNDS_256
+ # main hash loop for Skein_256
+Skein_256_block_loop:
+ movd bitAdd (%ebx),%xmm4
+ movq TWEAK+0(%edi),%xmm5
+ movq TWEAK+8(%edi),%xmm6
+ paddq %xmm4 ,%xmm5 #bump T0 by the bitAdd parameter
+ movq %xmm5,TWEAK(%edi) #save updated tweak value T0 (for next time)
+ movapd %xmm6,%xmm7
+ xorpd %xmm5,%xmm7 #compute overall tweak parity
+ movdqa %xmm5,ksTwk -F_O(%ebp)#save the expanded tweak schedule on the stack
+ movdqa %xmm6,ksTwk+16-F_O(%ebp)
+ movdqa %xmm7,ksTwk+32-F_O(%ebp)
+
+ movl blkPtr(%ebx),%esi #esi --> input block
+ movl $KW_PARITY_LO,%eax #init key schedule parity accumulator
+ movl $KW_PARITY_HI,%edx
+ movd %eax ,%xmm4
+ movd %edx ,%xmm0
+ unpcklps %xmm0,%xmm4 #replicate parity dword to 64 bits
+#
+ .irp _NN_,0,1,2,3 #copy in the chaining vars
+ movq X_VARS+8*\_NN_(%edi),%xmm\_NN_
+ xorpd %xmm\_NN_,%xmm4 #update overall parity
+ movdqa %xmm\_NN_,ksKey+16*_NN_-F_O(%ebp)
+ .endr
+ movdqa %xmm4,ksKey+16*WCNT-F_O(%ebp)#save overall parity at the end of the array
+#
+ paddq %xmm5,%xmm1 #inject the initial tweak words
+ paddq %xmm6,%xmm2
+#
+ .irp _NN_,0,1,2,3 #perform the initial key injection
+ movq 8*\_NN_(%esi),%xmm4#and save a copy of the input block on stack
+ movq %xmm4,8*\_NN_+Wcopy(%esp)
+ paddq %xmm4,%xmm\_NN_ #inject the key word
+ .endr
+#
+.if _SKEIN_DEBUG #debug dump of state at this point
+ Skein_Debug_Block 256
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL,SAVE_REGS
+.endif
+ addl $WCNT*8,%esi #skip to the next block
+ movl %esi,blkPtr(%ebx) #save the updated block pointer
+ #
+ # now the key schedule is computed. Start the rounds
+ #
+ xorl %edx,%edx #edx = iteration count
+.if SKEIN_ASM_UNROLL & 256
+_UNROLL_CNT = ROUNDS_256/8 #fully unrolled
+.else
+_UNROLL_CNT = SKEIN_UNROLL_256 #partial unroll count
+ .if ((ROUNDS_256/8) % _UNROLL_CNT)
+ .error "Invalid SKEIN_UNROLL_256" #sanity check
+ .endif
+ movl %ebp,%esi #use this as "rolling" pointer into ksTwk/ksKey
+Skein_256_round_loop: # (since there's no 16* scaled address mode)
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2 # here with X[0..3] in XMM0..XMM3
+ R_256_FourRounds _Rbase_
+_Rbase_ = _Rbase_+4
+.endr #rept _UNROLL_CNT*2
+#
+ .if _UNROLL_CNT <> (ROUNDS_256/8)
+ cmpl $2*(ROUNDS_256/8),%edx
+ jb Skein_256_round_loop
+ .endif
+ #----------------------------
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..3}
+ .irp _NN_,0,1,2,3
+ movq Wcopy+8*\_NN_(%esp),%xmm4
+ xorpd %xmm4,%xmm\_NN_
+ movq %xmm\_NN_,X_VARS+8*\_NN_(%edi)
+ .endr
+ andb $FIRST_MASK8,TWEAK +15(%edi)
+.if _SKEIN_DEBUG
+ Skein_Debug_Round 256,SKEIN_RND_FEED_FWD,SAVE_REGS
+.endif
+ # go back for more blocks, if needed
+ decl %ecx
+ jnz Skein_256_block_loop
+ Reset_Stack _Skein_256_Process_Block
+ ret
+#
+.ifdef _SKEIN_CODE_SIZE
+C_label Skein_256_Process_Block_CodeSize
+ movl $_Skein_256_Process_Block_CodeSize - _Skein_256_Process_Block,%eax
+ ret
+#
+C_label Skein_256_Unroll_Cnt
+ .if _UNROLL_CNT <> ROUNDS_256/8
+ movl $_UNROLL_CNT,%eax
+ .else
+ xorl %eax,%eax
+ .endif
+ ret
+.endif
+.endif #_USE_ASM_ & 256
+#
+#----------------------------------------------------------------
+#
+.if _USE_ASM_ & 512
+#
+# void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+# MACRO: one round
+#
+.macro R_512_Round _RR_, a0,a1,Ra, b0,b1,Rb, c0,c1,Rc, d0,d1,Rd
+ .irp _qq_,%((\_RR_) && 7)
+_Ra_ = RC_512_\_qq_&&_\Ra
+_Rb_ = RC_512_\_qq_&&_\Rb
+_Rc_ = RC_512_\_qq_&&_\Rc
+_Rd_ = RC_512_\_qq_&&_\Rd
+ .endr
+ paddq %xmm\a1 , %xmm\a0
+ _stX c0
+ movq %xmm\a1 , %xmm\c0
+ psllq $ _Ra_ , %xmm\a1
+ psrlq $64-_Ra_ , %xmm\c0
+ xorpd %xmm\c0 , %xmm\a1
+ xorpd %xmm\a0 , %xmm\a1
+
+ paddq %xmm\b1 , %xmm\b0
+ _stX a0
+ movq %xmm\b1 , %xmm\a0
+ psllq $ _Rb_ , %xmm\b1
+ psrlq $64-_Rb_ , %xmm\a0
+ xorpd %xmm\b0 , %xmm\b1
+ _ldX c0
+ xorpd %xmm\a0 , %xmm\b1
+
+ paddq %xmm\c1 , %xmm\c0
+ movq %xmm\c1 , %xmm\a0
+ psllq $ _Rc_ , %xmm\c1
+ psrlq $64-_Rc_ , %xmm\a0
+ xorpd %xmm\c0 , %xmm\c1
+ xorpd %xmm\a0 , %xmm\c1
+
+ paddq %xmm\d1 , %xmm\d0
+ movq %xmm\d1 , %xmm\a0
+ psllq $ _Rd_ , %xmm\d1
+ psrlq $64-_Rd_ , %xmm\a0
+ xorpd %xmm\a0 , %xmm\d1
+ _ldX a0
+ xorpd %xmm\d0 , %xmm\d1
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 512,%(_RR_+1),SAVE_REGS
+ .endif
+.endm
+#
+# MACRO: four rounds
+.macro R_512_FourRounds _RN_
+ R_512_Round %((_RN_) ), 0,1,0, 2,3,1, 4,5,2, 6,7,3
+ R_512_Round %((_RN_)+1), 2,1,0, 4,7,1, 6,5,2, 0,3,3
+ R_512_Round %((_RN_)+2), 4,1,0, 6,3,1, 0,5,2, 2,7,3
+ R_512_Round %((_RN_)+3), 6,1,0, 0,7,1, 2,5,2, 4,3,3
+
+ #inject key schedule
+.irp _NN_,0,1,2,3,4,5,6,7
+ .if _UNROLL_CNT == (ROUNDS_512/8)
+ paddq ksKey+16*((((\_RN_)/4)+(\_NN_)+1)%9)-F_O(%ebp),%xmm\_NN_
+ .else
+ paddq ksKey+16*((\_NN_)+1)-F_O(%esi),%xmm\_NN_
+ .endif
+.endr
+ _stX 0 #free up a register
+ incl %edx #bump round counter
+ movd %edx,%xmm0 #inject the tweak
+ .if _UNROLL_CNT == (ROUNDS_512/8)
+ paddq ksTwk+16*(((_RN_)+1) % 3)-F_O(%ebp),%xmm5
+ paddq ksTwk+16*(((_RN_)+2) % 3)-F_O(%ebp),%xmm6
+ paddq %xmm0 ,%xmm7
+ .else #looping version
+ paddq ksTwk+16*1-F_O(%esi),%xmm5
+ paddq ksTwk+16*2-F_O(%esi),%xmm6
+ paddq %xmm0 ,%xmm7
+ # "rotate" key schedule on the stack (for next time through)
+ movq ksKey -F_O(%esi),%xmm0
+ movq %xmm0,ksKey+16*(WCNT+1)-F_O(%esi)
+ movq ksTwk -F_O(%esi),%xmm0
+ movq %xmm0,ksTwk+16*3 -F_O(%esi)
+ addl $16,%esi #bump rolling pointer
+ .endif
+ _ldX 0 #restore X0
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT,SAVE_REGS
+ .endif
+.endm #R_512_FourRounds
+#################
+.if _SKEIN_DEBUG # macros for saving/restoring X_stk for debug routines
+_Put_XMM_512:
+ .irp _NN_,0,1,2,3,4,5,6,7
+ movq %xmm\_NN_,X_stk+4+\_NN_*8(%esp)
+ .endr
+ ret
+#
+_Get_XMM_512:
+ .irp _NN_,0,1,2,3,4,5,6,7
+ movq X_stk+4+\_NN_*8(%esp),%xmm\_NN_
+ .endr
+ ret
+.endif
+#
+#################
+#
+C_label Skein_512_Process_Block
+ WCNT = 8 #WCNT=8 for Skein-512
+ Setup_Stack WCNT,ROUNDS_512
+ # main hash loop for Skein_512
+Skein_512_block_loop:
+ movd bitAdd(%ebx) ,%xmm0
+ movq TWEAK+0(%edi),%xmm1
+ movq TWEAK+8(%edi),%xmm2
+ paddq %xmm0,%xmm1 #bump T0 by the bitAdd parameter
+ movq %xmm1,TWEAK(%edi) #save updated tweak value T0 (for next time)
+ movq %xmm2,%xmm0
+ xorpd %xmm1,%xmm0 #compute overall tweak parity
+ movdqa %xmm1,ksTwk -F_O(%ebp)#save the expanded tweak schedule on the stack
+ movdqa %xmm2,ksTwk+16*1-F_O(%ebp)
+ movdqa %xmm0,ksTwk+16*2-F_O(%ebp)
+
+ movl blkPtr(%ebx),%esi #esi --> input block
+ movl $KW_PARITY_LO,%eax #init key schedule parity accumulator
+ movl $KW_PARITY_HI,%edx
+ movd %eax ,%xmm0
+ movd %edx ,%xmm7
+ unpcklps %xmm7,%xmm0 #replicate parity dword to 64 bits
+#
+ .irp _NN_,7,6,5,4,3,2,1 #copy in the chaining vars (skip #0 for now)
+ movq X_VARS+8*\_NN_(%edi),%xmm\_NN_
+ xorpd %xmm\_NN_,%xmm0 #update overall parity
+ movdqa %xmm\_NN_,ksKey+16*\_NN_-F_O(%ebp)
+ .if \_NN_ == 5
+ paddq %xmm1,%xmm5 #inject the initial tweak words
+ paddq %xmm2,%xmm6 # (before they get trashed in %xmm1/2)
+ .endif
+ .endr
+ movq X_VARS(%edi),%xmm4 #handle #0 now
+ xorpd %xmm4,%xmm0 #update overall parity
+ movdqa %xmm4,ksKey+16* 0 -F_O(%ebp) #save the key value in slot #0
+ movdqa %xmm0,ksKey+16*WCNT-F_O(%ebp) #save overall parity at the end of the array
+#
+ movq %xmm4,%xmm0
+ .irp _NN_,7,6,5, 4,3,2,1,0 #perform the initial key injection (except #4)
+ movq 8*\_NN_(%esi),%xmm4 #and save a copy of the input block on stack
+ movq %xmm4,8*\_NN_+Wcopy(%esp)
+ paddq %xmm4,%xmm\_NN_
+ .endr
+ movq 8*4(%esi),%xmm4 #get input block word #4
+ movq %xmm4,8*4+Wcopy(%esp)
+ paddq ksKey+16*4-F_O(%ebp),%xmm4#inject the initial key
+#
+.if _SKEIN_DEBUG #debug dump of state at this point
+ Skein_Debug_Block 512
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL,SAVE_REGS
+.endif
+ addl $WCNT*8,%esi #skip to the next block
+ movl %esi,blkPtr(%ebx) #save the updated block pointer
+ #
+ # now the key schedule is computed. Start the rounds
+ #
+ xorl %edx,%edx #edx = round counter
+.if SKEIN_ASM_UNROLL & 512
+_UNROLL_CNT = ROUNDS_512/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_512
+ .if ((ROUNDS_512/8) % _UNROLL_CNT)
+ .error "Invalid SKEIN_UNROLL_512"
+ .endif
+ movl %ebp,%esi #use this as "rolling" pointer into ksTwk/ksKey
+Skein_512_round_loop: # (since there's no 16* scaled address mode)
+.endif
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+ R_512_FourRounds %_Rbase_
+_Rbase_ = _Rbase_+4
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 512) == 0
+ cmpl $2*(ROUNDS_512/8),%edx
+ jb Skein_512_round_loop
+.endif
+ #----------------------------
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..7}
+ andb $FIRST_MASK8,TWEAK +15(%edi)
+.irp _NN_,0,2,4,6 #do the aligned ones first
+ xorpd Wcopy+8*\_NN_(%esp),%xmm\_NN_
+ movq %xmm\_NN_,X_VARS+8*_NN_(%edi)
+.endr
+.irp _NN_,1,3,5,7 #now we have some register space available
+ movq Wcopy+8*\_NN_(%esp),%xmm0
+ xorpd %xmm0,%xmm&\_NN_
+ movq %xmm&\_NN_,X_VARS+8*\_NN_(%edi)
+.endr
+.if _SKEIN_DEBUG
+ Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+.endif
+ # go back for more blocks, if needed
+ decl %ecx
+ jnz Skein_512_block_loop
+
+ Reset_Stack _Skein_512_Process_Block
+ ret
+#
+.ifdef _SKEIN_CODE_SIZE
+C_label Skein_512_Process_Block_CodeSize
+ movl $(_Skein_512_Process_Block_CodeSize - _Skein_512_Process_Block),%eax
+ ret
+#
+C_label Skein_512_Unroll_Cnt
+ .if _UNROLL_CNT <> ROUNDS_512/8
+ movl $_UNROLL_CNT,%eax
+ .else
+ xorl %eax,%eax
+ .endif
+ ret
+.endif
+#
+.endif # _USE_ASM_ & 512
+#
+#----------------------------------------------------------------
+#
+.if _USE_ASM_ & 1024
+ .global _Skein1024_Process_Block
+#
+# void Skein_1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+R_1024_REGS = (5) #keep this many block variables in registers
+#
+################
+.if _SKEIN_DEBUG # macros for saving/restoring X_stk for debug routines
+_Put_XMM_1024:
+_NN_ = 0
+ .rept R_1024_REGS
+ .irp _rr_,%(_NN_)
+ movq %xmm\_rr_,X_stk+4+8*_NN_(%esp)
+ .endr
+_NN_ = _NN_+1
+ .endr
+ ret
+#
+_Get_XMM_1024:
+_NN_ = 0
+ .rept R_1024_REGS
+ .irp _rr_,%(_NN_)
+ movq X_stk+4+8*_NN_(%esp),%xmm\_rr_
+ .endr
+_NN_ = _NN_+1
+ .endr
+ ret
+.endif
+#
+#################
+# MACRO: one mix step
+.macro MixStep_1024 x0,x1,rotIdx0,rotIdx1,_debug_=0
+_r0_ = \x0 #default, if already loaded
+_r1_ = \x1
+ # load the regs (if necessary)
+ .if (\x0 >= R_1024_REGS)
+_r0_ = 5
+ movq X_stk+8*(\x0)(%esp),%xmm5
+ .endif
+ .if (\x1 >= R_1024_REGS)
+_r1_ = 6
+ movq X_stk+8*(\x1)(%esp),%xmm6
+ .endif
+ # do the mix
+ .irp _rx_,%((rotIdx0) && 7)
+_Rc_ = RC_1024_\_rx_&&_\rotIdx1 #rotation constant
+ .endr
+ .irp _x0_,%_r0_
+ .irp _x1_,%_r1_
+ paddq %xmm\_x1_,%xmm\_x0_
+ movq %xmm\_x1_,%xmm7
+ psllq $ _Rc_ ,%xmm\_x1_
+ psrlq $64-_Rc_ ,%xmm7
+ xorpd %xmm\_x0_,%xmm\_x1_
+ xorpd %xmm7 ,%xmm\_x1_
+ .endr
+ .endr
+ # save the regs (if necessary)
+ .if (\x0 >= R_1024_REGS)
+ movq %xmm5,X_stk+8*(\x0)(%esp)
+ .endif
+ .if (\x1 >= R_1024_REGS)
+ movq %xmm6,X_stk+8*(\x1)(%esp)
+ .endif
+ # debug output
+ .if _SKEIN_DEBUG && (\_debug_)
+ Skein_Debug_Round 1024,%((\RotIdx0)+1),SAVE_REGS
+ .endif
+.endm
+#################
+# MACRO: four rounds
+#
+.macro R_1024_FourRounds _RR_
+ #--------- round _RR_
+ MixStep_1024 0, 1,%((\_RR_)+0),0
+ MixStep_1024 2, 3,%((\_RR_)+0),1
+ MixStep_1024 4, 5,%((\_RR_)+0),2
+ MixStep_1024 6, 7,%((\_RR_)+0),3
+ MixStep_1024 8, 9,%((\_RR_)+0),4
+ MixStep_1024 10,11,%((\_RR_)+0),5
+ MixStep_1024 12,13,%((\_RR_)+0),6
+ MixStep_1024 14,15,%((\_RR_)+0),7,1
+ #--------- round _RR_+1
+ MixStep_1024 0, 9,%((\_RR_)+1),0
+ MixStep_1024 2,13,%((\_RR_)+1),1
+ MixStep_1024 6,11,%((\_RR_)+1),2
+ MixStep_1024 4,15,%((\_RR_)+1),3
+ MixStep_1024 10, 7,%((\_RR_)+1),4
+ MixStep_1024 12, 3,%((\_RR_)+1),5
+ MixStep_1024 14, 5,%((\_RR_)+1),6
+ MixStep_1024 8, 1,%((\_RR_)+1),7,1
+ #--------- round _RR_+2
+ MixStep_1024 0, 7,%((\_RR_)+2),0
+ MixStep_1024 2, 5,%((\_RR_)+2),1
+ MixStep_1024 4, 3,%((\_RR_)+2),2
+ MixStep_1024 6, 1,%((\_RR_)+2),3
+ MixStep_1024 12,15,%((\_RR_)+2),4
+ MixStep_1024 14,13,%((\_RR_)+2),5
+ MixStep_1024 8,11,%((\_RR_)+2),6
+ MixStep_1024 10, 9,%((\_RR_)+2),7,1
+ #--------- round _RR_+3
+ MixStep_1024 0,15,%((\_RR_)+3),0
+ MixStep_1024 2,11,%((\_RR_)+3),1
+ MixStep_1024 6,13,%((\_RR_)+3),2
+ MixStep_1024 4, 9,%((\_RR_)+3),3
+ MixStep_1024 14, 1,%((\_RR_)+3),4
+ MixStep_1024 8, 5,%((\_RR_)+3),5
+ MixStep_1024 10, 3,%((\_RR_)+3),6
+ MixStep_1024 12, 7,%((\_RR_)+3),7,1
+
+ incl %edx #edx = round number
+ movd %edx,%xmm7
+
+ #inject the key
+.irp _NN_,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
+ .if _UNROLL_CNT <> (ROUNDS_1024/8)
+ .if \_NN_ < R_1024_REGS
+ paddq ksKey+16*\_NN_+16-F_O(%esi),%xmm&\_NN_
+ .else
+ movq X_stk+ 8*\_NN_(%esp),%xmm6
+ .if \_NN_ == 15
+ paddq %xmm7,%xmm6
+ .elseif \_NN_ == 14
+ paddq ksTwk+16*2-F_O(%esi),%xmm6
+ .elseif \_NN_ == 13
+ paddq ksTwk+16*1-F_O(%esi),%xmm6
+ .endif
+ paddq ksKey+16*\_NN_+16-F_O(%esi),%xmm6
+ movq %xmm6,X_stk+ 8*\_NN_(%esp)
+ .endif
+ .else
+ .if \_NN_ < R_1024_REGS
+ paddq ksKey+16*(((_Rbase_/4)+(\_NN_)+1) % 17)-F_O(%ebp),%xmm&\_NN_
+ .else
+ movq X_stk+ 8*\_NN_(%esp), %xmm6
+ paddq ksKey+16*(((_Rbase_/4)+(\_NN_)+1) % 17)-F_O(%ebp),%xmm6
+ .if \_NN_ == 15
+ paddq %xmm7,%xmm6
+ .elseif \_NN_ == 14
+ paddq ksTwk+16*(((_Rbase_/4)+2) % 3)-F_O(%ebp),%xmm6
+ .elseif \_NN_ == 13
+ paddq ksTwk+16*(((_Rbase_/4)+1) % 3)-F_O(%ebp),%xmm6
+ .endif
+ movq %xmm6,X_stk+ 8*\_NN_(%esp)
+ .endif
+ .endif
+.endr
+ .if _UNROLL_CNT <> (ROUNDS_1024/8) #rotate the key schedule on the stack
+ movq ksKey-F_O(%esi), %xmm6
+ movq ksTwk-F_O(%esi), %xmm7
+ movq %xmm6,ksKey+16*(WCNT+1)-F_O(%esi)
+ movq %xmm7,ksTwk+16* 3 -F_O(%esi)
+ addl $16,%esi #bump rolling pointer
+ .endif
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT ,SAVE_REGS
+ .endif
+.endm #R_1024_FourRounds
+#
+################
+#
+C_label Skein1024_Process_Block
+#
+ WCNT = 16 #WCNT=16 for Skein-1024
+ Setup_Stack WCNT,ROUNDS_1024
+ addl $0x80,%edi #bias the edi ctxt offsets to keep them all short
+ # main hash loop for Skein1024
+Skein1024_block_loop:
+ movd bitAdd(%ebx) ,%xmm0
+ movq TWEAK+0-0x80(%edi),%xmm1
+ movq TWEAK+8-0x80(%edi),%xmm2
+ paddq %xmm0,%xmm1 #bump T0 by the bitAdd parameter
+ movq %xmm1,TWEAK-0x80(%edi) #save updated tweak value T0 (for next time)
+ movq %xmm2,%xmm0
+ xorpd %xmm1,%xmm0 #compute overall tweak parity
+ movdqa %xmm1,ksTwk -F_O(%ebp)#save the expanded tweak schedule on the stack
+ movdqa %xmm2,ksTwk+16-F_O(%ebp)
+ movdqa %xmm0,ksTwk+32-F_O(%ebp)
+
+ movl blkPtr(%ebx),%esi #esi --> input block
+ movl $KW_PARITY_LO,%eax #init key schedule parity accumulator
+ movl $KW_PARITY_HI,%edx
+ movd %eax ,%xmm7
+ movd %edx ,%xmm6
+ unpcklps %xmm6,%xmm7 #replicate parity dword to 64 bits
+#
+ leal 0x80(%esp),%eax #use short offsets for Wcopy, X_stk writes below
+.irp _NN_,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
+ movq X_VARS+8*\_NN_-0x80(%edi),%xmm6
+ xorpd %xmm6,%xmm7 #update overall parity
+ movdqa %xmm6,ksKey+16*\_NN_-F_O(%ebp) #save the key schedule on the stack
+ .if \_NN_ < R_1024_REGS
+ _rr_ = \_NN_
+ .else
+ _rr_ = R_1024_REGS
+ .endif
+ .irp _rn_,%(_rr_)
+ movq 8*\_NN_(%esi),%xmm\_rn_ #save copy of the input block on stack
+ movq %xmm\_rn_,Wcopy+8*\_NN_-0x80(%eax) #(for feedforward later)
+ paddq %xmm6,%xmm\_rn_ #inject the key into the block
+ .if \_NN_ == 13
+ paddq %xmm1,%xmm\_rn_ #inject the initial tweak words
+ .elseif \_NN_ == 14
+ paddq %xmm2,%xmm\_rn_
+ .endif
+ .if \_NN_ >= R_1024_REGS #only save X[5..15] on stack, leave X[0..4] in regs
+ movq %xmm\_rn_,X_stk+8*\_NN_-0x80(%eax)
+ .endif
+ .endr
+.endr
+ movdqa %xmm7,ksKey+16*WCNT-F_O(%ebp) #save overall key parity at the end of the array
+#
+.if _SKEIN_DEBUG #debug dump of state at this point
+ Skein_Debug_Block 1024
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL,SAVE_REGS
+.endif
+ addl $WCNT*8,%esi #skip to the next block
+ movl %esi,blkPtr(%ebx) #save the updated block pointer
+ #
+ # now the key schedule is computed. Start the rounds
+ #
+ xorl %edx,%edx #edx = round counter
+.if SKEIN_ASM_UNROLL & 1024
+_UNROLL_CNT = ROUNDS_1024/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_1024
+ .if ((ROUNDS_1024/8) % _UNROLL_CNT)
+ .error "Invalid SKEIN_UNROLL_1024"
+ .endif
+ movl %ebp,%esi #use this as "rolling" pointer into ksTwk/ksKey
+Skein_1024_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+ R_1024_FourRounds %_Rbase_
+_Rbase_ = _Rbase_+4
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 1024) == 0
+ cmp $2*(ROUNDS_1024/8),%edx
+ jb Skein_1024_round_loop
+.endif
+ andb $FIRST_MASK8,TWEAK +15-0x80(%edi) #clear tweak bit for next time thru
+ #----------------------------
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..15}
+ leal 0x80(%esp),%eax #allow short offsets to X_stk and Wcopy
+.irp _NN_,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
+ .if \_NN_ < R_1024_REGS
+ .if \_NN_ && 1 #already in regs: no load needed
+ movq Wcopy+ 8*\_NN_-0x80(%eax),%xmm7 #unaligned
+ xorpd %xmm7,%xmm\_NN_
+ .else
+ xorpd Wcopy+ 8*\_NN_-0x80(%eax),%xmm\_NN_ #aligned
+ .endif
+ movq %xmm\_NN_,X_VARS+8*\_NN_-0x80(%edi)
+ .else
+ movq X_stk+8*\_NN_-0x80(%eax),%xmm7 #load X value from stack
+ .if \_NN_ && 1
+ movq Wcopy+8*\_NN_-0x80(%eax),%xmm6 #unaligned
+ xorpd %xmm6,%xmm7
+ .else
+ xorpd Wcopy+8*\_NN_-0x80(%eax),%xmm7 #aligned
+ .endif
+ movq %xmm7,X_VARS+8*\_NN_-0x80(%edi)
+ .endif
+.endr
+.if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD #no need to save regs on stack here
+.endif
+ # go back for more blocks, if needed
+ decl %ecx
+ jnz Skein1024_block_loop
+
+ Reset_Stack _Skein1024_Process_Block
+ ret
+#
+.ifdef _SKEIN_CODE_SIZE
+C_label Skein1024_Process_Block_CodeSize
+ movl $(_Skein1024_Process_Block_CodeSize - _Skein1024_Process_Block),%eax
+ ret
+#
+C_label Skein1024_Unroll_Cnt
+ .if _UNROLL_CNT <> ROUNDS_1024/8
+ movl $_UNROLL_CNT,%eax
+ .else
+ xorl %eax,%eax
+ .endif
+ ret
+.endif
+#
+.endif # _USE_ASM_ & 1024
+#----------------------------------------------------------------
+ .end
Index: sys/contrib/skein/brg_endian.h
===================================================================
--- /dev/null
+++ sys/contrib/skein/brg_endian.h
@@ -0,0 +1,148 @@
+/*
+ ---------------------------------------------------------------------------
+ Copyright (c) 2003, Dr Brian Gladman, Worcester, UK. All rights reserved.
+
+ LICENSE TERMS
+
+ The free distribution and use of this software in both source and binary
+ form is allowed (with or without changes) provided that:
+
+ 1. distributions of this source code include the above copyright
+ notice, this list of conditions and the following disclaimer;
+
+ 2. distributions in binary form include the above copyright
+ notice, this list of conditions and the following disclaimer
+ in the documentation and/or other associated materials;
+
+ 3. the copyright holder's name is not used to endorse products
+ built using this software without specific written permission.
+
+ ALTERNATIVELY, provided that this notice is retained in full, this product
+ may be distributed under the terms of the GNU General Public License (GPL),
+ in which case the provisions of the GPL apply INSTEAD OF those given above.
+
+ DISCLAIMER
+
+ This software is provided 'as is' with no explicit or implied warranties
+ in respect of its properties, including, but not limited to, correctness
+ and/or fitness for purpose.
+ ---------------------------------------------------------------------------
+ Issue 20/10/2006
+*/
+
+#ifndef BRG_ENDIAN_H
+#define BRG_ENDIAN_H
+
+#define IS_BIG_ENDIAN 4321 /* byte 0 is most significant (mc68k) */
+#define IS_LITTLE_ENDIAN 1234 /* byte 0 is least significant (i386) */
+
+/* Include files where endian defines and byteswap functions may reside */
+#if defined( __FreeBSD__ ) || defined( __OpenBSD__ ) || defined( __NetBSD__ )
+# include <sys/endian.h>
+#elif defined( BSD ) && ( BSD >= 199103 ) || defined( __APPLE__ ) || \
+ defined( __CYGWIN32__ ) || defined( __DJGPP__ ) || defined( __osf__ )
+# include <machine/endian.h>
+#elif defined( __linux__ ) || defined( __GNUC__ ) || defined( __GNU_LIBRARY__ )
+# if !defined( __MINGW32__ ) && !defined(AVR)
+# include <endian.h>
+# if !defined( __BEOS__ )
+# include <byteswap.h>
+# endif
+# endif
+#endif
+
+/* Now attempt to set the define for platform byte order using any */
+/* of the four forms SYMBOL, _SYMBOL, __SYMBOL & __SYMBOL__, which */
+/* seem to encompass most endian symbol definitions */
+
+#if defined( BIG_ENDIAN ) && defined( LITTLE_ENDIAN )
+# if defined( BYTE_ORDER ) && BYTE_ORDER == BIG_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( BYTE_ORDER ) && BYTE_ORDER == LITTLE_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( BIG_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( LITTLE_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( _BIG_ENDIAN ) && defined( _LITTLE_ENDIAN )
+# if defined( _BYTE_ORDER ) && _BYTE_ORDER == _BIG_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( _BYTE_ORDER ) && _BYTE_ORDER == _LITTLE_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( _BIG_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( _LITTLE_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN ) && defined( __LITTLE_ENDIAN )
+# if defined( __BYTE_ORDER ) && __BYTE_ORDER == __BIG_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( __BYTE_ORDER ) && __BYTE_ORDER == __LITTLE_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( __BIG_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN__ ) && defined( __LITTLE_ENDIAN__ )
+# if defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __BIG_ENDIAN__
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __LITTLE_ENDIAN__
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( __BIG_ENDIAN__ )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN__ )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+/* if the platform byte order could not be determined, then try to */
+/* set this define using common machine defines */
+#if !defined(PLATFORM_BYTE_ORDER)
+
+#if defined( __alpha__ ) || defined( __alpha ) || defined( i386 ) || \
+ defined( __i386__ ) || defined( _M_I86 ) || defined( _M_IX86 ) || \
+ defined( __OS2__ ) || defined( sun386 ) || defined( __TURBOC__ ) || \
+ defined( vax ) || defined( vms ) || defined( VMS ) || \
+ defined( __VMS ) || defined( _M_X64 ) || defined( AVR )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+
+#elif defined( AMIGA ) || defined( applec ) || defined( __AS400__ ) || \
+ defined( _CRAY ) || defined( __hppa ) || defined( __hp9000 ) || \
+ defined( ibm370 ) || defined( mc68000 ) || defined( m68k ) || \
+ defined( __MRC__ ) || defined( __MVS__ ) || defined( __MWERKS__ ) || \
+ defined( sparc ) || defined( __sparc) || defined( SYMANTEC_C ) || \
+ defined( __VOS__ ) || defined( __TIGCC__ ) || defined( __TANDEM ) || \
+ defined( THINK_C ) || defined( __VMCMS__ )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+
+#elif 0 /* **** EDIT HERE IF NECESSARY **** */
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#elif 0 /* **** EDIT HERE IF NECESSARY **** */
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#else
+# error Please edit lines 126 or 128 in brg_endian.h to set the platform byte order
+#endif
+#endif
+
+/* special handler for IA64, which may be either endianness (?) */
+/* here we assume little-endian, but this may need to be changed */
+#if defined(__ia64) || defined(__ia64__) || defined(_M_IA64)
+# define PLATFORM_MUST_ALIGN (1)
+#ifndef PLATFORM_BYTE_ORDER
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+#endif
+
+#ifndef PLATFORM_MUST_ALIGN
+# define PLATFORM_MUST_ALIGN (0)
+#endif
+
+#endif /* ifndef BRG_ENDIAN_H */
Index: sys/contrib/skein/brg_types.h
===================================================================
--- /dev/null
+++ sys/contrib/skein/brg_types.h
@@ -0,0 +1,188 @@
+/*
+ ---------------------------------------------------------------------------
+ Copyright (c) 1998-2006, Brian Gladman, Worcester, UK. All rights reserved.
+
+ LICENSE TERMS
+
+ The free distribution and use of this software in both source and binary
+ form is allowed (with or without changes) provided that:
+
+ 1. distributions of this source code include the above copyright
+ notice, this list of conditions and the following disclaimer;
+
+ 2. distributions in binary form include the above copyright
+ notice, this list of conditions and the following disclaimer
+ in the documentation and/or other associated materials;
+
+ 3. the copyright holder's name is not used to endorse products
+ built using this software without specific written permission.
+
+ ALTERNATIVELY, provided that this notice is retained in full, this product
+ may be distributed under the terms of the GNU General Public License (GPL),
+ in which case the provisions of the GPL apply INSTEAD OF those given above.
+
+ DISCLAIMER
+
+ This software is provided 'as is' with no explicit or implied warranties
+ in respect of its properties, including, but not limited to, correctness
+ and/or fitness for purpose.
+ ---------------------------------------------------------------------------
+ Issue 09/09/2006
+
+ The unsigned integer types defined here are of the form uint_<nn>t where
+ <nn> is the length of the type; for example, the unsigned 32-bit type is
+ 'uint_32t'. These are NOT the same as the 'C99 integer types' that are
+ defined in the inttypes.h and stdint.h headers since attempts to use these
+ types have shown that support for them is still highly variable. However,
+ since the latter are of the form uint<nn>_t, a regular expression search
+ and replace (in VC++ search on 'uint_{:z}t' and replace with 'uint\1_t')
+ can be used to convert the types used here to the C99 standard types.
+*/
+
+#ifndef BRG_TYPES_H
+#define BRG_TYPES_H
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#include <limits.h>
+
+#ifndef BRG_UI8
+# define BRG_UI8
+# if UCHAR_MAX == 255u
+ typedef unsigned char uint_8t;
+# else
+# error Please define uint_8t as an 8-bit unsigned integer type in brg_types.h
+# endif
+#endif
+
+#ifndef BRG_UI16
+# define BRG_UI16
+# if USHRT_MAX == 65535u
+ typedef unsigned short uint_16t;
+# else
+# error Please define uint_16t as a 16-bit unsigned short type in brg_types.h
+# endif
+#endif
+
+#ifndef BRG_UI32
+# define BRG_UI32
+# if UINT_MAX == 4294967295u
+# define li_32(h) 0x##h##u
+ typedef unsigned int uint_32t;
+# elif ULONG_MAX == 4294967295u
+# define li_32(h) 0x##h##ul
+ typedef unsigned long uint_32t;
+# elif defined( _CRAY )
+# error This code needs 32-bit data types, which Cray machines do not provide
+# else
+# error Please define uint_32t as a 32-bit unsigned integer type in brg_types.h
+# endif
+#endif
+
+#ifndef BRG_UI64
+# if defined( __BORLANDC__ ) && !defined( __MSDOS__ )
+# define BRG_UI64
+# define li_64(h) 0x##h##ui64
+ typedef unsigned __int64 uint_64t;
+# elif defined( _MSC_VER ) && ( _MSC_VER < 1300 ) /* 1300 == VC++ 7.0 */
+# define BRG_UI64
+# define li_64(h) 0x##h##ui64
+ typedef unsigned __int64 uint_64t;
+# elif defined( __sun ) && defined(ULONG_MAX) && ULONG_MAX == 0xfffffffful
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# elif defined( UINT_MAX ) && UINT_MAX > 4294967295u
+# if UINT_MAX == 18446744073709551615u
+# define BRG_UI64
+# define li_64(h) 0x##h##u
+ typedef unsigned int uint_64t;
+# endif
+# elif defined( ULONG_MAX ) && ULONG_MAX > 4294967295u
+# if ULONG_MAX == 18446744073709551615ul
+# define BRG_UI64
+# define li_64(h) 0x##h##ul
+ typedef unsigned long uint_64t;
+# endif
+# elif defined( ULLONG_MAX ) && ULLONG_MAX > 4294967295u
+# if ULLONG_MAX == 18446744073709551615ull
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# endif
+# elif defined( ULONG_LONG_MAX ) && ULONG_LONG_MAX > 4294967295u
+# if ULONG_LONG_MAX == 18446744073709551615ull
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# endif
+# elif defined(__GNUC__) /* DLW: avoid mingw problem with -ansi */
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# endif
+#endif
+
+#if defined( NEED_UINT_64T ) && !defined( BRG_UI64 )
+# error Please define uint_64t as an unsigned 64 bit type in brg_types.h
+#endif
+
+#ifndef RETURN_VALUES
+# define RETURN_VALUES
+# if defined( DLL_EXPORT )
+# if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+# define VOID_RETURN __declspec( dllexport ) void __stdcall
+# define INT_RETURN __declspec( dllexport ) int __stdcall
+# elif defined( __GNUC__ )
+# define VOID_RETURN __declspec( __dllexport__ ) void
+# define INT_RETURN __declspec( __dllexport__ ) int
+# else
+# error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+# endif
+# elif defined( DLL_IMPORT )
+# if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+# define VOID_RETURN __declspec( dllimport ) void __stdcall
+# define INT_RETURN __declspec( dllimport ) int __stdcall
+# elif defined( __GNUC__ )
+# define VOID_RETURN __declspec( __dllimport__ ) void
+# define INT_RETURN __declspec( __dllimport__ ) int
+# else
+# error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+# endif
+# elif defined( __WATCOMC__ )
+# define VOID_RETURN void __cdecl
+# define INT_RETURN int __cdecl
+# else
+# define VOID_RETURN void
+# define INT_RETURN int
+# endif
+#endif
+
+/* These defines are used to declare buffers in a way that allows
+ faster operations on longer variables to be used. In all these
+ defines 'size' must be a power of 2 and >= 8
+
+ dec_unit_type(size,x) declares a variable 'x' of length
+ 'size' bits
+
+ dec_bufr_type(size,bsize,x) declares a buffer 'x' of length 'bsize'
+ bytes defined as an array of variables
+ each of 'size' bits (bsize must be a
+ multiple of size / 8)
+
+ ptr_cast(x,size) casts a pointer to a pointer to a
+ varaiable of length 'size' bits
+*/
+
+#define ui_type(size) uint_##size##t
+#define dec_unit_type(size,x) typedef ui_type(size) x
+#define dec_bufr_type(size,bsize,x) typedef ui_type(size) x[bsize / (size >> 3)]
+#define ptr_cast(x,size) ((ui_type(size)*)(x))
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif
Index: sys/contrib/skein/skein.h
===================================================================
--- /dev/null
+++ sys/contrib/skein/skein.h
@@ -0,0 +1,327 @@
+#ifndef _SKEIN_H_
+#define _SKEIN_H_ 1
+/**************************************************************************
+**
+** Interface declarations and 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
+**
+***************************************************************************/
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+#include <stddef.h> /* get size_t definition */
+#include "skein_port.h" /* get platform-specific definitions */
+
+enum
+ {
+ SKEIN_SUCCESS = 0, /* return codes from Skein calls */
+ SKEIN_FAIL = 1,
+ SKEIN_BAD_HASHLEN = 2
+ };
+
+#define SKEIN_MODIFIER_WORDS ( 2) /* number of modifier (tweak) words */
+
+#define SKEIN_256_STATE_WORDS ( 4)
+#define SKEIN_512_STATE_WORDS ( 8)
+#define SKEIN1024_STATE_WORDS (16)
+#define SKEIN_MAX_STATE_WORDS (16)
+
+#define SKEIN_256_STATE_BYTES ( 8*SKEIN_256_STATE_WORDS)
+#define SKEIN_512_STATE_BYTES ( 8*SKEIN_512_STATE_WORDS)
+#define SKEIN1024_STATE_BYTES ( 8*SKEIN1024_STATE_WORDS)
+
+#define SKEIN_256_STATE_BITS (64*SKEIN_256_STATE_WORDS)
+#define SKEIN_512_STATE_BITS (64*SKEIN_512_STATE_WORDS)
+#define SKEIN1024_STATE_BITS (64*SKEIN1024_STATE_WORDS)
+
+#define SKEIN_256_BLOCK_BYTES ( 8*SKEIN_256_STATE_WORDS)
+#define SKEIN_512_BLOCK_BYTES ( 8*SKEIN_512_STATE_WORDS)
+#define SKEIN1024_BLOCK_BYTES ( 8*SKEIN1024_STATE_WORDS)
+
+typedef struct
+ {
+ size_t hashBitLen; /* size of hash result, in bits */
+ size_t bCnt; /* current byte count in buffer b[] */
+ u64b_t T[SKEIN_MODIFIER_WORDS]; /* tweak words: T[0]=byte cnt, T[1]=flags */
+ } Skein_Ctxt_Hdr_t;
+
+typedef struct /* 256-bit Skein hash context structure */
+ {
+ Skein_Ctxt_Hdr_t h; /* common header context variables */
+ u64b_t X[SKEIN_256_STATE_WORDS]; /* chaining variables */
+ u08b_t b[SKEIN_256_BLOCK_BYTES]; /* partial block buffer (8-byte aligned) */
+ } Skein_256_Ctxt_t;
+
+typedef struct /* 512-bit Skein hash context structure */
+ {
+ Skein_Ctxt_Hdr_t h; /* common header context variables */
+ u64b_t X[SKEIN_512_STATE_WORDS]; /* chaining variables */
+ u08b_t b[SKEIN_512_BLOCK_BYTES]; /* partial block buffer (8-byte aligned) */
+ } Skein_512_Ctxt_t;
+
+typedef struct /* 1024-bit Skein hash context structure */
+ {
+ Skein_Ctxt_Hdr_t h; /* common header context variables */
+ u64b_t X[SKEIN1024_STATE_WORDS]; /* chaining variables */
+ u08b_t b[SKEIN1024_BLOCK_BYTES]; /* partial block buffer (8-byte aligned) */
+ } Skein1024_Ctxt_t;
+
+/* Skein APIs for (incremental) "straight hashing" */
+int Skein_256_Init (Skein_256_Ctxt_t *ctx, size_t hashBitLen);
+int Skein_512_Init (Skein_512_Ctxt_t *ctx, size_t hashBitLen);
+int Skein1024_Init (Skein1024_Ctxt_t *ctx, size_t hashBitLen);
+
+int Skein_256_Update(Skein_256_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+int Skein_512_Update(Skein_512_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+int Skein1024_Update(Skein1024_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+
+int Skein_256_Final (Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein_512_Final (Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein1024_Final (Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+
+/*
+** Skein APIs for "extended" initialization: MAC keys, tree hashing.
+** After an InitExt() call, just use Update/Final calls as with Init().
+**
+** Notes: Same parameters as _Init() calls, plus treeInfo/key/keyBytes.
+** When keyBytes == 0 and treeInfo == SKEIN_SEQUENTIAL,
+** the results of InitExt() are identical to calling Init().
+** The function Init() may be called once to "precompute" the IV for
+** a given hashBitLen value, then by saving a copy of the context
+** the IV computation may be avoided in later calls.
+** Similarly, the function InitExt() may be called once per MAC key
+** to precompute the MAC IV, then a copy of the context saved and
+** reused for each new MAC computation.
+**/
+int Skein_256_InitExt(Skein_256_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+int Skein_512_InitExt(Skein_512_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+int Skein1024_InitExt(Skein1024_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+
+/*
+** Skein APIs for MAC and tree hash:
+** Final_Pad: pad, do final block, but no OUTPUT type
+** Output: do just the output stage
+*/
+int Skein_256_Final_Pad(Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein_512_Final_Pad(Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein1024_Final_Pad(Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+
+#ifndef SKEIN_TREE_HASH
+#define SKEIN_TREE_HASH (1)
+#endif
+#if SKEIN_TREE_HASH
+int Skein_256_Output (Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein_512_Output (Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein1024_Output (Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+#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 */
+#define SKEIN_T1_BIT(BIT) ((BIT) - 64) /* offset 64 because it's the second word */
+
+#define SKEIN_T1_POS_TREE_LVL SKEIN_T1_BIT(112) /* bits 112..118: level in hash tree */
+#define SKEIN_T1_POS_BIT_PAD SKEIN_T1_BIT(119) /* bit 119 : partial final input byte */
+#define SKEIN_T1_POS_BLK_TYPE SKEIN_T1_BIT(120) /* bits 120..125: type field */
+#define SKEIN_T1_POS_FIRST SKEIN_T1_BIT(126) /* bits 126 : first block flag */
+#define SKEIN_T1_POS_FINAL SKEIN_T1_BIT(127) /* bit 127 : final block flag */
+
+/* tweak word T[1]: flag bit definition(s) */
+#define SKEIN_T1_FLAG_FIRST (((u64b_t) 1 ) << SKEIN_T1_POS_FIRST)
+#define SKEIN_T1_FLAG_FINAL (((u64b_t) 1 ) << SKEIN_T1_POS_FINAL)
+#define SKEIN_T1_FLAG_BIT_PAD (((u64b_t) 1 ) << SKEIN_T1_POS_BIT_PAD)
+
+/* tweak word T[1]: tree level bit field mask */
+#define SKEIN_T1_TREE_LVL_MASK (((u64b_t)0x7F) << SKEIN_T1_POS_TREE_LVL)
+#define SKEIN_T1_TREE_LEVEL(n) (((u64b_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 digital 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) (((u64b_t) (SKEIN_BLK_TYPE_##T)) << SKEIN_T1_POS_BLK_TYPE)
+#define SKEIN_T1_BLK_TYPE_KEY SKEIN_T1_BLK_TYPE(KEY) /* key, for MAC and KDF */
+#define SKEIN_T1_BLK_TYPE_CFG SKEIN_T1_BLK_TYPE(CFG) /* configuration block */
+#define SKEIN_T1_BLK_TYPE_PERS SKEIN_T1_BLK_TYPE(PERS) /* personalization string */
+#define SKEIN_T1_BLK_TYPE_PK SKEIN_T1_BLK_TYPE(PK) /* public key (for digital signature hashing) */
+#define SKEIN_T1_BLK_TYPE_KDF SKEIN_T1_BLK_TYPE(KDF) /* key identifier for KDF */
+#define SKEIN_T1_BLK_TYPE_NONCE SKEIN_T1_BLK_TYPE(NONCE)/* nonce for PRNG */
+#define SKEIN_T1_BLK_TYPE_MSG SKEIN_T1_BLK_TYPE(MSG) /* message processing */
+#define SKEIN_T1_BLK_TYPE_OUT SKEIN_T1_BLK_TYPE(OUT) /* output stage */
+#define SKEIN_T1_BLK_TYPE_MASK SKEIN_T1_BLK_TYPE(MASK) /* field bit 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) + (((u64b_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 (((u64b_t) 0xFF) << SKEIN_CFG_TREE_LEAF_SIZE_POS)
+#define SKEIN_CFG_TREE_NODE_SIZE_MSK (((u64b_t) 0xFF) << SKEIN_CFG_TREE_NODE_SIZE_POS)
+#define SKEIN_CFG_TREE_MAX_LEVEL_MSK (((u64b_t) 0xFF) << SKEIN_CFG_TREE_MAX_LEVEL_POS)
+
+#define SKEIN_CFG_TREE_INFO(leaf,node,maxLvl) \
+ ( (((u64b_t)(leaf )) << SKEIN_CFG_TREE_LEAF_SIZE_POS) | \
+ (((u64b_t)(node )) << SKEIN_CFG_TREE_NODE_SIZE_POS) | \
+ (((u64b_t)(maxLvl)) << SKEIN_CFG_TREE_MAX_LEVEL_POS) )
+
+#define SKEIN_CFG_TREE_INFO_SEQUENTIAL SKEIN_CFG_TREE_INFO(0,0,0) /* use as treeInfo in InitExt() call for sequential processing */
+
+/*
+** 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) {(ctxPtr)->h.T[TWK_NUM] = (tVal);}
+
+#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) \
+ { \
+ Skein_Set_T0(ctxPtr,(T0)); \
+ Skein_Set_T1(ctxPtr,(T1)); \
+ }
+
+#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) \
+ { Skein_Set_T0_T1(ctxPtr,0,SKEIN_T1_FLAG_FIRST | SKEIN_T1_BLK_TYPE_##BLK_TYPE); (ctxPtr)->h.bCnt=0; }
+
+#define Skein_Clear_First_Flag(hdr) { (hdr).T[1] &= ~SKEIN_T1_FLAG_FIRST; }
+#define Skein_Set_Bit_Pad_Flag(hdr) { (hdr).T[1] |= SKEIN_T1_FLAG_BIT_PAD; }
+
+#define Skein_Set_Tree_Level(hdr,height) { (hdr).T[1] |= SKEIN_T1_TREE_LEVEL(height);}
+
+/*****************************************************************
+** "Internal" Skein definitions for debugging and error checking
+******************************************************************/
+#ifdef SKEIN_DEBUG /* examine/display intermediate values? */
+#include "skein_debug.h"
+#else /* default is no callouts */
+#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)
+#endif
+
+#ifndef SKEIN_ERR_CHECK /* run-time checks (e.g., bad params, uninitialized context)? */
+#define Skein_Assert(x,retCode)/* default: ignore all Asserts, for performance */
+#define Skein_assert(x)
+#elif defined(SKEIN_ASSERT)
+#include <assert.h>
+#define Skein_Assert(x,retCode) assert(x)
+#define Skein_assert(x) assert(x)
+#else
+#include <assert.h>
+#define Skein_Assert(x,retCode) { if (!(x)) return retCode; } /* caller error */
+#define Skein_assert(x) assert(x) /* internal error */
+#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
+ };
+
+#ifndef SKEIN_ROUNDS
+#define SKEIN_256_ROUNDS_TOTAL (72) /* number of rounds for the different block sizes */
+#define SKEIN_512_ROUNDS_TOTAL (72)
+#define SKEIN1024_ROUNDS_TOTAL (80)
+#else /* allow command-line define in range 8*(5..14) */
+#define SKEIN_256_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS/100) + 5) % 10) + 5))
+#define SKEIN_512_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS/ 10) + 5) % 10) + 5))
+#define SKEIN1024_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS ) + 5) % 10) + 5))
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* ifndef _SKEIN_H_ */
Index: sys/contrib/skein/skein.c
===================================================================
--- /dev/null
+++ sys/contrib/skein/skein.c
@@ -0,0 +1,753 @@
+/***********************************************************************
+**
+** Implementation of the Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+
+#define SKEIN_PORT_CODE /* instantiate any code in skein_port.h */
+
+#include <string.h> /* get the memcpy/memset functions */
+#include "skein.h" /* get the Skein API definitions */
+#include "skein_iv.h" /* get precomputed IVs */
+
+/*****************************************************************/
+/* External function to process blkCnt (nonzero) full block(s) of data. */
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void Skein1024_Process_Block(Skein1024_Ctxt_t *ctx,const u08b_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
+ {
+ u08b_t b[SKEIN_256_STATE_BYTES];
+ u64b_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: memcpy(ctx->X,SKEIN_256_IV_256,sizeof(ctx->X)); break;
+ case 224: memcpy(ctx->X,SKEIN_256_IV_224,sizeof(ctx->X)); break;
+ case 160: memcpy(ctx->X,SKEIN_256_IV_160,sizeof(ctx->X)); break;
+ case 128: memcpy(ctx->X,SKEIN_256_IV_128,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) */
+ Skein_Start_New_Type(ctx,CFG_FINAL); /* set tweaks: T0=0; T1=CFG | FINAL */
+
+ cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER); /* set the schema, version */
+ cfg.w[1] = Skein_Swap64(hashBitLen); /* hash result length in bits */
+ cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+ memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+ /* compute the initial chaining values from config block */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the chaining variables */
+ 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,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+ {
+ union
+ {
+ u08b_t b[SKEIN_256_STATE_BYTES];
+ u64b_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? */
+ {
+ memset(ctx->X,0,sizeof(ctx->X)); /* no key: use all zeroes as key for config block */
+ }
+ else /* here to pre-process a key */
+ {
+ Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+ /* do a mini-Init right here */
+ ctx->h.hashBitLen=8*sizeof(ctx->X); /* set output hash bit count = state size */
+ Skein_Start_New_Type(ctx,KEY); /* set tweaks: T0 = 0; T1 = KEY type */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the initial chaining variables */
+ Skein_256_Update(ctx,key,keyBytes); /* hash the key */
+ Skein_256_Final_Pad(ctx,cfg.b); /* put result into cfg.b[] */
+ memcpy(ctx->X,cfg.b,sizeof(cfg.b)); /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+ {
+ uint_t i;
+ for (i=0;i<SKEIN_256_STATE_WORDS;i++) /* convert key bytes to context words */
+ 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);
+
+ memset(&cfg.w,0,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 */
+ cfg.w[2] = Skein_Swap64(treeInfo); /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+ 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 u08b_t *msg, size_t msgByteCnt)
+ {
+ size_t n;
+
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* process full blocks, if any */
+ if (msgByteCnt + ctx->h.bCnt > SKEIN_256_BLOCK_BYTES)
+ {
+ if (ctx->h.bCnt) /* finish up any buffered message data */
+ {
+ n = SKEIN_256_BLOCK_BYTES - ctx->h.bCnt; /* # bytes free in buffer b[] */
+ if (n)
+ {
+ Skein_assert(n < msgByteCnt); /* check on our logic here */
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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)
+ {
+ n = (msgByteCnt-1) / SKEIN_256_BLOCK_BYTES; /* number of full blocks to process */
+ 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);
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_256_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+
+ Skein_256_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_256_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_256_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_256_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_256_API_CodeSize(void)
+ {
+ return ((u08b_t *) Skein_256_API_CodeSize) -
+ ((u08b_t *) Skein_256_Init);
+ }
+#endif
+
+/*****************************************************************/
+/* 512-bit Skein */
+/*****************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a straight hashing operation */
+int Skein_512_Init(Skein_512_Ctxt_t *ctx, size_t hashBitLen)
+ {
+ union
+ {
+ u08b_t b[SKEIN_512_STATE_BYTES];
+ u64b_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: memcpy(ctx->X,SKEIN_512_IV_512,sizeof(ctx->X)); break;
+ case 384: memcpy(ctx->X,SKEIN_512_IV_384,sizeof(ctx->X)); break;
+ case 256: memcpy(ctx->X,SKEIN_512_IV_256,sizeof(ctx->X)); break;
+ case 224: memcpy(ctx->X,SKEIN_512_IV_224,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) */
+ Skein_Start_New_Type(ctx,CFG_FINAL); /* set tweaks: T0=0; T1=CFG | FINAL */
+
+ cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER); /* set the schema, version */
+ cfg.w[1] = Skein_Swap64(hashBitLen); /* hash result length in bits */
+ cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+ memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+ /* compute the initial chaining values from config block */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the chaining variables */
+ 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,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+ {
+ union
+ {
+ u08b_t b[SKEIN_512_STATE_BYTES];
+ u64b_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? */
+ {
+ memset(ctx->X,0,sizeof(ctx->X)); /* no key: use all zeroes as key for config block */
+ }
+ else /* here to pre-process a key */
+ {
+ Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+ /* do a mini-Init right here */
+ ctx->h.hashBitLen=8*sizeof(ctx->X); /* set output hash bit count = state size */
+ Skein_Start_New_Type(ctx,KEY); /* set tweaks: T0 = 0; T1 = KEY type */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the initial chaining variables */
+ Skein_512_Update(ctx,key,keyBytes); /* hash the key */
+ Skein_512_Final_Pad(ctx,cfg.b); /* put result into cfg.b[] */
+ memcpy(ctx->X,cfg.b,sizeof(cfg.b)); /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+ {
+ uint_t i;
+ for (i=0;i<SKEIN_512_STATE_WORDS;i++) /* convert key bytes to context words */
+ 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);
+
+ memset(&cfg.w,0,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 */
+ cfg.w[2] = Skein_Swap64(treeInfo); /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+ 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 u08b_t *msg, size_t msgByteCnt)
+ {
+ size_t n;
+
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* process full blocks, if any */
+ if (msgByteCnt + ctx->h.bCnt > SKEIN_512_BLOCK_BYTES)
+ {
+ if (ctx->h.bCnt) /* finish up any buffered message data */
+ {
+ n = SKEIN_512_BLOCK_BYTES - ctx->h.bCnt; /* # bytes free in buffer b[] */
+ if (n)
+ {
+ Skein_assert(n < msgByteCnt); /* check on our logic here */
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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)
+ {
+ n = (msgByteCnt-1) / SKEIN_512_BLOCK_BYTES; /* number of full blocks to process */
+ 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);
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_512_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+
+ Skein_512_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_512_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_512_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_512_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_512_API_CodeSize(void)
+ {
+ return ((u08b_t *) Skein_512_API_CodeSize) -
+ ((u08b_t *) Skein_512_Init);
+ }
+#endif
+
+/*****************************************************************/
+/* 1024-bit Skein */
+/*****************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a straight hashing operation */
+int Skein1024_Init(Skein1024_Ctxt_t *ctx, size_t hashBitLen)
+ {
+ union
+ {
+ u08b_t b[SKEIN1024_STATE_BYTES];
+ u64b_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: memcpy(ctx->X,SKEIN1024_IV_512 ,sizeof(ctx->X)); break;
+ case 384: memcpy(ctx->X,SKEIN1024_IV_384 ,sizeof(ctx->X)); break;
+ case 1024: memcpy(ctx->X,SKEIN1024_IV_1024,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) */
+ Skein_Start_New_Type(ctx,CFG_FINAL); /* set tweaks: T0=0; T1=CFG | FINAL */
+
+ cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER); /* set the schema, version */
+ cfg.w[1] = Skein_Swap64(hashBitLen); /* hash result length in bits */
+ cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+ memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+ /* compute the initial chaining values from config block */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the chaining variables */
+ 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,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+ {
+ union
+ {
+ u08b_t b[SKEIN1024_STATE_BYTES];
+ u64b_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? */
+ {
+ memset(ctx->X,0,sizeof(ctx->X)); /* no key: use all zeroes as key for config block */
+ }
+ else /* here to pre-process a key */
+ {
+ Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+ /* do a mini-Init right here */
+ ctx->h.hashBitLen=8*sizeof(ctx->X); /* set output hash bit count = state size */
+ Skein_Start_New_Type(ctx,KEY); /* set tweaks: T0 = 0; T1 = KEY type */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the initial chaining variables */
+ Skein1024_Update(ctx,key,keyBytes); /* hash the key */
+ Skein1024_Final_Pad(ctx,cfg.b); /* put result into cfg.b[] */
+ memcpy(ctx->X,cfg.b,sizeof(cfg.b)); /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+ {
+ uint_t i;
+ for (i=0;i<SKEIN1024_STATE_WORDS;i++) /* convert key bytes to context words */
+ 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);
+
+ memset(&cfg.w,0,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 */
+ cfg.w[2] = Skein_Swap64(treeInfo); /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+ 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 u08b_t *msg, size_t msgByteCnt)
+ {
+ size_t n;
+
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* process full blocks, if any */
+ if (msgByteCnt + ctx->h.bCnt > SKEIN1024_BLOCK_BYTES)
+ {
+ if (ctx->h.bCnt) /* finish up any buffered message data */
+ {
+ n = SKEIN1024_BLOCK_BYTES - ctx->h.bCnt; /* # bytes free in buffer b[] */
+ if (n)
+ {
+ Skein_assert(n < msgByteCnt); /* check on our logic here */
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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)
+ {
+ n = (msgByteCnt-1) / SKEIN1024_BLOCK_BYTES; /* number of full blocks to process */
+ 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);
+ memcpy(&ctx->b[ctx->h.bCnt],msg,msgByteCnt);
+ ctx->h.bCnt += msgByteCnt;
+ }
+
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the result */
+int Skein1024_Final(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN1024_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+
+ Skein1024_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN1024_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein1024_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN1024_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein1024_API_CodeSize(void)
+ {
+ return ((u08b_t *) Skein1024_API_CodeSize) -
+ ((u08b_t *) Skein1024_Init);
+ }
+#endif
+
+/**************** 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, u08b_t *hashVal)
+ {
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+ Skein_256_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN_256_BLOCK_BYTES); /* "output" the state 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, u08b_t *hashVal)
+ {
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+ Skein_512_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN_512_BLOCK_BYTES); /* "output" the state bytes */
+
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int Skein1024_Final_Pad(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+ Skein1024_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN1024_BLOCK_BYTES); /* "output" the state bytes */
+
+ return SKEIN_SUCCESS;
+ }
+
+#if SKEIN_TREE_HASH
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage */
+int Skein_256_Output(Skein_256_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_256_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_256_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_256_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_256_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage */
+int Skein_512_Output(Skein_512_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_512_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_512_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_512_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_512_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage */
+int Skein1024_Output(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN1024_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN1024_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein1024_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN1024_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+#endif
Index: sys/contrib/skein/skein_block.c
===================================================================
--- /dev/null
+++ sys/contrib/skein/skein_block.c
@@ -0,0 +1,689 @@
+/***********************************************************************
+**
+** 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]
+**
+************************************************************************/
+
+#include <string.h>
+#include "skein.h"
+
+#ifndef SKEIN_USE_ASM
+#define SKEIN_USE_ASM (0) /* default is all C code (no ASM) */
+#endif
+
+#ifndef SKEIN_LOOP
+#define SKEIN_LOOP 001 /* default: unroll 256 and 512, but not 1024 */
+#endif
+
+#define BLK_BITS (WCNT*64) /* some useful definitions for code here */
+#define KW_TWK_BASE (0)
+#define KW_KEY_BASE (3)
+#define ks (kw + KW_KEY_BASE)
+#define ts (kw + KW_TWK_BASE)
+
+#ifdef SKEIN_DEBUG
+#define DebugSaveTweak(ctx) { ctx->h.T[0] = ts[0]; ctx->h.T[1] = ts[1]; }
+#else
+#define DebugSaveTweak(ctx)
+#endif
+
+/***************************** Skein_256 ******************************/
+#if !(SKEIN_USE_ASM & 256)
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_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;
+ u64b_t kw[WCNT+4+RCNT*2]; /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+ u64b_t kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
+#endif
+ u64b_t X0,X1,X2,X3; /* local copy of context vars, for speed */
+ u64b_t w [WCNT]; /* local copy of input block */
+#ifdef SKEIN_DEBUG
+ const u64b_t *Xptr[4]; /* use for debugging (help compiler put Xn in registers) */
+ 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];
+
+ Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+ 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);
+
+ for (r=1;r < 2*RCNT;r+=2*SKEIN_UNROLL_256) /* loop thru it */
+#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 ((u08b_t *) Skein_256_Process_Block_CodeSize) -
+ ((u08b_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 u08b_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;
+ u64b_t kw[WCNT+4+RCNT*2]; /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+ u64b_t kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
+#endif
+ u64b_t X0,X1,X2,X3,X4,X5,X6,X7; /* local copy of vars, for speed */
+ u64b_t w [WCNT]; /* local copy of input block */
+#ifdef SKEIN_DEBUG
+ const u64b_t *Xptr[8]; /* use for debugging (help compiler put Xn in registers) */
+ 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];
+
+ Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+ 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);
+
+ for (r=1;r < 2*RCNT;r+=2*SKEIN_UNROLL_512) /* loop thru it */
+#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 ((u08b_t *) Skein_512_Process_Block_CodeSize) -
+ ((u08b_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 u08b_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;
+ u64b_t kw[WCNT+4+RCNT*2]; /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+ u64b_t kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
+#endif
+
+ u64b_t X00,X01,X02,X03,X04,X05,X06,X07, /* local copy of vars, for speed */
+ X08,X09,X10,X11,X12,X13,X14,X15;
+ u64b_t w [WCNT]; /* local copy of input block */
+#ifdef SKEIN_DEBUG
+ const u64b_t *Xptr[16]; /* use for debugging (help compiler put Xn in registers) */
+ 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];
+
+ Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+ 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);
+
+ for (r=1;r <= 2*RCNT;r+=2*SKEIN_UNROLL_1024) /* loop thru it */
+#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 ((u08b_t *) Skein1024_Process_Block_CodeSize) -
+ ((u08b_t *) Skein1024_Process_Block);
+ }
+uint_t Skein1024_Unroll_Cnt(void)
+ {
+ return SKEIN_UNROLL_1024;
+ }
+#endif
+#endif
Index: sys/contrib/skein/skein_debug.h
===================================================================
--- /dev/null
+++ sys/contrib/skein/skein_debug.h
@@ -0,0 +1,48 @@
+#ifndef _SKEIN_DEBUG_H_
+#define _SKEIN_DEBUG_H_
+/***********************************************************************
+**
+** Interface definitions for Skein hashing debug output.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+
+#ifdef SKEIN_DEBUG
+/* callout functions used inside Skein code */
+void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,const u08b_t *blkPtr,
+ const u64b_t *wPtr,const u64b_t *ksPtr,const u64b_t *tsPtr);
+void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X);
+void Skein_Show_R_Ptr(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X_ptr[]);
+void Skein_Show_Final(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t cnt,const u08b_t *outPtr);
+void Skein_Show_Key (uint_t bits,const Skein_Ctxt_Hdr_t *h,const u08b_t *key,size_t keyBytes);
+
+extern uint_t skein_DebugFlag; /* flags to control debug output (0 --> none) */
+
+#define SKEIN_RND_SPECIAL (1000u)
+#define SKEIN_RND_KEY_INITIAL (SKEIN_RND_SPECIAL+0u)
+#define SKEIN_RND_KEY_INJECT (SKEIN_RND_SPECIAL+1u)
+#define SKEIN_RND_FEED_FWD (SKEIN_RND_SPECIAL+2u)
+
+/* flag bits: skein_DebugFlag */
+#define SKEIN_DEBUG_KEY (1u << 1) /* show MAC key */
+#define SKEIN_DEBUG_CONFIG (1u << 2) /* show config block processing */
+#define SKEIN_DEBUG_STATE (1u << 3) /* show input state during Show_Block() */
+#define SKEIN_DEBUG_TWEAK (1u << 4) /* show input state during Show_Block() */
+#define SKEIN_DEBUG_KEYSCHED (1u << 5) /* show expanded key schedule */
+#define SKEIN_DEBUG_INPUT_64 (1u << 6) /* show input block as 64-bit words */
+#define SKEIN_DEBUG_INPUT_08 (1u << 7) /* show input block as 8-bit bytes */
+#define SKEIN_DEBUG_INJECT (1u << 8) /* show state after key injection & feedforward points */
+#define SKEIN_DEBUG_ROUNDS (1u << 9) /* show state after all rounds */
+#define SKEIN_DEBUG_FINAL (1u <<10) /* show final output of Skein */
+#define SKEIN_DEBUG_HDR (1u <<11) /* show block header */
+#define SKEIN_DEBUG_THREEFISH (1u <<12) /* use Threefish name instead of Skein */
+#define SKEIN_DEBUG_PERMUTE (1u <<13) /* use word permutations */
+#define SKEIN_DEBUG_ALL ((~0u) & ~(SKEIN_DEBUG_THREEFISH | SKEIN_DEBUG_PERMUTE))
+#define THREEFISH_DEBUG_ALL (SKEIN_DEBUG_ALL | SKEIN_DEBUG_THREEFISH)
+
+#endif /* SKEIN_DEBUG */
+
+#endif /* _SKEIN_DEBUG_H_ */
Index: sys/contrib/skein/skein_debug.c
===================================================================
--- /dev/null
+++ sys/contrib/skein/skein_debug.c
@@ -0,0 +1,247 @@
+/***********************************************************************
+**
+** Debug output functions for Skein hashing.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+#include <stdio.h>
+
+#ifdef SKEIN_DEBUG /* only instantiate this code if SKEIN_DEBUG is on */
+#include "skein.h"
+
+static const char INDENT[] = " "; /* how much to indent on new line */
+
+uint_t skein_DebugFlag = 0; /* off by default. Must be set externally */
+
+static void Show64_step(size_t cnt,const u64b_t *X,size_t step)
+ {
+ size_t i,j;
+ for (i=j=0;i < cnt;i++,j+=step)
+ {
+ if (i % 4 == 0) printf(INDENT);
+ printf(" %08X.%08X ",(uint_32t)(X[j] >> 32),(uint_32t)X[j]);
+ if (i % 4 == 3 || i==cnt-1) printf("\n");
+ fflush(stdout);
+ }
+ }
+
+#define Show64(cnt,X) Show64_step(cnt,X,1)
+
+static void Show64_flag(size_t cnt,const u64b_t *X)
+ {
+ size_t xptr = (size_t) X;
+ size_t step = (xptr & 1) ? 2 : 1;
+ if (step != 1)
+ {
+ X = (const u64b_t *) (xptr & ~1);
+ }
+ Show64_step(cnt,X,step);
+ }
+
+static void Show08(size_t cnt,const u08b_t *b)
+ {
+ size_t i;
+ for (i=0;i < cnt;i++)
+ {
+ if (i %16 == 0) printf(INDENT);
+ else if (i % 4 == 0) printf(" ");
+ printf(" %02X",b[i]);
+ if (i %16 == 15 || i==cnt-1) printf("\n");
+ fflush(stdout);
+ }
+ }
+
+static const char *AlgoHeader(uint_t bits)
+ {
+ if (skein_DebugFlag & SKEIN_DEBUG_THREEFISH)
+ switch (bits)
+ {
+ case 256: return ":Threefish-256: ";
+ case 512: return ":Threefish-512: ";
+ case 1024: return ":Threefish-1024:";
+ }
+ else
+ switch (bits)
+ {
+ case 256: return ":Skein-256: ";
+ case 512: return ":Skein-512: ";
+ case 1024: return ":Skein-1024:";
+ }
+ return NULL;
+ }
+
+void Skein_Show_Final(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t cnt,const u08b_t *outPtr)
+ {
+ if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+ if (skein_DebugFlag & SKEIN_DEBUG_FINAL)
+ {
+ printf("\n%s Final output=\n",AlgoHeader(bits));
+ Show08(cnt,outPtr);
+ printf(" ++++++++++\n");
+ fflush(stdout);
+ }
+ }
+
+/* show state after a round (or "pseudo-round") */
+void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X)
+ {
+ static uint_t injectNum=0; /* not multi-thread safe! */
+
+ if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+ if (skein_DebugFlag)
+ {
+ if (r >= SKEIN_RND_SPECIAL)
+ { /* a key injection (or feedforward) point */
+ injectNum = (r == SKEIN_RND_KEY_INITIAL) ? 0 : injectNum+1;
+ if ( skein_DebugFlag & SKEIN_DEBUG_INJECT ||
+ ((skein_DebugFlag & SKEIN_DEBUG_FINAL) && r == SKEIN_RND_FEED_FWD))
+ {
+ printf("\n%s",AlgoHeader(bits));
+ switch (r)
+ {
+ case SKEIN_RND_KEY_INITIAL:
+ printf(" [state after initial key injection]");
+ break;
+ case SKEIN_RND_KEY_INJECT:
+ printf(" [state after key injection #%02d]",injectNum);
+ break;
+ case SKEIN_RND_FEED_FWD:
+ printf(" [state after plaintext feedforward]");
+ injectNum = 0;
+ break;
+ }
+ printf("=\n");
+ Show64(bits/64,X);
+ if (r== SKEIN_RND_FEED_FWD)
+ printf(" ----------\n");
+ }
+ }
+ else if (skein_DebugFlag & SKEIN_DEBUG_ROUNDS)
+ {
+ uint_t j;
+ u64b_t p[SKEIN_MAX_STATE_WORDS];
+ const u08b_t *perm;
+ const static u08b_t PERM_256 [4][ 4] = { { 0,1,2,3 }, { 0,3,2,1 }, { 0,1,2,3 }, { 0,3,2,1 } };
+ const static u08b_t PERM_512 [4][ 8] = { { 0,1,2,3,4,5,6,7 },
+ { 2,1,4,7,6,5,0,3 },
+ { 4,1,6,3,0,5,2,7 },
+ { 6,1,0,7,2,5,4,3 }
+ };
+ const static u08b_t PERM_1024[4][16] = { { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,10,11,12,13,14,15 },
+ { 0, 9, 2,13, 6,11, 4,15,10, 7,12, 3,14, 5, 8, 1 },
+ { 0, 7, 2, 5, 4, 3, 6, 1,12,15,14,13, 8,11,10, 9 },
+ { 0,15, 2,11, 6,13, 4, 9,14, 1, 8, 5,10, 3,12, 7 }
+ };
+
+ if ((skein_DebugFlag & SKEIN_DEBUG_PERMUTE) && (r & 3))
+ {
+ printf("\n%s [state after round %2d (permuted)]=\n",AlgoHeader(bits),(int)r);
+ switch (bits)
+ {
+ case 256: perm = PERM_256 [r&3]; break;
+ case 512: perm = PERM_512 [r&3]; break;
+ default: perm = PERM_1024[r&3]; break;
+ }
+ for (j=0;j<bits/64;j++)
+ p[j] = X[perm[j]];
+ Show64(bits/64,p);
+ }
+ else
+ {
+ printf("\n%s [state after round %2d]=\n",AlgoHeader(bits),(int)r);
+ Show64(bits/64,X);
+ }
+ }
+ }
+ }
+
+/* show state after a round (or "pseudo-round"), given a list of pointers */
+void Skein_Show_R_Ptr(uint_t bits,const Skein_Ctxt_Hdr_t *h,size_t r,const u64b_t *X_ptr[])
+ {
+ uint_t i;
+ u64b_t X[SKEIN_MAX_STATE_WORDS];
+
+ for (i=0;i<bits/64;i++) /* copy over the words */
+ X[i] = X_ptr[i][0];
+ Skein_Show_Round(bits,h,r,X);
+ }
+
+
+/* show the state at the start of a block */
+void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,const u08b_t *blkPtr,
+ const u64b_t *wPtr, const u64b_t *ksPtr, const u64b_t *tsPtr)
+ {
+ uint_t n;
+ if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+ if (skein_DebugFlag)
+ {
+ if (skein_DebugFlag & SKEIN_DEBUG_HDR)
+ {
+ printf("\n%s Block: outBits=%4d. T0=%06X.",AlgoHeader(bits),(uint_t) h->hashBitLen,(uint_t)h->T[0]);
+ printf(" Type=");
+ n = (uint_t) ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) >> SKEIN_T1_POS_BLK_TYPE);
+ switch (n)
+ {
+ case SKEIN_BLK_TYPE_KEY: printf("KEY. "); break;
+ case SKEIN_BLK_TYPE_CFG: printf("CFG. "); break;
+ case SKEIN_BLK_TYPE_PERS: printf("PERS."); break;
+ case SKEIN_BLK_TYPE_PK : printf("PK. "); break;
+ case SKEIN_BLK_TYPE_KDF: printf("KDF. "); break;
+ case SKEIN_BLK_TYPE_MSG: printf("MSG. "); break;
+ case SKEIN_BLK_TYPE_OUT: printf("OUT. "); break;
+ default: printf("0x%02X.",n); break;
+ }
+ printf(" Flags=");
+ printf((h->T[1] & SKEIN_T1_FLAG_FIRST) ? " First":" ");
+ printf((h->T[1] & SKEIN_T1_FLAG_FINAL) ? " Final":" ");
+ printf((h->T[1] & SKEIN_T1_FLAG_BIT_PAD) ? " Pad" :" ");
+ n = (uint_t) ((h->T[1] & SKEIN_T1_TREE_LVL_MASK) >> SKEIN_T1_POS_TREE_LVL);
+ if (n)
+ printf(" TreeLevel = %02X",n);
+ printf("\n");
+ fflush(stdout);
+ }
+ if (skein_DebugFlag & SKEIN_DEBUG_TWEAK)
+ {
+ printf(" Tweak:\n");
+ Show64(2,h->T);
+ }
+ if (skein_DebugFlag & SKEIN_DEBUG_STATE)
+ {
+ printf(" %s words:\n",(skein_DebugFlag & SKEIN_DEBUG_THREEFISH)?"Key":"State");
+ Show64(bits/64,X);
+ }
+ if (skein_DebugFlag & SKEIN_DEBUG_KEYSCHED)
+ {
+ printf(" Tweak schedule:\n");
+ Show64_flag(3,tsPtr);
+ printf(" Key schedule:\n");
+ Show64_flag((bits/64)+1,ksPtr);
+ }
+ if (skein_DebugFlag & SKEIN_DEBUG_INPUT_64)
+ {
+ printf(" Input block (words):\n");
+ Show64(bits/64,wPtr);
+ }
+ if (skein_DebugFlag & SKEIN_DEBUG_INPUT_08)
+ {
+ printf(" Input block (bytes):\n");
+ Show08(bits/8,blkPtr);
+ }
+ }
+ }
+
+void Skein_Show_Key(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u08b_t *key,size_t keyBytes)
+ {
+ if (keyBytes)
+ if (skein_DebugFlag & SKEIN_DEBUG_CONFIG || ((h->T[1] & SKEIN_T1_BLK_TYPE_MASK) != SKEIN_T1_BLK_TYPE_CFG))
+ if (skein_DebugFlag & SKEIN_DEBUG_KEY)
+ {
+ printf("\n%s MAC key = %4u bytes\n",AlgoHeader(bits),(unsigned) keyBytes);
+ Show08(keyBytes,key);
+ }
+ }
+#endif
Index: sys/contrib/skein/skein_iv.h
===================================================================
--- /dev/null
+++ sys/contrib/skein/skein_iv.h
@@ -0,0 +1,199 @@
+#ifndef _SKEIN_IV_H_
+#define _SKEIN_IV_H_
+
+#include "skein.h" /* get Skein macros and types */
+
+/*
+***************** 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.
+**
+************************************************************
+**/
+
+#define MK_64 SKEIN_MK_64
+
+/* blkSize = 256 bits. hashSize = 128 bits */
+const u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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)
+ };
+
+#endif /* _SKEIN_IV_H_ */
Index: sys/contrib/skein/skein_port.h
===================================================================
--- /dev/null
+++ sys/contrib/skein/skein_port.h
@@ -0,0 +1,124 @@
+#ifndef _SKEIN_PORT_H_
+#define _SKEIN_PORT_H_
+/*******************************************************************
+**
+** 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.
+**
+********************************************************************/
+
+#include "brg_types.h" /* get integer type definitions */
+
+typedef unsigned int uint_t; /* native unsigned integer */
+typedef uint_8t u08b_t; /* 8-bit unsigned integer */
+typedef uint_64t u64b_t; /* 64-bit unsigned integer */
+
+#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 "brg_endian.h" /* get endianness selection */
+#if PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN
+ /* here for big-endian CPUs */
+#define SKEIN_NEED_SWAP (1)
+#elif PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN
+ /* 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) memcpy(dst08,src64,bCnt)
+#define Skein_Get64_LSB_First(dst64,src08,wCnt) memcpy(dst64,src08,8*(wCnt))
+#endif
+#else
+#error "Skein needs endianness setting!"
+#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) \
+ ( (( ((u64b_t)(w64)) & 0xFF) << 56) | \
+ (((((u64b_t)(w64)) >> 8) & 0xFF) << 48) | \
+ (((((u64b_t)(w64)) >>16) & 0xFF) << 40) | \
+ (((((u64b_t)(w64)) >>24) & 0xFF) << 32) | \
+ (((((u64b_t)(w64)) >>32) & 0xFF) << 24) | \
+ (((((u64b_t)(w64)) >>40) & 0xFF) << 16) | \
+ (((((u64b_t)(w64)) >>48) & 0xFF) << 8) | \
+ (((((u64b_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(u08b_t *dst,const u64b_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] = (u08b_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(u64b_t *dst,const u08b_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] = (((u64b_t) src[n ]) ) +
+ (((u64b_t) src[n+1]) << 8) +
+ (((u64b_t) src[n+2]) << 16) +
+ (((u64b_t) src[n+3]) << 24) +
+ (((u64b_t) src[n+4]) << 32) +
+ (((u64b_t) src[n+5]) << 40) +
+ (((u64b_t) src[n+6]) << 48) +
+ (((u64b_t) src[n+7]) << 56) ;
+ }
+#else
+ ; /* output only the function prototype */
+#endif
+#endif /* ifndef Skein_Get64_LSB_First */
+
+#endif /* ifndef _SKEIN_PORT_H_ */
Index: sys/crypto/skein/amd64/skein_block_asm.s
===================================================================
--- /dev/null
+++ sys/crypto/skein/amd64/skein_block_asm.s
@@ -0,0 +1,1328 @@
+#
+#----------------------------------------------------------------
+# 64-bit x86 assembler code (gnu as) for Skein block functions
+#
+# Author: Doug Whiting, Hifn/Exar
+#
+# This code is released to the public domain.
+#----------------------------------------------------------------
+#
+ .text
+ .altmacro
+ .psize 0,128 #list file has no page boundaries
+#
+_MASK_ALL_ = (256+512+1024) #all three algorithm bits
+_MAX_FRAME_ = 240
+#
+#################
+.ifndef SKEIN_USE_ASM
+_USE_ASM_ = _MASK_ALL_
+.else
+_USE_ASM_ = SKEIN_USE_ASM
+.endif
+#################
+.ifndef SKEIN_LOOP #configure loop unrolling
+_SKEIN_LOOP = 2 #default is fully unrolled for 256/512, twice for 1024
+.else
+_SKEIN_LOOP = SKEIN_LOOP
+ .irp _NN_,%_SKEIN_LOOP #only display loop unrolling if default changed on command line
+.print "+++ SKEIN_LOOP = \_NN_"
+ .endr
+.endif
+# the unroll counts (0 --> fully unrolled)
+SKEIN_UNROLL_256 = (_SKEIN_LOOP / 100) % 10
+SKEIN_UNROLL_512 = (_SKEIN_LOOP / 10) % 10
+SKEIN_UNROLL_1024 = (_SKEIN_LOOP ) % 10
+#
+SKEIN_ASM_UNROLL = 0
+ .irp _NN_,256,512,1024
+ .if (SKEIN_UNROLL_\_NN_) == 0
+SKEIN_ASM_UNROLL = SKEIN_ASM_UNROLL + \_NN_
+ .endif
+ .endr
+#################
+#
+.ifndef SKEIN_ROUNDS
+ROUNDS_256 = 72
+ROUNDS_512 = 72
+ROUNDS_1024 = 80
+.else
+ROUNDS_256 = 8*((((SKEIN_ROUNDS / 100) + 5) % 10) + 5)
+ROUNDS_512 = 8*((((SKEIN_ROUNDS / 10) + 5) % 10) + 5)
+ROUNDS_1024 = 8*((((SKEIN_ROUNDS ) + 5) % 10) + 5)
+# only display rounds if default size is changed on command line
+.irp _NN_,256,512,1024
+ .if _USE_ASM_ && \_NN_
+ .irp _RR_,%(ROUNDS_\_NN_)
+ .if _NN_ < 1024
+.print "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+ .else
+.print "+++ SKEIN_ROUNDS_\_NN_ = \_RR_"
+ .endif
+ .endr
+ .endif
+.endr
+.endif
+#################
+#
+.ifdef SKEIN_CODE_SIZE
+_SKEIN_CODE_SIZE = (1)
+.else
+.ifdef SKEIN_PERF #use code size if SKEIN_PERF is defined
+_SKEIN_CODE_SIZE = (1)
+.else
+_SKEIN_CODE_SIZE = (0)
+.endif
+.endif
+#
+#################
+#
+.ifndef SKEIN_DEBUG
+_SKEIN_DEBUG = 0
+.else
+_SKEIN_DEBUG = 1
+.endif
+#################
+#
+# define offsets of fields in hash context structure
+#
+HASH_BITS = 0 #bits of hash output
+BCNT = 8 + HASH_BITS #number of bytes in BUFFER[]
+TWEAK = 8 + BCNT #tweak values[0..1]
+X_VARS = 16 + TWEAK #chaining vars
+#
+#(Note: buffer[] in context structure is NOT needed here :-)
+#
+KW_PARITY = 0x1BD11BDAA9FC1A22 #overall parity of key schedule words
+FIRST_MASK = ~ (1 << 6)
+FIRST_MASK64= ~ (1 << 62)
+#
+# rotation constants for Skein
+#
+RC_256_0_0 = 14
+RC_256_0_1 = 16
+
+RC_256_1_0 = 52
+RC_256_1_1 = 57
+
+RC_256_2_0 = 23
+RC_256_2_1 = 40
+
+RC_256_3_0 = 5
+RC_256_3_1 = 37
+
+RC_256_4_0 = 25
+RC_256_4_1 = 33
+
+RC_256_5_0 = 46
+RC_256_5_1 = 12
+
+RC_256_6_0 = 58
+RC_256_6_1 = 22
+
+RC_256_7_0 = 32
+RC_256_7_1 = 32
+
+RC_512_0_0 = 46
+RC_512_0_1 = 36
+RC_512_0_2 = 19
+RC_512_0_3 = 37
+
+RC_512_1_0 = 33
+RC_512_1_1 = 27
+RC_512_1_2 = 14
+RC_512_1_3 = 42
+
+RC_512_2_0 = 17
+RC_512_2_1 = 49
+RC_512_2_2 = 36
+RC_512_2_3 = 39
+
+RC_512_3_0 = 44
+RC_512_3_1 = 9
+RC_512_3_2 = 54
+RC_512_3_3 = 56
+
+RC_512_4_0 = 39
+RC_512_4_1 = 30
+RC_512_4_2 = 34
+RC_512_4_3 = 24
+
+RC_512_5_0 = 13
+RC_512_5_1 = 50
+RC_512_5_2 = 10
+RC_512_5_3 = 17
+
+RC_512_6_0 = 25
+RC_512_6_1 = 29
+RC_512_6_2 = 39
+RC_512_6_3 = 43
+
+RC_512_7_0 = 8
+RC_512_7_1 = 35
+RC_512_7_2 = 56
+RC_512_7_3 = 22
+
+RC_1024_0_0 = 24
+RC_1024_0_1 = 13
+RC_1024_0_2 = 8
+RC_1024_0_3 = 47
+RC_1024_0_4 = 8
+RC_1024_0_5 = 17
+RC_1024_0_6 = 22
+RC_1024_0_7 = 37
+
+RC_1024_1_0 = 38
+RC_1024_1_1 = 19
+RC_1024_1_2 = 10
+RC_1024_1_3 = 55
+RC_1024_1_4 = 49
+RC_1024_1_5 = 18
+RC_1024_1_6 = 23
+RC_1024_1_7 = 52
+
+RC_1024_2_0 = 33
+RC_1024_2_1 = 4
+RC_1024_2_2 = 51
+RC_1024_2_3 = 13
+RC_1024_2_4 = 34
+RC_1024_2_5 = 41
+RC_1024_2_6 = 59
+RC_1024_2_7 = 17
+
+RC_1024_3_0 = 5
+RC_1024_3_1 = 20
+RC_1024_3_2 = 48
+RC_1024_3_3 = 41
+RC_1024_3_4 = 47
+RC_1024_3_5 = 28
+RC_1024_3_6 = 16
+RC_1024_3_7 = 25
+
+RC_1024_4_0 = 41
+RC_1024_4_1 = 9
+RC_1024_4_2 = 37
+RC_1024_4_3 = 31
+RC_1024_4_4 = 12
+RC_1024_4_5 = 47
+RC_1024_4_6 = 44
+RC_1024_4_7 = 30
+
+RC_1024_5_0 = 16
+RC_1024_5_1 = 34
+RC_1024_5_2 = 56
+RC_1024_5_3 = 51
+RC_1024_5_4 = 4
+RC_1024_5_5 = 53
+RC_1024_5_6 = 42
+RC_1024_5_7 = 41
+
+RC_1024_6_0 = 31
+RC_1024_6_1 = 44
+RC_1024_6_2 = 47
+RC_1024_6_3 = 46
+RC_1024_6_4 = 19
+RC_1024_6_5 = 42
+RC_1024_6_6 = 44
+RC_1024_6_7 = 25
+
+RC_1024_7_0 = 9
+RC_1024_7_1 = 48
+RC_1024_7_2 = 35
+RC_1024_7_3 = 52
+RC_1024_7_4 = 23
+RC_1024_7_5 = 31
+RC_1024_7_6 = 37
+RC_1024_7_7 = 20
+#
+# Input: reg
+# Output: <reg> <<< RC_BlkSize_roundNum_mixNum, BlkSize=256/512/1024
+#
+.macro RotL64 reg,BLK_SIZE,ROUND_NUM,MIX_NUM
+_RCNT_ = RC_\BLK_SIZE&_\ROUND_NUM&_\MIX_NUM
+ .if _RCNT_ #is there anything to do?
+ rolq $_RCNT_,%\reg
+ .endif
+.endm
+#
+#----------------------------------------------------------------
+#
+# MACROS: define local vars and configure stack
+#
+#----------------------------------------------------------------
+# declare allocated space on the stack
+.macro StackVar localName,localSize
+\localName = _STK_OFFS_
+_STK_OFFS_ = _STK_OFFS_+(\localSize)
+.endm #StackVar
+#
+#----------------------------------------------------------------
+#
+# MACRO: Configure stack frame, allocate local vars
+#
+.macro Setup_Stack BLK_BITS,KS_CNT,debugCnt
+ WCNT = (\BLK_BITS)/64
+#
+_PushCnt_ = 0 #save nonvolatile regs on stack
+ .irp _reg_,rbp,rbx,r12,r13,r14,r15
+ pushq %\_reg_
+_PushCnt_ = _PushCnt_ + 1 #track count to keep alignment
+ .endr
+#
+_STK_OFFS_ = 0 #starting offset from rsp
+ #---- local variables #<-- rsp
+ StackVar X_stk ,8*(WCNT) #local context vars
+ StackVar ksTwk ,8*3 #key schedule: tweak words
+ StackVar ksKey ,8*(WCNT)+8 #key schedule: key words
+ .if (SKEIN_ASM_UNROLL && (\BLK_BITS)) == 0
+ StackVar ksRot ,16*(\KS_CNT) #leave space for "rotation" to happen
+ .endif
+ StackVar Wcopy ,8*(WCNT) #copy of input block
+ .if _SKEIN_DEBUG
+ .if \debugCnt + 0 #temp location for debug X[] info
+ StackVar xDebug_\BLK_BITS ,8*(\debugCnt)
+ .endif
+ .endif
+ .if ((8*_PushCnt_ + _STK_OFFS_) % 8) == 0
+ StackVar align16,8 #keep 16-byte aligned (adjust for retAddr?)
+tmpStk_\BLK_BITS = align16 #use this
+ .endif
+ #---- saved caller parameters (from regs rdi, rsi, rdx, rcx)
+ StackVar ctxPtr ,8 #context ptr
+ StackVar blkPtr ,8 #pointer to block data
+ StackVar blkCnt ,8 #number of full blocks to process
+ StackVar bitAdd ,8 #bit count to add to tweak
+LOCAL_SIZE = _STK_OFFS_ #size of "local" vars
+ #----
+ StackVar savRegs,8*_PushCnt_ #saved registers
+ StackVar retAddr,8 #return address
+ #---- caller's stack frame (aligned mod 16)
+#
+# set up the stack frame pointer (rbp)
+#
+FRAME_OFFS = ksTwk + 128 #allow short (negative) offset to ksTwk, kwKey
+ .if FRAME_OFFS > _STK_OFFS_ #keep rbp in the "locals" range
+FRAME_OFFS = _STK_OFFS_
+ .endif
+F_O = -FRAME_OFFS
+#
+ #put some useful defines in the .lst file (for grep)
+__STK_LCL_SIZE_\BLK_BITS = LOCAL_SIZE
+__STK_TOT_SIZE_\BLK_BITS = _STK_OFFS_
+__STK_FRM_OFFS_\BLK_BITS = FRAME_OFFS
+#
+# Notes on stack frame setup:
+# * the most frequently used variable is X_stk[], based at [rsp+0]
+# * the next most used is the key schedule arrays, ksKey and ksTwk
+# so rbp is "centered" there, allowing short offsets to the key
+# schedule even in 1024-bit Skein case
+# * the Wcopy variables are infrequently accessed, but they have long
+# offsets from both rsp and rbp only in the 1024-bit case.
+# * all other local vars and calling parameters can be accessed
+# with short offsets, except in the 1024-bit case
+#
+ subq $LOCAL_SIZE,%rsp #make room for the locals
+ leaq FRAME_OFFS(%rsp),%rbp #maximize use of short offsets
+ movq %rdi, ctxPtr+F_O(%rbp) #save caller's parameters on the stack
+ movq %rsi, blkPtr+F_O(%rbp)
+ movq %rdx, blkCnt+F_O(%rbp)
+ movq %rcx, bitAdd+F_O(%rbp)
+#
+.endm #Setup_Stack
+#
+#----------------------------------------------------------------
+#
+.macro Reset_Stack
+ addq $LOCAL_SIZE,%rsp #get rid of locals (wipe??)
+ .irp _reg_,r15,r14,r13,r12,rbx,rbp
+ popq %\_reg_ #restore caller's regs
+_PushCnt_ = _PushCnt_ - 1
+ .endr
+ .if _PushCnt_
+ .error "Mismatched push/pops?"
+ .endif
+.endm # Reset_Stack
+#
+#----------------------------------------------------------------
+# macros to help debug internals
+#
+.if _SKEIN_DEBUG
+ .extern Skein_Show_Block #calls to C routines
+ .extern Skein_Show_Round
+#
+SKEIN_RND_SPECIAL = 1000
+SKEIN_RND_KEY_INITIAL = SKEIN_RND_SPECIAL+0
+SKEIN_RND_KEY_INJECT = SKEIN_RND_SPECIAL+1
+SKEIN_RND_FEED_FWD = SKEIN_RND_SPECIAL+2
+#
+.macro Skein_Debug_Block BLK_BITS
+#
+#void Skein_Show_Block(uint_t bits,const Skein_Ctxt_Hdr_t *h,const u64b_t *X,
+# const u08b_t *blkPtr, const u64b_t *wPtr,
+# const u64b_t *ksPtr,const u64b_t *tsPtr)
+#
+_NN_ = 0
+ .irp _reg_,rax,rcx,rdx,rsi,rdi,r8,r9,r10,r11
+ pushq %\_reg_ #save all volatile regs on tack before the call
+_NN_ = _NN_ + 1
+ .endr
+ # get and push call parameters
+ movq $\BLK_BITS ,%rdi #bits
+ movq ctxPtr+F_O(%rbp),%rsi #h (pointer)
+ leaq X_VARS (%rsi),%rdx #X (pointer)
+ movq blkPtr+F_O(%rbp),%rcx #blkPtr
+ leaq Wcopy +F_O(%rbp),%r8 #wPtr
+ leaq ksKey +F_O(%rbp),%r9 #key pointer
+ leaq ksTwk +F_O(%rbp),%rax #tweak pointer
+ pushq %rax # (pass on the stack)
+ call Skein_Show_Block #call external debug handler
+ addq $8*1,%rsp #discard parameters on stack
+ .if (_NN_ % 2 ) == 0 #check stack alignment
+ .error "Stack misalignment problem in Skein_Debug_Block_\_BLK_BITS"
+ .endif
+ .irp _reg_,r11,r10,r9,r8,rdi,rsi,rdx,rcx,rax
+ popq %\_reg_ #restore regs
+_NN_ = _NN_ - 1
+ .endr
+ .if _NN_
+ .error "Push/pop mismatch problem in Skein_Debug_Block_\_BLK_BITS"
+ .endif
+.endm # Skein_Debug_Block
+#
+# the macro to "call" to debug a round
+#
+.macro Skein_Debug_Round BLK_BITS,R,RDI_OFFS,afterOp
+ # call the appropriate (local) debug "function"
+ pushq %rdx #save rdx, so we can use it for round "number"
+ .if (SKEIN_ASM_UNROLL && \BLK_BITS) || (\R >= SKEIN_RND_SPECIAL)
+ movq $\R,%rdx
+ .else #compute round number using edi
+_rOffs_ = \RDI_OFFS + 0
+ .if \BLK_BITS == 1024
+ movq rIdx_offs+8(%rsp),%rdx #get rIdx off the stack (adjust for pushq rdx above)
+ leaq 1+(((\R)-1) && 3)+_rOffs_(,%rdx,4),%rdx
+ .else
+ leaq 1+(((\R)-1) && 3)+_rOffs_(,%rdi,4),%rdx
+ .endif
+ .endif
+ call Skein_Debug_Round_\BLK_BITS
+ popq %rdx #restore origianl rdx value
+#
+ afterOp
+.endm # Skein_Debug_Round
+.else #------- _SKEIN_DEBUG (dummy macros if debug not enabled)
+.macro Skein_Debug_Block BLK_BITS
+.endm
+#
+.macro Skein_Debug_Round BLK_BITS,R,RDI_OFFS,afterOp
+.endm
+#
+.endif # _SKEIN_DEBUG
+#
+#----------------------------------------------------------------
+#
+.macro addReg dstReg,srcReg_A,srcReg_B,useAddOp,immOffs
+ .if \immOffs + 0
+ leaq \immOffs(%\srcReg_A\srcReg_B,%\dstReg),%\dstReg
+ .elseif ((\useAddOp + 0) == 0)
+ .ifndef ASM_NO_LEA #lea seems to be faster on Core 2 Duo CPUs!
+ leaq (%\srcReg_A\srcReg_B,%\dstReg),%\dstReg
+ .else
+ addq %\srcReg_A\srcReg_B,%\dstReg
+ .endif
+ .else
+ addq %\srcReg_A\srcReg_B,%\dstReg
+ .endif
+.endm
+
+# keep Intel-style ordering here, to match addReg
+.macro xorReg dstReg,srcReg_A,srcReg_B
+ xorq %\srcReg_A\srcReg_B,%\dstReg
+.endm
+#
+#----------------------------------------------------------------
+#
+.macro C_label lName
+ \lName: #use both "genders" to work across linkage conventions
+_\lName:
+ .global \lName
+ .global _\lName
+.endm
+#
+#=================================== Skein_256 =============================================
+#
+.if _USE_ASM_ & 256
+#
+# void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+#
+# code
+#
+C_label Skein_256_Process_Block
+ Setup_Stack 256,((ROUNDS_256/8)+1)
+ movq TWEAK+8(%rdi),%r14
+ jmp Skein_256_block_loop
+ .p2align 4
+ # main hash loop for Skein_256
+Skein_256_block_loop:
+ #
+ # general register usage:
+ # RAX..RDX = X0..X3
+ # R08..R12 = ks[0..4]
+ # R13..R15 = ts[0..2]
+ # RSP, RBP = stack/frame pointers
+ # RDI = round counter or context pointer
+ # RSI = temp
+ #
+ movq TWEAK+0(%rdi) ,%r13
+ addq bitAdd+F_O(%rbp) ,%r13 #computed updated tweak value T0
+ movq %r14 ,%r15
+ xorq %r13 ,%r15 #now %r13.%r15 is set as the tweak
+
+ movq $KW_PARITY ,%r12
+ movq X_VARS+ 0(%rdi),%r8
+ movq X_VARS+ 8(%rdi),%r9
+ movq X_VARS+16(%rdi),%r10
+ movq X_VARS+24(%rdi),%r11
+ movq %r13,TWEAK+0(%rdi) #save updated tweak value ctx->h.T[0]
+ xorq %r8 ,%r12 #start accumulating overall parity
+
+ movq blkPtr +F_O(%rbp) ,%rsi #esi --> input block
+ xorq %r9 ,%r12
+ movq 0(%rsi) ,%rax #get X[0..3]
+ xorq %r10 ,%r12
+ movq 8(%rsi) ,%rbx
+ xorq %r11 ,%r12
+ movq 16(%rsi) ,%rcx
+ movq 24(%rsi) ,%rdx
+
+ movq %rax,Wcopy+ 0+F_O(%rbp) #save copy of input block
+ movq %rbx,Wcopy+ 8+F_O(%rbp)
+ movq %rcx,Wcopy+16+F_O(%rbp)
+ movq %rdx,Wcopy+24+F_O(%rbp)
+
+ addq %r8 ,%rax #initial key injection
+ addq %r9 ,%rbx
+ addq %r10,%rcx
+ addq %r11,%rdx
+ addq %r13,%rbx
+ addq %r14,%rcx
+
+.if _SKEIN_DEBUG
+ movq %r14,TWEAK+ 8(%rdi) #save updated tweak T[1] (start bit cleared?)
+ movq %r8 ,ksKey+ 0+F_O(%rbp) #save key schedule on stack for Skein_Debug_Block
+ movq %r9 ,ksKey+ 8+F_O(%rbp)
+ movq %r10,ksKey+16+F_O(%rbp)
+ movq %r11,ksKey+24+F_O(%rbp)
+ movq %r12,ksKey+32+F_O(%rbp)
+
+ movq %r13,ksTwk+ 0+F_O(%rbp)
+ movq %r14,ksTwk+ 8+F_O(%rbp)
+ movq %r15,ksTwk+16+F_O(%rbp)
+
+ movq %rax,X_stk + 0(%rsp) #save X[] on stack for Skein_Debug_Block
+ movq %rbx,X_stk + 8(%rsp)
+ movq %rcx,X_stk +16(%rsp)
+ movq %rdx,X_stk +24(%rsp)
+
+ Skein_Debug_Block 256 #debug dump
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INITIAL
+.endif
+#
+.if ((SKEIN_ASM_UNROLL & 256) == 0)
+ movq %r8 ,ksKey+40+F_O(%rbp) #save key schedule on stack for looping code
+ movq %r9 ,ksKey+ 8+F_O(%rbp)
+ movq %r10,ksKey+16+F_O(%rbp)
+ movq %r11,ksKey+24+F_O(%rbp)
+ movq %r12,ksKey+32+F_O(%rbp)
+
+ movq %r13,ksTwk+24+F_O(%rbp)
+ movq %r14,ksTwk+ 8+F_O(%rbp)
+ movq %r15,ksTwk+16+F_O(%rbp)
+.endif
+ addq $WCNT*8,%rsi #skip the block
+ movq %rsi,blkPtr +F_O(%rbp) #update block pointer
+ #
+ # now the key schedule is computed. Start the rounds
+ #
+.if SKEIN_ASM_UNROLL & 256
+_UNROLL_CNT = ROUNDS_256/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_256
+ .if ((ROUNDS_256/8) % _UNROLL_CNT)
+ .error "Invalid SKEIN_UNROLL_256"
+ .endif
+ xorq %rdi,%rdi #rdi = iteration count
+Skein_256_round_loop:
+.endif
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+ # all X and ks vars in regs # (ops to "rotate" ks vars, via mem, if not unrolled)
+ # round 4*_RBase_ + 0
+ addReg rax, rbx
+ RotL64 rbx, 256,%((4*_Rbase_+0) % 8),0
+ addReg rcx, rdx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*1+F_O(%rbp,%rdi,8),%r8
+ .endif
+ xorReg rbx, rax
+ RotL64 rdx, 256,%((4*_Rbase_+0) % 8),1
+ xorReg rdx, rcx
+ .if SKEIN_ASM_UNROLL & 256
+ .irp _r0_,%( 8+(_Rbase_+3) % 5)
+ .irp _r1_,%(13+(_Rbase_+2) % 3)
+ leaq (%r\_r0_,%r\_r1_),%rdi #precompute key injection value for %rcx
+ .endr
+ .endr
+ .endif
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksTwk+8*1+F_O(%rbp,%rdi,8),%r13
+ .endif
+ Skein_Debug_Round 256,%(4*_Rbase_+1)
+
+ # round 4*_Rbase_ + 1
+ addReg rax, rdx
+ RotL64 rdx, 256,%((4*_Rbase_+1) % 8),0
+ xorReg rdx, rax
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*2+F_O(%rbp,%rdi,8),%r9
+ .endif
+ addReg rcx, rbx
+ RotL64 rbx, 256,%((4*_Rbase_+1) % 8),1
+ xorReg rbx, rcx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*4+F_O(%rbp,%rdi,8),%r11
+ .endif
+ Skein_Debug_Round 256,%(4*_Rbase_+2)
+ .if SKEIN_ASM_UNROLL & 256
+ .irp _r0_,%( 8+(_Rbase_+2) % 5)
+ .irp _r1_,%(13+(_Rbase_+1) % 3)
+ leaq (%r\_r0_,%r\_r1_),%rsi #precompute key injection value for %rbx
+ .endr
+ .endr
+ .endif
+ # round 4*_Rbase_ + 2
+ addReg rax, rbx
+ RotL64 rbx, 256,%((4*_Rbase_+2) % 8),0
+ addReg rcx, rdx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq ksKey+8*3+F_O(%rbp,%rdi,8),%r10
+ .endif
+ xorReg rbx, rax
+ RotL64 rdx, 256,%((4*_Rbase_+2) % 8),1
+ xorReg rdx, rcx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ movq %r8,ksKey+8*6+F_O(%rbp,%rdi,8) #"rotate" the key
+ leaq 1(%r11,%rdi),%r11 #precompute key + tweak
+ .endif
+ Skein_Debug_Round 256,%(4*_Rbase_+3)
+ # round 4*_Rbase_ + 3
+ addReg rax, rdx
+ RotL64 rdx, 256,%((4*_Rbase_+3) % 8),0
+ addReg rcx, rbx
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ addq ksTwk+8*2+F_O(%rbp,%rdi,8),%r10 #precompute key + tweak
+ movq %r13,ksTwk+8*4+F_O(%rbp,%rdi,8) #"rotate" the tweak
+ .endif
+ xorReg rdx, rax
+ RotL64 rbx, 256,%((4*_Rbase_+3) % 8),1
+ xorReg rbx, rcx
+ Skein_Debug_Round 256,%(4*_Rbase_+4)
+ .if (SKEIN_ASM_UNROLL & 256) == 0
+ addReg r9 ,r13 #precompute key+tweak
+ .endif
+ #inject key schedule words
+_Rbase_ = _Rbase_+1
+ .if SKEIN_ASM_UNROLL & 256
+ addReg rax,r,%(8+((_Rbase_+0) % 5))
+ addReg rbx,rsi
+ addReg rcx,rdi
+ addReg rdx,r,%(8+((_Rbase_+3) % 5)),,_Rbase_
+ .else
+ incq %rdi
+ addReg rax,r8
+ addReg rcx,r10
+ addReg rbx,r9
+ addReg rdx,r11
+ .endif
+ Skein_Debug_Round 256,SKEIN_RND_KEY_INJECT
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 256) == 0
+ cmpq $2*(ROUNDS_256/8),%rdi
+ jb Skein_256_round_loop
+.endif # (SKEIN_ASM_UNROLL & 256) == 0
+ movq ctxPtr +F_O(%rbp),%rdi #restore rdi --> context
+
+ #----------------------------
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..3}
+ movq $FIRST_MASK64 ,%r14
+ xorq Wcopy + 0+F_O (%rbp),%rax
+ xorq Wcopy + 8+F_O (%rbp),%rbx
+ xorq Wcopy +16+F_O (%rbp),%rcx
+ xorq Wcopy +24+F_O (%rbp),%rdx
+ andq TWEAK + 8 (%rdi),%r14
+ movq %rax,X_VARS+ 0(%rdi) #store final result
+ movq %rbx,X_VARS+ 8(%rdi)
+ movq %rcx,X_VARS+16(%rdi)
+ movq %rdx,X_VARS+24(%rdi)
+
+ Skein_Debug_Round 256,SKEIN_RND_FEED_FWD
+
+ # go back for more blocks, if needed
+ decq blkCnt+F_O(%rbp)
+ jnz Skein_256_block_loop
+ movq %r14,TWEAK + 8(%rdi)
+ Reset_Stack
+ ret
+Skein_256_Process_Block_End:
+
+ .if _SKEIN_DEBUG
+Skein_Debug_Round_256: #here with rdx == round "number" from macro
+ pushq %rsi #save two regs for BLK_BITS-specific parms
+ pushq %rdi
+ movq 24(%rsp),%rdi #get back original rdx (pushed on stack in macro call) to rdi
+ movq %rax,X_stk+ 0+F_O(%rbp) #save X[] state on stack so debug routines can access it
+ movq %rbx,X_stk+ 8+F_O(%rbp) #(use FP_ since rsp has changed!)
+ movq %rcx,X_stk+16+F_O(%rbp)
+ movq %rdi,X_stk+24+F_O(%rbp)
+
+ movq ctxPtr+F_O(%rbp),%rsi #ctx_hdr_ptr
+ movq $256,%rdi #now <rdi,rsi,rdx> are set for the call
+ jmp Skein_Debug_Round_Common
+ .endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein_256_Process_Block_CodeSize
+ movq $(Skein_256_Process_Block_End-Skein_256_Process_Block),%rax
+ ret
+#
+C_label Skein_256_Unroll_Cnt
+ .if _UNROLL_CNT <> ROUNDS_256/8
+ movq $_UNROLL_CNT,%rax
+ .else
+ xorq %rax,%rax
+ .endif
+ ret
+.endif
+#
+.endif #_USE_ASM_ & 256
+#
+#=================================== Skein_512 =============================================
+#
+.if _USE_ASM_ & 512
+#
+# void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)
+#
+# X[i] == %r[8+i] #register assignments for X[] values during rounds (i=0..7)
+#
+#################
+# MACRO: one round for 512-bit blocks
+#
+.macro R_512_OneRound rn0,rn1,rn2,rn3,rn4,rn5,rn6,rn7,_Rn_,op1,op2,op3,op4
+#
+ addReg r\rn0, r\rn1
+ RotL64 r\rn1, 512,%((_Rn_) % 8),0
+ xorReg r\rn1, r\rn0
+ op1
+ addReg r\rn2, r\rn3
+ RotL64 r\rn3, 512,%((_Rn_) % 8),1
+ xorReg r\rn3, r\rn2
+ op2
+ addReg r\rn4, r\rn5
+ RotL64 r\rn5, 512,%((_Rn_) % 8),2
+ xorReg r\rn5, r\rn4
+ op3
+ addReg r\rn6, r\rn7
+ RotL64 r\rn7, 512,%((_Rn_) % 8),3
+ xorReg r\rn7, r\rn6
+ op4
+ Skein_Debug_Round 512,%(_Rn_+1),-4
+#
+.endm #R_512_OneRound
+#
+#################
+# MACRO: eight rounds for 512-bit blocks
+#
+.macro R_512_FourRounds _RR_ #RR = base round number (0 % 8)
+ .if (SKEIN_ASM_UNROLL && 512)
+ # here for fully unrolled case.
+ _II_ = ((_RR_)/4) + 1 #key injection counter
+ R_512_OneRound 8, 9,10,11,12,13,14,15,%((_RR_)+0),<movq ksKey+8*(((_II_)+3) % 9)+F_O(%rbp),%rax>,,<movq ksKey+8*(((_II_)+4) % 9)+F_O(%rbp),%rbx>
+ R_512_OneRound 10, 9,12,15,14,13, 8,11,%((_RR_)+1),<movq ksKey+8*(((_II_)+5) % 9)+F_O(%rbp),%rcx>,,<movq ksKey+8*(((_II_)+6) % 9)+F_O(%rbp),%rdx>
+ R_512_OneRound 12, 9,14,11, 8,13,10,15,%((_RR_)+2),<movq ksKey+8*(((_II_)+7) % 9)+F_O(%rbp),%rsi>,,<addq ksTwk+8*(((_II_)+0) % 3)+F_O(%rbp),%rcx>
+ R_512_OneRound 14, 9, 8,15,10,13,12,11,%((_RR_)+3),<addq ksTwk+8*(((_II_)+1) % 3)+F_O(%rbp),%rdx>,
+ # inject the key schedule
+ addq ksKey+8*(((_II_)+0)%9)+F_O(%rbp),%r8
+ addReg r11, rax
+ addq ksKey+8*(((_II_)+1)%9)+F_O(%rbp),%r9
+ addReg r12, rbx
+ addq ksKey+8*(((_II_)+2)%9)+F_O(%rbp),%r10
+ addReg r13, rcx
+ addReg r14, rdx
+ addReg r15, rsi,,,(_II_)
+ .else
+ # here for looping case #"rotate" key/tweak schedule (move up on stack)
+ incq %rdi #bump key injection counter
+ R_512_OneRound 8, 9,10,11,12,13,14,15,%((_RR_)+0),<movq ksKey+8*6+F_O(%rbp,%rdi,8),%rdx>,<movq ksTwk-8*1+F_O(%rbp,%rdi,8),%rax>,<movq ksKey-8*1+F_O(%rbp,%rdi,8),%rsi>
+ R_512_OneRound 10, 9,12,15,14,13, 8,11,%((_RR_)+1),<movq ksKey+8*5+F_O(%rbp,%rdi,8),%rcx>,<movq %rax,ksTwk+8*2+F_O(%rbp,%rdi,8) >,<movq %rsi,ksKey+8*8+F_O(%rbp,%rdi,8)>
+ R_512_OneRound 12, 9,14,11, 8,13,10,15,%((_RR_)+2),<movq ksKey+8*4+F_O(%rbp,%rdi,8),%rbx>,<addq ksTwk+8*1+F_O(%rbp,%rdi,8),%rdx>,<movq ksKey+8*7+F_O(%rbp,%rdi,8),%rsi>
+ R_512_OneRound 14, 9, 8,15,10,13,12,11,%((_RR_)+3),<movq ksKey+8*3+F_O(%rbp,%rdi,8),%rax>,<addq ksTwk+8*0+F_O(%rbp,%rdi,8),%rcx>
+ # inject the key schedule
+ addq ksKey+8*0+F_O(%rbp,%rdi,8),%r8
+ addReg r11, rax
+ addReg r12, rbx
+ addq ksKey+8*1+F_O(%rbp,%rdi,8),%r9
+ addReg r13, rcx
+ addReg r14, rdx
+ addq ksKey+8*2+F_O(%rbp,%rdi,8),%r10
+ addReg r15, rsi
+ addReg r15, rdi #inject the round number
+ .endif
+
+ #show the result of the key injection
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INJECT
+.endm #R_512_EightRounds
+#
+#################
+# instantiated code
+#
+C_label Skein_512_Process_Block
+ Setup_Stack 512,ROUNDS_512/8
+ movq TWEAK+ 8(%rdi),%rbx
+ jmp Skein_512_block_loop
+ .p2align 4
+ # main hash loop for Skein_512
+Skein_512_block_loop:
+ # general register usage:
+ # RAX..RDX = temps for key schedule pre-loads
+ # R8 ..R15 = X0..X7
+ # RSP, RBP = stack/frame pointers
+ # RDI = round counter or context pointer
+ # RSI = temp
+ #
+ movq TWEAK + 0(%rdi),%rax
+ addq bitAdd+F_O(%rbp),%rax #computed updated tweak value T0
+ movq %rbx,%rcx
+ xorq %rax,%rcx #%rax/%rbx/%rcx = tweak schedule
+ movq %rax,TWEAK+ 0 (%rdi) #save updated tweak value ctx->h.T[0]
+ movq %rax,ksTwk+ 0+F_O(%rbp)
+ movq $KW_PARITY,%rdx
+ movq blkPtr +F_O(%rbp),%rsi #%rsi --> input block
+ movq %rbx,ksTwk+ 8+F_O(%rbp)
+ movq %rcx,ksTwk+16+F_O(%rbp)
+ .irp _Rn_,8,9,10,11,12,13,14,15
+ movq X_VARS+8*(_Rn_-8)(%rdi),%r\_Rn_
+ xorq %r\_Rn_,%rdx #compute overall parity
+ movq %r\_Rn_,ksKey+8*(_Rn_-8)+F_O(%rbp)
+ .endr #load state into %r8 ..%r15, compute parity
+ movq %rdx,ksKey+8*(8)+F_O(%rbp)#save key schedule parity
+
+ addReg r13,rax #precompute key injection for tweak
+ addReg r14, rbx
+.if _SKEIN_DEBUG
+ movq %rbx,TWEAK+ 8(%rdi) #save updated tweak value ctx->h.T[1] for Skein_Debug_Block below
+.endif
+ movq 0(%rsi),%rax #load input block
+ movq 8(%rsi),%rbx
+ movq 16(%rsi),%rcx
+ movq 24(%rsi),%rdx
+ addReg r8 , rax #do initial key injection
+ addReg r9 , rbx
+ movq %rax,Wcopy+ 0+F_O(%rbp) #keep local copy for feedforward
+ movq %rbx,Wcopy+ 8+F_O(%rbp)
+ addReg r10, rcx
+ addReg r11, rdx
+ movq %rcx,Wcopy+16+F_O(%rbp)
+ movq %rdx,Wcopy+24+F_O(%rbp)
+
+ movq 32(%rsi),%rax
+ movq 40(%rsi),%rbx
+ movq 48(%rsi),%rcx
+ movq 56(%rsi),%rdx
+ addReg r12, rax
+ addReg r13, rbx
+ addReg r14, rcx
+ addReg r15, rdx
+ movq %rax,Wcopy+32+F_O(%rbp)
+ movq %rbx,Wcopy+40+F_O(%rbp)
+ movq %rcx,Wcopy+48+F_O(%rbp)
+ movq %rdx,Wcopy+56+F_O(%rbp)
+
+.if _SKEIN_DEBUG
+ .irp _Rn_,8,9,10,11,12,13,14,15 #save values on stack for debug output
+ movq %r\_Rn_,X_stk+8*(_Rn_-8)(%rsp)
+ .endr
+
+ Skein_Debug_Block 512 #debug dump
+ Skein_Debug_Round 512,SKEIN_RND_KEY_INITIAL
+.endif
+ addq $8*WCNT,%rsi #skip the block
+ movq %rsi,blkPtr+F_O(%rbp) #update block pointer
+ #
+ #################
+ # now the key schedule is computed. Start the rounds
+ #
+.if SKEIN_ASM_UNROLL & 512
+_UNROLL_CNT = ROUNDS_512/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_512
+ .if ((ROUNDS_512/8) % _UNROLL_CNT)
+ .err "Invalid SKEIN_UNROLL_512"
+ .endif
+ xorq %rdi,%rdi #rdi = round counter
+Skein_512_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2
+ R_512_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 512) == 0
+ cmpq $2*(ROUNDS_512/8),%rdi
+ jb Skein_512_round_loop
+ movq ctxPtr +F_O(%rbp),%rdi #restore rdi --> context
+.endif
+ # end of rounds
+ #################
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..7}
+ .irp _Rn_,8,9,10,11,12,13,14,15
+ .if (_Rn_ == 8)
+ movq $FIRST_MASK64,%rbx
+ .endif
+ xorq Wcopy+8*(_Rn_-8)+F_O(%rbp),%r\_Rn_ #feedforward XOR
+ movq %r\_Rn_,X_VARS+8*(_Rn_-8)(%rdi) #and store result
+ .if (_Rn_ == 14)
+ andq TWEAK+ 8(%rdi),%rbx
+ .endif
+ .endr
+ Skein_Debug_Round 512,SKEIN_RND_FEED_FWD
+
+ # go back for more blocks, if needed
+ decq blkCnt+F_O(%rbp)
+ jnz Skein_512_block_loop
+ movq %rbx,TWEAK + 8(%rdi)
+
+ Reset_Stack
+ ret
+Skein_512_Process_Block_End:
+#
+ .if _SKEIN_DEBUG
+# call here with rdx = "round number"
+Skein_Debug_Round_512:
+ pushq %rsi #save two regs for BLK_BITS-specific parms
+ pushq %rdi
+ .irp _Rn_,8,9,10,11,12,13,14,15 #save X[] state on stack so debug routines can access it
+ movq %r\_Rn_,X_stk+8*(_Rn_-8)+F_O(%rbp)
+ .endr
+ movq ctxPtr+F_O(%rbp),%rsi #ctx_hdr_ptr
+ movq $512,%rdi #now <rdi,rsi,rdx> are set for the call
+ jmp Skein_Debug_Round_Common
+ .endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein_512_Process_Block_CodeSize
+ movq $(Skein_512_Process_Block_End-Skein_512_Process_Block),%rax
+ ret
+#
+C_label Skein_512_Unroll_Cnt
+ .if _UNROLL_CNT <> (ROUNDS_512/8)
+ movq $_UNROLL_CNT,%rax
+ .else
+ xorq %rax,%rax
+ .endif
+ ret
+.endif
+#
+.endif # _USE_ASM_ & 512
+#
+#=================================== Skein1024 =============================================
+.if _USE_ASM_ & 1024
+#
+# void Skein1024_Process_Block(Skein_1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t bitcntAdd)#
+#
+#################
+# use details of permutation to make register assignments
+#
+o1K_rdi = 0 #offsets in X[] associated with each register
+o1K_rsi = 1
+o1K_rbp = 2
+o1K_rax = 3
+o1K_rcx = 4 #rcx is "shared" with X6, since X4/X6 alternate
+o1K_rbx = 5
+o1K_rdx = 7
+o1K_r8 = 8
+o1K_r9 = 9
+o1K_r10 = 10
+o1K_r11 = 11
+o1K_r12 = 12
+o1K_r13 = 13
+o1K_r14 = 14
+o1K_r15 = 15
+#
+rIdx_offs = tmpStk_1024
+#
+.macro r1024_Mix w0,w1,reg0,reg1,_RN0_,_Rn1_,op1
+ addReg \reg0 , \reg1 #perform the MIX
+ RotL64 \reg1 , 1024,%((_RN0_) % 8),_Rn1_
+ xorReg \reg1 , \reg0
+.if ((_RN0_) && 3) == 3 #time to do key injection?
+ .if _SKEIN_DEBUG
+ movq %\reg0 , xDebug_1024+8*w0(%rsp) #save intermediate values for Debug_Round
+ movq %\reg1 , xDebug_1024+8*w1(%rsp) # (before inline key injection)
+ .endif
+_II_ = ((_RN0_)/4)+1 #injection count
+ .if SKEIN_ASM_UNROLL && 1024 #here to do fully unrolled key injection
+ addq ksKey+ 8*((_II_+w0) % 17)(%rsp),%\reg0
+ addq ksKey+ 8*((_II_+w1) % 17)(%rsp),%\reg1
+ .if w1 == 13 #tweak injection
+ addq ksTwk+ 8*((_II_+ 0) % 3)(%rsp),%\reg1
+ .elseif w0 == 14
+ addq ksTwk+ 8*((_II_+ 1) % 3)(%rsp),%\reg0
+ .elseif w1 == 15
+ addq $_II_, %\reg1 #(injection counter)
+ .endif
+ .else #here to do looping key injection
+ .if (w0 == 0)
+ movq %rdi, X_stk+8*w0(%rsp) #if so, store N0 so we can use reg as index
+ movq rIdx_offs(%rsp),%rdi #get the injection counter index into rdi
+ .else
+ addq ksKey+8+8*w0(%rsp,%rdi,8),%\reg0 #even key injection
+ .endif
+ .if w1 == 13 #tweak injection
+ addq ksTwk+8+8* 0(%rsp,%rdi,8),%\reg1
+ .elseif w0 == 14
+ addq ksTwk+8+8* 1(%rsp,%rdi,8),%\reg0
+ .elseif w1 == 15
+ addReg \reg1,rdi,,,1 #(injection counter)
+ .endif
+ addq ksKey+8+8*w1(%rsp,%rdi,8),%\reg1 #odd key injection
+ .endif
+.endif
+ # insert the op provided, .if any
+ op1
+.endm
+#################
+# MACRO: four rounds for 1024-bit blocks
+#
+.macro r1024_FourRounds _RR_ #RR = base round number (0 mod 4)
+ # should be here with X4 set properly, X6 stored on stack
+_Rn_ = (_RR_) + 0
+ r1024_Mix 0, 1,rdi,rsi,_Rn_,0
+ r1024_Mix 2, 3,rbp,rax,_Rn_,1
+ r1024_Mix 4, 5,rcx,rbx,_Rn_,2,<movq %rcx,X_stk+8*4(%rsp)> #save X4 on stack (x4/x6 alternate)
+ r1024_Mix 8, 9,r8 ,r9 ,_Rn_,4,<movq X_stk+8*6(%rsp),%rcx> #load X6 from stack
+ r1024_Mix 10,11,r10,r11,_Rn_,5
+ r1024_Mix 12,13,r12,r13,_Rn_,6
+ r1024_Mix 6, 7,rcx,rdx,_Rn_,3
+ r1024_Mix 14,15,r14,r15,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+_Rn_ = (_RR_) + 1
+ r1024_Mix 0, 9,rdi,r9 ,_Rn_,0
+ r1024_Mix 2,13,rbp,r13,_Rn_,1
+ r1024_Mix 6,11,rcx,r11,_Rn_,2,<movq %rcx,X_stk+8*6(%rsp)> #save X6 on stack (x4/x6 alternate)
+ r1024_Mix 10, 7,r10,rdx,_Rn_,4,<movq X_stk+8*4(%rsp),%rcx> #load X4 from stack
+ r1024_Mix 12, 3,r12,rax,_Rn_,5
+ r1024_Mix 14, 5,r14,rbx,_Rn_,6
+ r1024_Mix 4,15,rcx,r15,_Rn_,3
+ r1024_Mix 8, 1,r8 ,rsi,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+_Rn_ = (_RR_) + 2
+ r1024_Mix 0, 7,rdi,rdx,_Rn_,0
+ r1024_Mix 2, 5,rbp,rbx,_Rn_,1
+ r1024_Mix 4, 3,rcx,rax,_Rn_,2,<movq %rcx,X_stk+8*4(%rsp)> #save X4 on stack (x4/x6 alternate)
+ r1024_Mix 12,15,r12,r15,_Rn_,4,<movq X_stk+8*6(%rsp),%rcx> #load X6 from stack
+ r1024_Mix 14,13,r14,r13,_Rn_,5
+ r1024_Mix 8,11,r8 ,r11,_Rn_,6
+ r1024_Mix 6, 1,rcx,rsi,_Rn_,3
+ r1024_Mix 10, 9,r10,r9 ,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+_Rn_ = (_RR_) + 3
+ r1024_Mix 0,15,rdi,r15,_Rn_,0
+ r1024_Mix 2,11,rbp,r11,_Rn_,1
+ r1024_Mix 6,13,rcx,r13,_Rn_,2,<movq %rcx,X_stk+8*6(%rsp)> #save X6 on stack (x4/x6 alternate)
+ r1024_Mix 14, 1,r14,rsi,_Rn_,4,<movq X_stk+8*4(%rsp),%rcx> #load X4 from stack
+ r1024_Mix 8, 5,r8 ,rbx,_Rn_,5
+ r1024_Mix 10, 3,r10,rax,_Rn_,6
+ r1024_Mix 4, 9,rcx,r9 ,_Rn_,3
+ r1024_Mix 12, 7,r12,rdx,_Rn_,7
+ .if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,%(_Rn_+1)
+ .endif
+
+ .if (SKEIN_ASM_UNROLL && 1024) == 0 #here with rdi == rIdx, X0 on stack
+ #"rotate" the key schedule on the stack
+i8 = o1K_r8
+i0 = o1K_rdi
+ movq %r8 , X_stk+8*i8(%rsp) #free up a register (save it on the stack)
+ movq ksKey+8* 0(%rsp,%rdi,8),%r8 #get key word
+ movq %r8 , ksKey+8*17(%rsp,%rdi,8) #rotate key (must do key first or tweak clobbers it!)
+ movq ksTwk+8* 0(%rsp,%rdi,8),%r8 #get tweak word
+ movq %r8 , ksTwk+8* 3(%rsp,%rdi,8) #rotate tweak (onto the stack)
+ movq X_stk+8*i8(%rsp) ,%r8 #get the reg back
+ incq %rdi #bump the index
+ movq %rdi, rIdx_offs (%rsp) #save rdi again
+ movq ksKey+8*i0(%rsp,%rdi,8),%rdi #get the key schedule word for X0 back
+ addq X_stk+8*i0(%rsp) ,%rdi #perform the X0 key injection
+ .endif
+ #show the result of the key injection
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INJECT
+.endm #r1024_FourRounds
+#
+################
+# code
+#
+C_label Skein1024_Process_Block
+#
+ Setup_Stack 1024,ROUNDS_1024/8,WCNT
+ movq TWEAK+ 8(%rdi),%r9
+ jmp Skein1024_block_loop
+ # main hash loop for Skein1024
+ .p2align 4
+Skein1024_block_loop:
+ # general register usage:
+ # RSP = stack pointer
+ # RAX..RDX,RSI,RDI = X1, X3..X7 (state words)
+ # R8 ..R15 = X8..X15 (state words)
+ # RBP = temp (used for X0 and X2)
+ #
+ .if (SKEIN_ASM_UNROLL & 1024) == 0
+ xorq %rax,%rax #init loop index on the stack
+ movq %rax,rIdx_offs(%rsp)
+ .endif
+ movq TWEAK+ 0(%rdi),%r8
+ addq bitAdd+ F_O(%rbp),%r8 #computed updated tweak value T0
+ movq %r9 ,%r10
+ xorq %r8 ,%r10 #%rax/%rbx/%rcx = tweak schedule
+ movq %r8 ,TWEAK+ 0(%rdi) #save updated tweak value ctx->h.T[0]
+ movq %r8 ,ksTwk+ 0+F_O(%rbp)
+ movq %r9 ,ksTwk+ 8+F_O(%rbp) #keep values in %r8 ,%r9 for initial tweak injection below
+ movq %r10,ksTwk+16+F_O(%rbp)
+ .if _SKEIN_DEBUG
+ movq %r9 ,TWEAK+ 8(%rdi) #save updated tweak value ctx->h.T[1] for Skein_Debug_Block
+ .endif
+ movq blkPtr +F_O(%rbp),%rsi # rsi --> input block
+ movq $KW_PARITY ,%rax #overall key schedule parity
+
+ # the logic here assumes the set {rdi,rsi,rbp,rax} = X[0,1,2,3]
+ .irp _rN_,0,1,2,3,4,6 #process the "initial" words, using r14/r15 as temps
+ movq X_VARS+8*_rN_(%rdi),%r14 #get state word
+ movq 8*_rN_(%rsi),%r15 #get msg word
+ xorq %r14,%rax #update key schedule overall parity
+ movq %r14,ksKey +8*_rN_+F_O(%rbp) #save key schedule word on stack
+ movq %r15,Wcopy +8*_rN_+F_O(%rbp) #save local msg Wcopy
+ addq %r15,%r14 #do the initial key injection
+ movq %r14,X_stk +8*_rN_ (%rsp) #save initial state var on stack
+ .endr
+ # now process the rest, using the "real" registers
+ # (MUST do it in reverse order to inject tweaks r8/r9 first)
+ .irp _rr_,r15,r14,r13,r12,r11,r10,r9,r8,rdx,rbx
+_oo_ = o1K_\_rr_ #offset assocated with the register
+ movq X_VARS+8*_oo_(%rdi),%\_rr_ #get key schedule word from context
+ movq 8*_oo_(%rsi),%rcx #get next input msg word
+ movq %\_rr_, ksKey +8*_oo_(%rsp) #save key schedule on stack
+ xorq %\_rr_, %rax #accumulate key schedule parity
+ movq %rcx,Wcopy+8*_oo_+F_O(%rbp) #save copy of msg word for feedforward
+ addq %rcx,%\_rr_ #do the initial key injection
+ .if _oo_ == 13 #do the initial tweak injection
+ addReg _rr_,r8 # (only in words 13/14)
+ .elseif _oo_ == 14
+ addReg _rr_,r9
+ .endif
+ .endr
+ movq %rax,ksKey+8*WCNT+F_O(%rbp) #save key schedule parity
+.if _SKEIN_DEBUG
+ Skein_Debug_Block 1024 #initial debug dump
+.endif
+ addq $8*WCNT,%rsi #bump the msg ptr
+ movq %rsi,blkPtr+F_O(%rbp) #save bumped msg ptr
+ # re-load words 0..4 from stack, enter the main loop
+ .irp _rr_,rdi,rsi,rbp,rax,rcx #(no need to re-load x6, already on stack)
+ movq X_stk+8*o1K_\_rr_(%rsp),%\_rr_ #re-load state and get ready to go!
+ .endr
+.if _SKEIN_DEBUG
+ Skein_Debug_Round 1024,SKEIN_RND_KEY_INITIAL #show state after initial key injection
+.endif
+ #
+ #################
+ # now the key schedule is computed. Start the rounds
+ #
+.if SKEIN_ASM_UNROLL & 1024
+_UNROLL_CNT = ROUNDS_1024/8
+.else
+_UNROLL_CNT = SKEIN_UNROLL_1024
+ .if ((ROUNDS_1024/8) % _UNROLL_CNT)
+ .error "Invalid SKEIN_UNROLL_1024"
+ .endif
+Skein1024_round_loop:
+.endif
+#
+_Rbase_ = 0
+.rept _UNROLL_CNT*2 #implement the rounds, 4 at a time
+ r1024_FourRounds %(4*_Rbase_+00)
+_Rbase_ = _Rbase_+1
+.endr #rept _UNROLL_CNT
+#
+.if (SKEIN_ASM_UNROLL & 1024) == 0
+ cmpq $2*(ROUNDS_1024/8),tmpStk_1024(%rsp) #see .if we are done
+ jb Skein1024_round_loop
+.endif
+ # end of rounds
+ #################
+ #
+ # feedforward: ctx->X[i] = X[i] ^ w[i], {i=0..15}
+ movq %rdx,X_stk+8*o1K_rdx(%rsp) #we need a register. x6 already on stack
+ movq ctxPtr(%rsp),%rdx
+
+ .irp _rr_,rdi,rsi,rbp,rax,rcx,rbx,r8,r9,r10,r11,r12,r13,r14,r15 #do all but x6,x7
+_oo_ = o1K_\_rr_
+ xorq Wcopy +8*_oo_(%rsp),%\_rr_ #feedforward XOR
+ movq %\_rr_,X_VARS+8*_oo_(%rdx) #save result into context
+ .if (_oo_ == 9)
+ movq $FIRST_MASK64 ,%r9
+ .endif
+ .if (_oo_ == 14)
+ andq TWEAK+ 8(%rdx),%r9
+ .endif
+ .endr
+ #
+ movq X_stk +8*6(%rsp),%rax #now process x6,x7 (skipped in .irp above)
+ movq X_stk +8*7(%rsp),%rbx
+ xorq Wcopy +8*6(%rsp),%rax
+ xorq Wcopy +8*7(%rsp),%rbx
+ movq %rax,X_VARS+8*6(%rdx)
+ decq blkCnt(%rsp) #set zero flag iff done
+ movq %rbx,X_VARS+8*7(%rdx)
+
+ Skein_Debug_Round 1024,SKEIN_RND_FEED_FWD,,<cmpq $0,blkCnt(%rsp)>
+ # go back for more blocks, if needed
+ movq ctxPtr(%rsp),%rdi #don't muck with the flags here!
+ lea FRAME_OFFS(%rsp),%rbp
+ jnz Skein1024_block_loop
+ movq %r9 ,TWEAK+ 8(%rdx)
+ Reset_Stack
+ ret
+#
+Skein1024_Process_Block_End:
+#
+.if _SKEIN_DEBUG
+Skein_Debug_Round_1024:
+ # call here with rdx = "round number",
+_SP_OFFS_ = 8*2 #stack "offset" here: rdx, return addr
+ #
+ #save rest of X[] state on stack so debug routines can access it
+ .irp _rr_,rsi,rbp,rax,rbx,r8,r9,r10,r11,r12,r13,r14,r15
+ movq %\_rr_,X_stk+8*o1K_\_rr_+_SP_OFFS_(%rsp)
+ .endr
+ # Figure out what to do with x0 (rdi). When rdx == 0 mod 4, it's already on stack
+ cmpq $SKEIN_RND_SPECIAL,%rdx #special rounds always save
+ jae save_x0
+ testq $3,%rdx #otherwise only if rdx != 0 mod 4
+ jz save_x0_not
+save_x0:
+ movq %rdi,X_stk+8*o1K_rdi+_SP_OFFS_(%rsp)
+save_x0_not:
+ #figure out the x4/x6 swapping state and save the correct one!
+ cmpq $SKEIN_RND_SPECIAL,%rdx #special rounds always do x4
+ jae save_x4
+ testq $1,%rdx #and even ones have r4 as well
+ jz save_x4
+ movq %rcx,X_stk+8*6+_SP_OFFS_(%rsp)
+ jmp debug_1024_go
+save_x4:
+ movq %rcx,X_stk+8*4+_SP_OFFS_(%rsp)
+debug_1024_go:
+ #now all is saved in Xstk[] except for rdx
+ push %rsi #save two regs for BLK_BITS-specific parms
+ push %rdi
+_SP_OFFS_ = _SP_OFFS_ + 16 #adjust stack offset accordingly (now 32)
+
+ movq _SP_OFFS_-8(%rsp),%rsi #get back original %rdx (pushed on stack in macro call)
+ movq %rsi,X_stk+8*o1K_rdx+_SP_OFFS_(%rsp) #and save it in its rightful place in X_stk[]
+
+ movq ctxPtr+_SP_OFFS_(%rsp),%rsi #rsi = ctx_hdr_ptr
+ movq $1024,%rdi #rdi = block size
+ jmp Skein_Debug_Round_Common
+.endif
+#
+.if _SKEIN_CODE_SIZE
+C_label Skein1024_Process_Block_CodeSize
+ movq $(Skein1024_Process_Block_End-Skein1024_Process_Block),%rax
+ ret
+#
+C_label Skein1024_Unroll_Cnt
+ .if _UNROLL_CNT <> (ROUNDS_1024/8)
+ movq $_UNROLL_CNT,%rax
+ .else
+ xorq %rax,%rax
+ .endif
+ ret
+.endif
+#
+.endif # _USE_ASM_ and 1024
+#
+.if _SKEIN_DEBUG
+#----------------------------------------------------------------
+#local debug routine to set up for calls to:
+# void Skein_Show_Round(uint_t bits,const Skein_Ctxt_Hdr_t *h,int r,const u64b_t *X)
+# [ rdi rsi rdx rcx]
+#
+# here with %rdx = round number
+# %rsi = ctx_hdr_ptr
+# %rdi = block size (256/512/1024)
+# on stack: saved rdi, saved rsi, retAddr, saved rdx
+#
+Skein_Debug_Round_Common:
+_SP_OFFS_ = 32 #account for four words on stack already
+ .irp _rr_,rax,rbx,rcx,rbp,r8,r9,r10,r11,r12,r13,r14,r15 #save the rest of the regs
+ pushq %\_rr_
+_SP_OFFS_ = _SP_OFFS_+8
+ .endr
+ .if (_SP_OFFS_ % 16) # make sure stack is still 16-byte aligned here
+ .error "Debug_Round_Common: stack alignment"
+ .endif
+ # compute %rcx = ptr to the X[] array on the stack (final parameter to call)
+ leaq X_stk+_SP_OFFS_(%rsp),%rcx #adjust for reg pushes, return address
+ cmpq $SKEIN_RND_FEED_FWD,%rdx #special handling for feedforward "round"?
+ jnz _got_rcxA
+ leaq X_VARS(%rsi),%rcx
+_got_rcxA:
+ .if _USE_ASM_ & 1024
+ # special handling for 1024-bit case
+ # (for rounds right before with key injection:
+ # use xDebug_1024[] instead of X_stk[])
+ cmpq $SKEIN_RND_SPECIAL,%rdx
+ jae _got_rcxB #must be a normal round
+ orq %rdx,%rdx
+ jz _got_rcxB #just before key injection
+ test $3,%rdx
+ jne _got_rcxB
+ cmp $1024,%rdi #only 1024-bit(s) for now
+ jne _got_rcxB
+ leaq xDebug_1024+_SP_OFFS_(%rsp),%rcx
+_got_rcxB:
+ .endif
+ call Skein_Show_Round #call external debug handler
+
+ .irp _rr_,r15,r14,r13,r12,r11,r10,r9,r8,rbp,rcx,rbx,rax #restore regs
+ popq %\_rr_
+_SP_OFFS_ = _SP_OFFS_-8
+ .endr
+ .if _SP_OFFS_ - 32
+ .error "Debug_Round_Common: push/pop misalignment!"
+ .endif
+ popq %rdi
+ popq %rsi
+ ret
+.endif
+#----------------------------------------------------------------
+ .end
Index: sys/crypto/skein/brg_endian.h
===================================================================
--- /dev/null
+++ sys/crypto/skein/brg_endian.h
@@ -0,0 +1,148 @@
+/*
+ ---------------------------------------------------------------------------
+ Copyright (c) 2003, Dr Brian Gladman, Worcester, UK. All rights reserved.
+
+ LICENSE TERMS
+
+ The free distribution and use of this software in both source and binary
+ form is allowed (with or without changes) provided that:
+
+ 1. distributions of this source code include the above copyright
+ notice, this list of conditions and the following disclaimer;
+
+ 2. distributions in binary form include the above copyright
+ notice, this list of conditions and the following disclaimer
+ in the documentation and/or other associated materials;
+
+ 3. the copyright holder's name is not used to endorse products
+ built using this software without specific written permission.
+
+ ALTERNATIVELY, provided that this notice is retained in full, this product
+ may be distributed under the terms of the GNU General Public License (GPL),
+ in which case the provisions of the GPL apply INSTEAD OF those given above.
+
+ DISCLAIMER
+
+ This software is provided 'as is' with no explicit or implied warranties
+ in respect of its properties, including, but not limited to, correctness
+ and/or fitness for purpose.
+ ---------------------------------------------------------------------------
+ Issue 20/10/2006
+*/
+
+#ifndef BRG_ENDIAN_H
+#define BRG_ENDIAN_H
+
+#define IS_BIG_ENDIAN 4321 /* byte 0 is most significant (mc68k) */
+#define IS_LITTLE_ENDIAN 1234 /* byte 0 is least significant (i386) */
+
+/* Include files where endian defines and byteswap functions may reside */
+#if defined( __FreeBSD__ ) || defined( __OpenBSD__ ) || defined( __NetBSD__ )
+# include <sys/endian.h>
+#elif defined( BSD ) && ( BSD >= 199103 ) || defined( __APPLE__ ) || \
+ defined( __CYGWIN32__ ) || defined( __DJGPP__ ) || defined( __osf__ )
+# include <machine/endian.h>
+#elif defined( __linux__ ) || defined( __GNUC__ ) || defined( __GNU_LIBRARY__ )
+# if !defined( __MINGW32__ ) && !defined(AVR)
+# include <endian.h>
+# if !defined( __BEOS__ )
+# include <byteswap.h>
+# endif
+# endif
+#endif
+
+/* Now attempt to set the define for platform byte order using any */
+/* of the four forms SYMBOL, _SYMBOL, __SYMBOL & __SYMBOL__, which */
+/* seem to encompass most endian symbol definitions */
+
+#if defined( BIG_ENDIAN ) && defined( LITTLE_ENDIAN )
+# if defined( BYTE_ORDER ) && BYTE_ORDER == BIG_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( BYTE_ORDER ) && BYTE_ORDER == LITTLE_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( BIG_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( LITTLE_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( _BIG_ENDIAN ) && defined( _LITTLE_ENDIAN )
+# if defined( _BYTE_ORDER ) && _BYTE_ORDER == _BIG_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( _BYTE_ORDER ) && _BYTE_ORDER == _LITTLE_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( _BIG_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( _LITTLE_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN ) && defined( __LITTLE_ENDIAN )
+# if defined( __BYTE_ORDER ) && __BYTE_ORDER == __BIG_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( __BYTE_ORDER ) && __BYTE_ORDER == __LITTLE_ENDIAN
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( __BIG_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN__ ) && defined( __LITTLE_ENDIAN__ )
+# if defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __BIG_ENDIAN__
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# elif defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __LITTLE_ENDIAN__
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif defined( __BIG_ENDIAN__ )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN__ )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+/* if the platform byte order could not be determined, then try to */
+/* set this define using common machine defines */
+#if !defined(PLATFORM_BYTE_ORDER)
+
+#if defined( __alpha__ ) || defined( __alpha ) || defined( i386 ) || \
+ defined( __i386__ ) || defined( _M_I86 ) || defined( _M_IX86 ) || \
+ defined( __OS2__ ) || defined( sun386 ) || defined( __TURBOC__ ) || \
+ defined( vax ) || defined( vms ) || defined( VMS ) || \
+ defined( __VMS ) || defined( _M_X64 ) || defined( AVR )
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+
+#elif defined( AMIGA ) || defined( applec ) || defined( __AS400__ ) || \
+ defined( _CRAY ) || defined( __hppa ) || defined( __hp9000 ) || \
+ defined( ibm370 ) || defined( mc68000 ) || defined( m68k ) || \
+ defined( __MRC__ ) || defined( __MVS__ ) || defined( __MWERKS__ ) || \
+ defined( sparc ) || defined( __sparc) || defined( SYMANTEC_C ) || \
+ defined( __VOS__ ) || defined( __TIGCC__ ) || defined( __TANDEM ) || \
+ defined( THINK_C ) || defined( __VMCMS__ )
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+
+#elif 0 /* **** EDIT HERE IF NECESSARY **** */
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#elif 0 /* **** EDIT HERE IF NECESSARY **** */
+# define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#else
+# error Please edit lines 126 or 128 in brg_endian.h to set the platform byte order
+#endif
+#endif
+
+/* special handler for IA64, which may be either endianness (?) */
+/* here we assume little-endian, but this may need to be changed */
+#if defined(__ia64) || defined(__ia64__) || defined(_M_IA64)
+# define PLATFORM_MUST_ALIGN (1)
+#ifndef PLATFORM_BYTE_ORDER
+# define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+#endif
+
+#ifndef PLATFORM_MUST_ALIGN
+# define PLATFORM_MUST_ALIGN (0)
+#endif
+
+#endif /* ifndef BRG_ENDIAN_H */
Index: sys/crypto/skein/brg_types.h
===================================================================
--- /dev/null
+++ sys/crypto/skein/brg_types.h
@@ -0,0 +1,188 @@
+/*
+ ---------------------------------------------------------------------------
+ Copyright (c) 1998-2006, Brian Gladman, Worcester, UK. All rights reserved.
+
+ LICENSE TERMS
+
+ The free distribution and use of this software in both source and binary
+ form is allowed (with or without changes) provided that:
+
+ 1. distributions of this source code include the above copyright
+ notice, this list of conditions and the following disclaimer;
+
+ 2. distributions in binary form include the above copyright
+ notice, this list of conditions and the following disclaimer
+ in the documentation and/or other associated materials;
+
+ 3. the copyright holder's name is not used to endorse products
+ built using this software without specific written permission.
+
+ ALTERNATIVELY, provided that this notice is retained in full, this product
+ may be distributed under the terms of the GNU General Public License (GPL),
+ in which case the provisions of the GPL apply INSTEAD OF those given above.
+
+ DISCLAIMER
+
+ This software is provided 'as is' with no explicit or implied warranties
+ in respect of its properties, including, but not limited to, correctness
+ and/or fitness for purpose.
+ ---------------------------------------------------------------------------
+ Issue 09/09/2006
+
+ The unsigned integer types defined here are of the form uint_<nn>t where
+ <nn> is the length of the type; for example, the unsigned 32-bit type is
+ 'uint_32t'. These are NOT the same as the 'C99 integer types' that are
+ defined in the inttypes.h and stdint.h headers since attempts to use these
+ types have shown that support for them is still highly variable. However,
+ since the latter are of the form uint<nn>_t, a regular expression search
+ and replace (in VC++ search on 'uint_{:z}t' and replace with 'uint\1_t')
+ can be used to convert the types used here to the C99 standard types.
+*/
+
+#ifndef BRG_TYPES_H
+#define BRG_TYPES_H
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#include <sys/limits.h>
+
+#ifndef BRG_UI8
+# define BRG_UI8
+# if UCHAR_MAX == 255u
+ typedef unsigned char uint_8t;
+# else
+# error Please define uint_8t as an 8-bit unsigned integer type in brg_types.h
+# endif
+#endif
+
+#ifndef BRG_UI16
+# define BRG_UI16
+# if USHRT_MAX == 65535u
+ typedef unsigned short uint_16t;
+# else
+# error Please define uint_16t as a 16-bit unsigned short type in brg_types.h
+# endif
+#endif
+
+#ifndef BRG_UI32
+# define BRG_UI32
+# if UINT_MAX == 4294967295u
+# define li_32(h) 0x##h##u
+ typedef unsigned int uint_32t;
+# elif ULONG_MAX == 4294967295u
+# define li_32(h) 0x##h##ul
+ typedef unsigned long uint_32t;
+# elif defined( _CRAY )
+# error This code needs 32-bit data types, which Cray machines do not provide
+# else
+# error Please define uint_32t as a 32-bit unsigned integer type in brg_types.h
+# endif
+#endif
+
+#ifndef BRG_UI64
+# if defined( __BORLANDC__ ) && !defined( __MSDOS__ )
+# define BRG_UI64
+# define li_64(h) 0x##h##ui64
+ typedef unsigned __int64 uint_64t;
+# elif defined( _MSC_VER ) && ( _MSC_VER < 1300 ) /* 1300 == VC++ 7.0 */
+# define BRG_UI64
+# define li_64(h) 0x##h##ui64
+ typedef unsigned __int64 uint_64t;
+# elif defined( __sun ) && defined(ULONG_MAX) && ULONG_MAX == 0xfffffffful
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# elif defined( UINT_MAX ) && UINT_MAX > 4294967295u
+# if UINT_MAX == 18446744073709551615u
+# define BRG_UI64
+# define li_64(h) 0x##h##u
+ typedef unsigned int uint_64t;
+# endif
+# elif defined( ULONG_MAX ) && ULONG_MAX > 4294967295u
+# if ULONG_MAX == 18446744073709551615ul
+# define BRG_UI64
+# define li_64(h) 0x##h##ul
+ typedef unsigned long uint_64t;
+# endif
+# elif defined( ULLONG_MAX ) && ULLONG_MAX > 4294967295u
+# if ULLONG_MAX == 18446744073709551615ull
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# endif
+# elif defined( ULONG_LONG_MAX ) && ULONG_LONG_MAX > 4294967295u
+# if ULONG_LONG_MAX == 18446744073709551615ull
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# endif
+# elif defined(__GNUC__) /* DLW: avoid mingw problem with -ansi */
+# define BRG_UI64
+# define li_64(h) 0x##h##ull
+ typedef unsigned long long uint_64t;
+# endif
+#endif
+
+#if defined( NEED_UINT_64T ) && !defined( BRG_UI64 )
+# error Please define uint_64t as an unsigned 64 bit type in brg_types.h
+#endif
+
+#ifndef RETURN_VALUES
+# define RETURN_VALUES
+# if defined( DLL_EXPORT )
+# if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+# define VOID_RETURN __declspec( dllexport ) void __stdcall
+# define INT_RETURN __declspec( dllexport ) int __stdcall
+# elif defined( __GNUC__ )
+# define VOID_RETURN __declspec( __dllexport__ ) void
+# define INT_RETURN __declspec( __dllexport__ ) int
+# else
+# error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+# endif
+# elif defined( DLL_IMPORT )
+# if defined( _MSC_VER ) || defined ( __INTEL_COMPILER )
+# define VOID_RETURN __declspec( dllimport ) void __stdcall
+# define INT_RETURN __declspec( dllimport ) int __stdcall
+# elif defined( __GNUC__ )
+# define VOID_RETURN __declspec( __dllimport__ ) void
+# define INT_RETURN __declspec( __dllimport__ ) int
+# else
+# error Use of the DLL is only available on the Microsoft, Intel and GCC compilers
+# endif
+# elif defined( __WATCOMC__ )
+# define VOID_RETURN void __cdecl
+# define INT_RETURN int __cdecl
+# else
+# define VOID_RETURN void
+# define INT_RETURN int
+# endif
+#endif
+
+/* These defines are used to declare buffers in a way that allows
+ faster operations on longer variables to be used. In all these
+ defines 'size' must be a power of 2 and >= 8
+
+ dec_unit_type(size,x) declares a variable 'x' of length
+ 'size' bits
+
+ dec_bufr_type(size,bsize,x) declares a buffer 'x' of length 'bsize'
+ bytes defined as an array of variables
+ each of 'size' bits (bsize must be a
+ multiple of size / 8)
+
+ ptr_cast(x,size) casts a pointer to a pointer to a
+ varaiable of length 'size' bits
+*/
+
+#define ui_type(size) uint_##size##t
+#define dec_unit_type(size,x) typedef ui_type(size) x
+#define dec_bufr_type(size,bsize,x) typedef ui_type(size) x[bsize / (size >> 3)]
+#define ptr_cast(x,size) ((ui_type(size)*)(x))
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif
Index: sys/crypto/skein/skein.h
===================================================================
--- /dev/null
+++ sys/crypto/skein/skein.h
@@ -0,0 +1,336 @@
+#ifndef _SKEIN_H_
+#define _SKEIN_H_ 1
+/**************************************************************************
+**
+** Interface declarations and 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
+**
+***************************************************************************/
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+#ifndef _KERNEL
+#include <stddef.h> /* get size_t definition */
+#endif
+#include "skein_port.h" /* get platform-specific definitions */
+
+enum
+ {
+ SKEIN_SUCCESS = 0, /* return codes from Skein calls */
+ SKEIN_FAIL = 1,
+ SKEIN_BAD_HASHLEN = 2
+ };
+
+#define SKEIN_MODIFIER_WORDS ( 2) /* number of modifier (tweak) words */
+
+#define SKEIN_256_STATE_WORDS ( 4)
+#define SKEIN_512_STATE_WORDS ( 8)
+#define SKEIN1024_STATE_WORDS (16)
+#define SKEIN_MAX_STATE_WORDS (16)
+
+#define SKEIN_256_STATE_BYTES ( 8*SKEIN_256_STATE_WORDS)
+#define SKEIN_512_STATE_BYTES ( 8*SKEIN_512_STATE_WORDS)
+#define SKEIN1024_STATE_BYTES ( 8*SKEIN1024_STATE_WORDS)
+
+#define SKEIN_256_STATE_BITS (64*SKEIN_256_STATE_WORDS)
+#define SKEIN_512_STATE_BITS (64*SKEIN_512_STATE_WORDS)
+#define SKEIN1024_STATE_BITS (64*SKEIN1024_STATE_WORDS)
+
+#define SKEIN_256_BLOCK_BYTES ( 8*SKEIN_256_STATE_WORDS)
+#define SKEIN_512_BLOCK_BYTES ( 8*SKEIN_512_STATE_WORDS)
+#define SKEIN1024_BLOCK_BYTES ( 8*SKEIN1024_STATE_WORDS)
+
+typedef struct
+ {
+ size_t hashBitLen; /* size of hash result, in bits */
+ size_t bCnt; /* current byte count in buffer b[] */
+ u64b_t T[SKEIN_MODIFIER_WORDS]; /* tweak words: T[0]=byte cnt, T[1]=flags */
+ } Skein_Ctxt_Hdr_t;
+
+typedef struct /* 256-bit Skein hash context structure */
+ {
+ Skein_Ctxt_Hdr_t h; /* common header context variables */
+ u64b_t X[SKEIN_256_STATE_WORDS]; /* chaining variables */
+ u08b_t b[SKEIN_256_BLOCK_BYTES]; /* partial block buffer (8-byte aligned) */
+ } Skein_256_Ctxt_t;
+
+typedef struct /* 512-bit Skein hash context structure */
+ {
+ Skein_Ctxt_Hdr_t h; /* common header context variables */
+ u64b_t X[SKEIN_512_STATE_WORDS]; /* chaining variables */
+ u08b_t b[SKEIN_512_BLOCK_BYTES]; /* partial block buffer (8-byte aligned) */
+ } Skein_512_Ctxt_t;
+
+typedef struct /* 1024-bit Skein hash context structure */
+ {
+ Skein_Ctxt_Hdr_t h; /* common header context variables */
+ u64b_t X[SKEIN1024_STATE_WORDS]; /* chaining variables */
+ u08b_t b[SKEIN1024_BLOCK_BYTES]; /* partial block buffer (8-byte aligned) */
+ } Skein1024_Ctxt_t;
+
+/* Skein APIs for (incremental) "straight hashing" */
+int Skein_256_Init (Skein_256_Ctxt_t *ctx, size_t hashBitLen);
+int Skein_512_Init (Skein_512_Ctxt_t *ctx, size_t hashBitLen);
+int Skein1024_Init (Skein1024_Ctxt_t *ctx, size_t hashBitLen);
+
+int Skein_256_Update(Skein_256_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+int Skein_512_Update(Skein_512_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+int Skein1024_Update(Skein1024_Ctxt_t *ctx, const u08b_t *msg, size_t msgByteCnt);
+
+int Skein_256_Final (Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein_512_Final (Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein1024_Final (Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+
+/*
+** Skein APIs for "extended" initialization: MAC keys, tree hashing.
+** After an InitExt() call, just use Update/Final calls as with Init().
+**
+** Notes: Same parameters as _Init() calls, plus treeInfo/key/keyBytes.
+** When keyBytes == 0 and treeInfo == SKEIN_SEQUENTIAL,
+** the results of InitExt() are identical to calling Init().
+** The function Init() may be called once to "precompute" the IV for
+** a given hashBitLen value, then by saving a copy of the context
+** the IV computation may be avoided in later calls.
+** Similarly, the function InitExt() may be called once per MAC key
+** to precompute the MAC IV, then a copy of the context saved and
+** reused for each new MAC computation.
+**/
+int Skein_256_InitExt(Skein_256_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+int Skein_512_InitExt(Skein_512_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+int Skein1024_InitExt(Skein1024_Ctxt_t *ctx, size_t hashBitLen, u64b_t treeInfo, const u08b_t *key, size_t keyBytes);
+
+/*
+** Skein APIs for MAC and tree hash:
+** Final_Pad: pad, do final block, but no OUTPUT type
+** Output: do just the output stage
+*/
+int Skein_256_Final_Pad(Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein_512_Final_Pad(Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein1024_Final_Pad(Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+
+#ifndef SKEIN_TREE_HASH
+#define SKEIN_TREE_HASH (1)
+#endif
+#if SKEIN_TREE_HASH
+int Skein_256_Output (Skein_256_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein_512_Output (Skein_512_Ctxt_t *ctx, u08b_t * hashVal);
+int Skein1024_Output (Skein1024_Ctxt_t *ctx, u08b_t * hashVal);
+#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 */
+#define SKEIN_T1_BIT(BIT) ((BIT) - 64) /* offset 64 because it's the second word */
+
+#define SKEIN_T1_POS_TREE_LVL SKEIN_T1_BIT(112) /* bits 112..118: level in hash tree */
+#define SKEIN_T1_POS_BIT_PAD SKEIN_T1_BIT(119) /* bit 119 : partial final input byte */
+#define SKEIN_T1_POS_BLK_TYPE SKEIN_T1_BIT(120) /* bits 120..125: type field */
+#define SKEIN_T1_POS_FIRST SKEIN_T1_BIT(126) /* bits 126 : first block flag */
+#define SKEIN_T1_POS_FINAL SKEIN_T1_BIT(127) /* bit 127 : final block flag */
+
+/* tweak word T[1]: flag bit definition(s) */
+#define SKEIN_T1_FLAG_FIRST (((u64b_t) 1 ) << SKEIN_T1_POS_FIRST)
+#define SKEIN_T1_FLAG_FINAL (((u64b_t) 1 ) << SKEIN_T1_POS_FINAL)
+#define SKEIN_T1_FLAG_BIT_PAD (((u64b_t) 1 ) << SKEIN_T1_POS_BIT_PAD)
+
+/* tweak word T[1]: tree level bit field mask */
+#define SKEIN_T1_TREE_LVL_MASK (((u64b_t)0x7F) << SKEIN_T1_POS_TREE_LVL)
+#define SKEIN_T1_TREE_LEVEL(n) (((u64b_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 digital 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) (((u64b_t) (SKEIN_BLK_TYPE_##T)) << SKEIN_T1_POS_BLK_TYPE)
+#define SKEIN_T1_BLK_TYPE_KEY SKEIN_T1_BLK_TYPE(KEY) /* key, for MAC and KDF */
+#define SKEIN_T1_BLK_TYPE_CFG SKEIN_T1_BLK_TYPE(CFG) /* configuration block */
+#define SKEIN_T1_BLK_TYPE_PERS SKEIN_T1_BLK_TYPE(PERS) /* personalization string */
+#define SKEIN_T1_BLK_TYPE_PK SKEIN_T1_BLK_TYPE(PK) /* public key (for digital signature hashing) */
+#define SKEIN_T1_BLK_TYPE_KDF SKEIN_T1_BLK_TYPE(KDF) /* key identifier for KDF */
+#define SKEIN_T1_BLK_TYPE_NONCE SKEIN_T1_BLK_TYPE(NONCE)/* nonce for PRNG */
+#define SKEIN_T1_BLK_TYPE_MSG SKEIN_T1_BLK_TYPE(MSG) /* message processing */
+#define SKEIN_T1_BLK_TYPE_OUT SKEIN_T1_BLK_TYPE(OUT) /* output stage */
+#define SKEIN_T1_BLK_TYPE_MASK SKEIN_T1_BLK_TYPE(MASK) /* field bit 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) + (((u64b_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 (((u64b_t) 0xFF) << SKEIN_CFG_TREE_LEAF_SIZE_POS)
+#define SKEIN_CFG_TREE_NODE_SIZE_MSK (((u64b_t) 0xFF) << SKEIN_CFG_TREE_NODE_SIZE_POS)
+#define SKEIN_CFG_TREE_MAX_LEVEL_MSK (((u64b_t) 0xFF) << SKEIN_CFG_TREE_MAX_LEVEL_POS)
+
+#define SKEIN_CFG_TREE_INFO(leaf,node,maxLvl) \
+ ( (((u64b_t)(leaf )) << SKEIN_CFG_TREE_LEAF_SIZE_POS) | \
+ (((u64b_t)(node )) << SKEIN_CFG_TREE_NODE_SIZE_POS) | \
+ (((u64b_t)(maxLvl)) << SKEIN_CFG_TREE_MAX_LEVEL_POS) )
+
+#define SKEIN_CFG_TREE_INFO_SEQUENTIAL SKEIN_CFG_TREE_INFO(0,0,0) /* use as treeInfo in InitExt() call for sequential processing */
+
+/*
+** 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) {(ctxPtr)->h.T[TWK_NUM] = (tVal);}
+
+#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) \
+ { \
+ Skein_Set_T0(ctxPtr,(T0)); \
+ Skein_Set_T1(ctxPtr,(T1)); \
+ }
+
+#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) \
+ { Skein_Set_T0_T1(ctxPtr,0,SKEIN_T1_FLAG_FIRST | SKEIN_T1_BLK_TYPE_##BLK_TYPE); (ctxPtr)->h.bCnt=0; }
+
+#define Skein_Clear_First_Flag(hdr) { (hdr).T[1] &= ~SKEIN_T1_FLAG_FIRST; }
+#define Skein_Set_Bit_Pad_Flag(hdr) { (hdr).T[1] |= SKEIN_T1_FLAG_BIT_PAD; }
+
+#define Skein_Set_Tree_Level(hdr,height) { (hdr).T[1] |= SKEIN_T1_TREE_LEVEL(height);}
+
+/*****************************************************************
+** "Internal" Skein definitions for debugging and error checking
+******************************************************************/
+#ifdef SKEIN_DEBUG /* examine/display intermediate values? */
+#include "skein_debug.h"
+#else /* default is no callouts */
+#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)
+#endif
+
+#ifndef SKEIN_ERR_CHECK /* run-time checks (e.g., bad params, uninitialized context)? */
+#define Skein_Assert(x,retCode)/* default: ignore all Asserts, for performance */
+#define Skein_assert(x)
+#elif defined(SKEIN_ASSERT)
+#include <assert.h>
+#define Skein_Assert(x,retCode) assert(x)
+#define Skein_assert(x) assert(x)
+#else
+#include <assert.h>
+#define Skein_Assert(x,retCode) { if (!(x)) return retCode; } /* caller error */
+#define Skein_assert(x) assert(x) /* internal error */
+#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
+ };
+
+#ifndef SKEIN_ROUNDS
+#define SKEIN_256_ROUNDS_TOTAL (72) /* number of rounds for the different block sizes */
+#define SKEIN_512_ROUNDS_TOTAL (72)
+#define SKEIN1024_ROUNDS_TOTAL (80)
+#else /* allow command-line define in range 8*(5..14) */
+#define SKEIN_256_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS/100) + 5) % 10) + 5))
+#define SKEIN_512_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS/ 10) + 5) % 10) + 5))
+#define SKEIN1024_ROUNDS_TOTAL (8*((((SKEIN_ROUNDS ) + 5) % 10) + 5))
+#endif
+
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void Skein1024_Process_Block(Skein1024_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+
+#ifdef __cplusplus
+}
+#endif
+
+/* Pull in FreeBSD specific shims */
+#include "skein_freebsd.h"
+
+#endif /* ifndef _SKEIN_H_ */
Index: sys/crypto/skein/skein.c
===================================================================
--- /dev/null
+++ sys/crypto/skein/skein.c
@@ -0,0 +1,858 @@
+/***********************************************************************
+**
+** Implementation of the Skein hash function.
+**
+** Source code author: Doug Whiting, 2008.
+**
+** This algorithm and source code is released to the public domain.
+**
+************************************************************************/
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/endian.h>
+#include <sys/types.h>
+
+/* get the memcpy/memset functions */
+#ifdef _KERNEL
+#include <sys/systm.h>
+#else
+#include <string.h>
+#endif
+
+#define SKEIN_PORT_CODE /* instantiate any code in skein_port.h */
+
+#include "skein.h" /* get the Skein API definitions */
+#include "skein_iv.h" /* get precomputed IVs */
+
+/*****************************************************************/
+/* External function to process blkCnt (nonzero) full block(s) of data. */
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx,const u08b_t *blkPtr,size_t blkCnt,size_t byteCntAdd);
+void Skein1024_Process_Block(Skein1024_Ctxt_t *ctx,const u08b_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
+ {
+ u08b_t b[SKEIN_256_STATE_BYTES];
+ u64b_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: memcpy(ctx->X,SKEIN_256_IV_256,sizeof(ctx->X)); break;
+ case 224: memcpy(ctx->X,SKEIN_256_IV_224,sizeof(ctx->X)); break;
+ case 160: memcpy(ctx->X,SKEIN_256_IV_160,sizeof(ctx->X)); break;
+ case 128: memcpy(ctx->X,SKEIN_256_IV_128,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) */
+ Skein_Start_New_Type(ctx,CFG_FINAL); /* set tweaks: T0=0; T1=CFG | FINAL */
+
+ cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER); /* set the schema, version */
+ cfg.w[1] = Skein_Swap64(hashBitLen); /* hash result length in bits */
+ cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+ memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+ /* compute the initial chaining values from config block */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the chaining variables */
+ 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,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+ {
+ union
+ {
+ u08b_t b[SKEIN_256_STATE_BYTES];
+ u64b_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? */
+ {
+ memset(ctx->X,0,sizeof(ctx->X)); /* no key: use all zeroes as key for config block */
+ }
+ else /* here to pre-process a key */
+ {
+ Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+ /* do a mini-Init right here */
+ ctx->h.hashBitLen=8*sizeof(ctx->X); /* set output hash bit count = state size */
+ Skein_Start_New_Type(ctx,KEY); /* set tweaks: T0 = 0; T1 = KEY type */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the initial chaining variables */
+ Skein_256_Update(ctx,key,keyBytes); /* hash the key */
+ Skein_256_Final_Pad(ctx,cfg.b); /* put result into cfg.b[] */
+ memcpy(ctx->X,cfg.b,sizeof(cfg.b)); /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+ {
+ uint_t i;
+ for (i=0;i<SKEIN_256_STATE_WORDS;i++) /* convert key bytes to context words */
+ 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);
+
+ memset(&cfg.w,0,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 */
+ cfg.w[2] = Skein_Swap64(treeInfo); /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+ 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 u08b_t *msg, size_t msgByteCnt)
+ {
+ size_t n;
+
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* process full blocks, if any */
+ if (msgByteCnt + ctx->h.bCnt > SKEIN_256_BLOCK_BYTES)
+ {
+ if (ctx->h.bCnt) /* finish up any buffered message data */
+ {
+ n = SKEIN_256_BLOCK_BYTES - ctx->h.bCnt; /* # bytes free in buffer b[] */
+ if (n)
+ {
+ Skein_assert(n < msgByteCnt); /* check on our logic here */
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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)
+ {
+ n = (msgByteCnt-1) / SKEIN_256_BLOCK_BYTES; /* number of full blocks to process */
+ 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);
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_256_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+
+ Skein_256_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_256_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_256_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_256_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_256_API_CodeSize(void)
+ {
+ return ((u08b_t *) Skein_256_API_CodeSize) -
+ ((u08b_t *) Skein_256_Init);
+ }
+#endif
+
+/*****************************************************************/
+/* 512-bit Skein */
+/*****************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a straight hashing operation */
+int Skein_512_Init(Skein_512_Ctxt_t *ctx, size_t hashBitLen)
+ {
+ union
+ {
+ u08b_t b[SKEIN_512_STATE_BYTES];
+ u64b_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: memcpy(ctx->X,SKEIN_512_IV_512,sizeof(ctx->X)); break;
+ case 384: memcpy(ctx->X,SKEIN_512_IV_384,sizeof(ctx->X)); break;
+ case 256: memcpy(ctx->X,SKEIN_512_IV_256,sizeof(ctx->X)); break;
+ case 224: memcpy(ctx->X,SKEIN_512_IV_224,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) */
+ Skein_Start_New_Type(ctx,CFG_FINAL); /* set tweaks: T0=0; T1=CFG | FINAL */
+
+ cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER); /* set the schema, version */
+ cfg.w[1] = Skein_Swap64(hashBitLen); /* hash result length in bits */
+ cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+ memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+ /* compute the initial chaining values from config block */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the chaining variables */
+ 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,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+ {
+ union
+ {
+ u08b_t b[SKEIN_512_STATE_BYTES];
+ u64b_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? */
+ {
+ memset(ctx->X,0,sizeof(ctx->X)); /* no key: use all zeroes as key for config block */
+ }
+ else /* here to pre-process a key */
+ {
+ Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+ /* do a mini-Init right here */
+ ctx->h.hashBitLen=8*sizeof(ctx->X); /* set output hash bit count = state size */
+ Skein_Start_New_Type(ctx,KEY); /* set tweaks: T0 = 0; T1 = KEY type */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the initial chaining variables */
+ Skein_512_Update(ctx,key,keyBytes); /* hash the key */
+ Skein_512_Final_Pad(ctx,cfg.b); /* put result into cfg.b[] */
+ memcpy(ctx->X,cfg.b,sizeof(cfg.b)); /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+ {
+ uint_t i;
+ for (i=0;i<SKEIN_512_STATE_WORDS;i++) /* convert key bytes to context words */
+ 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);
+
+ memset(&cfg.w,0,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 */
+ cfg.w[2] = Skein_Swap64(treeInfo); /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+ 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 u08b_t *msg, size_t msgByteCnt)
+ {
+ size_t n;
+
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* process full blocks, if any */
+ if (msgByteCnt + ctx->h.bCnt > SKEIN_512_BLOCK_BYTES)
+ {
+ if (ctx->h.bCnt) /* finish up any buffered message data */
+ {
+ n = SKEIN_512_BLOCK_BYTES - ctx->h.bCnt; /* # bytes free in buffer b[] */
+ if (n)
+ {
+ Skein_assert(n < msgByteCnt); /* check on our logic here */
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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)
+ {
+ n = (msgByteCnt-1) / SKEIN_512_BLOCK_BYTES; /* number of full blocks to process */
+ 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);
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_512_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+
+ Skein_512_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_512_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_512_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_512_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein_512_API_CodeSize(void)
+ {
+ return ((u08b_t *) Skein_512_API_CodeSize) -
+ ((u08b_t *) Skein_512_Init);
+ }
+#endif
+
+/*****************************************************************/
+/* 1024-bit Skein */
+/*****************************************************************/
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* init the context for a straight hashing operation */
+int Skein1024_Init(Skein1024_Ctxt_t *ctx, size_t hashBitLen)
+ {
+ union
+ {
+ u08b_t b[SKEIN1024_STATE_BYTES];
+ u64b_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: memcpy(ctx->X,SKEIN1024_IV_512 ,sizeof(ctx->X)); break;
+ case 384: memcpy(ctx->X,SKEIN1024_IV_384 ,sizeof(ctx->X)); break;
+ case 1024: memcpy(ctx->X,SKEIN1024_IV_1024,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) */
+ Skein_Start_New_Type(ctx,CFG_FINAL); /* set tweaks: T0=0; T1=CFG | FINAL */
+
+ cfg.w[0] = Skein_Swap64(SKEIN_SCHEMA_VER); /* set the schema, version */
+ cfg.w[1] = Skein_Swap64(hashBitLen); /* hash result length in bits */
+ cfg.w[2] = Skein_Swap64(SKEIN_CFG_TREE_INFO_SEQUENTIAL);
+ memset(&cfg.w[3],0,sizeof(cfg) - 3*sizeof(cfg.w[0])); /* zero pad config block */
+
+ /* compute the initial chaining values from config block */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the chaining variables */
+ 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,u64b_t treeInfo, const u08b_t *key, size_t keyBytes)
+ {
+ union
+ {
+ u08b_t b[SKEIN1024_STATE_BYTES];
+ u64b_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? */
+ {
+ memset(ctx->X,0,sizeof(ctx->X)); /* no key: use all zeroes as key for config block */
+ }
+ else /* here to pre-process a key */
+ {
+ Skein_assert(sizeof(cfg.b) >= sizeof(ctx->X));
+ /* do a mini-Init right here */
+ ctx->h.hashBitLen=8*sizeof(ctx->X); /* set output hash bit count = state size */
+ Skein_Start_New_Type(ctx,KEY); /* set tweaks: T0 = 0; T1 = KEY type */
+ memset(ctx->X,0,sizeof(ctx->X)); /* zero the initial chaining variables */
+ Skein1024_Update(ctx,key,keyBytes); /* hash the key */
+ Skein1024_Final_Pad(ctx,cfg.b); /* put result into cfg.b[] */
+ memcpy(ctx->X,cfg.b,sizeof(cfg.b)); /* copy over into ctx->X[] */
+#if SKEIN_NEED_SWAP
+ {
+ uint_t i;
+ for (i=0;i<SKEIN1024_STATE_WORDS;i++) /* convert key bytes to context words */
+ 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);
+
+ memset(&cfg.w,0,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 */
+ cfg.w[2] = Skein_Swap64(treeInfo); /* tree hash config info (or SKEIN_CFG_TREE_INFO_SEQUENTIAL) */
+
+ 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 u08b_t *msg, size_t msgByteCnt)
+ {
+ size_t n;
+
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* process full blocks, if any */
+ if (msgByteCnt + ctx->h.bCnt > SKEIN1024_BLOCK_BYTES)
+ {
+ if (ctx->h.bCnt) /* finish up any buffered message data */
+ {
+ n = SKEIN1024_BLOCK_BYTES - ctx->h.bCnt; /* # bytes free in buffer b[] */
+ if (n)
+ {
+ Skein_assert(n < msgByteCnt); /* check on our logic here */
+ memcpy(&ctx->b[ctx->h.bCnt],msg,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)
+ {
+ n = (msgByteCnt-1) / SKEIN1024_BLOCK_BYTES; /* number of full blocks to process */
+ 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);
+ memcpy(&ctx->b[ctx->h.bCnt],msg,msgByteCnt);
+ ctx->h.bCnt += msgByteCnt;
+ }
+
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the result */
+int Skein1024_Final(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN1024_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+
+ Skein1024_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN1024_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein1024_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN1024_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+#if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
+size_t Skein1024_API_CodeSize(void)
+ {
+ return ((u08b_t *) Skein1024_API_CodeSize) -
+ ((u08b_t *) Skein1024_Init);
+ }
+#endif
+
+/**************** 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, u08b_t *hashVal)
+ {
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_256_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_256_BLOCK_BYTES - ctx->h.bCnt);
+ Skein_256_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN_256_BLOCK_BYTES); /* "output" the state 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, u08b_t *hashVal)
+ {
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
+ Skein_512_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN_512_BLOCK_BYTES); /* "output" the state bytes */
+
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* finalize the hash computation and output the block, no OUTPUT stage */
+int Skein1024_Final_Pad(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL; /* tag as the final block */
+ if (ctx->h.bCnt < SKEIN1024_BLOCK_BYTES) /* zero pad b[] if necessary */
+ memset(&ctx->b[ctx->h.bCnt],0,SKEIN1024_BLOCK_BYTES - ctx->h.bCnt);
+ Skein1024_Process_Block(ctx,ctx->b,1,ctx->h.bCnt); /* process the final block */
+
+ Skein_Put64_LSB_First(hashVal,ctx->X,SKEIN1024_BLOCK_BYTES); /* "output" the state bytes */
+
+ return SKEIN_SUCCESS;
+ }
+
+#if SKEIN_TREE_HASH
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage */
+int Skein_256_Output(Skein_256_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_256_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_256_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_256_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_256_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_256_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage */
+int Skein_512_Output(Skein_512_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN_512_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN_512_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN_512_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein_512_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN_512_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+/*++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
+/* just do the OUTPUT stage */
+int Skein1024_Output(Skein1024_Ctxt_t *ctx, u08b_t *hashVal)
+ {
+ size_t i,n,byteCnt;
+ u64b_t X[SKEIN1024_STATE_WORDS];
+ Skein_Assert(ctx->h.bCnt <= SKEIN1024_BLOCK_BYTES,SKEIN_FAIL); /* catch uninitialized context */
+
+ /* now output the result */
+ byteCnt = (ctx->h.hashBitLen + 7) >> 3; /* total number of output bytes */
+
+ /* run Threefish in "counter mode" to generate output */
+ memset(ctx->b,0,sizeof(ctx->b)); /* zero out b[], so it can hold the counter */
+ memcpy(X,ctx->X,sizeof(X)); /* keep a local copy of counter mode "key" */
+ for (i=0;i*SKEIN1024_BLOCK_BYTES < byteCnt;i++)
+ {
+ ((u64b_t *)ctx->b)[0]= Skein_Swap64((u64b_t) i); /* build the counter block */
+ Skein_Start_New_Type(ctx,OUT_FINAL);
+ Skein1024_Process_Block(ctx,ctx->b,1,sizeof(u64b_t)); /* run "counter mode" */
+ n = byteCnt - i*SKEIN1024_BLOCK_BYTES; /* number of output bytes left to go */
+ 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);
+ memcpy(ctx->X,X,sizeof(X)); /* restore the counter mode key for next time */
+ }
+ return SKEIN_SUCCESS;
+ }
+
+
+/* Adapt the functions to match the prototype expected by libmd */
+void
+SKEIN256_Init(SKEIN256_CTX * ctx)
+{
+
+ Skein_256_Init(ctx, 256);
+}
+
+void
+SKEIN512_Init(SKEIN512_CTX * ctx)
+{
+
+ Skein_512_Init(ctx, 512);
+}
+
+void
+SKEIN1024_Init(SKEIN1024_CTX * ctx)
+{
+
+ Skein1024_Init(ctx, 1024);
+}
+
+void
+SKEIN256_Update(SKEIN256_CTX * ctx, const void *in, size_t len)
+{
+
+ Skein_256_Update(ctx, in, len);
+}
+
+void
+SKEIN512_Update(SKEIN512_CTX * ctx, const void *in, size_t len)
+{
+
+ Skein_512_Update(ctx, in, len);
+}
+
+void
+SKEIN1024_Update(SKEIN1024_CTX * ctx, const void *in, size_t len)
+{
+
+ Skein1024_Update(ctx, in, len);
+}
+
+void
+SKEIN256_Final(unsigned char digest[SKEIN_256_BLOCK_BYTES], SKEIN256_CTX * ctx)
+{
+
+ Skein_256_Final(ctx, digest);
+}
+
+void
+SKEIN512_Final(unsigned char digest[SKEIN_512_BLOCK_BYTES], SKEIN512_CTX * ctx)
+{
+
+ Skein_512_Final(ctx, digest);
+}
+
+void
+SKEIN1024_Final(unsigned char digest[SKEIN1024_BLOCK_BYTES], SKEIN1024_CTX * ctx)
+{
+
+ Skein1024_Final(ctx, digest);
+}
+
+#ifdef WEAK_REFS
+/* When building libmd, provide weak references. Note: this is not
+ activated in the context of compiling these sources for internal
+ use in libcrypt.
+ */
+#undef Skein_256_Init
+__weak_reference(_libmd_SKEIN256_Init, Skein_256_Init);
+#undef Skein_256_Update
+__weak_reference(_libmd_SKEIN256_Update, Skein_256_Update);
+#undef Skein_256_Final
+__weak_reference(_libmd_SKEIN256_Final, Skein_256_Final);
+
+#undef Skein_512_Init
+__weak_reference(_libmd_SKEIN512_Init, Skein_512_Init);
+#undef Skein_512_Update
+__weak_reference(_libmd_SKEIN512_Update, Skein_512_Update);
+#undef Skein_512_Final
+__weak_reference(_libmd_SKEIN512_Final, Skein_512_Final);
+
+#undef Skein1024_Init
+__weak_reference(_libmd_SKEIN1024_Init, Skein1024_Init);
+#undef Skein1024_Update
+__weak_reference(_libmd_SKEIN1024_Update, Skein1024_Update);
+#undef Skein1024_Final
+__weak_reference(_libmd_SKEIN1024_Final, Skein1024_Final);
+#endif
+
+#endif
Index: sys/crypto/skein/skein_block.c
===================================================================
--- /dev/null
+++ sys/crypto/skein/skein_block.c
@@ -0,0 +1,700 @@
+/***********************************************************************
+**
+** 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]
+**
+************************************************************************/
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/endian.h>
+#include <sys/types.h>
+
+#ifdef _KERNEL
+#include <sys/systm.h>
+#else
+#include <string.h>
+#endif
+
+#include "skein.h"
+
+#ifndef SKEIN_USE_ASM
+#define SKEIN_USE_ASM (0) /* default is all C code (no ASM) */
+#endif
+
+#ifndef SKEIN_LOOP
+#define SKEIN_LOOP 001 /* default: unroll 256 and 512, but not 1024 */
+#endif
+
+#define BLK_BITS (WCNT*64) /* some useful definitions for code here */
+#define KW_TWK_BASE (0)
+#define KW_KEY_BASE (3)
+#define ks (kw + KW_KEY_BASE)
+#define ts (kw + KW_TWK_BASE)
+
+#ifdef SKEIN_DEBUG
+#define DebugSaveTweak(ctx) { ctx->h.T[0] = ts[0]; ctx->h.T[1] = ts[1]; }
+#else
+#define DebugSaveTweak(ctx)
+#endif
+
+/***************************** Skein_256 ******************************/
+#if !(SKEIN_USE_ASM & 256)
+void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx,const u08b_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;
+ u64b_t kw[WCNT+4+RCNT*2]; /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+ u64b_t kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
+#endif
+ u64b_t X0,X1,X2,X3; /* local copy of context vars, for speed */
+ u64b_t w [WCNT]; /* local copy of input block */
+#ifdef SKEIN_DEBUG
+ const u64b_t *Xptr[4]; /* use for debugging (help compiler put Xn in registers) */
+ 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];
+
+ Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+ 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);
+
+ for (r=1;r < 2*RCNT;r+=2*SKEIN_UNROLL_256) /* loop thru it */
+#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 ((u08b_t *) Skein_256_Process_Block_CodeSize) -
+ ((u08b_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 u08b_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;
+ u64b_t kw[WCNT+4+RCNT*2]; /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+ u64b_t kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
+#endif
+ u64b_t X0,X1,X2,X3,X4,X5,X6,X7; /* local copy of vars, for speed */
+ u64b_t w [WCNT]; /* local copy of input block */
+#ifdef SKEIN_DEBUG
+ const u64b_t *Xptr[8]; /* use for debugging (help compiler put Xn in registers) */
+ 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];
+
+ Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+ 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);
+
+ for (r=1;r < 2*RCNT;r+=2*SKEIN_UNROLL_512) /* loop thru it */
+#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 ((u08b_t *) Skein_512_Process_Block_CodeSize) -
+ ((u08b_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 u08b_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;
+ u64b_t kw[WCNT+4+RCNT*2]; /* key schedule words : chaining vars + tweak + "rotation"*/
+#else
+ u64b_t kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
+#endif
+
+ u64b_t X00,X01,X02,X03,X04,X05,X06,X07, /* local copy of vars, for speed */
+ X08,X09,X10,X11,X12,X13,X14,X15;
+ u64b_t w [WCNT]; /* local copy of input block */
+#ifdef SKEIN_DEBUG
+ const u64b_t *Xptr[16]; /* use for debugging (help compiler put Xn in registers) */
+ 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];
+
+ Skein_Get64_LSB_First(w,blkPtr,WCNT); /* get input block in little-endian format */
+ 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);
+
+ for (r=1;r <= 2*RCNT;r+=2*SKEIN_UNROLL_1024) /* loop thru it */
+#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 ((u08b_t *) Skein1024_Process_Block_CodeSize) -
+ ((u08b_t *) Skein1024_Process_Block);
+ }
+uint_t Skein1024_Unroll_Cnt(void)
+ {
+ return SKEIN_UNROLL_1024;
+ }
+#endif
+#endif
Index: sys/crypto/skein/skein_freebsd.h
===================================================================
--- /dev/null
+++ sys/crypto/skein/skein_freebsd.h
@@ -0,0 +1,51 @@
+#ifndef _SKEIN_FREEBSD_H_
+#define _SKEIN_FREEBSD_H_
+
+#define SKEIN_256_BLOCK_BYTES ( 8*SKEIN_256_STATE_WORDS)
+#define SKEIN_512_BLOCK_BYTES ( 8*SKEIN_512_STATE_WORDS)
+#define SKEIN1024_BLOCK_BYTES ( 8*SKEIN1024_STATE_WORDS)
+
+#define SKEIN256_BLOCK_LENGTH SKEIN_256_BLOCK_BYTES
+#define SKEIN256_DIGEST_LENGTH 32
+#define SKEIN256_DIGEST_STRING_LENGTH (SKEIN256_DIGEST_LENGTH * 2 + 1)
+#define SKEIN512_BLOCK_LENGTH SKEIN_512_BLOCK_BYTES
+#define SKEIN512_DIGEST_LENGTH 64
+#define SKEIN512_DIGEST_STRING_LENGTH (SKEIN512_DIGEST_LENGTH * 2 + 1)
+#define SKEIN1024_BLOCK_LENGTH SKEIN1024_BLOCK_BYTES
+#define SKEIN1024_DIGEST_LENGTH 128
+#define SKEIN1024_DIGEST_STRING_LENGTH (SKEIN1024_DIGEST_LENGTH * 2 + 1)
+
+/* Make the context types look like the other hashes on FreeBSD */
+typedef Skein_256_Ctxt_t SKEIN256_CTX;
+typedef Skein_512_Ctxt_t SKEIN512_CTX;
+typedef Skein1024_Ctxt_t SKEIN1024_CTX;
+
+/* Make the prototypes look like the other hashes */
+void SKEIN256_Init (SKEIN256_CTX *ctx);
+void SKEIN512_Init (SKEIN512_CTX *ctx);
+void SKEIN1024_Init (SKEIN1024_CTX *ctx);
+
+void SKEIN256_Update(SKEIN256_CTX * ctx, const void *in, size_t len);
+void SKEIN512_Update(SKEIN512_CTX * ctx, const void *in, size_t len);
+void SKEIN1024_Update(SKEIN1024_CTX * ctx, const void *in, size_t len);
+
+void SKEIN256_Final(unsigned char digest[SKEIN256_DIGEST_LENGTH], SKEIN256_CTX * ctx);
+void SKEIN512_Final(unsigned char digest[SKEIN512_DIGEST_LENGTH], SKEIN512_CTX * ctx);
+void SKEIN1024_Final(unsigned char digest[SKEIN1024_DIGEST_LENGTH], SKEIN1024_CTX * ctx);
+
+#ifndef _KERNEL
+char *SKEIN256_End(SKEIN256_CTX *, char *);
+char *SKEIN512_End(SKEIN512_CTX *, char *);
+char *SKEIN1024_End(SKEIN1024_CTX *, char *);
+char *SKEIN256_Data(const void *, unsigned int, char *);
+char *SKEIN512_Data(const void *, unsigned int, char *);
+char *SKEIN1024_Data(const void *, unsigned int, char *);
+char *SKEIN256_File(const char *, char *);
+char *SKEIN512_File(const char *, char *);
+char *SKEIN1024_File(const char *, char *);
+char *SKEIN256_FileChunk(const char *, char *, off_t, off_t);
+char *SKEIN512_FileChunk(const char *, char *, off_t, off_t);
+char *SKEIN1024_FileChunk(const char *, char *, off_t, off_t);
+#endif
+
+#endif /* ifndef _SKEIN_FREEBSD_H_ */
Index: sys/crypto/skein/skein_iv.h
===================================================================
--- /dev/null
+++ sys/crypto/skein/skein_iv.h
@@ -0,0 +1,199 @@
+#ifndef _SKEIN_IV_H_
+#define _SKEIN_IV_H_
+
+#include "skein.h" /* get Skein macros and types */
+
+/*
+***************** 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.
+**
+************************************************************
+**/
+
+#define MK_64 SKEIN_MK_64
+
+/* blkSize = 256 bits. hashSize = 128 bits */
+const u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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 u64b_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)
+ };
+
+#endif /* _SKEIN_IV_H_ */
Index: sys/crypto/skein/skein_port.h
===================================================================
--- /dev/null
+++ sys/crypto/skein/skein_port.h
@@ -0,0 +1,167 @@
+#ifndef _SKEIN_PORT_H_
+#define _SKEIN_PORT_H_
+/*******************************************************************
+**
+** 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.
+**
+********************************************************************/
+
+#ifndef _KERNEL
+#include <sys/types.h>
+#endif
+
+#include "brg_types.h" /* get integer type definitions */
+
+typedef unsigned int uint_t; /* native unsigned integer */
+typedef uint_8t u08b_t; /* 8-bit unsigned integer */
+typedef uint_64t u64b_t; /* 64-bit unsigned integer */
+
+#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 "brg_endian.h" /* get endianness selection */
+#if PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN
+ /* here for big-endian CPUs */
+#define SKEIN_NEED_SWAP (1)
+#elif PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN
+ /* 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) memcpy(dst08,src64,bCnt)
+#define Skein_Get64_LSB_First(dst64,src08,wCnt) memcpy(dst64,src08,8*(wCnt))
+#endif
+#else
+#error "Skein needs endianness setting!"
+#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) \
+ ( (( ((u64b_t)(w64)) & 0xFF) << 56) | \
+ (((((u64b_t)(w64)) >> 8) & 0xFF) << 48) | \
+ (((((u64b_t)(w64)) >>16) & 0xFF) << 40) | \
+ (((((u64b_t)(w64)) >>24) & 0xFF) << 32) | \
+ (((((u64b_t)(w64)) >>32) & 0xFF) << 24) | \
+ (((((u64b_t)(w64)) >>40) & 0xFF) << 16) | \
+ (((((u64b_t)(w64)) >>48) & 0xFF) << 8) | \
+ (((((u64b_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(u08b_t *dst,const u64b_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] = (u08b_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(u64b_t *dst,const u08b_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] = (((u64b_t) src[n ]) ) +
+ (((u64b_t) src[n+1]) << 8) +
+ (((u64b_t) src[n+2]) << 16) +
+ (((u64b_t) src[n+3]) << 24) +
+ (((u64b_t) src[n+4]) << 32) +
+ (((u64b_t) src[n+5]) << 40) +
+ (((u64b_t) src[n+6]) << 48) +
+ (((u64b_t) src[n+7]) << 56) ;
+ }
+#else
+ ; /* output only the function prototype */
+#endif
+#endif /* ifndef Skein_Get64_LSB_First */
+
+/* Start FreeBSD libmd shims */
+
+/* Ensure libmd symbols do not clash with libcrypto */
+#ifndef Skein_256_Init
+#define Skein_256_Init _libmd_SKEIN256_Init
+#define Skein_512_Init _libmd_SKEIN512_Init
+#define Skein_1024_Init _libmd_SKEIN1024_Init
+#endif
+#ifndef Skein_256_Update
+#define Skein_256_Update _libmd_SKEIN256_Update
+#define Skein_512_Update _libmd_SKEIN512_Update
+#define Skein_1024_Update _libmd_SKEIN1024_Update
+#endif
+#ifndef Skein_256_Final
+#define Skein_256_Final _libmd_SKEIN256_Final
+#define Skein_512_Final _libmd_SKEIN512_Final
+#define Skein1024_Final _libmd_SKEIN1024_Final
+#endif
+#ifndef Skein_256_End
+#define Skein_256_End _libmd_SKEIN256_End
+#define Skein_512_End _libmd_SKEIN512_End
+#define Skein_1024_End _libmd_SKEIN1024_End
+#endif
+#ifndef Skein_256_File
+#define Skein_256_File _libmd_SKEIN256_File
+#define Skein_512_File _libmd_SKEIN512_File
+#define Skein_1024_File _libmd_SKEIN1024_File
+#endif
+#ifndef Skein_256_FileChunk
+#define Skein_256_FileChunk _libmd_SKEIN256_FileChunk
+#define Skein_512_FileChunk _libmd_SKEIN512_FileChunk
+#define Skein_1024_FileChunk _libmd_SKEIN1024_FileChunk
+#endif
+#ifndef Skein_256_Data
+#define Skein_256_Data _libmd_SKEIN256_Data
+#define Skein_512_Data _libmd_SKEIN512_Data
+#define Skein_1024_Data _libmd_SKEIN1024_Data
+#endif
+
+#endif /* ifndef _SKEIN_PORT_H_ */
Index: sys/modules/crypto/Makefile
===================================================================
--- sys/modules/crypto/Makefile
+++ sys/modules/crypto/Makefile
@@ -8,6 +8,7 @@
.PATH: ${.CURDIR}/../../crypto/rijndael
.PATH: ${.CURDIR}/../../crypto/sha2
.PATH: ${.CURDIR}/../../crypto/siphash
+.PATH: ${.CURDIR}/../../crypto/skein
KMOD = crypto
SRCS = crypto.c cryptodev_if.c
@@ -17,6 +18,13 @@
SRCS += camellia.c camellia-api.c
SRCS += des_ecb.c des_enc.c des_setkey.c
SRCS += sha1.c sha256c.c sha512c.c
+SRCS += skein.c skein_block.c
+.if exists(${MACHINE_ARCH}/skein_block_asm.s)
+.PATH: ${.CURDIR}/../../crypto/skein/${MACHINE_ARCH}
+SRCS += skein_block_asm.s
+CFLAGS += -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792
+ACFLAGS += -DELF -Wa,--noexecstack
+.endif
SRCS += siphash.c
SRCS += gmac.c gfmult.c
SRCS += opt_param.h cryptodev_if.h bus_if.h device_if.h

File Metadata

Mime Type
text/plain
Expires
Mon, Oct 5, 2:48 AM (3 h, 16 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
40223132
Default Alt Text
D6166.id15779.diff (499 KB)

Event Timeline