cipher: xts s390x

cipher/xts_asm.go

@@ -1,4 +1,4 @@
-//go:build (amd64 || arm64) && !purego
+//go:build (amd64 || arm64 || s390x) && !purego
 
 package cipher
 

cipher/xts_asm_test.go

@@ -1,4 +1,4 @@
-//go:build (amd64 || arm64) && !purego
+//go:build (amd64 || arm64 || s390x) && !purego
 
 package cipher
 

cipher/xts_generic.go

@@ -1,4 +1,4 @@
-//go:build purego || !(amd64 || arm64)
+//go:build purego || !(amd64 || arm64 || s390x)
 
 package cipher
 

cipher/xts_s390x.s

@@ -0,0 +1,119 @@
+// Copyright 2024 Sun Yimin. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+//go:build !purego
+
+#include "textflag.h"
+
+DATA xtsMask<>+0x00(SB)/8, $0x0f0e0d0c0b0a0908 // byte swap BE -> LE
+DATA xtsMask<>+0x08(SB)/8, $0x0706050403020100
+GLOBL xtsMask<>(SB), (NOPTR+RODATA), $16
+
+#define BSWAP V0
+#define POLY V1
+#define B0 V2
+#define T0 V3
+#define T1 V4
+#define CPOOL R3
+
+#define doubleTweak(B0, BSWAP, POLY, T0, T1) \
+	VPERM B0, B0, BSWAP, B0  \// BE -> LE
+	\ // Multiply by 2
+	VESRAF $31, B0, T0 \
+	VLREPF $0, T0, T0  \
+	VN POLY, T0, T0    \    // T0 for reduction
+	VREPIB $1, T1      \
+	VSL T1, B0, T1     \
+	VX T1, T0, B0      \
+	\
+	VPERM B0, B0, BSWAP, B0
+
+#define gbDoubleTweak(B0, POLY, T0, T1) \
+	VESLF $31, B0, T0   \
+	VESRAF $31, T0, T0  \
+	VLREPF $3, T0, T0   \
+	VN POLY, T0, T0     \ // T0 for reduction
+	\
+	VREPIB $1, T1       \
+	VSRL T1, B0, T1     \
+	VX T1, T0, B0
+
+// func mul2(tweak *[blockSize]byte, isGB bool)
+TEXT ·mul2(SB),NOSPLIT,$0
+	MOVD tweak+0(FP), R1
+	MOVB isGB+8(FP), R2
+
+	CMPBEQ R2, $1, gb_alg
+
+	MOVD $xtsMask<>+0x00(SB), CPOOL
+	VL  (CPOOL), BSWAP
+	
+	// Load polynomial for reduction
+	VZERO POLY
+	VLEIB $15, $0x87, POLY
+
+	// Load tweak
+	VL 0(R1), B0
+	doubleTweak(B0, BSWAP, POLY, T0, T1)
+	VST B0, 0(R1)
+
+gb_alg:
+	// Load polynomial for reduction
+	VZERO POLY
+	VLEIB $0, $0xe1, POLY
+
+	// Load tweak
+	VL 0(R1), B0
+	gbDoubleTweak(B0, POLY, T0, T1)
+	VST B0, 0(R1)
+
+	RET
+
+// func doubleTweaks(tweak *[blockSize]byte, tweaks []byte, isGB bool)
+TEXT ·doubleTweaks(SB),NOSPLIT,$0
+	MOVD tweak+0(FP), R1
+	MOVD tweaks+8(FP), R2
+	MOVD tweaks_len+16(FP), R3
+	MOVB isGB+32(FP), R4
+
+	AND	$-16, R3
+	LAY	(R2)(R3*1), R5
+
+	VL 0(R1), B0
+
+	CMPBEQ R4, $1, gb_alg
+
+	MOVD $xtsMask<>+0x00(SB), CPOOL
+	VL  (CPOOL), BSWAP
+	
+	// Load polynomial for reduction
+	VZERO POLY
+	VLEIB $15, $0x87, POLY
+
+loop:
+	VST B0, 0(R2)
+
+	doubleTweak(B0, BSWAP, POLY, T0, T1)
+
+	LA	16(R2), R2
+	CMPBLT	R2, R5, loop
+
+	VST B0, 0(R1)
+	RET
+
+gb_alg:	
+	// Load polynomial for reduction
+	VZERO POLY
+	VLEIB $0, $0xe1, POLY
+
+gb_alg_loop:
+	VST B0, 0(R2)
+
+	gbDoubleTweak(B0, POLY, T0, T1)
+
+	LA	16(R2), R2
+	CMPBLT	R2, R5, gb_alg_loop
+
+	VST B0, 0(R1)
+	RET