From 3e32a52bee01eeb0171460f148ca2db501693d0e Mon Sep 17 00:00:00 2001 From: Emman Date: Thu, 13 Jan 2022 08:40:32 +0800 Subject: [PATCH] [SM4] try arm64 gcmSm4Init --- sm4/gcm_arm64.s | 125 ++++++++++++++++++++----------------------- sm4/sm4_gcm_arm64.go | 6 +-- sm4/sm4_gcm_test.go | 4 +- 3 files changed, 63 insertions(+), 72 deletions(-) diff --git a/sm4/gcm_arm64.s b/sm4/gcm_arm64.s index 6563de6..298d2b7 100644 --- a/sm4/gcm_arm64.s +++ b/sm4/gcm_arm64.s @@ -107,71 +107,45 @@ TEXT ·gcmSm4Finish(SB),NOSPLIT,$0 #undef dlen #define SM4_SBOX(x, y, z, z1, z2) \ - MOVD $0x0F0F0F0F0F0F0F0F, R19; \ - VMOV R19, z1.D2; \ // nibble mask - VAND x.B16, z1.B16, z2.B16; \ - MOVD $0x9197E2E474720701, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0xC7C1B4B222245157, R19; \ - VMOV R19, z.D[1]; \ // m1 low - VTBL z2.B16, [z.B16], y.B16; \ - VUSHR $4, x.D2, x.D2; \ - VAND x.B16, z1.B16, z2.B16; \ - MOVD $0xE240AB09EB49A200, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0xF052B91BF95BB012, R19; \ - VMOV R19, z.D[1]; \ // m1 high - VTBL z2.B16, [z.B16], z2.B16; \ - VEOR y.B16, z2.B16, x.B16; \ - MOVD $0x0B0E0104070A0D00, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0x0306090C0F020508, R19; \ - VMOV R19, z.D[1]; \ // inverse shift row - VTBL z.B16, [x.B16], x.B16; \ - AESE ZERO.B16, x.B16; \ - VAND x.B16, z1.B16, z2.B16; \ - MOVD $0x5B67F2CEA19D0834, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0xEDD14478172BBE82, R19; \ - VMOV R19, z.D[1]; \ // m2 low - VTBL z2.B16, [z.B16], y.B16; \ - VUSHR $4, x.D2, x.D2; \ - VAND x.B16, z1.B16, z2.B16; \ - MOVD $0xAE7201DD73AFDC00, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0x11CDBE62CC1063BF, R19; \ - VMOV R19, z.D[1]; \ // m2 high - VTBL z2.B16, [z.B16], z2.B16; \ + VMOVQ $0x0F0F0F0F0F0F0F0F, $0x0F0F0F0F0F0F0F0F, z1; \ // nibble mask + VAND x.B16, z1.B16, z2.B16; \ + VMOVQ $0xC7C1B4B222245157, $0x9197E2E474720701, z; \ + VTBL z2.B16, [z.B16], y.B16; \ + VUSHR $4, x.D2, x.D2; \ + VAND x.B16, z1.B16, z2.B16; \ + VMOVQ $0xF052B91BF95BB012, $0xE240AB09EB49A200, z; \ + VTBL z2.B16, [z.B16], z2.B16; \ + VEOR y.B16, z2.B16, x.B16; \ + VMOVQ $0x0306090C0F020508, $0x0B0E0104070A0D00, z; \ + VTBL z.B16, [x.B16], x.B16; \ + AESE ZERO.B16, x.B16; \ + VAND x.B16, z1.B16, z2.B16; \ + VMOVQ $0xEDD14478172BBE82, $0x5B67F2CEA19D0834, z; \ + VTBL z2.B16, [z.B16], y.B16; \ + VUSHR $4, x.D2, x.D2; \ + VAND x.B16, z1.B16, z2.B16; \ + VMOVQ $0x11CDBE62CC1063BF, $0xAE7201DD73AFDC00, z; \ + VTBL z2.B16, [z.B16], z2.B16; \ VEOR y.B16, z2.B16, x.B16 #define SM4_TAO_L1(x, y, z, z1, z2) \ - SM4_SBOX(x, y, z, z1, z2); \ - ; \ - MOVD $0x0605040702010003, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0x0E0D0C0F0A09080B, R19; \ - VMOV R19, z.D[1]; \ // r08 mask - VTBL z.B16, [x.B16], y.B16; \ - VEOR y.B16, x.B16, y.B16; \ - MOVD $0x0504070601000302, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0x0D0C0F0E09080B0A, R19; \ - VMOV R19, z.D[1]; \ // r16 mask - VTBL z.B16, [x.B16], z.B16; \ - VEOR z.B16, y.B16, y.B16; \ - VSHL $2, y.S4, z.S4; \ - VUSHR $30, y.S4, y.S4; \ - VORR y.B16, z.B16, y.B16; \ - MOVD $0x0407060500030201, R19; \ - VMOV R19, z.D[0]; \ - MOVD $0x0C0F0E0D080B0A09, R19; \ - VMOV R19, z.D[1]; \ // r24 mask - VTBL z.B16, [x.B16], z.B16; \ - VEOR z.B16, x.B16, x.B16; \ + SM4_SBOX(x, y, z, z1, z2); \ + VMOVQ $0x0E0D0C0F0A09080B, $0x0605040702010003, z; \ + VTBL z.B16, [x.B16], y.B16; \ + VEOR y.B16, x.B16, y.B16; \ + VMOVQ $0x0D0C0F0E09080B0A, $0x0504070601000302, z; \ + VTBL z.B16, [x.B16], z.B16; \ + VEOR z.B16, y.B16, y.B16; \ + VSHL $2, y.S4, z.S4; \ + VUSHR $30, y.S4, y.S4; \ + VORR y.B16, z.B16, y.B16; \ + VMOVQ $0x0C0F0E0D080B0A09, $0x0407060500030201, z; \ + VTBL z.B16, [x.B16], z.B16; \ + VEOR z.B16, x.B16, x.B16; \ VEOR y.B16, x.B16, x.B16 -#define SM4_ROUND(RK, x, y, z, z1, z2, t0, t1, t2, t3) \ - MOVW.P 4(RK), R19; \ +#define SM4_ROUND(index, RK, x, y, z, z1, z2, t0, t1, t2, t3) \ + MOVW (index*4)(RK), R19; \ VMOV R19, x.S4; \ VEOR t1.B16, x.B16, x.B16; \ VEOR t2.B16, x.B16, x.B16; \ @@ -179,10 +153,10 @@ TEXT ·gcmSm4Finish(SB),NOSPLIT,$0 SM4_TAO_L1(x, y, z, z1, z2); \ VEOR x.B16, t0.B16, t0.B16 -// func precomputeTableAsm(productTable *[256]byte, src *[16]byte) -TEXT ·precomputeTableAsm(SB),NOSPLIT,$0 +// func gcmSm4Init(productTable *[256]byte, rk []uint32) +TEXT ·gcmSm4Init(SB),NOSPLIT,$0 #define pTbl R0 -#define SRC R1 +#define RK R1 #define I R2 MOVD productTable+0(FP), pTbl @@ -195,8 +169,27 @@ TEXT ·precomputeTableAsm(SB),NOSPLIT,$0 VMOV I, POLY.D[1] VEOR ZERO.B16, ZERO.B16, ZERO.B16 - VLD1 (SRC), [B0.B16] - VREV64 B0.B16, B0.B16 + // Encrypt block 0 with the SM4 keys to generate the hash key H + VEOR B0.B16, B0.B16, B0.B16 + VEOR B1.B16, B1.B16, B1.B16 + VEOR B2.B16, B2.B16, B2.B16 + VEOR B3.B16, B3.B16, B3.B16 + EOR R3, R3 + +sm4InitEncLoop: + SM4_ROUND(0, RK, K0, K1, K2, K3, K4, B0, B1, B2, B3) + SM4_ROUND(1, RK, K0, K1, K2, K3, K4, B1, B2, B3, B0) + SM4_ROUND(2, RK, K0, K1, K2, K3, K4, B2, B3, B0, B1) + SM4_ROUND(3, RK, K0, K1, K2, K3, K4, B3, B0, B1, B2) + + ADD $16, R3 + CMP $128, R3 + BNE sm4InitEncLoop + + VMOV B0.S[0], B0.S[2] + VMOV B1.S[0], B0.S[3] + VMOV B2.S[0], B0.S[0] + VMOV B3.S[0], B0.S[1] // Multiply by 2 modulo P VMOV B0.D[0], I @@ -254,7 +247,7 @@ initLoop: BNE initLoop RET #undef I -#undef SRC +#undef RK #undef pTbl // func gcmSm4Data(productTable *[256]byte, data []byte, T *[16]byte) diff --git a/sm4/sm4_gcm_arm64.go b/sm4/sm4_gcm_arm64.go index ed4dd8b..9949176 100644 --- a/sm4/sm4_gcm_arm64.go +++ b/sm4/sm4_gcm_arm64.go @@ -21,7 +21,7 @@ type sm4CipherGCM struct { var _ gcmAble = (*sm4CipherGCM)(nil) //go:noescape -func precomputeTableAsm(productTable *[256]byte, src *[16]byte) +func gcmSm4Init(productTable *[256]byte, rk []uint32) //go:noescape func gcmSm4Data(productTable *[256]byte, data []byte, T *[16]byte) @@ -41,9 +41,7 @@ func (c *sm4CipherGCM) NewGCM(nonceSize, tagSize int) (cipher.AEAD, error) { g.cipher = &c.sm4CipherAsm g.nonceSize = nonceSize g.tagSize = tagSize - var key [gcmBlockSize]byte - c.Encrypt(key[:], key[:]) - precomputeTableAsm(&g.bytesProductTable, &key) + gcmSm4Init(&g.bytesProductTable, g.cipher.enc) return g, nil } diff --git a/sm4/sm4_gcm_test.go b/sm4/sm4_gcm_test.go index 4b7c7f1..75acca9 100644 --- a/sm4/sm4_gcm_test.go +++ b/sm4/sm4_gcm_test.go @@ -1,5 +1,5 @@ -//go:build amd64 -// +build amd64 +//go:build amd64 || arm64 +// +build amd64 arm64 package sm4