diff --git a/sm4/cbc_cipher_asm.go b/sm4/cbc_cipher_asm.go index 889eed4..e584a09 100644 --- a/sm4/cbc_cipher_asm.go +++ b/sm4/cbc_cipher_asm.go @@ -45,22 +45,20 @@ func (x *cbc) CryptBlocks(dst, src []byte) { } end := len(src) copy(x.tmp, src[end-BlockSize:end]) - start := end - FourBlocksSize - var temp []byte = make([]byte, FourBlocksSize) - var src64 []byte = make([]byte, FourBlocksSize) + start := end - x.b.blocksSize + var temp []byte = make([]byte, x.b.blocksSize) + var batchSrc []byte = make([]byte, x.b.blocksSize) for start > 0 { encryptBlocksAsm(&x.b.dec[0], &temp[0], &src[start:end][0]) - xor.XorBytes(dst[end-BlockSize:end], temp[FourBlocksSize-BlockSize:FourBlocksSize], src[end-2*BlockSize:end-BlockSize]) - xor.XorBytes(dst[end-2*BlockSize:end-BlockSize], temp[FourBlocksSize-2*BlockSize:FourBlocksSize-BlockSize], src[end-3*BlockSize:end-2*BlockSize]) - xor.XorBytes(dst[end-3*BlockSize:end-2*BlockSize], temp[FourBlocksSize-3*BlockSize:FourBlocksSize-2*BlockSize], src[end-4*BlockSize:end-3*BlockSize]) - xor.XorBytes(dst[end-4*BlockSize:end-3*BlockSize], temp[:BlockSize], src[end-5*BlockSize:end-4*BlockSize]) - + for i := 0; i < x.b.batchBlocks; i++ { + xor.XorBytes(dst[end-(i+1)*BlockSize:end-i*BlockSize], temp[x.b.blocksSize-(i+1)*BlockSize:x.b.blocksSize-i*BlockSize], src[end-(i+2)*BlockSize:end-(i+1)*BlockSize]) + } end = start - start -= FourBlocksSize + start -= x.b.blocksSize } - copy(src64, src[:end]) - encryptBlocksAsm(&x.b.dec[0], &temp[0], &src[:end][0]) + copy(batchSrc, src[:end]) + encryptBlocksAsm(&x.b.dec[0], &temp[0], &batchSrc[0]) count := end / BlockSize for i := count; i > 1; i-- { xor.XorBytes(dst[end-BlockSize:end], temp[end-BlockSize:end], src[end-2*BlockSize:end-BlockSize]) diff --git a/sm4/cipher_asm.go b/sm4/cipher_asm.go index 98cb55f..d835007 100644 --- a/sm4/cipher_asm.go +++ b/sm4/cipher_asm.go @@ -25,13 +25,15 @@ func expandKeyAsm(key *byte, ck, enc, dec *uint32) type sm4CipherAsm struct { sm4Cipher + batchBlocks int + blocksSize int } func newCipher(key []byte) (cipher.Block, error) { if !supportsAES { return newCipherGeneric(key) } - c := sm4CipherAsm{sm4Cipher{make([]uint32, rounds), make([]uint32, rounds)}} + c := sm4CipherAsm{sm4Cipher{make([]uint32, rounds), make([]uint32, rounds)}, 4, 4 * BlockSize} expandKeyAsm(&key[0], &ck[0], &c.enc[0], &c.dec[0]) if supportsAES && supportsGFMUL { return &sm4CipherGCM{c}, nil @@ -39,13 +41,9 @@ func newCipher(key []byte) (cipher.Block, error) { return &c, nil } -const FourBlocksSize = 64 - -const BatchBlocks = 4 - func (c *sm4CipherAsm) BlockSize() int { return BlockSize } -func (c *sm4CipherAsm) Concurrency() int { return BatchBlocks } +func (c *sm4CipherAsm) Concurrency() int { return c.batchBlocks } func (c *sm4CipherAsm) Encrypt(dst, src []byte) { if len(src) < BlockSize { @@ -61,13 +59,13 @@ func (c *sm4CipherAsm) Encrypt(dst, src []byte) { } func (c *sm4CipherAsm) EncryptBlocks(dst, src []byte) { - if len(src) < FourBlocksSize { + if len(src) < c.blocksSize { panic("sm4: input not full blocks") } - if len(dst) < FourBlocksSize { + if len(dst) < c.blocksSize { panic("sm4: output not full blocks") } - if subtle.InexactOverlap(dst[:FourBlocksSize], src[:FourBlocksSize]) { + if subtle.InexactOverlap(dst[:c.blocksSize], src[:c.blocksSize]) { panic("sm4: invalid buffer overlap") } encryptBlocksAsm(&c.enc[0], &dst[0], &src[0]) @@ -87,13 +85,13 @@ func (c *sm4CipherAsm) Decrypt(dst, src []byte) { } func (c *sm4CipherAsm) DecryptBlocks(dst, src []byte) { - if len(src) < FourBlocksSize { + if len(src) < c.blocksSize { panic("sm4: input not full blocks") } - if len(dst) < FourBlocksSize { + if len(dst) < c.blocksSize { panic("sm4: output not full blocks") } - if subtle.InexactOverlap(dst[:FourBlocksSize], src[:FourBlocksSize]) { + if subtle.InexactOverlap(dst[:c.blocksSize], src[:c.blocksSize]) { panic("sm4: invalid buffer overlap") } encryptBlocksAsm(&c.dec[0], &dst[0], &src[0]) diff --git a/sm4/ctr_cipher_asm.go b/sm4/ctr_cipher_asm.go index 57703a3..337699e 100644 --- a/sm4/ctr_cipher_asm.go +++ b/sm4/ctr_cipher_asm.go @@ -34,14 +34,14 @@ func (c *sm4CipherAsm) NewCTR(iv []byte) cipher.Stream { } s := &ctr{ b: c, - ctr: make([]byte, 4*len(iv)), + ctr: make([]byte, c.batchBlocks*len(iv)), out: make([]byte, 0, bufSize), outUsed: 0, } copy(s.ctr, iv) - s.genCtr(BlockSize) - s.genCtr(2 * BlockSize) - s.genCtr(3 * BlockSize) + for i := 1; i < c.batchBlocks; i++ { + s.genCtr(i * BlockSize) + } return s } @@ -66,15 +66,14 @@ func (x *ctr) refill() { remain := len(x.out) - x.outUsed copy(x.out, x.out[x.outUsed:]) x.out = x.out[:cap(x.out)] - for remain <= len(x.out)-FourBlocksSize { + for remain <= len(x.out)-x.b.blocksSize { encryptBlocksAsm(&x.b.enc[0], &x.out[remain:][0], &x.ctr[0]) - remain += FourBlocksSize + remain += x.b.blocksSize // Increment counter - x.genCtr(0) - x.genCtr(BlockSize) - x.genCtr(2 * BlockSize) - x.genCtr(3 * BlockSize) + for i := 0; i < x.b.batchBlocks; i++ { + x.genCtr(i * BlockSize) + } } x.out = x.out[:remain] x.outUsed = 0 diff --git a/sm4/gcm_cipher_asm.go b/sm4/gcm_cipher_asm.go index 5845243..5e50b29 100644 --- a/sm4/gcm_cipher_asm.go +++ b/sm4/gcm_cipher_asm.go @@ -265,32 +265,27 @@ func gcmInc32(counterBlock *[16]byte) { // counterCrypt crypts in to out using g.cipher in counter mode. func (g *gcm) counterCrypt(out, in []byte, counter *[gcmBlockSize]byte) { - var mask [FourBlocksSize]byte - var couters [FourBlocksSize]byte + mask := make([]byte, g.cipher.blocksSize) + counters := make([]byte, g.cipher.blocksSize) - for len(in) >= FourBlocksSize { - copy(couters[:], counter[:]) - gcmInc32(counter) - copy(couters[gcmBlockSize:], counter[:]) - gcmInc32(counter) - copy(couters[2*gcmBlockSize:], counter[:]) - gcmInc32(counter) - copy(couters[3*gcmBlockSize:], counter[:]) - - encryptBlocksAsm(&g.cipher.enc[0], &mask[0], &couters[0]) - gcmInc32(counter) + for len(in) >= g.cipher.blocksSize { + for i := 0; i < g.cipher.batchBlocks; i++ { + copy(counters[i*gcmBlockSize:(i+1)*gcmBlockSize], counter[:]) + gcmInc32(counter) + } + encryptBlocksAsm(&g.cipher.enc[0], &mask[0], &counters[0]) xor.XorWords(out, in, mask[:]) - out = out[FourBlocksSize:] - in = in[FourBlocksSize:] + out = out[g.cipher.blocksSize:] + in = in[g.cipher.blocksSize:] } if len(in) > 0 { blocks := (len(in) + gcmBlockSize - 1) / gcmBlockSize for i := 0; i < blocks; i++ { - copy(couters[i*gcmBlockSize:], counter[:]) + copy(counters[i*gcmBlockSize:], counter[:]) gcmInc32(counter) } - encryptBlocksAsm(&g.cipher.enc[0], &mask[0], &couters[0]) + encryptBlocksAsm(&g.cipher.enc[0], &mask[0], &counters[0]) xor.XorBytes(out, in, mask[:blocks*gcmBlockSize]) } } diff --git a/sm4/sm4_gcm_test.go b/sm4/sm4_gcm_test.go index c16d7bb..d892f65 100644 --- a/sm4/sm4_gcm_test.go +++ b/sm4/sm4_gcm_test.go @@ -10,7 +10,7 @@ import ( func genPrecomputeTable() *gcmAsm { key := []byte{0x01, 0x23, 0x45, 0x67, 0x89, 0xab, 0xcd, 0xef, 0xfe, 0xdc, 0xba, 0x98, 0x76, 0x54, 0x32, 0x10} - c := sm4CipherAsm{sm4Cipher{make([]uint32, rounds), make([]uint32, rounds)}} + c := sm4CipherAsm{sm4Cipher{make([]uint32, rounds), make([]uint32, rounds)}, 4, 64} expandKeyAsm(&key[0], &ck[0], &c.enc[0], &c.dec[0]) c1 := &sm4CipherGCM{c} g := &gcmAsm{} diff --git a/sm4_test/cbc_sm4_test.go b/sm4_test/cbc_sm4_test.go index 72ba303..9b90131 100644 --- a/sm4_test/cbc_sm4_test.go +++ b/sm4_test/cbc_sm4_test.go @@ -75,6 +75,37 @@ var cbcSM4Tests = []struct { 0x34, 0x6e, 0x9d, 0xad, 0xe1, 0x8a, 0xf4, 0xa1, 0x83, 0x69, 0x57, 0xb9, 0x37, 0x26, 0x7e, 0x03, }, }, + { + "Five blocks", + []byte("0123456789ABCDEF"), + []byte("0123456789ABCDEF"), + []byte("Hello World Hello World Hello World Hello World Hello World Hello World Hello Wo"), + []byte{ + 0xd3, 0x1e, 0x36, 0x83, 0xe4, 0xfc, 0x9b, 0x51, 0x6a, 0x2c, 0x0f, 0x98, 0x36, 0x76, 0xa9, 0xeb, + 0x1f, 0xdc, 0xc3, 0x2a, 0xf3, 0x84, 0x08, 0x97, 0x81, 0x57, 0xa2, 0x06, 0x5d, 0xe3, 0x4c, 0x6a, + 0xe0, 0x02, 0xd6, 0xe4, 0xf5, 0x66, 0x87, 0xc4, 0xcc, 0x54, 0x1d, 0x1f, 0x1c, 0xc4, 0x2f, 0xe6, + 0xe5, 0x1d, 0xea, 0x52, 0xb8, 0x0c, 0xc8, 0xbe, 0xae, 0xcc, 0x44, 0xa8, 0x51, 0x81, 0x08, 0x60, + 0xb6, 0x09, 0x7b, 0xb8, 0x7e, 0xdb, 0x53, 0x4b, 0xea, 0x2a, 0xc6, 0xa1, 0xe5, 0xa0, 0x2a, 0xe9, + 0x62, 0xb5, 0xe7, 0x50, 0x44, 0xea, 0x24, 0xcc, 0x9b, 0x5e, 0x07, 0x48, 0x04, 0x89, 0xa2, 0x74, + }, + }, + { + "9 blocks", + []byte("0123456789ABCDEF"), + []byte("0123456789ABCDEF"), + []byte("Hello World Hello World Hello World Hello World Hello World Hello World Hello World Hello World Hello World Hello World Hello World"), + []byte{ + 0xd3, 0x1e, 0x36, 0x83, 0xe4, 0xfc, 0x9b, 0x51, 0x6a, 0x2c, 0x0f, 0x98, 0x36, 0x76, 0xa9, 0xeb, + 0x1f, 0xdc, 0xc3, 0x2a, 0xf3, 0x84, 0x08, 0x97, 0x81, 0x57, 0xa2, 0x06, 0x5d, 0xe3, 0x4c, 0x6a, + 0xe0, 0x02, 0xd6, 0xe4, 0xf5, 0x66, 0x87, 0xc4, 0xcc, 0x54, 0x1d, 0x1f, 0x1c, 0xc4, 0x2f, 0xe6, + 0xe5, 0x1d, 0xea, 0x52, 0xb8, 0x0c, 0xc8, 0xbe, 0xae, 0xcc, 0x44, 0xa8, 0x51, 0x81, 0x08, 0x60, + 0xb6, 0x09, 0x7b, 0xb8, 0x7e, 0xdb, 0x53, 0x4b, 0xea, 0x2a, 0xc6, 0xa1, 0xe5, 0xa0, 0x2a, 0xe9, + 0x22, 0x65, 0x5b, 0xa3, 0xb9, 0xcc, 0x63, 0x92, 0x16, 0x0e, 0x2f, 0xf4, 0x3b, 0x93, 0x06, 0x82, + 0xb3, 0x8c, 0x26, 0x2e, 0x06, 0x51, 0x34, 0x2c, 0xe4, 0x3d, 0xd0, 0xc7, 0x2b, 0x8f, 0x31, 0x15, + 0x30, 0xa8, 0x96, 0x1c, 0xbc, 0x8e, 0xf7, 0x4f, 0x6b, 0x69, 0x9d, 0xc9, 0x40, 0x89, 0xd7, 0xe8, + 0xf7, 0x90, 0x47, 0x74, 0xaf, 0x40, 0xfd, 0x72, 0xc6, 0x17, 0xeb, 0xc0, 0x8b, 0x01, 0x71, 0x5c, + }, + }, { "A.1", []byte{0x01, 0x23, 0x45, 0x67, 0x89, 0xab, 0xcd, 0xef, 0xfe, 0xdc, 0xba, 0x98, 0x76, 0x54, 0x32, 0x10}, @@ -103,6 +134,12 @@ func TestCBCEncrypterSM4(t *testing.T) { encrypter.CryptBlocks(data, data) if !bytes.Equal(test.out, data) { t.Errorf("%s: CBCEncrypter\nhave %s\nwant %x", test.name, hex.EncodeToString(data), test.out) + for i := 0; i < len(data); i++ { + fmt.Printf("0x%02x, ", data[i]) + if (i+1)%16 == 0 { + fmt.Println() + } + } } } }