arne
/
cryptopp
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Fix ARMv8 AES Encryption 

ARMv8 AES decryption is not working at the moment. This check-in will allow us to test the current changes more widespread. We expected AES decryption failures only
pull/461/head
Jeffrey Walton 2017-08-06 21:25:36 -04:00
parent fd97121e8a
commit 701ec3aa1f
No known key found for this signature in database
GPG Key ID: B36AB348921B1838
3 changed files with 132 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
config.h
View File

@ -578,9 +578,6 @@ NAMESPACE_END
# endif
#endif
// TODO...
#undef CRYPTOPP_ARM_AES_AVAILABLE
#endif // ARM32, ARM64
// ***************** Miscellaneous ********************

105
rijndael-simd.cpp
View File

@ -10,6 +10,10 @@
#include "config.h"
#include "misc.h"
// TODO: Remove after debugging
#include <stdio.h>
#include <stdlib.h>
// Clang and GCC hoops...
#if !(defined(__ARM_FEATURE_CRYPTO) || defined(_MSC_VER))
# undef CRYPTOPP_ARM_AES_AVAILABLE
@ -122,6 +126,97 @@ bool CPU_TryAES_ARMV8()
#endif // ARM32 or ARM64
#if (CRYPTOPP_ARM_AES_AVAILABLE)
void PrintMessage(const byte *inBlock)
{
printf("M: ");
for (unsigned int j=0; j<16; ++j)
printf("%02X", inBlock[j]);
printf("\n");
}
void PrintCipher(const byte *outBlock)
{
printf("C: ");
for (unsigned int j=0; j<16; ++j)
printf("%02X", outBlock[j]);
printf("\n");
}
void PrintSubKeys(const word32 *keys, unsigned int rounds)
{
const byte* k = (const byte*)keys;
for (unsigned int i=0; i<rounds+1; ++i)
{
printf("R%d: ", i);
for (unsigned int j=0; j<16; ++j)
printf("%02X", *(k+(i*16)+j));
printf("\n");
}
}
void Rijndael_Enc_ProcessAndXorBlock_ARMV8(const byte *inBlock, const byte *xorBlock, byte *outBlock,
const word32 *subKeys, unsigned int rounds)
{
//PrintMessage(inBlock);
//PrintSubKeys(subKeys, rounds);
uint8x16_t data = vld1q_u8(inBlock);
const byte *keys = reinterpret_cast<const byte*>(subKeys);
unsigned int i;
for (i=0; i<rounds-1; ++i)
{
// AES single round encryption
data = vaeseq_u8(data, vld1q_u8(keys+i*16));
// AES mix columns
data = vaesmcq_u8(data);
}
// One round of encryption: AES, no Mix
data = vaeseq_u8(data, vld1q_u8(keys+i*16));
// Final Add (bitwise Xor)
data = veorq_u8(data, vld1q_u8(keys+(i+1)*16));
if (xorBlock)
vst1q_u8(outBlock, veorq_u8(data, vld1q_u8(xorBlock)));
else
vst1q_u8(outBlock, data);
//PrintCipher(outBlock);
}
void Rijndael_Dec_ProcessAndXorBlock_ARMV8(const byte *inBlock, const byte *xorBlock, byte *outBlock,
const word32 *subKeys, unsigned int rounds)
{
//PrintSubKeys(subKeys, rounds);
//PrintSubKeys(subKeys, rounds);
uint8x16_t data = vld1q_u8(inBlock);
const byte *keys = reinterpret_cast<const byte*>(subKeys);
// AES single round decryption
data = vaesdq_u8(data, vld1q_u8(keys));
unsigned int i;
for (i=0; i<rounds-1; ++i)
{
// AES mix columns
data = vaesmcq_u8(data);
// AES single round decryption
data = vaesdq_u8(data, vld1q_u8(keys+i*16));
}
// Final Add (bitwise Xor)
data = veorq_u8(data, vld1q_u8(keys+i*16));
if (xorBlock)
vst1q_u8(outBlock, veorq_u8(data, vld1q_u8(xorBlock)));
else
vst1q_u8(outBlock, data);
//PrintCipher(outBlock);
}
#endif // CRYPTOPP_ARM_AES_AVAILABLE
#if (CRYPTOPP_AESNI_AVAILABLE)
@ -200,7 +295,7 @@ static const word32 s_one[] = {0, 0, 0, 1<<24};
template <typename F1, typename F4>
inline size_t Rijndael_AdvancedProcessBlocks_AESNI(F1 func1, F4 func4,
MAYBE_CONST __m128i *subkeys, unsigned int rounds, const byte *inBlocks,
MAYBE_CONST __m128i *subkeys, size_t rounds, const byte *inBlocks,
const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
{
size_t blockSize = 16;
@ -310,20 +405,20 @@ inline size_t Rijndael_AdvancedProcessBlocks_AESNI(F1 func1, F4 func4,
return length;
}
size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(MAYBE_CONST word32 *subkeys, unsigned int rounds,
size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(MAYBE_CONST word32 *subkeys, size_t rounds,
const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
{
MAYBE_CONST __m128i* keys = reinterpret_cast<MAYBE_CONST __m128i*>(subkeys);
return Rijndael_AdvancedProcessBlocks_AESNI(AESNI_Enc_Block, AESNI_Enc_4_Blocks,
keys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
keys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
}
size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(MAYBE_CONST word32 *subkeys, unsigned int rounds,
size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(MAYBE_CONST word32 *subkeys, size_t rounds,
const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
{
MAYBE_CONST __m128i* keys = reinterpret_cast<MAYBE_CONST __m128i*>(subkeys);
return Rijndael_AdvancedProcessBlocks_AESNI(AESNI_Dec_Block, AESNI_Dec_4_Blocks,
keys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
keys, rounds, inBlocks, xorBlocks, outBlocks, length, flags);
}
void Rijndael_UncheckedSetKey_SSE4_AESNI(const byte *userKey, size_t keyLen, word32 *rk)

44
rijndael.cpp
View File

@ -74,12 +74,6 @@ being unloaded from L1 cache, until that round is finished.
#include "misc.h"
#include "cpu.h"
// TODO: remove...
#if (CRYPTOPP_ARM_AES_AVAILABLE)
# include "arm_neon.h"
# include "arm_acle.h"
#endif
NAMESPACE_BEGIN(CryptoPP)
// Hack for http://github.com/weidai11/cryptopp/issues/42 and http://github.com/weidai11/cryptopp/issues/132
@ -224,12 +218,19 @@ void Rijndael::Base::FillDecTable()
extern void Rijndael_UncheckedSetKey_SSE4_AESNI(const byte *userKey, size_t keyLen, word32* rk);
extern void Rijndael_UncheckedSetKeyRev_SSE4_AESNI(word32 *key, unsigned int rounds);
extern size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(const word32 *subkeys, unsigned int rounds,
extern size_t Rijndael_AdvancedProcessBlocks_Enc_AESNI(const word32 *subkeys, size_t rounds,
const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags);
extern size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(const word32 *subkeys, unsigned int rounds,
extern size_t Rijndael_AdvancedProcessBlocks_Dec_AESNI(const word32 *subkeys, size_t rounds,
const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags);
#endif
#if (CRYPTOPP_ARM_AES_AVAILABLE)
extern void Rijndael_Enc_ProcessAndXorBlock_ARMV8(const byte *inBlock, const byte *xorBlock, byte *outBlock,
const word32 *subKeys, unsigned int rounds);
extern void Rijndael_Dec_ProcessAndXorBlock_ARMV8(const byte *inBlock, const byte *xorBlock, byte *outBlock,
const word32 *subKeys, unsigned int rounds);
#endif
void Rijndael::Base::UncheckedSetKey(const byte *userKey, unsigned int keyLen, const NameValuePairs &)
{
AssertValidKeyLength(keyLen);
@ -327,22 +328,34 @@ void Rijndael::Base::UncheckedSetKey(const byte *userKey, unsigned int keyLen, c
if (HasAESNI())
ConditionalByteReverse(BIG_ENDIAN_ORDER, rk+4, rk+4, (m_rounds-1)*16);
#endif
#if CRYPTOPP_ARM_AES_AVAILABLE
if (HasAES())
ConditionalByteReverse(BIG_ENDIAN_ORDER, rk+4, rk+4, (m_rounds-1)*16);
#endif
}
void Rijndael::Enc::ProcessAndXorBlock(const byte *inBlock, const byte *xorBlock, byte *outBlock) const
{
#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE) || CRYPTOPP_AESNI_AVAILABLE
#if (CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)) && !defined(CRYPTOPP_DISABLE_RIJNDAEL_ASM)
# if (CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)) && !defined(CRYPTOPP_DISABLE_RIJNDAEL_ASM)
if (HasSSE2())
#else
# else
if (HasAESNI())
#endif
# endif
{
(void)Rijndael::Enc::AdvancedProcessBlocks(inBlock, xorBlock, outBlock, 16, 0);
return;
}
#endif
#if (CRYPTOPP_ARM_AES_AVAILABLE)
if (HasAES())
{
Rijndael_Enc_ProcessAndXorBlock_ARMV8(inBlock, xorBlock, outBlock, m_key.begin(), m_rounds);
return;
}
#endif
typedef BlockGetAndPut<word32, NativeByteOrder> Block;
word32 s0, s1, s2, s3, t0, t1, t2, t3;
@ -421,6 +434,14 @@ void Rijndael::Dec::ProcessAndXorBlock(const byte *inBlock, const byte *xorBlock
}
#endif
#if (CRYPTOPP_ARM_AES_AVAILABLE) && 0
if (HasAES())
{
Rijndael_Dec_ProcessAndXorBlock_ARMV8(inBlock, xorBlock, outBlock, m_key.begin(), m_rounds);
return;
}
#endif
typedef BlockGetAndPut<word32, NativeByteOrder> Block;
word32 s0, s1, s2, s3, t0, t1, t2, t3;
@ -1049,7 +1070,6 @@ size_t Rijndael::Enc::AdvancedProcessBlocks(const byte *inBlocks, const byte *xo
#if CRYPTOPP_AESNI_AVAILABLE
if (HasAESNI())
return Rijndael_AdvancedProcessBlocks_Enc_AESNI(m_key.begin(), m_rounds, inBlocks, xorBlocks, outBlocks, length, flags);
#endif
#if (CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)) && !defined(CRYPTOPP_DISABLE_RIJNDAEL_ASM)