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

Fixup under-aligned buffers when using AES on AltiVec and Power8 

This commit supports the upcoming AltiVec and Power8 processor. This commit affects a number of classes due to the ubiquitous use of AES. The commit provides the data alignment requirements.
pull/489/head
Jeffrey Walton 2017-09-04 11:21:47 -04:00
parent 32cc92e048
commit 75aef9bded
No known key found for this signature in database
GPG Key ID: B36AB348921B1838
3 changed files with 16 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
rijndael-simd.cpp
View File

@ -10,6 +10,9 @@
// Skip Hovsmith and Barry O'Rourke for the mbedTLS project. Stepping // Skip Hovsmith and Barry O'Rourke for the mbedTLS project. Stepping
// mbedTLS under a debugger was helped for us to determine problems // mbedTLS under a debugger was helped for us to determine problems
// with our subkey generation and scheduling. // with our subkey generation and scheduling.
//
// AltiVec and Power8 code based on "POWER8 in-core cryptography."
// http://www.ibm.com/developerworks/library/se-power8-in-core-cryptography/index.html
#include "pch.h" #include "pch.h"
#include "config.h" #include "config.h"
@ -24,7 +27,7 @@
// Hack... We are supposed to use <nmmintrin.h>. GCC 4.8, LLVM Clang 3.5 // Hack... We are supposed to use <nmmintrin.h>. GCC 4.8, LLVM Clang 3.5
// and Apple Clang 6.0 conflates SSE4.1 and SSE4.2. If we use <nmmintrin.h> // and Apple Clang 6.0 conflates SSE4.1 and SSE4.2. If we use <nmmintrin.h>
// then compile fails with "SSE4.2 instruction set not enabled". Also see // then compile fails with "SSE4.2 instruction set not enabled". Also see
// https://gcc.gnu.org/ml/gcc-help/2017-08/msg00015.html. // http://gcc.gnu.org/ml/gcc-help/2017-08/msg00015.html.
# include "smmintrin.h" # include "smmintrin.h"
# include "wmmintrin.h" # include "wmmintrin.h"
#endif #endif

9
rijndael.cpp
View File

@ -220,6 +220,15 @@ void Rijndael::Base::FillDecTable()
s_TdFilled = true; s_TdFilled = true;
} }
unsigned int Rijndael::Base::OptimalDataAlignment() const
{
#if CRYPTOPP_BOOL_ALIGN16
return 16;
#else
return GetAlignmentOf<word32>();
#endif
}
#if (CRYPTOPP_AESNI_AVAILABLE) #if (CRYPTOPP_AESNI_AVAILABLE)
extern void Rijndael_UncheckedSetKey_SSE4_AESNI(const byte *userKey, size_t keyLen, word32* rk); extern void Rijndael_UncheckedSetKey_SSE4_AESNI(const byte *userKey, size_t keyLen, word32* rk);
extern void Rijndael_UncheckedSetKeyRev_AESNI(word32 *key, unsigned int rounds); extern void Rijndael_UncheckedSetKeyRev_AESNI(word32 *key, unsigned int rounds);

3
rijndael.h
View File

@ -39,6 +39,9 @@ class CRYPTOPP_DLL Rijndael : public Rijndael_Info, public BlockCipherDocumentat
class CRYPTOPP_DLL CRYPTOPP_NO_VTABLE Base : public BlockCipherImpl<Rijndael_Info> class CRYPTOPP_DLL CRYPTOPP_NO_VTABLE Base : public BlockCipherImpl<Rijndael_Info>
{ {
public: public:
// Intel and ARM SIMD units can handle unaligned loads, but AltiVec and Power8 cannot.
unsigned int OptimalDataAlignment() const;
void UncheckedSetKey(const byte *userKey, unsigned int keyLength, const NameValuePairs &params); void UncheckedSetKey(const byte *userKey, unsigned int keyLength, const NameValuePairs &params);
protected: protected: