Remove CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS support (GH #682)

We were able to gut CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS for everything except Rijndael. Rijndael uses unaligned accesses on x86 to harden against timing attacks.
There's a little more to CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS and Rijndael. If we remove unaligned access then AliasedWithTable hangs in an endless loop on non-AESNI machines. So care must be taken when trying to remove the vestige from Rijndael.

1 files changed, 16 insertions, 15 deletions

diff --git a/rijndael.cpp b/rijndael.cpp
index 100d2a6c..1553e02e 100644
--- a/rijndael.cpp
+++ b/rijndael.cpp
@@ -30,8 +30,9 @@ x86 assembly code, doing an 8-bit register move to minimize the number of
 register spills. Also switched to compressed tables and copying round keys to

 the stack.

 

-The C++ implementation now uses compressed tables if

-CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS is defined.

+The C++ implementation uses compressed tables if

+CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS is defined.

+It is defined on x86 platforms by default but no others.

 */

 

 /*

@@ -95,7 +96,7 @@ NAMESPACE_BEGIN(CryptoPP)
 #endif

 

 // Hack for http://github.com/weidai11/cryptopp/issues/42 and http://github.com/weidai11/cryptopp/issues/132

-#if (CRYPTOPP_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)) && !defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS)

+#if (CRYPTOPP_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE))

 # define CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS 1

 #endif

 

@@ -103,7 +104,7 @@ NAMESPACE_BEGIN(CryptoPP)
 #define M128I_CAST(x) ((__m128i *)(void *)(x))

 #define CONST_M128I_CAST(x) ((const __m128i *)(const void *)(x))

 

-#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 # if (CRYPTOPP_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)) && !defined(CRYPTOPP_DISABLE_RIJNDAEL_ASM)

 namespace rdtable {CRYPTOPP_ALIGN_DATA(16) word64 Te[256+2];}

 using namespace rdtable;

@@ -111,14 +112,14 @@ using namespace rdtable;
 static word64 Te[256];

 # endif

 static word64 Td[256];

-#else // Not CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS

+#else // Not CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS

 # if defined(CRYPTOPP_X64_MASM_AVAILABLE)

 // Unused; avoids linker error on Microsoft X64 non-AESNI platforms

 namespace rdtable {CRYPTOPP_ALIGN_DATA(16) word64 Te[256+2];}

 # endif

 CRYPTOPP_ALIGN_DATA(16) static word32 Te[256*4];

 CRYPTOPP_ALIGN_DATA(16) static word32 Td[256*4];

-#endif // CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS

+#endif // CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS

 

 static volatile bool s_TeFilled = false, s_TdFilled = false;

 

@@ -199,7 +200,7 @@ ANONYMOUS_NAMESPACE_END
 	tempBlock[c] = ((byte *)(Te+byte(t)))[1]; t >>= 8;\

 	tempBlock[d] = ((byte *)(Te+t))[1];

 

-#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 	#define QUARTER_ROUND_LD(t, a, b, c, d)	\

 		tempBlock[a] = ((byte *)(Td+byte(t)))[GetNativeByteOrder()*7]; t >>= 8;\

 		tempBlock[b] = ((byte *)(Td+byte(t)))[GetNativeByteOrder()*7]; t >>= 8;\

@@ -219,7 +220,7 @@ ANONYMOUS_NAMESPACE_END
 #ifdef CRYPTOPP_LITTLE_ENDIAN

 	#define QUARTER_ROUND_FE(t, a, b, c, d)		QUARTER_ROUND(TL_F, Te, t, d, c, b, a)

 	#define QUARTER_ROUND_FD(t, a, b, c, d)		QUARTER_ROUND(TL_F, Td, t, d, c, b, a)

-	#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+	#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 		#define TL_F(T, i, x)	(*(word32 *)(void *)((byte *)T + x*8 + (6-i)%4+1))

 		#define TL_M(T, i, x)	(*(word32 *)(void *)((byte *)T + x*8 + (i+3)%4+1))

 	#else

@@ -229,7 +230,7 @@ ANONYMOUS_NAMESPACE_END
 #else

 	#define QUARTER_ROUND_FE(t, a, b, c, d)		QUARTER_ROUND(TL_F, Te, t, a, b, c, d)

 	#define QUARTER_ROUND_FD(t, a, b, c, d)		QUARTER_ROUND(TL_F, Td, t, a, b, c, d)

-	#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+	#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 		#define TL_F(T, i, x)	(*(word32 *)(void *)((byte *)T + x*8 + (4-i)%4))

 		#define TL_M			TL_F

 	#else

@@ -272,7 +273,7 @@ void Rijndael::Base::FillEncTable()
 	for (int i=0; i<256; i++)

 	{

 		byte x = Se[i];

-#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 		word32 y = word32(x)<<8 | word32(x)<<16 | word32(f2(x))<<24;

 		Te[i] = word64(y | f3(x))<<32 | y;

 #else

@@ -295,7 +296,7 @@ void Rijndael::Base::FillDecTable()
 	for (int i=0; i<256; i++)

 	{

 		byte x = Sd[i];

-#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 		word32 y = word32(fd(x))<<8 | word32(f9(x))<<16 | word32(fe(x))<<24;

 		Td[i] = word64(y | fb(x))<<32 | y | x;

 #else

@@ -591,7 +592,7 @@ void Rijndael::Enc::ProcessAndXorBlock(const byte *inBlock, const byte *xorBlock
 	unsigned int i;

 	volatile word32 _u = 0;

 	word32 u = _u;

-#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 	for (i=0; i<2048; i+=cacheLineSize)

 #else

 	for (i=0; i<1024; i+=cacheLineSize)

@@ -693,7 +694,7 @@ void Rijndael::Dec::ProcessAndXorBlock(const byte *inBlock, const byte *xorBlock
 	unsigned int i;

 	volatile word32 _u = 0;

 	word32 u = _u;

-#if defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

+#if defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS)

 	for (i=0; i<2048; i+=cacheLineSize)

 #else

 	for (i=0; i<1024; i+=cacheLineSize)

@@ -728,9 +729,9 @@ void Rijndael::Dec::ProcessAndXorBlock(const byte *inBlock, const byte *xorBlock
 		rk += 8;

 	} while (--r);

 

-#if !(defined(CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS) || defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS))

+#if !(defined(CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS))

 	// timing attack countermeasure. see comments at top for more details

-	// If CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS is defined,

+	// If CRYPTOPP_ALLOW_RIJNDAEL_UNALIGNED_DATA_ACCESS is defined,

 	// QUARTER_ROUND_LD will use Td, which is already preloaded.

 	u = _u;

 	for (i=0; i<256; i+=cacheLineSize)