Add test for AES, SHA1 and SHA2 from ARM Crypto extension from arm-neon branch

cpu.cpp

@@ -94,8 +94,7 @@ bool CpuId(word32 input, word32 output[4])
 	return true;
 #else
 	// longjmp and clobber warnings. Volatile is required.
-	// http://github.com/weidai11/cryptopp/issues/24
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
-	// http://stackoverflow.com/q/7721854
 	volatile bool result = true;
 
 	SigHandler oldHandler = signal(SIGILL, SigIllHandlerCPUID);
@@ -149,8 +148,7 @@ static bool TrySSE2()
 	return true;
 #else
 	// longjmp and clobber warnings. Volatile is required.
-	// http://github.com/weidai11/cryptopp/issues/24
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
-	// http://stackoverflow.com/q/7721854
 	volatile bool result = true;
 
 	SigHandler oldHandler = signal(SIGILL, SigIllHandlerSSE2);
@@ -262,7 +260,7 @@ void DetectX86Features()
 #elif (CRYPTOPP_BOOL_ARM32 || CRYPTOPP_BOOL_ARM64)
 
 bool g_ArmDetectionDone = false;
-bool g_hasNEON = false, g_hasCRC32 = false, g_hasCrypto = false;
+bool g_hasNEON = false, g_hasCRC32 = false, g_hasAES = false, g_hasSHA1 = false, g_hasSHA2 = false;
 
 word32 g_cacheLineSize = CRYPTOPP_L1_CACHE_LINE_SIZE;
 
@@ -292,10 +290,22 @@ extern "C" {
 		longjmp(s_jmpNoCRC32, 1);
 	}
 
-	static jmp_buf s_jmpNoCrypto;
+	static jmp_buf s_jmpNoAES;
-	static void SigIllHandlerCrypto(int)
+	static void SigIllHandlerAES(int)
 	{
-		longjmp(s_jmpNoCrypto, 1);
+		longjmp(s_jmpNoAES, 1);
+	}
+
+	static jmp_buf s_jmpNoSHA1;
+	static void SigIllHandlerSHA1(int)
+	{
+		longjmp(s_jmpNoSHA1, 1);
+	}
+
+	static jmp_buf s_jmpNoSHA2;
+	static void SigIllHandlerSHA2(int)
+	{
+		longjmp(s_jmpNoSHA2, 1);
 	}
 };
 #endif  // Not CRYPTOPP_MS_STYLE_INLINE_ASSEMBLY
@@ -325,8 +335,7 @@ static bool TryNEON()
 	return true;
 # else
 	// longjmp and clobber warnings. Volatile is required.
-	// http://github.com/weidai11/cryptopp/issues/24
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
-	// http://stackoverflow.com/q/7721854
 	volatile bool result = true;
 
 	SigHandler oldHandler = signal(SIGILL, SigIllHandlerNEON);
@@ -376,8 +385,7 @@ static bool TryCRC32()
 	return true;
 # else
 	// longjmp and clobber warnings. Volatile is required.
-	// http://github.com/weidai11/cryptopp/issues/24
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
-	// http://stackoverflow.com/q/7721854
 	volatile bool result = true;
 
 	SigHandler oldHandler = signal(SIGILL, SigIllHandlerCRC32);
@@ -402,7 +410,7 @@ static bool TryCRC32()
 #endif  // CRYPTOPP_BOOL_ARM_CRC32_INTRINSICS_AVAILABLE
 }
 
-static bool TryCrypto()
+static bool TryAES()
 {
 #if (CRYPTOPP_BOOL_ARM_CRYPTO_INTRINSICS_AVAILABLE)
 # if defined(CRYPTOPP_MS_STYLE_INLINE_ASSEMBLY)
@@ -412,8 +420,7 @@ static bool TryCrypto()
 		static const uint8x16_t data = vdupq_n_u8(0), key = vdupq_n_u8(0); 
 		uint8x16_t r1 = vaeseq_u8(data, key);
 		uint8x16_t r2 = vaesdq_u8(data, key);
-
+		CRYPTOPP_UNUSED(r1), CRYPTOPP_UNUSED(r2);
-		// 
 	}
 	__except (EXCEPTION_EXECUTE_HANDLER)
 	{
@@ -422,21 +429,121 @@ static bool TryCrypto()
 	return true;
 # else
 	// longjmp and clobber warnings. Volatile is required.
-	// http://github.com/weidai11/cryptopp/issues/24
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
-	// http://stackoverflow.com/q/7721854
 	volatile bool result = true;
 
-	SigHandler oldHandler = signal(SIGILL, SigIllHandlerCrypto);
+	SigHandler oldHandler = signal(SIGILL, SigIllHandlerAES);
 	if (oldHandler == SIG_ERR)
 		result = false;
 
-	if (setjmp(s_jmpNoCrypto))
+	if (setjmp(s_jmpNoAES))
 		result = false;
 	else
 	{
 		static const uint8x16_t data = vdupq_n_u8(0), key = vdupq_n_u8(0); 
 		uint8x16_t r1 = vaeseq_u8(data, key);
 		uint8x16_t r2 = vaesdq_u8(data, key);
+		CRYPTOPP_UNUSED(r1), CRYPTOPP_UNUSED(r2);
+	}
+
+	signal(SIGILL, oldHandler);
+	return result;
+# endif
+#else
+	return false;
+#endif  // CRYPTOPP_BOOL_ARM_CRYPTO_INTRINSICS_AVAILABLE
+}
+
+static bool TrySHA1()
+{
+#if (CRYPTOPP_BOOL_ARM_CRYPTO_INTRINSICS_AVAILABLE)
+# if defined(CRYPTOPP_MS_STYLE_INLINE_ASSEMBLY)
+	__try
+	{
+		static const uint32x4_t data = vdupq_n_u32(0); 
+		static const uint32_t hash = 0x0;
+
+	    uint32x4_t r1 = vsha1cq_u32 (data, hash, data);
+		uint32x4_t r2 = vsha1mq_u32 (data, hash, data);
+		uint32x4_t r3 = vsha1pq_u32 (data, hash, data);
+		CRYPTOPP_UNUSED(r1), CRYPTOPP_UNUSED(r2), CRYPTOPP_UNUSED(r3);
+	}
+	__except (EXCEPTION_EXECUTE_HANDLER)
+	{
+		return false;
+	}
+	return true;
+# else
+	// longjmp and clobber warnings. Volatile is required.
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
+	volatile bool result = true;
+
+	SigHandler oldHandler = signal(SIGILL, SigIllHandlerSHA1);
+	if (oldHandler == SIG_ERR)
+		result = false;
+
+	if (setjmp(s_jmpNoSHA1))
+		result = false;
+	else
+	{
+		static const uint32x4_t data = vdupq_n_u32(0); 
+		static const uint32_t hash = 0x0;
+
+	    uint32x4_t r1 = vsha1cq_u32 (data, hash, data);
+		uint32x4_t r2 = vsha1mq_u32 (data, hash, data);
+		uint32x4_t r3 = vsha1pq_u32 (data, hash, data);
+		CRYPTOPP_UNUSED(r1), CRYPTOPP_UNUSED(r2), CRYPTOPP_UNUSED(r3);
+	}
+
+	signal(SIGILL, oldHandler);
+	return result;
+# endif
+#else
+	return false;
+#endif  // CRYPTOPP_BOOL_ARM_CRYPTO_INTRINSICS_AVAILABLE
+}
+
+static bool TrySHA2()
+{
+#if (CRYPTOPP_BOOL_ARM_CRYPTO_INTRINSICS_AVAILABLE)
+# if defined(CRYPTOPP_MS_STYLE_INLINE_ASSEMBLY)
+	__try
+	{
+		static const uint32x4_t data = vdupq_n_u32(0); 
+		static const uint32x4_t hash = vdupq_n_u32(0);
+
+		uint32x4_t r1 = vsha256hq_u32 (hash, hash, data);
+		uint32x4_t r2 = vsha256h2q_u32 (hash, hash, data);
+		uint32x4_t r3 = vsha256su0q_u32 (data, data);
+		uint32x4_t r4 = vsha256su1q_u32 (data, data, data);
+		CRYPTOPP_UNUSED(r1), CRYPTOPP_UNUSED(r2), CRYPTOPP_UNUSED(r3), CRYPTOPP_UNUSED(r4);
+	}
+	__except (EXCEPTION_EXECUTE_HANDLER)
+	{
+		return false;
+	}
+	return true;
+# else
+	// longjmp and clobber warnings. Volatile is required.
+	// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
+	volatile bool result = true;
+
+	SigHandler oldHandler = signal(SIGILL, SigIllHandlerSHA2);
+	if (oldHandler == SIG_ERR)
+		result = false;
+
+	if (setjmp(s_jmpNoSHA2))
+		result = false;
+	else
+	{
+		static const uint32x4_t data = vdupq_n_u32(0); 
+		static const uint32x4_t hash = vdupq_n_u32(0);
+
+		uint32x4_t r1 = vsha256hq_u32 (hash, hash, data);
+		uint32x4_t r2 = vsha256h2q_u32 (hash, hash, data);
+		uint32x4_t r3 = vsha256su0q_u32 (data, data);
+		uint32x4_t r4 = vsha256su1q_u32 (data, data, data);
+		CRYPTOPP_UNUSED(r1), CRYPTOPP_UNUSED(r2), CRYPTOPP_UNUSED(r3), CRYPTOPP_UNUSED(r4);
 	}
 
 	signal(SIGILL, oldHandler);
@@ -457,7 +564,9 @@ void DetectArmFeatures()
 {
 	g_hasNEON = TryNEON();
 	g_hasCRC32 = TryCRC32();
-	g_hasCrypto = TryCrypto();
+	g_hasAES = TryAES();
+	g_hasSHA1 = TrySHA1();
+	g_hasSHA2 = TrySHA2();
 
 	*((volatile bool*)&g_ArmDetectionDone) = true;
 }

cpu.h

@@ -240,7 +240,7 @@ inline int GetCacheLineSize()
 #elif (CRYPTOPP_BOOL_ARM32 || CRYPTOPP_BOOL_ARM64)
 
 extern bool g_ArmDetectionDone;
-extern bool g_hasNEON, g_hasCRC32, g_hasCrypto;
+extern bool g_hasNEON, g_hasCRC32, g_hasAES, g_hasSHA1, g_hasSHA2;
 void CRYPTOPP_API DetectArmFeatures();
 
 //! \brief Determine if an ARM processor has Advanced SIMD available
@@ -271,19 +271,49 @@ inline bool HasCRC32()
 	return g_hasCRC32;
 }
 
-//! \brief Determine if an ARM processor has Crypto available
+//! \brief Determine if an ARM processor has AES available
-//! \returns true if the hardware is capable of Crypto at runtime, false otherwise.
+//! \returns true if the hardware is capable of AES at runtime, false otherwise.
-//! \details Crypto instructions provide access to the processor's AES, SHA-1, SHA-224 and SHA-256 intructions.
+//! \details AES is part of the Crypto extensions from ARM C Language Extensions 2.0 (ACLE 2.0)
-//!   They are provided by ARM C Language Extensions 2.0 (ACLE 2.0) and available under Aarch64
+//!   and available under Aarch64 (ARM-64) and Aarch32 (ARM-32) running on Aarch64 (i.e., an
-//!   (ARM-64) and Aarch32 (ARM-32) running on Aarch64 (i.e., an AArch32 execution environment).
+//!   AArch32 execution environment).
 //! \details Runtime support requires compile time support. When compiling with GCC, you may
 //!   need to compile with <tt>-march=armv8-a+crypto</tt>; while Apple requires
 //!   <tt>-arch arm64</tt>. Also see ARM's <tt>__ARM_FEATURE_CRYPTO</tt> preprocessor macro.
-inline bool HasCrypto()
+inline bool HasAES()
 {
 	if (!g_ArmDetectionDone)
 		DetectArmFeatures();
-	return g_hasCrypto;
+	return g_hasAES;
+}
+
+//! \brief Determine if an ARM processor has SHA1 available
+//! \returns true if the hardware is capable of SHA1 at runtime, false otherwise.
+//! \details SHA1 is part of the Crypto extensions from ARM C Language Extensions 2.0 (ACLE 2.0)
+//!   and available under Aarch64 (ARM-64) and Aarch32 (ARM-32) running on Aarch64 (i.e., an
+//!   AArch32 execution environment).
+//! \details Runtime support requires compile time support. When compiling with GCC, you may
+//!   need to compile with <tt>-march=armv8-a+crypto</tt>; while Apple requires
+//!   <tt>-arch arm64</tt>. Also see ARM's <tt>__ARM_FEATURE_CRYPTO</tt> preprocessor macro.
+inline bool HasSHA1()
+{
+	if (!g_ArmDetectionDone)
+		DetectArmFeatures();
+	return g_hasSHA1;
+}
+
+//! \brief Determine if an ARM processor has SHA2 available
+//! \returns true if the hardware is capable of SHA2 at runtime, false otherwise.
+//! \details SHA2 is part of the Crypto extensions from ARM C Language Extensions 2.0 (ACLE 2.0)
+//!   and available under Aarch64 (ARM-64) and Aarch32 (ARM-32) running on Aarch64 (i.e., an
+//!   AArch32 execution environment).
+//! \details Runtime support requires compile time support. When compiling with GCC, you may
+//!   need to compile with <tt>-march=armv8-a+crypto</tt>; while Apple requires
+//!   <tt>-arch arm64</tt>. Also see ARM's <tt>__ARM_FEATURE_CRYPTO</tt> preprocessor macro.
+inline bool HasSHA2()
+{
+	if (!g_ArmDetectionDone)
+		DetectArmFeatures();
+	return g_hasSHA2;
 }
 
 //! \brief Provides the cache line size at runtime

validat1.cpp

@@ -311,6 +311,16 @@ bool TestSettings()
 
 	cout << "hasMMX == " << hasMMX << ", hasISSE == " << hasISSE << ", hasSSE2 == " << hasSSE2 << ", hasSSSE3 == " << hasSSE4 << ", hasSSE4 == " << hasSSSE3 << ", hasAESNI == " << HasAESNI() << ", hasRDRAND == " << HasRDRAND() << ", hasRDSEED == " << HasRDSEED() << ", hasCLMUL == " << HasCLMUL() << ", isP4 == " << isP4 << ", cacheLineSize == " << cacheLineSize;
 	cout << ", AESNI_INTRINSICS == " << CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE << endl;
+
+#elif (CRYPTOPP_BOOL_ARM32 || CRYPTOPP_BOOL_ARM64)
+	bool hasNEON = HasNEON();
+	bool hasCRC32 = HasCRC32();
+	bool hasAES = HasAES();
+	bool hasSHA1 = HasSHA1();
+	bool hasSHA2 = HasSHA2();
+
+	cout << "passed:  ";
+	cout << "hasNEON == " << hasNEON << ", hasCRC32 == " << hasCRC32 << ", hasAES == " << hasAES << ", hasSHA1 == " << hasSHA1 << ", hasSHA2 == " << hasSHA2 << endl;
 #endif
 
 	if (!pass)