Windows/Linux/macOS: implement AES hardware support on ARM64 (ARMv8)

src/Build/Include/Makefile.inc

@@ -14,7 +14,7 @@ $(NAME): $(NAME).a
 
 clean:
 	@echo Cleaning $(NAME)
-	rm -f $(APPNAME) $(NAME).a $(OBJS) $(OBJSEX) $(OBJSNOOPT) $(OBJSHANI) $(OBJSSSE41) $(OBJSSSSE3) $(OBJS:.o=.d) $(OBJSEX:.oo=.d) $(OBJSNOOPT:.o0=.d) $(OBJSHANI:.oshani=.d) $(OBJSSSE41:.osse41=.d) $(OBJSSSSE3:.ossse3=.d) *.gch
+	rm -f $(APPNAME) $(NAME).a $(OBJS) $(OBJSEX) $(OBJSNOOPT) $(OBJSHANI) $(OBJSSSE41) $(OBJSSSSE3) $(OBJARMV8CRYPTO) $(OBJS:.o=.d) $(OBJSEX:.oo=.d) $(OBJSNOOPT:.o0=.d) $(OBJSHANI:.oshani=.d) $(OBJSSSE41:.osse41=.d) $(OBJSSSSE3:.ossse3=.d) $(OBJARMV8CRYPTO:.oarmv8crypto=.d) *.gch
 
 %.o: %.c
 	@echo Compiling $(<F)
@@ -36,6 +36,10 @@ clean:
 	@echo Compiling $(<F)
 	$(CC) $(CFLAGS) -mssse3 -c $< -o $@
 
+%.oarmv8crypto: %.c
+	@echo Compiling $(<F)
+	$(CC) $(CFLAGS) -march=armv8-a+crypto -c $< -o $@
+
 %.o: %.cpp
 	@echo Compiling $(<F)
 	$(CXX) $(CXXFLAGS) -c $< -o $@
@@ -96,10 +100,10 @@ TR_SED_BIN := tr '\n' ' ' | tr -s ' ' ',' | sed -e 's/^,//g' -e 's/,$$/n/' | tr
 
 
 # Dependencies
--include $(OBJS:.o=.d) $(OBJSEX:.oo=.d) $(OBJSNOOPT:.o0=.d) $(OBJSHANI:.oshani=.d) $(OBJSSSE41:.osse41=.d) $(OBJSSSSE3:.ossse3=.d)
+-include $(OBJS:.o=.d) $(OBJSEX:.oo=.d) $(OBJSNOOPT:.o0=.d) $(OBJSHANI:.oshani=.d) $(OBJSSSE41:.osse41=.d) $(OBJSSSSE3:.ossse3=.d) $(OBJARMV8CRYPTO:.oarmv8crypto=.d)
 
 
-$(NAME).a: $(OBJS) $(OBJSEX) $(OBJSNOOPT) $(OBJSHANI) $(OBJSSSE41) $(OBJSSSSE3)
+$(NAME).a: $(OBJS) $(OBJSEX) $(OBJSNOOPT) $(OBJSHANI) $(OBJSSSE41) $(OBJSSSSE3) $(OBJARMV8CRYPTO)
 	@echo Updating library $@
-	$(AR) $(AFLAGS) -rc $@ $(OBJS) $(OBJSEX) $(OBJSNOOPT) $(OBJSHANI) $(OBJSSSE41) $(OBJSSSSE3)
+	$(AR) $(AFLAGS) -rc $@ $(OBJS) $(OBJSEX) $(OBJSNOOPT) $(OBJSHANI) $(OBJSSSE41) $(OBJSSSSE3) $(OBJARMV8CRYPTO)
 	$(RANLIB) $@

src/Common/Crypto.c

@@ -1164,8 +1164,6 @@ BOOL IsAesHwCpuSupported ()
 	}
 
 	return state && !HwEncryptionDisabled;
-#elif defined (_M_ARM64) || defined(__arm__) || defined (__arm64__) || defined (__aarch64__)
-	return 0;
 #else
 	return (HasAESNI() && !HwEncryptionDisabled)? TRUE : FALSE;
 #endif
@@ -1483,29 +1481,3 @@ void VcUnprotectKeys (PCRYPTO_INFO pCryptoInfo, uint64 encID)
 
 #endif
 
-#if defined(_M_ARM64) || defined(__arm__) || defined (__arm64__) || defined (__aarch64__)
-/* dummy implementation that should never be called */
-void aes_hw_cpu_decrypt(const uint8* ks, uint8* data)
-{
-	ks = ks;
-	data = data;
-}
-
-void aes_hw_cpu_decrypt_32_blocks(const uint8* ks, uint8* data)
-{
-	ks = ks;
-	data = data;
-}
-
-void aes_hw_cpu_encrypt(const uint8* ks, uint8* data)
-{
-	ks = ks;
-	data = data;
-}
-
-void aes_hw_cpu_encrypt_32_blocks(const uint8* ks, uint8* data)
-{
-	ks = ks;
-	data = data;
-}
-#endif

src/Common/Dlgcode.c

@@ -1046,7 +1046,7 @@ BOOL IsOSVersionAtLeast (OSVersionEnum reqMinOS, int reqMinServicePack)
 		>= (major << 16 | minor << 8 | reqMinServicePack));
 }
 
-BOOL IsWin10BuildAtLeast(DWORD minBuild)
+BOOL IsWin10BuildAtLeast(int minBuild)
 {
 	// Must first be recognized as Windows 10 or higher  
 	if (nCurrentOS < WIN_10)
@@ -14883,6 +14883,7 @@ void GetAppRandomSeed (unsigned char* pbRandSeed, size_t cbRandSeed)
 			}
 		}
 
+#ifndef _M_ARM64
 		// use RDSEED or RDRAND from CPU as source of entropy if enabled
 		if (	IsCpuRngEnabled() && 
 			(	(HasRDSEED() && RDSEED_getBytes (digest, sizeof (digest)))
@@ -14891,6 +14892,7 @@ void GetAppRandomSeed (unsigned char* pbRandSeed, size_t cbRandSeed)
 		{
 			WHIRLPOOL_add (digest, sizeof(digest), &tctx);
 		}
+#endif
 		WHIRLPOOL_finalize (&tctx, digest);
 
 		count = VC_MIN (cbRandSeed, sizeof (digest));

src/Common/Dlgcode.h

@@ -503,7 +503,7 @@ void Debug (char *format, ...);
 void DebugMsgBox (char *format, ...);
 BOOL IsOSAtLeast (OSVersionEnum reqMinOS);
 BOOL IsOSVersionAtLeast (OSVersionEnum reqMinOS, int reqMinServicePack);
-BOOL IsWin10BuildAtLeast(DWORD minBuild);
+BOOL IsWin10BuildAtLeast(int minBuild);
 BOOL IsSupportedOS ();
 BOOL Is64BitOs ();
 BOOL IsARM();

src/Common/Random.c

@@ -878,6 +878,7 @@ BOOL SlowPoll (void)
 		}
 	}
 
+#ifndef _M_ARM64
 	// use RDSEED or RDRAND from CPU as source of entropy if present
 	if (	IsCpuRngEnabled() &&
 		(	(HasRDSEED() && RDSEED_getBytes (buffer, sizeof (buffer)))
@@ -886,6 +887,7 @@ BOOL SlowPoll (void)
 	{
 		RandaddBuf (buffer, sizeof (buffer));
 	}
+#endif
 
 	burn(buffer, sizeof (buffer));
 
@@ -1011,6 +1013,7 @@ BOOL FastPoll (void)
 		return FALSE;
 	}
 
+#ifndef _M_ARM64
 	// use RDSEED or RDRAND from CPU as source of entropy if enabled
 	if (	IsCpuRngEnabled() &&
 		(	(HasRDSEED() && RDSEED_getBytes (buffer, sizeof (buffer)))
@@ -1019,6 +1022,7 @@ BOOL FastPoll (void)
 	{
 		RandaddBuf (buffer, sizeof (buffer));
 	}
+#endif
 
 	burn (buffer, sizeof(buffer));
 

src/Common/Tests.c

@@ -1490,7 +1490,9 @@ BOOL AutoTestAlgorithms (void)
 	{
 		/* unexepected exception raised. Disable all CPU extended feature and try again */
 		EnableHwEncryption (hwEncryptionEnabled);
+#ifndef _M_ARM64
 		DisableCPUExtendedFeatures ();
+#endif
 		__try
 		{
 			result = DoAutoTestAlgorithms();

src/Crypto/Aes_hw_armv8.c

@@ -0,0 +1,316 @@
+/*
+* AES using ARMv8
+* Contributed by Jeffrey Walton
+*
+* Further changes
+* (C) 2017,2018 Jack Lloyd
+*
+* Botan is released under the Simplified BSD License (see license.txt)
+*/
+
+/* Modified and adapted for VeraCrypt */
+
+#include "Common/Tcdefs.h"
+#include "Aes_hw_cpu.h"
+#if !defined(_UEFI)
+#include <memory.h>
+#include <stdlib.h>
+#endif
+#include "cpu.h"
+#include "misc.h"
+
+#if CRYPTOPP_ARM_AES_AVAILABLE
+
+#include <arm_neon.h>
+
+// Single block encryption operations
+VC_INLINE void aes_enc_block(uint8x16_t* B, uint8x16_t K)
+{
+    *B = vaesmcq_u8(vaeseq_u8(*B, K));
+}
+
+VC_INLINE void aes_enc_block_last(uint8x16_t* B, uint8x16_t K, uint8x16_t K2)
+{
+    *B = veorq_u8(vaeseq_u8(*B, K), K2);
+}
+
+// 4-block parallel encryption operations
+VC_INLINE void aes_enc_4_blocks(uint8x16_t* B0, uint8x16_t* B1, 
+                                  uint8x16_t* B2, uint8x16_t* B3, uint8x16_t K)
+{
+    *B0 = vaesmcq_u8(vaeseq_u8(*B0, K));
+    *B1 = vaesmcq_u8(vaeseq_u8(*B1, K));
+    *B2 = vaesmcq_u8(vaeseq_u8(*B2, K));
+    *B3 = vaesmcq_u8(vaeseq_u8(*B3, K));
+}
+
+VC_INLINE void aes_enc_4_blocks_last(uint8x16_t* B0, uint8x16_t* B1, 
+                                       uint8x16_t* B2, uint8x16_t* B3,
+                                       uint8x16_t K, uint8x16_t K2)
+{
+    *B0 = veorq_u8(vaeseq_u8(*B0, K), K2);
+    *B1 = veorq_u8(vaeseq_u8(*B1, K), K2);
+    *B2 = veorq_u8(vaeseq_u8(*B2, K), K2);
+    *B3 = veorq_u8(vaeseq_u8(*B3, K), K2);
+}
+
+// Single block decryption operations
+VC_INLINE void aes_dec_block(uint8x16_t* B, uint8x16_t K)
+{
+    *B = vaesimcq_u8(vaesdq_u8(*B, K));
+}
+
+VC_INLINE void aes_dec_block_last(uint8x16_t* B, uint8x16_t K, uint8x16_t K2)
+{
+    *B = veorq_u8(vaesdq_u8(*B, K), K2);
+}
+
+// 4-block parallel decryption operations
+VC_INLINE void aes_dec_4_blocks(uint8x16_t* B0, uint8x16_t* B1,
+                                  uint8x16_t* B2, uint8x16_t* B3, uint8x16_t K)
+{
+    *B0 = vaesimcq_u8(vaesdq_u8(*B0, K));
+    *B1 = vaesimcq_u8(vaesdq_u8(*B1, K));
+    *B2 = vaesimcq_u8(vaesdq_u8(*B2, K));
+    *B3 = vaesimcq_u8(vaesdq_u8(*B3, K));
+}
+
+VC_INLINE void aes_dec_4_blocks_last(uint8x16_t* B0, uint8x16_t* B1,
+                                       uint8x16_t* B2, uint8x16_t* B3,
+                                       uint8x16_t K, uint8x16_t K2)
+{
+    *B0 = veorq_u8(vaesdq_u8(*B0, K), K2);
+    *B1 = veorq_u8(vaesdq_u8(*B1, K), K2);
+    *B2 = veorq_u8(vaesdq_u8(*B2, K), K2);
+    *B3 = veorq_u8(vaesdq_u8(*B3, K), K2);
+}
+
+VC_INLINE void aes256_hw_encrypt_blocks(uint8 buffer[], size_t blocks, const uint8* ks)
+{
+    const uint8x16_t K0 = vld1q_u8(ks + 0 * 16);
+    const uint8x16_t K1 = vld1q_u8(ks + 1 * 16);
+    const uint8x16_t K2 = vld1q_u8(ks + 2 * 16);
+    const uint8x16_t K3 = vld1q_u8(ks + 3 * 16);
+    const uint8x16_t K4 = vld1q_u8(ks + 4 * 16);
+    const uint8x16_t K5 = vld1q_u8(ks + 5 * 16);
+    const uint8x16_t K6 = vld1q_u8(ks + 6 * 16);
+    const uint8x16_t K7 = vld1q_u8(ks + 7 * 16);
+    const uint8x16_t K8 = vld1q_u8(ks + 8 * 16);
+    const uint8x16_t K9 = vld1q_u8(ks + 9 * 16);
+    const uint8x16_t K10 = vld1q_u8(ks + 10 * 16);
+    const uint8x16_t K11 = vld1q_u8(ks + 11 * 16);
+    const uint8x16_t K12 = vld1q_u8(ks + 12 * 16);
+    const uint8x16_t K13 = vld1q_u8(ks + 13 * 16);
+    const uint8x16_t K14 = vld1q_u8(ks + 14 * 16);
+
+    while(blocks >= 4) {
+        uint8x16_t B0 = vld1q_u8(buffer);
+        uint8x16_t B1 = vld1q_u8(buffer + 16);
+        uint8x16_t B2 = vld1q_u8(buffer + 32);
+        uint8x16_t B3 = vld1q_u8(buffer + 48);
+
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K0);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K1);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K2);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K3);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K4);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K5);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K6);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K7);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K8);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K9);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K10);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K11);
+        aes_enc_4_blocks(&B0, &B1, &B2, &B3, K12);
+        aes_enc_4_blocks_last(&B0, &B1, &B2, &B3, K13, K14);
+
+        vst1q_u8(buffer, B0);
+        vst1q_u8(buffer + 16, B1); 
+        vst1q_u8(buffer + 32, B2);
+        vst1q_u8(buffer + 48, B3);
+
+        buffer += 16 * 4;
+        blocks -= 4;
+    }
+
+    for(size_t i = 0; i != blocks; ++i) {
+        uint8x16_t B = vld1q_u8(buffer + 16 * i);
+        aes_enc_block(&B, K0);
+        aes_enc_block(&B, K1);
+        aes_enc_block(&B, K2);
+        aes_enc_block(&B, K3);
+        aes_enc_block(&B, K4);
+        aes_enc_block(&B, K5);
+        aes_enc_block(&B, K6);
+        aes_enc_block(&B, K7);
+        aes_enc_block(&B, K8);
+        aes_enc_block(&B, K9);
+        aes_enc_block(&B, K10);
+        aes_enc_block(&B, K11);
+        aes_enc_block(&B, K12);
+        aes_enc_block_last(&B, K13, K14);
+        vst1q_u8(buffer + 16 * i, B);
+    }
+}
+
+VC_INLINE void aes256_hw_encrypt_block(uint8 buffer[], const uint8* ks)
+{
+    const uint8x16_t K0 = vld1q_u8(ks + 0 * 16);
+    const uint8x16_t K1 = vld1q_u8(ks + 1 * 16);
+    const uint8x16_t K2 = vld1q_u8(ks + 2 * 16);
+    const uint8x16_t K3 = vld1q_u8(ks + 3 * 16);
+    const uint8x16_t K4 = vld1q_u8(ks + 4 * 16);
+    const uint8x16_t K5 = vld1q_u8(ks + 5 * 16);
+    const uint8x16_t K6 = vld1q_u8(ks + 6 * 16);
+    const uint8x16_t K7 = vld1q_u8(ks + 7 * 16);
+    const uint8x16_t K8 = vld1q_u8(ks + 8 * 16);
+    const uint8x16_t K9 = vld1q_u8(ks + 9 * 16);
+    const uint8x16_t K10 = vld1q_u8(ks + 10 * 16);
+    const uint8x16_t K11 = vld1q_u8(ks + 11 * 16);
+    const uint8x16_t K12 = vld1q_u8(ks + 12 * 16);
+    const uint8x16_t K13 = vld1q_u8(ks + 13 * 16);
+    const uint8x16_t K14 = vld1q_u8(ks + 14 * 16);
+
+    uint8x16_t B = vld1q_u8(buffer);
+    aes_enc_block(&B, K0);
+    aes_enc_block(&B, K1);
+    aes_enc_block(&B, K2);
+    aes_enc_block(&B, K3);
+    aes_enc_block(&B, K4);
+    aes_enc_block(&B, K5);
+    aes_enc_block(&B, K6);
+    aes_enc_block(&B, K7);
+    aes_enc_block(&B, K8);
+    aes_enc_block(&B, K9);
+    aes_enc_block(&B, K10);
+    aes_enc_block(&B, K11);
+    aes_enc_block(&B, K12);
+    aes_enc_block_last(&B, K13, K14);
+    vst1q_u8(buffer, B);
+}
+
+VC_INLINE void aes256_hw_decrypt_blocks(uint8 buffer[], size_t blocks, const uint8* ks)
+{
+    const uint8x16_t K0 = vld1q_u8(ks + 0 * 16);
+    const uint8x16_t K1 = vld1q_u8(ks + 1 * 16);
+    const uint8x16_t K2 = vld1q_u8(ks + 2 * 16);
+    const uint8x16_t K3 = vld1q_u8(ks + 3 * 16);
+    const uint8x16_t K4 = vld1q_u8(ks + 4 * 16);
+    const uint8x16_t K5 = vld1q_u8(ks + 5 * 16);
+    const uint8x16_t K6 = vld1q_u8(ks + 6 * 16);
+    const uint8x16_t K7 = vld1q_u8(ks + 7 * 16);
+    const uint8x16_t K8 = vld1q_u8(ks + 8 * 16);
+    const uint8x16_t K9 = vld1q_u8(ks + 9 * 16);
+    const uint8x16_t K10 = vld1q_u8(ks + 10 * 16);
+    const uint8x16_t K11 = vld1q_u8(ks + 11 * 16);
+    const uint8x16_t K12 = vld1q_u8(ks + 12 * 16);
+    const uint8x16_t K13 = vld1q_u8(ks + 13 * 16);
+    const uint8x16_t K14 = vld1q_u8(ks + 14 * 16);
+
+    while(blocks >= 4) {
+        uint8x16_t B0 = vld1q_u8(buffer);
+        uint8x16_t B1 = vld1q_u8(buffer + 16);
+        uint8x16_t B2 = vld1q_u8(buffer + 32);
+        uint8x16_t B3 = vld1q_u8(buffer + 48);
+
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K0);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K1);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K2);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K3);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K4);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K5);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K6);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K7);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K8);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K9);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K10);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K11);
+        aes_dec_4_blocks(&B0, &B1, &B2, &B3, K12);
+        aes_dec_4_blocks_last(&B0, &B1, &B2, &B3, K13, K14);
+
+        vst1q_u8(buffer, B0);
+        vst1q_u8(buffer + 16, B1);
+        vst1q_u8(buffer + 32, B2);
+        vst1q_u8(buffer + 48, B3);
+
+        buffer += 16 * 4;
+        blocks -= 4;
+    }
+
+    for(size_t i = 0; i != blocks; ++i) {
+        uint8x16_t B = vld1q_u8(buffer + 16 * i);
+        aes_dec_block(&B, K0);
+        aes_dec_block(&B, K1);
+        aes_dec_block(&B, K2);
+        aes_dec_block(&B, K3);
+        aes_dec_block(&B, K4);
+        aes_dec_block(&B, K5);
+        aes_dec_block(&B, K6);
+        aes_dec_block(&B, K7);
+        aes_dec_block(&B, K8);
+        aes_dec_block(&B, K9);
+        aes_dec_block(&B, K10);
+        aes_dec_block(&B, K11);
+        aes_dec_block(&B, K12);
+        aes_dec_block_last(&B, K13, K14);
+        vst1q_u8(buffer + 16 * i, B);
+    }
+}
+
+VC_INLINE void aes256_hw_decrypt_block(uint8 buffer[], const uint8* ks)
+{
+    const uint8x16_t K0 = vld1q_u8(ks + 0 * 16);
+    const uint8x16_t K1 = vld1q_u8(ks + 1 * 16);
+    const uint8x16_t K2 = vld1q_u8(ks + 2 * 16);
+    const uint8x16_t K3 = vld1q_u8(ks + 3 * 16);
+    const uint8x16_t K4 = vld1q_u8(ks + 4 * 16);
+    const uint8x16_t K5 = vld1q_u8(ks + 5 * 16);
+    const uint8x16_t K6 = vld1q_u8(ks + 6 * 16);
+    const uint8x16_t K7 = vld1q_u8(ks + 7 * 16);
+    const uint8x16_t K8 = vld1q_u8(ks + 8 * 16);
+    const uint8x16_t K9 = vld1q_u8(ks + 9 * 16);
+    const uint8x16_t K10 = vld1q_u8(ks + 10 * 16);
+    const uint8x16_t K11 = vld1q_u8(ks + 11 * 16);
+    const uint8x16_t K12 = vld1q_u8(ks + 12 * 16);
+    const uint8x16_t K13 = vld1q_u8(ks + 13 * 16);
+    const uint8x16_t K14 = vld1q_u8(ks + 14 * 16);
+
+    uint8x16_t B = vld1q_u8(buffer);
+    aes_dec_block(&B, K0);
+    aes_dec_block(&B, K1);
+    aes_dec_block(&B, K2);
+    aes_dec_block(&B, K3);
+    aes_dec_block(&B, K4);
+    aes_dec_block(&B, K5);
+    aes_dec_block(&B, K6);
+    aes_dec_block(&B, K7);
+    aes_dec_block(&B, K8);
+    aes_dec_block(&B, K9);
+    aes_dec_block(&B, K10);
+    aes_dec_block(&B, K11);
+    aes_dec_block(&B, K12);
+    aes_dec_block_last(&B, K13, K14);
+    vst1q_u8(buffer, B);
+}
+
+void aes_hw_cpu_decrypt (const uint8 *ks, uint8 *data)
+{
+    aes256_hw_decrypt_block(data, ks);
+}
+
+void aes_hw_cpu_decrypt_32_blocks (const uint8 *ks, uint8 *data)
+{
+    aes256_hw_decrypt_blocks(data, 32, ks);
+}
+
+void aes_hw_cpu_encrypt (const uint8 *ks, uint8 *data)
+{
+    aes256_hw_encrypt_block(data, ks);
+}
+
+void aes_hw_cpu_encrypt_32_blocks (const uint8 *ks, uint8 *data)
+{
+    aes256_hw_encrypt_blocks(data, 32, ks);
+}
+
+#endif

src/Crypto/Aes_hw_cpu.h

@@ -22,8 +22,8 @@ extern "C"
 
 #if defined (TC_WINDOWS_BOOT)
 uint8 is_aes_hw_cpu_supported ();
-#endif
 void aes_hw_cpu_enable_sse ();
+#endif
 void aes_hw_cpu_decrypt (const uint8 *ks, uint8 *data);
 void VC_CDECL aes_hw_cpu_decrypt_32_blocks (const uint8 *ks, uint8 *data);
 void aes_hw_cpu_encrypt (const uint8 *ks, uint8 *data);

src/Crypto/Crypto.vcxproj

@@ -226,6 +226,10 @@
     </ClCompile>
     <ClCompile Include="Aeskey.c" />
     <ClCompile Include="Aestab.c" />
+    <ClCompile Include="Aes_hw_armv8.c">
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
+    </ClCompile>
     <ClCompile Include="blake2s.c" />
     <ClCompile Include="blake2s_SSE2.c" />
     <ClCompile Include="blake2s_SSE41.c" />

src/Crypto/Crypto.vcxproj.filters

@@ -90,6 +90,12 @@
     <ClCompile Include="Sha2Intel.c">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="Aescrypt.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="Aes_hw_armv8.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="Aes.h">
@@ -167,15 +173,6 @@
     <ClInclude Include="t1ha_selfcheck.h">
       <Filter>Header Files</Filter>
     </ClInclude>
-    <ClInclude Include="blake2s-load-sse2.h">
-      <Filter>Header Files</Filter>
-    </ClInclude>
-    <ClInclude Include="blake2s-load-sse41.h">
-      <Filter>Header Files</Filter>
-    </ClInclude>
-    <ClInclude Include="blake2s-round.h">
-      <Filter>Header Files</Filter>
-    </ClInclude>
   </ItemGroup>
   <ItemGroup>
     <CustomBuild Include="Aes_hw_cpu.asm">

src/Crypto/config.h

@@ -29,6 +29,11 @@
 	#define CRYPTOPP_CLANG_INTEGRATED_ASSEMBLER 1
 #endif
 
+#if defined(_MSC_VER) && !defined(__clang__)
+# undef CRYPTOPP_LLVM_CLANG_VERSION
+# define CRYPTOPP_MSC_VERSION (_MSC_VER)
+#endif
+
 // Clang due to "Inline assembly operands don't work with .intel_syntax", http://llvm.org/bugs/show_bug.cgi?id=24232
 //   TODO: supply the upper version when LLVM fixes it. We set it to 20.0 for compilation purposes.
 #if (defined(CRYPTOPP_LLVM_CLANG_VERSION) && CRYPTOPP_LLVM_CLANG_VERSION <= 200000) || (defined(CRYPTOPP_APPLE_CLANG_VERSION) && CRYPTOPP_APPLE_CLANG_VERSION <= 200000) || defined(CRYPTOPP_CLANG_INTEGRATED_ASSEMBLER)
@@ -201,6 +206,40 @@
     #define CRYPTOPP_BOOL_X64 0
 #endif
 
+#if defined(__arm64__) || defined(__aarch64__) || defined(_M_ARM64)
+	#define CRYPTOPP_BOOL_ARMV8 1
+	#define CRYPTOPP_BOOL_ARM64 1
+#else
+	#define CRYPTOPP_BOOL_ARMV8 0
+	#define CRYPTOPP_BOOL_ARM64 0
+#endif
+
+// ARMv8 and ASIMD. -march=armv8-a or above must be present
+// Requires GCC 4.8, Clang 3.3 or Visual Studio 2017
+// Do not use APPLE_CLANG_VERSION; use __ARM_FEATURE_XXX instead.
+#if !defined(CRYPTOPP_ARM_ASIMD_AVAILABLE) && !defined(CRYPTOPP_DISABLE_ARM_ASIMD)
+# if defined(__aarch32__) || defined(__aarch64__) || defined(__arm64__) || defined(_M_ARM64)
+#  if defined(__ARM_NEON) || defined(__ARM_ASIMD) || defined(__ARM_FEATURE_NEON) || defined(__ARM_FEATURE_ASIMD) || \
+      (CRYPTOPP_GCC_VERSION >= 40800) || (CRYPTOPP_LLVM_CLANG_VERSION >= 30300) || \
+      (CRYPTOPP_APPLE_CLANG_VERSION >= 40000) || (CRYPTOPP_MSC_VERSION >= 1916)
+#   define CRYPTOPP_ARM_NEON_AVAILABLE 1
+#   define CRYPTOPP_ARM_ASIMD_AVAILABLE 1
+#  endif  // Compilers
+# endif  // Platforms
+#endif
+
+// ARMv8 and AES. -march=armv8-a+crypto or above must be present
+// Requires GCC 4.8, Clang 3.3 or Visual Studio 2017
+#if !defined(CRYPTOPP_ARM_AES_AVAILABLE) && !defined(CRYPTOPP_DISABLE_ARM_AES)
+# if defined(__aarch32__) || defined(__aarch64__) || defined(_M_ARM64)
+#  if defined(__ARM_FEATURE_CRYPTO) || (CRYPTOPP_GCC_VERSION >= 40800) || \
+      (CRYPTOPP_LLVM_CLANG_VERSION >= 30300) || (CRYPTOPP_APPLE_CLANG_VERSION >= 40300) || \
+      (CRYPTOPP_MSC_VERSION >= 1916)
+#   define CRYPTOPP_ARM_AES_AVAILABLE 1
+#  endif  // Compilers
+# endif  // Platforms
+#endif
+
 // Undo the ASM and Intrinsic related defines due to X32.
 #if CRYPTOPP_BOOL_X32
 # undef CRYPTOPP_BOOL_X64

src/Crypto/cpu.c

@@ -469,3 +469,41 @@ void DisableCPUExtendedFeatures ()
 
 #endif
 
+#if CRYPTOPP_BOOL_ARMV8
+
+volatile int g_hasAESARM = 0;
+
+#ifndef HWCAP_AES
+# define HWCAP_AES (1 << 3)
+#endif
+
+inline int CPU_QueryAES()
+{
+#if defined(CRYPTOPP_ARM_AES_AVAILABLE)
+#if defined(__linux__) && defined(__aarch64__)
+	if ((getauxval(AT_HWCAP) & HWCAP_AES) != 0)
+		return 1;
+#elif defined(__APPLE__) && defined(__aarch64__)
+	// Apple Sillcon (M1) and later
+	return 1;
+#elif defined(_WIN32) && defined(_M_ARM64)
+#ifdef TC_WINDOWS_DRIVER
+	if (ExIsProcessorFeaturePresent(PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE) != 0)
+		return 1;
+#else
+	if (IsProcessorFeaturePresent(PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE) != 0)
+		return 1;
+#endif
+#endif
+	return 0;
+#else
+	return 0;
+#endif
+}
+
+void DetectArmFeatures()
+{
+	g_hasAESARM  = CPU_QueryAES();
+}
+
+#endif

src/Crypto/cpu.h

@@ -288,6 +288,24 @@ void DisableCPUExtendedFeatures ();
 }
 #endif
 
+#elif CRYPTOPP_BOOL_ARMV8
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if !defined(CRYPTOPP_DISABLE_AESNI) && !defined(WOLFCRYPT_BACKEND)
+#define TC_AES_HW_CPU
+#endif
+
+extern volatile int g_hasAESARM;
+void DetectArmFeatures();
+
+#define HasAESNI() g_hasAESARM
+
+#if defined(__cplusplus)
+}
+#endif
+
 #else
 
 #define HasSSE2()	0

src/Driver/Driver.vcxproj

@@ -228,6 +228,10 @@ copy $(OutDir)veracrypt.inf "$(SolutionDir)Debug\Setup Files\veracrypt.inf"</Com
     </ClCompile>
     <ClCompile Include="..\Crypto\Aeskey.c" />
     <ClCompile Include="..\Crypto\Aestab.c" />
+    <ClCompile Include="..\Crypto\Aes_hw_armv8.c">
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
+    </ClCompile>
     <ClCompile Include="..\Crypto\blake2s.c" />
     <ClCompile Include="..\Crypto\blake2s_SSE2.c">
       <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>

src/Driver/Driver.vcxproj.filters

@@ -165,6 +165,9 @@
     <ClCompile Include="..\Driver\VolumeFilter.c">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="..\Crypto\Aes_hw_armv8.c">
+      <Filter>Crypto\Source Files</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="..\Common\Tcdefs.h">

src/Driver/Ntdriver.c

@@ -232,7 +232,7 @@ void GetDriverRandomSeed (unsigned char* pbRandSeed, size_t cbRandSeed)
 				jent_entropy_collector_free (ec);
 			}
 		}
-
+#ifndef _M_ARM64
 		// use RDSEED or RDRAND from CPU as source of entropy if enabled
 		if (	IsCpuRngEnabled() && 
 			(	(HasRDSEED() && RDSEED_getBytes (digest, sizeof (digest)))
@@ -241,6 +241,7 @@ void GetDriverRandomSeed (unsigned char* pbRandSeed, size_t cbRandSeed)
 		{
 			WHIRLPOOL_add (digest, sizeof(digest), &tctx);
 		}
+#endif
 		WHIRLPOOL_finalize (&tctx, digest);
 
 		count = VC_MIN (cbRandSeed, sizeof (digest));
@@ -266,7 +267,11 @@ NTSTATUS DriverEntry(PDRIVER_OBJECT DriverObject, PUNICODE_STRING RegistryPath)
 
 	Dump("DriverEntry " TC_APP_NAME " " VERSION_STRING VERSION_STRING_SUFFIX "\n");
 
+#ifndef _M_ARM64
 	DetectX86Features();
+#else
+	DetectArmFeatures();
+#endif
 
 	PsGetVersion(&OsMajorVersion, &OsMinorVersion, NULL, NULL);
 
@@ -293,7 +298,11 @@ NTSTATUS DriverEntry(PDRIVER_OBJECT DriverObject, PUNICODE_STRING RegistryPath)
 			{
 				// in case of system encryption, if self-tests fail, disable all extended CPU
 				// features and try again in order to workaround faulty configurations
+#ifndef _M_ARM64
 				DisableCPUExtendedFeatures();
+#else
+				EnableHwEncryption(FALSE);
+#endif
 				SelfTestsPassed = AutoTestAlgorithms();
 
 				// BUG CHECK if the self-tests still fail

src/ExpandVolume/WinMain.cpp

@@ -1093,7 +1093,11 @@ int WINAPI wWinMain (HINSTANCE hInstance, HINSTANCE hPrevInstance, wchar_t *lpsz
 	/* application title */
 	lpszTitle = L"VeraCrypt Expander";
 
-	DetectX86Features ();
+#ifndef _M_ARM64
+	DetectX86Features();
+#else
+	DetectArmFeatures();
+#endif
 
 	status = DriverAttach ();
 	if (status != 0)

src/Format/Tcformat.c

@@ -10588,7 +10588,11 @@ int WINAPI wWinMain (HINSTANCE hInstance, HINSTANCE hPrevInstance, wchar_t *lpsz
 	VirtualLock (&szFileName, sizeof(szFileName));
 	VirtualLock (&szDiskFile, sizeof(szDiskFile));
 
-	DetectX86Features ();
+#ifndef _M_ARM64
+	DetectX86Features();
+#else
+	DetectArmFeatures();
+#endif
 
 	try
 	{

src/Main/UserInterface.cpp

@@ -555,6 +555,9 @@ namespace VeraCrypt
 
 #ifdef CRYPTOPP_CPUID_AVAILABLE
 		DetectX86Features ();
+#endif
+#if CRYPTOPP_BOOL_ARMV8
+		DetectArmFeatures();
 #endif
 		LangString.Init();
 		Core->Init();
@@ -972,7 +975,7 @@ const FileManager fileManagers[] = {
 		}
 
 		ShowWarning(wxT("Unable to find a file manager to open the mounted volume.\n"
-				"Please install xdg-utils or set a default file manager."));
+					"Please install xdg-utils or set a default file manager."));
 #endif
 	}
 

src/Makefile

@@ -162,6 +162,9 @@ else ifneq (,$(filter x86_64 x86-64 amd64 x64,$(ARCH)))
 else ifneq (,$(filter armv7l,$(ARCH)))
 	PLATFORM_ARCH := armv7
 	CPU_ARCH = armv7
+else ifneq (,$(filter aarch64 arm64 armv8l,$(ARCH)))
+	PLATFORM_ARCH := arm64
+	CPU_ARCH = arm64
 endif
 
 ifeq "$(origin NOASM)" "command line"
@@ -338,6 +341,9 @@ $(error Specified SDK version was not found, ensure your active developer direct
 		VC_FUSE_PACKAGE := fuse-t
 	endif
 
+	export CFLAGS_ARM64 := $(CFLAGS) $(C_CXX_FLAGS) -arch arm64 -march=armv8-a+crypto
+	export CFLAGS_X64 := $(CFLAGS) $(C_CXX_FLAGS) -arch x86_64
+
 	# Set x86 assembly flags (-msse2, -mssse3, -msse4.1)
 	# Apply flags if SIMD_SUPPORTED is 1 or if not in local development build (we are creating universal binary in this case)
 	ifneq "$(LOCAL_DEVELOPMENT_BUILD)" "true"

src/Mount/Mount.c

@@ -10132,7 +10132,11 @@ int WINAPI wWinMain (HINSTANCE hInstance, HINSTANCE hPrevInstance, wchar_t *lpsz
 	VirtualLock (&defaultMountOptions, sizeof (defaultMountOptions));
 	VirtualLock (&szFileName, sizeof(szFileName));	
 
-	DetectX86Features ();
+#ifndef _M_ARM64
+	DetectX86Features();
+#else
+	DetectArmFeatures();
+#endif
 
 	try
 	{

src/Volume/Volume.make

@@ -43,6 +43,7 @@ ifeq "$(PLATFORM)" "MacOSX"
 ifneq "$(COMPILE_ASM)" "false"
 	OBJSEX += ../Crypto/Aes_asm.oo
 	OBJS += ../Crypto/Aes_hw_cpu.o
+	OBJSEX += ../Crypto/Aes_hw_armv8.oo
 	OBJS += ../Crypto/Aescrypt.o
 	OBJSEX += ../Crypto/Twofish_asm.oo
 	OBJSEX += ../Crypto/Camellia_asm.oo
@@ -78,6 +79,9 @@ else ifeq "$(CPU_ARCH)" "x64"
 	OBJS += ../Crypto/sha512_avx1_x64.o
 	OBJS += ../Crypto/sha512_avx2_x64.o
 	OBJS += ../Crypto/sha512_sse4_x64.o
+else ifeq "$(CPU_ARCH)" "arm64"
+	OBJARMV8CRYPTO += ../Crypto/Aes_hw_armv8.oarmv8crypto
+	OBJS += ../Crypto/Aescrypt.o
 else
 	OBJS += ../Crypto/Aescrypt.o
 endif
@@ -140,6 +144,12 @@ VolumeLibrary: Volume.a
 ifeq "$(ENABLE_WOLFCRYPT)" "0"
 ifeq "$(PLATFORM)" "MacOSX"
 ifneq "$(COMPILE_ASM)" "false"
+../Crypto/Aes_hw_armv8.oo: ../Crypto/Aes_hw_armv8.c
+	@echo Compiling $(<F)
+	$(CC) $(CFLAGS_ARM64) -c ../Crypto/Aes_hw_armv8.c -o ../Crypto/Aes_hw_armv8_arm64.o
+	$(CC) $(CFLAGS_X64) -c ../Crypto/Aes_hw_armv8.c -o ../Crypto/Aes_hw_armv8_x64.o
+	lipo -create ../Crypto/Aes_hw_armv8_arm64.o ../Crypto/Aes_hw_armv8_x64.o -output ../Crypto/Aes_hw_armv8.oo
+	rm -fr ../Crypto/Aes_hw_armv8_arm64.o ../Crypto/Aes_hw_armv8_x64.o
 ../Crypto/Aes_asm.oo: ../Crypto/Aes_x86.asm ../Crypto/Aes_x64.asm
 	@echo Assembling $(<F)
 	$(AS) $(ASFLAGS32) -o ../Crypto/Aes_x86.o ../Crypto/Aes_x86.asm