Add ARM32 asm code for X25519 and Ed25519

Improvements to performance of ARM64 code for X25519 and Ed25519

src/include.am

@@ -375,6 +375,7 @@ if BUILD_INTELASM
 src_libwolfssl_la_SOURCES += wolfcrypt/src/fe_x25519_asm.S
 else
 if BUILD_ARMASM
+src_libwolfssl_la_SOURCES += wolfcrypt/src/port/arm/armv7-curve25519.S
 src_libwolfssl_la_SOURCES += wolfcrypt/src/port/arm/armv8-curve25519.S
 else
 src_libwolfssl_la_SOURCES += wolfcrypt/src/fe_operations.c
@@ -393,6 +394,7 @@ if BUILD_INTELASM
 src_libwolfssl_la_SOURCES += wolfcrypt/src/fe_x25519_asm.S
 else
 if BUILD_ARMASM
+src_libwolfssl_la_SOURCES += wolfcrypt/src/port/arm/armv7-curve25519.S
 src_libwolfssl_la_SOURCES += wolfcrypt/src/port/arm/armv8-curve25519.S
 else
 src_libwolfssl_la_SOURCES += wolfcrypt/src/fe_operations.c

wolfcrypt/src/fe_operations.c

@@ -42,7 +42,9 @@
 #endif
 
 #ifdef CURVED25519_X64
-/* Assumbly code in fe_x25519_asm.* */
+/* Assembly code in fe_x25519_asm.* */
+#elif defined(WOLFSSL_ARMASM)
+/* Assembly code in fe_armv[78]_x25519.* */
 #elif defined(CURVED25519_128BIT)
 #include "fe_x25519_128.i"
 #else

wolfcrypt/src/include.am

@@ -52,6 +52,8 @@ EXTRA_DIST += wolfcrypt/src/port/ti/ti-aes.c \
               wolfcrypt/src/port/arm/armv8-sha256.c \
               wolfcrypt/src/port/arm/armv8-curve25519.c \
               wolfcrypt/src/port/arm/armv8-curve25519.S \
+              wolfcrypt/src/port/arm/armv7-curve25519.c \
+              wolfcrypt/src/port/arm/armv7-curve25519.S \
               wolfcrypt/src/port/nxp/ksdk_port.c \
               wolfcrypt/src/port/atmel/README.md \
               wolfcrypt/src/port/xilinx/xil-sha3.c \

wolfssl/wolfcrypt/fe_operations.h

@@ -39,8 +39,17 @@
     #define CURVED25519_128BIT
 #endif
 
-#if defined(CURVED25519_X64) || defined(WOLFSSL_ARMASM)
+#if defined(CURVED25519_X64)
     #define CURVED25519_ASM_64BIT
+    #define CURVED25519_ASM
+#endif
+#if defined(WOLFSSL_ARMASM)
+    #ifdef __aarch64__
+        #define CURVED25519_ASM_64BIT
+    #else
+        #define CURVED25519_ASM_32BIT
+    #endif
+    #define CURVED25519_ASM
 #endif
 
 /*
@@ -78,6 +87,8 @@ WOLFSSL_LOCAL int  curve25519(byte * q, byte * n, byte * p);
 
 #ifdef CURVED25519_ASM_64BIT
     typedef int64_t  fe[4];
+#elif defined(CURVED25519_ASM_32BIT)
+    typedef int32_t  fe[8];
 #elif defined(CURVED25519_128BIT)
     typedef int64_t  fe[5];
 #else
@@ -112,7 +123,7 @@ WOLFSSL_LOCAL void fe_pow22523(fe,const fe);
 WOLFSSL_LOCAL uint64_t load_3(const unsigned char *in);
 WOLFSSL_LOCAL uint64_t load_4(const unsigned char *in);
 
-#ifdef CURVED25519_ASM_64BIT
+#ifdef CURVED25519_ASM
 WOLFSSL_LOCAL void fe_ge_to_p2(fe rx, fe ry, fe rz, const fe px, const fe py,
                                const fe pz, const fe pt);
 WOLFSSL_LOCAL void fe_ge_to_p3(fe rx, fe ry, fe rz, fe rt, const fe px,
@@ -136,7 +147,7 @@ WOLFSSL_LOCAL void fe_ge_sub(fe rx, fe ry, fe rz, fe rt, const fe px,
                              const fe qt2d, const fe qyplusx,
                              const fe qyminusx);
 WOLFSSL_LOCAL void fe_cmov_table(fe* r, fe* base, signed char b);
-#endif /* CURVED25519_ASM_64BIT */
+#endif /* CURVED25519_ASM */
 #endif /* !CURVE25519_SMALL || !ED25519_SMALL */
 
 /* Use less memory and only 32bit types or less, but is slower

wolfssl/wolfcrypt/ge_operations.h

@@ -49,6 +49,8 @@ Representations:
   typedef byte     ge[F25519_SIZE];
 #elif defined(CURVED25519_ASM_64BIT)
   typedef int64_t  ge[4];
+#elif defined(CURVED25519_ASM_32BIT)
+  typedef int32_t  ge[8];
 #elif defined(CURVED25519_128BIT)
   typedef int64_t  ge[5];
 #else