fixes for gating and ARM32 alignment defects:

wolfcrypt/src/port/arm/armv8-aes.c: in the WOLFSSL_ARMASM_NO_HW_CRYPTO version of wc_AesSetKey(), copy the supplied userKey to a properly aligned buffer if necessary before calling AES_set_encrypt_key();

src/dtls13.c: in Dtls13GetRnMask(), if defined(WOLFSSL_LINUXKM)), return retval of wc_AesEncryptDirect();

wolfcrypt/src/misc.c: add readUnalignedWord32(), writeUnalignedWord32(), readUnalignedWords32(), and writeUnalignedWords32();

wolfcrypt/src/siphash.c: use readUnalignedWord64(), readUnalignedWord32(), and writeUnalignedWord64(), to avoid unaligned access faults, and fix cast in byte-reversing version of GET_U32().

src/dtls13.c

@@ -260,7 +260,8 @@ static int Dtls13GetRnMask(WOLFSSL* ssl, const byte* ciphertext, byte* mask,
         if (c->aes == NULL)
             return BAD_STATE_E;
 #if !defined(HAVE_SELFTEST) && \
-    (!defined(HAVE_FIPS) || (defined(FIPS_VERSION_GE) && FIPS_VERSION_GE(5,3)))
+    (!defined(HAVE_FIPS) || (defined(FIPS_VERSION_GE) && FIPS_VERSION_GE(5,3)) \
+    || defined(WOLFSSL_LINUXKM))
         return wc_AesEncryptDirect(c->aes, mask, ciphertext);
 #else
         wc_AesEncryptDirect(c->aes, mask, ciphertext);

src/ssl_crypto.c

@@ -3002,7 +3002,8 @@ void wolfSSL_AES_encrypt(const unsigned char* input, unsigned char* output,
     }
     else
 #if !defined(HAVE_SELFTEST) && \
-    (!defined(HAVE_FIPS) || (defined(FIPS_VERSION_GE) && FIPS_VERSION_GE(5,3)))
+    (!defined(HAVE_FIPS) || (defined(FIPS_VERSION_GE) && FIPS_VERSION_GE(5,3)) \
+    || defined(WOLFSSL_LINUXKM))
     /* Encrypt a block with wolfCrypt AES. */
     if (wc_AesEncryptDirect((Aes*)key, output, input) != 0) {
         WOLFSSL_MSG("wc_AesEncryptDirect failed");

wolfcrypt/src/misc.c

@@ -209,6 +209,53 @@ WC_MISC_STATIC WC_INLINE void ByteReverseWords(word32* out, const word32* in,
 #endif
 }
 
+WC_MISC_STATIC WC_INLINE word32 readUnalignedWord32(const byte *in)
+{
+    if (((wc_ptr_t)in & (wc_ptr_t)(sizeof(word32) - 1U)) == (wc_ptr_t)0)
+        return *(word32 *)in;
+    else {
+        word32 out;
+        XMEMCPY(&out, in, sizeof(word32));
+        return out;
+    }
+}
+
+WC_MISC_STATIC WC_INLINE word32 writeUnalignedWord32(void *out, word32 in)
+{
+    if (((wc_ptr_t)out & (wc_ptr_t)(sizeof(word32) - 1U)) == (wc_ptr_t)0)
+        *(word32 *)out = in;
+    else {
+        XMEMCPY(out, &in, sizeof(word32));
+    }
+    return in;
+}
+
+WC_MISC_STATIC WC_INLINE void readUnalignedWords32(word32 *out, const byte *in,
+                                                   size_t count)
+{
+    if (((wc_ptr_t)in & (wc_ptr_t)(sizeof(word32) - 1U)) == (wc_ptr_t)0) {
+        const word32 *in_word32 = (const word32 *)in;
+        while (count-- > 0)
+            *out++ = *in_word32++;
+    }
+    else {
+        XMEMCPY(out, in, count * sizeof(word32));
+    }
+}
+
+WC_MISC_STATIC WC_INLINE void writeUnalignedWords32(byte *out, const word32 *in,
+                                                    size_t count)
+{
+    if (((wc_ptr_t)out & (wc_ptr_t)(sizeof(word32) - 1U)) == (wc_ptr_t)0) {
+        word32 *out_word32 = (word32 *)out;
+        while (count-- > 0)
+            *out_word32++ = *in++;
+    }
+    else {
+        XMEMCPY(out, in, count * sizeof(word32));
+    }
+}
+
 #if defined(WORD64_AVAILABLE) && !defined(WOLFSSL_NO_WORD64_OPS)
 
 WC_MISC_STATIC WC_INLINE word64 readUnalignedWord64(const byte *in)

wolfcrypt/src/port/arm/armv8-aes.c

@@ -16561,6 +16561,7 @@ int wc_AesSetKey(Aes* aes, const byte* userKey, word32 keylen,
 {
 #if defined(AES_MAX_KEY_SIZE)
     const word32 max_key_len = (AES_MAX_KEY_SIZE / 8);
+    word32 userKey_aligned[AES_MAX_KEY_SIZE / WOLFSSL_BIT_SIZE / sizeof(word32)];
 #endif
 
     if (((keylen != 16) && (keylen != 24) && (keylen != 32)) ||
@@ -16574,6 +16575,14 @@ int wc_AesSetKey(Aes* aes, const byte* userKey, word32 keylen,
         return BAD_FUNC_ARG;
     }
 #endif
+
+#if !defined(AES_MAX_KEY_SIZE)
+    /* Check alignment */
+    if ((unsigned long)userKey & (sizeof(aes->key[0]) - 1U)) {
+        return BAD_FUNC_ARG;
+    }
+#endif
+
 #ifdef WOLF_CRYPTO_CB
     if (aes->devId != INVALID_DEVID) {
         if (keylen > sizeof(aes->devKey)) {
@@ -16590,7 +16599,17 @@ int wc_AesSetKey(Aes* aes, const byte* userKey, word32 keylen,
     aes->keylen = keylen;
     aes->rounds = keylen/4 + 6;
 
-    AES_set_encrypt_key(userKey, keylen * 8, (byte*)aes->key);
+#if defined(AES_MAX_KEY_SIZE)
+    if ((unsigned long)userKey & (sizeof(aes->key[0]) - 1U)) {
+        XMEMCPY(userKey_aligned, userKey, keylen);
+        AES_set_encrypt_key((byte *)userKey_aligned, keylen * 8, (byte*)aes->key);
+    }
+    else
+#endif
+    {
+        AES_set_encrypt_key(userKey, keylen * 8, (byte*)aes->key);
+    }
+
 #ifdef HAVE_AES_DECRYPT
     if (dir == AES_DECRYPTION) {
         AES_invert_key((byte*)aes->key, aes->rounds);

wolfcrypt/src/siphash.c

@@ -69,14 +69,14 @@
  * @param [in] a  Little-endian byte array.
  * @return 64-bit number.
  */
-#define GET_U64(a)      (*(word64*)(a))
+#define GET_U64(a)      readUnalignedWord64(a)
 /**
  * Decode little-endian byte array to 32-bit number.
  *
  * @param [in] a  Little-endian byte array.
  * @return 32-bit number.
  */
-#define GET_U32(a)      (*(word32*)(a))
+#define GET_U32(a)      readUnalignedWord32(a)
 /**
  * Decode little-endian byte array to 16-bit number.
  *
@@ -90,7 +90,7 @@
  * @param [out] a  Byte array to write into.
  * @param [in]  n  Number to encode.
  */
-#define SET_U64(a, n)   ((*(word64*)(a)) = (n))
+#define SET_U64(a, n)   writeUnalignedWord64(a, n)
 #else
 /**
  * Decode little-endian byte array to 64-bit number.
@@ -112,7 +112,7 @@
  * @param [in] a  Little-endian byte array.
  * @return 32-bit number.
  */
-#define GET_U32(a)      (((word64)((a)[3]) << 24) |     \
+#define GET_U32(a)      (((word32)((a)[3]) << 24) |     \
                          ((word32)((a)[2]) << 16) |     \
                          ((word32)((a)[1]) <<  8) |     \
                          ((word32)((a)[0])      ))