Fix undefined behaviour in shifts.

Td4 and Te4 are arrays of u8. A u8 << int promotes the u8 to an int first then shifts.
If the mathematical result of a shift (as modelled by lhs * 2^{rhs}) is not representable
in an integer, behaviour is undefined. In other words, you can't shift into the sign bit
of a signed integer. Fix this by casting to u32 whenever we're shifting left by 24.

(For consistency, cast other shifts, too.)

Caught by -fsanitize=shift

Submitted by Nick Lewycky (Google)

Reviewed-by: Andy Polyakov <appro@openssl.org>
(cherry picked from commit 8b37e5c14f0eddb10c7f91ef91004622d90ef361)

2 files changed, 144 insertions, 144 deletions

diff --git a/crypto/aes/aes_core.c b/crypto/aes/aes_core.c
index ff0d1643fd..2ddb0860d7 100644
--- a/crypto/aes/aes_core.c
+++ b/crypto/aes/aes_core.c
@@ -1130,31 +1130,31 @@ void AES_decrypt(const unsigned char *in, unsigned char *out,
      * map cipher state to byte array block:
      */
     s0 =
-        (Td4[(t0 >> 24)       ] << 24) ^
-        (Td4[(t3 >> 16) & 0xff] << 16) ^
-        (Td4[(t2 >>  8) & 0xff] <<  8) ^
-        (Td4[(t1      ) & 0xff])       ^
+        ((u32)Td4[(t0 >> 24)       ] << 24) ^
+        ((u32)Td4[(t3 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(t2 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(t1      ) & 0xff])       ^
         rk[0];
     PUTU32(out     , s0);
     s1 =
-        (Td4[(t1 >> 24)       ] << 24) ^
-        (Td4[(t0 >> 16) & 0xff] << 16) ^
-        (Td4[(t3 >>  8) & 0xff] <<  8) ^
-        (Td4[(t2      ) & 0xff])       ^
+        ((u32)Td4[(t1 >> 24)       ] << 24) ^
+        ((u32)Td4[(t0 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(t3 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(t2      ) & 0xff])       ^
         rk[1];
     PUTU32(out +  4, s1);
     s2 =
-        (Td4[(t2 >> 24)       ] << 24) ^
-        (Td4[(t1 >> 16) & 0xff] << 16) ^
-        (Td4[(t0 >>  8) & 0xff] <<  8) ^
-        (Td4[(t3      ) & 0xff])       ^
+        ((u32)Td4[(t2 >> 24)       ] << 24) ^
+        ((u32)Td4[(t1 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(t0 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(t3      ) & 0xff])       ^
         rk[2];
     PUTU32(out +  8, s2);
     s3 =
-        (Td4[(t3 >> 24)       ] << 24) ^
-        (Td4[(t2 >> 16) & 0xff] << 16) ^
-        (Td4[(t1 >>  8) & 0xff] <<  8) ^
-        (Td4[(t0      ) & 0xff])       ^
+        ((u32)Td4[(t3 >> 24)       ] << 24) ^
+        ((u32)Td4[(t2 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(t1 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(t0      ) & 0xff])       ^
         rk[3];
     PUTU32(out + 12, s3);
 }
@@ -1233,10 +1233,10 @@ int private_AES_set_encrypt_key(const unsigned char *userKey, const int bits,
         while (1) {
             temp  = rk[3];
             rk[4] = rk[0] ^
-                (Te4[(temp >> 16) & 0xff] << 24) ^
-                (Te4[(temp >>  8) & 0xff] << 16) ^
-                (Te4[(temp      ) & 0xff] << 8) ^
-                (Te4[(temp >> 24)       ]) ^
+                ((u32)Te4[(temp >> 16) & 0xff] << 24) ^
+                ((u32)Te4[(temp >>  8) & 0xff] << 16) ^
+                ((u32)Te4[(temp      ) & 0xff] << 8) ^
+                ((u32)Te4[(temp >> 24)       ]) ^
                 rcon[i];
             rk[5] = rk[1] ^ rk[4];
             rk[6] = rk[2] ^ rk[5];
@@ -1253,10 +1253,10 @@ int private_AES_set_encrypt_key(const unsigned char *userKey, const int bits,
         while (1) {
             temp = rk[ 5];
             rk[ 6] = rk[ 0] ^
-                (Te4[(temp >> 16) & 0xff] << 24) ^
-                (Te4[(temp >>  8) & 0xff] << 16) ^
-                (Te4[(temp      ) & 0xff] << 8) ^
-                (Te4[(temp >> 24)       ]) ^
+                ((u32)Te4[(temp >> 16) & 0xff] << 24) ^
+                ((u32)Te4[(temp >>  8) & 0xff] << 16) ^
+                ((u32)Te4[(temp      ) & 0xff] << 8) ^
+                ((u32)Te4[(temp >> 24)       ]) ^
                 rcon[i];
             rk[ 7] = rk[ 1] ^ rk[ 6];
             rk[ 8] = rk[ 2] ^ rk[ 7];
@@ -1275,10 +1275,10 @@ int private_AES_set_encrypt_key(const unsigned char *userKey, const int bits,
         while (1) {
             temp = rk[ 7];
             rk[ 8] = rk[ 0] ^
-                (Te4[(temp >> 16) & 0xff] << 24) ^
-                (Te4[(temp >>  8) & 0xff] << 16) ^
-                (Te4[(temp      ) & 0xff] << 8) ^
-                (Te4[(temp >> 24)       ]) ^
+                ((u32)Te4[(temp >> 16) & 0xff] << 24) ^
+                ((u32)Te4[(temp >>  8) & 0xff] << 16) ^
+                ((u32)Te4[(temp      ) & 0xff] << 8) ^
+                ((u32)Te4[(temp >> 24)       ]) ^
                 rcon[i];
             rk[ 9] = rk[ 1] ^ rk[ 8];
             rk[10] = rk[ 2] ^ rk[ 9];
@@ -1288,10 +1288,10 @@ int private_AES_set_encrypt_key(const unsigned char *userKey, const int bits,
             }
             temp = rk[11];
             rk[12] = rk[ 4] ^
-                (Te4[(temp >> 24)       ] << 24) ^
-                (Te4[(temp >> 16) & 0xff] << 16) ^
-                (Te4[(temp >>  8) & 0xff] << 8) ^
-                (Te4[(temp      ) & 0xff]);
+                ((u32)Te4[(temp >> 24)       ] << 24) ^
+                ((u32)Te4[(temp >> 16) & 0xff] << 16) ^
+                ((u32)Te4[(temp >>  8) & 0xff] << 8) ^
+                ((u32)Te4[(temp      ) & 0xff]);
             rk[13] = rk[ 5] ^ rk[12];
             rk[14] = rk[ 6] ^ rk[13];
             rk[15] = rk[ 7] ^ rk[14];
diff --git a/crypto/aes/aes_x86core.c b/crypto/aes/aes_x86core.c
index f651059054..1defbb1abf 100644
--- a/crypto/aes/aes_x86core.c
+++ b/crypto/aes/aes_x86core.c
@@ -497,10 +497,10 @@ int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
         while (1) {
             temp  = rk[3];
             rk[4] = rk[0] ^
-                (Te4[(temp >>  8) & 0xff]      ) ^
-                (Te4[(temp >> 16) & 0xff] <<  8) ^
-                (Te4[(temp >> 24)       ] << 16) ^
-                (Te4[(temp      ) & 0xff] << 24) ^
+                ((u32)Te4[(temp >>  8) & 0xff]      ) ^
+                ((u32)Te4[(temp >> 16) & 0xff] <<  8) ^
+                ((u32)Te4[(temp >> 24)       ] << 16) ^
+                ((u32)Te4[(temp      ) & 0xff] << 24) ^
                 rcon[i];
             rk[5] = rk[1] ^ rk[4];
             rk[6] = rk[2] ^ rk[5];
@@ -517,10 +517,10 @@ int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
         while (1) {
             temp = rk[ 5];
             rk[ 6] = rk[ 0] ^
-                (Te4[(temp >>  8) & 0xff]      ) ^
-                (Te4[(temp >> 16) & 0xff] <<  8) ^
-                (Te4[(temp >> 24)       ] << 16) ^
-                (Te4[(temp      ) & 0xff] << 24) ^
+                ((u32)Te4[(temp >>  8) & 0xff]      ) ^
+                ((u32)Te4[(temp >> 16) & 0xff] <<  8) ^
+                ((u32)Te4[(temp >> 24)       ] << 16) ^
+                ((u32)Te4[(temp      ) & 0xff] << 24) ^
                 rcon[i];
             rk[ 7] = rk[ 1] ^ rk[ 6];
             rk[ 8] = rk[ 2] ^ rk[ 7];
@@ -539,10 +539,10 @@ int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
         while (1) {
             temp = rk[ 7];
             rk[ 8] = rk[ 0] ^
-                (Te4[(temp >>  8) & 0xff]      ) ^
-                (Te4[(temp >> 16) & 0xff] <<  8) ^
-                (Te4[(temp >> 24)       ] << 16) ^
-                (Te4[(temp      ) & 0xff] << 24) ^
+                ((u32)Te4[(temp >>  8) & 0xff]      ) ^
+                ((u32)Te4[(temp >> 16) & 0xff] <<  8) ^
+                ((u32)Te4[(temp >> 24)       ] << 16) ^
+                ((u32)Te4[(temp      ) & 0xff] << 24) ^
                 rcon[i];
             rk[ 9] = rk[ 1] ^ rk[ 8];
             rk[10] = rk[ 2] ^ rk[ 9];
@@ -552,10 +552,10 @@ int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
             }
             temp = rk[11];
             rk[12] = rk[ 4] ^
-                (Te4[(temp      ) & 0xff]      ) ^
-                (Te4[(temp >>  8) & 0xff] <<  8) ^
-                (Te4[(temp >> 16) & 0xff] << 16) ^
-                (Te4[(temp >> 24)       ] << 24);
+                ((u32)Te4[(temp      ) & 0xff]      ) ^
+                ((u32)Te4[(temp >>  8) & 0xff] <<  8) ^
+                ((u32)Te4[(temp >> 16) & 0xff] << 16) ^
+                ((u32)Te4[(temp >> 24)       ] << 24);
             rk[13] = rk[ 5] ^ rk[12];
             rk[14] = rk[ 6] ^ rk[13];
             rk[15] = rk[ 7] ^ rk[14];
@@ -674,22 +674,22 @@ void AES_encrypt(const unsigned char *in, unsigned char *out,
 #if defined(AES_COMPACT_IN_OUTER_ROUNDS)
     prefetch256(Te4);
 
-    t[0] =  Te4[(s0      ) & 0xff]       ^
-        Te4[(s1 >>  8) & 0xff] <<  8 ^
-        Te4[(s2 >> 16) & 0xff] << 16 ^
-        Te4[(s3 >> 24)       ] << 24;
-    t[1] =  Te4[(s1      ) & 0xff]       ^
-        Te4[(s2 >>  8) & 0xff] <<  8 ^
-        Te4[(s3 >> 16) & 0xff] << 16 ^
-        Te4[(s0 >> 24)       ] << 24;
-    t[2] =  Te4[(s2      ) & 0xff]       ^
-        Te4[(s3 >>  8) & 0xff] <<  8 ^
-        Te4[(s0 >> 16) & 0xff] << 16 ^
-        Te4[(s1 >> 24)       ] << 24;
-    t[3] =  Te4[(s3      ) & 0xff]       ^
-        Te4[(s0 >>  8) & 0xff] <<  8 ^
-        Te4[(s1 >> 16) & 0xff] << 16 ^
-        Te4[(s2 >> 24)       ] << 24;
+    t[0] = (u32)Te4[(s0      ) & 0xff]       ^
+           (u32)Te4[(s1 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s2 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s3 >> 24)       ] << 24;
+    t[1] = (u32)Te4[(s1      ) & 0xff]       ^
+           (u32)Te4[(s2 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s3 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s0 >> 24)       ] << 24;
+    t[2] = (u32)Te4[(s2      ) & 0xff]       ^
+           (u32)Te4[(s3 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s0 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s1 >> 24)       ] << 24;
+    t[3] = (u32)Te4[(s3      ) & 0xff]       ^
+           (u32)Te4[(s0 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s1 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s2 >> 24)       ] << 24;
 
     /* now do the linear transform using words */
     {   int i;
@@ -740,22 +740,22 @@ void AES_encrypt(const unsigned char *in, unsigned char *out,
      */
     for (rk+=8,r=key->rounds-2; r>0; rk+=4,r--) {
 #if defined(AES_COMPACT_IN_INNER_ROUNDS)
-        t[0] =  Te4[(s0      ) & 0xff]       ^
-            Te4[(s1 >>  8) & 0xff] <<  8 ^
-            Te4[(s2 >> 16) & 0xff] << 16 ^
-            Te4[(s3 >> 24)       ] << 24;
-        t[1] =  Te4[(s1      ) & 0xff]       ^
-            Te4[(s2 >>  8) & 0xff] <<  8 ^
-            Te4[(s3 >> 16) & 0xff] << 16 ^
-            Te4[(s0 >> 24)       ] << 24;
-        t[2] =  Te4[(s2      ) & 0xff]       ^
-            Te4[(s3 >>  8) & 0xff] <<  8 ^
-            Te4[(s0 >> 16) & 0xff] << 16 ^
-            Te4[(s1 >> 24)       ] << 24;
-        t[3] =  Te4[(s3      ) & 0xff]       ^
-            Te4[(s0 >>  8) & 0xff] <<  8 ^
-            Te4[(s1 >> 16) & 0xff] << 16 ^
-            Te4[(s2 >> 24)       ] << 24;
+        t[0] = (u32)Te4[(s0      ) & 0xff]       ^
+               (u32)Te4[(s1 >>  8) & 0xff] <<  8 ^
+               (u32)Te4[(s2 >> 16) & 0xff] << 16 ^
+               (u32)Te4[(s3 >> 24)       ] << 24;
+        t[1] = (u32)Te4[(s1      ) & 0xff]       ^
+               (u32)Te4[(s2 >>  8) & 0xff] <<  8 ^
+               (u32)Te4[(s3 >> 16) & 0xff] << 16 ^
+               (u32)Te4[(s0 >> 24)       ] << 24;
+        t[2] = (u32)Te4[(s2      ) & 0xff]       ^
+               (u32)Te4[(s3 >>  8) & 0xff] <<  8 ^
+               (u32)Te4[(s0 >> 16) & 0xff] << 16 ^
+               (u32)Te4[(s1 >> 24)       ] << 24;
+        t[3] = (u32)Te4[(s3      ) & 0xff]       ^
+               (u32)Te4[(s0 >>  8) & 0xff] <<  8 ^
+               (u32)Te4[(s1 >> 16) & 0xff] << 16 ^
+               (u32)Te4[(s2 >> 24)       ] << 24;
 
         /* now do the linear transform using words */
         {
@@ -810,28 +810,28 @@ void AES_encrypt(const unsigned char *in, unsigned char *out,
     prefetch256(Te4);
 
     *(u32*)(out+0) =
-        Te4[(s0      ) & 0xff]       ^
-        Te4[(s1 >>  8) & 0xff] <<  8 ^
-        Te4[(s2 >> 16) & 0xff] << 16 ^
-        Te4[(s3 >> 24)       ] << 24 ^
+           (u32)Te4[(s0      ) & 0xff]       ^
+           (u32)Te4[(s1 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s2 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s3 >> 24)       ] << 24 ^
         rk[0];
     *(u32*)(out+4) =
-        Te4[(s1      ) & 0xff]       ^
-        Te4[(s2 >>  8) & 0xff] <<  8 ^
-        Te4[(s3 >> 16) & 0xff] << 16 ^
-        Te4[(s0 >> 24)       ] << 24 ^
+           (u32)Te4[(s1      ) & 0xff]       ^
+           (u32)Te4[(s2 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s3 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s0 >> 24)       ] << 24 ^
         rk[1];
     *(u32*)(out+8) =
-        Te4[(s2      ) & 0xff]       ^
-        Te4[(s3 >>  8) & 0xff] <<  8 ^
-        Te4[(s0 >> 16) & 0xff] << 16 ^
-        Te4[(s1 >> 24)       ] << 24 ^
+           (u32)Te4[(s2      ) & 0xff]       ^
+           (u32)Te4[(s3 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s0 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s1 >> 24)       ] << 24 ^
         rk[2];
     *(u32*)(out+12) =
-        Te4[(s3      ) & 0xff]       ^
-        Te4[(s0 >>  8) & 0xff] <<  8 ^
-        Te4[(s1 >> 16) & 0xff] << 16 ^
-        Te4[(s2 >> 24)       ] << 24 ^
+           (u32)Te4[(s3      ) & 0xff]       ^
+           (u32)Te4[(s0 >>  8) & 0xff] <<  8 ^
+           (u32)Te4[(s1 >> 16) & 0xff] << 16 ^
+           (u32)Te4[(s2 >> 24)       ] << 24 ^
         rk[3];
 #else
     *(u32*)(out+0) =
@@ -888,22 +888,22 @@ void AES_decrypt(const unsigned char *in, unsigned char *out,
 #if defined(AES_COMPACT_IN_OUTER_ROUNDS)
     prefetch256(Td4);
 
-    t[0] =  Td4[(s0      ) & 0xff]       ^
-    Td4[(s3 >>  8) & 0xff] <<  8 ^
-    Td4[(s2 >> 16) & 0xff] << 16 ^
-    Td4[(s1 >> 24)       ] << 24;
-    t[1] =  Td4[(s1      ) & 0xff]       ^
-    Td4[(s0 >>  8) & 0xff] <<  8 ^
-    Td4[(s3 >> 16) & 0xff] << 16 ^
-    Td4[(s2 >> 24)       ] << 24;
-    t[2] =  Td4[(s2      ) & 0xff]       ^
-    Td4[(s1 >>  8) & 0xff] <<  8 ^
-    Td4[(s0 >> 16) & 0xff] << 16 ^
-    Td4[(s3 >> 24)       ] << 24;
-    t[3] =  Td4[(s3      ) & 0xff]       ^
-    Td4[(s2 >>  8) & 0xff] <<  8 ^
-    Td4[(s1 >> 16) & 0xff] << 16 ^
-    Td4[(s0 >> 24)       ] << 24;
+    t[0] = (u32)Td4[(s0      ) & 0xff]       ^
+           (u32)Td4[(s3 >>  8) & 0xff] <<  8 ^
+           (u32)Td4[(s2 >> 16) & 0xff] << 16 ^
+           (u32)Td4[(s1 >> 24)       ] << 24;
+    t[1] = (u32)Td4[(s1      ) & 0xff]       ^
+           (u32)Td4[(s0 >>  8) & 0xff] <<  8 ^
+           (u32)Td4[(s3 >> 16) & 0xff] << 16 ^
+           (u32)Td4[(s2 >> 24)       ] << 24;
+    t[2] = (u32)Td4[(s2      ) & 0xff]       ^
+           (u32)Td4[(s1 >>  8) & 0xff] <<  8 ^
+           (u32)Td4[(s0 >> 16) & 0xff] << 16 ^
+           (u32)Td4[(s3 >> 24)       ] << 24;
+    t[3] = (u32)Td4[(s3      ) & 0xff]       ^
+           (u32)Td4[(s2 >>  8) & 0xff] <<  8 ^
+           (u32)Td4[(s1 >> 16) & 0xff] << 16 ^
+           (u32)Td4[(s0 >> 24)       ] << 24;
 
     /* now do the linear transform using words */ 
     {
@@ -965,22 +965,22 @@ void AES_decrypt(const unsigned char *in, unsigned char *out,
      */
     for (rk+=8,r=key->rounds-2; r>0; rk+=4,r--) {
 #if defined(AES_COMPACT_IN_INNER_ROUNDS)
-        t[0] =  Td4[(s0      ) & 0xff]       ^
-        Td4[(s3 >>  8) & 0xff] <<  8 ^
-        Td4[(s2 >> 16) & 0xff] << 16 ^
-        Td4[(s1 >> 24)       ] << 24;
-        t[1] =  Td4[(s1      ) & 0xff]       ^
-        Td4[(s0 >>  8) & 0xff] <<  8 ^
-        Td4[(s3 >> 16) & 0xff] << 16 ^
-        Td4[(s2 >> 24)       ] << 24;
-        t[2] =  Td4[(s2      ) & 0xff]       ^
-        Td4[(s1 >>  8) & 0xff] <<  8 ^
-        Td4[(s0 >> 16) & 0xff] << 16 ^
-        Td4[(s3 >> 24)       ] << 24;
-        t[3] =  Td4[(s3      ) & 0xff]       ^
-        Td4[(s2 >>  8) & 0xff] <<  8 ^
-        Td4[(s1 >> 16) & 0xff] << 16 ^
-        Td4[(s0 >> 24)       ] << 24;
+        t[0] = (u32)Td4[(s0      ) & 0xff]       ^
+               (u32)Td4[(s3 >>  8) & 0xff] <<  8 ^
+               (u32)Td4[(s2 >> 16) & 0xff] << 16 ^
+               (u32)Td4[(s1 >> 24)       ] << 24;
+        t[1] = (u32)Td4[(s1      ) & 0xff]       ^
+               (u32)Td4[(s0 >>  8) & 0xff] <<  8 ^
+               (u32)Td4[(s3 >> 16) & 0xff] << 16 ^
+               (u32)Td4[(s2 >> 24)       ] << 24;
+        t[2] = (u32)Td4[(s2      ) & 0xff]       ^
+               (u32)Td4[(s1 >>  8) & 0xff] <<  8 ^
+               (u32)Td4[(s0 >> 16) & 0xff] << 16 ^
+               (u32)Td4[(s3 >> 24)       ] << 24;
+        t[3] = (u32)Td4[(s3      ) & 0xff]       ^
+               (u32)Td4[(s2 >>  8) & 0xff] <<  8 ^
+               (u32)Td4[(s1 >> 16) & 0xff] << 16 ^
+               (u32)Td4[(s0 >> 24)       ] << 24;
 
     /* now do the linear transform using words */ 
     {
@@ -1044,27 +1044,27 @@ void AES_decrypt(const unsigned char *in, unsigned char *out,
     prefetch256(Td4);
 
     *(u32*)(out+0) =
-        (Td4[(s0      ) & 0xff])    ^
-        (Td4[(s3 >>  8) & 0xff] <<  8) ^
-        (Td4[(s2 >> 16) & 0xff] << 16) ^
-        (Td4[(s1 >> 24)       ] << 24) ^
+        ((u32)Td4[(s0      ) & 0xff])    ^
+        ((u32)Td4[(s3 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(s2 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(s1 >> 24)       ] << 24) ^
         rk[0];
     *(u32*)(out+4) =
-        (Td4[(s1      ) & 0xff])     ^
-        (Td4[(s0 >>  8) & 0xff] <<  8) ^
-        (Td4[(s3 >> 16) & 0xff] << 16) ^
-        (Td4[(s2 >> 24)       ] << 24) ^
+        ((u32)Td4[(s1      ) & 0xff])     ^
+        ((u32)Td4[(s0 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(s3 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(s2 >> 24)       ] << 24) ^
         rk[1];
     *(u32*)(out+8) =
-        (Td4[(s2      ) & 0xff])     ^
-        (Td4[(s1 >>  8) & 0xff] <<  8) ^
-        (Td4[(s0 >> 16) & 0xff] << 16) ^
-        (Td4[(s3 >> 24)       ] << 24) ^
+        ((u32)Td4[(s2      ) & 0xff])     ^
+        ((u32)Td4[(s1 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(s0 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(s3 >> 24)       ] << 24) ^
         rk[2];
     *(u32*)(out+12) =
-        (Td4[(s3      ) & 0xff])     ^
-        (Td4[(s2 >>  8) & 0xff] <<  8) ^
-        (Td4[(s1 >> 16) & 0xff] << 16) ^
-        (Td4[(s0 >> 24)       ] << 24) ^
+        ((u32)Td4[(s3      ) & 0xff])     ^
+        ((u32)Td4[(s2 >>  8) & 0xff] <<  8) ^
+        ((u32)Td4[(s1 >> 16) & 0xff] << 16) ^
+        ((u32)Td4[(s0 >> 24)       ] << 24) ^
         rk[3];
 }