Workaround added for AMDGPU-Pro OpenCL runtime: AES encrypt and decrypt Invertkey function was calculated wrong in certain cases

OpenCL/m06600.cl

@@ -732,7 +732,7 @@ static void AES128_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES128_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 40; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 40; i < 20; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m06800.cl

@@ -759,7 +759,7 @@ static void AES256_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m08800.cl

@@ -726,7 +726,7 @@ static void AES128_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES128_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 40; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 40; i < 20; i += 4, j -= 4)
   {
     u32 temp;
 
@@ -896,7 +896,7 @@ static void AES256_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m09400.cl

@@ -729,7 +729,7 @@ static void AES128_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES128_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 40; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 40; i < 20; i += 4, j -= 4)
   {
     u32 temp;
 
@@ -967,7 +967,7 @@ static void AES256_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m09500.cl

@@ -729,7 +729,7 @@ static void AES128_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES128_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 40; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 40; i < 20; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m09600.cl

@@ -753,7 +753,7 @@ static void AES256_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m11300.cl

@@ -756,7 +756,7 @@ static void AES256_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m11600.cl

@@ -759,7 +759,7 @@ static void AES256_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m12500.cl

@@ -740,7 +740,7 @@ static void AES128_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES128_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 40; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 40; i < 20; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m12700.cl

@@ -759,7 +759,7 @@ static void AES256_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m13200.cl

@@ -729,7 +729,7 @@ static void AES128_ExpandKey (u32 *userkey, u32 *rek, __local u32 *s_te0, __loca
 
 static void AES128_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1, __local u32 *s_td2, __local u32 *s_td3, __local u32 *s_td4, __local u32 *s_te0, __local u32 *s_te1, __local u32 *s_te2, __local u32 *s_te3, __local u32 *s_te4)
 {
-  for (u32 i = 0, j = 40; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 40; i < 20; i += 4, j -= 4)
   {
     u32 temp;
 

OpenCL/m13400.cl

@@ -748,7 +748,7 @@ static void AES256_InvertKey (u32 *rdk, __local u32 *s_td0, __local u32 *s_td1,
   #ifdef _unroll
   #pragma unroll
   #endif
-  for (u32 i = 0, j = 56; i < j; i += 4, j -= 4)
+  for (u32 i = 0, j = 56; i < 28; i += 4, j -= 4)
   {
     u32 temp;
 

docs/changes.txt

@@ -1,3 +1,11 @@
+* changes v3.30 -> v3.xx:
+
+##
+## Workarounds
+##
+
+- Workaround added for AMDGPU-Pro OpenCL runtime: AES encrypt and decrypt Invertkey function was calculated wrong in certain cases
+
 * changes v3.20 -> v3.30:
 
 ##