Fix bugs in intel-gcm-x86-masm.asm and re-enable the Intel AES assembly

nss/lib/freebl/intel-gcm-x86-masm.asm

Index: nss/lib/freebl/intel-gcm-x86-masm.asm
diff --git a/nss/lib/freebl/intel-gcm-x86-masm.asm b/nss/lib/freebl/intel-gcm-x86-masm.asm
index d8ba9b4961cf26a073905d760111755efe6f7058..917b58f64495a06256ec4c6cac5a5ec22d20765d 100644
--- a/nss/lib/freebl/intel-gcm-x86-masm.asm
+++ b/nss/lib/freebl/intel-gcm-x86-masm.asm
@@ -840,11 +840,12 @@ LEncDataTail:
     vmovdqa XMMWORD PTR[esp], TMP2
 ; copy as many bytes as needed
     xor KS, KS
+    mov aluTMP, edx
 @@:
         cmp len, KS
         je  @f
-        mov di, [PT + KS]
-        mov [esp + KS], di
+        mov dl, BYTE PTR[PT + KS]
+        mov BYTE PTR[esp + KS], dl
         inc KS
         jmp @b
 @@:
@@ -854,8 +855,8 @@ LEncDataTail:
 @@:
         cmp len, KS
         je  @f
-        mov di, [esp + KS]
-        mov [CT + KS], di
+        mov dl, BYTE PTR[esp + KS]
+        mov BYTE PTR[CT + KS], dl
         inc KS
         jmp @b
 @@:
@@ -865,6 +866,7 @@ LEncDataTail:
         inc KS
         jmp @b
 @@:
+    mov edx, aluTMP
     vmovdqa TMP1, XMMWORD PTR[esp]
 
     vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
@@ -1150,11 +1152,12 @@ LDecDataTail:
 
 ; copy as many bytes as needed
     xor KS, KS
+    mov aluTMP, edx
 @@:
         cmp len, KS
         je  @f
-        mov di, [CT + KS]
-        mov [esp + KS], di
+        mov dl, BYTE PTR[CT + KS]
+        mov BYTE PTR[esp + KS], dl
         inc KS
         jmp @b
 @@:
@@ -1164,7 +1167,7 @@ LDecDataTail:
         inc KS
         jmp @b
 @@:
-
+    mov edx, aluTMP
     vmovdqa TMP1, XMMWORD PTR[esp]
     vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
     vpxor   TMP1, TMP1, T
@@ -1173,24 +1176,19 @@ LDecDataTail:
     GFMUL   TMP1, TMP1, TMP0, TMP5, TMP2, TMP3, TMP4
     vmovdqu T, TMP1
 
-
     vpxor   xmm7, xmm7, XMMWORD PTR[esp]
     vmovdqa XMMWORD PTR[esp], xmm7
     xor     KS, KS
+    mov aluTMP, edx
 @@:
         cmp len, KS
         je  @f
-        mov di, [esp + KS]
-        mov [PT + KS], di
-        inc KS
-        jmp @b
-@@:
-        cmp KS, 16

[agl, 2014/04/30 21:30:44]

Just highlighting that this version of the change also removes this zero-filling
to the next 16 byte granularity.

[wtc, 2014/04/30 21:36:58]

This removal is also in the previous version of the CL. This loop writes 0 bytes
beyond the end of the PT buffer, corrupting memory.

-        je  @f
-        mov BYTE PTR[PT + KS], 0
+        mov dl, BYTE PTR[esp + KS]
+        mov BYTE PTR[PT + KS], dl
         inc KS
         jmp @b
 @@:
+    mov edx, aluTMP
 
 LDecDataEnd: