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

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

 ; LICENSE:
 ; This submission to NSS is to be made available under the terms of the
 ; Mozilla Public License, v. 2.0. You can obtain one at http:
 ; //mozilla.org/MPL/2.0/.
 ;###############################################################################
 ; Copyright(c) 2014, Intel Corp.
 ; Developers and authors:
 ; Shay Gueron and Vlad Krasnov
 ; Intel Corporation, Israel Development Centre, Haifa, Israel
 ; Please send feedback directly to crypto.feedback.alias@intel.com
@@ skipping 822 matching lines @@
     vaesenc TMP1, TMP1, XMMWORD PTR[12*16 + KS]
     vaesenc TMP1, TMP1, XMMWORD PTR[13*16 + KS]
     vmovdqu TMP2, XMMWORD PTR[14*16 + KS]
 @@:
     vaesenclast TMP1, TMP1, TMP2
 ; zero a temp location
     vpxor   TMP2, TMP2, TMP2
     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
 @@:
     vpxor   TMP1, TMP1, XMMWORD PTR[esp]
     vmovdqa XMMWORD PTR[esp], TMP1
     xor KS, KS
 @@:
         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
 @@:
         cmp KS, 16
         je  @f
         mov BYTE PTR[esp + KS], 0
         inc KS
         jmp @b
 @@:
+    mov edx, aluTMP
     vmovdqa TMP1, XMMWORD PTR[esp]
 
     vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
     vpxor   TMP1, TMP1, T
 
     vmovdqu TMP0, XMMWORD PTR[Htbl]
     GFMUL   TMP1, TMP1, TMP0, TMP5, TMP2, TMP3, TMP4
     vmovdqu T, TMP1
 
 LEncDataEnd:
@@ skipping 265 matching lines @@
     cmp NR, 12
     je  @f
     vaesenc TMP1, TMP1, XMMWORD PTR[12*16 + KS]
     vaesenc TMP1, TMP1, XMMWORD PTR[13*16 + KS]
     vmovdqu TMP2, XMMWORD PTR[14*16 + KS]
 @@:
     vaesenclast xmm7, TMP1, TMP2
 
 ; 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
 @@:
         cmp KS, 16
         je  @f
         mov BYTE PTR[esp + KS], 0
         inc KS
         jmp @b
 @@:
-
+    mov edx, aluTMP
     vmovdqa TMP1, XMMWORD PTR[esp]
     vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
     vpxor   TMP1, TMP1, T
 
     vmovdqu TMP0, XMMWORD PTR[Htbl]
     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
+        mov dl, BYTE PTR[esp + KS]
+        mov BYTE PTR[PT + KS], dl
         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
-        inc KS
-        jmp @b
-@@:
+    mov edx, aluTMP
 
 LDecDataEnd:
 
     bswap   aluCTR
     mov     [16*16 + 2*16 + 3*4 + Gctx], aluCTR
 
     mov esp, ebp
     pop edi
     pop esi
     pop ebx
     pop ebp
 
     vzeroupper
 
     ret
 intel_aes_gcmDEC ENDP
 
 
 END