Bug#14211271 ISSUES WITH SSL ON DEBIAN WHEEZY I386 AND KFREEBSD-I386

Problem:
It was reported that on Debian and KFreeBSD platforms, i386 architecture 
machines certain SSL tests are failing. main.ssl_connect  rpl.rpl_heartbeat_ssl
rpl.rpl_ssl1 rpl.rpl_ssl main.ssl_cipher, main.func_encrypt were the tests that
 were reportedly failing (crashing). The reason for the crashes are said to be
due to the assembly code of yaSSL.

Solution:
There was initially a workaround suggested i.e., to enable 
-DTAOCRYPT_DISABLE_X86ASM flag which would prevent the crash, but at an expense
 of 4X reduction of speed. Since this was unacceptable, the fix was the 
functions using assembly, now input variables from the function call using 
extended inline assembly on GCC instead of relying on direct assembly code.

12 files changed, 282 insertions, 179 deletions

diff --git a/extra/yassl/taocrypt/include/integer.hpp b/extra/yassl/taocrypt/include/integer.hpp
index e1a2ed5f426..68f3c4bbf39 100644
--- a/extra/yassl/taocrypt/include/integer.hpp
+++ b/extra/yassl/taocrypt/include/integer.hpp
@@ -46,6 +46,14 @@
 
 
 #ifdef TAOCRYPT_X86ASM_AVAILABLE
+    #if defined(__GNUC__) && (__GNUC__ >= 4)
+        // GCC 4 or greater optimizes too much inline on recursive for bigint,
+        // -O3 just as fast without asm here anyway
+        #undef TAOCRYPT_X86ASM_AVAILABLE
+    #endif
+#endif
+
+#ifdef TAOCRYPT_X86ASM_AVAILABLE
 
 #ifdef _M_IX86
     #if (defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 500)) || \
diff --git a/extra/yassl/taocrypt/include/misc.hpp b/extra/yassl/taocrypt/include/misc.hpp
index 1382d3690fb..f68abbda8da 100644
--- a/extra/yassl/taocrypt/include/misc.hpp
+++ b/extra/yassl/taocrypt/include/misc.hpp
@@ -124,15 +124,22 @@ void CleanUp();
 
 
 // no gas on these systems ?, disable for now
-#if defined(__sun__) || defined (__APPLE__)
+#if defined(__sun__)
+    #undef  TAOCRYPT_DISABLE_X86ASM
     #define TAOCRYPT_DISABLE_X86ASM
 #endif
 
 // icc problem with -03 and integer, disable for now
 #if defined(__INTEL_COMPILER)
+    #undef  TAOCRYPT_DISABLE_X86ASM
     #define TAOCRYPT_DISABLE_X86ASM
 #endif
 
+// indpedent of build system, unless ia32 asm is enabled disable it
+#if !defined(TAOCRYPT_ENABLE_X86ASM)
+    #undef  TAOCRYPT_DISABLE_X86ASM
+    #define TAOCRYPT_DISABLE_X86ASM
+#endif
 
 // Turn on ia32 ASM for Big Integer
 // CodeWarrior defines _MSC_VER
diff --git a/extra/yassl/taocrypt/src/aes.cpp b/extra/yassl/taocrypt/src/aes.cpp
index 9b04e7b0dbc..d04ef69c19f 100644
--- a/extra/yassl/taocrypt/src/aes.cpp
+++ b/extra/yassl/taocrypt/src/aes.cpp
@@ -52,7 +52,7 @@ void AES::Process(byte* out, const byte* in, word32 sz)
             in  += BLOCK_SIZE;
         }
     else if (mode_ == CBC) {
-        if (dir_ == ENCRYPTION)
+        if (dir_ == ENCRYPTION) {
             while (blocks--) {
                 r_[0] ^= *(word32*)in;
                 r_[1] ^= *(word32*)(in +  4);
@@ -65,7 +65,8 @@ void AES::Process(byte* out, const byte* in, word32 sz)
                 out += BLOCK_SIZE;
                 in  += BLOCK_SIZE;
             }
-        else
+        }
+    else {
             while (blocks--) {
                 AsmDecrypt(in, out, (void*)Td0);
                 
@@ -78,6 +79,7 @@ void AES::Process(byte* out, const byte* in, word32 sz)
                 out += BLOCK_SIZE;
                 in  += BLOCK_SIZE;
             }
+       }
    }
 }
 
@@ -452,27 +454,31 @@ void AES::decrypt(const byte* inBlock, const byte* xorBlock,
 
 #if defined(DO_AES_ASM)
     #ifdef __GNUC__
-        #define AS1(x)    asm(#x);
-        #define AS2(x, y) asm(#x ", " #y);
+        #define AS1(x)    #x ";"
+        #define AS2(x, y) #x ", " #y ";"
 
         #define PROLOG()  \
-            asm(".intel_syntax noprefix"); \
-            AS2(    movd  mm3, edi                      )   \
-            AS2(    movd  mm4, ebx                      )   \
-            AS2(    sub   esp, 4                        )   \
-            AS2(    movd  mm7, ebp                      )   \
-            AS2(    mov   [ebp - 4], esi                )   \
-            AS2(    mov   ecx, DWORD PTR [ebp +  8]     )   \
-            AS2(    mov   esi, DWORD PTR [ebp + 12]     )   \
-            AS2(    mov   ebp, DWORD PTR [ebp + 20]     )
+        __asm__ __volatile__ \
+        ( \
+            ".intel_syntax noprefix;" \
+            "push ebx;" \
+            "push ebp;" \
+            "movd mm7, ebp;" \
+            "movd mm4, eax;" \
+            "mov  ebp, edx;"  \
+            "sub  esp, 4;"
 
         #define EPILOG()  \
-            AS2(    mov  esi, [ebp - 4]             )   \
-            AS2(    mov  esp, ebp                   )   \
-            AS2(    movd ebx, mm4                   )   \
-            AS2(    movd edi, mm3                   )   \
-            AS1(    emms                            )   \
-            asm(".att_syntax");
+            "add esp, 4;" \
+            "pop ebp;" \
+            "pop ebx;" \
+                   "emms;" \
+                   ".att_syntax;" \
+                : \
+                : "c" (this), "S" (inBlock), "d" (boxes), "a" (outBlock) \
+                : "%edi", "memory", "cc" \
+        );
+
     #else
         #define AS1(x)    __asm x
         #define AS2(x, y) __asm x, y
@@ -504,6 +510,8 @@ void AES::decrypt(const byte* inBlock, const byte* xorBlock,
 
 #ifdef _MSC_VER
     __declspec(naked) 
+#else
+    __attribute__ ((noinline))
 #endif
 void AES::AsmEncrypt(const byte* inBlock, byte* outBlock, void* boxes) const
 {
@@ -537,7 +545,11 @@ void AES::AsmEncrypt(const byte* inBlock, byte* outBlock, void* boxes) const
     AS2(    xor   ecx, DWORD PTR [edi +  8]          )   // s2
     AS2(    xor   edx, DWORD PTR [edi + 12]          )   // s3
 
-    AS1(loop1:                                                          )
+#ifdef _MSC_VER
+    AS1( loop1: )  // loop1
+#else
+    AS1(1:  )      // loop1
+#endif
             /* Put0 (mm0) =  
                 Te0[get0,rs 24] ^
                 Te1[get1,rs 16] ^
@@ -652,7 +664,11 @@ void AES::AsmEncrypt(const byte* inBlock, byte* outBlock, void* boxes) const
     AS1(    dec   edi                                                   )
     AS2(    movd  mm5, edi                                              )
 
-    AS1(    jnz   loop1                                                 )
+#ifdef _MSC_VER
+    AS1(    jnz   loop1)  // loop1
+#else
+    AS1(    jnz   1b )    // loop1
+#endif
 
             // last round
             /*
@@ -799,9 +815,9 @@ void AES::AsmEncrypt(const byte* inBlock, byte* outBlock, void* boxes) const
 
             // store
     #ifdef __GNUC__
-        AS2(    mov esi, DWORD PTR [ebp + 16]       )   //  outBlock
+        AS2(    movd esi, mm4                       )   //  outBlock
     #else
-        AS2(    mov esi, DWORD PTR [ebp + 12]       )   //  outBlock
+        AS2(    mov  esi, DWORD PTR [ebp + 12]      )   //  outBlock
     #endif
 
     AS1(    bswap ecx                                                   )
@@ -818,7 +834,9 @@ void AES::AsmEncrypt(const byte* inBlock, byte* outBlock, void* boxes) const
 
 
 #ifdef _MSC_VER
-    __declspec(naked) 
+    __declspec(naked)
+#else
+    __attribute__ ((noinline)) 
 #endif
 void AES::AsmDecrypt(const byte* inBlock, byte* outBlock, void* boxes) const
 {
@@ -853,7 +871,11 @@ void AES::AsmDecrypt(const byte* inBlock, byte* outBlock, void* boxes) const
     AS2(    xor   edx, DWORD PTR [edi + 12]          )   // s3
 
 
-    AS1(loop2:                                                          )
+#ifdef _MSC_VER
+    AS1( loop2: )  // loop2
+#else
+    AS1(2:  )      // loop2
+#endif
        /*   Put0 (mm0) =
             Td0[GETBYTE(get0, rs24)] ^
             Td1[GETBYTE(get3, rs16)] ^
@@ -964,7 +986,11 @@ void AES::AsmDecrypt(const byte* inBlock, byte* outBlock, void* boxes) const
     AS1(    dec   edi                                                   )
     AS2(    movd  mm5, edi                                              )
 
-    AS1(    jnz   loop2                                                 )
+#ifdef _MSC_VER
+    AS1(    jnz   loop2)  // loop2
+#else
+    AS1(    jnz   2b )    // loop2
+#endif
 
             // last round
             /*
@@ -1114,9 +1140,9 @@ void AES::AsmDecrypt(const byte* inBlock, byte* outBlock, void* boxes) const
 
             // store
     #ifdef __GNUC__
-        AS2(    mov esi, DWORD PTR [ebp + 16]       )   //  outBlock
+        AS2(    movd esi, mm4                        )   //  outBlock
     #else
-        AS2(    mov esi, DWORD PTR [ebp + 12]       )   //  outBlock
+        AS2(    mov esi,  DWORD PTR [ebp + 12]       )   //  outBlock
     #endif
     AS2(    mov DWORD PTR [esi],      eax                               )
     AS2(    mov DWORD PTR [esi +  4], ebx                               )
diff --git a/extra/yassl/taocrypt/src/arc4.cpp b/extra/yassl/taocrypt/src/arc4.cpp
index a76a37904f9..f5794ec2566 100644
--- a/extra/yassl/taocrypt/src/arc4.cpp
+++ b/extra/yassl/taocrypt/src/arc4.cpp
@@ -111,28 +111,27 @@ void ARC4::Process(byte* out, const byte* in, word32 length)
 void ARC4::AsmProcess(byte* out, const byte* in, word32 length)
 {
 #ifdef __GNUC__
-    #define AS1(x)    asm(#x);
-    #define AS2(x, y) asm(#x ", " #y);
+    #define AS1(x)    #x ";"
+    #define AS2(x, y) #x ", " #y ";"
 
     #define PROLOG()  \
-        asm(".intel_syntax noprefix"); \
-        AS2(    movd  mm3, edi                      )   \
-        AS2(    movd  mm4, ebx                      )   \
-        AS2(    movd  mm5, esi                      )   \
-        AS2(    movd  mm6, ebp                      )   \
-        AS2(    mov   ecx, DWORD PTR [ebp +  8]     )   \
-        AS2(    mov   edi, DWORD PTR [ebp + 12]     )   \
-        AS2(    mov   esi, DWORD PTR [ebp + 16]     )   \
-        AS2(    mov   ebp, DWORD PTR [ebp + 20]     )
+    __asm__ __volatile__ \
+    ( \
+        ".intel_syntax noprefix;" \
+        "push ebx;" \
+        "push ebp;" \
+        "mov ebp, eax;"
 
     #define EPILOG()  \
-        AS2(    movd  ebp, mm6                  )   \
-        AS2(    movd  esi, mm5                  )   \
-        AS2(    movd  ebx, mm4                  )   \
-        AS2(    mov   esp, ebp                  )   \
-        AS2(    movd  edi, mm3                  )   \
-        AS1(    emms                            )   \
-        asm(".att_syntax");
+        "pop ebp;" \
+        "pop ebx;" \
+               "emms;" \
+               ".att_syntax;" \
+            : \
+            : "c" (this), "D" (out), "S" (in), "a" (length) \
+            : "%edx", "memory", "cc" \
+    );
+
 #else
     #define AS1(x)    __asm x
     #define AS2(x, y) __asm x, y
@@ -178,7 +177,11 @@ void ARC4::AsmProcess(byte* out, const byte* in, word32 length)
     AS2(    movzx  eax, BYTE PTR [ebp + ecx]    )
 
 
-AS1( begin:                             )
+#ifdef _MSC_VER
+    AS1( loopStart: )  // loopStart
+#else
+    AS1( 0: )          // loopStart for some gas (need numeric for jump back
+#endif
 
     // y = (y+a) & 0xff;
     AS2(    add    edx, eax                     )
@@ -215,7 +218,11 @@ AS1( begin:                             )
     AS1(    inc    edi                          )
 
     AS1(    dec    DWORD PTR [esp]              )
-    AS1(    jnz    begin                        )
+#ifdef _MSC_VER
+    AS1(    jnz   loopStart )  // loopStart
+#else
+    AS1(    jnz   0b )         // loopStart
+#endif
 
 
     // write back to x_ and y_
@@ -225,6 +232,8 @@ AS1( begin:                             )
 
 AS1( nothing:                           )
 
+    // inline adjust
+    AS2(    add   esp, 4               )   // fix room on stack
 
     EPILOG()
 }
diff --git a/extra/yassl/taocrypt/src/blowfish.cpp b/extra/yassl/taocrypt/src/blowfish.cpp
index cedae6af281..5cfd81e369c 100644
--- a/extra/yassl/taocrypt/src/blowfish.cpp
+++ b/extra/yassl/taocrypt/src/blowfish.cpp
@@ -54,7 +54,7 @@ void Blowfish::Process(byte* out, const byte* in, word32 sz)
             in  += BLOCK_SIZE;
         }
     else if (mode_ == CBC) {
-        if (dir_ == ENCRYPTION)
+        if (dir_ == ENCRYPTION) {
             while (blocks--) {
                 r_[0] ^= *(word32*)in;
                 r_[1] ^= *(word32*)(in + 4);
@@ -66,7 +66,8 @@ void Blowfish::Process(byte* out, const byte* in, word32 sz)
                 out += BLOCK_SIZE;
                 in  += BLOCK_SIZE;
             }
-        else
+        }
+        else {
             while (blocks--) {
                 AsmProcess(in, out);
                 
@@ -78,6 +79,7 @@ void Blowfish::Process(byte* out, const byte* in, word32 sz)
                 out += BLOCK_SIZE;
                 in  += BLOCK_SIZE;
             }
+        }
     }
 }
 
@@ -222,23 +224,26 @@ void Blowfish::ProcessAndXorBlock(const byte* in, const byte* xOr, byte* out)
 
 #if defined(DO_BLOWFISH_ASM)
     #ifdef __GNUC__
-        #define AS1(x)    asm(#x);
-        #define AS2(x, y) asm(#x ", " #y);
+        #define AS1(x)    #x ";"
+        #define AS2(x, y) #x ", " #y ";"
 
         #define PROLOG()  \
-            asm(".intel_syntax noprefix"); \
-            AS2(    movd  mm3, edi                      )   \
-            AS2(    movd  mm4, ebx                      )   \
-            AS2(    movd  mm5, esi                      )   \
-            AS2(    mov   ecx, DWORD PTR [ebp +  8]     )   \
-            AS2(    mov   esi, DWORD PTR [ebp + 12]     )
-
+        __asm__ __volatile__ \
+        ( \
+            ".intel_syntax noprefix;" \
+            "push ebx;" \
+            "push ebp;" \
+            "movd mm3, eax;"
         #define EPILOG()  \
-            AS2(    movd esi, mm5                  )   \
-            AS2(    movd ebx, mm4                  )   \
-            AS2(    movd edi, mm3                  )   \
-            AS1(    emms                           )   \
-            asm(".att_syntax");
+            "pop ebp;" \
+            "pop ebx;" \
+                   "emms;" \
+                   ".att_syntax;" \
+                : \
+                : "c" (this), "S" (inBlock), "a" (outBlock) \
+                : "%edi", "%edx", "memory", "cc" \
+        );
+
     #else
         #define AS1(x)    __asm x
         #define AS2(x, y) __asm x, y
@@ -286,7 +291,9 @@ void Blowfish::ProcessAndXorBlock(const byte* in, const byte* xOr, byte* out)
 
 
 #ifdef _MSC_VER
-    __declspec(naked) 
+    __declspec(naked)
+#else
+    __attribute__ ((noinline))
 #endif
 void Blowfish::AsmProcess(const byte* inBlock, byte* outBlock) const
 {
@@ -335,7 +342,7 @@ void Blowfish::AsmProcess(const byte* inBlock, byte* outBlock) const
     #endif
 
     #ifdef __GNUC__
-        AS2(    mov   edi, [ebp + 16]           ) // outBlock
+        AS2(    movd  edi, mm3                  ) // outBlock
     #else
         AS2(    mov   edi, [ebp + 12]           ) // outBlock
     #endif
diff --git a/extra/yassl/taocrypt/src/des.cpp b/extra/yassl/taocrypt/src/des.cpp
index 5cffeaea9c9..2b9a9d9019e 100644
--- a/extra/yassl/taocrypt/src/des.cpp
+++ b/extra/yassl/taocrypt/src/des.cpp
@@ -473,7 +473,7 @@ void DES_EDE3::ProcessAndXorBlock(const byte* in, const byte* xOr,
 
    uses ecx
 */
-#define AsmIPERM() {\
+#define AsmIPERM() \
     AS2(    rol   ebx, 4                        )   \
     AS2(    mov   ecx, eax                      )   \
     AS2(    xor   ecx, ebx                      )   \
@@ -504,7 +504,7 @@ void DES_EDE3::ProcessAndXorBlock(const byte* in, const byte* xOr,
     AS2(    and   ecx, 0xaaaaaaaa               )   \
     AS2(    xor   eax, ecx                      )   \
     AS2(    rol   eax, 1                        )   \
-    AS2(    xor   ebx, ecx                      ) }
+    AS2(    xor   ebx, ecx                      )
 
 
 /* Uses FPERM algorithm from above
@@ -514,7 +514,7 @@ void DES_EDE3::ProcessAndXorBlock(const byte* in, const byte* xOr,
 
    uses ecx
 */
-#define AsmFPERM()    {\
+#define AsmFPERM()    \
     AS2(    ror  ebx, 1                     )    \
     AS2(    mov  ecx, eax                   )    \
     AS2(    xor  ecx, ebx                   )    \
@@ -545,7 +545,7 @@ void DES_EDE3::ProcessAndXorBlock(const byte* in, const byte* xOr,
     AS2(    and  ecx, 0xf0f0f0f0            )    \
     AS2(    xor  eax, ecx                   )    \
     AS2(    xor  ebx, ecx                   )    \
-    AS2(    ror  eax, 4                     ) }
+    AS2(    ror  eax, 4                     )
 
 
 
@@ -641,32 +641,34 @@ void DES_EDE3::ProcessAndXorBlock(const byte* in, const byte* xOr,
 
 
 #ifdef _MSC_VER
-    __declspec(naked) 
+    __declspec(naked)
+#else
+    __attribute__ ((noinline)) 
 #endif
 void DES_EDE3::AsmProcess(const byte* in, byte* out, void* box) const
 {
 #ifdef __GNUC__
-    #define AS1(x)    asm(#x);
-    #define AS2(x, y) asm(#x ", " #y);
-
-    asm(".intel_syntax noprefix");
+    #define AS1(x)    #x ";"
+    #define AS2(x, y) #x ", " #y ";"
 
     #define PROLOG()  \
-        AS2(    movd  mm3, edi                      )   \
-        AS2(    movd  mm4, ebx                      )   \
-        AS2(    movd  mm5, esi                      )   \
-        AS2(    movd  mm6, ebp                      )   \
-        AS2(    mov   edx, DWORD PTR [ebp +  8]     )   \
-        AS2(    mov   esi, DWORD PTR [ebp + 12]     )   \
-        AS2(    mov   ebp, DWORD PTR [ebp + 20]     )
-
-    // ebp restored at end
-    #define EPILOG()    \
-        AS2(    movd  edi, mm3                      )   \
-        AS2(    movd  ebx, mm4                      )   \
-        AS2(    movd  esi, mm5                      )   \
-        AS1(    emms                                )   \
-        asm(".att_syntax");
+    __asm__ __volatile__ \
+    ( \
+        ".intel_syntax noprefix;" \
+        "push ebx;" \
+        "push ebp;" \
+        "movd mm6, ebp;" \
+        "movd mm7, ecx;" \
+        "mov  ebp, eax;"
+    #define EPILOG()  \
+        "pop ebp;" \
+        "pop ebx;" \
+               "emms;" \
+               ".att_syntax;" \
+            :  \
+            : "d" (this), "S" (in), "a" (box), "c" (out) \
+            : "%edi", "memory", "cc" \
+    );
 
 #else
     #define AS1(x)      __asm x
@@ -756,7 +758,7 @@ void DES_EDE3::AsmProcess(const byte* in, byte* out, void* box) const
     AS1(    bswap eax                           )
 
 #ifdef __GNUC__
-    AS2(    mov   esi, DWORD PTR [ebp +  16]    )   // outBlock
+    AS2(    movd  esi, mm7   )   // outBlock
 #else
     AS2(    mov   esi, DWORD PTR [ebp +  12]    )   // outBlock
 #endif
diff --git a/extra/yassl/taocrypt/src/integer.cpp b/extra/yassl/taocrypt/src/integer.cpp
index bfced6d0c74..c981b1bf786 100644
--- a/extra/yassl/taocrypt/src/integer.cpp
+++ b/extra/yassl/taocrypt/src/integer.cpp
@@ -55,7 +55,9 @@
         #include <emmintrin.h>
     #endif
 #elif defined(_MSC_VER) && defined(_M_IX86)
-    #pragma message("You do not seem to have the Visual C++ Processor Pack ")
+/*    #pragma message("You do not seem to have the Visual C++ Processor Pack ")
+     #pragma message("installed, so use of SSE2 intrinsics will be disabled.")
+*/
     #pragma message("installed, so use of SSE2 intrinsics will be disabled.")
 #elif defined(__GNUC__) && defined(__i386__)
 /*   #warning You do not have GCC 3.3 or later, or did not specify the -msse2 \
@@ -73,7 +75,7 @@ template <class T>
 CPP_TYPENAME AlignedAllocator<T>::pointer AlignedAllocator<T>::allocate(
                                            size_type n, const void *)
 {
-    if (n > max_size())
+    if (n > this->max_size())
         return 0;
     if (n == 0)
         return 0;
diff --git a/extra/yassl/taocrypt/src/md5.cpp b/extra/yassl/taocrypt/src/md5.cpp
index ee0d64acc17..e9a9e8fe517 100644
--- a/extra/yassl/taocrypt/src/md5.cpp
+++ b/extra/yassl/taocrypt/src/md5.cpp
@@ -223,32 +223,32 @@ void MD5::Update(const byte* data, word32 len)
 
 
 #ifdef _MSC_VER
-    __declspec(naked) 
+    __declspec(naked)
+#else
+    __attribute__ ((noinline))
 #endif
 void MD5::AsmTransform(const byte* data, word32 times)
 {
 #ifdef __GNUC__
-    #define AS1(x)    asm(#x);
-    #define AS2(x, y) asm(#x ", " #y);
+    #define AS1(x)    #x ";"
+    #define AS2(x, y) #x ", " #y ";"
 
     #define PROLOG()  \
-        asm(".intel_syntax noprefix"); \
-        AS2(    movd  mm3, edi                      )   \
-        AS2(    movd  mm4, ebx                      )   \
-        AS2(    movd  mm5, esi                      )   \
-        AS2(    movd  mm6, ebp                      )   \
-        AS2(    mov   ecx, DWORD PTR [ebp +  8]     )   \
-        AS2(    mov   edi, DWORD PTR [ebp + 12]     )   \
-        AS2(    mov   eax, DWORD PTR [ebp + 16]     )
-
+    __asm__ __volatile__ \
+    ( \
+        ".intel_syntax noprefix;" \
+        "push ebx;" \
+        "push ebp;"
     #define EPILOG()  \
-        AS2(    movd  ebp, mm6                  )   \
-        AS2(    movd  esi, mm5                  )   \
-        AS2(    movd  ebx, mm4                  )   \
-        AS2(    mov   esp, ebp                  )   \
-        AS2(    movd  edi, mm3                  )   \
-        AS1(    emms                            )   \
-        asm(".att_syntax");
+        "pop ebp;" \
+        "pop ebx;" \
+               "emms;" \
+               ".att_syntax;" \
+            : \
+            : "c" (this), "D" (data), "a" (times) \
+            : "%esi", "%edx", "memory", "cc" \
+    );
+
 #else
     #define AS1(x)    __asm x
     #define AS2(x, y) __asm x, y
@@ -294,7 +294,11 @@ void MD5::AsmTransform(const byte* data, word32 times)
     AS2(    mov   ecx, [esi +  8]       )   // c
     AS2(    mov   edx, [esi + 12]       )   // d
   
-AS1(loopStart:)
+#ifdef _MSC_VER
+    AS1( loopStart: )  // loopStart
+#else
+    AS1( 0: )          // loopStart for some gas (need numeric for jump back
+#endif
 
     // set up
     AS2(    mov   esi, ecx      )
@@ -389,7 +393,11 @@ AS1(loopStart:)
     AS2(    movd  ebp, mm2              )   // times
     AS1(    dec   ebp                   )
     AS2(    movd  mm2, ebp              )
-    AS1(    jnz   loopStart             )
+#ifdef _MSC_VER
+    AS1(    jnz   loopStart )  // loopStart
+#else
+    AS1(    jnz   0b )         // loopStart
+#endif
 
 
     EPILOG()
diff --git a/extra/yassl/taocrypt/src/rabbit.cpp b/extra/yassl/taocrypt/src/rabbit.cpp
index 071a25191f8..89e6a207a1b 100644
--- a/extra/yassl/taocrypt/src/rabbit.cpp
+++ b/extra/yassl/taocrypt/src/rabbit.cpp
@@ -198,7 +198,6 @@ void Rabbit::Process(byte* output, const byte* input, word32 msglen)
 {
     /* Temporary variables */
     word32 i;
-    byte buffer[16];
 
     /* Encrypt/decrypt all full blocks */
     while (msglen >= 16) {
@@ -227,17 +226,23 @@ void Rabbit::Process(byte* output, const byte* input, word32 msglen)
 
     /* Encrypt/decrypt remaining data */
     if (msglen) {
+
+        word32 tmp[4];
+        byte*  buffer = (byte*)tmp;
+
+        memset(tmp, 0, sizeof(tmp));   /* help static analysis */
+
         /* Iterate the system */
         NextState(Work);
 
         /* Generate 16 bytes of pseudo-random data */
-        *(word32*)(buffer+ 0) = LITTLE32(workCtx_.x[0] ^
+        tmp[0] = LITTLE32(workCtx_.x[0] ^ 
                   (workCtx_.x[5]>>16) ^ U32V(workCtx_.x[3]<<16));
-        *(word32*)(buffer+ 4) = LITTLE32(workCtx_.x[2] ^ 
+        tmp[1] = LITTLE32(workCtx_.x[2] ^ 
                   (workCtx_.x[7]>>16) ^ U32V(workCtx_.x[5]<<16));
-        *(word32*)(buffer+ 8) = LITTLE32(workCtx_.x[4] ^ 
+        tmp[2] = LITTLE32(workCtx_.x[4] ^ 
                   (workCtx_.x[1]>>16) ^ U32V(workCtx_.x[7]<<16));
-        *(word32*)(buffer+12) = LITTLE32(workCtx_.x[6] ^ 
+        tmp[3] = LITTLE32(workCtx_.x[6] ^ 
                   (workCtx_.x[3]>>16) ^ U32V(workCtx_.x[1]<<16));
 
         /* Encrypt/decrypt the data */
diff --git a/extra/yassl/taocrypt/src/ripemd.cpp b/extra/yassl/taocrypt/src/ripemd.cpp
index f77010ca94d..b670a9eca86 100644
--- a/extra/yassl/taocrypt/src/ripemd.cpp
+++ b/extra/yassl/taocrypt/src/ripemd.cpp
@@ -511,27 +511,26 @@ void RIPEMD160::Transform()
 void RIPEMD160::AsmTransform(const byte* data, word32 times)
 {
 #ifdef __GNUC__
-    #define AS1(x)    asm(#x);
-    #define AS2(x, y) asm(#x ", " #y);
+    #define AS1(x)    #x ";"
+    #define AS2(x, y) #x ", " #y ";"
 
     #define PROLOG()  \
-        asm(".intel_syntax noprefix"); \
-        AS2(    movd  mm3, edi                      )   \
-        AS2(    movd  mm4, ebx                      )   \
-        AS2(    movd  mm5, esi                      )   \
-        AS2(    movd  mm6, ebp                      )   \
-        AS2(    mov   ecx, DWORD PTR [ebp +  8]     )   \
-        AS2(    mov   edi, DWORD PTR [ebp + 12]     )   \
-        AS2(    mov   edx, DWORD PTR [ebp + 16]     )
+    __asm__ __volatile__ \
+    ( \
+        ".intel_syntax noprefix;" \
+        "push ebx;" \
+        "push ebp;"
 
     #define EPILOG()  \
-        AS2(    movd  ebp, mm6                  )   \
-        AS2(    movd  esi, mm5                  )   \
-        AS2(    movd  ebx, mm4                  )   \
-        AS2(    mov   esp, ebp                  )   \
-        AS2(    movd  edi, mm3                  )   \
-        AS1(    emms                            )   \
-        asm(".att_syntax");
+        "pop ebp;" \
+        "pop ebx;" \
+               "emms;" \
+               ".att_syntax;" \
+            : \
+            : "c" (this), "D" (data), "d" (times) \
+            : "%esi", "%eax", "memory", "cc" \
+    );
+
 #else
     #define AS1(x)    __asm x
     #define AS2(x, y) __asm x, y
@@ -569,7 +568,11 @@ void RIPEMD160::AsmTransform(const byte* data, word32 times)
     AS2(    sub   esp, 24               )   // make room for tmp a1 - e1
     AS2(    movd  mm1, esi              )   // store digest_
     
-AS1( loopStart: )
+#ifdef _MSC_VER
+    AS1( loopStart: )  // loopStart
+#else
+    AS1( 0: )          // loopStart for some gas (need numeric for jump back
+#endif
 
     AS2(    movd  mm2, edx              )   // store times_
 
@@ -821,8 +824,14 @@ AS1( loopStart: )
     AS2(    movd  edx, mm2              )   // times
     AS2(    movd  edi, mm0              )   // data, already advanced
     AS1(    dec   edx                   )
-    AS1(    jnz   loopStart             )
+#ifdef _MSC_VER
+    AS1(    jnz   loopStart )  // loopStart
+#else
+    AS1(    jnz   0b )         // loopStart
+#endif
 
+    // inline adjust
+    AS2(    add   esp, 24               )   // fix room on stack
 
     EPILOG()
 }
diff --git a/extra/yassl/taocrypt/src/sha.cpp b/extra/yassl/taocrypt/src/sha.cpp
index f67a4f55067..0d3491eb83d 100644
--- a/extra/yassl/taocrypt/src/sha.cpp
+++ b/extra/yassl/taocrypt/src/sha.cpp
@@ -760,32 +760,33 @@ void SHA384::Transform()
 
 
 #ifdef _MSC_VER
-    __declspec(naked) 
+    __declspec(naked)
+#else
+    __attribute__ ((noinline))
 #endif
 void SHA::AsmTransform(const byte* data, word32 times)
 {
 #ifdef __GNUC__
-    #define AS1(x)    asm(#x);
-    #define AS2(x, y) asm(#x ", " #y);
+    #define AS1(x)    #x ";"
+    #define AS2(x, y) #x ", " #y ";"
 
     #define PROLOG()  \
-        asm(".intel_syntax noprefix"); \
-        AS2(    movd  mm3, edi                      )   \
-        AS2(    movd  mm4, ebx                      )   \
-        AS2(    movd  mm5, esi                      )   \
-        AS2(    movd  mm6, ebp                      )   \
-        AS2(    mov   ecx, DWORD PTR [ebp +  8]     )   \
-        AS2(    mov   edi, DWORD PTR [ebp + 12]     )   \
-        AS2(    mov   eax, DWORD PTR [ebp + 16]     )
+    __asm__ __volatile__ \
+    ( \
+        ".intel_syntax noprefix;" \
+        "push ebx;" \
+        "push ebp;"
 
     #define EPILOG()  \
-        AS2(    movd  ebp, mm6                  )   \
-        AS2(    movd  esi, mm5                  )   \
-        AS2(    movd  ebx, mm4                  )   \
-        AS2(    mov   esp, ebp                  )   \
-        AS2(    movd  edi, mm3                  )   \
-        AS1(    emms                            )   \
-        asm(".att_syntax");
+        "pop ebp;" \
+        "pop ebx;" \
+               "emms;" \
+               ".att_syntax;" \
+            : \
+            : "c" (this), "D" (data), "a" (times) \
+            : "%esi", "%edx", "memory", "cc" \
+    );
+
 #else
     #define AS1(x)    __asm x
     #define AS2(x, y) __asm x, y
@@ -826,7 +827,11 @@ void SHA::AsmTransform(const byte* data, word32 times)
 
     AS2(    sub   esp, 68               )   // make room on stack
 
-AS1( loopStart: )
+#ifdef _MSC_VER
+    AS1( loopStart: )  // loopStart
+#else
+    AS1( 0: )          // loopStart for some gas (need numeric for jump back
+#endif
 
     // byte reverse 16 words of input, 4 at a time, put on stack for W[]
 
@@ -1011,8 +1016,14 @@ AS1( loopStart: )
     
     AS1(    dec   ebp                   )
     AS2(    movd  mm2, ebp              )
-    AS1(    jnz   loopStart             )
+#ifdef _MSC_VER
+    AS1(    jnz   loopStart )  // loopStart
+#else
+    AS1(    jnz   0b )         // loopStart
+#endif
 
+    // inline adjust
+    AS2(    add   esp, 68               )   // fix room on stack
 
     EPILOG()
 }
diff --git a/extra/yassl/taocrypt/src/twofish.cpp b/extra/yassl/taocrypt/src/twofish.cpp
index 235b712b7c5..3a93ac11476 100644
--- a/extra/yassl/taocrypt/src/twofish.cpp
+++ b/extra/yassl/taocrypt/src/twofish.cpp
@@ -55,7 +55,7 @@ void Twofish::Process(byte* out, const byte* in, word32 sz)
             in  += BLOCK_SIZE;
         }
     else if (mode_ == CBC) {
-        if (dir_ == ENCRYPTION)
+        if (dir_ == ENCRYPTION) {
             while (blocks--) {
                 r_[0] ^= *(word32*)in;
                 r_[1] ^= *(word32*)(in +  4);
@@ -68,7 +68,8 @@ void Twofish::Process(byte* out, const byte* in, word32 sz)
                 out += BLOCK_SIZE;
                 in  += BLOCK_SIZE;
             }
-        else
+        }
+        else {
             while (blocks--) {
                 AsmDecrypt(in, out);
                
@@ -82,6 +83,7 @@ void Twofish::Process(byte* out, const byte* in, word32 sz)
                 out += BLOCK_SIZE;
                 in  += BLOCK_SIZE;
             }
+        }
     }
 }
 
@@ -272,25 +274,28 @@ void Twofish::decrypt(const byte* inBlock, const byte* xorBlock,
 
 #if defined(DO_TWOFISH_ASM)
     #ifdef __GNUC__
-        #define AS1(x)    asm(#x);
-        #define AS2(x, y) asm(#x ", " #y);
+        #define AS1(x)    #x ";"
+        #define AS2(x, y) #x ", " #y ";"
 
         #define PROLOG()  \
-            asm(".intel_syntax noprefix"); \
-            AS2(    movd  mm3, edi                      )   \
-            AS2(    movd  mm4, ebx                      )   \
-            AS2(    movd  mm5, esi                      )   \
-            AS2(    movd  mm6, ebp                      )   \
-            AS2(    mov   edi, DWORD PTR [ebp +  8]     )   \
-            AS2(    mov   esi, DWORD PTR [ebp + 12]     )
+        __asm__ __volatile__ \
+        ( \
+            ".intel_syntax noprefix;" \
+            "push ebx;" \
+            "push ebp;" \
+            "movd mm3, eax;" \
+            "movd mm6, ebp;"
 
         #define EPILOG()  \
-            AS2(    movd esp, mm6                  )   \
-            AS2(    movd esi, mm5                  )   \
-            AS2(    movd ebx, mm4                  )   \
-            AS2(    movd edi, mm3                  )   \
-            AS1(    emms                           )   \
-            asm(".att_syntax");
+            "pop ebp;" \
+            "pop ebx;" \
+                   "emms;" \
+                   ".att_syntax;" \
+                : \
+                : "D" (this), "S" (inBlock), "a" (outBlock) \
+                : "%ecx", "%edx", "memory", "cc" \
+        );
+
     #else
         #define AS1(x)    __asm x
         #define AS2(x, y) __asm x, y
@@ -424,6 +429,8 @@ void Twofish::decrypt(const byte* inBlock, const byte* xorBlock,
 
 #ifdef _MSC_VER
     __declspec(naked) 
+#else
+    __attribute__ ((noinline))
 #endif
 void Twofish::AsmEncrypt(const byte* inBlock, byte* outBlock) const
 {
@@ -472,7 +479,7 @@ void Twofish::AsmEncrypt(const byte* inBlock, byte* outBlock) const
     AS2(    movd  ebp, mm6                      )
     AS2(    movd  esi, mm0                      ) // k_
     #ifdef __GNUC__
-        AS2(    mov   edi, [ebp + 16]           ) // outBlock
+       AS2(    movd  edi, mm3                  ) // outBlock
     #else
         AS2(    mov   edi, [ebp + 12]           ) // outBlock
     #endif
@@ -493,7 +500,9 @@ void Twofish::AsmEncrypt(const byte* inBlock, byte* outBlock) const
 
 
 #ifdef _MSC_VER
-    __declspec(naked) 
+    __declspec(naked)
+#else
+    __attribute__ ((noinline))
 #endif
 void Twofish::AsmDecrypt(const byte* inBlock, byte* outBlock) const
 {
@@ -542,7 +551,7 @@ void Twofish::AsmDecrypt(const byte* inBlock, byte* outBlock) const
     AS2(    movd  ebp, mm6                      )
     AS2(    movd  esi, mm0                      ) // k_
     #ifdef __GNUC__
-        AS2(    mov   edi, [ebp + 16]           ) // outBlock
+        AS2(    movd   edi, mm3                 ) // outBlock
     #else
         AS2(    mov   edi, [ebp + 12]           ) // outBlock
     #endif