Fix Solaris 11/Sparc crash in SHA-384 (GH #689, GH #403)

I believe Andrew Marlow first reported it. At the time we could not get our hands on hardware to fully test things. Instead we were using -xmemalign=4i option as a band-aide to avoid running afoul of the Sparc instruction that moves 64-bits of data in one shot.

GNUmakefile

@@ -566,10 +566,6 @@ CXXFLAGS += -KPIC
 endif
 # Add to all Solaris
 CXXFLAGS += -template=no%extdef
-# http://github.com/weidai11/cryptopp/issues/403
-ifneq ($(IS_SPARC32)$(IS_SPARC64),00)
-CXXFLAGS += -xmemalign=4i
-endif
 SUN_CC10_BUGGY := $(shell $(CXX) -V 2>&1 | $(GREP) -c -E "CC: Sun .* 5\.10 .* (2009|2010/0[1-4])")
 ifneq ($(SUN_CC10_BUGGY),0)
 # -DCRYPTOPP_INCLUDE_VECTOR_CC is needed for Sun Studio 12u1 Sun C++ 5.10 SunOS_i386 128229-02 2009/09/21 and was fixed in May 2010

sha.cpp

@@ -1166,7 +1166,7 @@ ANONYMOUS_NAMESPACE_BEGIN
 #define g(i) T[(6-i)&7]
 #define h(i) T[(7-i)&7]
 
-#define blk0(i) (W[i]=data[i])
+#define blk0(i) (W[i]=D[i])
 #define blk2(i) (W[i&15]+=s1(W[(i-2)&15])+W[(i-7)&15]+s0(W[(i-15)&15]))
 
 #define Ch(x,y,z) (z^(x&(y^z)))
@@ -1185,9 +1185,14 @@ void SHA512_HashBlock_CXX(word64 *state, const word64 *data)
     CRYPTOPP_ASSERT(state);
     CRYPTOPP_ASSERT(data);
 
-    word64 W[16]={0}, T[8];
+    word64 W[16]={0}, D[16], T[8];
+
     /* Copy context->state[] to working vars */
-    memcpy(T, state, sizeof(T));
+    std::memcpy(T, state, sizeof(T));
+    
+    /* Solaris/Sparc64 crash */
+    std::memcpy(D, data, sizeof(D));
+
     /* 80 operations, partially loop unrolled */
     for (unsigned int j=0; j<80; j+=16)
     {
@@ -1196,6 +1201,10 @@ void SHA512_HashBlock_CXX(word64 *state, const word64 *data)
         R( 8); R( 9); R(10); R(11);
         R(12); R(13); R(14); R(15);
     }
+
+    /* Solaris 11/Sparc64 crash */
+    if (IsAligned<word64>(state) == true)
+    {
         /* Add the working vars back into context.state[] */
         state[0] += a(0);
         state[1] += b(0);
@@ -1206,29 +1215,24 @@ void SHA512_HashBlock_CXX(word64 *state, const word64 *data)
         state[6] += g(0);
         state[7] += h(0);
     }
+    else
+    {
+        /* Reuse W[] */
+        std::memcpy(W, state, 8 * sizeof(W[0]));
 
-#undef Ch
+        /* Add the working vars back into context.state[] */
-#undef Maj
+        W[0] += a(0);
+        W[1] += b(0);
+        W[2] += c(0);
+        W[3] += d(0);
+        W[4] += e(0);
+        W[5] += f(0);
+        W[6] += g(0);
+        W[7] += h(0);
 
-#undef s0
+        std::memcpy(state, W, 8 * sizeof(W[0]));
-#undef s1
+    }
-#undef S0
+}
-#undef S1
-
-#undef blk0
-#undef blk1
-#undef blk2
-
-#undef R
-
-#undef a
-#undef b
-#undef c
-#undef d
-#undef e
-#undef f
-#undef g
-#undef h
 
 ANONYMOUS_NAMESPACE_END
 
@@ -1255,6 +1259,29 @@ void SHA512::Transform(word64 *state, const word64 *data)
     SHA512_HashBlock_CXX(state, data);
 }
 
+#undef Ch
+#undef Maj
+
+#undef s0
+#undef s1
+#undef S0
+#undef S1
+
+#undef blk0
+#undef blk1
+#undef blk2
+
+#undef R
+
+#undef a
+#undef b
+#undef c
+#undef d
+#undef e
+#undef f
+#undef g
+#undef h
+
 NAMESPACE_END
 
 #endif    // Not CRYPTOPP_GENERATE_X64_MASM

tweetnacl.cpp

@@ -759,7 +759,7 @@ int crypto_sign_sk2pk(byte *pk, const byte *sk)
 {
   byte d[64];
   gf p[4];
-  int i;
+  // int i;
 
   // randombytes(sk, 32);
   crypto_hash(d, sk, 32);