crypto/modes: even more strict aliasing fixes [and fix bug in cbc128.c from

previous cbc128.c commit].

4 files changed, 51 insertions, 48 deletions

diff --git a/crypto/modes/cbc128.c b/crypto/modes/cbc128.c
index 34b2a2ef21..0e54f75470 100644
--- a/crypto/modes/cbc128.c
+++ b/crypto/modes/cbc128.c
@@ -137,11 +137,13 @@ void CRYPTO_cbc128_decrypt(const unsigned char *in, unsigned char *out,
 				out += 16;
 			}
 		}
-		else {
+		else  if (16%sizeof(size_t) == 0) { /* always true */
 			while (len>=16) {
+				size_t *out_t=(size_t *)out, *iv_t=(size_t *)iv;
+
 				(*block)(in, out, key);
-				for(n=0; n<16; n+=sizeof(size_t))
-					*(size_t *)(out+n) ^= *(size_t *)(iv+n);
+				for(n=0; n<16/sizeof(size_t); n++)
+					out_t[n] ^= iv_t[n];
 				iv = in;
 				len -= 16;
 				in  += 16;
@@ -166,18 +168,19 @@ void CRYPTO_cbc128_decrypt(const unsigned char *in, unsigned char *out,
 			}
 		}
 		else if (16%sizeof(size_t) == 0) { /* always true */
-			size_t c, *out_t=(size_t *)out, *ivec_t=(size_t *)ivec;
-			const size_t *in_t=(const size_t *)in;
 			while (len>=16) {
-				(*block)((const unsigned char *)in_t, tmp.c, key);
+				size_t c, *out_t=(size_t *)out, *ivec_t=(size_t *)ivec;
+				const size_t *in_t=(const size_t *)in;
+
+				(*block)(in, tmp.c, key);
 				for(n=0; n<16/sizeof(size_t); n++) {
 					c = in_t[n];
 					out_t[n] = tmp.t[n] ^ ivec_t[n];
 					ivec_t[n] = c;
 				}
 				len -= 16;
-				in_t  += 16/sizeof(size_t);
-				out_t += 16/sizeof(size_t);
+				in  += 16;
+				out += 16;
 			}
 		}
 	}
diff --git a/crypto/modes/ccm128.c b/crypto/modes/ccm128.c
index c9b35e5b35..3ce11d0d98 100644
--- a/crypto/modes/ccm128.c
+++ b/crypto/modes/ccm128.c
@@ -87,7 +87,7 @@ int CRYPTO_ccm128_setiv(CCM128_CONTEXT *ctx,
 		ctx->nonce.c[11] = (u8)(mlen>>(32%(sizeof(mlen)*8)));
 	}
 	else
-		*(u32*)(&ctx->nonce.c[8]) = 0;
+		ctx->nonce.u[1] = 0;
 
 	ctx->nonce.c[12] = (u8)(mlen>>24);
 	ctx->nonce.c[13] = (u8)(mlen>>16);
diff --git a/crypto/modes/cts128.c b/crypto/modes/cts128.c
index c0e1f3696c..2d583de6f6 100644
--- a/crypto/modes/cts128.c
+++ b/crypto/modes/cts128.c
@@ -108,12 +108,8 @@ size_t CRYPTO_cts128_encrypt(const unsigned char *in, unsigned char *out,
 	(*cbc)(in,out-16,residue,key,ivec,1);
 	memcpy(out,tmp.c,residue);
 #else
-	{
-	size_t n;
-	for (n=0; n<16; n+=sizeof(size_t))
-		*(size_t *)(tmp.c+n) = 0;
+	memset(tmp.c,0,sizeof(tmp));
 	memcpy(tmp.c,in,residue);
-	}
 	memcpy(out,out-16,residue);
 	(*cbc)(tmp.c,out-16,16,key,ivec,1);
 #endif
@@ -144,12 +140,8 @@ size_t CRYPTO_nistcts128_encrypt(const unsigned char *in, unsigned char *out,
 #if defined(CBC_HANDLES_TRUNCATED_IO)
 	(*cbc)(in,out-16+residue,residue,key,ivec,1);
 #else
-	{
-	size_t n;
-	for (n=0; n<16; n+=sizeof(size_t))
-		*(size_t *)(tmp.c+n) = 0;
+	memset(tmp.c,0,sizeof(tmp));
 	memcpy(tmp.c,in,residue);
-	}
 	(*cbc)(tmp.c,out-16+residue,16,key,ivec,1);
 #endif
 	return len+residue;
@@ -177,8 +169,7 @@ size_t CRYPTO_cts128_decrypt_block(const unsigned char *in, unsigned char *out,
 
 	(*block)(in,tmp.c+16,key);
 
-	for (n=0; n<16; n+=sizeof(size_t))
-		*(size_t *)(tmp.c+n) = *(size_t *)(tmp.c+16+n);
+	memcpy(tmp.c,tmp.c+16,16);
 	memcpy(tmp.c,in+16,residue);
 	(*block)(tmp.c,tmp.c,key);
 
@@ -220,8 +211,7 @@ size_t CRYPTO_nistcts128_decrypt_block(const unsigned char *in, unsigned char *o
 
 	(*block)(in+residue,tmp.c+16,key);
 
-	for (n=0; n<16; n+=sizeof(size_t))
-		*(size_t *)(tmp.c+n) = *(size_t *)(tmp.c+16+n);
+	memcpy(tmp.c,tmp.c+16,16);
 	memcpy(tmp.c,in,residue);
 	(*block)(tmp.c,tmp.c,key);
 
@@ -240,7 +230,7 @@ size_t CRYPTO_nistcts128_decrypt_block(const unsigned char *in, unsigned char *o
 size_t CRYPTO_cts128_decrypt(const unsigned char *in, unsigned char *out,
 			size_t len, const void *key,
 			unsigned char ivec[16], cbc128_f cbc)
-{	size_t residue, n;
+{	size_t residue;
 	union { size_t align; unsigned char c[32]; } tmp;
 
 	assert (in && out && key && ivec);
@@ -257,8 +247,7 @@ size_t CRYPTO_cts128_decrypt(const unsigned char *in, unsigned char *out,
 		out += len;
 	}
 
-	for (n=16; n<32; n+=sizeof(size_t))
-		*(size_t *)(tmp.c+n) = 0;
+	memset(tmp.c,0,sizeof(tmp));
 	/* this places in[16] at &tmp.c[16] and decrypted block at &tmp.c[0] */
 	(*cbc)(in,tmp.c,16,key,tmp.c+16,0);
 
@@ -275,7 +264,7 @@ size_t CRYPTO_cts128_decrypt(const unsigned char *in, unsigned char *out,
 size_t CRYPTO_nistcts128_decrypt(const unsigned char *in, unsigned char *out,
 			size_t len, const void *key,
 			unsigned char ivec[16], cbc128_f cbc)
-{	size_t residue, n;
+{	size_t residue;
 	union { size_t align; unsigned char c[32]; } tmp;
 
 	assert (in && out && key && ivec);
@@ -297,8 +286,7 @@ size_t CRYPTO_nistcts128_decrypt(const unsigned char *in, unsigned char *out,
 		out += len;
 	}
 
-	for (n=16; n<32; n+=sizeof(size_t))
-		*(size_t *)(tmp.c+n) = 0;
+	memset(tmp.c,0,sizeof(tmp));
 	/* this places in[16] at &tmp.c[16] and decrypted block at &tmp.c[0] */
 	(*cbc)(in+residue,tmp.c,16,key,tmp.c+16,0);
 
diff --git a/crypto/modes/gcm128.c b/crypto/modes/gcm128.c
index 8f5ce98858..e7d1736786 100644
--- a/crypto/modes/gcm128.c
+++ b/crypto/modes/gcm128.c
@@ -962,8 +962,9 @@ int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx,
 		    size_t j=GHASH_CHUNK;
 
 		    while (j) {
-			size_t *out_t=(size_t *)out, *ivec_t=(size_t *)ivec;
-			const size_t *in_t=(const size_t *)in;
+		    	size_t *out_t=(size_t *)out;
+		    	const size_t *in_t=(const size_t *)in;
+
 			(*block)(ctx->Yi.c,ctx->EKi.c,key);
 			++ctr;
 			if (is_endian.little)
@@ -983,15 +984,17 @@ int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx,
 		    size_t j=i;
 
 		    while (len>=16) {
+		    	size_t *out_t=(size_t *)out;
+		    	const size_t *in_t=(const size_t *)in;
+
 			(*block)(ctx->Yi.c,ctx->EKi.c,key);
 			++ctr;
 			if (is_endian.little)
 				PUTU32(ctx->Yi.c+12,ctr);
 			else
 				ctx->Yi.d[3] = ctr;
-			for (i=0; i<16; i+=sizeof(size_t))
-				*(size_t *)(out+i) =
-				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
+			for (i=0; i<16/sizeof(size_t); ++i)
+				out_t[i] = in_t[i] ^ ctx->EKi.t[i];
 			out += 16;
 			in  += 16;
 			len -= 16;
@@ -1000,16 +1003,18 @@ int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx,
 		}
 #else
 		while (len>=16) {
+		    	size_t *out_t=(size_t *)out;
+		    	const size_t *in_t=(const size_t *)in;
+
 			(*block)(ctx->Yi.c,ctx->EKi.c,key);
 			++ctr;
 			if (is_endian.little)
 				PUTU32(ctx->Yi.c+12,ctr);
 			else
 				ctx->Yi.d[3] = ctr;
-			for (i=0; i<16; i+=sizeof(size_t))
-				*(size_t *)(ctx->Xi.c+i) ^=
-				*(size_t *)(out+i) =
-				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
+			for (i=0; i<16/sizeof(size_t); ++i)
+				ctx->Xi.t[i] ^=
+				out_t[i] = in_t[i]^ctx->EKi.t[i];
 			GCM_MUL(ctx,Xi);
 			out += 16;
 			in  += 16;
@@ -1113,15 +1118,17 @@ int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx,
 
 		    GHASH(ctx,in,GHASH_CHUNK);
 		    while (j) {
+		    	size_t *out_t=(size_t *)out;
+		    	const size_t *in_t=(const size_t *)in;
+
 			(*block)(ctx->Yi.c,ctx->EKi.c,key);
 			++ctr;
 			if (is_endian.little)
 				PUTU32(ctx->Yi.c+12,ctr);
 			else
 				ctx->Yi.d[3] = ctr;
-			for (i=0; i<16; i+=sizeof(size_t))
-				*(size_t *)(out+i) =
-				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
+			for (i=0; i<16/sizeof(size_t); ++i)
+				out_t[i] = in_t[i]^ctx->EKi.t[i];
 			out += 16;
 			in  += 16;
 			j   -= 16;
@@ -1131,15 +1138,17 @@ int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx,
 		if ((i = (len&(size_t)-16))) {
 		    GHASH(ctx,in,i);
 		    while (len>=16) {
+		    	size_t *out_t=(size_t *)out;
+		    	const size_t *in_t=(const size_t *)in;
+
 			(*block)(ctx->Yi.c,ctx->EKi.c,key);
 			++ctr;
 			if (is_endian.little)
 				PUTU32(ctx->Yi.c+12,ctr);
 			else
 				ctx->Yi.d[3] = ctr;
-			for (i=0; i<16; i+=sizeof(size_t))
-				*(size_t *)(out+i) =
-				*(size_t *)(in+i)^*(size_t *)(ctx->EKi.c+i);
+			for (i=0; i<16/sizeof(size_t); ++i)
+				out_t[i] = in_t[i]^ctx->EKi.t[i];
 			out += 16;
 			in  += 16;
 			len -= 16;
@@ -1147,16 +1156,19 @@ int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx,
 		}
 #else
 		while (len>=16) {
+		    	size_t *out_t=(size_t *)out;
+		    	const size_t *in_t=(const size_t *)in;
+
 			(*block)(ctx->Yi.c,ctx->EKi.c,key);
 			++ctr;
 			if (is_endian.little)
 				PUTU32(ctx->Yi.c+12,ctr);
 			else
 				ctx->Yi.d[3] = ctr;
-			for (i=0; i<16; i+=sizeof(size_t)) {
-				size_t c = *(size_t *)(in+i);
-				*(size_t *)(out+i) = c^*(size_t *)(ctx->EKi.c+i);
-				*(size_t *)(ctx->Xi.c+i) ^= c;
+			for (i=0; i<16/sizeof(size_t); ++i) {
+				size_t c = in[i];
+				out[i] = c^ctx->EKi.t[i];
+				ctx->Xi.t[i] ^= c;
 			}
 			GCM_MUL(ctx,Xi);
 			out += 16;