Fix for CVE-2014-0076

Fix for the attack described in the paper "Recovering OpenSSL
ECDSA Nonces Using the FLUSH+RELOAD Cache Side-channel Attack"
by Yuval Yarom and Naomi Benger. Details can be obtained from:
http://eprint.iacr.org/2014/140

Thanks to Yuval Yarom and Naomi Benger for discovering this
flaw and to Yuval Yarom for supplying a fix.
(cherry picked from commit 2198be3483259de374f91e57d247d0fc667aef29)

Conflicts:

	CHANGES

4 files changed, 88 insertions, 11 deletions

diff --git a/CHANGES b/CHANGES
index 58ac884d80..99aeefb4c2 100644
--- a/CHANGES
+++ b/CHANGES
@@ -4,6 +4,15 @@
 
  Changes between 1.0.1f and 1.0.1g [xx XXX xxxx]
 
+  *) Fix for the attack described in the paper "Recovering OpenSSL
+     ECDSA Nonces Using the FLUSH+RELOAD Cache Side-channel Attack"
+     by Yuval Yarom and Naomi Benger. Details can be obtained from:
+     http://eprint.iacr.org/2014/140
+
+     Thanks to Yuval Yarom and Naomi Benger for discovering this
+     flaw and to Yuval Yarom for supplying a fix (CVE-2014-0076)
+     [Yuval Yarom and Naomi Benger]
+
   *) TLS pad extension: draft-agl-tls-padding-02
 
      Workaround for the "TLS hang bug" (see FAQ and PR#2771): if the
diff --git a/crypto/bn/bn.h b/crypto/bn/bn.h
index f34248ec4f..21a1a3fe35 100644
--- a/crypto/bn/bn.h
+++ b/crypto/bn/bn.h
@@ -538,6 +538,8 @@ BIGNUM *BN_mod_inverse(BIGNUM *ret,
 BIGNUM *BN_mod_sqrt(BIGNUM *ret,
 	const BIGNUM *a, const BIGNUM *n,BN_CTX *ctx);
 
+void	BN_consttime_swap(BN_ULONG swap, BIGNUM *a, BIGNUM *b, int nwords);
+
 /* Deprecated versions */
 #ifndef OPENSSL_NO_DEPRECATED
 BIGNUM *BN_generate_prime(BIGNUM *ret,int bits,int safe,
@@ -774,11 +776,20 @@ int RAND_pseudo_bytes(unsigned char *buf,int num);
 
 #define bn_fix_top(a)		bn_check_top(a)
 
+#define bn_check_size(bn, bits) bn_wcheck_size(bn, ((bits+BN_BITS2-1))/BN_BITS2)
+#define bn_wcheck_size(bn, words) \
+	do { \
+		const BIGNUM *_bnum2 = (bn); \
+		assert(words <= (_bnum2)->dmax && words >= (_bnum2)->top); \
+	} while(0)
+
 #else /* !BN_DEBUG */
 
 #define bn_pollute(a)
 #define bn_check_top(a)
 #define bn_fix_top(a)		bn_correct_top(a)
+#define bn_check_size(bn, bits)
+#define bn_wcheck_size(bn, words)
 
 #endif
 
diff --git a/crypto/bn/bn_lib.c b/crypto/bn/bn_lib.c
index 7a5676de69..5461e6ee7d 100644
--- a/crypto/bn/bn_lib.c
+++ b/crypto/bn/bn_lib.c
@@ -824,3 +824,55 @@ int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b,
 		}
 	return bn_cmp_words(a,b,cl);
 	}
+
+/* 
+ * Constant-time conditional swap of a and b.  
+ * a and b are swapped if condition is not 0.  The code assumes that at most one bit of condition is set.
+ * nwords is the number of words to swap.  The code assumes that at least nwords are allocated in both a and b,
+ * and that no more than nwords are used by either a or b.
+ * a and b cannot be the same number
+ */
+void BN_consttime_swap(BN_ULONG condition, BIGNUM *a, BIGNUM *b, int nwords)
+	{
+	BN_ULONG t;
+	int i;
+
+	bn_wcheck_size(a, nwords);
+	bn_wcheck_size(b, nwords);
+
+	assert(a != b);
+	assert((condition & (condition - 1)) == 0);
+	assert(sizeof(BN_ULONG) >= sizeof(int));
+
+	condition = ((condition - 1) >> (BN_BITS2 - 1)) - 1;
+
+	t = (a->top^b->top) & condition;
+	a->top ^= t;
+	b->top ^= t;
+
+#define BN_CONSTTIME_SWAP(ind) \
+	do { \
+		t = (a->d[ind] ^ b->d[ind]) & condition; \
+		a->d[ind] ^= t; \
+		b->d[ind] ^= t; \
+	} while (0)
+
+
+	switch (nwords) {
+	default:
+		for (i = 10; i < nwords; i++) 
+			BN_CONSTTIME_SWAP(i);
+		/* Fallthrough */
+	case 10: BN_CONSTTIME_SWAP(9); /* Fallthrough */
+	case 9: BN_CONSTTIME_SWAP(8); /* Fallthrough */
+	case 8: BN_CONSTTIME_SWAP(7); /* Fallthrough */
+	case 7: BN_CONSTTIME_SWAP(6); /* Fallthrough */
+	case 6: BN_CONSTTIME_SWAP(5); /* Fallthrough */
+	case 5: BN_CONSTTIME_SWAP(4); /* Fallthrough */
+	case 4: BN_CONSTTIME_SWAP(3); /* Fallthrough */
+	case 3: BN_CONSTTIME_SWAP(2); /* Fallthrough */
+	case 2: BN_CONSTTIME_SWAP(1); /* Fallthrough */
+	case 1: BN_CONSTTIME_SWAP(0);
+	}
+#undef BN_CONSTTIME_SWAP
+}
diff --git a/crypto/ec/ec2_mult.c b/crypto/ec/ec2_mult.c
index 26f4a783fc..1c575dc47a 100644
--- a/crypto/ec/ec2_mult.c
+++ b/crypto/ec/ec2_mult.c
@@ -208,11 +208,15 @@ static int gf2m_Mxy(const EC_GROUP *group, const BIGNUM *x, const BIGNUM *y, BIG
 	return ret;
 	}
 
+
 /* Computes scalar*point and stores the result in r.
  * point can not equal r.
- * Uses algorithm 2P of
+ * Uses a modified algorithm 2P of
  *     Lopez, J. and Dahab, R.  "Fast multiplication on elliptic curves over 
  *     GF(2^m) without precomputation" (CHES '99, LNCS 1717).
+ *
+ * To protect against side-channel attack the function uses constant time swap,
+ * avoiding conditional branches.
  */
 static int ec_GF2m_montgomery_point_multiply(const EC_GROUP *group, EC_POINT *r, const BIGNUM *scalar,
 	const EC_POINT *point, BN_CTX *ctx)
@@ -246,6 +250,11 @@ static int ec_GF2m_montgomery_point_multiply(const EC_GROUP *group, EC_POINT *r,
 	x2 = &r->X;
 	z2 = &r->Y;
 
+	bn_wexpand(x1, group->field.top);
+	bn_wexpand(z1, group->field.top);
+	bn_wexpand(x2, group->field.top);
+	bn_wexpand(z2, group->field.top);
+
 	if (!BN_GF2m_mod_arr(x1, &point->X, group->poly)) goto err; /* x1 = x */
 	if (!BN_one(z1)) goto err; /* z1 = 1 */
 	if (!group->meth->field_sqr(group, z2, x1, ctx)) goto err; /* z2 = x1^2 = x^2 */
@@ -270,16 +279,12 @@ static int ec_GF2m_montgomery_point_multiply(const EC_GROUP *group, EC_POINT *r,
 		word = scalar->d[i];
 		while (mask)
 			{
-			if (word & mask)
-				{
-				if (!gf2m_Madd(group, &point->X, x1, z1, x2, z2, ctx)) goto err;
-				if (!gf2m_Mdouble(group, x2, z2, ctx)) goto err;
-				}
-			else
-				{
-				if (!gf2m_Madd(group, &point->X, x2, z2, x1, z1, ctx)) goto err;
-				if (!gf2m_Mdouble(group, x1, z1, ctx)) goto err;
-				}
+			BN_consttime_swap(word & mask, x1, x2, group->field.top);
+			BN_consttime_swap(word & mask, z1, z2, group->field.top);
+			if (!gf2m_Madd(group, &point->X, x2, z2, x1, z1, ctx)) goto err;
+			if (!gf2m_Mdouble(group, x1, z1, ctx)) goto err;
+			BN_consttime_swap(word & mask, x1, x2, group->field.top);
+			BN_consttime_swap(word & mask, z1, z2, group->field.top);
 			mask >>= 1;
 			}
 		mask = BN_TBIT;