diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 4d61e32..bcf5064 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -90,7 +90,8 @@
 	props->device_cap_flags    = IB_DEVICE_CHANGE_PHY_PORT |
 		IB_DEVICE_PORT_ACTIVE_EVENT		|
 		IB_DEVICE_SYS_IMAGE_GUID		|
-		IB_DEVICE_RC_RNR_NAK_GEN;
+		IB_DEVICE_RC_RNR_NAK_GEN		|
+		IB_DEVICE_BLOCK_MULTICAST_LOOPBACK;
 	if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_PKEY_CNTR)
 		props->device_cap_flags |= IB_DEVICE_BAD_PKEY_CNTR;
 	if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_QKEY_CNTR)
@@ -437,7 +438,9 @@
 static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 {
 	return mlx4_multicast_attach(to_mdev(ibqp->device)->dev,
-				     &to_mqp(ibqp)->mqp, gid->raw);
+				     &to_mqp(ibqp)->mqp, gid->raw,
+				     !!(to_mqp(ibqp)->flags &
+					MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK));
 }
 
 static int mlx4_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index 5cf9947..c4cf5b6 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -101,7 +101,8 @@
 };
 
 enum mlx4_ib_qp_flags {
-	MLX4_IB_QP_LSO		= 1 << 0
+	MLX4_IB_QP_LSO				= 1 << 0,
+	MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK	= 1 << 1,
 };
 
 struct mlx4_ib_qp {
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 44bbd6c..91590e7 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -511,6 +511,9 @@
 	} else {
 		qp->sq_no_prefetch = 0;
 
+		if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK)
+			qp->flags |= MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK;
+
 		if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO)
 			qp->flags |= MLX4_IB_QP_LSO;
 
@@ -684,10 +687,15 @@
 	struct mlx4_ib_qp *qp;
 	int err;
 
-	/* We only support LSO, and only for kernel UD QPs. */
-	if (init_attr->create_flags & ~IB_QP_CREATE_IPOIB_UD_LSO)
+	/*
+	 * We only support LSO and multicast loopback blocking, and
+	 * only for kernel UD QPs.
+	 */
+	if (init_attr->create_flags & ~(IB_QP_CREATE_IPOIB_UD_LSO |
+					IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK))
 		return ERR_PTR(-EINVAL);
-	if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO &&
+
+	if (init_attr->create_flags &&
 	    (pd->uobject || init_attr->qp_type != IB_QPT_UD))
 		return ERR_PTR(-EINVAL);
 
@@ -1844,6 +1852,13 @@
 
 	qp_init_attr->cap	     = qp_attr->cap;
 
+	qp_init_attr->create_flags = 0;
+	if (qp->flags & MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK)
+		qp_init_attr->create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK;
+
+	if (qp->flags & MLX4_IB_QP_LSO)
+		qp_init_attr->create_flags |= IB_QP_CREATE_IPOIB_UD_LSO;
+
 out:
 	mutex_unlock(&qp->mutex);
 	return err;
diff --git a/drivers/net/mlx4/mcg.c b/drivers/net/mlx4/mcg.c
index 57f7f1f..b4b5787 100644
--- a/drivers/net/mlx4/mcg.c
+++ b/drivers/net/mlx4/mcg.c
@@ -38,6 +38,9 @@
 
 #include "mlx4.h"
 
+#define MGM_QPN_MASK       0x00FFFFFF
+#define MGM_BLCK_LB_BIT    30
+
 struct mlx4_mgm {
 	__be32			next_gid_index;
 	__be32			members_count;
@@ -153,7 +156,8 @@
 	return err;
 }
 
-int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
+int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
+			  int block_mcast_loopback)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct mlx4_cmd_mailbox *mailbox;
@@ -202,13 +206,18 @@
 	}
 
 	for (i = 0; i < members_count; ++i)
-		if (mgm->qp[i] == cpu_to_be32(qp->qpn)) {
+		if ((be32_to_cpu(mgm->qp[i]) & MGM_QPN_MASK) == qp->qpn) {
 			mlx4_dbg(dev, "QP %06x already a member of MGM\n", qp->qpn);
 			err = 0;
 			goto out;
 		}
 
-	mgm->qp[members_count++] = cpu_to_be32(qp->qpn);
+	if (block_mcast_loopback)
+		mgm->qp[members_count++] = cpu_to_be32((qp->qpn & MGM_QPN_MASK) |
+						       (1 << MGM_BLCK_LB_BIT));
+	else
+		mgm->qp[members_count++] = cpu_to_be32(qp->qpn & MGM_QPN_MASK);
+
 	mgm->members_count       = cpu_to_be32(members_count);
 
 	err = mlx4_WRITE_MCG(dev, index, mailbox);
@@ -283,7 +292,7 @@
 
 	members_count = be32_to_cpu(mgm->members_count);
 	for (loc = -1, i = 0; i < members_count; ++i)
-		if (mgm->qp[i] == cpu_to_be32(qp->qpn))
+		if ((be32_to_cpu(mgm->qp[i]) & MGM_QPN_MASK) == qp->qpn)
 			loc = i;
 
 	if (loc == -1) {
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index a744383..81b3dd5 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -398,7 +398,8 @@
 int mlx4_INIT_PORT(struct mlx4_dev *dev, int port);
 int mlx4_CLOSE_PORT(struct mlx4_dev *dev, int port);
 
-int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16]);
+int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
+			  int block_mcast_loopback);
 int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16]);
 
 int mlx4_map_phys_fmr(struct mlx4_dev *dev, struct mlx4_fmr *fmr, u64 *page_list,