Index: sys/ofed/drivers/net/mlx4/en_netdev.c
===================================================================
--- sys/ofed/drivers/net/mlx4/en_netdev.c
+++ sys/ofed/drivers/net/mlx4/en_netdev.c
@@ -1617,6 +1617,7 @@
 		priv->tx_ring[i]->bytes = 0;
 		priv->tx_ring[i]->packets = 0;
 		priv->tx_ring[i]->tx_csum = 0;
+		priv->tx_ring[i]->oversized_packets = 0;
 	}
 	for (i = 0; i < priv->rx_ring_num; i++) {
 		priv->rx_ring[i]->bytes = 0;
@@ -1803,8 +1804,11 @@
                 EVENTHANDLER_DEREGISTER(vlan_unconfig, priv->vlan_detach);
 
 	/* Unregister device - this will close the port if it was up */
-	if (priv->registered)
+	if (priv->registered) {
+		mutex_lock(&mdev->state_lock);
 		ether_ifdetach(dev);
+		mutex_unlock(&mdev->state_lock);
+	}
 
 	if (priv->allocated)
 		mlx4_free_hwq_res(mdev->dev, &priv->res, MLX4_EN_PAGE_SIZE);
@@ -2012,22 +2016,22 @@
 		error = -mlx4_en_change_mtu(dev, ifr->ifr_mtu);
 		break;
 	case SIOCSIFFLAGS:
+		mutex_lock(&mdev->state_lock);
 		if (dev->if_flags & IFF_UP) {
 			if ((dev->if_drv_flags & IFF_DRV_RUNNING) == 0) {
-				mutex_lock(&mdev->state_lock);
 				mlx4_en_start_port(dev);
-				mutex_unlock(&mdev->state_lock);
 			} else {
+				mutex_unlock(&mdev->state_lock);
 				mlx4_en_set_rx_mode(dev);
+				break;
 			}
 		} else {
-			mutex_lock(&mdev->state_lock);
 			if (dev->if_drv_flags & IFF_DRV_RUNNING) {
 				mlx4_en_stop_port(dev);
 				if_link_state_change(dev, LINK_STATE_DOWN);
 			}
-			mutex_unlock(&mdev->state_lock);
 		}
+		mutex_unlock(&mdev->state_lock);
 		break;
 	case SIOCADDMULTI:
 	case SIOCDELMULTI:
@@ -2040,14 +2044,58 @@
 	case SIOCSIFCAP:
 		mutex_lock(&mdev->state_lock);
 		mask = ifr->ifr_reqcap ^ dev->if_capenable;
-		if (mask & IFCAP_HWCSUM)
-			dev->if_capenable ^= IFCAP_HWCSUM;
-		if (mask & IFCAP_TSO4)
+		if (mask & IFCAP_TXCSUM) {
+			dev->if_capenable ^= IFCAP_TXCSUM;
+			dev->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
+
+			if (IFCAP_TSO4 & dev->if_capenable &&
+			    !(IFCAP_TXCSUM & dev->if_capenable)) {
+				dev->if_capenable &= ~IFCAP_TSO4;
+				dev->if_hwassist &= ~CSUM_IP_TSO;
+				if_printf(dev,
+				    "tso4 disabled due to -txcsum.\n");
+			}
+		}
+		if (mask & IFCAP_TXCSUM_IPV6) {
+			dev->if_capenable ^= IFCAP_TXCSUM_IPV6;
+			dev->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
+
+			if (IFCAP_TSO6 & dev->if_capenable &&
+			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
+				dev->if_capenable &= ~IFCAP_TSO6;
+				dev->if_hwassist &= ~CSUM_IP6_TSO;
+				if_printf(dev,
+				    "tso6 disabled due to -txcsum6.\n");
+			}
+		}
+		if (mask & IFCAP_RXCSUM)
+			dev->if_capenable ^= IFCAP_RXCSUM;
+		if (mask & IFCAP_RXCSUM_IPV6)
+			dev->if_capenable ^= IFCAP_RXCSUM_IPV6;
+
+		if (mask & IFCAP_TSO4) {
+			if (!(IFCAP_TSO4 & dev->if_capenable) &&
+			    !(IFCAP_TXCSUM & dev->if_capenable)) {
+				if_printf(dev, "enable txcsum first.\n");
+				error = EAGAIN;
+				goto out;
+			}
 			dev->if_capenable ^= IFCAP_TSO4;
-		if (mask & IFCAP_TSO6)
+			dev->if_hwassist ^= CSUM_IP_TSO;
+		}
+		if (mask & IFCAP_TSO6) {
+			if (!(IFCAP_TSO6 & dev->if_capenable) &&
+			    !(IFCAP_TXCSUM_IPV6 & dev->if_capenable)) {
+				if_printf(dev, "enable txcsum6 first.\n");
+				error = EAGAIN;
+				goto out;
+			}
 			dev->if_capenable ^= IFCAP_TSO6;
+			dev->if_hwassist ^= CSUM_IP6_TSO;
+		}
 		if (mask & IFCAP_LRO)
 			dev->if_capenable ^= IFCAP_LRO;
+
 		if (mask & IFCAP_VLAN_HWTAGGING)
 			dev->if_capenable ^= IFCAP_VLAN_HWTAGGING;
 		if (mask & IFCAP_VLAN_HWFILTER)
@@ -2056,6 +2104,7 @@
 			dev->if_capenable ^= IFCAP_WOL_MAGIC;
 		if (dev->if_drv_flags & IFF_DRV_RUNNING)
 			mlx4_en_start_port(dev);
+out:
 		mutex_unlock(&mdev->state_lock);
 		VLAN_CAPABILITIES(dev);
 		break;
@@ -2315,7 +2364,7 @@
 	/*
 	 * Set driver features
 	 */
-	dev->if_capabilities |= IFCAP_RXCSUM | IFCAP_TXCSUM;
+	dev->if_capabilities |= IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6;
 	dev->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING;
 	dev->if_capabilities |= IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWFILTER;
 	dev->if_capabilities |= IFCAP_LINKSTATE | IFCAP_JUMBO_MTU;
@@ -2340,6 +2389,8 @@
 		dev->if_hwassist |= CSUM_TSO;
 	if (dev->if_capenable & IFCAP_TXCSUM)
 		dev->if_hwassist |= (CSUM_TCP | CSUM_UDP | CSUM_IP);
+	if (dev->if_capenable & IFCAP_TXCSUM_IPV6)
+		dev->if_hwassist |= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
 
 
         /* Register for VLAN events */
@@ -3161,6 +3212,8 @@
 	    &priv->port_stats.wake_queue, "Queue resumed after full");
 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_timeout", CTLFLAG_RD,
 	    &priv->port_stats.tx_timeout, "Transmit timeouts");
+	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "tx_oversized_packets", CTLFLAG_RD,
+	    &priv->port_stats.oversized_packets, "TX oversized packets, m_defrag failed");
 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_alloc_failed", CTLFLAG_RD,
 	    &priv->port_stats.rx_alloc_failed, "RX failed to allocate mbuf");
 	SYSCTL_ADD_ULONG(ctx, node_list, OID_AUTO, "rx_chksum_good", CTLFLAG_RD,
Index: sys/ofed/drivers/net/mlx4/en_port.c
===================================================================
--- sys/ofed/drivers/net/mlx4/en_port.c
+++ sys/ofed/drivers/net/mlx4/en_port.c
@@ -198,6 +198,7 @@
 		priv->port_stats.tx_chksum_offload += priv->tx_ring[i]->tx_csum;
 		priv->port_stats.queue_stopped += priv->tx_ring[i]->queue_stopped;
 		priv->port_stats.wake_queue += priv->tx_ring[i]->wake_queue;
+		priv->port_stats.oversized_packets += priv->tx_ring[i]->oversized_packets;
 	}
 	/* RX Statistics */
 	priv->pkstats.rx_packets = be64_to_cpu(mlx4_en_stats->RTOT_prio_0) +
Index: sys/ofed/drivers/net/mlx4/en_rx.c
===================================================================
--- sys/ofed/drivers/net/mlx4/en_rx.c
+++ sys/ofed/drivers/net/mlx4/en_rx.c
@@ -611,7 +611,8 @@
 			mb->m_pkthdr.ether_vtag = be16_to_cpu(cqe->sl_vid);
 			mb->m_flags |= M_VLANTAG;
 		}
-		if (likely(dev->if_capabilities & IFCAP_RXCSUM) &&
+		if (likely(dev->if_capenable &
+		    (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6)) &&
 		    (cqe->status & cpu_to_be16(MLX4_CQE_STATUS_IPOK)) &&
 		    (cqe->checksum == cpu_to_be16(0xffff))) {
 			priv->port_stats.rx_chksum_good++;
@@ -692,6 +693,7 @@
         // Because there is no NAPI in freeBSD
         done = mlx4_en_poll_rx_cq(cq, MLX4_EN_RX_BUDGET);
 	if (priv->port_up  && (done == MLX4_EN_RX_BUDGET) ) {
+		cq->curr_poll_rx_cpu_id = curcpu;
 		taskqueue_enqueue(cq->tq, &cq->cq_task);
         }
 	else {
@@ -702,8 +704,15 @@
 void mlx4_en_rx_que(void *context, int pending)
 {
         struct mlx4_en_cq *cq;
+	struct thread *td;
 
         cq = context;
+	td = curthread;
+
+	thread_lock(td);
+	sched_bind(td, cq->curr_poll_rx_cpu_id);
+	thread_unlock(td);
+
         while (mlx4_en_poll_rx_cq(cq, MLX4_EN_RX_BUDGET)
                         == MLX4_EN_RX_BUDGET);
         mlx4_en_arm_cq(cq->dev->if_softc, cq);
Index: sys/ofed/drivers/net/mlx4/en_tx.c
===================================================================
--- sys/ofed/drivers/net/mlx4/en_tx.c
+++ sys/ofed/drivers/net/mlx4/en_tx.c
@@ -813,14 +813,6 @@
 	/* Optimize the common case when there are no wraparounds */
 	if (likely((void *) tx_desc + tx_info->nr_txbb * TXBB_SIZE <= end)) {
 		if (!tx_info->inl) {
-			if (tx_info->linear) {
-				dma_unmap_single(priv->ddev,
-					(dma_addr_t) be64_to_cpu(data->addr),
-					 be32_to_cpu(data->byte_count),
-					 PCI_DMA_TODEVICE);
-				++data;
-			}
-
 			for (i = 0; i < frags; i++) {
                                 pci_unmap_single(mdev->pdev,
                                                 (dma_addr_t) be64_to_cpu(data[i].addr),
@@ -832,15 +824,6 @@
 			if ((void *) data >= end) {
 				data = ring->buf + ((void *)data - end);
 			}
-
-			if (tx_info->linear) {
-				dma_unmap_single(priv->ddev,
-					(dma_addr_t) be64_to_cpu(data->addr),
-					 be32_to_cpu(data->byte_count),
-					 PCI_DMA_TODEVICE);
-				++data;
-			}
-
 			for (i = 0; i < frags; i++) {
 				/* Check for wraparound before unmapping */
 				if ((void *) data >= end)
@@ -1271,11 +1254,7 @@
 	        up = (vlan_tag >> 13) % MLX4_EN_NUM_UP;
 	}
 #endif
-	/* check if flowid is set */
-	if (M_HASHTYPE_GET(mb) != M_HASHTYPE_NONE)
-		queue_index = mb->m_pkthdr.flowid;
-	else
-		queue_index = m_ether_tcpip_hash(MBUF_HASHFLAG_L3 | MBUF_HASHFLAG_L4, mb, hashrandom);
+	queue_index = m_ether_tcpip_hash(MBUF_HASHFLAG_L3 | MBUF_HASHFLAG_L4, mb, hashrandom);
 
 	return ((queue_index % rings_p_up) + (up * rings_p_up));
 }
@@ -1351,13 +1330,15 @@
                         mb = m_defrag(*mbp, M_NOWAIT);
                         if (mb == NULL) {
                                 mb = *mbp;
+				ring->oversized_packets++;
                                 goto tx_drop;
                         }
                         *mbp = mb;
                         defrag = 0;
                         goto retry;
                 }
-		en_warn(priv, "Oversized header or SG list\n");
+		/*Oversized header or SG list*/
+		ring->oversized_packets++;
 		goto tx_drop;
 	}
 
@@ -1622,13 +1603,15 @@
 	}
 
 	enqueued = 0;
-	if (m != NULL) {
-		if ((err = drbr_enqueue(dev, ring->br, m)) != 0)
-			return (err);
-	}
+	if (m != NULL)
+		/* If we can't insert mbuf into drbr, try to xmit anyway.
+		 * We keep the error we got so we could return that after xmit.
+		 */
+		err = drbr_enqueue(dev, ring->br, m);
+
 	/* Process the queue */
 	while ((next = drbr_peek(dev, ring->br)) != NULL) {
-		if ((err = mlx4_en_xmit(dev, tx_ind, &next)) != 0) {
+		if (mlx4_en_xmit(dev, tx_ind, &next) != 0) {
 			if (next == NULL) {
 				drbr_advance(dev, ring->br);
 			} else {
@@ -1694,7 +1677,12 @@
 	else
 #endif
 	/* Compute which queue to use */
-	i = mlx4_en_select_queue(dev, m);
+	if (M_HASHTYPE_GET(m) != M_HASHTYPE_NONE) {
+		i = m->m_pkthdr.flowid % priv->tx_ring_num;
+	}
+	else {
+		i = mlx4_en_select_queue(dev, m);
+	}
 
 #ifdef CONFIG_RATELIMIT
 lock_and_transmit:
Index: sys/ofed/drivers/net/mlx4/mlx4_en.h
===================================================================
--- sys/ofed/drivers/net/mlx4/mlx4_en.h
+++ sys/ofed/drivers/net/mlx4/mlx4_en.h
@@ -260,7 +260,6 @@
         struct mbuf *mb;
         u32 nr_txbb;
 	u32 nr_bytes;
-        u8 linear;
         u8 nr_segs;
         u8 data_offset;
         u8 inl;
@@ -334,6 +333,7 @@
 	unsigned long packets;
 	unsigned long tx_csum;
 	unsigned long queue_stopped;
+	unsigned long oversized_packets;
 	unsigned long wake_queue;
 	struct mlx4_bf bf;
 	bool bf_enabled;
@@ -437,6 +437,7 @@
 #define MLX4_EN_OPCODE_ERROR	0x1e
 	u32 tot_rx;
 	u32 tot_tx;
+	u32 curr_poll_rx_cpu_id;
 
 #ifdef CONFIG_NET_RX_BUSY_POLL
 	unsigned int state;
Index: sys/ofed/drivers/net/mlx4/mlx4_stats.h
===================================================================
--- sys/ofed/drivers/net/mlx4/mlx4_stats.h
+++ sys/ofed/drivers/net/mlx4/mlx4_stats.h
@@ -124,6 +124,7 @@
 	unsigned long queue_stopped;
 	unsigned long wake_queue;
 	unsigned long tx_timeout;
+	unsigned long oversized_packets;
 	unsigned long rx_alloc_failed;
 	unsigned long rx_chksum_good;
 	unsigned long rx_chksum_none;