Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next

Pull networking changes from David S Miller:

 1) Remove the ipv4 routing cache.  Now lookups go directly into the FIB
    trie and use prebuilt routes cached there.

    No more garbage collection, no more rDOS attacks on the routing
    cache.  Instead we now get predictable and consistent performance,
    no matter what the pattern of traffic we service.

    This has been almost 2 years in the making.  Special thanks to
    Julian Anastasov, Eric Dumazet, Steffen Klassert, and others who
    have helped along the way.

    I'm sure that with a change of this magnitude there will be some
    kind of fallout, but such things ought the be simple to fix at this
    point.  Luckily I'm not European so I'll be around all of August to
    fix things :-)

    The major stages of this work here are each fronted by a forced
    merge commit whose commit message contains a top-level description
    of the motivations and implementation issues.

 2) Pre-demux of established ipv4 TCP sockets, saves a route demux on
    input.

 3) TCP SYN/ACK performance tweaks from Eric Dumazet.

 4) Add namespace support for netfilter L4 conntrack helpers, from Gao
    Feng.

 5) Add config mechanism for Energy Efficient Ethernet to ethtool, from
    Yuval Mintz.

 6) Remove quadratic behavior from /proc/net/unix, from Eric Dumazet.

 7) Support for connection tracker helpers in userspace, from Pablo
    Neira Ayuso.

 8) Allow userspace driven TX load balancing functions in TEAM driver,
    from Jiri Pirko.

 9) Kill off NLMSG_PUT and RTA_PUT macros, more gross stuff with
    embedded gotos.

10) TCP Small Queues, essentially minimize the amount of TCP data queued
    up in the packet scheduler layer.  Whereas the existing BQL (Byte
    Queue Limits) limits the pkt_sched --> netdevice queuing levels,
    this controls the TCP --> pkt_sched queueing levels.

    From Eric Dumazet.

11) Reduce the number of get_page/put_page ops done on SKB fragments,
    from Alexander Duyck.

12) Implement protection against blind resets in TCP (RFC 5961), from
    Eric Dumazet.

13) Support the client side of TCP Fast Open, basically the ability to
    send data in the SYN exchange, from Yuchung Cheng.

    Basically, the sender queues up data with a sendmsg() call using
    MSG_FASTOPEN, then they do the connect() which emits the queued up
    fastopen data.

14) Avoid all the problems we get into in TCP when timers or PMTU events
    hit a locked socket.  The TCP Small Queues changes added a
    tcp_release_cb() that allows us to queue work up to the
    release_sock() caller, and that's what we use here too.  From Eric
    Dumazet.

15) Zero copy on TX support for TUN driver, from Michael S. Tsirkin.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1870 commits)
  genetlink: define lockdep_genl_is_held() when CONFIG_LOCKDEP
  r8169: revert "add byte queue limit support".
  ipv4: Change rt->rt_iif encoding.
  net: Make skb->skb_iif always track skb->dev
  ipv4: Prepare for change of rt->rt_iif encoding.
  ipv4: Remove all RTCF_DIRECTSRC handliing.
  ipv4: Really ignore ICMP address requests/replies.
  decnet: Don't set RTCF_DIRECTSRC.
  net/ipv4/ip_vti.c: Fix __rcu warnings detected by sparse.
  ipv4: Remove redundant assignment
  rds: set correct msg_namelen
  openvswitch: potential NULL deref in sample()
  tcp: dont drop MTU reduction indications
  bnx2x: Add new 57840 device IDs
  tcp: avoid oops in tcp_metrics and reset tcpm_stamp
  niu: Change niu_rbr_fill() to use unlikely() to check niu_rbr_add_page() return value
  niu: Fix to check for dma mapping errors.
  net: Fix references to out-of-scope variables in put_cmsg_compat()
  net: ethernet: davinci_emac: add pm_runtime support
  net: ethernet: davinci_emac: Remove unnecessary #include
  ...

1 files changed, 173 insertions, 79 deletions

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
index 8098eea9704d..e879e19eb0d6 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
@@ -40,12 +40,19 @@
  * Makes sure the contents of the bp->fp[to].napi is kept
  * intact. This is done by first copying the napi struct from
  * the target to the source, and then mem copying the entire
- * source onto the target
+ * source onto the target. Update txdata pointers and related
+ * content.
  */
 static inline void bnx2x_move_fp(struct bnx2x *bp, int from, int to)
 {
 	struct bnx2x_fastpath *from_fp = &bp->fp[from];
 	struct bnx2x_fastpath *to_fp = &bp->fp[to];
+	struct bnx2x_sp_objs *from_sp_objs = &bp->sp_objs[from];
+	struct bnx2x_sp_objs *to_sp_objs = &bp->sp_objs[to];
+	struct bnx2x_fp_stats *from_fp_stats = &bp->fp_stats[from];
+	struct bnx2x_fp_stats *to_fp_stats = &bp->fp_stats[to];
+	int old_max_eth_txqs, new_max_eth_txqs;
+	int old_txdata_index = 0, new_txdata_index = 0;
 
 	/* Copy the NAPI object as it has been already initialized */
 	from_fp->napi = to_fp->napi;
@@ -53,6 +60,30 @@ static inline void bnx2x_move_fp(struct bnx2x *bp, int from, int to)
 	/* Move bnx2x_fastpath contents */
 	memcpy(to_fp, from_fp, sizeof(*to_fp));
 	to_fp->index = to;
+
+	/* move sp_objs contents as well, as their indices match fp ones */
+	memcpy(to_sp_objs, from_sp_objs, sizeof(*to_sp_objs));
+
+	/* move fp_stats contents as well, as their indices match fp ones */
+	memcpy(to_fp_stats, from_fp_stats, sizeof(*to_fp_stats));
+
+	/* Update txdata pointers in fp and move txdata content accordingly:
+	 * Each fp consumes 'max_cos' txdata structures, so the index should be
+	 * decremented by max_cos x delta.
+	 */
+
+	old_max_eth_txqs = BNX2X_NUM_ETH_QUEUES(bp) * (bp)->max_cos;
+	new_max_eth_txqs = (BNX2X_NUM_ETH_QUEUES(bp) - from + to) *
+				(bp)->max_cos;
+	if (from == FCOE_IDX(bp)) {
+		old_txdata_index = old_max_eth_txqs + FCOE_TXQ_IDX_OFFSET;
+		new_txdata_index = new_max_eth_txqs + FCOE_TXQ_IDX_OFFSET;
+	}
+
+	memcpy(&bp->bnx2x_txq[old_txdata_index],
+	       &bp->bnx2x_txq[new_txdata_index],
+	       sizeof(struct bnx2x_fp_txdata));
+	to_fp->txdata_ptr[0] = &bp->bnx2x_txq[new_txdata_index];
 }
 
 int load_count[2][3] = { {0} }; /* per-path: 0-common, 1-port0, 2-port1 */
@@ -190,7 +221,7 @@ int bnx2x_tx_int(struct bnx2x *bp, struct bnx2x_fp_txdata *txdata)
 
 		if ((netif_tx_queue_stopped(txq)) &&
 		    (bp->state == BNX2X_STATE_OPEN) &&
-		    (bnx2x_tx_avail(bp, txdata) >= MAX_SKB_FRAGS + 4))
+		    (bnx2x_tx_avail(bp, txdata) >= MAX_DESC_PER_TX_PKT))
 			netif_tx_wake_queue(txq);
 
 		__netif_tx_unlock(txq);
@@ -264,12 +295,20 @@ static inline void bnx2x_update_sge_prod(struct bnx2x_fastpath *fp,
  * CQE (calculated by HW).
  */
 static u32 bnx2x_get_rxhash(const struct bnx2x *bp,
-			    const struct eth_fast_path_rx_cqe *cqe)
+			    const struct eth_fast_path_rx_cqe *cqe,
+			    bool *l4_rxhash)
 {
 	/* Set Toeplitz hash from CQE */
 	if ((bp->dev->features & NETIF_F_RXHASH) &&
-	    (cqe->status_flags & ETH_FAST_PATH_RX_CQE_RSS_HASH_FLG))
+	    (cqe->status_flags & ETH_FAST_PATH_RX_CQE_RSS_HASH_FLG)) {
+		enum eth_rss_hash_type htype;
+
+		htype = cqe->status_flags & ETH_FAST_PATH_RX_CQE_RSS_HASH_TYPE;
+		*l4_rxhash = (htype == TCP_IPV4_HASH_TYPE) ||
+			     (htype == TCP_IPV6_HASH_TYPE);
 		return le32_to_cpu(cqe->rss_hash_result);
+	}
+	*l4_rxhash = false;
 	return 0;
 }
 
@@ -323,7 +362,7 @@ static void bnx2x_tpa_start(struct bnx2x_fastpath *fp, u16 queue,
 	tpa_info->tpa_state = BNX2X_TPA_START;
 	tpa_info->len_on_bd = le16_to_cpu(cqe->len_on_bd);
 	tpa_info->placement_offset = cqe->placement_offset;
-	tpa_info->rxhash = bnx2x_get_rxhash(bp, cqe);
+	tpa_info->rxhash = bnx2x_get_rxhash(bp, cqe, &tpa_info->l4_rxhash);
 	if (fp->mode == TPA_MODE_GRO) {
 		u16 gro_size = le16_to_cpu(cqe->pkt_len_or_gro_seg_len);
 		tpa_info->full_page =
@@ -479,7 +518,7 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 		   where we are and drop the whole packet */
 		err = bnx2x_alloc_rx_sge(bp, fp, sge_idx);
 		if (unlikely(err)) {
-			fp->eth_q_stats.rx_skb_alloc_failed++;
+			bnx2x_fp_qstats(bp, fp)->rx_skb_alloc_failed++;
 			return err;
 		}
 
@@ -558,6 +597,7 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 		skb_reserve(skb, pad + NET_SKB_PAD);
 		skb_put(skb, len);
 		skb->rxhash = tpa_info->rxhash;
+		skb->l4_rxhash = tpa_info->l4_rxhash;
 
 		skb->protocol = eth_type_trans(skb, bp->dev);
 		skb->ip_summed = CHECKSUM_UNNECESSARY;
@@ -584,7 +624,7 @@ drop:
 	/* drop the packet and keep the buffer in the bin */
 	DP(NETIF_MSG_RX_STATUS,
 	   "Failed to allocate or map a new skb - dropping packet!\n");
-	fp->eth_q_stats.rx_skb_alloc_failed++;
+	bnx2x_fp_stats(bp, fp)->eth_q_stats.rx_skb_alloc_failed++;
 }
 
 static int bnx2x_alloc_rx_data(struct bnx2x *bp,
@@ -617,8 +657,10 @@ static int bnx2x_alloc_rx_data(struct bnx2x *bp,
 	return 0;
 }
 
-static void bnx2x_csum_validate(struct sk_buff *skb, union eth_rx_cqe *cqe,
-				struct bnx2x_fastpath *fp)
+static
+void bnx2x_csum_validate(struct sk_buff *skb, union eth_rx_cqe *cqe,
+				 struct bnx2x_fastpath *fp,
+				 struct bnx2x_eth_q_stats *qstats)
 {
 	/* Do nothing if no IP/L4 csum validation was done */
 
@@ -632,7 +674,7 @@ static void bnx2x_csum_validate(struct sk_buff *skb, union eth_rx_cqe *cqe,
 	if (cqe->fast_path_cqe.type_error_flags &
 	    (ETH_FAST_PATH_RX_CQE_IP_BAD_XSUM_FLG |
 	     ETH_FAST_PATH_RX_CQE_L4_BAD_XSUM_FLG))
-		fp->eth_q_stats.hw_csum_err++;
+		qstats->hw_csum_err++;
 	else
 		skb->ip_summed = CHECKSUM_UNNECESSARY;
 }
@@ -679,6 +721,7 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 		enum eth_rx_cqe_type cqe_fp_type;
 		u16 len, pad, queue;
 		u8 *data;
+		bool l4_rxhash;
 
 #ifdef BNX2X_STOP_ON_ERROR
 		if (unlikely(bp->panic))
@@ -776,7 +819,7 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 			DP(NETIF_MSG_RX_ERR | NETIF_MSG_RX_STATUS,
 			   "ERROR  flags %x  rx packet %u\n",
 			   cqe_fp_flags, sw_comp_cons);
-			fp->eth_q_stats.rx_err_discard_pkt++;
+			bnx2x_fp_qstats(bp, fp)->rx_err_discard_pkt++;
 			goto reuse_rx;
 		}
 
@@ -789,7 +832,7 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 			if (skb == NULL) {
 				DP(NETIF_MSG_RX_ERR | NETIF_MSG_RX_STATUS,
 				   "ERROR  packet dropped because of alloc failure\n");
-				fp->eth_q_stats.rx_skb_alloc_failed++;
+				bnx2x_fp_qstats(bp, fp)->rx_skb_alloc_failed++;
 				goto reuse_rx;
 			}
 			memcpy(skb->data, data + pad, len);
@@ -803,14 +846,15 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 				skb = build_skb(data, 0);
 				if (unlikely(!skb)) {
 					kfree(data);
-					fp->eth_q_stats.rx_skb_alloc_failed++;
+					bnx2x_fp_qstats(bp, fp)->
+							rx_skb_alloc_failed++;
 					goto next_rx;
 				}
 				skb_reserve(skb, pad);
 			} else {
 				DP(NETIF_MSG_RX_ERR | NETIF_MSG_RX_STATUS,
 				   "ERROR  packet dropped because of alloc failure\n");
-				fp->eth_q_stats.rx_skb_alloc_failed++;
+				bnx2x_fp_qstats(bp, fp)->rx_skb_alloc_failed++;
 reuse_rx:
 				bnx2x_reuse_rx_data(fp, bd_cons, bd_prod);
 				goto next_rx;
@@ -821,13 +865,14 @@ reuse_rx:
 		skb->protocol = eth_type_trans(skb, bp->dev);
 
 		/* Set Toeplitz hash for a none-LRO skb */
-		skb->rxhash = bnx2x_get_rxhash(bp, cqe_fp);
+		skb->rxhash = bnx2x_get_rxhash(bp, cqe_fp, &l4_rxhash);
+		skb->l4_rxhash = l4_rxhash;
 
 		skb_checksum_none_assert(skb);
 
 		if (bp->dev->features & NETIF_F_RXCSUM)
-			bnx2x_csum_validate(skb, cqe, fp);
-
+			bnx2x_csum_validate(skb, cqe, fp,
+					    bnx2x_fp_qstats(bp, fp));
 
 		skb_record_rx_queue(skb, fp->rx_queue);
 
@@ -888,7 +933,7 @@ static irqreturn_t bnx2x_msix_fp_int(int irq, void *fp_cookie)
 	prefetch(fp->rx_cons_sb);
 
 	for_each_cos_in_tx_queue(fp, cos)
-		prefetch(fp->txdata[cos].tx_cons_sb);
+		prefetch(fp->txdata_ptr[cos]->tx_cons_sb);
 
 	prefetch(&fp->sb_running_index[SM_RX_ID]);
 	napi_schedule(&bnx2x_fp(bp, fp->index, napi));
@@ -1205,7 +1250,7 @@ static void bnx2x_free_tx_skbs(struct bnx2x *bp)
 	for_each_tx_queue(bp, i) {
 		struct bnx2x_fastpath *fp = &bp->fp[i];
 		for_each_cos_in_tx_queue(fp, cos) {
-			struct bnx2x_fp_txdata *txdata = &fp->txdata[cos];
+			struct bnx2x_fp_txdata *txdata = fp->txdata_ptr[cos];
 			unsigned pkts_compl = 0, bytes_compl = 0;
 
 			u16 sw_prod = txdata->tx_pkt_prod;
@@ -1217,7 +1262,8 @@ static void bnx2x_free_tx_skbs(struct bnx2x *bp)
 				sw_cons++;
 			}
 			netdev_tx_reset_queue(
-			    netdev_get_tx_queue(bp->dev, txdata->txq_index));
+				netdev_get_tx_queue(bp->dev,
+						    txdata->txq_index));
 		}
 	}
 }
@@ -1325,7 +1371,7 @@ void bnx2x_free_irq(struct bnx2x *bp)
 		free_irq(bp->dev->irq, bp->dev);
 }
 
-int __devinit bnx2x_enable_msix(struct bnx2x *bp)
+int bnx2x_enable_msix(struct bnx2x *bp)
 {
 	int msix_vec = 0, i, rc, req_cnt;
 
@@ -1579,6 +1625,8 @@ void bnx2x_set_num_queues(struct bnx2x *bp)
 #endif
 	/* Add special queues */
 	bp->num_queues += NON_ETH_CONTEXT_USE;
+
+	BNX2X_DEV_INFO("set number of queues to %d\n", bp->num_queues);
 }
 
 /**
@@ -1607,8 +1655,8 @@ static int bnx2x_set_real_num_queues(struct bnx2x *bp)
 {
 	int rc, tx, rx;
 
-	tx = MAX_TXQS_PER_COS * bp->max_cos;
-	rx = BNX2X_NUM_ETH_QUEUES(bp);
+	tx = BNX2X_NUM_ETH_QUEUES(bp) * bp->max_cos;
+	rx = BNX2X_NUM_QUEUES(bp) - NON_ETH_CONTEXT_USE;
 
 /* account for fcoe queue */
 #ifdef BCM_CNIC
@@ -1666,14 +1714,13 @@ static void bnx2x_set_rx_buf_size(struct bnx2x *bp)
 static int bnx2x_init_rss_pf(struct bnx2x *bp)
 {
 	int i;
-	u8 ind_table[T_ETH_INDIRECTION_TABLE_SIZE] = {0};
 	u8 num_eth_queues = BNX2X_NUM_ETH_QUEUES(bp);
 
 	/* Prepare the initial contents fo the indirection table if RSS is
 	 * enabled
 	 */
-	for (i = 0; i < sizeof(ind_table); i++)
-		ind_table[i] =
+	for (i = 0; i < sizeof(bp->rss_conf_obj.ind_table); i++)
+		bp->rss_conf_obj.ind_table[i] =
 			bp->fp->cl_id +
 			ethtool_rxfh_indir_default(i, num_eth_queues);
 
@@ -1685,12 +1732,11 @@ static int bnx2x_init_rss_pf(struct bnx2x *bp)
 	 * For 57712 and newer on the other hand it's a per-function
 	 * configuration.
 	 */
-	return bnx2x_config_rss_eth(bp, ind_table,
-				    bp->port.pmf || !CHIP_IS_E1x(bp));
+	return bnx2x_config_rss_eth(bp, bp->port.pmf || !CHIP_IS_E1x(bp));
 }
 
 int bnx2x_config_rss_pf(struct bnx2x *bp, struct bnx2x_rss_config_obj *rss_obj,
-			u8 *ind_table, bool config_hash)
+			bool config_hash)
 {
 	struct bnx2x_config_rss_params params = {NULL};
 	int i;
@@ -1713,11 +1759,15 @@ int bnx2x_config_rss_pf(struct bnx2x *bp, struct bnx2x_rss_config_obj *rss_obj,
 	__set_bit(BNX2X_RSS_IPV4_TCP, &params.rss_flags);
 	__set_bit(BNX2X_RSS_IPV6, &params.rss_flags);
 	__set_bit(BNX2X_RSS_IPV6_TCP, &params.rss_flags);
+	if (rss_obj->udp_rss_v4)
+		__set_bit(BNX2X_RSS_IPV4_UDP, &params.rss_flags);
+	if (rss_obj->udp_rss_v6)
+		__set_bit(BNX2X_RSS_IPV6_UDP, &params.rss_flags);
 
 	/* Hash bits */
 	params.rss_result_mask = MULTI_MASK;
 
-	memcpy(params.ind_table, ind_table, sizeof(params.ind_table));
+	memcpy(params.ind_table, rss_obj->ind_table, sizeof(params.ind_table));
 
 	if (config_hash) {
 		/* RSS keys */
@@ -1754,7 +1804,7 @@ static void bnx2x_squeeze_objects(struct bnx2x *bp)
 	int rc;
 	unsigned long ramrod_flags = 0, vlan_mac_flags = 0;
 	struct bnx2x_mcast_ramrod_params rparam = {NULL};
-	struct bnx2x_vlan_mac_obj *mac_obj = &bp->fp->mac_obj;
+	struct bnx2x_vlan_mac_obj *mac_obj = &bp->sp_objs->mac_obj;
 
 	/***************** Cleanup MACs' object first *************************/
 
@@ -1765,7 +1815,7 @@ static void bnx2x_squeeze_objects(struct bnx2x *bp)
 
 	/* Clean ETH primary MAC */
 	__set_bit(BNX2X_ETH_MAC, &vlan_mac_flags);
-	rc = mac_obj->delete_all(bp, &bp->fp->mac_obj, &vlan_mac_flags,
+	rc = mac_obj->delete_all(bp, &bp->sp_objs->mac_obj, &vlan_mac_flags,
 				 &ramrod_flags);
 	if (rc != 0)
 		BNX2X_ERR("Failed to clean ETH MACs: %d\n", rc);
@@ -1851,11 +1901,16 @@ bool bnx2x_test_firmware_version(struct bnx2x *bp, bool is_err)
 static void bnx2x_bz_fp(struct bnx2x *bp, int index)
 {
 	struct bnx2x_fastpath *fp = &bp->fp[index];
+	struct bnx2x_fp_stats *fp_stats = &bp->fp_stats[index];
+
+	int cos;
 	struct napi_struct orig_napi = fp->napi;
+	struct bnx2x_agg_info *orig_tpa_info = fp->tpa_info;
 	/* bzero bnx2x_fastpath contents */
-	if (bp->stats_init)
+	if (bp->stats_init) {
+		memset(fp->tpa_info, 0, sizeof(*fp->tpa_info));
 		memset(fp, 0, sizeof(*fp));
-	else {
+	} else {
 		/* Keep Queue statistics */
 		struct bnx2x_eth_q_stats *tmp_eth_q_stats;
 		struct bnx2x_eth_q_stats_old *tmp_eth_q_stats_old;
@@ -1863,26 +1918,27 @@ static void bnx2x_bz_fp(struct bnx2x *bp, int index)
 		tmp_eth_q_stats = kzalloc(sizeof(struct bnx2x_eth_q_stats),
 					  GFP_KERNEL);
 		if (tmp_eth_q_stats)
-			memcpy(tmp_eth_q_stats, &fp->eth_q_stats,
+			memcpy(tmp_eth_q_stats, &fp_stats->eth_q_stats,
 			       sizeof(struct bnx2x_eth_q_stats));
 
 		tmp_eth_q_stats_old =
 			kzalloc(sizeof(struct bnx2x_eth_q_stats_old),
 				GFP_KERNEL);
 		if (tmp_eth_q_stats_old)
-			memcpy(tmp_eth_q_stats_old, &fp->eth_q_stats_old,
+			memcpy(tmp_eth_q_stats_old, &fp_stats->eth_q_stats_old,
 			       sizeof(struct bnx2x_eth_q_stats_old));
 
+		memset(fp->tpa_info, 0, sizeof(*fp->tpa_info));
 		memset(fp, 0, sizeof(*fp));
 
 		if (tmp_eth_q_stats) {
-			memcpy(&fp->eth_q_stats, tmp_eth_q_stats,
-				   sizeof(struct bnx2x_eth_q_stats));
+			memcpy(&fp_stats->eth_q_stats, tmp_eth_q_stats,
+			       sizeof(struct bnx2x_eth_q_stats));
 			kfree(tmp_eth_q_stats);
 		}
 
 		if (tmp_eth_q_stats_old) {
-			memcpy(&fp->eth_q_stats_old, tmp_eth_q_stats_old,
+			memcpy(&fp_stats->eth_q_stats_old, tmp_eth_q_stats_old,
 			       sizeof(struct bnx2x_eth_q_stats_old));
 			kfree(tmp_eth_q_stats_old);
 		}
@@ -1891,7 +1947,7 @@ static void bnx2x_bz_fp(struct bnx2x *bp, int index)
 
 	/* Restore the NAPI object as it has been already initialized */
 	fp->napi = orig_napi;
-
+	fp->tpa_info = orig_tpa_info;
 	fp->bp = bp;
 	fp->index = index;
 	if (IS_ETH_FP(fp))
@@ -1900,6 +1956,16 @@ static void bnx2x_bz_fp(struct bnx2x *bp, int index)
 		/* Special queues support only one CoS */
 		fp->max_cos = 1;
 
+	/* Init txdata pointers */
+#ifdef BCM_CNIC
+	if (IS_FCOE_FP(fp))
+		fp->txdata_ptr[0] = &bp->bnx2x_txq[FCOE_TXQ_IDX(bp)];
+#endif
+	if (IS_ETH_FP(fp))
+		for_each_cos_in_tx_queue(fp, cos)
+			fp->txdata_ptr[cos] = &bp->bnx2x_txq[cos *
+				BNX2X_NUM_ETH_QUEUES(bp) + index];
+
 	/*
 	 * set the tpa flag for each queue. The tpa flag determines the queue
 	 * minimal size so it must be set prior to queue memory allocation
@@ -1949,11 +2015,13 @@ int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 	/*
 	 * Zero fastpath structures preserving invariants like napi, which are
 	 * allocated only once, fp index, max_cos, bp pointer.
-	 * Also set fp->disable_tpa.
+	 * Also set fp->disable_tpa and txdata_ptr.
 	 */
 	DP(NETIF_MSG_IFUP, "num queues: %d", bp->num_queues);
 	for_each_queue(bp, i)
 		bnx2x_bz_fp(bp, i);
+	memset(bp->bnx2x_txq, 0, bp->bnx2x_txq_size *
+	       sizeof(struct bnx2x_fp_txdata));
 
 
 	/* Set the receive queues buffer size */
@@ -2176,6 +2244,7 @@ int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 		break;
 
 	case LOAD_DIAG:
+	case LOAD_LOOPBACK_EXT:
 		bp->state = BNX2X_STATE_DIAG;
 		break;
 
@@ -2195,6 +2264,7 @@ int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 	/* re-read iscsi info */
 	bnx2x_get_iscsi_info(bp);
 	bnx2x_setup_cnic_irq_info(bp);
+	bnx2x_setup_cnic_info(bp);
 	if (bp->state == BNX2X_STATE_OPEN)
 		bnx2x_cnic_notify(bp, CNIC_CTL_START_CMD);
 #endif
@@ -2215,7 +2285,10 @@ int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 		return -EBUSY;
 	}
 
-	bnx2x_dcbx_init(bp);
+	/* If PMF - send ADMIN DCBX msg to MFW to initiate DCBX FSM */
+	if (bp->port.pmf && (bp->state != BNX2X_STATE_DIAG))
+		bnx2x_dcbx_init(bp, false);
+
 	return 0;
 
 #ifndef BNX2X_STOP_ON_ERROR
@@ -2298,6 +2371,7 @@ int bnx2x_nic_unload(struct bnx2x *bp, int unload_mode)
 
 	/* Stop Tx */
 	bnx2x_tx_disable(bp);
+	netdev_reset_tc(bp->dev);
 
 #ifdef BCM_CNIC
 	bnx2x_cnic_notify(bp, CNIC_CTL_STOP_CMD);
@@ -2456,8 +2530,8 @@ int bnx2x_poll(struct napi_struct *napi, int budget)
 #endif
 
 		for_each_cos_in_tx_queue(fp, cos)
-			if (bnx2x_tx_queue_has_work(&fp->txdata[cos]))
-				bnx2x_tx_int(bp, &fp->txdata[cos]);
+			if (bnx2x_tx_queue_has_work(fp->txdata_ptr[cos]))
+				bnx2x_tx_int(bp, fp->txdata_ptr[cos]);
 
 
 		if (bnx2x_has_rx_work(fp)) {
@@ -2834,7 +2908,6 @@ netdev_tx_t bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
 {
 	struct bnx2x *bp = netdev_priv(dev);
 
-	struct bnx2x_fastpath *fp;
 	struct netdev_queue *txq;
 	struct bnx2x_fp_txdata *txdata;
 	struct sw_tx_bd *tx_buf;
@@ -2844,7 +2917,7 @@ netdev_tx_t bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	struct eth_tx_parse_bd_e2 *pbd_e2 = NULL;
 	u32 pbd_e2_parsing_data = 0;
 	u16 pkt_prod, bd_prod;
-	int nbd, txq_index, fp_index, txdata_index;
+	int nbd, txq_index;
 	dma_addr_t mapping;
 	u32 xmit_type = bnx2x_xmit_type(bp, skb);
 	int i;
@@ -2863,39 +2936,22 @@ netdev_tx_t bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	BUG_ON(txq_index >= MAX_ETH_TXQ_IDX(bp) + FCOE_PRESENT);
 
-	/* decode the fastpath index and the cos index from the txq */
-	fp_index = TXQ_TO_FP(txq_index);
-	txdata_index = TXQ_TO_COS(txq_index);
-
-#ifdef BCM_CNIC
-	/*
-	 * Override the above for the FCoE queue:
-	 *   - FCoE fp entry is right after the ETH entries.
-	 *   - FCoE L2 queue uses bp->txdata[0] only.
-	 */
-	if (unlikely(!NO_FCOE(bp) && (txq_index ==
-				      bnx2x_fcoe_tx(bp, txq_index)))) {
-		fp_index = FCOE_IDX;
-		txdata_index = 0;
-	}
-#endif
+	txdata = &bp->bnx2x_txq[txq_index];
 
 	/* enable this debug print to view the transmission queue being used
 	DP(NETIF_MSG_TX_QUEUED, "indices: txq %d, fp %d, txdata %d\n",
 	   txq_index, fp_index, txdata_index); */
 
-	/* locate the fastpath and the txdata */
-	fp = &bp->fp[fp_index];
-	txdata = &fp->txdata[txdata_index];
-
 	/* enable this debug print to view the tranmission details
 	DP(NETIF_MSG_TX_QUEUED,
 	   "transmitting packet cid %d fp index %d txdata_index %d tx_data ptr %p fp pointer %p\n",
 	   txdata->cid, fp_index, txdata_index, txdata, fp); */
 
 	if (unlikely(bnx2x_tx_avail(bp, txdata) <
-		     (skb_shinfo(skb)->nr_frags + 3))) {
-		fp->eth_q_stats.driver_xoff++;
+			skb_shinfo(skb)->nr_frags +
+			BDS_PER_TX_PKT +
+			NEXT_CNT_PER_TX_PKT(MAX_BDS_PER_TX_PKT))) {
+		bnx2x_fp_qstats(bp, txdata->parent_fp)->driver_xoff++;
 		netif_tx_stop_queue(txq);
 		BNX2X_ERR("BUG! Tx ring full when queue awake!\n");
 		return NETDEV_TX_BUSY;
@@ -3169,7 +3225,7 @@ netdev_tx_t bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	txdata->tx_bd_prod += nbd;
 
-	if (unlikely(bnx2x_tx_avail(bp, txdata) < MAX_SKB_FRAGS + 4)) {
+	if (unlikely(bnx2x_tx_avail(bp, txdata) < MAX_DESC_PER_TX_PKT)) {
 		netif_tx_stop_queue(txq);
 
 		/* paired memory barrier is in bnx2x_tx_int(), we have to keep
@@ -3177,8 +3233,8 @@ netdev_tx_t bnx2x_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		 * fp->bd_tx_cons */
 		smp_mb();
 
-		fp->eth_q_stats.driver_xoff++;
-		if (bnx2x_tx_avail(bp, txdata) >= MAX_SKB_FRAGS + 4)
+		bnx2x_fp_qstats(bp, txdata->parent_fp)->driver_xoff++;
+		if (bnx2x_tx_avail(bp, txdata) >= MAX_DESC_PER_TX_PKT)
 			netif_tx_wake_queue(txq);
 	}
 	txdata->tx_pkt++;
@@ -3243,7 +3299,7 @@ int bnx2x_setup_tc(struct net_device *dev, u8 num_tc)
 	/* configure traffic class to transmission queue mapping */
 	for (cos = 0; cos < bp->max_cos; cos++) {
 		count = BNX2X_NUM_ETH_QUEUES(bp);
-		offset = cos * MAX_TXQS_PER_COS;
+		offset = cos * BNX2X_NUM_NON_CNIC_QUEUES(bp);
 		netdev_set_tc_queue(dev, cos, count, offset);
 		DP(BNX2X_MSG_SP | NETIF_MSG_IFUP,
 		   "mapping tc %d to offset %d count %d\n",
@@ -3342,7 +3398,7 @@ static void bnx2x_free_fp_mem_at(struct bnx2x *bp, int fp_index)
 	if (!skip_tx_queue(bp, fp_index)) {
 		/* fastpath tx rings: tx_buf tx_desc */
 		for_each_cos_in_tx_queue(fp, cos) {
-			struct bnx2x_fp_txdata *txdata = &fp->txdata[cos];
+			struct bnx2x_fp_txdata *txdata = fp->txdata_ptr[cos];
 
 			DP(NETIF_MSG_IFDOWN,
 			   "freeing tx memory of fp %d cos %d cid %d\n",
@@ -3414,7 +3470,7 @@ static int bnx2x_alloc_rx_bds(struct bnx2x_fastpath *fp,
 			       cqe_ring_prod);
 	fp->rx_pkt = fp->rx_calls = 0;
 
-	fp->eth_q_stats.rx_skb_alloc_failed += failure_cnt;
+	bnx2x_fp_stats(bp, fp)->eth_q_stats.rx_skb_alloc_failed += failure_cnt;
 
 	return i - failure_cnt;
 }
@@ -3499,7 +3555,7 @@ static int bnx2x_alloc_fp_mem_at(struct bnx2x *bp, int index)
 	if (!skip_tx_queue(bp, index)) {
 		/* fastpath tx rings: tx_buf tx_desc */
 		for_each_cos_in_tx_queue(fp, cos) {
-			struct bnx2x_fp_txdata *txdata = &fp->txdata[cos];
+			struct bnx2x_fp_txdata *txdata = fp->txdata_ptr[cos];
 
 			DP(NETIF_MSG_IFUP,
 			   "allocating tx memory of fp %d cos %d\n",
@@ -3582,7 +3638,7 @@ int bnx2x_alloc_fp_mem(struct bnx2x *bp)
 #ifdef BCM_CNIC
 	if (!NO_FCOE(bp))
 		/* FCoE */
-		if (bnx2x_alloc_fp_mem_at(bp, FCOE_IDX))
+		if (bnx2x_alloc_fp_mem_at(bp, FCOE_IDX(bp)))
 			/* we will fail load process instead of mark
 			 * NO_FCOE_FLAG
 			 */
@@ -3607,7 +3663,7 @@ int bnx2x_alloc_fp_mem(struct bnx2x *bp)
 		 */
 
 		/* move FCoE fp even NO_FCOE_FLAG is on */
-		bnx2x_move_fp(bp, FCOE_IDX, FCOE_IDX - delta);
+		bnx2x_move_fp(bp, FCOE_IDX(bp), FCOE_IDX(bp) - delta);
 #endif
 		bp->num_queues -= delta;
 		BNX2X_ERR("Adjusted num of queues from %d to %d\n",
@@ -3619,7 +3675,11 @@ int bnx2x_alloc_fp_mem(struct bnx2x *bp)
 
 void bnx2x_free_mem_bp(struct bnx2x *bp)
 {
+	kfree(bp->fp->tpa_info);
 	kfree(bp->fp);
+	kfree(bp->sp_objs);
+	kfree(bp->fp_stats);
+	kfree(bp->bnx2x_txq);
 	kfree(bp->msix_table);
 	kfree(bp->ilt);
 }
@@ -3630,6 +3690,8 @@ int __devinit bnx2x_alloc_mem_bp(struct bnx2x *bp)
 	struct msix_entry *tbl;
 	struct bnx2x_ilt *ilt;
 	int msix_table_size = 0;
+	int fp_array_size;
+	int i;
 
 	/*
 	 * The biggest MSI-X table we might need is as a maximum number of fast
@@ -3638,12 +3700,44 @@ int __devinit bnx2x_alloc_mem_bp(struct bnx2x *bp)
 	msix_table_size = bp->igu_sb_cnt + 1;
 
 	/* fp array: RSS plus CNIC related L2 queues */
-	fp = kcalloc(BNX2X_MAX_RSS_COUNT(bp) + NON_ETH_CONTEXT_USE,
-		     sizeof(*fp), GFP_KERNEL);
+	fp_array_size = BNX2X_MAX_RSS_COUNT(bp) + NON_ETH_CONTEXT_USE;
+	BNX2X_DEV_INFO("fp_array_size %d", fp_array_size);
+
+	fp = kcalloc(fp_array_size, sizeof(*fp), GFP_KERNEL);
 	if (!fp)
 		goto alloc_err;
+	for (i = 0; i < fp_array_size; i++) {
+		fp[i].tpa_info =
+			kcalloc(ETH_MAX_AGGREGATION_QUEUES_E1H_E2,
+				sizeof(struct bnx2x_agg_info), GFP_KERNEL);
+		if (!(fp[i].tpa_info))
+			goto alloc_err;
+	}
+
 	bp->fp = fp;
 
+	/* allocate sp objs */
+	bp->sp_objs = kcalloc(fp_array_size, sizeof(struct bnx2x_sp_objs),
+			      GFP_KERNEL);
+	if (!bp->sp_objs)
+		goto alloc_err;
+
+	/* allocate fp_stats */
+	bp->fp_stats = kcalloc(fp_array_size, sizeof(struct bnx2x_fp_stats),
+			       GFP_KERNEL);
+	if (!bp->fp_stats)
+		goto alloc_err;
+
+	/* Allocate memory for the transmission queues array */
+	bp->bnx2x_txq_size = BNX2X_MAX_RSS_COUNT(bp) * BNX2X_MULTI_TX_COS;
+#ifdef BCM_CNIC
+	bp->bnx2x_txq_size++;
+#endif
+	bp->bnx2x_txq = kcalloc(bp->bnx2x_txq_size,
+				sizeof(struct bnx2x_fp_txdata), GFP_KERNEL);
+	if (!bp->bnx2x_txq)
+		goto alloc_err;
+
 	/* msix table */
 	tbl = kcalloc(msix_table_size, sizeof(*tbl), GFP_KERNEL);
 	if (!tbl)