update Intel copyright years to 2014
[dpdk.git] / lib / librte_pmd_ixgbe / ixgbe_rxtx.c
index 9b7ba2c..3ff22a7 100644 (file)
@@ -1,35 +1,34 @@
 /*-
  *   BSD LICENSE
  * 
- *   Copyright(c) 2010-2013 Intel Corporation. All rights reserved.
+ *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
  *   All rights reserved.
  * 
- *   Redistribution and use in source and binary forms, with or without 
- *   modification, are permitted provided that the following conditions 
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
  *   are met:
  * 
- *     * Redistributions of source code must retain the above copyright 
+ *     * Redistributions of source code must retain the above copyright
  *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright 
- *       notice, this list of conditions and the following disclaimer in 
- *       the documentation and/or other materials provided with the 
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
  *       distribution.
- *     * Neither the name of Intel Corporation nor the names of its 
- *       contributors may be used to endorse or promote products derived 
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
  *       from this software without specific prior written permission.
  * 
- *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 
- *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 
- *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 
- *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 
- *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 
- *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 
- *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 
- *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 
- *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 
- *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- * 
  */
 
 #include <sys/queue.h>
@@ -141,6 +140,7 @@ struct igb_rx_queue {
 #endif
        uint16_t            rx_free_thresh; /**< max free RX desc to hold. */
        uint16_t            queue_id; /**< RX queue index. */
+       uint16_t            reg_idx;  /**< RX queue register index. */
        uint8_t             port_id;  /**< Device port identifier. */
        uint8_t             crc_len;  /**< 0 if CRC stripped, 4 otherwise. */
        uint8_t             drop_en;  /**< If not 0, set SRRCTL.Drop_En. */
@@ -194,6 +194,7 @@ struct igb_tx_queue {
        uint16_t tx_next_dd; /**< next desc to scan for DD bit */
        uint16_t tx_next_rs; /**< next desc to set RS bit */
        uint16_t            queue_id;      /**< TX queue index. */
+       uint16_t            reg_idx;       /**< TX queue register index. */
        uint8_t             port_id;       /**< Device port identifier. */
        uint8_t             pthresh;       /**< Prefetch threshold register. */
        uint8_t             hthresh;       /**< Host threshold register. */
@@ -1796,12 +1797,14 @@ ixgbe_dev_tx_queue_release(void *txq)
 static void
 ixgbe_reset_tx_queue(struct igb_tx_queue *txq)
 {
+       static const union ixgbe_adv_tx_desc zeroed_desc = { .read = {
+                       .buffer_addr = 0}};
        struct igb_tx_entry *txe = txq->sw_ring;
        uint16_t prev, i;
 
        /* Zero out HW ring memory */
-       for (i = 0; i < sizeof(union ixgbe_adv_tx_desc) * txq->nb_tx_desc; i++) {
-               ((volatile char *)txq->tx_ring)[i] = 0;
+       for (i = 0; i < txq->nb_tx_desc; i++) {
+               txq->tx_ring[i] = zeroed_desc;
        }
 
        /* Initialize SW ring entries */
@@ -1810,9 +1813,9 @@ ixgbe_reset_tx_queue(struct igb_tx_queue *txq)
                volatile union ixgbe_adv_tx_desc *txd = &txq->tx_ring[i];
                txd->wb.status = IXGBE_TXD_STAT_DD;
                txe[i].mbuf = NULL;
-               txe[i].last_id = i;
-               txe[prev].next_id = i;
-               prev = i;
+               txe[i].last_id = (uint16_t)i;
+               txe[prev].next_id = (uint16_t)i;
+               prev = (uint16_t)i;
        }
 
        txq->tx_next_dd = (uint16_t)(txq->tx_rs_thresh - 1);
@@ -1956,6 +1959,8 @@ ixgbe_dev_tx_queue_setup(struct rte_eth_dev *dev,
        txq->hthresh = tx_conf->tx_thresh.hthresh;
        txq->wthresh = tx_conf->tx_thresh.wthresh;
        txq->queue_id = queue_idx;
+       txq->reg_idx = (uint16_t)((RTE_ETH_DEV_SRIOV(dev).active == 0) ? 
+               queue_idx : RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx + queue_idx);
        txq->port_id = dev->data->port_id;
        txq->txq_flags = tx_conf->txq_flags;
 
@@ -1965,7 +1970,7 @@ ixgbe_dev_tx_queue_setup(struct rte_eth_dev *dev,
        if (hw->mac.type == ixgbe_mac_82599_vf)
                txq->tdt_reg_addr = IXGBE_PCI_REG_ADDR(hw, IXGBE_VFTDT(queue_idx));
        else
-               txq->tdt_reg_addr = IXGBE_PCI_REG_ADDR(hw, IXGBE_TDT(queue_idx));
+               txq->tdt_reg_addr = IXGBE_PCI_REG_ADDR(hw, IXGBE_TDT(txq->reg_idx));
 
        txq->tx_ring_phys_addr = (uint64_t) tz->phys_addr;
        txq->tx_ring = (union ixgbe_adv_tx_desc *) tz->addr;
@@ -2088,6 +2093,8 @@ check_rx_burst_bulk_alloc_preconditions(__rte_unused struct igb_rx_queue *rxq)
 static void
 ixgbe_reset_rx_queue(struct igb_rx_queue *rxq)
 {
+       static const union ixgbe_adv_rx_desc zeroed_desc = { .read = {
+                       .pkt_addr = 0}};
        unsigned i;
        uint16_t len;
 
@@ -2115,8 +2122,8 @@ ixgbe_reset_rx_queue(struct igb_rx_queue *rxq)
         * the H/W ring so look-ahead logic in Rx Burst bulk alloc function
         * reads extra memory as zeros.
         */
-       for (i = 0; i < len * sizeof(union ixgbe_adv_rx_desc); i++) {
-               ((volatile char *)rxq->rx_ring)[i] = 0;
+       for (i = 0; i < len; i++) {
+               rxq->rx_ring[i] = zeroed_desc;
        }
 
 #ifdef RTE_LIBRTE_IXGBE_RX_ALLOW_BULK_ALLOC
@@ -2180,6 +2187,8 @@ ixgbe_dev_rx_queue_setup(struct rte_eth_dev *dev,
        rxq->nb_rx_desc = nb_desc;
        rxq->rx_free_thresh = rx_conf->rx_free_thresh;
        rxq->queue_id = queue_idx;
+       rxq->reg_idx = (uint16_t)((RTE_ETH_DEV_SRIOV(dev).active == 0) ? 
+               queue_idx : RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx + queue_idx);
        rxq->port_id = dev->data->port_id;
        rxq->crc_len = (uint8_t) ((dev->data->dev_conf.rxmode.hw_strip_crc) ?
                                                        0 : ETHER_CRC_LEN);
@@ -2200,10 +2209,18 @@ ixgbe_dev_rx_queue_setup(struct rte_eth_dev *dev,
        /*
         * Modified to setup VFRDT for Virtual Function
         */
-       if (hw->mac.type == ixgbe_mac_82599_vf)
-               rxq->rdt_reg_addr = IXGBE_PCI_REG_ADDR(hw, IXGBE_VFRDT(queue_idx));
-       else
-               rxq->rdt_reg_addr = IXGBE_PCI_REG_ADDR(hw, IXGBE_RDT(queue_idx));
+       if (hw->mac.type == ixgbe_mac_82599_vf) {
+               rxq->rdt_reg_addr =
+                       IXGBE_PCI_REG_ADDR(hw, IXGBE_VFRDT(queue_idx));
+               rxq->rdh_reg_addr =
+                       IXGBE_PCI_REG_ADDR(hw, IXGBE_VFRDH(queue_idx));
+       }
+       else {
+               rxq->rdt_reg_addr =
+                       IXGBE_PCI_REG_ADDR(hw, IXGBE_RDT(rxq->reg_idx));
+               rxq->rdh_reg_addr =
+                       IXGBE_PCI_REG_ADDR(hw, IXGBE_RDH(rxq->reg_idx));
+       }
 
        rxq->rx_ring_phys_addr = (uint64_t) rz->phys_addr;
        rxq->rx_ring = (union ixgbe_adv_rx_desc *) rz->addr;
@@ -2257,29 +2274,49 @@ ixgbe_dev_rx_queue_setup(struct rte_eth_dev *dev,
        return 0;
 }
 
-uint32_t ixgbe_dev_rx_queue_count(struct rte_eth_dev *dev,
-                                       uint16_t rx_queue_id)
+uint32_t
+ixgbe_dev_rx_queue_count(struct rte_eth_dev *dev, uint16_t rx_queue_id)
 {
+#define IXGBE_RXQ_SCAN_INTERVAL 4
+       volatile union ixgbe_adv_rx_desc *rxdp;
        struct igb_rx_queue *rxq;
-       uint32_t nb_pkts_available;
-       uint32_t rx_rdh;
-       uint32_t rx_id;
+       uint32_t desc = 0;
 
        if (rx_queue_id >= dev->data->nb_rx_queues) {
-           PMD_RX_LOG(DEBUG,"Invalid RX queue_id=%d\n", rx_queue_id);
-           return 0;
+               PMD_RX_LOG(ERR, "Invalid RX queue id=%d\n", rx_queue_id);
+               return 0;
        }
 
        rxq = dev->data->rx_queues[rx_queue_id];
-       rx_id = (uint16_t)((rxq->rx_tail == 0) ?
-               (rxq->nb_rx_desc - 1) : (rxq->rx_tail - 1));
-       rx_rdh = IXGBE_PCI_REG(rxq->rdh_reg_addr);
-       if (rx_rdh > rx_id) 
-           nb_pkts_available = rx_rdh - rx_id;
-       else 
-           nb_pkts_available = rx_rdh - rx_id + rxq->nb_rx_desc;
-       return (nb_pkts_available);
+       rxdp = &(rxq->rx_ring[rxq->rx_tail]);
+
+       while ((desc < rxq->nb_rx_desc) &&
+               (rxdp->wb.upper.status_error & IXGBE_RXDADV_STAT_DD)) {
+               desc += IXGBE_RXQ_SCAN_INTERVAL;
+               rxdp += IXGBE_RXQ_SCAN_INTERVAL;
+               if (rxq->rx_tail + desc >= rxq->nb_rx_desc)
+                       rxdp = &(rxq->rx_ring[rxq->rx_tail +
+                               desc - rxq->nb_rx_desc]);
+       }
+
+       return desc;
+}
+
+int
+ixgbe_dev_rx_descriptor_done(void *rx_queue, uint16_t offset)
+{
+       volatile union ixgbe_adv_rx_desc *rxdp;
+       struct igb_rx_queue *rxq = rx_queue;
+       uint32_t desc;
+
+       if (unlikely(offset >= rxq->nb_rx_desc))
+               return 0;
+       desc = rxq->rx_tail + offset;
+       if (desc >= rxq->nb_rx_desc)
+               desc -= rxq->nb_rx_desc;
+
+       rxdp = &rxq->rx_ring[desc];
+       return !!(rxdp->wb.upper.status_error & IXGBE_RXDADV_STAT_DD);
 }
 
 void
@@ -2477,6 +2514,7 @@ ixgbe_vmdq_dcb_configure(struct rte_eth_dev *dev)
        } else {
                vt_ctl |= IXGBE_VT_CTL_DIS_DEFPL;
        }
+
        IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vt_ctl);
 
        /* RTRUP2TC: mapping user priorities to traffic classes (TCs) */
@@ -2832,7 +2870,7 @@ ixgbe_dcb_hw_configure(struct rte_eth_dev *dev,
                        IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 
        switch(dev->data->dev_conf.rxmode.mq_mode){
-       case ETH_VMDQ_DCB:
+       case ETH_MQ_RX_VMDQ_DCB:
                dcb_config->vt_mode = true;
                if (hw->mac.type != ixgbe_mac_82598EB) {
                        config_dcb_rx = DCB_RX_CONFIG;
@@ -2845,7 +2883,7 @@ ixgbe_dcb_hw_configure(struct rte_eth_dev *dev,
                        ixgbe_vmdq_dcb_configure(dev);
                }
                break;
-       case ETH_DCB_RX:
+       case ETH_MQ_RX_DCB:
                dcb_config->vt_mode = false;
                config_dcb_rx = DCB_RX_CONFIG;
                /* Get dcb TX configuration parameters from rte_eth_conf */
@@ -2858,7 +2896,7 @@ ixgbe_dcb_hw_configure(struct rte_eth_dev *dev,
                break;
        }
        switch (dev->data->dev_conf.txmode.mq_mode) {
-       case ETH_VMDQ_DCB_TX:
+       case ETH_MQ_TX_VMDQ_DCB:
                dcb_config->vt_mode = true;
                config_dcb_tx = DCB_TX_CONFIG;
                /* get DCB and VT TX configuration parameters from rte_eth_conf */
@@ -2867,9 +2905,9 @@ ixgbe_dcb_hw_configure(struct rte_eth_dev *dev,
                ixgbe_vmdq_dcb_hw_tx_config(dev,dcb_config);
                break;
 
-       case ETH_DCB_TX:
+       case ETH_MQ_TX_DCB:
                dcb_config->vt_mode = false;
-               config_dcb_tx = DCB_RX_CONFIG;
+               config_dcb_tx = DCB_TX_CONFIG;
                /*get DCB TX configuration parameters from rte_eth_conf*/
                ixgbe_dcb_tx_config(dev,dcb_config);
                /*Configure general DCB TX parameters*/
@@ -2996,15 +3034,135 @@ void ixgbe_configure_dcb(struct rte_eth_dev *dev)
 {
        struct ixgbe_dcb_config *dcb_cfg =
                        IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private); 
+       struct rte_eth_conf *dev_conf = &(dev->data->dev_conf);
        
        PMD_INIT_FUNC_TRACE();  
+       
+       /* check support mq_mode for DCB */
+       if ((dev_conf->rxmode.mq_mode != ETH_MQ_RX_VMDQ_DCB) && 
+           (dev_conf->rxmode.mq_mode != ETH_MQ_RX_DCB)) 
+               return;
+
+       if (dev->data->nb_rx_queues != ETH_DCB_NUM_QUEUES)
+               return;
+
        /** Configure DCB hardware **/
-       if(((dev->data->dev_conf.rxmode.mq_mode != ETH_RSS) && 
-               (dev->data->nb_rx_queues == ETH_DCB_NUM_QUEUES))||
-                       ((dev->data->dev_conf.txmode.mq_mode != ETH_DCB_NONE) && 
-                           (dev->data->nb_tx_queues == ETH_DCB_NUM_QUEUES))) {
-               ixgbe_dcb_hw_configure(dev,dcb_cfg);
+       ixgbe_dcb_hw_configure(dev,dcb_cfg);
+       
+       return;
+}
+
+/*
+ * VMDq only support for 10 GbE NIC.
+ */
+static void
+ixgbe_vmdq_rx_hw_configure(struct rte_eth_dev *dev)
+{
+       struct rte_eth_vmdq_rx_conf *cfg;
+       struct ixgbe_hw *hw;
+       enum rte_eth_nb_pools num_pools;
+       uint32_t mrqc, vt_ctl, vlanctrl;
+       int i;
+
+       PMD_INIT_FUNC_TRACE();
+       hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       cfg = &dev->data->dev_conf.rx_adv_conf.vmdq_rx_conf;
+       num_pools = cfg->nb_queue_pools;
+
+       ixgbe_rss_disable(dev);
+
+       /* MRQC: enable vmdq */
+       mrqc = IXGBE_MRQC_VMDQEN;
+       IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
+
+       /* PFVTCTL: turn on virtualisation and set the default pool */
+       vt_ctl = IXGBE_VT_CTL_VT_ENABLE | IXGBE_VT_CTL_REPLEN;
+       if (cfg->enable_default_pool)
+               vt_ctl |= (cfg->default_pool << IXGBE_VT_CTL_POOL_SHIFT);
+       else
+               vt_ctl |= IXGBE_VT_CTL_DIS_DEFPL;
+
+       IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vt_ctl);
+
+       /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
+       vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
+       vlanctrl |= IXGBE_VLNCTRL_VFE ; /* enable vlan filters */
+       IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
+
+       /* VFTA - enable all vlan filters */
+       for (i = 0; i < NUM_VFTA_REGISTERS; i++) 
+               IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), UINT32_MAX);
+
+       /* VFRE: pool enabling for receive - 64 */
+       IXGBE_WRITE_REG(hw, IXGBE_VFRE(0), UINT32_MAX);
+       if (num_pools == ETH_64_POOLS)
+               IXGBE_WRITE_REG(hw, IXGBE_VFRE(1), UINT32_MAX);
+
+       /*
+        * MPSAR - allow pools to read specific mac addresses
+        * In this case, all pools should be able to read from mac addr 0
+        */
+       IXGBE_WRITE_REG(hw, IXGBE_MPSAR_LO(0), UINT32_MAX);
+       IXGBE_WRITE_REG(hw, IXGBE_MPSAR_HI(0), UINT32_MAX);
+
+       /* PFVLVF, PFVLVFB: set up filters for vlan tags as configured */
+       for (i = 0; i < cfg->nb_pool_maps; i++) {
+               /* set vlan id in VF register and set the valid bit */
+               IXGBE_WRITE_REG(hw, IXGBE_VLVF(i), (IXGBE_VLVF_VIEN | \
+                               (cfg->pool_map[i].vlan_id & IXGBE_RXD_VLAN_ID_MASK)));
+               /*
+                * Put the allowed pools in VFB reg. As we only have 16 or 64
+                * pools, we only need to use the first half of the register
+                * i.e. bits 0-31
+                */
+               if (((cfg->pool_map[i].pools >> 32) & UINT32_MAX) == 0) 
+                       IXGBE_WRITE_REG(hw, IXGBE_VLVFB(i*2), \
+                                       (cfg->pool_map[i].pools & UINT32_MAX));
+               else
+                       IXGBE_WRITE_REG(hw, IXGBE_VLVFB((i*2+1)), \
+                                       ((cfg->pool_map[i].pools >> 32) \
+                                       & UINT32_MAX));
+
        }
+
+       IXGBE_WRITE_FLUSH(hw);
+}
+
+/*
+ * ixgbe_dcb_config_tx_hw_config - Configure general VMDq TX parameters
+ * @hw: pointer to hardware structure
+ */
+static void 
+ixgbe_vmdq_tx_hw_configure(struct ixgbe_hw *hw)
+{
+       uint32_t reg;
+       uint32_t q;
+       
+       PMD_INIT_FUNC_TRACE();
+       /*PF VF Transmit Enable*/
+       IXGBE_WRITE_REG(hw, IXGBE_VFTE(0), UINT32_MAX);
+       IXGBE_WRITE_REG(hw, IXGBE_VFTE(1), UINT32_MAX);
+
+       /* Disable the Tx desc arbiter so that MTQC can be changed */
+       reg = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
+       reg |= IXGBE_RTTDCS_ARBDIS;
+       IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, reg);
+
+       reg = IXGBE_MTQC_VT_ENA | IXGBE_MTQC_64VF;
+       IXGBE_WRITE_REG(hw, IXGBE_MTQC, reg);
+
+       /* Disable drop for all queues */
+       for (q = 0; q < IXGBE_MAX_RX_QUEUE_NUM; q++)
+               IXGBE_WRITE_REG(hw, IXGBE_QDE,
+                 (IXGBE_QDE_WRITE | (q << IXGBE_QDE_IDX_SHIFT)));
+
+       /* Enable the Tx desc arbiter */
+       reg = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
+       reg &= ~IXGBE_RTTDCS_ARBDIS;
+       IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, reg);
+
+       IXGBE_WRITE_FLUSH(hw);
+
        return;
 }
 
@@ -3043,6 +3201,123 @@ ixgbe_alloc_rx_queue_mbufs(struct igb_rx_queue *rxq)
        return 0;
 }
 
+static int
+ixgbe_dev_mq_rx_configure(struct rte_eth_dev *dev)
+{
+       struct ixgbe_hw *hw = 
+               IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+       if (hw->mac.type == ixgbe_mac_82598EB)
+               return 0;
+
+       if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
+               /* 
+                * SRIOV inactive scheme
+                * any DCB/RSS w/o VMDq multi-queue setting
+                */
+               if (dev->data->nb_rx_queues > 1)
+                       switch (dev->data->dev_conf.rxmode.mq_mode) {
+                       case ETH_MQ_RX_NONE:
+                               /* if mq_mode not assign, we use rss mode.*/
+                       case ETH_MQ_RX_RSS:
+                               ixgbe_rss_configure(dev);
+                               break;
+
+                       case ETH_MQ_RX_VMDQ_DCB:
+                               ixgbe_vmdq_dcb_configure(dev);
+                               break;
+       
+                       case ETH_MQ_RX_VMDQ_ONLY:
+                               ixgbe_vmdq_rx_hw_configure(dev);
+                               break;
+                       
+                       default: ixgbe_rss_disable(dev);
+                       }
+               else
+                       ixgbe_rss_disable(dev);
+       } else {
+               switch (RTE_ETH_DEV_SRIOV(dev).active) {
+               /*
+                * SRIOV active scheme
+                * FIXME if support DCB/RSS together with VMDq & SRIOV
+                */
+               case ETH_64_POOLS:
+                       IXGBE_WRITE_REG(hw, IXGBE_MRQC, IXGBE_MRQC_VMDQEN);
+                       break;
+
+               case ETH_32_POOLS:
+                       IXGBE_WRITE_REG(hw, IXGBE_MRQC, IXGBE_MRQC_VMDQRT4TCEN);
+                       break;
+               
+               case ETH_16_POOLS:
+                       IXGBE_WRITE_REG(hw, IXGBE_MRQC, IXGBE_MRQC_VMDQRT8TCEN);
+                       break;
+               default:
+                       RTE_LOG(ERR, PMD, "invalid pool number in IOV mode\n");
+               }
+       }
+
+       return 0;
+}
+
+static int
+ixgbe_dev_mq_tx_configure(struct rte_eth_dev *dev)
+{
+       struct ixgbe_hw *hw = 
+               IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       uint32_t mtqc;
+       uint32_t rttdcs;
+
+       if (hw->mac.type == ixgbe_mac_82598EB)
+               return 0;
+
+       /* disable arbiter before setting MTQC */
+       rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
+       rttdcs |= IXGBE_RTTDCS_ARBDIS;
+       IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
+
+       if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
+               /* 
+                * SRIOV inactive scheme
+                * any DCB w/o VMDq multi-queue setting
+                */
+               if (dev->data->dev_conf.txmode.mq_mode == ETH_MQ_TX_VMDQ_ONLY)
+                       ixgbe_vmdq_tx_hw_configure(hw);
+               else {
+                       mtqc = IXGBE_MTQC_64Q_1PB;
+                       IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
+               }
+       } else {
+               switch (RTE_ETH_DEV_SRIOV(dev).active) {
+
+               /*
+                * SRIOV active scheme
+                * FIXME if support DCB together with VMDq & SRIOV
+                */
+               case ETH_64_POOLS:
+                       mtqc = IXGBE_MTQC_VT_ENA | IXGBE_MTQC_64VF;
+                       break;
+               case ETH_32_POOLS:
+                       mtqc = IXGBE_MTQC_VT_ENA | IXGBE_MTQC_32VF;
+                       break;
+               case ETH_16_POOLS:
+                       mtqc = IXGBE_MTQC_VT_ENA | IXGBE_MTQC_RT_ENA | 
+                               IXGBE_MTQC_8TC_8TQ;
+                       break;
+               default:
+                       mtqc = IXGBE_MTQC_64Q_1PB;
+                       RTE_LOG(ERR, PMD, "invalid pool number in IOV mode\n");
+               }
+               IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
+       }
+
+       /* re-enable arbiter */
+       rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
+       IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
+
+       return 0;
+}
+
 /*
  * Initializes Receive Unit.
  */
@@ -3063,7 +3338,7 @@ ixgbe_dev_rx_init(struct rte_eth_dev *dev)
        uint16_t buf_size;
        uint16_t i;
        int ret;
-
+       
        PMD_INIT_FUNC_TRACE();
        hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 
@@ -3123,14 +3398,14 @@ ixgbe_dev_rx_init(struct rte_eth_dev *dev)
 
                /* Setup the Base and Length of the Rx Descriptor Rings */
                bus_addr = rxq->rx_ring_phys_addr;
-               IXGBE_WRITE_REG(hw, IXGBE_RDBAL(i),
+               IXGBE_WRITE_REG(hw, IXGBE_RDBAL(rxq->reg_idx),
                                (uint32_t)(bus_addr & 0x00000000ffffffffULL));
-               IXGBE_WRITE_REG(hw, IXGBE_RDBAH(i),
+               IXGBE_WRITE_REG(hw, IXGBE_RDBAH(rxq->reg_idx),
                                (uint32_t)(bus_addr >> 32));
-               IXGBE_WRITE_REG(hw, IXGBE_RDLEN(i),
+               IXGBE_WRITE_REG(hw, IXGBE_RDLEN(rxq->reg_idx),
                                rxq->nb_rx_desc * sizeof(union ixgbe_adv_rx_desc));
-               IXGBE_WRITE_REG(hw, IXGBE_RDH(i), 0);
-               IXGBE_WRITE_REG(hw, IXGBE_RDT(i), 0);
+               IXGBE_WRITE_REG(hw, IXGBE_RDH(rxq->reg_idx), 0);
+               IXGBE_WRITE_REG(hw, IXGBE_RDT(rxq->reg_idx), 0);
 
                /* Configure the SRRCTL register */
 #ifdef RTE_HEADER_SPLIT_ENABLE
@@ -3145,7 +3420,7 @@ ixgbe_dev_rx_init(struct rte_eth_dev *dev)
                                        IXGBE_PSRTYPE_UDPHDR   |
                                        IXGBE_PSRTYPE_IPV4HDR  |
                                        IXGBE_PSRTYPE_IPV6HDR;
-                               IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(i), psrtype);
+                               IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(rxq->reg_idx), psrtype);
                        }
                        srrctl = ((dev->data->dev_conf.rxmode.split_hdr_size <<
                                   IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT) &
@@ -3171,36 +3446,23 @@ ixgbe_dev_rx_init(struct rte_eth_dev *dev)
                                       RTE_PKTMBUF_HEADROOM);
                srrctl |= ((buf_size >> IXGBE_SRRCTL_BSIZEPKT_SHIFT) &
                           IXGBE_SRRCTL_BSIZEPKT_MASK);
-               IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(i), srrctl);
+               IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxq->reg_idx), srrctl);
 
                buf_size = (uint16_t) ((srrctl & IXGBE_SRRCTL_BSIZEPKT_MASK) <<
                                       IXGBE_SRRCTL_BSIZEPKT_SHIFT);
-               if (dev->data->dev_conf.rxmode.max_rx_pkt_len +
-                               IXGBE_RX_BUF_THRESHOLD > buf_size){
+
+               /* It adds dual VLAN length for supporting dual VLAN */
+               if ((dev->data->dev_conf.rxmode.max_rx_pkt_len +
+                               2 * IXGBE_VLAN_TAG_SIZE) > buf_size){
                        dev->data->scattered_rx = 1;
                        dev->rx_pkt_burst = ixgbe_recv_scattered_pkts;
                }
        }
 
        /*
-        * Configure RSS if device configured with multiple RX queues.
+        * Device configured with multiple RX queues.
         */
-       if (hw->mac.type == ixgbe_mac_82599EB) {
-               if (dev->data->nb_rx_queues > 1)
-                       switch (dev->data->dev_conf.rxmode.mq_mode) {
-                               case ETH_RSS:
-                                       ixgbe_rss_configure(dev);
-                                       break;
-
-                               case ETH_VMDQ_DCB:
-                                       ixgbe_vmdq_dcb_configure(dev);
-                                       break;
-
-                               default: ixgbe_rss_disable(dev);
-                       }
-               else
-                       ixgbe_rss_disable(dev);
-       }
+       ixgbe_dev_mq_rx_configure(dev);
 
        /*
         * Setup the Checksum Register.
@@ -3225,7 +3487,7 @@ ixgbe_dev_rx_init(struct rte_eth_dev *dev)
                rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
                IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
        }
-
+       
        return 0;
 }
 
@@ -3240,7 +3502,6 @@ ixgbe_dev_tx_init(struct rte_eth_dev *dev)
        uint64_t bus_addr;
        uint32_t hlreg0;
        uint32_t txctrl;
-       uint32_t rttdcs;
        uint16_t i;
 
        PMD_INIT_FUNC_TRACE();
@@ -3256,15 +3517,15 @@ ixgbe_dev_tx_init(struct rte_eth_dev *dev)
                txq = dev->data->tx_queues[i];
 
                bus_addr = txq->tx_ring_phys_addr;
-               IXGBE_WRITE_REG(hw, IXGBE_TDBAL(i),
+               IXGBE_WRITE_REG(hw, IXGBE_TDBAL(txq->reg_idx),
                                (uint32_t)(bus_addr & 0x00000000ffffffffULL));
-               IXGBE_WRITE_REG(hw, IXGBE_TDBAH(i),
+               IXGBE_WRITE_REG(hw, IXGBE_TDBAH(txq->reg_idx),
                                (uint32_t)(bus_addr >> 32));
-               IXGBE_WRITE_REG(hw, IXGBE_TDLEN(i),
+               IXGBE_WRITE_REG(hw, IXGBE_TDLEN(txq->reg_idx),
                                txq->nb_tx_desc * sizeof(union ixgbe_adv_tx_desc));
                /* Setup the HW Tx Head and TX Tail descriptor pointers */
-               IXGBE_WRITE_REG(hw, IXGBE_TDH(i), 0);
-               IXGBE_WRITE_REG(hw, IXGBE_TDT(i), 0);
+               IXGBE_WRITE_REG(hw, IXGBE_TDH(txq->reg_idx), 0);
+               IXGBE_WRITE_REG(hw, IXGBE_TDT(txq->reg_idx), 0);
 
                /*
                 * Disable Tx Head Writeback RO bit, since this hoses
@@ -3273,9 +3534,9 @@ ixgbe_dev_tx_init(struct rte_eth_dev *dev)
                switch (hw->mac.type) {
                        case ixgbe_mac_82598EB:
                                txctrl = IXGBE_READ_REG(hw,
-                                                       IXGBE_DCA_TXCTRL(i));
+                                                       IXGBE_DCA_TXCTRL(txq->reg_idx));
                                txctrl &= ~IXGBE_DCA_TXCTRL_DESC_WRO_EN;
-                               IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(i),
+                               IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(txq->reg_idx),
                                                txctrl);
                                break;
 
@@ -3283,26 +3544,16 @@ ixgbe_dev_tx_init(struct rte_eth_dev *dev)
                        case ixgbe_mac_X540:
                        default:
                                txctrl = IXGBE_READ_REG(hw,
-                                               IXGBE_DCA_TXCTRL_82599(i));
+                                               IXGBE_DCA_TXCTRL_82599(txq->reg_idx));
                                txctrl &= ~IXGBE_DCA_TXCTRL_DESC_WRO_EN;
-                               IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(i),
+                               IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(txq->reg_idx),
                                                txctrl);
                                break;
                }
        }
 
-       if (hw->mac.type != ixgbe_mac_82598EB) {
-               /* disable arbiter before setting MTQC */
-               rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
-               rttdcs |= IXGBE_RTTDCS_ARBDIS;
-               IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
-
-               IXGBE_WRITE_REG(hw, IXGBE_MTQC, IXGBE_MTQC_64Q_1PB);
-
-               /* re-enable arbiter */
-               rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
-               IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
-       }
+       /* Device configured with multiple TX queues. */
+       ixgbe_dev_mq_tx_configure(dev);
 }
 
 /*
@@ -3327,11 +3578,11 @@ ixgbe_dev_rxtx_start(struct rte_eth_dev *dev)
        for (i = 0; i < dev->data->nb_tx_queues; i++) {
                txq = dev->data->tx_queues[i];
                /* Setup Transmit Threshold Registers */
-               txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
+               txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txq->reg_idx));
                txdctl |= txq->pthresh & 0x7F;
                txdctl |= ((txq->hthresh & 0x7F) << 8);
                txdctl |= ((txq->wthresh & 0x7F) << 16);
-               IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(i), txdctl);
+               IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(txq->reg_idx), txdctl);
        }
 
        if (hw->mac.type != ixgbe_mac_82598EB) {
@@ -3341,16 +3592,17 @@ ixgbe_dev_rxtx_start(struct rte_eth_dev *dev)
        }
 
        for (i = 0; i < dev->data->nb_tx_queues; i++) {
-               txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
+               txq = dev->data->tx_queues[i];
+               txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txq->reg_idx));
                txdctl |= IXGBE_TXDCTL_ENABLE;
-               IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(i), txdctl);
+               IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(txq->reg_idx), txdctl);
 
                /* Wait until TX Enable ready */
                if (hw->mac.type == ixgbe_mac_82599EB) {
                        poll_ms = 10;
                        do {
                                rte_delay_ms(1);
-                               txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
+                               txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txq->reg_idx));
                        } while (--poll_ms && !(txdctl & IXGBE_TXDCTL_ENABLE));
                        if (!poll_ms)
                                PMD_INIT_LOG(ERR, "Could not enable "
@@ -3359,21 +3611,21 @@ ixgbe_dev_rxtx_start(struct rte_eth_dev *dev)
        }
        for (i = 0; i < dev->data->nb_rx_queues; i++) {
                rxq = dev->data->rx_queues[i];
-               rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
+               rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
                rxdctl |= IXGBE_RXDCTL_ENABLE;
-               IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), rxdctl);
+               IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), rxdctl);
 
                /* Wait until RX Enable ready */
                poll_ms = 10;
                do {
                        rte_delay_ms(1);
-                       rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
+                       rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
                } while (--poll_ms && !(rxdctl & IXGBE_RXDCTL_ENABLE));
                if (!poll_ms)
                        PMD_INIT_LOG(ERR, "Could not enable "
                                     "Rx Queue %d\n", i);
                rte_wmb();
-               IXGBE_WRITE_REG(hw, IXGBE_RDT(i), rxq->nb_rx_desc - 1);
+               IXGBE_WRITE_REG(hw, IXGBE_RDT(rxq->reg_idx), rxq->nb_rx_desc - 1);
        }
 
        /* Enable Receive engine */
@@ -3474,7 +3726,10 @@ ixgbevf_dev_rx_init(struct rte_eth_dev *dev)
 
                buf_size = (uint16_t) ((srrctl & IXGBE_SRRCTL_BSIZEPKT_MASK) <<
                                       IXGBE_SRRCTL_BSIZEPKT_SHIFT);
-               if (dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size){
+
+               /* It adds dual VLAN length for supporting dual VLAN */
+               if ((dev->data->dev_conf.rxmode.max_rx_pkt_len +
+                               2 * IXGBE_VLAN_TAG_SIZE) > buf_size) {
                        dev->data->scattered_rx = 1;
                        dev->rx_pkt_burst = ixgbe_recv_scattered_pkts;
                }