net/ice: support Rx queue interruption
authorWenzhuo Lu <wenzhuo.lu@intel.com>
Tue, 18 Dec 2018 08:46:35 +0000 (16:46 +0800)
committerFerruh Yigit <ferruh.yigit@intel.com>
Fri, 21 Dec 2018 15:22:41 +0000 (16:22 +0100)
Add below ops,
rx_queue_intr_enable
rx_queue_intr_disable

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Signed-off-by: Qiming Yang <qiming.yang@intel.com>
Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
doc/guides/nics/features/ice.ini
drivers/net/ice/ice_ethdev.c

index b492ccd..4d6ca4f 100644 (file)
@@ -7,6 +7,7 @@
 Speed capabilities   = Y
 Link status          = Y
 Link status event    = Y
+Rx interrupt         = Y
 Queue start/stop     = Y
 MTU update           = Y
 Jumbo frame          = Y
index a5e8da6..7c9ddcb 100644 (file)
@@ -48,6 +48,10 @@ static int ice_macaddr_add(struct rte_eth_dev *dev,
                           __rte_unused uint32_t index,
                           uint32_t pool);
 static void ice_macaddr_remove(struct rte_eth_dev *dev, uint32_t index);
+static int ice_rx_queue_intr_enable(struct rte_eth_dev *dev,
+                                   uint16_t queue_id);
+static int ice_rx_queue_intr_disable(struct rte_eth_dev *dev,
+                                    uint16_t queue_id);
 static int ice_vlan_pvid_set(struct rte_eth_dev *dev,
                             uint16_t pvid, int on);
 
@@ -86,6 +90,8 @@ static const struct eth_dev_ops ice_eth_dev_ops = {
        .reta_query                   = ice_rss_reta_query,
        .rss_hash_update              = ice_rss_hash_update,
        .rss_hash_conf_get            = ice_rss_hash_conf_get,
+       .rx_queue_intr_enable         = ice_rx_queue_intr_enable,
+       .rx_queue_intr_disable        = ice_rx_queue_intr_disable,
        .vlan_pvid_set                = ice_vlan_pvid_set,
        .rxq_info_get                 = ice_rxq_info_get,
        .txq_info_get                 = ice_txq_info_get,
@@ -1263,11 +1269,40 @@ ice_release_vsi(struct ice_vsi *vsi)
        return 0;
 }
 
+static void
+ice_vsi_disable_queues_intr(struct ice_vsi *vsi)
+{
+       struct rte_eth_dev *dev = vsi->adapter->eth_dev;
+       struct rte_pci_device *pci_dev = ICE_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct ice_hw *hw = ICE_VSI_TO_HW(vsi);
+       uint16_t msix_intr, i;
+
+       /* disable interrupt and also clear all the exist config */
+       for (i = 0; i < vsi->nb_qps; i++) {
+               ICE_WRITE_REG(hw, QINT_TQCTL(vsi->base_queue + i), 0);
+               ICE_WRITE_REG(hw, QINT_RQCTL(vsi->base_queue + i), 0);
+               rte_wmb();
+       }
+
+       if (rte_intr_allow_others(intr_handle))
+               /* vfio-pci */
+               for (i = 0; i < vsi->nb_msix; i++) {
+                       msix_intr = vsi->msix_intr + i;
+                       ICE_WRITE_REG(hw, GLINT_DYN_CTL(msix_intr),
+                                     GLINT_DYN_CTL_WB_ON_ITR_M);
+               }
+       else
+               /* igb_uio */
+               ICE_WRITE_REG(hw, GLINT_DYN_CTL(0), GLINT_DYN_CTL_WB_ON_ITR_M);
+}
+
 static void
 ice_dev_stop(struct rte_eth_dev *dev)
 {
        struct rte_eth_dev_data *data = dev->data;
        struct ice_pf *pf = ICE_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+       struct ice_vsi *main_vsi = pf->main_vsi;
        struct rte_pci_device *pci_dev = ICE_DEV_TO_PCI(dev);
        struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
        uint16_t i;
@@ -1284,6 +1319,9 @@ ice_dev_stop(struct rte_eth_dev *dev)
        for (i = 0; i < data->nb_tx_queues; i++)
                ice_tx_queue_stop(dev, i);
 
+       /* disable all queue interrupts */
+       ice_vsi_disable_queues_intr(main_vsi);
+
        /* Clear all queues and release mbufs */
        ice_clear_queues(dev);
 
@@ -1411,6 +1449,158 @@ static int ice_init_rss(struct ice_pf *pf)
        return 0;
 }
 
+static void
+__vsi_queues_bind_intr(struct ice_vsi *vsi, uint16_t msix_vect,
+                      int base_queue, int nb_queue)
+{
+       struct ice_hw *hw = ICE_VSI_TO_HW(vsi);
+       uint32_t val, val_tx;
+       int i;
+
+       for (i = 0; i < nb_queue; i++) {
+               /*do actual bind*/
+               val = (msix_vect & QINT_RQCTL_MSIX_INDX_M) |
+                     (0 < QINT_RQCTL_ITR_INDX_S) | QINT_RQCTL_CAUSE_ENA_M;
+               val_tx = (msix_vect & QINT_TQCTL_MSIX_INDX_M) |
+                        (0 < QINT_TQCTL_ITR_INDX_S) | QINT_TQCTL_CAUSE_ENA_M;
+
+               PMD_DRV_LOG(INFO, "queue %d is binding to vect %d",
+                           base_queue + i, msix_vect);
+               /* set ITR0 value */
+               ICE_WRITE_REG(hw, GLINT_ITR(0, msix_vect), 0x10);
+               ICE_WRITE_REG(hw, QINT_RQCTL(base_queue + i), val);
+               ICE_WRITE_REG(hw, QINT_TQCTL(base_queue + i), val_tx);
+       }
+}
+
+static void
+ice_vsi_queues_bind_intr(struct ice_vsi *vsi)
+{
+       struct rte_eth_dev *dev = vsi->adapter->eth_dev;
+       struct rte_pci_device *pci_dev = ICE_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct ice_hw *hw = ICE_VSI_TO_HW(vsi);
+       uint16_t msix_vect = vsi->msix_intr;
+       uint16_t nb_msix = RTE_MIN(vsi->nb_msix, intr_handle->nb_efd);
+       uint16_t queue_idx = 0;
+       int record = 0;
+       int i;
+
+       /* clear Rx/Tx queue interrupt */
+       for (i = 0; i < vsi->nb_used_qps; i++) {
+               ICE_WRITE_REG(hw, QINT_TQCTL(vsi->base_queue + i), 0);
+               ICE_WRITE_REG(hw, QINT_RQCTL(vsi->base_queue + i), 0);
+       }
+
+       /* PF bind interrupt */
+       if (rte_intr_dp_is_en(intr_handle)) {
+               queue_idx = 0;
+               record = 1;
+       }
+
+       for (i = 0; i < vsi->nb_used_qps; i++) {
+               if (nb_msix <= 1) {
+                       if (!rte_intr_allow_others(intr_handle))
+                               msix_vect = ICE_MISC_VEC_ID;
+
+                       /* uio mapping all queue to one msix_vect */
+                       __vsi_queues_bind_intr(vsi, msix_vect,
+                                              vsi->base_queue + i,
+                                              vsi->nb_used_qps - i);
+
+                       for (; !!record && i < vsi->nb_used_qps; i++)
+                               intr_handle->intr_vec[queue_idx + i] =
+                                       msix_vect;
+                       break;
+               }
+
+               /* vfio 1:1 queue/msix_vect mapping */
+               __vsi_queues_bind_intr(vsi, msix_vect,
+                                      vsi->base_queue + i, 1);
+
+               if (!!record)
+                       intr_handle->intr_vec[queue_idx + i] = msix_vect;
+
+               msix_vect++;
+               nb_msix--;
+       }
+}
+
+static void
+ice_vsi_enable_queues_intr(struct ice_vsi *vsi)
+{
+       struct rte_eth_dev *dev = vsi->adapter->eth_dev;
+       struct rte_pci_device *pci_dev = ICE_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct ice_hw *hw = ICE_VSI_TO_HW(vsi);
+       uint16_t msix_intr, i;
+
+       if (rte_intr_allow_others(intr_handle))
+               for (i = 0; i < vsi->nb_used_qps; i++) {
+                       msix_intr = vsi->msix_intr + i;
+                       ICE_WRITE_REG(hw, GLINT_DYN_CTL(msix_intr),
+                                     GLINT_DYN_CTL_INTENA_M |
+                                     GLINT_DYN_CTL_CLEARPBA_M |
+                                     GLINT_DYN_CTL_ITR_INDX_M |
+                                     GLINT_DYN_CTL_WB_ON_ITR_M);
+               }
+       else
+               ICE_WRITE_REG(hw, GLINT_DYN_CTL(0),
+                             GLINT_DYN_CTL_INTENA_M |
+                             GLINT_DYN_CTL_CLEARPBA_M |
+                             GLINT_DYN_CTL_ITR_INDX_M |
+                             GLINT_DYN_CTL_WB_ON_ITR_M);
+}
+
+static int
+ice_rxq_intr_setup(struct rte_eth_dev *dev)
+{
+       struct ice_pf *pf = ICE_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+       struct rte_pci_device *pci_dev = ICE_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct ice_vsi *vsi = pf->main_vsi;
+       uint32_t intr_vector = 0;
+
+       rte_intr_disable(intr_handle);
+
+       /* check and configure queue intr-vector mapping */
+       if ((rte_intr_cap_multiple(intr_handle) ||
+            !RTE_ETH_DEV_SRIOV(dev).active) &&
+           dev->data->dev_conf.intr_conf.rxq != 0) {
+               intr_vector = dev->data->nb_rx_queues;
+               if (intr_vector > ICE_MAX_INTR_QUEUE_NUM) {
+                       PMD_DRV_LOG(ERR, "At most %d intr queues supported",
+                                   ICE_MAX_INTR_QUEUE_NUM);
+                       return -ENOTSUP;
+               }
+               if (rte_intr_efd_enable(intr_handle, intr_vector))
+                       return -1;
+       }
+
+       if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
+               intr_handle->intr_vec =
+               rte_zmalloc(NULL, dev->data->nb_rx_queues * sizeof(int),
+                           0);
+               if (!intr_handle->intr_vec) {
+                       PMD_DRV_LOG(ERR,
+                                   "Failed to allocate %d rx_queues intr_vec",
+                                   dev->data->nb_rx_queues);
+                       return -ENOMEM;
+               }
+       }
+
+       /* Map queues with MSIX interrupt */
+       vsi->nb_used_qps = dev->data->nb_rx_queues;
+       ice_vsi_queues_bind_intr(vsi);
+
+       /* Enable interrupts for all the queues */
+       ice_vsi_enable_queues_intr(vsi);
+
+       rte_intr_enable(intr_handle);
+
+       return 0;
+}
+
 static int
 ice_dev_start(struct rte_eth_dev *dev)
 {
@@ -1447,6 +1637,10 @@ ice_dev_start(struct rte_eth_dev *dev)
 
        ice_set_rx_function(dev);
 
+       /* enable Rx interrput and mapping Rx queue to interrupt vector */
+       if (ice_rxq_intr_setup(dev))
+               return -EIO;
+
        ret = ice_aq_set_event_mask(hw, hw->port_info->lport,
                                    ((u16)(ICE_AQ_LINK_EVENT_LINK_FAULT |
                                     ICE_AQ_LINK_EVENT_PHY_TEMP_ALARM |
@@ -2252,6 +2446,42 @@ ice_rss_hash_conf_get(struct rte_eth_dev *dev,
        return 0;
 }
 
+static int ice_rx_queue_intr_enable(struct rte_eth_dev *dev,
+                                   uint16_t queue_id)
+{
+       struct rte_pci_device *pci_dev = ICE_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct ice_hw *hw = ICE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       uint32_t val;
+       uint16_t msix_intr;
+
+       msix_intr = intr_handle->intr_vec[queue_id];
+
+       val = GLINT_DYN_CTL_INTENA_M | GLINT_DYN_CTL_CLEARPBA_M |
+             GLINT_DYN_CTL_ITR_INDX_M;
+       val &= ~GLINT_DYN_CTL_WB_ON_ITR_M;
+
+       ICE_WRITE_REG(hw, GLINT_DYN_CTL(msix_intr), val);
+       rte_intr_enable(&pci_dev->intr_handle);
+
+       return 0;
+}
+
+static int ice_rx_queue_intr_disable(struct rte_eth_dev *dev,
+                                    uint16_t queue_id)
+{
+       struct rte_pci_device *pci_dev = ICE_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct ice_hw *hw = ICE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       uint16_t msix_intr;
+
+       msix_intr = intr_handle->intr_vec[queue_id];
+
+       ICE_WRITE_REG(hw, GLINT_DYN_CTL(msix_intr), GLINT_DYN_CTL_WB_ON_ITR_M);
+
+       return 0;
+}
+
 static int
 ice_vsi_vlan_pvid_set(struct ice_vsi *vsi, struct ice_vsi_vlan_pvid_info *info)
 {