net/i40e: fix binding interrupt without MSI-X vector
[dpdk.git] / drivers / net / i40e / i40e_ethdev.c
index 03ddb8e..0c32e45 100644 (file)
@@ -26,6 +26,7 @@
 #include <rte_dev.h>
 #include <rte_tailq.h>
 #include <rte_hash_crc.h>
+#include <rte_bitmap.h>
 
 #include "i40e_logs.h"
 #include "base/i40e_prototype.h"
@@ -398,18 +399,6 @@ static void i40e_tunnel_filter_restore(struct i40e_pf *pf);
 static void i40e_filter_restore(struct i40e_pf *pf);
 static void i40e_notify_all_vfs_link_status(struct rte_eth_dev *dev);
 
-int i40e_logtype_init;
-int i40e_logtype_driver;
-#ifdef RTE_LIBRTE_I40E_DEBUG_RX
-int i40e_logtype_rx;
-#endif
-#ifdef RTE_LIBRTE_I40E_DEBUG_TX
-int i40e_logtype_tx;
-#endif
-#ifdef RTE_LIBRTE_I40E_DEBUG_TX_FREE
-int i40e_logtype_tx_free;
-#endif
-
 static const char *const valid_keys[] = {
        ETH_I40E_FLOATING_VEB_ARG,
        ETH_I40E_FLOATING_VEB_LIST_ARG,
@@ -443,6 +432,7 @@ static const struct rte_pci_id pci_id_i40e_map[] = {
        { RTE_PCI_DEVICE(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_X710_N3000) },
        { RTE_PCI_DEVICE(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_XXV710_N3000) },
        { RTE_PCI_DEVICE(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_10G_BASE_T_BC) },
+       { RTE_PCI_DEVICE(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_5G_BASE_T_BC) },
        { RTE_PCI_DEVICE(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_10G_B) },
        { RTE_PCI_DEVICE(I40E_INTEL_VENDOR_ID, I40E_DEV_ID_10G_SFP) },
        { .vendor_id = 0, /* sentinel */ },
@@ -1056,8 +1046,15 @@ static int
 i40e_init_fdir_filter_list(struct rte_eth_dev *dev)
 {
        struct i40e_pf *pf = I40E_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
        struct i40e_fdir_info *fdir_info = &pf->fdir;
        char fdir_hash_name[RTE_HASH_NAMESIZE];
+       uint32_t alloc = hw->func_caps.fd_filters_guaranteed;
+       uint32_t best = hw->func_caps.fd_filters_best_effort;
+       struct rte_bitmap *bmp = NULL;
+       uint32_t bmp_size;
+       void *mem = NULL;
+       uint32_t i = 0;
        int ret;
 
        struct rte_hash_parameters fdir_hash_params = {
@@ -1078,6 +1075,7 @@ i40e_init_fdir_filter_list(struct rte_eth_dev *dev)
                PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
                return -EINVAL;
        }
+
        fdir_info->hash_map = rte_zmalloc("i40e_fdir_hash_map",
                                          sizeof(struct i40e_fdir_filter *) *
                                          I40E_MAX_FDIR_FILTER_NUM,
@@ -1088,8 +1086,74 @@ i40e_init_fdir_filter_list(struct rte_eth_dev *dev)
                ret = -ENOMEM;
                goto err_fdir_hash_map_alloc;
        }
+
+       fdir_info->fdir_filter_array = rte_zmalloc("fdir_filter",
+                       sizeof(struct i40e_fdir_filter) *
+                       I40E_MAX_FDIR_FILTER_NUM,
+                       0);
+
+       if (!fdir_info->fdir_filter_array) {
+               PMD_INIT_LOG(ERR,
+                            "Failed to allocate memory for fdir filter array!");
+               ret = -ENOMEM;
+               goto err_fdir_filter_array_alloc;
+       }
+
+       fdir_info->fdir_space_size = alloc + best;
+       fdir_info->fdir_actual_cnt = 0;
+       fdir_info->fdir_guarantee_total_space = alloc;
+       fdir_info->fdir_guarantee_free_space =
+               fdir_info->fdir_guarantee_total_space;
+
+       PMD_DRV_LOG(INFO, "FDIR guarantee space: %u, best_effort space %u.", alloc, best);
+
+       fdir_info->fdir_flow_pool.pool =
+                       rte_zmalloc("i40e_fdir_entry",
+                               sizeof(struct i40e_fdir_entry) *
+                               fdir_info->fdir_space_size,
+                               0);
+
+       if (!fdir_info->fdir_flow_pool.pool) {
+               PMD_INIT_LOG(ERR,
+                            "Failed to allocate memory for bitmap flow!");
+               ret = -ENOMEM;
+               goto err_fdir_bitmap_flow_alloc;
+       }
+
+       for (i = 0; i < fdir_info->fdir_space_size; i++)
+               fdir_info->fdir_flow_pool.pool[i].idx = i;
+
+       bmp_size =
+               rte_bitmap_get_memory_footprint(fdir_info->fdir_space_size);
+       mem = rte_zmalloc("fdir_bmap", bmp_size, RTE_CACHE_LINE_SIZE);
+       if (mem == NULL) {
+               PMD_INIT_LOG(ERR,
+                            "Failed to allocate memory for fdir bitmap!");
+               ret = -ENOMEM;
+               goto err_fdir_mem_alloc;
+       }
+       bmp = rte_bitmap_init(fdir_info->fdir_space_size, mem, bmp_size);
+       if (bmp == NULL) {
+               PMD_INIT_LOG(ERR,
+                            "Failed to initialization fdir bitmap!");
+               ret = -ENOMEM;
+               goto err_fdir_bmp_alloc;
+       }
+       for (i = 0; i < fdir_info->fdir_space_size; i++)
+               rte_bitmap_set(bmp, i);
+
+       fdir_info->fdir_flow_pool.bitmap = bmp;
+
        return 0;
 
+err_fdir_bmp_alloc:
+       rte_free(mem);
+err_fdir_mem_alloc:
+       rte_free(fdir_info->fdir_flow_pool.pool);
+err_fdir_bitmap_flow_alloc:
+       rte_free(fdir_info->fdir_filter_array);
+err_fdir_filter_array_alloc:
+       rte_free(fdir_info->hash_map);
 err_fdir_hash_map_alloc:
        rte_hash_free(fdir_info->hash_table);
 
@@ -1112,6 +1176,30 @@ i40e_init_customized_info(struct i40e_pf *pf)
        pf->esp_support = false;
 }
 
+static void
+i40e_init_filter_invalidation(struct i40e_pf *pf)
+{
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       struct i40e_fdir_info *fdir_info = &pf->fdir;
+       uint32_t glqf_ctl_reg = 0;
+
+       glqf_ctl_reg = i40e_read_rx_ctl(hw, I40E_GLQF_CTL);
+       if (!pf->support_multi_driver) {
+               fdir_info->fdir_invalprio = 1;
+               glqf_ctl_reg |= I40E_GLQF_CTL_INVALPRIO_MASK;
+               PMD_DRV_LOG(INFO, "FDIR INVALPRIO set to guaranteed first");
+               i40e_write_rx_ctl(hw, I40E_GLQF_CTL, glqf_ctl_reg);
+       } else {
+               if (glqf_ctl_reg & I40E_GLQF_CTL_INVALPRIO_MASK) {
+                       fdir_info->fdir_invalprio = 1;
+                       PMD_DRV_LOG(INFO, "FDIR INVALPRIO is: guaranteed first");
+               } else {
+                       fdir_info->fdir_invalprio = 0;
+                       PMD_DRV_LOG(INFO, "FDIR INVALPRIO is: shared first");
+               }
+       }
+}
+
 void
 i40e_init_queue_region_conf(struct rte_eth_dev *dev)
 {
@@ -1656,12 +1744,18 @@ eth_i40e_dev_init(struct rte_eth_dev *dev, void *init_params __rte_unused)
        /* initialize mirror rule list */
        TAILQ_INIT(&pf->mirror_list);
 
+       /* initialize RSS rule list */
+       TAILQ_INIT(&pf->rss_config_list);
+
        /* initialize Traffic Manager configuration */
        i40e_tm_conf_init(dev);
 
        /* Initialize customized information */
        i40e_init_customized_info(pf);
 
+       /* Initialize the filter invalidation configuration */
+       i40e_init_filter_invalidation(pf);
+
        ret = i40e_init_ethtype_filter_list(dev);
        if (ret < 0)
                goto err_init_ethtype_filter_list;
@@ -1675,7 +1769,7 @@ eth_i40e_dev_init(struct rte_eth_dev *dev, void *init_params __rte_unused)
        /* initialize queue region configuration */
        i40e_init_queue_region_conf(dev);
 
-       /* initialize rss configuration from rte_flow */
+       /* initialize RSS configuration from rte_flow */
        memset(&pf->rss_info, 0,
                sizeof(struct i40e_rte_flow_rss_conf));
 
@@ -1757,16 +1851,30 @@ i40e_rm_fdir_filter_list(struct i40e_pf *pf)
        struct i40e_fdir_info *fdir_info;
 
        fdir_info = &pf->fdir;
-       /* Remove all flow director rules and hash */
+
+       /* Remove all flow director rules */
+       while ((p_fdir = TAILQ_FIRST(&fdir_info->fdir_list)))
+               TAILQ_REMOVE(&fdir_info->fdir_list, p_fdir, rules);
+}
+
+static void
+i40e_fdir_memory_cleanup(struct i40e_pf *pf)
+{
+       struct i40e_fdir_info *fdir_info;
+
+       fdir_info = &pf->fdir;
+
+       /* flow director memory cleanup */
        if (fdir_info->hash_map)
                rte_free(fdir_info->hash_map);
        if (fdir_info->hash_table)
                rte_hash_free(fdir_info->hash_table);
-
-       while ((p_fdir = TAILQ_FIRST(&fdir_info->fdir_list))) {
-               TAILQ_REMOVE(&fdir_info->fdir_list, p_fdir, rules);
-               rte_free(p_fdir);
-       }
+       if (fdir_info->fdir_flow_pool.bitmap)
+               rte_bitmap_free(fdir_info->fdir_flow_pool.bitmap);
+       if (fdir_info->fdir_flow_pool.pool)
+               rte_free(fdir_info->fdir_flow_pool.pool);
+       if (fdir_info->fdir_filter_array)
+               rte_free(fdir_info->fdir_filter_array);
 }
 
 void i40e_flex_payload_reg_set_default(struct i40e_hw *hw)
@@ -2009,7 +2117,7 @@ __vsi_queues_bind_intr(struct i40e_vsi *vsi, uint16_t msix_vect,
        I40E_WRITE_FLUSH(hw);
 }
 
-void
+int
 i40e_vsi_queues_bind_intr(struct i40e_vsi *vsi, uint16_t itr_idx)
 {
        struct rte_eth_dev *dev = vsi->adapter->eth_dev;
@@ -2029,10 +2137,14 @@ i40e_vsi_queues_bind_intr(struct i40e_vsi *vsi, uint16_t itr_idx)
 
        /* VF bind interrupt */
        if (vsi->type == I40E_VSI_SRIOV) {
+               if (vsi->nb_msix == 0) {
+                       PMD_DRV_LOG(ERR, "No msix resource");
+                       return -EINVAL;
+               }
                __vsi_queues_bind_intr(vsi, msix_vect,
                                       vsi->base_queue, vsi->nb_qps,
                                       itr_idx);
-               return;
+               return 0;
        }
 
        /* PF & VMDq bind interrupt */
@@ -2049,7 +2161,10 @@ i40e_vsi_queues_bind_intr(struct i40e_vsi *vsi, uint16_t itr_idx)
        }
 
        for (i = 0; i < vsi->nb_used_qps; i++) {
-               if (nb_msix <= 1) {
+               if (vsi->nb_msix == 0) {
+                       PMD_DRV_LOG(ERR, "No msix resource");
+                       return -EINVAL;
+               } else if (nb_msix <= 1) {
                        if (!rte_intr_allow_others(intr_handle))
                                /* allow to share MISC_VEC_ID */
                                msix_vect = I40E_MISC_VEC_ID;
@@ -2074,9 +2189,11 @@ i40e_vsi_queues_bind_intr(struct i40e_vsi *vsi, uint16_t itr_idx)
                msix_vect++;
                nb_msix--;
        }
+
+       return 0;
 }
 
-static void
+void
 i40e_vsi_enable_queues_intr(struct i40e_vsi *vsi)
 {
        struct rte_eth_dev *dev = vsi->adapter->eth_dev;
@@ -2103,7 +2220,7 @@ i40e_vsi_enable_queues_intr(struct i40e_vsi *vsi)
        I40E_WRITE_FLUSH(hw);
 }
 
-static void
+void
 i40e_vsi_disable_queues_intr(struct i40e_vsi *vsi)
 {
        struct rte_eth_dev *dev = vsi->adapter->eth_dev;
@@ -2276,6 +2393,7 @@ i40e_dev_start(struct rte_eth_dev *dev)
        struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
        uint32_t intr_vector = 0;
        struct i40e_vsi *vsi;
+       uint16_t nb_rxq, nb_txq;
 
        hw->adapter_stopped = 0;
 
@@ -2307,35 +2425,38 @@ i40e_dev_start(struct rte_eth_dev *dev)
        ret = i40e_dev_rxtx_init(pf);
        if (ret != I40E_SUCCESS) {
                PMD_DRV_LOG(ERR, "Failed to init rx/tx queues");
-               goto err_up;
+               return ret;
        }
 
        /* Map queues with MSIX interrupt */
        main_vsi->nb_used_qps = dev->data->nb_rx_queues -
                pf->nb_cfg_vmdq_vsi * RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM;
-       i40e_vsi_queues_bind_intr(main_vsi, I40E_ITR_INDEX_DEFAULT);
+       ret = i40e_vsi_queues_bind_intr(main_vsi, I40E_ITR_INDEX_DEFAULT);
+       if (ret < 0)
+               return ret;
        i40e_vsi_enable_queues_intr(main_vsi);
 
        /* Map VMDQ VSI queues with MSIX interrupt */
        for (i = 0; i < pf->nb_cfg_vmdq_vsi; i++) {
                pf->vmdq[i].vsi->nb_used_qps = RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM;
-               i40e_vsi_queues_bind_intr(pf->vmdq[i].vsi,
-                                         I40E_ITR_INDEX_DEFAULT);
+               ret = i40e_vsi_queues_bind_intr(pf->vmdq[i].vsi,
+                                               I40E_ITR_INDEX_DEFAULT);
+               if (ret < 0)
+                       return ret;
                i40e_vsi_enable_queues_intr(pf->vmdq[i].vsi);
        }
 
-       /* enable FDIR MSIX interrupt */
-       if (pf->fdir.fdir_vsi) {
-               i40e_vsi_queues_bind_intr(pf->fdir.fdir_vsi,
-                                         I40E_ITR_INDEX_NONE);
-               i40e_vsi_enable_queues_intr(pf->fdir.fdir_vsi);
+       /* Enable all queues which have been configured */
+       for (nb_rxq = 0; nb_rxq < dev->data->nb_rx_queues; nb_rxq++) {
+               ret = i40e_dev_rx_queue_start(dev, nb_rxq);
+               if (ret)
+                       goto rx_err;
        }
 
-       /* Enable all queues which have been configured */
-       ret = i40e_dev_switch_queues(pf, TRUE);
-       if (ret != I40E_SUCCESS) {
-               PMD_DRV_LOG(ERR, "Failed to enable VSI");
-               goto err_up;
+       for (nb_txq = 0; nb_txq < dev->data->nb_tx_queues; nb_txq++) {
+               ret = i40e_dev_tx_queue_start(dev, nb_txq);
+               if (ret)
+                       goto tx_err;
        }
 
        /* Enable receiving broadcast packets */
@@ -2365,7 +2486,7 @@ i40e_dev_start(struct rte_eth_dev *dev)
                ret = i40e_aq_set_lb_modes(hw, dev->data->dev_conf.lpbk_mode, NULL);
                if (ret != I40E_SUCCESS) {
                        PMD_DRV_LOG(ERR, "fail to set loopback link");
-                       goto err_up;
+                       goto tx_err;
                }
        }
 
@@ -2373,7 +2494,7 @@ i40e_dev_start(struct rte_eth_dev *dev)
        ret = i40e_apply_link_speed(dev);
        if (I40E_SUCCESS != ret) {
                PMD_DRV_LOG(ERR, "Fail to apply link setting");
-               goto err_up;
+               goto tx_err;
        }
 
        if (!rte_intr_allow_others(intr_handle)) {
@@ -2416,9 +2537,12 @@ i40e_dev_start(struct rte_eth_dev *dev)
 
        return I40E_SUCCESS;
 
-err_up:
-       i40e_dev_switch_queues(pf, FALSE);
-       i40e_dev_clear_queues(dev);
+tx_err:
+       for (i = 0; i < nb_txq; i++)
+               i40e_dev_tx_queue_stop(dev, i);
+rx_err:
+       for (i = 0; i < nb_rxq; i++)
+               i40e_dev_rx_queue_stop(dev, i);
 
        return ret;
 }
@@ -2442,7 +2566,11 @@ i40e_dev_stop(struct rte_eth_dev *dev)
        }
 
        /* Disable all queues */
-       i40e_dev_switch_queues(pf, FALSE);
+       for (i = 0; i < dev->data->nb_tx_queues; i++)
+               i40e_dev_tx_queue_stop(dev, i);
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++)
+               i40e_dev_rx_queue_stop(dev, i);
 
        /* un-map queues with interrupt registers */
        i40e_vsi_disable_queues_intr(main_vsi);
@@ -2453,10 +2581,6 @@ i40e_dev_stop(struct rte_eth_dev *dev)
                i40e_vsi_queues_unbind_intr(pf->vmdq[i].vsi);
        }
 
-       if (pf->fdir.fdir_vsi) {
-               i40e_vsi_queues_unbind_intr(pf->fdir.fdir_vsi);
-               i40e_vsi_disable_queues_intr(pf->fdir.fdir_vsi);
-       }
        /* Clear all queues and release memory */
        i40e_dev_clear_queues(dev);
 
@@ -2612,9 +2736,14 @@ i40e_dev_close(struct rte_eth_dev *dev)
        /* Remove all flows */
        while ((p_flow = TAILQ_FIRST(&pf->flow_list))) {
                TAILQ_REMOVE(&pf->flow_list, p_flow, node);
-               rte_free(p_flow);
+               /* Do not free FDIR flows since they are static allocated */
+               if (p_flow->filter_type != RTE_ETH_FILTER_FDIR)
+                       rte_free(p_flow);
        }
 
+       /* release the fdir static allocated memory */
+       i40e_fdir_memory_cleanup(pf);
+
        /* Remove all Traffic Manager configuration */
        i40e_tm_conf_uninit(dev);
 
@@ -3859,11 +3988,6 @@ i40e_vlan_offload_set(struct rte_eth_dev *dev, int mask)
        struct i40e_vsi *vsi = pf->main_vsi;
        struct rte_eth_rxmode *rxmode;
 
-       if (mask & ETH_QINQ_STRIP_MASK) {
-               PMD_DRV_LOG(ERR, "Strip qinq is not supported.");
-               return -ENOTSUP;
-       }
-
        rxmode = &dev->data->dev_conf.rxmode;
        if (mask & ETH_VLAN_FILTER_MASK) {
                if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
@@ -4931,6 +5055,7 @@ i40e_res_pool_free(struct i40e_res_pool_info *pool,
 {
        struct pool_entry *entry, *next, *prev, *valid_entry = NULL;
        uint32_t pool_offset;
+       uint16_t len;
        int insert;
 
        if (pool == NULL) {
@@ -4969,12 +5094,13 @@ i40e_res_pool_free(struct i40e_res_pool_info *pool,
        }
 
        insert = 0;
+       len = valid_entry->len;
        /* Try to merge with next one*/
        if (next != NULL) {
                /* Merge with next one */
-               if (valid_entry->base + valid_entry->len == next->base) {
+               if (valid_entry->base + len == next->base) {
                        next->base = valid_entry->base;
-                       next->len += valid_entry->len;
+                       next->len += len;
                        rte_free(valid_entry);
                        valid_entry = next;
                        insert = 1;
@@ -4984,13 +5110,15 @@ i40e_res_pool_free(struct i40e_res_pool_info *pool,
        if (prev != NULL) {
                /* Merge with previous one */
                if (prev->base + prev->len == valid_entry->base) {
-                       prev->len += valid_entry->len;
+                       prev->len += len;
                        /* If it merge with next one, remove next node */
                        if (insert == 1) {
                                LIST_REMOVE(valid_entry, next);
                                rte_free(valid_entry);
+                               valid_entry = NULL;
                        } else {
                                rte_free(valid_entry);
+                               valid_entry = NULL;
                                insert = 1;
                        }
                }
@@ -5006,8 +5134,8 @@ i40e_res_pool_free(struct i40e_res_pool_info *pool,
                        LIST_INSERT_HEAD(&pool->free_list, valid_entry, next);
        }
 
-       pool->num_free += valid_entry->len;
-       pool->num_alloc -= valid_entry->len;
+       pool->num_free += len;
+       pool->num_alloc -= len;
 
        return 0;
 }
@@ -6278,33 +6406,6 @@ i40e_switch_tx_queue(struct i40e_hw *hw, uint16_t q_idx, bool on)
        return I40E_SUCCESS;
 }
 
-/* Swith on or off the tx queues */
-static int
-i40e_dev_switch_tx_queues(struct i40e_pf *pf, bool on)
-{
-       struct rte_eth_dev_data *dev_data = pf->dev_data;
-       struct i40e_tx_queue *txq;
-       struct rte_eth_dev *dev = pf->adapter->eth_dev;
-       uint16_t i;
-       int ret;
-
-       for (i = 0; i < dev_data->nb_tx_queues; i++) {
-               txq = dev_data->tx_queues[i];
-               /* Don't operate the queue if not configured or
-                * if starting only per queue */
-               if (!txq || !txq->q_set || (on && txq->tx_deferred_start))
-                       continue;
-               if (on)
-                       ret = i40e_dev_tx_queue_start(dev, i);
-               else
-                       ret = i40e_dev_tx_queue_stop(dev, i);
-               if ( ret != I40E_SUCCESS)
-                       return ret;
-       }
-
-       return I40E_SUCCESS;
-}
-
 int
 i40e_switch_rx_queue(struct i40e_hw *hw, uint16_t q_idx, bool on)
 {
@@ -6356,59 +6457,6 @@ i40e_switch_rx_queue(struct i40e_hw *hw, uint16_t q_idx, bool on)
 
        return I40E_SUCCESS;
 }
-/* Switch on or off the rx queues */
-static int
-i40e_dev_switch_rx_queues(struct i40e_pf *pf, bool on)
-{
-       struct rte_eth_dev_data *dev_data = pf->dev_data;
-       struct i40e_rx_queue *rxq;
-       struct rte_eth_dev *dev = pf->adapter->eth_dev;
-       uint16_t i;
-       int ret;
-
-       for (i = 0; i < dev_data->nb_rx_queues; i++) {
-               rxq = dev_data->rx_queues[i];
-               /* Don't operate the queue if not configured or
-                * if starting only per queue */
-               if (!rxq || !rxq->q_set || (on && rxq->rx_deferred_start))
-                       continue;
-               if (on)
-                       ret = i40e_dev_rx_queue_start(dev, i);
-               else
-                       ret = i40e_dev_rx_queue_stop(dev, i);
-               if (ret != I40E_SUCCESS)
-                       return ret;
-       }
-
-       return I40E_SUCCESS;
-}
-
-/* Switch on or off all the rx/tx queues */
-int
-i40e_dev_switch_queues(struct i40e_pf *pf, bool on)
-{
-       int ret;
-
-       if (on) {
-               /* enable rx queues before enabling tx queues */
-               ret = i40e_dev_switch_rx_queues(pf, on);
-               if (ret) {
-                       PMD_DRV_LOG(ERR, "Failed to switch rx queues");
-                       return ret;
-               }
-               ret = i40e_dev_switch_tx_queues(pf, on);
-       } else {
-               /* Stop tx queues before stopping rx queues */
-               ret = i40e_dev_switch_tx_queues(pf, on);
-               if (ret) {
-                       PMD_DRV_LOG(ERR, "Failed to switch tx queues");
-                       return ret;
-               }
-               ret = i40e_dev_switch_rx_queues(pf, on);
-       }
-
-       return ret;
-}
 
 /* Initialize VSI for TX */
 static int
@@ -8014,6 +8062,13 @@ i40e_dev_tunnel_filter_set(struct i40e_pf *pf,
 #define I40E_TR_GRE_KEY_MASK                   0x400
 #define I40E_TR_GRE_KEY_WITH_XSUM_MASK         0x800
 #define I40E_TR_GRE_NO_KEY_MASK                        0x8000
+#define I40E_AQC_REPLACE_CLOUD_CMD_INPUT_PORT_TR_WORD0 0x49
+#define I40E_AQC_REPLACE_CLOUD_CMD_INPUT_DIRECTION_WORD0 0x41
+#define I40E_AQC_REPLACE_CLOUD_CMD_INPUT_INGRESS_WORD0 0x80
+#define I40E_DIRECTION_INGRESS_KEY             0x8000
+#define I40E_TR_L4_TYPE_TCP                    0x2
+#define I40E_TR_L4_TYPE_UDP                    0x4
+#define I40E_TR_L4_TYPE_SCTP                   0x8
 
 static enum
 i40e_status_code i40e_replace_mpls_l1_filter(struct i40e_pf *pf)
@@ -8312,6 +8367,132 @@ i40e_status_code i40e_replace_gtp_cloud_filter(struct i40e_pf *pf)
        return status;
 }
 
+static enum i40e_status_code
+i40e_replace_port_l1_filter(struct i40e_pf *pf,
+                           enum i40e_l4_port_type l4_port_type)
+{
+       struct i40e_aqc_replace_cloud_filters_cmd_buf  filter_replace_buf;
+       struct i40e_aqc_replace_cloud_filters_cmd  filter_replace;
+       enum i40e_status_code status = I40E_SUCCESS;
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       struct rte_eth_dev *dev = ((struct i40e_adapter *)hw->back)->eth_dev;
+
+       if (pf->support_multi_driver) {
+               PMD_DRV_LOG(ERR, "Replace l1 filter is not supported.");
+               return I40E_NOT_SUPPORTED;
+       }
+
+       memset(&filter_replace, 0,
+              sizeof(struct i40e_aqc_replace_cloud_filters_cmd));
+       memset(&filter_replace_buf, 0,
+              sizeof(struct i40e_aqc_replace_cloud_filters_cmd_buf));
+
+       /* create L1 filter */
+       if (l4_port_type == I40E_L4_PORT_TYPE_SRC) {
+               filter_replace.old_filter_type =
+                       I40E_AQC_REPLACE_CLOUD_CMD_INPUT_FV_TUNNLE_KEY;
+               filter_replace.new_filter_type = I40E_AQC_ADD_L1_FILTER_0X11;
+               filter_replace_buf.data[8] =
+                       I40E_AQC_REPLACE_CLOUD_CMD_INPUT_FV_SRC_PORT;
+       } else {
+               filter_replace.old_filter_type =
+                       I40E_AQC_REPLACE_CLOUD_CMD_INPUT_FV_STAG_IVLAN;
+               filter_replace.new_filter_type = I40E_AQC_ADD_L1_FILTER_0X10;
+               filter_replace_buf.data[8] =
+                       I40E_AQC_REPLACE_CLOUD_CMD_INPUT_FV_DST_PORT;
+       }
+
+       filter_replace.tr_bit = 0;
+       /* Prepare the buffer, 3 entries */
+       filter_replace_buf.data[0] =
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_DIRECTION_WORD0;
+       filter_replace_buf.data[0] |=
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_VALIDATED;
+       filter_replace_buf.data[2] = 0x00;
+       filter_replace_buf.data[3] =
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_INGRESS_WORD0;
+       filter_replace_buf.data[4] =
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_PORT_TR_WORD0;
+       filter_replace_buf.data[4] |=
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_VALIDATED;
+       filter_replace_buf.data[5] = 0x00;
+       filter_replace_buf.data[6] = I40E_TR_L4_TYPE_UDP |
+               I40E_TR_L4_TYPE_TCP |
+               I40E_TR_L4_TYPE_SCTP;
+       filter_replace_buf.data[7] = 0x00;
+       filter_replace_buf.data[8] |=
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_VALIDATED;
+       filter_replace_buf.data[9] = 0x00;
+       filter_replace_buf.data[10] = 0xFF;
+       filter_replace_buf.data[11] = 0xFF;
+
+       status = i40e_aq_replace_cloud_filters(hw, &filter_replace,
+                                              &filter_replace_buf);
+       if (!status && filter_replace.old_filter_type !=
+           filter_replace.new_filter_type)
+               PMD_DRV_LOG(WARNING, "i40e device %s changed cloud l1 type."
+                           " original: 0x%x, new: 0x%x",
+                           dev->device->name,
+                           filter_replace.old_filter_type,
+                           filter_replace.new_filter_type);
+
+       return status;
+}
+
+static enum i40e_status_code
+i40e_replace_port_cloud_filter(struct i40e_pf *pf,
+                              enum i40e_l4_port_type l4_port_type)
+{
+       struct i40e_aqc_replace_cloud_filters_cmd_buf  filter_replace_buf;
+       struct i40e_aqc_replace_cloud_filters_cmd  filter_replace;
+       enum i40e_status_code status = I40E_SUCCESS;
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       struct rte_eth_dev *dev = ((struct i40e_adapter *)hw->back)->eth_dev;
+
+       if (pf->support_multi_driver) {
+               PMD_DRV_LOG(ERR, "Replace cloud filter is not supported.");
+               return I40E_NOT_SUPPORTED;
+       }
+
+       memset(&filter_replace, 0,
+              sizeof(struct i40e_aqc_replace_cloud_filters_cmd));
+       memset(&filter_replace_buf, 0,
+              sizeof(struct i40e_aqc_replace_cloud_filters_cmd_buf));
+
+       if (l4_port_type == I40E_L4_PORT_TYPE_SRC) {
+               filter_replace.old_filter_type = I40E_AQC_ADD_CLOUD_FILTER_IIP;
+               filter_replace.new_filter_type =
+                       I40E_AQC_ADD_CLOUD_FILTER_0X11;
+               filter_replace_buf.data[4] = I40E_AQC_ADD_CLOUD_FILTER_0X11;
+       } else {
+               filter_replace.old_filter_type = I40E_AQC_ADD_CLOUD_FILTER_OIP;
+               filter_replace.new_filter_type =
+                       I40E_AQC_ADD_CLOUD_FILTER_0X10;
+               filter_replace_buf.data[4] = I40E_AQC_ADD_CLOUD_FILTER_0X10;
+       }
+
+       filter_replace.valid_flags = I40E_AQC_REPLACE_CLOUD_FILTER;
+       filter_replace.tr_bit = 0;
+       /* Prepare the buffer, 2 entries */
+       filter_replace_buf.data[0] = I40E_AQC_REPLACE_CLOUD_CMD_INPUT_FV_STAG;
+       filter_replace_buf.data[0] |=
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_VALIDATED;
+       filter_replace_buf.data[4] |=
+               I40E_AQC_REPLACE_CLOUD_CMD_INPUT_VALIDATED;
+       status = i40e_aq_replace_cloud_filters(hw, &filter_replace,
+                                              &filter_replace_buf);
+
+       if (!status && filter_replace.old_filter_type !=
+           filter_replace.new_filter_type)
+               PMD_DRV_LOG(WARNING, "i40e device %s changed cloud filter type."
+                           " original: 0x%x, new: 0x%x",
+                           dev->device->name,
+                           filter_replace.old_filter_type,
+                           filter_replace.new_filter_type);
+
+       return status;
+}
+
 int
 i40e_dev_consistent_tunnel_filter_set(struct i40e_pf *pf,
                      struct i40e_tunnel_filter_conf *tunnel_filter,
@@ -8459,6 +8640,62 @@ i40e_dev_consistent_tunnel_filter_set(struct i40e_pf *pf,
                pfilter->general_fields[0] = tunnel_filter->inner_vlan;
                pfilter->general_fields[1] = tunnel_filter->outer_vlan;
                big_buffer = 1;
+               break;
+       case I40E_CLOUD_TYPE_UDP:
+       case I40E_CLOUD_TYPE_TCP:
+       case I40E_CLOUD_TYPE_SCTP:
+               if (tunnel_filter->l4_port_type == I40E_L4_PORT_TYPE_SRC) {
+                       if (!pf->sport_replace_flag) {
+                               i40e_replace_port_l1_filter(pf,
+                                               tunnel_filter->l4_port_type);
+                               i40e_replace_port_cloud_filter(pf,
+                                               tunnel_filter->l4_port_type);
+                               pf->sport_replace_flag = 1;
+                       }
+                       teid_le = rte_cpu_to_le_32(tunnel_filter->tenant_id);
+                       pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X11_WORD0] =
+                               I40E_DIRECTION_INGRESS_KEY;
+
+                       if (tunnel_filter->tunnel_type == I40E_CLOUD_TYPE_UDP)
+                               pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X11_WORD1] =
+                                       I40E_TR_L4_TYPE_UDP;
+                       else if (tunnel_filter->tunnel_type == I40E_CLOUD_TYPE_TCP)
+                               pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X11_WORD1] =
+                                       I40E_TR_L4_TYPE_TCP;
+                       else
+                               pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X11_WORD1] =
+                                       I40E_TR_L4_TYPE_SCTP;
+
+                       pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X11_WORD2] =
+                               (teid_le >> 16) & 0xFFFF;
+                       big_buffer = 1;
+               } else {
+                       if (!pf->dport_replace_flag) {
+                               i40e_replace_port_l1_filter(pf,
+                                               tunnel_filter->l4_port_type);
+                               i40e_replace_port_cloud_filter(pf,
+                                               tunnel_filter->l4_port_type);
+                               pf->dport_replace_flag = 1;
+                       }
+                       teid_le = rte_cpu_to_le_32(tunnel_filter->tenant_id);
+                       pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X10_WORD0] =
+                               I40E_DIRECTION_INGRESS_KEY;
+
+                       if (tunnel_filter->tunnel_type == I40E_CLOUD_TYPE_UDP)
+                               pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X10_WORD1] =
+                                       I40E_TR_L4_TYPE_UDP;
+                       else if (tunnel_filter->tunnel_type == I40E_CLOUD_TYPE_TCP)
+                               pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X10_WORD1] =
+                                       I40E_TR_L4_TYPE_TCP;
+                       else
+                               pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X10_WORD1] =
+                                       I40E_TR_L4_TYPE_SCTP;
+
+                       pfilter->general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X10_WORD2] =
+                               (teid_le >> 16) & 0xFFFF;
+                       big_buffer = 1;
+               }
+
                break;
        default:
                /* Other tunnel types is not supported. */
@@ -8482,7 +8719,16 @@ i40e_dev_consistent_tunnel_filter_set(struct i40e_pf *pf,
        else if (tunnel_filter->tunnel_type == I40E_TUNNEL_TYPE_QINQ)
                pfilter->element.flags |=
                        I40E_AQC_ADD_CLOUD_FILTER_0X10;
-       else {
+       else if (tunnel_filter->tunnel_type == I40E_CLOUD_TYPE_UDP ||
+                tunnel_filter->tunnel_type == I40E_CLOUD_TYPE_TCP ||
+                tunnel_filter->tunnel_type == I40E_CLOUD_TYPE_SCTP) {
+               if (tunnel_filter->l4_port_type == I40E_L4_PORT_TYPE_SRC)
+                       pfilter->element.flags |=
+                               I40E_AQC_ADD_CLOUD_FILTER_0X11;
+               else
+                       pfilter->element.flags |=
+                               I40E_AQC_ADD_CLOUD_FILTER_0X10;
+       } else {
                val = i40e_dev_get_filter_type(tunnel_filter->filter_type,
                                                &pfilter->element.flags);
                if (val < 0) {
@@ -8845,7 +9091,7 @@ i40e_tunnel_filter_param_check(struct i40e_pf *pf,
 
 #define I40E_GL_PRS_FVBM_MSK_ENA 0x80000000
 #define I40E_GL_PRS_FVBM(_i)     (0x00269760 + ((_i) * 4))
-static int
+int
 i40e_dev_set_gre_key_len(struct i40e_hw *hw, uint8_t len)
 {
        struct i40e_pf *pf = &((struct i40e_adapter *)hw->back)->pf;
@@ -12151,7 +12397,7 @@ static int i40e_get_module_eeprom(struct rte_eth_dev *dev,
                }
                status = i40e_aq_get_phy_register(hw,
                                I40E_AQ_PHY_REG_ACCESS_EXTERNAL_MODULE,
-                               addr, offset, 1, &value, NULL);
+                               addr, 1, offset, &value, NULL);
                if (status)
                        return -EIO;
                data[i] = (uint8_t)value;
@@ -12328,14 +12574,16 @@ i40e_tunnel_filter_restore(struct i40e_pf *pf)
        }
 }
 
-/* Restore rss filter */
+/* Restore RSS filter */
 static inline void
 i40e_rss_filter_restore(struct i40e_pf *pf)
 {
-       struct i40e_rte_flow_rss_conf *conf =
-                                       &pf->rss_info;
-       if (conf->conf.queue_num)
-               i40e_config_rss_filter(pf, conf, TRUE);
+       struct i40e_rss_conf_list *list = &pf->rss_config_list;
+       struct i40e_rss_filter *filter;
+
+       TAILQ_FOREACH(filter, list, next) {
+               i40e_config_rss_filter(pf, &filter->rss_filter_info, TRUE);
+       }
 }
 
 static void
@@ -12945,45 +13193,303 @@ i40e_rss_conf_init(struct i40e_rte_flow_rss_conf *out,
        return 0;
 }
 
-int
-i40e_action_rss_same(const struct rte_flow_action_rss *comp,
-                    const struct rte_flow_action_rss *with)
+/* Write HENA register to enable hash */
+static int
+i40e_rss_hash_set(struct i40e_pf *pf, struct i40e_rte_flow_rss_conf *rss_conf)
 {
-       return (comp->func == with->func &&
-               comp->level == with->level &&
-               comp->types == with->types &&
-               comp->key_len == with->key_len &&
-               comp->queue_num == with->queue_num &&
-               !memcmp(comp->key, with->key, with->key_len) &&
-               !memcmp(comp->queue, with->queue,
-                       sizeof(*with->queue) * with->queue_num));
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       uint8_t *key = (void *)(uintptr_t)rss_conf->conf.key;
+       uint64_t hena;
+       int ret;
+
+       ret = i40e_set_rss_key(pf->main_vsi, key,
+                              rss_conf->conf.key_len);
+       if (ret)
+               return ret;
+
+       hena = i40e_config_hena(pf->adapter, rss_conf->conf.types);
+       i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (uint32_t)hena);
+       i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (uint32_t)(hena >> 32));
+       I40E_WRITE_FLUSH(hw);
+
+       return 0;
 }
 
-int
-i40e_config_rss_filter(struct i40e_pf *pf,
-               struct i40e_rte_flow_rss_conf *conf, bool add)
+/* Configure hash input set */
+static int
+i40e_rss_conf_hash_inset(struct i40e_pf *pf, uint64_t types)
 {
        struct i40e_hw *hw = I40E_PF_TO_HW(pf);
-       uint32_t i, lut = 0;
-       uint16_t j, num;
-       struct rte_eth_rss_conf rss_conf = {
-               .rss_key = conf->conf.key_len ?
-                       (void *)(uintptr_t)conf->conf.key : NULL,
-               .rss_key_len = conf->conf.key_len,
-               .rss_hf = conf->conf.types,
+       struct rte_eth_input_set_conf conf;
+       uint64_t mask0;
+       int ret = 0;
+       uint32_t j;
+       int i;
+       static const struct {
+               uint64_t type;
+               enum rte_eth_input_set_field field;
+       } inset_match_table[] = {
+               {ETH_RSS_FRAG_IPV4 | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP4},
+               {ETH_RSS_FRAG_IPV4 | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP4},
+               {ETH_RSS_FRAG_IPV4 | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
+               {ETH_RSS_FRAG_IPV4 | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
+
+               {ETH_RSS_NONFRAG_IPV4_TCP | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP4},
+               {ETH_RSS_NONFRAG_IPV4_TCP | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP4},
+               {ETH_RSS_NONFRAG_IPV4_TCP | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L4_TCP_SRC_PORT},
+               {ETH_RSS_NONFRAG_IPV4_TCP | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L4_TCP_DST_PORT},
+
+               {ETH_RSS_NONFRAG_IPV4_UDP | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP4},
+               {ETH_RSS_NONFRAG_IPV4_UDP | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP4},
+               {ETH_RSS_NONFRAG_IPV4_UDP | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L4_UDP_SRC_PORT},
+               {ETH_RSS_NONFRAG_IPV4_UDP | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L4_UDP_DST_PORT},
+
+               {ETH_RSS_NONFRAG_IPV4_SCTP | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP4},
+               {ETH_RSS_NONFRAG_IPV4_SCTP | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP4},
+               {ETH_RSS_NONFRAG_IPV4_SCTP | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L4_SCTP_SRC_PORT},
+               {ETH_RSS_NONFRAG_IPV4_SCTP | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L4_SCTP_DST_PORT},
+
+               {ETH_RSS_NONFRAG_IPV4_OTHER | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP4},
+               {ETH_RSS_NONFRAG_IPV4_OTHER | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP4},
+               {ETH_RSS_NONFRAG_IPV4_OTHER | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
+               {ETH_RSS_NONFRAG_IPV4_OTHER | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
+
+               {ETH_RSS_FRAG_IPV6 | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP6},
+               {ETH_RSS_FRAG_IPV6 | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP6},
+               {ETH_RSS_FRAG_IPV6 | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
+               {ETH_RSS_FRAG_IPV6 | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
+
+               {ETH_RSS_NONFRAG_IPV6_TCP | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP6},
+               {ETH_RSS_NONFRAG_IPV6_TCP | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP6},
+               {ETH_RSS_NONFRAG_IPV6_TCP | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L4_TCP_SRC_PORT},
+               {ETH_RSS_NONFRAG_IPV6_TCP | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L4_TCP_DST_PORT},
+
+               {ETH_RSS_NONFRAG_IPV6_UDP | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP6},
+               {ETH_RSS_NONFRAG_IPV6_UDP | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP6},
+               {ETH_RSS_NONFRAG_IPV6_UDP | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L4_UDP_SRC_PORT},
+               {ETH_RSS_NONFRAG_IPV6_UDP | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L4_UDP_DST_PORT},
+
+               {ETH_RSS_NONFRAG_IPV6_SCTP | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP6},
+               {ETH_RSS_NONFRAG_IPV6_SCTP | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP6},
+               {ETH_RSS_NONFRAG_IPV6_SCTP | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L4_SCTP_SRC_PORT},
+               {ETH_RSS_NONFRAG_IPV6_SCTP | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L4_SCTP_DST_PORT},
+
+               {ETH_RSS_NONFRAG_IPV6_OTHER | ETH_RSS_L3_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_L3_SRC_IP6},
+               {ETH_RSS_NONFRAG_IPV6_OTHER | ETH_RSS_L3_DST_ONLY,
+                       RTE_ETH_INPUT_SET_L3_DST_IP6},
+               {ETH_RSS_NONFRAG_IPV6_OTHER | ETH_RSS_L4_SRC_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
+               {ETH_RSS_NONFRAG_IPV6_OTHER | ETH_RSS_L4_DST_ONLY,
+                       RTE_ETH_INPUT_SET_UNKNOWN},
        };
-       struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
 
-       if (!add) {
-               if (i40e_action_rss_same(&rss_info->conf, &conf->conf)) {
-                       i40e_pf_disable_rss(pf);
-                       memset(rss_info, 0,
-                               sizeof(struct i40e_rte_flow_rss_conf));
+       mask0 = types & pf->adapter->flow_types_mask;
+       conf.op = RTE_ETH_INPUT_SET_SELECT;
+       conf.inset_size = 0;
+       for (i = RTE_ETH_FLOW_UNKNOWN + 1; i < RTE_ETH_FLOW_MAX; i++) {
+               if (mask0 & (1ULL << i)) {
+                       conf.flow_type = i;
+                       break;
+               }
+       }
+
+       for (j = 0; j < RTE_DIM(inset_match_table); j++) {
+               if ((types & inset_match_table[j].type) ==
+                   inset_match_table[j].type) {
+                       if (inset_match_table[j].field ==
+                           RTE_ETH_INPUT_SET_UNKNOWN)
+                               return -EINVAL;
+
+                       conf.field[conf.inset_size] =
+                               inset_match_table[j].field;
+                       conf.inset_size++;
+               }
+       }
+
+       if (conf.inset_size) {
+               ret = i40e_hash_filter_inset_select(hw, &conf);
+               if (ret)
+                       return ret;
+       }
+
+       return ret;
+}
+
+/* Look up the conflicted rule then mark it as invalid */
+static void
+i40e_rss_mark_invalid_rule(struct i40e_pf *pf,
+               struct i40e_rte_flow_rss_conf *conf)
+{
+       struct i40e_rss_filter *rss_item;
+       uint64_t rss_inset;
+
+       /* Clear input set bits before comparing the pctype */
+       rss_inset = ~(ETH_RSS_L3_SRC_ONLY | ETH_RSS_L3_DST_ONLY |
+               ETH_RSS_L4_SRC_ONLY | ETH_RSS_L4_DST_ONLY);
+
+       /* Look up the conflicted rule then mark it as invalid */
+       TAILQ_FOREACH(rss_item, &pf->rss_config_list, next) {
+               if (!rss_item->rss_filter_info.valid)
+                       continue;
+
+               if (conf->conf.queue_num &&
+                   rss_item->rss_filter_info.conf.queue_num)
+                       rss_item->rss_filter_info.valid = false;
+
+               if (conf->conf.types &&
+                   (rss_item->rss_filter_info.conf.types &
+                   rss_inset) ==
+                   (conf->conf.types & rss_inset))
+                       rss_item->rss_filter_info.valid = false;
+
+               if (conf->conf.func ==
+                   RTE_ETH_HASH_FUNCTION_SIMPLE_XOR &&
+                   rss_item->rss_filter_info.conf.func ==
+                   RTE_ETH_HASH_FUNCTION_SIMPLE_XOR)
+                       rss_item->rss_filter_info.valid = false;
+       }
+}
+
+/* Configure RSS hash function */
+static int
+i40e_rss_config_hash_function(struct i40e_pf *pf,
+               struct i40e_rte_flow_rss_conf *conf)
+{
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       uint32_t reg, i;
+       uint64_t mask0;
+       uint16_t j;
+
+       if (conf->conf.func == RTE_ETH_HASH_FUNCTION_SIMPLE_XOR) {
+               reg = i40e_read_rx_ctl(hw, I40E_GLQF_CTL);
+               if (!(reg & I40E_GLQF_CTL_HTOEP_MASK)) {
+                       PMD_DRV_LOG(DEBUG, "Hash function already set to Simple XOR");
+                       I40E_WRITE_FLUSH(hw);
+                       i40e_rss_mark_invalid_rule(pf, conf);
+
                        return 0;
                }
+               reg &= ~I40E_GLQF_CTL_HTOEP_MASK;
+
+               i40e_write_global_rx_ctl(hw, I40E_GLQF_CTL, reg);
+               I40E_WRITE_FLUSH(hw);
+               i40e_rss_mark_invalid_rule(pf, conf);
+       } else if (conf->conf.func ==
+                  RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ) {
+               mask0 = conf->conf.types & pf->adapter->flow_types_mask;
+
+               i40e_set_symmetric_hash_enable_per_port(hw, 1);
+               for (i = RTE_ETH_FLOW_UNKNOWN + 1; i < UINT64_BIT; i++) {
+                       if (mask0 & (1UL << i))
+                               break;
+               }
+
+               if (i == UINT64_BIT)
+                       return -EINVAL;
+
+               for (j = I40E_FILTER_PCTYPE_INVALID + 1;
+                    j < I40E_FILTER_PCTYPE_MAX; j++) {
+                       if (pf->adapter->pctypes_tbl[i] & (1ULL << j))
+                               i40e_write_global_rx_ctl(hw,
+                                       I40E_GLQF_HSYM(j),
+                                       I40E_GLQF_HSYM_SYMH_ENA_MASK);
+               }
+       }
+
+       return 0;
+}
+
+/* Enable RSS according to the configuration */
+static int
+i40e_rss_enable_hash(struct i40e_pf *pf,
+               struct i40e_rte_flow_rss_conf *conf)
+{
+       struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
+       struct i40e_rte_flow_rss_conf rss_conf;
+
+       if (!(conf->conf.types & pf->adapter->flow_types_mask))
+               return -ENOTSUP;
+
+       memset(&rss_conf, 0, sizeof(rss_conf));
+       rte_memcpy(&rss_conf, conf, sizeof(rss_conf));
+
+       /* Configure hash input set */
+       if (i40e_rss_conf_hash_inset(pf, conf->conf.types))
                return -EINVAL;
+
+       if (rss_conf.conf.key == NULL || rss_conf.conf.key_len <
+           (I40E_PFQF_HKEY_MAX_INDEX + 1) * sizeof(uint32_t)) {
+               /* Random default keys */
+               static uint32_t rss_key_default[] = {0x6b793944,
+                       0x23504cb5, 0x5bea75b6, 0x309f4f12, 0x3dc0a2b8,
+                       0x024ddcdf, 0x339b8ca0, 0x4c4af64a, 0x34fac605,
+                       0x55d85839, 0x3a58997d, 0x2ec938e1, 0x66031581};
+
+               rss_conf.conf.key = (uint8_t *)rss_key_default;
+               rss_conf.conf.key_len = (I40E_PFQF_HKEY_MAX_INDEX + 1) *
+                               sizeof(uint32_t);
+               PMD_DRV_LOG(INFO,
+                       "No valid RSS key config for i40e, using default\n");
        }
 
+       rss_conf.conf.types |= rss_info->conf.types;
+       i40e_rss_hash_set(pf, &rss_conf);
+
+       if (conf->conf.func == RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ)
+               i40e_rss_config_hash_function(pf, conf);
+
+       i40e_rss_mark_invalid_rule(pf, conf);
+
+       return 0;
+}
+
+/* Configure RSS queue region */
+static int
+i40e_rss_config_queue_region(struct i40e_pf *pf,
+               struct i40e_rte_flow_rss_conf *conf)
+{
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       uint32_t lut = 0;
+       uint16_t j, num;
+       uint32_t i;
+
        /* If both VMDQ and RSS enabled, not all of PF queues are configured.
         * It's necessary to calculate the actual PF queues that are configured.
         */
@@ -13013,60 +13519,213 @@ i40e_config_rss_filter(struct i40e_pf *pf,
                        I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i >> 2), lut);
        }
 
-       if ((rss_conf.rss_hf & pf->adapter->flow_types_mask) == 0) {
-               i40e_pf_disable_rss(pf);
-               return 0;
+       i40e_rss_mark_invalid_rule(pf, conf);
+
+       return 0;
+}
+
+/* Configure RSS hash function to default */
+static int
+i40e_rss_clear_hash_function(struct i40e_pf *pf,
+               struct i40e_rte_flow_rss_conf *conf)
+{
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       uint32_t i, reg;
+       uint64_t mask0;
+       uint16_t j;
+
+       if (conf->conf.func == RTE_ETH_HASH_FUNCTION_SIMPLE_XOR) {
+               reg = i40e_read_rx_ctl(hw, I40E_GLQF_CTL);
+               if (reg & I40E_GLQF_CTL_HTOEP_MASK) {
+                       PMD_DRV_LOG(DEBUG,
+                               "Hash function already set to Toeplitz");
+                       I40E_WRITE_FLUSH(hw);
+
+                       return 0;
+               }
+               reg |= I40E_GLQF_CTL_HTOEP_MASK;
+
+               i40e_write_global_rx_ctl(hw, I40E_GLQF_CTL, reg);
+               I40E_WRITE_FLUSH(hw);
+       } else if (conf->conf.func ==
+                  RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ) {
+               mask0 = conf->conf.types & pf->adapter->flow_types_mask;
+
+               for (i = RTE_ETH_FLOW_UNKNOWN + 1; i < UINT64_BIT; i++) {
+                       if (mask0 & (1UL << i))
+                               break;
+               }
+
+               if (i == UINT64_BIT)
+                       return -EINVAL;
+
+               for (j = I40E_FILTER_PCTYPE_INVALID + 1;
+                    j < I40E_FILTER_PCTYPE_MAX; j++) {
+                       if (pf->adapter->pctypes_tbl[i] & (1ULL << j))
+                               i40e_write_global_rx_ctl(hw,
+                                       I40E_GLQF_HSYM(j),
+                                       0);
+               }
        }
-       if (rss_conf.rss_key == NULL || rss_conf.rss_key_len <
-               (I40E_PFQF_HKEY_MAX_INDEX + 1) * sizeof(uint32_t)) {
-               /* Random default keys */
-               static uint32_t rss_key_default[] = {0x6b793944,
-                       0x23504cb5, 0x5bea75b6, 0x309f4f12, 0x3dc0a2b8,
-                       0x024ddcdf, 0x339b8ca0, 0x4c4af64a, 0x34fac605,
-                       0x55d85839, 0x3a58997d, 0x2ec938e1, 0x66031581};
 
-               rss_conf.rss_key = (uint8_t *)rss_key_default;
-               rss_conf.rss_key_len = (I40E_PFQF_HKEY_MAX_INDEX + 1) *
-                                                       sizeof(uint32_t);
-               PMD_DRV_LOG(INFO,
-                       "No valid RSS key config for i40e, using default\n");
+       return 0;
+}
+
+/* Disable RSS hash and configure default input set */
+static int
+i40e_rss_disable_hash(struct i40e_pf *pf,
+               struct i40e_rte_flow_rss_conf *conf)
+{
+       struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       struct i40e_rte_flow_rss_conf rss_conf;
+       uint32_t i;
+
+       memset(&rss_conf, 0, sizeof(rss_conf));
+       rte_memcpy(&rss_conf, conf, sizeof(rss_conf));
+
+       /* Disable RSS hash */
+       rss_conf.conf.types = rss_info->conf.types & ~(conf->conf.types);
+       i40e_rss_hash_set(pf, &rss_conf);
+
+       for (i = RTE_ETH_FLOW_IPV4; i <= RTE_ETH_FLOW_L2_PAYLOAD; i++) {
+               if (!(pf->adapter->flow_types_mask & (1ULL << i)) ||
+                   !(conf->conf.types & (1ULL << i)))
+                       continue;
+
+               /* Configure default input set */
+               struct rte_eth_input_set_conf input_conf = {
+                       .op = RTE_ETH_INPUT_SET_SELECT,
+                       .flow_type = i,
+                       .inset_size = 1,
+               };
+               input_conf.field[0] = RTE_ETH_INPUT_SET_DEFAULT;
+               i40e_hash_filter_inset_select(hw, &input_conf);
        }
 
-       i40e_hw_rss_hash_set(pf, &rss_conf);
+       rss_info->conf.types = rss_conf.conf.types;
 
-       if (i40e_rss_conf_init(rss_info, &conf->conf))
-               return -EINVAL;
+       i40e_rss_clear_hash_function(pf, conf);
 
        return 0;
 }
 
-RTE_INIT(i40e_init_log)
+/* Configure RSS queue region to default */
+static int
+i40e_rss_clear_queue_region(struct i40e_pf *pf)
 {
-       i40e_logtype_init = rte_log_register("pmd.net.i40e.init");
-       if (i40e_logtype_init >= 0)
-               rte_log_set_level(i40e_logtype_init, RTE_LOG_NOTICE);
-       i40e_logtype_driver = rte_log_register("pmd.net.i40e.driver");
-       if (i40e_logtype_driver >= 0)
-               rte_log_set_level(i40e_logtype_driver, RTE_LOG_NOTICE);
+       struct i40e_hw *hw = I40E_PF_TO_HW(pf);
+       struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
+       uint16_t queue[I40E_MAX_Q_PER_TC];
+       uint32_t num_rxq, i;
+       uint32_t lut = 0;
+       uint16_t j, num;
+
+       num_rxq = RTE_MIN(pf->dev_data->nb_rx_queues, I40E_MAX_Q_PER_TC);
 
+       for (j = 0; j < num_rxq; j++)
+               queue[j] = j;
+
+       /* If both VMDQ and RSS enabled, not all of PF queues are configured.
+        * It's necessary to calculate the actual PF queues that are configured.
+        */
+       if (pf->dev_data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_VMDQ_FLAG)
+               num = i40e_pf_calc_configured_queues_num(pf);
+       else
+               num = pf->dev_data->nb_rx_queues;
+
+       num = RTE_MIN(num, num_rxq);
+       PMD_DRV_LOG(INFO, "Max of contiguous %u PF queues are configured",
+                       num);
+
+       if (num == 0) {
+               PMD_DRV_LOG(ERR,
+                       "No PF queues are configured to enable RSS for port %u",
+                       pf->dev_data->port_id);
+               return -ENOTSUP;
+       }
+
+       /* Fill in redirection table */
+       for (i = 0, j = 0; i < hw->func_caps.rss_table_size; i++, j++) {
+               if (j == num)
+                       j = 0;
+               lut = (lut << 8) | (queue[j] & ((0x1 <<
+                       hw->func_caps.rss_table_entry_width) - 1));
+               if ((i & 3) == 3)
+                       I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i >> 2), lut);
+       }
+
+       rss_info->conf.queue_num = 0;
+       memset(&rss_info->conf.queue, 0, sizeof(uint16_t));
+
+       return 0;
+}
+
+int
+i40e_config_rss_filter(struct i40e_pf *pf,
+               struct i40e_rte_flow_rss_conf *conf, bool add)
+{
+       struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
+       struct rte_flow_action_rss update_conf = rss_info->conf;
+       int ret = 0;
+
+       if (add) {
+               if (conf->conf.queue_num) {
+                       /* Configure RSS queue region */
+                       ret = i40e_rss_config_queue_region(pf, conf);
+                       if (ret)
+                               return ret;
+
+                       update_conf.queue_num = conf->conf.queue_num;
+                       update_conf.queue = conf->conf.queue;
+               } else if (conf->conf.func ==
+                          RTE_ETH_HASH_FUNCTION_SIMPLE_XOR) {
+                       /* Configure hash function */
+                       ret = i40e_rss_config_hash_function(pf, conf);
+                       if (ret)
+                               return ret;
+
+                       update_conf.func = conf->conf.func;
+               } else {
+                       /* Configure hash enable and input set */
+                       ret = i40e_rss_enable_hash(pf, conf);
+                       if (ret)
+                               return ret;
+
+                       update_conf.types |= conf->conf.types;
+                       update_conf.key = conf->conf.key;
+                       update_conf.key_len = conf->conf.key_len;
+               }
+
+               /* Update RSS info in pf */
+               if (i40e_rss_conf_init(rss_info, &update_conf))
+                       return -EINVAL;
+       } else {
+               if (!conf->valid)
+                       return 0;
+
+               if (conf->conf.queue_num)
+                       i40e_rss_clear_queue_region(pf);
+               else if (conf->conf.func == RTE_ETH_HASH_FUNCTION_SIMPLE_XOR)
+                       i40e_rss_clear_hash_function(pf, conf);
+               else
+                       i40e_rss_disable_hash(pf, conf);
+       }
+
+       return 0;
+}
+
+RTE_LOG_REGISTER(i40e_logtype_init, pmd.net.i40e.init, NOTICE);
+RTE_LOG_REGISTER(i40e_logtype_driver, pmd.net.i40e.driver, NOTICE);
 #ifdef RTE_LIBRTE_I40E_DEBUG_RX
-       i40e_logtype_rx = rte_log_register("pmd.net.i40e.rx");
-       if (i40e_logtype_rx >= 0)
-               rte_log_set_level(i40e_logtype_rx, RTE_LOG_DEBUG);
+RTE_LOG_REGISTER(i40e_logtype_rx, pmd.net.i40e.rx, DEBUG);
 #endif
-
 #ifdef RTE_LIBRTE_I40E_DEBUG_TX
-       i40e_logtype_tx = rte_log_register("pmd.net.i40e.tx");
-       if (i40e_logtype_tx >= 0)
-               rte_log_set_level(i40e_logtype_tx, RTE_LOG_DEBUG);
+RTE_LOG_REGISTER(i40e_logtype_tx, pmd.net.i40e.tx, DEBUG);
 #endif
-
 #ifdef RTE_LIBRTE_I40E_DEBUG_TX_FREE
-       i40e_logtype_tx_free = rte_log_register("pmd.net.i40e.tx_free");
-       if (i40e_logtype_tx_free >= 0)
-               rte_log_set_level(i40e_logtype_tx_free, RTE_LOG_DEBUG);
+RTE_LOG_REGISTER(i40e_logtype_tx_free, pmd.net.i40e.tx_free, DEBUG);
 #endif
-}
 
 RTE_PMD_REGISTER_PARAM_STRING(net_i40e,
                              ETH_I40E_FLOATING_VEB_ARG "=1"