net/hns3: support Tx push quick doorbell for performance
[dpdk.git] / drivers / net / hns3 / hns3_ethdev.c
index ee47117..e515125 100644 (file)
@@ -16,9 +16,6 @@
 #include "hns3_dcb.h"
 #include "hns3_mp.h"
 
-#define HNS3_DEFAULT_PORT_CONF_BURST_SIZE      32
-#define HNS3_DEFAULT_PORT_CONF_QUEUES_NUM      1
-
 #define HNS3_SERVICE_INTERVAL          1000000 /* us */
 #define HNS3_SERVICE_QUICK_INTERVAL    10
 #define HNS3_INVALID_PVID              0xFFFF
@@ -105,6 +102,7 @@ static int hns3_remove_mc_addr(struct hns3_hw *hw,
 static int hns3_restore_fec(struct hns3_hw *hw);
 static int hns3_query_dev_fec_info(struct hns3_hw *hw);
 static int hns3_do_stop(struct hns3_adapter *hns);
+static int hns3_check_port_speed(struct hns3_hw *hw, uint32_t link_speeds);
 
 void hns3_ether_format_addr(char *buf, uint16_t size,
                            const struct rte_ether_addr *ether_addr)
@@ -255,8 +253,8 @@ static void
 hns3_clear_all_event_cause(struct hns3_hw *hw)
 {
        uint32_t vector0_int_stats;
-       vector0_int_stats = hns3_read_dev(hw, HNS3_VECTOR0_OTHER_INT_STS_REG);
 
+       vector0_int_stats = hns3_read_dev(hw, HNS3_VECTOR0_OTHER_INT_STS_REG);
        if (BIT(HNS3_VECTOR0_IMPRESET_INT_B) & vector0_int_stats)
                hns3_warn(hw, "Probe during IMP reset interrupt");
 
@@ -279,7 +277,7 @@ hns3_handle_mac_tnl(struct hns3_hw *hw)
        uint32_t status;
        int ret;
 
-       /* query and clear mac tnl interruptions */
+       /* query and clear mac tnl interrupt */
        hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_QUERY_MAC_TNL_INT, true);
        ret = hns3_cmd_send(hw, &desc, 1);
        if (ret) {
@@ -324,10 +322,8 @@ hns3_interrupt_handler(void *param)
                hns3_warn(hw, "received interrupt: vector0_int_stat:0x%x "
                          "ras_int_stat:0x%x cmdq_int_stat:0x%x",
                          vector0_int, ras_int, cmdq_int);
-               hns3_handle_msix_error(hns, &hw->reset.request);
-               hns3_handle_ras_error(hns, &hw->reset.request);
                hns3_handle_mac_tnl(hw);
-               hns3_schedule_reset(hns);
+               hns3_handle_error(hns);
        } else if (event_cause == HNS3_VECTOR0_EVENT_RST) {
                hns3_warn(hw, "received reset interrupt");
                hns3_schedule_reset(hns);
@@ -463,7 +459,7 @@ hns3_vlan_filter_configure(struct hns3_adapter *hns, uint16_t vlan_id, int on)
         * When port base vlan enabled, we use port base vlan as the vlan
         * filter condition. In this case, we don't update vlan filter table
         * when user add new vlan or remove exist vlan, just update the
-        * vlan list. The vlan id in vlan list will be writen in vlan filter
+        * vlan list. The vlan id in vlan list will be written in vlan filter
         * table until port base vlan disabled
         */
        if (hw->port_base_vlan_cfg.state == HNS3_PORT_BASE_VLAN_DISABLE) {
@@ -644,7 +640,8 @@ hns3_en_hw_strip_rxvtag(struct hns3_adapter *hns, bool enable)
 
        ret = hns3_set_vlan_rx_offload_cfg(hns, &rxvlan_cfg);
        if (ret) {
-               hns3_err(hw, "enable strip rx vtag failed, ret =%d", ret);
+               hns3_err(hw, "%s strip rx vtag failed, ret = %d.",
+                               enable ? "enable" : "disable", ret);
                return ret;
        }
 
@@ -2224,24 +2221,17 @@ hns3_check_mq_mode(struct rte_eth_dev *dev)
        int max_tc = 0;
        int i;
 
-       dcb_rx_conf = &dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
-       dcb_tx_conf = &dev->data->dev_conf.tx_adv_conf.dcb_tx_conf;
-
-       if (rx_mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
-               hns3_err(hw, "ETH_MQ_RX_VMDQ_DCB_RSS is not supported. "
-                        "rx_mq_mode = %d", rx_mq_mode);
-               return -EINVAL;
-       }
-
-       if (rx_mq_mode == ETH_MQ_RX_VMDQ_DCB ||
-           tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
-               hns3_err(hw, "ETH_MQ_RX_VMDQ_DCB and ETH_MQ_TX_VMDQ_DCB "
-                        "is not supported. rx_mq_mode = %d, tx_mq_mode = %d",
+       if ((rx_mq_mode & ETH_MQ_RX_VMDQ_FLAG) ||
+           (tx_mq_mode == ETH_MQ_TX_VMDQ_DCB ||
+            tx_mq_mode == ETH_MQ_TX_VMDQ_ONLY)) {
+               hns3_err(hw, "VMDQ is not supported, rx_mq_mode = %d, tx_mq_mode = %d.",
                         rx_mq_mode, tx_mq_mode);
-               return -EINVAL;
+               return -EOPNOTSUPP;
        }
 
-       if (rx_mq_mode == ETH_MQ_RX_DCB_RSS) {
+       dcb_rx_conf = &dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
+       dcb_tx_conf = &dev->data->dev_conf.tx_adv_conf.dcb_tx_conf;
+       if (rx_mq_mode & ETH_MQ_RX_DCB_FLAG) {
                if (dcb_rx_conf->nb_tcs > pf->tc_max) {
                        hns3_err(hw, "nb_tcs(%u) > max_tc(%u) driver supported.",
                                 dcb_rx_conf->nb_tcs, pf->tc_max);
@@ -2284,25 +2274,6 @@ hns3_check_mq_mode(struct rte_eth_dev *dev)
        return 0;
 }
 
-static int
-hns3_check_dcb_cfg(struct rte_eth_dev *dev)
-{
-       struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-       if (!hns3_dev_dcb_supported(hw)) {
-               hns3_err(hw, "this port does not support dcb configurations.");
-               return -EOPNOTSUPP;
-       }
-
-       if (hw->current_fc_status == HNS3_FC_STATUS_MAC_PAUSE) {
-               hns3_err(hw, "MAC pause enabled, cannot config dcb info.");
-               return -EOPNOTSUPP;
-       }
-
-       /* Check multiple queue mode */
-       return hns3_check_mq_mode(dev);
-}
-
 static int
 hns3_bind_ring_with_vector(struct hns3_hw *hw, uint16_t vector_id, bool en,
                           enum hns3_ring_type queue_type, uint16_t queue_id)
@@ -2438,6 +2409,70 @@ hns3_refresh_mtu(struct rte_eth_dev *dev, struct rte_eth_conf *conf)
        return 0;
 }
 
+static int
+hns3_setup_dcb(struct rte_eth_dev *dev)
+{
+       struct hns3_adapter *hns = dev->data->dev_private;
+       struct hns3_hw *hw = &hns->hw;
+       int ret;
+
+       if (!hns3_dev_dcb_supported(hw)) {
+               hns3_err(hw, "this port does not support dcb configurations.");
+               return -EOPNOTSUPP;
+       }
+
+       if (hw->current_fc_status == HNS3_FC_STATUS_MAC_PAUSE) {
+               hns3_err(hw, "MAC pause enabled, cannot config dcb info.");
+               return -EOPNOTSUPP;
+       }
+
+       ret = hns3_dcb_configure(hns);
+       if (ret)
+               hns3_err(hw, "failed to config dcb: %d", ret);
+
+       return ret;
+}
+
+static int
+hns3_check_link_speed(struct hns3_hw *hw, uint32_t link_speeds)
+{
+       int ret;
+
+       /*
+        * Some hardware doesn't support auto-negotiation, but users may not
+        * configure link_speeds (default 0), which means auto-negotiation.
+        * In this case, a warning message need to be printed, instead of
+        * an error.
+        */
+       if (link_speeds == ETH_LINK_SPEED_AUTONEG &&
+           hw->mac.support_autoneg == 0) {
+               hns3_warn(hw, "auto-negotiation is not supported, use default fixed speed!");
+               return 0;
+       }
+
+       if (link_speeds != ETH_LINK_SPEED_AUTONEG) {
+               ret = hns3_check_port_speed(hw, link_speeds);
+               if (ret)
+                       return ret;
+       }
+
+       return 0;
+}
+
+static int
+hns3_check_dev_conf(struct rte_eth_dev *dev)
+{
+       struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       struct rte_eth_conf *conf = &dev->data->dev_conf;
+       int ret;
+
+       ret = hns3_check_mq_mode(dev);
+       if (ret)
+               return ret;
+
+       return hns3_check_link_speed(hw, conf->link_speeds);
+}
+
 static int
 hns3_dev_configure(struct rte_eth_dev *dev)
 {
@@ -2463,24 +2498,20 @@ hns3_dev_configure(struct rte_eth_dev *dev)
         * work as usual. But these fake queues are imperceptible, and can not
         * be used by upper applications.
         */
-       if (!hns3_dev_indep_txrx_supported(hw)) {
-               ret = hns3_set_fake_rx_or_tx_queues(dev, nb_rx_q, nb_tx_q);
-               if (ret) {
-                       hns3_err(hw, "fail to set Rx/Tx fake queues, ret = %d.",
-                                ret);
-                       return ret;
-               }
+       ret = hns3_set_fake_rx_or_tx_queues(dev, nb_rx_q, nb_tx_q);
+       if (ret) {
+               hns3_err(hw, "fail to set Rx/Tx fake queues, ret = %d.", ret);
+               hw->cfg_max_queues = 0;
+               return ret;
        }
 
        hw->adapter_state = HNS3_NIC_CONFIGURING;
-       if (conf->link_speeds & ETH_LINK_SPEED_FIXED) {
-               hns3_err(hw, "setting link speed/duplex not supported");
-               ret = -EINVAL;
+       ret = hns3_check_dev_conf(dev);
+       if (ret)
                goto cfg_err;
-       }
 
        if ((uint32_t)mq_mode & ETH_MQ_RX_DCB_FLAG) {
-               ret = hns3_check_dcb_cfg(dev);
+               ret = hns3_setup_dcb(dev);
                if (ret)
                        goto cfg_err;
        }
@@ -2519,6 +2550,7 @@ hns3_dev_configure(struct rte_eth_dev *dev)
        return 0;
 
 cfg_err:
+       hw->cfg_max_queues = 0;
        (void)hns3_set_fake_rx_or_tx_queues(dev, 0, 0);
        hw->adapter_state = HNS3_NIC_INITIALIZED;
 
@@ -2617,6 +2649,67 @@ hns3_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
        return 0;
 }
 
+static uint32_t
+hns3_get_copper_port_speed_capa(uint32_t supported_speed)
+{
+       uint32_t speed_capa = 0;
+
+       if (supported_speed & HNS3_PHY_LINK_SPEED_10M_HD_BIT)
+               speed_capa |= ETH_LINK_SPEED_10M_HD;
+       if (supported_speed & HNS3_PHY_LINK_SPEED_10M_BIT)
+               speed_capa |= ETH_LINK_SPEED_10M;
+       if (supported_speed & HNS3_PHY_LINK_SPEED_100M_HD_BIT)
+               speed_capa |= ETH_LINK_SPEED_100M_HD;
+       if (supported_speed & HNS3_PHY_LINK_SPEED_100M_BIT)
+               speed_capa |= ETH_LINK_SPEED_100M;
+       if (supported_speed & HNS3_PHY_LINK_SPEED_1000M_BIT)
+               speed_capa |= ETH_LINK_SPEED_1G;
+
+       return speed_capa;
+}
+
+static uint32_t
+hns3_get_firber_port_speed_capa(uint32_t supported_speed)
+{
+       uint32_t speed_capa = 0;
+
+       if (supported_speed & HNS3_FIBER_LINK_SPEED_1G_BIT)
+               speed_capa |= ETH_LINK_SPEED_1G;
+       if (supported_speed & HNS3_FIBER_LINK_SPEED_10G_BIT)
+               speed_capa |= ETH_LINK_SPEED_10G;
+       if (supported_speed & HNS3_FIBER_LINK_SPEED_25G_BIT)
+               speed_capa |= ETH_LINK_SPEED_25G;
+       if (supported_speed & HNS3_FIBER_LINK_SPEED_40G_BIT)
+               speed_capa |= ETH_LINK_SPEED_40G;
+       if (supported_speed & HNS3_FIBER_LINK_SPEED_50G_BIT)
+               speed_capa |= ETH_LINK_SPEED_50G;
+       if (supported_speed & HNS3_FIBER_LINK_SPEED_100G_BIT)
+               speed_capa |= ETH_LINK_SPEED_100G;
+       if (supported_speed & HNS3_FIBER_LINK_SPEED_200G_BIT)
+               speed_capa |= ETH_LINK_SPEED_200G;
+
+       return speed_capa;
+}
+
+static uint32_t
+hns3_get_speed_capa(struct hns3_hw *hw)
+{
+       struct hns3_mac *mac = &hw->mac;
+       uint32_t speed_capa;
+
+       if (mac->media_type == HNS3_MEDIA_TYPE_COPPER)
+               speed_capa =
+                       hns3_get_copper_port_speed_capa(mac->supported_speed);
+       else
+               speed_capa =
+                       hns3_get_firber_port_speed_capa(mac->supported_speed);
+
+       if (mac->support_autoneg == 0)
+               speed_capa |= ETH_LINK_SPEED_FIXED;
+
+       return speed_capa;
+}
+
 int
 hns3_dev_infos_get(struct rte_eth_dev *eth_dev, struct rte_eth_dev_info *info)
 {
@@ -2688,6 +2781,7 @@ hns3_dev_infos_get(struct rte_eth_dev *eth_dev, struct rte_eth_dev_info *info)
                .nb_mtu_seg_max = hw->max_non_tso_bd_num,
        };
 
+       info->speed_capa = hns3_get_speed_capa(hw);
        info->default_rxconf = (struct rte_eth_rxconf) {
                .rx_free_thresh = HNS3_DEFAULT_RX_FREE_THRESH,
                /*
@@ -2703,8 +2797,6 @@ hns3_dev_infos_get(struct rte_eth_dev *eth_dev, struct rte_eth_dev_info *info)
                .offloads = 0,
        };
 
-       info->vmdq_queue_num = 0;
-
        info->reta_size = hw->rss_ind_tbl_size;
        info->hash_key_size = HNS3_RSS_KEY_SIZE;
        info->flow_type_rss_offloads = HNS3_ETH_RSS_SUPPORT;
@@ -2737,8 +2829,11 @@ hns3_fw_version_get(struct rte_eth_dev *eth_dev, char *fw_version,
                                      HNS3_FW_VERSION_BYTE1_S),
                       hns3_get_field(version, HNS3_FW_VERSION_BYTE0_M,
                                      HNS3_FW_VERSION_BYTE0_S));
+       if (ret < 0)
+               return -EINVAL;
+
        ret += 1; /* add the size of '\0' */
-       if (fw_size < (uint32_t)ret)
+       if (fw_size < (size_t)ret)
                return ret;
        else
                return 0;
@@ -2776,20 +2871,21 @@ hns3_setup_linkstatus(struct rte_eth_dev *eth_dev,
        case ETH_SPEED_NUM_50G:
        case ETH_SPEED_NUM_100G:
        case ETH_SPEED_NUM_200G:
-               new_link->link_speed = mac->link_speed;
+               if (mac->link_status)
+                       new_link->link_speed = mac->link_speed;
                break;
        default:
                if (mac->link_status)
                        new_link->link_speed = ETH_SPEED_NUM_UNKNOWN;
-               else
-                       new_link->link_speed = ETH_SPEED_NUM_NONE;
                break;
        }
 
+       if (!mac->link_status)
+               new_link->link_speed = ETH_SPEED_NUM_NONE;
+
        new_link->link_duplex = mac->link_duplex;
        new_link->link_status = mac->link_status ? ETH_LINK_UP : ETH_LINK_DOWN;
-       new_link->link_autoneg =
-           !(eth_dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED);
+       new_link->link_autoneg = mac->link_autoneg;
 }
 
 static int
@@ -2804,6 +2900,15 @@ hns3_dev_link_update(struct rte_eth_dev *eth_dev, int wait_to_complete)
        struct rte_eth_link new_link;
        int ret;
 
+       /* When port is stopped, report link down. */
+       if (eth_dev->data->dev_started == 0) {
+               new_link.link_autoneg = mac->link_autoneg;
+               new_link.link_duplex = mac->link_duplex;
+               new_link.link_speed = ETH_SPEED_NUM_NONE;
+               new_link.link_status = ETH_LINK_DOWN;
+               goto out;
+       }
+
        do {
                ret = hns3_update_port_link_info(eth_dev);
                if (ret) {
@@ -2821,6 +2926,7 @@ hns3_dev_link_update(struct rte_eth_dev *eth_dev, int wait_to_complete)
        memset(&new_link, 0, sizeof(new_link));
        hns3_setup_linkstatus(eth_dev, &new_link);
 
+out:
        return rte_eth_linkstatus_set(eth_dev, &new_link);
 }
 
@@ -2973,8 +3079,6 @@ hns3_parse_cfg(struct hns3_cfg *cfg, struct hns3_cmd_desc *desc)
        req = (struct hns3_cfg_param_cmd *)desc[0].data;
 
        /* get the configuration */
-       cfg->vmdq_vport_num = hns3_get_field(rte_le_to_cpu_32(req->param[0]),
-                                            HNS3_CFG_VMDQ_M, HNS3_CFG_VMDQ_S);
        cfg->tc_num = hns3_get_field(rte_le_to_cpu_32(req->param[0]),
                                     HNS3_CFG_TC_NUM_M, HNS3_CFG_TC_NUM_S);
        cfg->tqp_desc_num = hns3_get_field(rte_le_to_cpu_32(req->param[0]),
@@ -3023,7 +3127,6 @@ hns3_parse_cfg(struct hns3_cfg *cfg, struct hns3_cmd_desc *desc)
        ext_rss_size_max = hns3_get_field(rte_le_to_cpu_32(req->param[2]),
                                               HNS3_CFG_EXT_RSS_SIZE_M,
                                               HNS3_CFG_EXT_RSS_SIZE_S);
-
        /*
         * Field ext_rss_size_max obtained from firmware will be more flexible
         * for future changes and expansions, which is an exponent of 2, instead
@@ -3742,7 +3845,6 @@ hns3_drop_nopfc_buf_till_fit(struct hns3_hw *hw,
        for (i = HNS3_MAX_TC_NUM - 1; i >= 0; i--) {
                priv = &buf_alloc->priv_buf[i];
                mask = BIT((uint8_t)i);
-
                if (hw->hw_tc_map & mask &&
                    !(hw->dcb_info.hw_pfc_map & mask)) {
                        /* Clear the no pfc TC private buffer */
@@ -3828,7 +3930,6 @@ hns3_only_alloc_priv_buff(struct hns3_hw *hw,
                        COMPENSATE_HALF_MPS_NUM * half_mps;
        min_rx_priv = roundup(min_rx_priv, HNS3_BUF_SIZE_UNIT);
        rx_priv = rounddown(rx_priv, HNS3_BUF_SIZE_UNIT);
-
        if (rx_priv < min_rx_priv)
                return false;
 
@@ -3893,8 +3994,8 @@ hns3_rx_buffer_calc(struct hns3_hw *hw, struct hns3_pkt_buf_alloc *buf_alloc)
         * For different application scenes, the enabled port number, TC number
         * and no_drop TC number are different. In order to obtain the better
         * performance, software could allocate the buffer size and configure
-        * the waterline by tring to decrease the private buffer size according
-        * to the order, namely, waterline of valided tc, pfc disabled tc, pfc
+        * the waterline by trying to decrease the private buffer size according
+        * to the order, namely, waterline of valid tc, pfc disabled tc, pfc
         * enabled tc.
         */
        if (hns3_rx_buf_calc_all(hw, false, buf_alloc))
@@ -4117,28 +4218,6 @@ hns3_buffer_alloc(struct hns3_hw *hw)
        return ret;
 }
 
-static int
-hns3_firmware_compat_config(struct hns3_hw *hw, bool is_init)
-{
-       struct hns3_firmware_compat_cmd *req;
-       struct hns3_cmd_desc desc;
-       uint32_t compat = 0;
-
-       hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_FIRMWARE_COMPAT_CFG, false);
-       req = (struct hns3_firmware_compat_cmd *)desc.data;
-
-       if (is_init) {
-               hns3_set_bit(compat, HNS3_LINK_EVENT_REPORT_EN_B, 1);
-               hns3_set_bit(compat, HNS3_NCSI_ERROR_REPORT_EN_B, 0);
-               if (hw->mac.media_type == HNS3_MEDIA_TYPE_COPPER)
-                       hns3_set_bit(compat, HNS3_FIRMWARE_PHY_DRIVER_EN_B, 1);
-       }
-
-       req->compat = rte_cpu_to_le_32(compat);
-
-       return hns3_cmd_send(hw, &desc, 1);
-}
-
 static int
 hns3_mac_init(struct hns3_hw *hw)
 {
@@ -4529,24 +4608,48 @@ hns3_dev_promisc_restore(struct hns3_adapter *hns)
 }
 
 static int
-hns3_get_sfp_speed(struct hns3_hw *hw, uint32_t *speed)
+hns3_get_sfp_info(struct hns3_hw *hw, struct hns3_mac *mac_info)
 {
-       struct hns3_sfp_speed_cmd *resp;
+       struct hns3_sfp_info_cmd *resp;
        struct hns3_cmd_desc desc;
        int ret;
 
-       hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_SFP_GET_SPEED, true);
-       resp = (struct hns3_sfp_speed_cmd *)desc.data;
+       hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_GET_SFP_INFO, true);
+       resp = (struct hns3_sfp_info_cmd *)desc.data;
+       resp->query_type = HNS3_ACTIVE_QUERY;
+
        ret = hns3_cmd_send(hw, &desc, 1);
        if (ret == -EOPNOTSUPP) {
-               hns3_err(hw, "IMP do not support get SFP speed %d", ret);
+               hns3_warn(hw, "firmware does not support get SFP info,"
+                         " ret = %d.", ret);
                return ret;
        } else if (ret) {
-               hns3_err(hw, "get sfp speed failed %d", ret);
+               hns3_err(hw, "get sfp info failed, ret = %d.", ret);
                return ret;
        }
 
-       *speed = resp->sfp_speed;
+       /*
+        * In some case, the speed of MAC obtained from firmware may be 0, it
+        * shouldn't be set to mac->speed.
+        */
+       if (!rte_le_to_cpu_32(resp->sfp_speed))
+               return 0;
+
+       mac_info->link_speed = rte_le_to_cpu_32(resp->sfp_speed);
+       /*
+        * if resp->supported_speed is 0, it means it's an old version
+        * firmware, do not update these params.
+        */
+       if (resp->supported_speed) {
+               mac_info->query_type = HNS3_ACTIVE_QUERY;
+               mac_info->supported_speed =
+                                       rte_le_to_cpu_32(resp->supported_speed);
+               mac_info->support_autoneg = resp->autoneg_ability;
+               mac_info->link_autoneg = (resp->autoneg == 0) ? ETH_LINK_FIXED
+                                       : ETH_LINK_AUTONEG;
+       } else {
+               mac_info->query_type = HNS3_DEFAULT_QUERY;
+       }
 
        return 0;
 }
@@ -4588,25 +4691,51 @@ static int
 hns3_update_fiber_link_info(struct hns3_hw *hw)
 {
        struct hns3_pf *pf = HNS3_DEV_HW_TO_PF(hw);
-       uint32_t speed;
+       struct hns3_mac *mac = &hw->mac;
+       struct hns3_mac mac_info;
        int ret;
 
-       /* If IMP do not support get SFP/qSFP speed, return directly */
+       /* If firmware do not support get SFP/qSFP speed, return directly */
        if (!pf->support_sfp_query)
                return 0;
 
-       ret = hns3_get_sfp_speed(hw, &speed);
+       memset(&mac_info, 0, sizeof(struct hns3_mac));
+       ret = hns3_get_sfp_info(hw, &mac_info);
        if (ret == -EOPNOTSUPP) {
                pf->support_sfp_query = false;
                return ret;
        } else if (ret)
                return ret;
 
-       if (speed == ETH_SPEED_NUM_NONE)
-               return 0; /* do nothing if no SFP */
+       /* Do nothing if no SFP */
+       if (mac_info.link_speed == ETH_SPEED_NUM_NONE)
+               return 0;
+
+       /*
+        * If query_type is HNS3_ACTIVE_QUERY, it is no need
+        * to reconfigure the speed of MAC. Otherwise, it indicates
+        * that the current firmware only supports to obtain the
+        * speed of the SFP, and the speed of MAC needs to reconfigure.
+        */
+       mac->query_type = mac_info.query_type;
+       if (mac->query_type == HNS3_ACTIVE_QUERY) {
+               if (mac_info.link_speed != mac->link_speed) {
+                       ret = hns3_port_shaper_update(hw, mac_info.link_speed);
+                       if (ret)
+                               return ret;
+               }
+
+               mac->link_speed = mac_info.link_speed;
+               mac->supported_speed = mac_info.supported_speed;
+               mac->support_autoneg = mac_info.support_autoneg;
+               mac->link_autoneg = mac_info.link_autoneg;
+
+               return 0;
+       }
 
        /* Config full duplex for SFP */
-       return hns3_cfg_mac_speed_dup(hw, speed, ETH_LINK_FULL_DUPLEX);
+       return hns3_cfg_mac_speed_dup(hw, mac_info.link_speed,
+                                     ETH_LINK_FULL_DUPLEX);
 }
 
 static void
@@ -4774,8 +4903,6 @@ hns3_update_link_status(struct hns3_hw *hw)
        if (state != hw->mac.link_status) {
                hw->mac.link_status = state;
                hns3_warn(hw, "Link status change to %s!", state ? "up" : "down");
-               hns3_config_mac_tnl_int(hw,
-                                       state == ETH_LINK_UP ? true : false);
                return true;
        }
 
@@ -4815,28 +4942,6 @@ hns3_service_handler(void *param)
        rte_eal_alarm_set(HNS3_SERVICE_INTERVAL, hns3_service_handler, eth_dev);
 }
 
-static void
-hns3_update_dev_lsc_cap(struct hns3_hw *hw,
-                       int fw_compact_cmd_result)
-{
-       struct rte_eth_dev *dev = &rte_eth_devices[hw->data->port_id];
-
-       if (hw->adapter_state != HNS3_NIC_UNINITIALIZED)
-               return;
-
-       if (fw_compact_cmd_result != 0) {
-               /*
-                * If fw_compact_cmd_result is not zero, it means firmware don't
-                * support link status change interrupt.
-                * Framework already set RTE_ETH_DEV_INTR_LSC bit because driver
-                * declared RTE_PCI_DRV_INTR_LSC in drv_flags. It need to clear
-                * the RTE_ETH_DEV_INTR_LSC capability when detect firmware
-                * don't support link status change interrupt.
-                */
-               dev->data->dev_flags &= ~RTE_ETH_DEV_INTR_LSC;
-       }
-}
-
 static int
 hns3_init_hardware(struct hns3_adapter *hns)
 {
@@ -4916,16 +5021,6 @@ hns3_init_hardware(struct hns3_adapter *hns)
                goto err_mac_init;
        }
 
-       /*
-        * Requiring firmware to enable some features, driver can
-        * still work without it.
-        */
-       ret = hns3_firmware_compat_config(hw, true);
-       if (ret)
-               PMD_INIT_LOG(WARNING, "firmware compatible features not "
-                            "supported, ret = %d.", ret);
-       hns3_update_dev_lsc_cap(hw, ret);
-
        return 0;
 
 err_mac_init:
@@ -4959,13 +5054,96 @@ hns3_config_all_msix_error(struct hns3_hw *hw, bool enable)
         * and belong to a different type from the MSI-x errors processed
         * by the network driver.
         *
-        * Network driver should open the new error report on initialition
+        * Network driver should open the new error report on initialization.
         */
        val = hns3_read_dev(hw, HNS3_VECTOR0_OTER_EN_REG);
        hns3_set_bit(val, HNS3_VECTOR0_ALL_MSIX_ERR_B, enable ? 1 : 0);
        hns3_write_dev(hw, HNS3_VECTOR0_OTER_EN_REG, val);
 }
 
+static uint32_t
+hns3_set_firber_default_support_speed(struct hns3_hw *hw)
+{
+       struct hns3_mac *mac = &hw->mac;
+
+       switch (mac->link_speed) {
+       case ETH_SPEED_NUM_1G:
+               return HNS3_FIBER_LINK_SPEED_1G_BIT;
+       case ETH_SPEED_NUM_10G:
+               return HNS3_FIBER_LINK_SPEED_10G_BIT;
+       case ETH_SPEED_NUM_25G:
+               return HNS3_FIBER_LINK_SPEED_25G_BIT;
+       case ETH_SPEED_NUM_40G:
+               return HNS3_FIBER_LINK_SPEED_40G_BIT;
+       case ETH_SPEED_NUM_50G:
+               return HNS3_FIBER_LINK_SPEED_50G_BIT;
+       case ETH_SPEED_NUM_100G:
+               return HNS3_FIBER_LINK_SPEED_100G_BIT;
+       case ETH_SPEED_NUM_200G:
+               return HNS3_FIBER_LINK_SPEED_200G_BIT;
+       default:
+               hns3_warn(hw, "invalid speed %u Mbps.", mac->link_speed);
+               return 0;
+       }
+}
+
+/*
+ * Validity of supported_speed for firber and copper media type can be
+ * guaranteed by the following policy:
+ * Copper:
+ *       Although the initialization of the phy in the firmware may not be
+ *       completed, the firmware can guarantees that the supported_speed is
+ *       an valid value.
+ * Firber:
+ *       If the version of firmware supports the acitive query way of the
+ *       HNS3_OPC_GET_SFP_INFO opcode, the supported_speed can be obtained
+ *       through it. If unsupported, use the SFP's speed as the value of the
+ *       supported_speed.
+ */
+static int
+hns3_get_port_supported_speed(struct rte_eth_dev *eth_dev)
+{
+       struct hns3_adapter *hns = eth_dev->data->dev_private;
+       struct hns3_hw *hw = &hns->hw;
+       struct hns3_mac *mac = &hw->mac;
+       int ret;
+
+       ret = hns3_update_link_info(eth_dev);
+       if (ret)
+               return ret;
+
+       if (mac->media_type == HNS3_MEDIA_TYPE_FIBER) {
+               /*
+                * Some firmware does not support the report of supported_speed,
+                * and only report the effective speed of SFP. In this case, it
+                * is necessary to use the SFP's speed as the supported_speed.
+                */
+               if (mac->supported_speed == 0)
+                       mac->supported_speed =
+                               hns3_set_firber_default_support_speed(hw);
+       }
+
+       return 0;
+}
+
+static void
+hns3_get_fc_autoneg_capability(struct hns3_adapter *hns)
+{
+       struct hns3_mac *mac = &hns->hw.mac;
+
+       if (mac->media_type == HNS3_MEDIA_TYPE_COPPER) {
+               hns->pf.support_fc_autoneg = true;
+               return;
+       }
+
+       /*
+        * Flow control auto-negotiation requires the cooperation of the driver
+        * and firmware. Currently, the optical port does not support flow
+        * control auto-negotiation.
+        */
+       hns->pf.support_fc_autoneg = false;
+}
+
 static int
 hns3_init_pf(struct rte_eth_dev *eth_dev)
 {
@@ -4996,6 +5174,8 @@ hns3_init_pf(struct rte_eth_dev *eth_dev)
                goto err_cmd_init;
        }
 
+       hns3_tx_push_init(eth_dev);
+
        /*
         * To ensure that the hardware environment is clean during
         * initialization, the driver actively clear the hardware environment
@@ -5066,14 +5246,24 @@ hns3_init_pf(struct rte_eth_dev *eth_dev)
                goto err_enable_intr;
        }
 
+       ret = hns3_get_port_supported_speed(eth_dev);
+       if (ret) {
+               PMD_INIT_LOG(ERR, "failed to get speed capabilities supported "
+                            "by device, ret = %d.", ret);
+               goto err_supported_speed;
+       }
+
+       hns3_get_fc_autoneg_capability(hns);
+
        hns3_tm_conf_init(eth_dev);
 
        return 0;
 
+err_supported_speed:
+       (void)hns3_enable_hw_error_intr(hns, false);
 err_enable_intr:
        hns3_fdir_filter_uninit(hns);
 err_fdir:
-       (void)hns3_firmware_compat_config(hw, false);
        hns3_uninit_umv_space(hw);
 err_init_hw:
        hns3_tqp_stats_uninit(hw);
@@ -5108,7 +5298,6 @@ hns3_uninit_pf(struct rte_eth_dev *eth_dev)
        (void)hns3_config_gro(hw, false);
        hns3_promisc_uninit(hw);
        hns3_fdir_filter_uninit(hns);
-       (void)hns3_firmware_compat_config(hw, false);
        hns3_uninit_umv_space(hw);
        hns3_tqp_stats_uninit(hw);
        hns3_config_mac_tnl_int(hw, false);
@@ -5122,20 +5311,282 @@ hns3_uninit_pf(struct rte_eth_dev *eth_dev)
        hw->io_base = NULL;
 }
 
+static uint32_t
+hns3_convert_link_speeds2bitmap_copper(uint32_t link_speeds)
+{
+       uint32_t speed_bit;
+
+       switch (link_speeds & ~ETH_LINK_SPEED_FIXED) {
+       case ETH_LINK_SPEED_10M:
+               speed_bit = HNS3_PHY_LINK_SPEED_10M_BIT;
+               break;
+       case ETH_LINK_SPEED_10M_HD:
+               speed_bit = HNS3_PHY_LINK_SPEED_10M_HD_BIT;
+               break;
+       case ETH_LINK_SPEED_100M:
+               speed_bit = HNS3_PHY_LINK_SPEED_100M_BIT;
+               break;
+       case ETH_LINK_SPEED_100M_HD:
+               speed_bit = HNS3_PHY_LINK_SPEED_100M_HD_BIT;
+               break;
+       case ETH_LINK_SPEED_1G:
+               speed_bit = HNS3_PHY_LINK_SPEED_1000M_BIT;
+               break;
+       default:
+               speed_bit = 0;
+               break;
+       }
+
+       return speed_bit;
+}
+
+static uint32_t
+hns3_convert_link_speeds2bitmap_fiber(uint32_t link_speeds)
+{
+       uint32_t speed_bit;
+
+       switch (link_speeds & ~ETH_LINK_SPEED_FIXED) {
+       case ETH_LINK_SPEED_1G:
+               speed_bit = HNS3_FIBER_LINK_SPEED_1G_BIT;
+               break;
+       case ETH_LINK_SPEED_10G:
+               speed_bit = HNS3_FIBER_LINK_SPEED_10G_BIT;
+               break;
+       case ETH_LINK_SPEED_25G:
+               speed_bit = HNS3_FIBER_LINK_SPEED_25G_BIT;
+               break;
+       case ETH_LINK_SPEED_40G:
+               speed_bit = HNS3_FIBER_LINK_SPEED_40G_BIT;
+               break;
+       case ETH_LINK_SPEED_50G:
+               speed_bit = HNS3_FIBER_LINK_SPEED_50G_BIT;
+               break;
+       case ETH_LINK_SPEED_100G:
+               speed_bit = HNS3_FIBER_LINK_SPEED_100G_BIT;
+               break;
+       case ETH_LINK_SPEED_200G:
+               speed_bit = HNS3_FIBER_LINK_SPEED_200G_BIT;
+               break;
+       default:
+               speed_bit = 0;
+               break;
+       }
+
+       return speed_bit;
+}
+
 static int
-hns3_do_start(struct hns3_adapter *hns, bool reset_queue)
+hns3_check_port_speed(struct hns3_hw *hw, uint32_t link_speeds)
 {
-       struct hns3_hw *hw = &hns->hw;
+       struct hns3_mac *mac = &hw->mac;
+       uint32_t supported_speed = mac->supported_speed;
+       uint32_t speed_bit = 0;
+
+       if (mac->media_type == HNS3_MEDIA_TYPE_COPPER)
+               speed_bit = hns3_convert_link_speeds2bitmap_copper(link_speeds);
+       else if (mac->media_type == HNS3_MEDIA_TYPE_FIBER)
+               speed_bit = hns3_convert_link_speeds2bitmap_fiber(link_speeds);
+
+       if (!(speed_bit & supported_speed)) {
+               hns3_err(hw, "link_speeds(0x%x) exceeds the supported speed capability or is incorrect.",
+                        link_speeds);
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static inline uint32_t
+hns3_get_link_speed(uint32_t link_speeds)
+{
+       uint32_t speed = ETH_SPEED_NUM_NONE;
+
+       if (link_speeds & ETH_LINK_SPEED_10M ||
+           link_speeds & ETH_LINK_SPEED_10M_HD)
+               speed = ETH_SPEED_NUM_10M;
+       if (link_speeds & ETH_LINK_SPEED_100M ||
+           link_speeds & ETH_LINK_SPEED_100M_HD)
+               speed = ETH_SPEED_NUM_100M;
+       if (link_speeds & ETH_LINK_SPEED_1G)
+               speed = ETH_SPEED_NUM_1G;
+       if (link_speeds & ETH_LINK_SPEED_10G)
+               speed = ETH_SPEED_NUM_10G;
+       if (link_speeds & ETH_LINK_SPEED_25G)
+               speed = ETH_SPEED_NUM_25G;
+       if (link_speeds & ETH_LINK_SPEED_40G)
+               speed = ETH_SPEED_NUM_40G;
+       if (link_speeds & ETH_LINK_SPEED_50G)
+               speed = ETH_SPEED_NUM_50G;
+       if (link_speeds & ETH_LINK_SPEED_100G)
+               speed = ETH_SPEED_NUM_100G;
+       if (link_speeds & ETH_LINK_SPEED_200G)
+               speed = ETH_SPEED_NUM_200G;
+
+       return speed;
+}
+
+static uint8_t
+hns3_get_link_duplex(uint32_t link_speeds)
+{
+       if ((link_speeds & ETH_LINK_SPEED_10M_HD) ||
+           (link_speeds & ETH_LINK_SPEED_100M_HD))
+               return ETH_LINK_HALF_DUPLEX;
+       else
+               return ETH_LINK_FULL_DUPLEX;
+}
+
+static int
+hns3_set_copper_port_link_speed(struct hns3_hw *hw,
+                               struct hns3_set_link_speed_cfg *cfg)
+{
+       struct hns3_cmd_desc desc[HNS3_PHY_PARAM_CFG_BD_NUM];
+       struct hns3_phy_params_bd0_cmd *req;
+       uint16_t i;
+
+       for (i = 0; i < HNS3_PHY_PARAM_CFG_BD_NUM - 1; i++) {
+               hns3_cmd_setup_basic_desc(&desc[i], HNS3_OPC_PHY_PARAM_CFG,
+                                         false);
+               desc[i].flag |= rte_cpu_to_le_16(HNS3_CMD_FLAG_NEXT);
+       }
+       hns3_cmd_setup_basic_desc(&desc[i], HNS3_OPC_PHY_PARAM_CFG, false);
+       req = (struct hns3_phy_params_bd0_cmd *)desc[0].data;
+       req->autoneg = cfg->autoneg;
+
+       /*
+        * The full speed capability is used to negotiate when
+        * auto-negotiation is enabled.
+        */
+       if (cfg->autoneg) {
+               req->advertising = HNS3_PHY_LINK_SPEED_10M_BIT |
+                                   HNS3_PHY_LINK_SPEED_10M_HD_BIT |
+                                   HNS3_PHY_LINK_SPEED_100M_BIT |
+                                   HNS3_PHY_LINK_SPEED_100M_HD_BIT |
+                                   HNS3_PHY_LINK_SPEED_1000M_BIT;
+       } else {
+               req->speed = cfg->speed;
+               req->duplex = cfg->duplex;
+       }
+
+       return hns3_cmd_send(hw, desc, HNS3_PHY_PARAM_CFG_BD_NUM);
+}
+
+static int
+hns3_set_autoneg(struct hns3_hw *hw, bool enable)
+{
+       struct hns3_config_auto_neg_cmd *req;
+       struct hns3_cmd_desc desc;
+       uint32_t flag = 0;
        int ret;
 
-       ret = hns3_dcb_cfg_update(hns);
+       hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_CONFIG_AN_MODE, false);
+
+       req = (struct hns3_config_auto_neg_cmd *)desc.data;
+       if (enable)
+               hns3_set_bit(flag, HNS3_MAC_CFG_AN_EN_B, 1);
+       req->cfg_an_cmd_flag = rte_cpu_to_le_32(flag);
+
+       ret = hns3_cmd_send(hw, &desc, 1);
        if (ret)
-               return ret;
+               hns3_err(hw, "autoneg set cmd failed, ret = %d.", ret);
+
+       return ret;
+}
+
+static int
+hns3_set_fiber_port_link_speed(struct hns3_hw *hw,
+                              struct hns3_set_link_speed_cfg *cfg)
+{
+       int ret;
+
+       if (hw->mac.support_autoneg) {
+               ret = hns3_set_autoneg(hw, cfg->autoneg);
+               if (ret) {
+                       hns3_err(hw, "failed to configure auto-negotiation.");
+                       return ret;
+               }
+
+               /*
+                * To enable auto-negotiation, we only need to open the switch
+                * of auto-negotiation, then firmware sets all speed
+                * capabilities.
+                */
+               if (cfg->autoneg)
+                       return 0;
+       }
 
        /*
-        * The hns3_dcb_cfg_update may configure TM module, so
-        * hns3_tm_conf_update must called later.
+        * Some hardware doesn't support auto-negotiation, but users may not
+        * configure link_speeds (default 0), which means auto-negotiation.
+        * In this case, it should return success.
         */
+       if (cfg->autoneg)
+               return 0;
+
+       return hns3_cfg_mac_speed_dup(hw, cfg->speed, cfg->duplex);
+}
+
+static int
+hns3_set_port_link_speed(struct hns3_hw *hw,
+                        struct hns3_set_link_speed_cfg *cfg)
+{
+       int ret;
+
+       if (hw->mac.media_type == HNS3_MEDIA_TYPE_COPPER) {
+#if defined(RTE_HNS3_ONLY_1630_FPGA)
+               struct hns3_pf *pf = HNS3_DEV_HW_TO_PF(hw);
+               if (pf->is_tmp_phy)
+                       return 0;
+#endif
+
+               ret = hns3_set_copper_port_link_speed(hw, cfg);
+               if (ret) {
+                       hns3_err(hw, "failed to set copper port link speed,"
+                                "ret = %d.", ret);
+                       return ret;
+               }
+       } else if (hw->mac.media_type == HNS3_MEDIA_TYPE_FIBER) {
+               ret = hns3_set_fiber_port_link_speed(hw, cfg);
+               if (ret) {
+                       hns3_err(hw, "failed to set fiber port link speed,"
+                                "ret = %d.", ret);
+                       return ret;
+               }
+       }
+
+       return 0;
+}
+
+static int
+hns3_apply_link_speed(struct hns3_hw *hw)
+{
+       struct rte_eth_conf *conf = &hw->data->dev_conf;
+       struct hns3_set_link_speed_cfg cfg;
+
+       memset(&cfg, 0, sizeof(struct hns3_set_link_speed_cfg));
+       cfg.autoneg = (conf->link_speeds == ETH_LINK_SPEED_AUTONEG) ?
+                       ETH_LINK_AUTONEG : ETH_LINK_FIXED;
+       if (cfg.autoneg != ETH_LINK_AUTONEG) {
+               cfg.speed = hns3_get_link_speed(conf->link_speeds);
+               cfg.duplex = hns3_get_link_duplex(conf->link_speeds);
+       }
+
+       return hns3_set_port_link_speed(hw, &cfg);
+}
+
+static int
+hns3_do_start(struct hns3_adapter *hns, bool reset_queue)
+{
+       struct hns3_hw *hw = &hns->hw;
+       int ret;
+
+       ret = hns3_update_queue_map_configure(hns);
+       if (ret) {
+               hns3_err(hw, "failed to update queue mapping configuration, ret = %d",
+                        ret);
+               return ret;
+       }
+
+       /* Note: hns3_tm_conf_update must be called after configuring DCB. */
        ret = hns3_tm_conf_update(hw);
        if (ret) {
                PMD_INIT_LOG(ERR, "failed to update tm conf, ret = %d.", ret);
@@ -5155,8 +5606,16 @@ hns3_do_start(struct hns3_adapter *hns, bool reset_queue)
                PMD_INIT_LOG(ERR, "failed to enable MAC, ret = %d", ret);
                goto err_config_mac_mode;
        }
+
+       ret = hns3_apply_link_speed(hw);
+       if (ret)
+               goto err_set_link_speed;
+
        return 0;
 
+err_set_link_speed:
+       (void)hns3_cfg_mac_mode(hw, false);
+
 err_config_mac_mode:
        hns3_dev_release_mbufs(hns);
        /*
@@ -5435,7 +5894,7 @@ hns3_dev_stop(struct rte_eth_dev *dev)
        /* Disable datapath on secondary process. */
        hns3_mp_req_stop_rxtx(dev);
        /* Prevent crashes when queues are still in use. */
-       rte_delay_ms(hw->tqps_num);
+       rte_delay_ms(hw->cfg_max_queues);
 
        rte_spinlock_lock(&hw->lock);
        if (__atomic_load_n(&hw->reset.resetting, __ATOMIC_RELAXED) == 0) {
@@ -5488,16 +5947,102 @@ hns3_dev_close(struct rte_eth_dev *eth_dev)
        return ret;
 }
 
+static void
+hns3_get_autoneg_rxtx_pause_copper(struct hns3_hw *hw, bool *rx_pause,
+                                  bool *tx_pause)
+{
+       struct hns3_mac *mac = &hw->mac;
+       uint32_t advertising = mac->advertising;
+       uint32_t lp_advertising = mac->lp_advertising;
+       *rx_pause = false;
+       *tx_pause = false;
+
+       if (advertising & lp_advertising & HNS3_PHY_LINK_MODE_PAUSE_BIT) {
+               *rx_pause = true;
+               *tx_pause = true;
+       } else if (advertising & lp_advertising &
+                  HNS3_PHY_LINK_MODE_ASYM_PAUSE_BIT) {
+               if (advertising & HNS3_PHY_LINK_MODE_PAUSE_BIT)
+                       *rx_pause = true;
+               else if (lp_advertising & HNS3_PHY_LINK_MODE_PAUSE_BIT)
+                       *tx_pause = true;
+       }
+}
+
+static enum hns3_fc_mode
+hns3_get_autoneg_fc_mode(struct hns3_hw *hw)
+{
+       enum hns3_fc_mode current_mode;
+       bool rx_pause = false;
+       bool tx_pause = false;
+
+       switch (hw->mac.media_type) {
+       case HNS3_MEDIA_TYPE_COPPER:
+               hns3_get_autoneg_rxtx_pause_copper(hw, &rx_pause, &tx_pause);
+               break;
+
+       /*
+        * Flow control auto-negotiation is not supported for fiber and
+        * backpalne media type.
+        */
+       case HNS3_MEDIA_TYPE_FIBER:
+       case HNS3_MEDIA_TYPE_BACKPLANE:
+               hns3_err(hw, "autoneg FC mode can't be obtained, but flow control auto-negotiation is enabled.");
+               current_mode = hw->requested_fc_mode;
+               goto out;
+       default:
+               hns3_err(hw, "autoneg FC mode can't be obtained for unknown media type(%u).",
+                        hw->mac.media_type);
+               current_mode = HNS3_FC_NONE;
+               goto out;
+       }
+
+       if (rx_pause && tx_pause)
+               current_mode = HNS3_FC_FULL;
+       else if (rx_pause)
+               current_mode = HNS3_FC_RX_PAUSE;
+       else if (tx_pause)
+               current_mode = HNS3_FC_TX_PAUSE;
+       else
+               current_mode = HNS3_FC_NONE;
+
+out:
+       return current_mode;
+}
+
+static enum hns3_fc_mode
+hns3_get_current_fc_mode(struct rte_eth_dev *dev)
+{
+       struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       struct hns3_pf *pf = HNS3_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+       struct hns3_mac *mac = &hw->mac;
+
+       /*
+        * When the flow control mode is obtained, the device may not complete
+        * auto-negotiation. It is necessary to wait for link establishment.
+        */
+       (void)hns3_dev_link_update(dev, 1);
+
+       /*
+        * If the link auto-negotiation of the nic is disabled, or the flow
+        * control auto-negotiation is not supported, the forced flow control
+        * mode is used.
+        */
+       if (mac->link_autoneg == 0 || !pf->support_fc_autoneg)
+               return hw->requested_fc_mode;
+
+       return hns3_get_autoneg_fc_mode(hw);
+}
+
 static int
 hns3_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
 {
        struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
        struct hns3_pf *pf = HNS3_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+       enum hns3_fc_mode current_mode;
 
-       fc_conf->pause_time = pf->pause_time;
-
-       /* return fc current mode */
-       switch (hw->current_mode) {
+       current_mode = hns3_get_current_fc_mode(dev);
+       switch (current_mode) {
        case HNS3_FC_FULL:
                fc_conf->mode = RTE_FC_FULL;
                break;
@@ -5513,38 +6058,51 @@ hns3_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
                break;
        }
 
+       fc_conf->pause_time = pf->pause_time;
+       fc_conf->autoneg = pf->support_fc_autoneg ? hw->mac.link_autoneg : 0;
+
        return 0;
 }
 
-static void
-hns3_get_fc_mode(struct hns3_hw *hw, enum rte_eth_fc_mode mode)
+static int
+hns3_check_fc_autoneg_valid(struct hns3_hw *hw, uint8_t autoneg)
 {
-       switch (mode) {
-       case RTE_FC_NONE:
-               hw->requested_mode = HNS3_FC_NONE;
-               break;
-       case RTE_FC_RX_PAUSE:
-               hw->requested_mode = HNS3_FC_RX_PAUSE;
-               break;
-       case RTE_FC_TX_PAUSE:
-               hw->requested_mode = HNS3_FC_TX_PAUSE;
-               break;
-       case RTE_FC_FULL:
-               hw->requested_mode = HNS3_FC_FULL;
-               break;
-       default:
-               hw->requested_mode = HNS3_FC_NONE;
-               hns3_warn(hw, "fc_mode(%u) exceeds member scope and is "
-                         "configured to RTE_FC_NONE", mode);
-               break;
+       struct hns3_pf *pf = HNS3_DEV_HW_TO_PF(hw);
+
+       if (!pf->support_fc_autoneg) {
+               if (autoneg != 0) {
+                       hns3_err(hw, "unsupported fc auto-negotiation setting.");
+                       return -EOPNOTSUPP;
+               }
+
+               /*
+                * Flow control auto-negotiation of the NIC is not supported,
+                * but other auto-negotiation features may be supported.
+                */
+               if (autoneg != hw->mac.link_autoneg) {
+                       hns3_err(hw, "please use 'link_speeds' in struct rte_eth_conf to disable autoneg!");
+                       return -EOPNOTSUPP;
+               }
+
+               return 0;
+       }
+
+       /*
+        * If flow control auto-negotiation of the NIC is supported, all
+        * auto-negotiation features are supported.
+        */
+       if (autoneg != hw->mac.link_autoneg) {
+               hns3_err(hw, "please use 'link_speeds' in struct rte_eth_conf to change autoneg!");
+               return -EOPNOTSUPP;
        }
+
+       return 0;
 }
 
 static int
 hns3_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
 {
        struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-       struct hns3_pf *pf = HNS3_DEV_PRIVATE_TO_PF(dev->data->dev_private);
        int ret;
 
        if (fc_conf->high_water || fc_conf->low_water ||
@@ -5556,10 +6114,11 @@ hns3_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
                         fc_conf->send_xon, fc_conf->mac_ctrl_frame_fwd);
                return -EINVAL;
        }
-       if (fc_conf->autoneg) {
-               hns3_err(hw, "Unsupported fc auto-negotiation setting.");
-               return -EINVAL;
-       }
+
+       ret = hns3_check_fc_autoneg_valid(hw, fc_conf->autoneg);
+       if (ret)
+               return ret;
+
        if (!fc_conf->pause_time) {
                hns3_err(hw, "Invalid pause time %u setting.",
                         fc_conf->pause_time);
@@ -5578,11 +6137,6 @@ hns3_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
                return -EOPNOTSUPP;
        }
 
-       hns3_get_fc_mode(hw, fc_conf->mode);
-       if (hw->requested_mode == hw->current_mode &&
-           pf->pause_time == fc_conf->pause_time)
-               return 0;
-
        rte_spinlock_lock(&hw->lock);
        ret = hns3_fc_enable(dev, fc_conf);
        rte_spinlock_unlock(&hw->lock);
@@ -5595,8 +6149,6 @@ hns3_priority_flow_ctrl_set(struct rte_eth_dev *dev,
                            struct rte_eth_pfc_conf *pfc_conf)
 {
        struct hns3_hw *hw = HNS3_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-       struct hns3_pf *pf = HNS3_DEV_PRIVATE_TO_PF(dev->data->dev_private);
-       uint8_t priority;
        int ret;
 
        if (!hns3_dev_dcb_supported(hw)) {
@@ -5631,13 +6183,6 @@ hns3_priority_flow_ctrl_set(struct rte_eth_dev *dev,
                return -EOPNOTSUPP;
        }
 
-       priority = pfc_conf->priority;
-       hns3_get_fc_mode(hw, pfc_conf->fc.mode);
-       if (hw->dcb_info.pfc_en & BIT(priority) &&
-           hw->requested_mode == hw->current_mode &&
-           pfc_conf->fc.pause_time == pf->pause_time)
-               return 0;
-
        rte_spinlock_lock(&hw->lock);
        ret = hns3_dcb_pfc_enable(dev, pfc_conf);
        rte_spinlock_unlock(&hw->lock);
@@ -5751,12 +6296,14 @@ hns3_is_reset_pending(struct hns3_adapter *hns)
 
        hns3_check_event_cause(hns, NULL);
        reset = hns3_get_reset_level(hns, &hw->reset.pending);
-       if (hw->reset.level != HNS3_NONE_RESET && hw->reset.level < reset) {
+       if (reset != HNS3_NONE_RESET && hw->reset.level != HNS3_NONE_RESET &&
+           hw->reset.level < reset) {
                hns3_warn(hw, "High level reset %d is pending", reset);
                return true;
        }
        reset = hns3_get_reset_level(hns, &hw->reset.request);
-       if (hw->reset.level != HNS3_NONE_RESET && hw->reset.level < reset) {
+       if (reset != HNS3_NONE_RESET && hw->reset.level != HNS3_NONE_RESET &&
+           hw->reset.level < reset) {
                hns3_warn(hw, "High level reset %d is request", reset);
                return true;
        }
@@ -5773,7 +6320,7 @@ hns3_wait_hardware_ready(struct hns3_adapter *hns)
        if (wait_data->result == HNS3_WAIT_SUCCESS)
                return 0;
        else if (wait_data->result == HNS3_WAIT_TIMEOUT) {
-               gettimeofday(&tv, NULL);
+               hns3_clock_gettime(&tv);
                hns3_warn(hw, "Reset step4 hardware not ready after reset time=%ld.%.6ld",
                          tv.tv_sec, tv.tv_usec);
                return -ETIME;
@@ -5783,7 +6330,7 @@ hns3_wait_hardware_ready(struct hns3_adapter *hns)
        wait_data->hns = hns;
        wait_data->check_completion = is_pf_reset_done;
        wait_data->end_ms = (uint64_t)HNS3_RESET_WAIT_CNT *
-                                     HNS3_RESET_WAIT_MS + get_timeofday_ms();
+                               HNS3_RESET_WAIT_MS + hns3_clock_gettime_ms();
        wait_data->interval = HNS3_RESET_WAIT_MS * USEC_PER_MSEC;
        wait_data->count = HNS3_RESET_WAIT_CNT;
        wait_data->result = HNS3_WAIT_REQUEST;
@@ -5822,7 +6369,7 @@ hns3_msix_process(struct hns3_adapter *hns, enum hns3_reset_level reset_level)
        struct timeval tv;
        uint32_t val;
 
-       gettimeofday(&tv, NULL);
+       hns3_clock_gettime(&tv);
        if (hns3_read_dev(hw, HNS3_GLOBAL_RESET_REG) ||
            hns3_read_dev(hw, HNS3_FUN_RST_ING)) {
                hns3_warn(hw, "Don't process msix during resetting time=%ld.%.6ld",
@@ -5962,7 +6509,7 @@ hns3_stop_service(struct hns3_adapter *hns)
        rte_wmb();
        /* Disable datapath on secondary process. */
        hns3_mp_req_stop_rxtx(eth_dev);
-       rte_delay_ms(hw->tqps_num);
+       rte_delay_ms(hw->cfg_max_queues);
 
        rte_spinlock_lock(&hw->lock);
        if (hns->hw.adapter_state == HNS3_NIC_STARTED ||
@@ -6130,14 +6677,13 @@ hns3_reset_service(void *param)
         */
        reset_level = hns3_get_reset_level(hns, &hw->reset.pending);
        if (reset_level != HNS3_NONE_RESET) {
-               gettimeofday(&tv_start, NULL);
+               hns3_clock_gettime(&tv_start);
                ret = hns3_reset_process(hns, reset_level);
-               gettimeofday(&tv, NULL);
+               hns3_clock_gettime(&tv);
                timersub(&tv, &tv_start, &tv_delta);
-               msec = tv_delta.tv_sec * MSEC_PER_SEC +
-                      tv_delta.tv_usec / USEC_PER_MSEC;
+               msec = hns3_clock_calctime_ms(&tv_delta);
                if (msec > HNS3_RESET_PROCESS_MS)
-                       hns3_err(hw, "%d handle long time delta %" PRIx64
+                       hns3_err(hw, "%d handle long time delta %" PRIu64
                                     " ms time=%ld.%.6ld",
                                 hw->reset.level, msec,
                                 tv.tv_sec, tv.tv_usec);
@@ -6264,8 +6810,7 @@ get_current_fec_auto_state(struct hns3_hw *hw, uint8_t *state)
 static int
 hns3_fec_get_internal(struct hns3_hw *hw, uint32_t *fec_capa)
 {
-#define QUERY_ACTIVE_SPEED     1
-       struct hns3_sfp_speed_cmd *resp;
+       struct hns3_sfp_info_cmd *resp;
        uint32_t tmp_fec_capa;
        uint8_t auto_state;
        struct hns3_cmd_desc desc;
@@ -6287,9 +6832,9 @@ hns3_fec_get_internal(struct hns3_hw *hw, uint32_t *fec_capa)
                }
        }
 
-       hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_SFP_GET_SPEED, true);
-       resp = (struct hns3_sfp_speed_cmd *)desc.data;
-       resp->query_type = QUERY_ACTIVE_SPEED;
+       hns3_cmd_setup_basic_desc(&desc, HNS3_OPC_GET_SFP_INFO, true);
+       resp = (struct hns3_sfp_info_cmd *)desc.data;
+       resp->query_type = HNS3_ACTIVE_QUERY;
 
        ret = hns3_cmd_send(hw, &desc, 1);
        if (ret == -EOPNOTSUPP) {
@@ -6422,9 +6967,11 @@ hns3_fec_set(struct rte_eth_dev *dev, uint32_t mode)
                return ret;
 
        /* HNS3 PMD driver only support one bit set mode, e.g. 0x1, 0x4 */
-       if (!is_fec_mode_one_bit_set(mode))
-               hns3_err(hw, "FEC mode(0x%x) not supported in HNS3 PMD,"
+       if (!is_fec_mode_one_bit_set(mode)) {
+               hns3_err(hw, "FEC mode(0x%x) not supported in HNS3 PMD, "
                             "FEC mode should be only one bit set", mode);
+               return -EINVAL;
+       }
 
        /*
         * Check whether the configured mode is within the FEC capability.
@@ -6636,6 +7183,39 @@ hns3_get_module_info(struct rte_eth_dev *dev,
        return 0;
 }
 
+void
+hns3_clock_gettime(struct timeval *tv)
+{
+#ifdef CLOCK_MONOTONIC_RAW /* Defined in glibc bits/time.h */
+#define CLOCK_TYPE CLOCK_MONOTONIC_RAW
+#else
+#define CLOCK_TYPE CLOCK_MONOTONIC
+#endif
+#define NSEC_TO_USEC_DIV 1000
+
+       struct timespec spec;
+       (void)clock_gettime(CLOCK_TYPE, &spec);
+
+       tv->tv_sec = spec.tv_sec;
+       tv->tv_usec = spec.tv_nsec / NSEC_TO_USEC_DIV;
+}
+
+uint64_t
+hns3_clock_calctime_ms(struct timeval *tv)
+{
+       return (uint64_t)tv->tv_sec * MSEC_PER_SEC +
+               tv->tv_usec / USEC_PER_MSEC;
+}
+
+uint64_t
+hns3_clock_gettime_ms(void)
+{
+       struct timeval tv;
+
+       hns3_clock_gettime(&tv);
+       return hns3_clock_calctime_ms(&tv);
+}
+
 static int
 hns3_parse_io_hint_func(const char *key, const char *value, void *extra_args)
 {
@@ -6676,6 +7256,19 @@ hns3_get_io_hint_func_name(uint32_t hint)
        }
 }
 
+static int
+hns3_parse_dev_caps_mask(const char *key, const char *value, void *extra_args)
+{
+       uint64_t val;
+
+       RTE_SET_USED(key);
+
+       val = strtoull(value, NULL, 16);
+       *(uint64_t *)extra_args = val;
+
+       return 0;
+}
+
 void
 hns3_parse_devargs(struct rte_eth_dev *dev)
 {
@@ -6683,6 +7276,7 @@ hns3_parse_devargs(struct rte_eth_dev *dev)
        uint32_t rx_func_hint = HNS3_IO_FUNC_HINT_NONE;
        uint32_t tx_func_hint = HNS3_IO_FUNC_HINT_NONE;
        struct hns3_hw *hw = &hns->hw;
+       uint64_t dev_caps_mask = 0;
        struct rte_kvargs *kvlist;
 
        if (dev->device->devargs == NULL)
@@ -6692,10 +7286,12 @@ hns3_parse_devargs(struct rte_eth_dev *dev)
        if (!kvlist)
                return;
 
-       rte_kvargs_process(kvlist, HNS3_DEVARG_RX_FUNC_HINT,
+       (void)rte_kvargs_process(kvlist, HNS3_DEVARG_RX_FUNC_HINT,
                           &hns3_parse_io_hint_func, &rx_func_hint);
-       rte_kvargs_process(kvlist, HNS3_DEVARG_TX_FUNC_HINT,
+       (void)rte_kvargs_process(kvlist, HNS3_DEVARG_TX_FUNC_HINT,
                           &hns3_parse_io_hint_func, &tx_func_hint);
+       (void)rte_kvargs_process(kvlist, HNS3_DEVARG_DEV_CAPS_MASK,
+                          &hns3_parse_dev_caps_mask, &dev_caps_mask);
        rte_kvargs_free(kvlist);
 
        if (rx_func_hint != HNS3_IO_FUNC_HINT_NONE)
@@ -6706,6 +7302,11 @@ hns3_parse_devargs(struct rte_eth_dev *dev)
                hns3_warn(hw, "parsed %s = %s.", HNS3_DEVARG_TX_FUNC_HINT,
                          hns3_get_io_hint_func_name(tx_func_hint));
        hns->tx_func_hint = tx_func_hint;
+
+       if (dev_caps_mask != 0)
+               hns3_warn(hw, "parsed %s = 0x%" PRIx64 ".",
+                         HNS3_DEVARG_DEV_CAPS_MASK, dev_caps_mask);
+       hns->dev_caps_mask = dev_caps_mask;
 }
 
 static const struct eth_dev_ops hns3_eth_dev_ops = {
@@ -6806,8 +7407,8 @@ hns3_dev_init(struct rte_eth_dev *eth_dev)
                PMD_INIT_LOG(ERR, "Failed to alloc memory for process private");
                return -ENOMEM;
        }
-       /* initialize flow filter lists */
-       hns3_filterlist_init(eth_dev);
+
+       hns3_flow_init(eth_dev);
 
        hns3_set_rxtx_function(eth_dev);
        eth_dev->dev_ops = &hns3_eth_dev_ops;
@@ -6819,8 +7420,8 @@ hns3_dev_init(struct rte_eth_dev *eth_dev)
                                     "process, ret = %d", ret);
                        goto err_mp_init_secondary;
                }
-
                hw->secondary_cnt++;
+               hns3_tx_push_init(eth_dev);
                return 0;
        }
 
@@ -6973,6 +7574,7 @@ RTE_PMD_REGISTER_PCI_TABLE(net_hns3, pci_id_hns3_map);
 RTE_PMD_REGISTER_KMOD_DEP(net_hns3, "* igb_uio | vfio-pci");
 RTE_PMD_REGISTER_PARAM_STRING(net_hns3,
                HNS3_DEVARG_RX_FUNC_HINT "=vec|sve|simple|common "
-               HNS3_DEVARG_TX_FUNC_HINT "=vec|sve|simple|common ");
-RTE_LOG_REGISTER(hns3_logtype_init, pmd.net.hns3.init, NOTICE);
-RTE_LOG_REGISTER(hns3_logtype_driver, pmd.net.hns3.driver, NOTICE);
+               HNS3_DEVARG_TX_FUNC_HINT "=vec|sve|simple|common "
+               HNS3_DEVARG_DEV_CAPS_MASK "=<1-65535> ");
+RTE_LOG_REGISTER_SUFFIX(hns3_logtype_init, init, NOTICE);
+RTE_LOG_REGISTER_SUFFIX(hns3_logtype_driver, driver, NOTICE);