net/i40e: fix link down and negotiation
[dpdk.git] / drivers / net / i40e / i40e_ethdev.c
index 3f463ac..f3339b5 100644 (file)
@@ -515,6 +515,7 @@ static const struct eth_dev_ops i40e_eth_dev_ops = {
        .get_eeprom                   = i40e_get_eeprom,
        .mac_addr_set                 = i40e_set_default_mac_addr,
        .mtu_set                      = i40e_dev_mtu_set,
+       .tm_ops_get                   = i40e_tm_ops_get,
 };
 
 /* store statistics names and its offset in stats structure */
@@ -1295,6 +1296,9 @@ eth_i40e_dev_init(struct rte_eth_dev *dev)
        /* initialize mirror rule list */
        TAILQ_INIT(&pf->mirror_list);
 
+       /* initialize Traffic Manager configuration */
+       i40e_tm_conf_init(dev);
+
        ret = i40e_init_ethtype_filter_list(dev);
        if (ret < 0)
                goto err_init_ethtype_filter_list;
@@ -1457,6 +1461,9 @@ eth_i40e_dev_uninit(struct rte_eth_dev *dev)
                rte_free(p_flow);
        }
 
+       /* Remove all Traffic Manager configuration */
+       i40e_tm_conf_uninit(dev);
+
        return 0;
 }
 
@@ -1807,11 +1814,15 @@ i40e_parse_link_speeds(uint16_t link_speeds)
 static int
 i40e_phy_conf_link(struct i40e_hw *hw,
                   uint8_t abilities,
-                  uint8_t force_speed)
+                  uint8_t force_speed,
+                  bool is_up)
 {
        enum i40e_status_code status;
        struct i40e_aq_get_phy_abilities_resp phy_ab;
        struct i40e_aq_set_phy_config phy_conf;
+       enum i40e_aq_phy_type cnt;
+       uint32_t phy_type_mask = 0;
+
        const uint8_t mask = I40E_AQ_PHY_FLAG_PAUSE_TX |
                        I40E_AQ_PHY_FLAG_PAUSE_RX |
                        I40E_AQ_PHY_FLAG_PAUSE_RX |
@@ -1829,6 +1840,10 @@ i40e_phy_conf_link(struct i40e_hw *hw,
        if (status)
                return ret;
 
+       /* If link already up, no need to set up again */
+       if (is_up && phy_ab.phy_type != 0)
+               return I40E_SUCCESS;
+
        memset(&phy_conf, 0, sizeof(phy_conf));
 
        /* bits 0-2 use the values from get_phy_abilities_resp */
@@ -1839,13 +1854,21 @@ i40e_phy_conf_link(struct i40e_hw *hw,
        if (abilities & I40E_AQ_PHY_AN_ENABLED)
                phy_conf.link_speed = advt;
        else
-               phy_conf.link_speed = force_speed;
+               phy_conf.link_speed = is_up ? force_speed : phy_ab.link_speed;
 
        phy_conf.abilities = abilities;
 
+
+
+       /* To enable link, phy_type mask needs to include each type */
+       for (cnt = I40E_PHY_TYPE_SGMII; cnt < I40E_PHY_TYPE_MAX; cnt++)
+               phy_type_mask |= 1 << cnt;
+
        /* use get_phy_abilities_resp value for the rest */
-       phy_conf.phy_type = phy_ab.phy_type;
-       phy_conf.phy_type_ext = phy_ab.phy_type_ext;
+       phy_conf.phy_type = is_up ? cpu_to_le32(phy_type_mask) : 0;
+       phy_conf.phy_type_ext = is_up ? (I40E_AQ_PHY_TYPE_EXT_25G_KR |
+               I40E_AQ_PHY_TYPE_EXT_25G_CR | I40E_AQ_PHY_TYPE_EXT_25G_SR |
+               I40E_AQ_PHY_TYPE_EXT_25G_LR) : 0;
        phy_conf.fec_config = phy_ab.fec_cfg_curr_mod_ext_info;
        phy_conf.eee_capability = phy_ab.eee_capability;
        phy_conf.eeer = phy_ab.eeer_val;
@@ -1877,13 +1900,7 @@ i40e_apply_link_speed(struct rte_eth_dev *dev)
                abilities |= I40E_AQ_PHY_AN_ENABLED;
        abilities |= I40E_AQ_PHY_LINK_ENABLED;
 
-       /* Skip changing speed on 40G interfaces, FW does not support */
-       if (I40E_PHY_TYPE_SUPPORT_40G(hw->phy.phy_types)) {
-               speed =  I40E_LINK_SPEED_UNKNOWN;
-               abilities |= I40E_AQ_PHY_AN_ENABLED;
-       }
-
-       return i40e_phy_conf_link(hw, abilities, speed);
+       return i40e_phy_conf_link(hw, abilities, speed, true);
 }
 
 static int
@@ -2009,7 +2026,7 @@ i40e_dev_start(struct rte_eth_dev *dev)
                if (dev->data->dev_conf.intr_conf.lsc != 0)
                        PMD_INIT_LOG(INFO,
                                "lsc won't enable because of no intr multiplex");
-       } else if (dev->data->dev_conf.intr_conf.lsc != 0) {
+       } else {
                ret = i40e_aq_set_phy_int_mask(hw,
                                               ~(I40E_AQ_EVENT_LINK_UPDOWN |
                                               I40E_AQ_EVENT_MODULE_QUAL_FAIL |
@@ -2017,7 +2034,7 @@ i40e_dev_start(struct rte_eth_dev *dev)
                if (ret != I40E_SUCCESS)
                        PMD_DRV_LOG(WARNING, "Fail to set phy mask");
 
-               /* Call get_link_info aq commond to enable LSE */
+               /* Call get_link_info aq commond to enable/disable LSE */
                i40e_dev_link_update(dev, 0);
        }
 
@@ -2026,6 +2043,11 @@ i40e_dev_start(struct rte_eth_dev *dev)
 
        i40e_filter_restore(pf);
 
+       if (pf->tm_conf.root && !pf->tm_conf.committed)
+               PMD_DRV_LOG(WARNING,
+                           "please call hierarchy_commit() "
+                           "before starting the port");
+
        return I40E_SUCCESS;
 
 err_up:
@@ -2039,12 +2061,15 @@ static void
 i40e_dev_stop(struct rte_eth_dev *dev)
 {
        struct i40e_pf *pf = I40E_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+       struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
        struct i40e_vsi *main_vsi = pf->main_vsi;
        struct i40e_mirror_rule *p_mirror;
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
        struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
        int i;
 
+       if (hw->adapter_stopped == 1)
+               return;
        /* Disable all queues */
        i40e_dev_switch_queues(pf, FALSE);
 
@@ -2086,6 +2111,11 @@ i40e_dev_stop(struct rte_eth_dev *dev)
                rte_free(intr_handle->intr_vec);
                intr_handle->intr_vec = NULL;
        }
+
+       /* reset hierarchy commit */
+       pf->tm_conf.committed = false;
+
+       hw->adapter_stopped = 1;
 }
 
 static void
@@ -2101,7 +2131,6 @@ i40e_dev_close(struct rte_eth_dev *dev)
        PMD_INIT_FUNC_TRACE();
 
        i40e_dev_stop(dev);
-       hw->adapter_stopped = 1;
        i40e_dev_free_queues(dev);
 
        /* Disable interrupt */
@@ -2226,7 +2255,7 @@ i40e_dev_set_link_down(struct rte_eth_dev *dev)
        struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 
        abilities = I40E_AQ_PHY_ENABLE_ATOMIC_LINK;
-       return i40e_phy_conf_link(hw, abilities, speed);
+       return i40e_phy_conf_link(hw, abilities, speed, false);
 }
 
 int
@@ -2353,9 +2382,6 @@ i40e_update_vsi_stats(struct i40e_vsi *vsi)
        i40e_stat_update_48(hw, I40E_GLV_BPTCH(idx), I40E_GLV_BPTCL(idx),
                            vsi->offset_loaded,  &oes->tx_broadcast,
                            &nes->tx_broadcast);
-       /* exclude CRC bytes */
-       nes->tx_bytes -= (nes->tx_unicast + nes->tx_multicast +
-               nes->tx_broadcast) * ETHER_CRC_LEN;
        /* GLV_TDPC not supported */
        i40e_stat_update_32(hw, I40E_GLV_TEPC(idx), vsi->offset_loaded,
                            &oes->tx_errors, &nes->tx_errors);
@@ -2391,14 +2417,35 @@ i40e_read_stats_registers(struct i40e_pf *pf, struct i40e_hw *hw)
        i40e_stat_update_48(hw, I40E_GLV_GORCH(hw->port),
                        I40E_GLV_GORCL(hw->port),
                        pf->offset_loaded,
-                       &pf->internal_rx_bytes_offset,
-                       &pf->internal_rx_bytes);
+                       &pf->internal_stats_offset.rx_bytes,
+                       &pf->internal_stats.rx_bytes);
 
        i40e_stat_update_48(hw, I40E_GLV_GOTCH(hw->port),
                        I40E_GLV_GOTCL(hw->port),
                        pf->offset_loaded,
-                       &pf->internal_tx_bytes_offset,
-                       &pf->internal_tx_bytes);
+                       &pf->internal_stats_offset.tx_bytes,
+                       &pf->internal_stats.tx_bytes);
+       /* Get total internal rx packet count */
+       i40e_stat_update_48(hw, I40E_GLV_UPRCH(hw->port),
+                           I40E_GLV_UPRCL(hw->port),
+                           pf->offset_loaded,
+                           &pf->internal_stats_offset.rx_unicast,
+                           &pf->internal_stats.rx_unicast);
+       i40e_stat_update_48(hw, I40E_GLV_MPRCH(hw->port),
+                           I40E_GLV_MPRCL(hw->port),
+                           pf->offset_loaded,
+                           &pf->internal_stats_offset.rx_multicast,
+                           &pf->internal_stats.rx_multicast);
+       i40e_stat_update_48(hw, I40E_GLV_BPRCH(hw->port),
+                           I40E_GLV_BPRCL(hw->port),
+                           pf->offset_loaded,
+                           &pf->internal_stats_offset.rx_broadcast,
+                           &pf->internal_stats.rx_broadcast);
+
+       /* exclude CRC size */
+       pf->internal_stats.rx_bytes -= (pf->internal_stats.rx_unicast +
+               pf->internal_stats.rx_multicast +
+               pf->internal_stats.rx_broadcast) * ETHER_CRC_LEN;
 
        /* Get statistics of struct i40e_eth_stats */
        i40e_stat_update_48(hw, I40E_GLPRT_GORCH(hw->port),
@@ -2421,7 +2468,17 @@ i40e_read_stats_registers(struct i40e_pf *pf, struct i40e_hw *hw)
         * so subtract ETHER_CRC_LEN from the byte counter for each rx packet.
         */
        ns->eth.rx_bytes -= (ns->eth.rx_unicast + ns->eth.rx_multicast +
-               ns->eth.rx_broadcast) * ETHER_CRC_LEN + pf->internal_rx_bytes;
+               ns->eth.rx_broadcast) * ETHER_CRC_LEN;
+
+       /* Workaround: it is possible I40E_GLV_GORCH[H/L] is updated before
+        * I40E_GLPRT_GORCH[H/L], so there is a small window that cause negtive
+        * value.
+        */
+       if (ns->eth.rx_bytes < pf->internal_stats.rx_bytes)
+               ns->eth.rx_bytes = 0;
+       /* exlude internal rx bytes */
+       else
+               ns->eth.rx_bytes -= pf->internal_stats.rx_bytes;
 
        i40e_stat_update_32(hw, I40E_GLPRT_RDPC(hw->port),
                            pf->offset_loaded, &os->eth.rx_discards,
@@ -2449,7 +2506,14 @@ i40e_read_stats_registers(struct i40e_pf *pf, struct i40e_hw *hw)
                            pf->offset_loaded, &os->eth.tx_broadcast,
                            &ns->eth.tx_broadcast);
        ns->eth.tx_bytes -= (ns->eth.tx_unicast + ns->eth.tx_multicast +
-               ns->eth.tx_broadcast) * ETHER_CRC_LEN + pf->internal_tx_bytes;
+               ns->eth.tx_broadcast) * ETHER_CRC_LEN;
+
+       /* exclude internal tx bytes */
+       if (ns->eth.tx_bytes < pf->internal_stats.tx_bytes)
+               ns->eth.tx_bytes = 0;
+       else
+               ns->eth.tx_bytes -= pf->internal_stats.tx_bytes;
+
        /* GLPRT_TEPC not supported */
 
        /* additional port specific stats */
@@ -2974,71 +3038,93 @@ i40e_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 }
 
 static int
-i40e_vlan_tpid_set(struct rte_eth_dev *dev,
-                  enum rte_vlan_type vlan_type,
-                  uint16_t tpid)
+i40e_vlan_tpid_set_by_registers(struct rte_eth_dev *dev,
+                               enum rte_vlan_type vlan_type,
+                               uint16_t tpid, int qinq)
 {
        struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-       uint64_t reg_r = 0, reg_w = 0;
-       uint16_t reg_id = 0;
-       int ret = 0;
-       int qinq = dev->data->dev_conf.rxmode.hw_vlan_extend;
+       uint64_t reg_r = 0;
+       uint64_t reg_w = 0;
+       uint16_t reg_id = 3;
+       int ret;
 
-       switch (vlan_type) {
-       case ETH_VLAN_TYPE_OUTER:
-               if (qinq)
+       if (qinq) {
+               if (vlan_type == ETH_VLAN_TYPE_OUTER)
                        reg_id = 2;
-               else
-                       reg_id = 3;
-               break;
-       case ETH_VLAN_TYPE_INNER:
-               if (qinq)
-                       reg_id = 3;
-               else {
-                       ret = -EINVAL;
-                       PMD_DRV_LOG(ERR,
-                               "Unsupported vlan type in single vlan.");
-                       return ret;
-               }
-               break;
-       default:
-               ret = -EINVAL;
-               PMD_DRV_LOG(ERR, "Unsupported vlan type %d", vlan_type);
-               return ret;
        }
+
        ret = i40e_aq_debug_read_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
                                          &reg_r, NULL);
        if (ret != I40E_SUCCESS) {
                PMD_DRV_LOG(ERR,
                           "Fail to debug read from I40E_GL_SWT_L2TAGCTRL[%d]",
                           reg_id);
-               ret = -EIO;
-               return ret;
+               return -EIO;
        }
        PMD_DRV_LOG(DEBUG,
-               "Debug read from I40E_GL_SWT_L2TAGCTRL[%d]: 0x%08"PRIx64,
-               reg_id, reg_r);
+                   "Debug read from I40E_GL_SWT_L2TAGCTRL[%d]: 0x%08"PRIx64,
+                   reg_id, reg_r);
 
        reg_w = reg_r & (~(I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_MASK));
        reg_w |= ((uint64_t)tpid << I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT);
        if (reg_r == reg_w) {
-               ret = 0;
                PMD_DRV_LOG(DEBUG, "No need to write");
-               return ret;
+               return 0;
        }
 
        ret = i40e_aq_debug_write_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
                                           reg_w, NULL);
        if (ret != I40E_SUCCESS) {
-               ret = -EIO;
                PMD_DRV_LOG(ERR,
-                       "Fail to debug write to I40E_GL_SWT_L2TAGCTRL[%d]",
-                       reg_id);
-               return ret;
+                           "Fail to debug write to I40E_GL_SWT_L2TAGCTRL[%d]",
+                           reg_id);
+               return -EIO;
        }
        PMD_DRV_LOG(DEBUG,
-               "Debug write 0x%08"PRIx64" to I40E_GL_SWT_L2TAGCTRL[%d]",
-               reg_w, reg_id);
+                   "Debug write 0x%08"PRIx64" to I40E_GL_SWT_L2TAGCTRL[%d]",
+                   reg_w, reg_id);
+
+       return 0;
+}
+
+static int
+i40e_vlan_tpid_set(struct rte_eth_dev *dev,
+                  enum rte_vlan_type vlan_type,
+                  uint16_t tpid)
+{
+       struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       int qinq = dev->data->dev_conf.rxmode.hw_vlan_extend;
+       int ret = 0;
+
+       if ((vlan_type != ETH_VLAN_TYPE_INNER &&
+            vlan_type != ETH_VLAN_TYPE_OUTER) ||
+           (!qinq && vlan_type == ETH_VLAN_TYPE_INNER)) {
+               PMD_DRV_LOG(ERR,
+                           "Unsupported vlan type.");
+               return -EINVAL;
+       }
+       /* 802.1ad frames ability is added in NVM API 1.7*/
+       if (hw->flags & I40E_HW_FLAG_802_1AD_CAPABLE) {
+               if (qinq) {
+                       if (vlan_type == ETH_VLAN_TYPE_OUTER)
+                               hw->first_tag = rte_cpu_to_le_16(tpid);
+                       else if (vlan_type == ETH_VLAN_TYPE_INNER)
+                               hw->second_tag = rte_cpu_to_le_16(tpid);
+               } else {
+                       if (vlan_type == ETH_VLAN_TYPE_OUTER)
+                               hw->second_tag = rte_cpu_to_le_16(tpid);
+               }
+               ret = i40e_aq_set_switch_config(hw, 0, 0, NULL);
+               if (ret != I40E_SUCCESS) {
+                       PMD_DRV_LOG(ERR,
+                                   "Set switch config failed aq_err: %d",
+                                   hw->aq.asq_last_status);
+                       ret = -EIO;
+               }
+       } else
+               /* If NVM API < 1.7, keep the register setting */
+               ret = i40e_vlan_tpid_set_by_registers(dev, vlan_type,
+                                                     tpid, qinq);
 
        return ret;
 }
@@ -3067,7 +3153,7 @@ i40e_vlan_offload_set(struct rte_eth_dev *dev, int mask)
        if (mask & ETH_VLAN_EXTEND_MASK) {
                if (dev->data->dev_conf.rxmode.hw_vlan_extend) {
                        i40e_vsi_config_double_vlan(vsi, TRUE);
-                       /* Set global registers with default ether type value */
+                       /* Set global registers with default ethertype. */
                        i40e_vlan_tpid_set(dev, ETH_VLAN_TYPE_OUTER,
                                           ETHER_TYPE_VLAN);
                        i40e_vlan_tpid_set(dev, ETH_VLAN_TYPE_INNER,
@@ -4285,6 +4371,8 @@ i40e_vsi_config_tc_queue_mapping(struct i40e_vsi *vsi,
        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
                if (enabled_tcmap & (1 << i))
                        total_tc++;
+       if (total_tc == 0)
+               total_tc = 1;
        vsi->enabled_tc = enabled_tcmap;
 
        /* Number of queues per enabled TC */
@@ -5217,10 +5305,8 @@ i40e_pf_setup(struct i40e_pf *pf)
        pf->offset_loaded = FALSE;
        memset(&pf->stats, 0, sizeof(struct i40e_hw_port_stats));
        memset(&pf->stats_offset, 0, sizeof(struct i40e_hw_port_stats));
-       pf->internal_rx_bytes = 0;
-       pf->internal_tx_bytes = 0;
-       pf->internal_rx_bytes_offset = 0;
-       pf->internal_tx_bytes_offset = 0;
+       memset(&pf->internal_stats, 0, sizeof(struct i40e_eth_stats));
+       memset(&pf->internal_stats_offset, 0, sizeof(struct i40e_eth_stats));
 
        ret = i40e_pf_get_switch_config(pf);
        if (ret != I40E_SUCCESS) {
@@ -9151,8 +9237,9 @@ i40e_pctype_to_flowtype(enum i40e_filter_pctype pctype)
  */
 
 /* For both X710 and XL710 */
-#define I40E_GL_SWR_PRI_JOIN_MAP_0_VALUE 0x10000200
-#define I40E_GL_SWR_PRI_JOIN_MAP_0       0x26CE00
+#define I40E_GL_SWR_PRI_JOIN_MAP_0_VALUE_1     0x10000200
+#define I40E_GL_SWR_PRI_JOIN_MAP_0_VALUE_2     0x20000200
+#define I40E_GL_SWR_PRI_JOIN_MAP_0             0x26CE00
 
 #define I40E_GL_SWR_PRI_JOIN_MAP_2_VALUE 0x011f0200
 #define I40E_GL_SWR_PRI_JOIN_MAP_2       0x26CE08
@@ -9207,8 +9294,12 @@ i40e_configure_registers(struct i40e_hw *hw)
                                reg_table[i].val =
                                        I40E_X722_GL_SWR_PRI_JOIN_MAP_0_VALUE;
                        else /* For X710/XL710/XXV710 */
-                               reg_table[i].val =
-                                       I40E_GL_SWR_PRI_JOIN_MAP_0_VALUE;
+                               if (hw->aq.fw_maj_ver < 6)
+                                       reg_table[i].val =
+                                            I40E_GL_SWR_PRI_JOIN_MAP_0_VALUE_1;
+                               else
+                                       reg_table[i].val =
+                                            I40E_GL_SWR_PRI_JOIN_MAP_0_VALUE_2;
                }
 
                if (reg_table[i].addr == I40E_GL_SWR_PRI_JOIN_MAP_2) {