ethdev: add device flag to bypass auto-filled queue xstats
[dpdk.git] / drivers / net / bnxt / bnxt_ethdev.c
index d0d11a6..6c12369 100644 (file)
@@ -864,7 +864,11 @@ uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
                speed_capa |= ETH_LINK_SPEED_50G;
        if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
                speed_capa |= ETH_LINK_SPEED_100G;
-       if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_200GB)
+       if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
+               speed_capa |= ETH_LINK_SPEED_50G;
+       if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
+               speed_capa |= ETH_LINK_SPEED_100G;
+       if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
                speed_capa |= ETH_LINK_SPEED_200G;
 
        if (bp->link_info->auto_mode ==
@@ -916,7 +920,9 @@ static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
        dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
        if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
                dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
-       dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT;
+       dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
+       dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
+                                   dev_info->tx_queue_offload_capa;
        dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
 
        dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
@@ -1187,6 +1193,7 @@ bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
 {
 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
 #ifndef RTE_LIBRTE_IEEE1588
+       uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
        struct bnxt *bp = eth_dev->data->dev_private;
 
        /*
@@ -1194,7 +1201,7 @@ bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
         * or tx offloads.
         */
        if (!eth_dev->data->scattered_rx &&
-           !eth_dev->data->dev_conf.txmode.offloads &&
+           !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
            !BNXT_TRUFLOW_EN(bp)) {
                PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
                            eth_dev->data->port_id);
@@ -1206,7 +1213,7 @@ bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
                    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
                    eth_dev->data->port_id,
                    eth_dev->data->scattered_rx,
-                   eth_dev->data->dev_conf.txmode.offloads);
+                   offloads);
 #endif
 #endif
        return bnxt_xmit_pkts;
@@ -1275,7 +1282,7 @@ static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
        eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
        eth_dev->data->dev_started = 1;
 
-       bnxt_link_update(eth_dev, 1, ETH_LINK_UP);
+       bnxt_link_update_op(eth_dev, 1);
 
        if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
                vlan_mask |= ETH_VLAN_FILTER_MASK;
@@ -1338,11 +1345,13 @@ static void bnxt_free_switch_domain(struct bnxt *bp)
 }
 
 /* Unload the driver, release resources */
-static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
+static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 {
        struct bnxt *bp = eth_dev->data->dev_private;
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
        struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct rte_eth_link link;
+       int ret;
 
        eth_dev->data->dev_started = 0;
        eth_dev->data->scattered_rx = 0;
@@ -1357,7 +1366,9 @@ static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
        rte_intr_disable(intr_handle);
 
        /* Stop the child representors for this device */
-       bnxt_rep_stop_all(bp);
+       ret = bnxt_rep_stop_all(bp);
+       if (ret != 0)
+               return ret;
 
        /* delete the bnxt ULP port details */
        bnxt_ulp_port_deinit(bp);
@@ -1365,15 +1376,15 @@ static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
        bnxt_cancel_fw_health_check(bp);
 
        /* Do not bring link down during reset recovery */
-       if (!is_bnxt_in_error(bp))
+       if (!is_bnxt_in_error(bp)) {
                bnxt_dev_set_link_down_op(eth_dev);
-
-       /* Wait for link to be reset and the async notification to process.
-        * During reset recovery, there is no need to wait and
-        * VF/NPAR functions do not have privilege to change PHY config.
-        */
-       if (!is_bnxt_in_error(bp) && BNXT_SINGLE_PF(bp))
-               bnxt_link_update(eth_dev, 1, ETH_LINK_DOWN);
+               /* Wait for link to be reset */
+               if (BNXT_SINGLE_PF(bp))
+                       rte_delay_ms(500);
+               /* clear the recorded link status */
+               memset(&link, 0, sizeof(link));
+               rte_eth_linkstatus_set(eth_dev, &link);
+       }
 
        /* Clean queue intr-vector mapping */
        rte_intr_efd_disable(intr_handle);
@@ -1398,11 +1409,17 @@ static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
        /* All filters are deleted on a port stop. */
        if (BNXT_FLOW_XSTATS_EN(bp))
                bp->flow_stat->flow_count = 0;
+
+       return 0;
 }
 
-static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
+static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 {
        struct bnxt *bp = eth_dev->data->dev_private;
+       int ret = 0;
+
+       if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+               return 0;
 
        /* cancel the recovery handler before remove dev */
        rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
@@ -1410,7 +1427,7 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
        bnxt_cancel_fc_thread(bp);
 
        if (eth_dev->data->dev_started)
-               bnxt_dev_stop_op(eth_dev);
+               ret = bnxt_dev_stop_op(eth_dev);
 
        bnxt_free_switch_domain(bp);
 
@@ -1422,10 +1439,6 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
        bnxt_free_pf_info(bp);
        bnxt_free_parent_info(bp);
 
-       eth_dev->dev_ops = NULL;
-       eth_dev->rx_pkt_burst = NULL;
-       eth_dev->tx_pkt_burst = NULL;
-
        rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
        bp->tx_mem_zone = NULL;
        rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
@@ -1435,6 +1448,8 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 
        rte_free(bp->grp_info);
        bp->grp_info = NULL;
+
+       return ret;
 }
 
 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
@@ -1548,14 +1563,13 @@ static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
        return rc;
 }
 
-int bnxt_link_update(struct rte_eth_dev *eth_dev, int wait_to_complete,
-                    bool exp_link_status)
+int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
 {
        int rc = 0;
        struct bnxt *bp = eth_dev->data->dev_private;
        struct rte_eth_link new;
-       int cnt = exp_link_status ? BNXT_LINK_UP_WAIT_CNT :
-                 BNXT_LINK_DOWN_WAIT_CNT;
+       int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
+                       BNXT_MIN_LINK_WAIT_CNT;
 
        rc = is_bnxt_in_error(bp);
        if (rc)
@@ -1573,12 +1587,18 @@ int bnxt_link_update(struct rte_eth_dev *eth_dev, int wait_to_complete,
                        goto out;
                }
 
-               if (!wait_to_complete || new.link_status == exp_link_status)
+               if (!wait_to_complete || new.link_status)
                        break;
 
                rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
        } while (cnt--);
 
+       /* Only single function PF can bring phy down.
+        * When port is stopped, report link down for VF/MH/NPAR functions.
+        */
+       if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
+               memset(&new, 0, sizeof(new));
+
 out:
        /* Timed out or success */
        if (new.link_status != eth_dev->data->dev_link.link_status ||
@@ -1595,12 +1615,6 @@ out:
        return rc;
 }
 
-int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
-                       int wait_to_complete)
-{
-       return bnxt_link_update(eth_dev, wait_to_complete, ETH_LINK_UP);
-}
-
 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
 {
        struct bnxt *bp = eth_dev->data->dev_private;
@@ -1889,6 +1903,9 @@ static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
        /* Update the default RSS VNIC(s) */
        vnic = BNXT_GET_DEFAULT_VNIC(bp);
        vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
+       vnic->hash_mode =
+               bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
+                                           ETH_RSS_LEVEL(rss_conf->rss_hf));
 
        /*
         * If hashkey is not specified, use the previously configured
@@ -1959,6 +1976,10 @@ static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
                        hash_types &=
                                ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
                }
+
+               rss_conf->rss_hf |=
+                       bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
+
                if (hash_types) {
                        PMD_DRV_LOG(ERR,
                                "Unknown RSS config from firmware (%08x), RSS disabled",
@@ -2173,14 +2194,6 @@ bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
        }
 
        rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
-       if (!rc) {
-               if (tunnel_type ==
-                   HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN)
-                       bp->vxlan_port = 0;
-               if (tunnel_type ==
-                   HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE)
-                       bp->geneve_port = 0;
-       }
        return rc;
 }
 
@@ -2645,6 +2658,7 @@ bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
        qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
        qinfo->conf.rx_drop_en = rxq->drop_en;
        qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
+       qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
 }
 
 static void
@@ -2668,6 +2682,7 @@ bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
        qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
        qinfo->conf.tx_rs_thresh = 0;
        qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
+       qinfo->conf.offloads = txq->offloads;
 }
 
 static const struct {
@@ -5187,37 +5202,14 @@ static void bnxt_config_vf_req_fwd(struct bnxt *bp)
        if (!BNXT_PF(bp))
                return;
 
-#define ALLOW_FUNC(x)  \
-       { \
-               uint32_t arg = (x); \
-               bp->pf->vf_req_fwd[((arg) >> 5)] &= \
-               ~rte_cpu_to_le_32(1 << ((arg) & 0x1f)); \
-       }
-
-       /* Forward all requests if firmware is new enough */
-       if (((bp->fw_ver >= ((20 << 24) | (6 << 16) | (100 << 8))) &&
-            (bp->fw_ver < ((20 << 24) | (7 << 16)))) ||
-           ((bp->fw_ver >= ((20 << 24) | (8 << 16))))) {
-               memset(bp->pf->vf_req_fwd, 0xff, sizeof(bp->pf->vf_req_fwd));
-       } else {
-               PMD_DRV_LOG(WARNING,
-                           "Firmware too old for VF mailbox functionality\n");
-               memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
-       }
+       memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
 
-       /*
-        * The following are used for driver cleanup. If we disallow these,
-        * VF drivers can't clean up cleanly.
-        */
-       ALLOW_FUNC(HWRM_FUNC_DRV_UNRGTR);
-       ALLOW_FUNC(HWRM_VNIC_FREE);
-       ALLOW_FUNC(HWRM_RING_FREE);
-       ALLOW_FUNC(HWRM_RING_GRP_FREE);
-       ALLOW_FUNC(HWRM_VNIC_RSS_COS_LB_CTX_FREE);
-       ALLOW_FUNC(HWRM_CFA_L2_FILTER_FREE);
-       ALLOW_FUNC(HWRM_STAT_CTX_FREE);
-       ALLOW_FUNC(HWRM_PORT_PHY_QCFG);
-       ALLOW_FUNC(HWRM_VNIC_TPA_CFG);
+       if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
+               BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
+       BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
+       BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
+       BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
+       BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
 }
 
 uint16_t
@@ -6005,6 +5997,7 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
                return 0;
 
        rte_eth_copy_pci_info(eth_dev, pci_dev);
+       eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
 
        bp = eth_dev->data->dev_private;
 
@@ -6068,11 +6061,6 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
 
        bnxt_alloc_switch_domain(bp);
 
-       /* Pass the information to the rte_eth_dev_close() that it should also
-        * release the private port resources.
-        */
-       eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
-
        PMD_DRV_LOG(INFO,
                    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
                    pci_dev->mem_resource[0].phys_addr,
@@ -6173,7 +6161,10 @@ bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
 
        bnxt_free_int(bp);
        bnxt_free_mem(bp, reconfig_dev);
+
        bnxt_hwrm_func_buf_unrgtr(bp);
+       rte_free(bp->pf->vf_req_buf);
+
        rc = bnxt_hwrm_func_driver_unregister(bp, 0);
        bp->flags &= ~BNXT_FLAG_REGISTERED;
        bnxt_free_ctx_mem(bp);
@@ -6466,6 +6457,10 @@ static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
        }
        PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
                    backing_eth_dev->data->port_id);
+
+       if (!num_rep)
+               return ret;
+
        /* probe representor ports now */
        ret = bnxt_rep_port_probe(pci_dev, eth_da, backing_eth_dev,
                                  pci_dev->device.devargs->args);