ethdev: add namespace
[dpdk.git] / drivers / net / pfe / pfe_ethdev.c
index 7f238fc..047010e 100644 (file)
@@ -1,27 +1,65 @@
 /* SPDX-License-Identifier: BSD-3-Clause
- * Copyright 2019 NXP
+ * Copyright 2018-2019 NXP
  */
 
+#include <sys/ioctl.h>
+#include <sys/epoll.h>
 #include <rte_kvargs.h>
-#include <rte_ethdev_vdev.h>
+#include <ethdev_vdev.h>
 #include <rte_bus_vdev.h>
+#include <rte_ether.h>
 #include <dpaa_of.h>
 
+#include "pfe_logs.h"
 #include "pfe_mod.h"
 
-#define PFE_MAX_MACS 1 /*we can support upto 4 MACs per IF*/
+#define PFE_MAX_MACS 1 /* we can support up to 4 MACs per IF */
 #define PFE_VDEV_GEM_ID_ARG    "intf"
 
 struct pfe_vdev_init_params {
        int8_t  gem_id;
 };
 static struct pfe *g_pfe;
+/* Supported Rx offloads */
+static uint64_t dev_rx_offloads_sup =
+               RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
+               RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
+               RTE_ETH_RX_OFFLOAD_TCP_CKSUM;
+
+/* Supported Tx offloads */
+static uint64_t dev_tx_offloads_sup =
+               RTE_ETH_TX_OFFLOAD_IPV4_CKSUM |
+               RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
+               RTE_ETH_TX_OFFLOAD_TCP_CKSUM;
 
 /* TODO: make pfe_svr a runtime option.
  * Driver should be able to get the SVR
  * information from HW.
  */
 unsigned int pfe_svr = SVR_LS1012A_REV1;
+static void *cbus_emac_base[3];
+static void *cbus_gpi_base[3];
+
+/* pfe_gemac_init
+ */
+static int
+pfe_gemac_init(struct pfe_eth_priv_s *priv)
+{
+       struct gemac_cfg cfg;
+
+       cfg.speed = SPEED_1000M;
+       cfg.duplex = DUPLEX_FULL;
+
+       gemac_set_config(priv->EMAC_baseaddr, &cfg);
+       gemac_allow_broadcast(priv->EMAC_baseaddr);
+       gemac_enable_1536_rx(priv->EMAC_baseaddr);
+       gemac_enable_stacked_vlan(priv->EMAC_baseaddr);
+       gemac_enable_pause_rx(priv->EMAC_baseaddr);
+       gemac_set_bus_width(priv->EMAC_baseaddr, 64);
+       gemac_enable_rx_checksum_offload(priv->EMAC_baseaddr);
+
+       return 0;
+}
 
 static void
 pfe_soc_version_get(void)
@@ -29,18 +67,278 @@ pfe_soc_version_get(void)
        FILE *svr_file = NULL;
        unsigned int svr_ver = 0;
 
+       PMD_INIT_FUNC_TRACE();
+
        svr_file = fopen(PFE_SOC_ID_FILE, "r");
-       if (!svr_file)
+       if (!svr_file) {
+               PFE_PMD_ERR("Unable to open SoC device");
                return; /* Not supported on this infra */
+       }
 
        if (fscanf(svr_file, "svr:%x", &svr_ver) > 0)
                pfe_svr = svr_ver;
        else
-               printf("Unable to read SoC device");
+               PFE_PMD_ERR("Unable to read SoC device");
 
        fclose(svr_file);
 }
 
+static int pfe_eth_start(struct pfe_eth_priv_s *priv)
+{
+       gpi_enable(priv->GPI_baseaddr);
+       gemac_enable(priv->EMAC_baseaddr);
+
+       return 0;
+}
+
+static void
+pfe_eth_flush_txQ(struct pfe_eth_priv_s *priv, int tx_q_num, int
+                 __rte_unused from_tx, __rte_unused int n_desc)
+{
+       struct rte_mbuf *mbuf;
+       unsigned int flags;
+
+       /* Clean HIF and client queue */
+       while ((mbuf = hif_lib_tx_get_next_complete(&priv->client,
+                                                  tx_q_num, &flags,
+                                                  HIF_TX_DESC_NT))) {
+               if (mbuf) {
+                       mbuf->next = NULL;
+                       mbuf->nb_segs = 1;
+                       rte_pktmbuf_free(mbuf);
+               }
+       }
+}
+
+
+static void
+pfe_eth_flush_tx(struct pfe_eth_priv_s *priv)
+{
+       unsigned int ii;
+
+       for (ii = 0; ii < emac_txq_cnt; ii++)
+               pfe_eth_flush_txQ(priv, ii, 0, 0);
+}
+
+static int
+pfe_eth_event_handler(void *data, int event, __rte_unused int qno)
+{
+       struct pfe_eth_priv_s *priv = data;
+
+       switch (event) {
+       case EVENT_TXDONE_IND:
+               pfe_eth_flush_tx(priv);
+               hif_lib_event_handler_start(&priv->client, EVENT_TXDONE_IND, 0);
+               break;
+       case EVENT_HIGH_RX_WM:
+       default:
+               break;
+       }
+
+       return 0;
+}
+
+static uint16_t
+pfe_recv_pkts_on_intr(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+       struct hif_client_rx_queue *queue = rxq;
+       struct pfe_eth_priv_s *priv = queue->priv;
+       struct epoll_event epoll_ev;
+       uint64_t ticks = 1;  /* 1 msec */
+       int ret;
+       int have_something, work_done;
+
+#define RESET_STATUS (HIF_INT | HIF_RXPKT_INT)
+
+       /*TODO can we remove this cleanup from here?*/
+       pfe_tx_do_cleanup(priv->pfe);
+       have_something = pfe_hif_rx_process(priv->pfe, nb_pkts);
+       work_done = hif_lib_receive_pkt(rxq, priv->pfe->hif.shm->pool,
+                       rx_pkts, nb_pkts);
+
+       if (!have_something || !work_done) {
+               writel(RESET_STATUS, HIF_INT_SRC);
+               writel(readl(HIF_INT_ENABLE) | HIF_RXPKT_INT, HIF_INT_ENABLE);
+               ret = epoll_wait(priv->pfe->hif.epoll_fd, &epoll_ev, 1, ticks);
+               if (ret < 0 && errno != EINTR)
+                       PFE_PMD_ERR("epoll_wait fails with %d\n", errno);
+       }
+
+       return work_done;
+}
+
+static uint16_t
+pfe_recv_pkts(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+       struct hif_client_rx_queue *queue = rxq;
+       struct pfe_eth_priv_s *priv = queue->priv;
+       struct rte_mempool *pool;
+
+       /*TODO can we remove this cleanup from here?*/
+       pfe_tx_do_cleanup(priv->pfe);
+       pfe_hif_rx_process(priv->pfe, nb_pkts);
+       pool = priv->pfe->hif.shm->pool;
+
+       return hif_lib_receive_pkt(rxq, pool, rx_pkts, nb_pkts);
+}
+
+static uint16_t
+pfe_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+       struct hif_client_tx_queue *queue = tx_queue;
+       struct pfe_eth_priv_s *priv = queue->priv;
+       struct rte_eth_stats *stats = &priv->stats;
+       int i;
+
+       for (i = 0; i < nb_pkts; i++) {
+               if (tx_pkts[i]->nb_segs > 1) {
+                       struct rte_mbuf *mbuf;
+                       int j;
+
+                       hif_lib_xmit_pkt(&priv->client, queue->queue_id,
+                               (void *)(size_t)rte_pktmbuf_iova(tx_pkts[i]),
+                               tx_pkts[i]->buf_addr + tx_pkts[i]->data_off,
+                               tx_pkts[i]->data_len, 0x0, HIF_FIRST_BUFFER,
+                               tx_pkts[i]);
+
+                       mbuf = tx_pkts[i]->next;
+                       for (j = 0; j < (tx_pkts[i]->nb_segs - 2); j++) {
+                               hif_lib_xmit_pkt(&priv->client, queue->queue_id,
+                                       (void *)(size_t)rte_pktmbuf_iova(mbuf),
+                                       mbuf->buf_addr + mbuf->data_off,
+                                       mbuf->data_len,
+                                       0x0, 0x0, mbuf);
+                               mbuf = mbuf->next;
+                       }
+
+                       hif_lib_xmit_pkt(&priv->client, queue->queue_id,
+                                       (void *)(size_t)rte_pktmbuf_iova(mbuf),
+                                       mbuf->buf_addr + mbuf->data_off,
+                                       mbuf->data_len,
+                                       0x0, HIF_LAST_BUFFER | HIF_DATA_VALID,
+                                       mbuf);
+               } else {
+                       hif_lib_xmit_pkt(&priv->client, queue->queue_id,
+                               (void *)(size_t)rte_pktmbuf_iova(tx_pkts[i]),
+                               tx_pkts[i]->buf_addr + tx_pkts[i]->data_off,
+                               tx_pkts[i]->pkt_len, 0 /*ctrl*/,
+                               HIF_FIRST_BUFFER | HIF_LAST_BUFFER |
+                               HIF_DATA_VALID,
+                               tx_pkts[i]);
+               }
+               stats->obytes += tx_pkts[i]->pkt_len;
+               hif_tx_dma_start();
+       }
+       stats->opackets += nb_pkts;
+       pfe_tx_do_cleanup(priv->pfe);
+
+       return nb_pkts;
+}
+
+static uint16_t
+pfe_dummy_xmit_pkts(__rte_unused void *tx_queue,
+               __rte_unused struct rte_mbuf **tx_pkts,
+               __rte_unused uint16_t nb_pkts)
+{
+       return 0;
+}
+
+static uint16_t
+pfe_dummy_recv_pkts(__rte_unused void *rxq,
+               __rte_unused struct rte_mbuf **rx_pkts,
+               __rte_unused uint16_t nb_pkts)
+{
+       return 0;
+}
+
+static int
+pfe_eth_open(struct rte_eth_dev *dev)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+       struct hif_client_s *client;
+       struct hif_shm *hif_shm;
+       int rc;
+
+       /* Register client driver with HIF */
+       client = &priv->client;
+
+       if (client->pfe) {
+               hif_shm = client->pfe->hif.shm;
+               /* TODO please remove the below code of if block, once we add
+                * the proper cleanup in eth_close
+                */
+               if (!test_bit(PFE_CL_GEM0 + priv->id,
+                             &hif_shm->g_client_status[0])) {
+                       /* Register client driver with HIF */
+                       memset(client, 0, sizeof(*client));
+                       client->id = PFE_CL_GEM0 + priv->id;
+                       client->tx_qn = emac_txq_cnt;
+                       client->rx_qn = EMAC_RXQ_CNT;
+                       client->priv = priv;
+                       client->pfe = priv->pfe;
+                       client->port_id = dev->data->port_id;
+                       client->event_handler = pfe_eth_event_handler;
+
+                       client->tx_qsize = EMAC_TXQ_DEPTH;
+                       client->rx_qsize = EMAC_RXQ_DEPTH;
+
+                       rc = hif_lib_client_register(client);
+                       if (rc) {
+                               PFE_PMD_ERR("hif_lib_client_register(%d)"
+                                           " failed", client->id);
+                               goto err0;
+                       }
+               } else {
+                       /* Freeing the packets if already exists */
+                       int ret = 0;
+                       struct rte_mbuf *rx_pkts[32];
+                       /* TODO multiqueue support */
+                       ret = hif_lib_receive_pkt(&client->rx_q[0],
+                                                 hif_shm->pool, rx_pkts, 32);
+                       while (ret) {
+                               int i;
+                               for (i = 0; i < ret; i++)
+                                       rte_pktmbuf_free(rx_pkts[i]);
+                               ret = hif_lib_receive_pkt(&client->rx_q[0],
+                                                         hif_shm->pool,
+                                                         rx_pkts, 32);
+                       }
+               }
+       } else {
+               /* Register client driver with HIF */
+               memset(client, 0, sizeof(*client));
+               client->id = PFE_CL_GEM0 + priv->id;
+               client->tx_qn = emac_txq_cnt;
+               client->rx_qn = EMAC_RXQ_CNT;
+               client->priv = priv;
+               client->pfe = priv->pfe;
+               client->port_id = dev->data->port_id;
+               client->event_handler = pfe_eth_event_handler;
+
+               client->tx_qsize = EMAC_TXQ_DEPTH;
+               client->rx_qsize = EMAC_RXQ_DEPTH;
+
+               rc = hif_lib_client_register(client);
+               if (rc) {
+                       PFE_PMD_ERR("hif_lib_client_register(%d) failed",
+                                   client->id);
+                       goto err0;
+               }
+       }
+       rc = pfe_eth_start(priv);
+       dev->rx_pkt_burst = &pfe_recv_pkts;
+       dev->tx_pkt_burst = &pfe_xmit_pkts;
+       /* If no prefetch is configured. */
+       if (getenv("PFE_INTR_SUPPORT")) {
+               dev->rx_pkt_burst = &pfe_recv_pkts_on_intr;
+               PFE_PMD_INFO("PFE INTERRUPT Mode enabled");
+       }
+
+
+err0:
+       return rc;
+}
+
 static int
 pfe_eth_open_cdev(struct pfe_eth_priv_s *priv)
 {
@@ -51,6 +349,9 @@ pfe_eth_open_cdev(struct pfe_eth_priv_s *priv)
 
        pfe_cdev_fd = open(PFE_CDEV_PATH, O_RDONLY);
        if (pfe_cdev_fd < 0) {
+               PFE_PMD_WARN("Unable to open PFE device file (%s).\n",
+                            PFE_CDEV_PATH);
+               PFE_PMD_WARN("Link status update will not be available.\n");
                priv->link_fd = PFE_CDEV_INVALID_FD;
                return -1;
        }
@@ -72,34 +373,432 @@ pfe_eth_close_cdev(struct pfe_eth_priv_s *priv)
        }
 }
 
-static void
-pfe_eth_exit(struct rte_eth_dev *dev, struct pfe *pfe)
+static int
+pfe_eth_stop(struct rte_eth_dev *dev/*, int wake*/)
 {
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+
+       dev->data->dev_started = 0;
+
+       gemac_disable(priv->EMAC_baseaddr);
+       gpi_disable(priv->GPI_baseaddr);
+
+       dev->rx_pkt_burst = &pfe_dummy_recv_pkts;
+       dev->tx_pkt_burst = &pfe_dummy_xmit_pkts;
+
+       return 0;
+}
+
+static int
+pfe_eth_close(struct rte_eth_dev *dev)
+{
+       int ret;
+       PMD_INIT_FUNC_TRACE();
+
+       if (!dev)
+               return -1;
+
+       if (!g_pfe)
+               return -1;
+
+       if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+               return 0;
+
+       ret = pfe_eth_stop(dev);
        /* Close the device file for link status */
        pfe_eth_close_cdev(dev->data->dev_private);
 
-       rte_free(dev->data->mac_addrs);
-       rte_eth_dev_release_port(dev);
-       pfe->nb_devs--;
+       munmap(g_pfe->cbus_baseaddr, g_pfe->cbus_size);
+       g_pfe->nb_devs--;
+
+       if (g_pfe->nb_devs == 0) {
+               pfe_hif_exit(g_pfe);
+               pfe_hif_lib_exit(g_pfe);
+               rte_free(g_pfe);
+               g_pfe = NULL;
+       }
+
+       return ret;
+}
+
+static int
+pfe_eth_configure(struct rte_eth_dev *dev __rte_unused)
+{
+       return 0;
+}
+
+static int
+pfe_eth_info(struct rte_eth_dev *dev,
+               struct rte_eth_dev_info *dev_info)
+{
+       dev_info->max_mac_addrs = PFE_MAX_MACS;
+       dev_info->max_rx_queues = dev->data->nb_rx_queues;
+       dev_info->max_tx_queues = dev->data->nb_tx_queues;
+       dev_info->min_rx_bufsize = HIF_RX_PKT_MIN_SIZE;
+       dev_info->min_mtu = RTE_ETHER_MIN_MTU;
+       dev_info->rx_offload_capa = dev_rx_offloads_sup;
+       dev_info->tx_offload_capa = dev_tx_offloads_sup;
+       if (pfe_svr == SVR_LS1012A_REV1) {
+               dev_info->max_rx_pktlen = MAX_MTU_ON_REV1 + PFE_ETH_OVERHEAD;
+               dev_info->max_mtu = MAX_MTU_ON_REV1;
+       } else {
+               dev_info->max_rx_pktlen = JUMBO_FRAME_SIZE;
+               dev_info->max_mtu = JUMBO_FRAME_SIZE - PFE_ETH_OVERHEAD;
+       }
+
+       return 0;
+}
+
+/* Only first mb_pool given on first call of this API will be used
+ * in whole system, also nb_rx_desc and rx_conf are unused params
+ */
+static int
+pfe_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+               __rte_unused uint16_t nb_rx_desc,
+               __rte_unused unsigned int socket_id,
+               __rte_unused const struct rte_eth_rxconf *rx_conf,
+               struct rte_mempool *mb_pool)
+{
+       int rc = 0;
+       struct pfe *pfe;
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+
+       pfe = priv->pfe;
+
+       if (queue_idx >= EMAC_RXQ_CNT) {
+               PFE_PMD_ERR("Invalid queue idx = %d, Max queues = %d",
+                               queue_idx, EMAC_RXQ_CNT);
+               return -1;
+       }
+
+       if (!pfe->hif.setuped) {
+               rc = pfe_hif_shm_init(pfe->hif.shm, mb_pool);
+               if (rc) {
+                       PFE_PMD_ERR("Could not allocate buffer descriptors");
+                       return -1;
+               }
+
+               pfe->hif.shm->pool = mb_pool;
+               if (pfe_hif_init_buffers(&pfe->hif)) {
+                       PFE_PMD_ERR("Could not initialize buffer descriptors");
+                       return -1;
+               }
+               hif_init();
+               hif_rx_enable();
+               hif_tx_enable();
+               pfe->hif.setuped = 1;
+       }
+       dev->data->rx_queues[queue_idx] = &priv->client.rx_q[queue_idx];
+       priv->client.rx_q[queue_idx].queue_id = queue_idx;
+
+       return 0;
+}
+
+static int
+pfe_tx_queue_setup(struct rte_eth_dev *dev,
+                  uint16_t queue_idx,
+                  __rte_unused uint16_t nb_desc,
+                  __rte_unused unsigned int socket_id,
+                  __rte_unused const struct rte_eth_txconf *tx_conf)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+
+       if (queue_idx >= emac_txq_cnt) {
+               PFE_PMD_ERR("Invalid queue idx = %d, Max queues = %d",
+                               queue_idx, emac_txq_cnt);
+               return -1;
+       }
+       dev->data->tx_queues[queue_idx] = &priv->client.tx_q[queue_idx];
+       priv->client.tx_q[queue_idx].queue_id = queue_idx;
+       return 0;
+}
+
+static const uint32_t *
+pfe_supported_ptypes_get(struct rte_eth_dev *dev)
+{
+       static const uint32_t ptypes[] = {
+               /*todo -= add more types */
+               RTE_PTYPE_L2_ETHER,
+               RTE_PTYPE_L3_IPV4,
+               RTE_PTYPE_L3_IPV4_EXT,
+               RTE_PTYPE_L3_IPV6,
+               RTE_PTYPE_L3_IPV6_EXT,
+               RTE_PTYPE_L4_TCP,
+               RTE_PTYPE_L4_UDP,
+               RTE_PTYPE_L4_SCTP
+       };
+
+       if (dev->rx_pkt_burst == pfe_recv_pkts ||
+                       dev->rx_pkt_burst == pfe_recv_pkts_on_intr)
+               return ptypes;
+       return NULL;
+}
+
+static inline int
+pfe_eth_atomic_read_link_status(struct rte_eth_dev *dev,
+                               struct rte_eth_link *link)
+{
+       struct rte_eth_link *dst = link;
+       struct rte_eth_link *src = &dev->data->dev_link;
+
+       if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
+                               *(uint64_t *)src) == 0)
+               return -1;
+
+       return 0;
+}
+
+static inline int
+pfe_eth_atomic_write_link_status(struct rte_eth_dev *dev,
+                                struct rte_eth_link *link)
+{
+       struct rte_eth_link *dst = &dev->data->dev_link;
+       struct rte_eth_link *src = link;
+
+       if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
+                               *(uint64_t *)src) == 0)
+               return -1;
+
+       return 0;
+}
+
+static int
+pfe_eth_link_update(struct rte_eth_dev *dev, int wait_to_complete __rte_unused)
+{
+       int ret, ioctl_cmd = 0;
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+       struct rte_eth_link link, old;
+       unsigned int lstatus = 1;
+
+       memset(&old, 0, sizeof(old));
+       memset(&link, 0, sizeof(struct rte_eth_link));
+
+       pfe_eth_atomic_read_link_status(dev, &old);
+
+       /* Read from PFE CDEV, status of link, if file was successfully
+        * opened.
+        */
+       if (priv->link_fd != PFE_CDEV_INVALID_FD) {
+               if (priv->id == 0)
+                       ioctl_cmd = PFE_CDEV_ETH0_STATE_GET;
+               if (priv->id == 1)
+                       ioctl_cmd = PFE_CDEV_ETH1_STATE_GET;
+
+               ret = ioctl(priv->link_fd, ioctl_cmd, &lstatus);
+               if (ret != 0) {
+                       PFE_PMD_ERR("Unable to fetch link status (ioctl)\n");
+                       /* use dummy link value */
+                       link.link_status = 1;
+               }
+               PFE_PMD_DEBUG("Fetched link state (%d) for dev %d.\n",
+                             lstatus, priv->id);
+       }
+
+       if (old.link_status == lstatus) {
+               /* no change in status */
+               PFE_PMD_DEBUG("No change in link status; Not updating.\n");
+               return -1;
+       }
+
+       link.link_status = lstatus;
+       link.link_speed = RTE_ETH_LINK_SPEED_1G;
+       link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
+       link.link_autoneg = RTE_ETH_LINK_AUTONEG;
+
+       pfe_eth_atomic_write_link_status(dev, &link);
+
+       PFE_PMD_INFO("Port (%d) link is %s\n", dev->data->port_id,
+                    link.link_status ? "up" : "down");
+
+       return 0;
 }
 
+static int
+pfe_promiscuous_enable(struct rte_eth_dev *dev)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+
+       priv->promisc = 1;
+       dev->data->promiscuous = 1;
+       gemac_enable_copy_all(priv->EMAC_baseaddr);
+
+       return 0;
+}
+
+static int
+pfe_promiscuous_disable(struct rte_eth_dev *dev)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+
+       priv->promisc = 0;
+       dev->data->promiscuous = 0;
+       gemac_disable_copy_all(priv->EMAC_baseaddr);
+
+       return 0;
+}
+
+static int
+pfe_allmulticast_enable(struct rte_eth_dev *dev)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+       struct pfe_mac_addr    hash_addr; /* hash register structure */
+
+       /* Set the hash to rx all multicast frames */
+       hash_addr.bottom = 0xFFFFFFFF;
+       hash_addr.top = 0xFFFFFFFF;
+       gemac_set_hash(priv->EMAC_baseaddr, &hash_addr);
+       dev->data->all_multicast = 1;
+
+       return 0;
+}
+
+static int
+pfe_link_down(struct rte_eth_dev *dev)
+{
+       return pfe_eth_stop(dev);
+}
+
+static int
+pfe_link_up(struct rte_eth_dev *dev)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+
+       pfe_eth_start(priv);
+       return 0;
+}
+
+static int
+pfe_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+       uint16_t frame_size = mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN;
+
+       /*TODO Support VLAN*/
+       return gemac_set_rx(priv->EMAC_baseaddr, frame_size);
+}
+
+/* pfe_eth_enet_addr_byte_mac
+ */
+static int
+pfe_eth_enet_addr_byte_mac(u8 *enet_byte_addr,
+                          struct pfe_mac_addr *enet_addr)
+{
+       if (!enet_byte_addr || !enet_addr) {
+               return -1;
+
+       } else {
+               enet_addr->bottom = enet_byte_addr[0] |
+                       (enet_byte_addr[1] << 8) |
+                       (enet_byte_addr[2] << 16) |
+                       (enet_byte_addr[3] << 24);
+               enet_addr->top = enet_byte_addr[4] |
+                       (enet_byte_addr[5] << 8);
+               return 0;
+       }
+}
+
+static int
+pfe_dev_set_mac_addr(struct rte_eth_dev *dev,
+                      struct rte_ether_addr *addr)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+       struct pfe_mac_addr spec_addr;
+       int ret;
+
+       ret = pfe_eth_enet_addr_byte_mac(addr->addr_bytes, &spec_addr);
+       if (ret)
+               return ret;
+
+       gemac_set_laddrN(priv->EMAC_baseaddr,
+                        (struct pfe_mac_addr *)&spec_addr, 1);
+       rte_ether_addr_copy(addr, &dev->data->mac_addrs[0]);
+       return 0;
+}
+
+static int
+pfe_stats_get(struct rte_eth_dev *dev,
+             struct rte_eth_stats *stats)
+{
+       struct pfe_eth_priv_s *priv = dev->data->dev_private;
+       struct rte_eth_stats *eth_stats = &priv->stats;
+
+       if (stats == NULL)
+               return -1;
+
+       memset(stats, 0, sizeof(struct rte_eth_stats));
+
+       stats->ipackets = eth_stats->ipackets;
+       stats->ibytes = eth_stats->ibytes;
+       stats->opackets = eth_stats->opackets;
+       stats->obytes = eth_stats->obytes;
+
+       return 0;
+}
+
+static const struct eth_dev_ops ops = {
+       .dev_start = pfe_eth_open,
+       .dev_stop = pfe_eth_stop,
+       .dev_close = pfe_eth_close,
+       .dev_configure = pfe_eth_configure,
+       .dev_infos_get = pfe_eth_info,
+       .rx_queue_setup = pfe_rx_queue_setup,
+       .tx_queue_setup = pfe_tx_queue_setup,
+       .dev_supported_ptypes_get = pfe_supported_ptypes_get,
+       .link_update  = pfe_eth_link_update,
+       .promiscuous_enable   = pfe_promiscuous_enable,
+       .promiscuous_disable  = pfe_promiscuous_disable,
+       .allmulticast_enable  = pfe_allmulticast_enable,
+       .dev_set_link_down    = pfe_link_down,
+       .dev_set_link_up      = pfe_link_up,
+       .mtu_set              = pfe_mtu_set,
+       .mac_addr_set         = pfe_dev_set_mac_addr,
+       .stats_get            = pfe_stats_get,
+};
+
 static int
 pfe_eth_init(struct rte_vdev_device *vdev, struct pfe *pfe, int id)
 {
        struct rte_eth_dev *eth_dev = NULL;
        struct pfe_eth_priv_s *priv = NULL;
+       struct ls1012a_eth_platform_data *einfo;
+       struct ls1012a_pfe_platform_data *pfe_info;
+       struct rte_ether_addr addr;
        int err;
 
        eth_dev = rte_eth_vdev_allocate(vdev, sizeof(*priv));
        if (eth_dev == NULL)
                return -ENOMEM;
 
+       /* Extract pltform data */
+       pfe_info = (struct ls1012a_pfe_platform_data *)&pfe->platform_data;
+       if (!pfe_info) {
+               PFE_PMD_ERR("pfe missing additional platform data");
+               err = -ENODEV;
+               goto err0;
+       }
+
+       einfo = (struct ls1012a_eth_platform_data *)pfe_info->ls1012a_eth_pdata;
+
+       /* einfo never be NULL, but no harm in having this check */
+       if (!einfo) {
+               PFE_PMD_ERR("pfe missing additional gemacs platform data");
+               err = -ENODEV;
+               goto err0;
+       }
+
        priv = eth_dev->data->dev_private;
        priv->ndev = eth_dev;
+       priv->id = einfo[id].gem_id;
        priv->pfe = pfe;
 
        pfe->eth.eth_priv[id] = priv;
 
+       /* Set the info in the priv to the current info */
+       priv->einfo = &einfo[id];
+       priv->EMAC_baseaddr = cbus_emac_base[id];
+       priv->PHY_baseaddr = cbus_emac_base[id];
+       priv->GPI_baseaddr = cbus_gpi_base[id];
+
 #define HIF_GEMAC_TMUQ_BASE    6
        priv->low_tmu_q = HIF_GEMAC_TMUQ_BASE + (id * 2);
        priv->high_tmu_q = priv->low_tmu_q + 1;
@@ -110,11 +809,24 @@ pfe_eth_init(struct rte_vdev_device *vdev, struct pfe *pfe, int id)
        eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
                        ETHER_ADDR_LEN * PFE_MAX_MACS, 0);
        if (eth_dev->data->mac_addrs == NULL) {
+               PFE_PMD_ERR("Failed to allocate mem %d to store MAC addresses",
+                       ETHER_ADDR_LEN * PFE_MAX_MACS);
                err = -ENOMEM;
                goto err0;
        }
 
+       memcpy(addr.addr_bytes, priv->einfo->mac_addr,
+                      ETH_ALEN);
+
+       pfe_dev_set_mac_addr(eth_dev, &addr);
+       rte_ether_addr_copy(&addr, &eth_dev->data->mac_addrs[0]);
+
        eth_dev->data->mtu = 1500;
+       eth_dev->dev_ops = &ops;
+       err = pfe_eth_stop(eth_dev);
+       if (err != 0)
+               goto err0;
+       pfe_gemac_init(priv);
 
        eth_dev->data->nb_rx_queues = 1;
        eth_dev->data->nb_tx_queues = 1;
@@ -132,6 +844,58 @@ err0:
        return err;
 }
 
+static int
+pfe_get_gemac_if_proprties(struct pfe *pfe,
+               __rte_unused const struct device_node *parent,
+               unsigned int port, unsigned int if_cnt,
+               struct ls1012a_pfe_platform_data *pdata)
+{
+       const struct device_node *gem = NULL;
+       size_t size;
+       unsigned int ii = 0, phy_id = 0;
+       const u32 *addr;
+       const void *mac_addr;
+
+       for (ii = 0; ii < if_cnt; ii++) {
+               gem = of_get_next_child(parent, gem);
+               if (!gem)
+                       goto err;
+               addr = of_get_property(gem, "reg", &size);
+               if (addr && (rte_be_to_cpu_32((unsigned int)*addr) == port))
+                       break;
+       }
+
+       if (ii >= if_cnt) {
+               PFE_PMD_ERR("Failed to find interface = %d", if_cnt);
+               goto err;
+       }
+
+       pdata->ls1012a_eth_pdata[port].gem_id = port;
+
+       mac_addr = of_get_mac_address(gem);
+
+       if (mac_addr) {
+               memcpy(pdata->ls1012a_eth_pdata[port].mac_addr, mac_addr,
+                      ETH_ALEN);
+       }
+
+       addr = of_get_property(gem, "fsl,mdio-mux-val", &size);
+       if (!addr) {
+               PFE_PMD_ERR("Invalid mdio-mux-val....");
+       } else {
+               phy_id = rte_be_to_cpu_32((unsigned int)*addr);
+               pdata->ls1012a_eth_pdata[port].mdio_muxval = phy_id;
+       }
+       if (pdata->ls1012a_eth_pdata[port].phy_id < 32)
+               pfe->mdio_muxval[pdata->ls1012a_eth_pdata[port].phy_id] =
+                        pdata->ls1012a_eth_pdata[port].mdio_muxval;
+
+       return 0;
+
+err:
+       return -1;
+}
+
 /* Parse integer from integer argument */
 static int
 parse_integer_arg(const char *key __rte_unused,
@@ -142,8 +906,10 @@ parse_integer_arg(const char *key __rte_unused,
        errno = 0;
 
        i = strtol(value, &end, 10);
-       if (*end != 0 || errno != 0 || i < 0 || i > 1)
+       if (*end != 0 || errno != 0 || i < 0 || i > 1) {
+               PFE_PMD_ERR("Supported Port IDS are 0 and 1");
                return -EINVAL;
+       }
 
        *((uint32_t *)extra_args) = i;
 
@@ -188,7 +954,7 @@ pmd_pfe_probe(struct rte_vdev_device *vdev)
        const uint32_t *addr;
        uint64_t cbus_addr, ddr_size, cbus_size;
        int rc = -1, fd = -1, gem_id;
-       unsigned int interface_count = 0;
+       unsigned int ii, interface_count = 0;
        size_t size = 0;
        struct pfe_vdev_init_params init_params = {
                .gem_id = -1
@@ -199,10 +965,15 @@ pmd_pfe_probe(struct rte_vdev_device *vdev)
        if (rc < 0)
                return -EINVAL;
 
+       PFE_PMD_LOG(INFO, "Initializing pmd_pfe for %s Given gem-id %d",
+               name, init_params.gem_id);
+
        if (g_pfe) {
-               if (g_pfe->nb_devs >= g_pfe->max_intf)
+               if (g_pfe->nb_devs >= g_pfe->max_intf) {
+                       PFE_PMD_ERR("PFE %d dev already created Max is %d",
+                               g_pfe->nb_devs, g_pfe->max_intf);
                        return -EINVAL;
-
+               }
                goto eth_init;
        }
 
@@ -212,32 +983,42 @@ pmd_pfe_probe(struct rte_vdev_device *vdev)
 
        /* Load the device-tree driver */
        rc = of_init();
-       if (rc)
+       if (rc) {
+               PFE_PMD_ERR("of_init failed with ret: %d", rc);
                goto err;
+       }
 
        np = of_find_compatible_node(NULL, NULL, "fsl,pfe");
        if (!np) {
+               PFE_PMD_ERR("Invalid device node");
                rc = -EINVAL;
                goto err;
        }
 
        addr = of_get_address(np, 0, &cbus_size, NULL);
-       if (!addr)
+       if (!addr) {
+               PFE_PMD_ERR("of_get_address cannot return qman address\n");
                goto err;
-
+       }
        cbus_addr = of_translate_address(np, addr);
-       if (!cbus_addr)
+       if (!cbus_addr) {
+               PFE_PMD_ERR("of_translate_address failed\n");
                goto err;
-
+       }
 
        addr = of_get_address(np, 1, &ddr_size, NULL);
-       if (!addr)
+       if (!addr) {
+               PFE_PMD_ERR("of_get_address cannot return qman address\n");
                goto err;
+       }
 
        g_pfe->ddr_phys_baseaddr = of_translate_address(np, addr);
-       if (!g_pfe->ddr_phys_baseaddr)
+       if (!g_pfe->ddr_phys_baseaddr) {
+               PFE_PMD_ERR("of_translate_address failed\n");
                goto err;
+       }
 
+       g_pfe->ddr_baseaddr = pfe_mem_ptov(g_pfe->ddr_phys_baseaddr);
        g_pfe->ddr_size = ddr_size;
        g_pfe->cbus_size = cbus_size;
 
@@ -246,6 +1027,7 @@ pmd_pfe_probe(struct rte_vdev_device *vdev)
                                        MAP_SHARED, fd, cbus_addr);
        close(fd);
        if (g_pfe->cbus_baseaddr == MAP_FAILED) {
+               PFE_PMD_ERR("Can not map cbus base");
                rc = -EINVAL;
                goto err;
        }
@@ -253,16 +1035,57 @@ pmd_pfe_probe(struct rte_vdev_device *vdev)
        /* Read interface count */
        prop = of_get_property(np, "fsl,pfe-num-interfaces", &size);
        if (!prop) {
+               PFE_PMD_ERR("Failed to read number of interfaces");
                rc = -ENXIO;
                goto err_prop;
        }
 
        interface_count = rte_be_to_cpu_32((unsigned int)*prop);
        if (interface_count <= 0) {
+               PFE_PMD_ERR("No ethernet interface count : %d",
+                               interface_count);
                rc = -ENXIO;
                goto err_prop;
        }
+       PFE_PMD_INFO("num interfaces = %d ", interface_count);
+
        g_pfe->max_intf  = interface_count;
+       g_pfe->platform_data.ls1012a_mdio_pdata[0].phy_mask = 0xffffffff;
+
+       for (ii = 0; ii < interface_count; ii++) {
+               pfe_get_gemac_if_proprties(g_pfe, np, ii, interface_count,
+                                          &g_pfe->platform_data);
+       }
+
+       pfe_lib_init(g_pfe->cbus_baseaddr, g_pfe->ddr_baseaddr,
+                    g_pfe->ddr_phys_baseaddr, g_pfe->ddr_size);
+
+       PFE_PMD_INFO("CLASS version: %x", readl(CLASS_VERSION));
+       PFE_PMD_INFO("TMU version: %x", readl(TMU_VERSION));
+
+       PFE_PMD_INFO("BMU1 version: %x", readl(BMU1_BASE_ADDR + BMU_VERSION));
+       PFE_PMD_INFO("BMU2 version: %x", readl(BMU2_BASE_ADDR + BMU_VERSION));
+
+       PFE_PMD_INFO("EGPI1 version: %x", readl(EGPI1_BASE_ADDR + GPI_VERSION));
+       PFE_PMD_INFO("EGPI2 version: %x", readl(EGPI2_BASE_ADDR + GPI_VERSION));
+       PFE_PMD_INFO("HGPI version: %x", readl(HGPI_BASE_ADDR + GPI_VERSION));
+
+       PFE_PMD_INFO("HIF version: %x", readl(HIF_VERSION));
+       PFE_PMD_INFO("HIF NOPCY version: %x", readl(HIF_NOCPY_VERSION));
+
+       cbus_emac_base[0] = EMAC1_BASE_ADDR;
+       cbus_emac_base[1] = EMAC2_BASE_ADDR;
+
+       cbus_gpi_base[0] = EGPI1_BASE_ADDR;
+       cbus_gpi_base[1] = EGPI2_BASE_ADDR;
+
+       rc = pfe_hif_lib_init(g_pfe);
+       if (rc < 0)
+               goto err_hif_lib;
+
+       rc = pfe_hif_init(g_pfe);
+       if (rc < 0)
+               goto err_hif;
        pfe_soc_version_get();
 eth_init:
        if (init_params.gem_id < 0)
@@ -270,7 +1093,7 @@ eth_init:
        else
                gem_id = init_params.gem_id;
 
-       RTE_LOG(INFO, PMD, "Init pmd_pfe for %s gem-id %d(given =%d)\n",
+       PFE_PMD_LOG(INFO, "Init pmd_pfe for %s gem-id %d(given =%d)",
                name, gem_id, init_params.gem_id);
 
        rc = pfe_eth_init(vdev, g_pfe, gem_id);
@@ -282,6 +1105,12 @@ eth_init:
        return 0;
 
 err_eth:
+       pfe_hif_exit(g_pfe);
+
+err_hif:
+       pfe_hif_lib_exit(g_pfe);
+
+err_hif_lib:
 err_prop:
        munmap(g_pfe->cbus_baseaddr, cbus_size);
 err:
@@ -294,22 +1123,24 @@ pmd_pfe_remove(struct rte_vdev_device *vdev)
 {
        const char *name;
        struct rte_eth_dev *eth_dev = NULL;
+       int ret = 0;
 
        name = rte_vdev_device_name(vdev);
        if (name == NULL)
                return -EINVAL;
 
+       PFE_PMD_INFO("Closing eventdev sw device %s", name);
+
        if (!g_pfe)
                return 0;
 
        eth_dev = rte_eth_dev_allocated(name);
-       if (eth_dev == NULL)
-               return -ENODEV;
-
-       pfe_eth_exit(eth_dev, g_pfe);
-       munmap(g_pfe->cbus_baseaddr, g_pfe->cbus_size);
+       if (eth_dev) {
+               pfe_eth_close(eth_dev);
+               ret = rte_eth_dev_release_port(eth_dev);
+       }
 
-       return 0;
+       return ret;
 }
 
 static
@@ -320,3 +1151,4 @@ struct rte_vdev_driver pmd_pfe_drv = {
 
 RTE_PMD_REGISTER_VDEV(PFE_NAME_PMD, pmd_pfe_drv);
 RTE_PMD_REGISTER_PARAM_STRING(PFE_NAME_PMD, PFE_VDEV_GEM_ID_ARG "=<int> ");
+RTE_LOG_REGISTER_DEFAULT(pfe_logtype_pmd, NOTICE);