staging: fsl-dpaa2/ethsw: Add Freescale DPAA2 Ethernet Switch driver
authorRazvan Stefanescu <razvan.stefanescu@nxp.com>
Wed, 14 Mar 2018 15:55:54 +0000 (10:55 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 14 Mar 2018 16:36:09 +0000 (17:36 +0100)
Introduce the DPAA2 Ethernet Switch driver, which manages Datapath Switch
(DPSW) objects discovered on the MC bus.

Suggested-by: Alexandru Marginean <alexandru.marginean@nxp.com>
Signed-off-by: Razvan Stefanescu <razvan.stefanescu@nxp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/fsl-dpaa2/ethsw/Makefile
drivers/staging/fsl-dpaa2/ethsw/ethsw.c [new file with mode: 0644]
drivers/staging/fsl-dpaa2/ethsw/ethsw.h [new file with mode: 0644]

index 9846e61c4ffd67809b70cbeddb0244dd524e920e..7755603b1462e8b46b07d002dd32875836d9430c 100644 (file)
@@ -7,4 +7,4 @@
 
 obj-$(CONFIG_FSL_DPAA2_ETHSW) += dpaa2-ethsw.o
 
-dpaa2-ethsw-objs := dpsw.o
+dpaa2-ethsw-objs := ethsw.o dpsw.o
diff --git a/drivers/staging/fsl-dpaa2/ethsw/ethsw.c b/drivers/staging/fsl-dpaa2/ethsw/ethsw.c
new file mode 100644 (file)
index 0000000..5aa6e95
--- /dev/null
@@ -0,0 +1,1507 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * DPAA2 Ethernet Switch driver
+ *
+ * Copyright 2014-2016 Freescale Semiconductor Inc.
+ * Copyright 2017-2018 NXP
+ *
+ */
+
+#include <linux/module.h>
+
+#include <linux/interrupt.h>
+#include <linux/msi.h>
+#include <linux/kthread.h>
+#include <linux/workqueue.h>
+
+#include <linux/fsl/mc.h>
+
+#include "ethsw.h"
+
+static struct workqueue_struct *ethsw_owq;
+
+/* Minimal supported DPSW version */
+#define DPSW_MIN_VER_MAJOR             8
+#define DPSW_MIN_VER_MINOR             0
+
+#define DEFAULT_VLAN_ID                        1
+
+static int ethsw_add_vlan(struct ethsw_core *ethsw, u16 vid)
+{
+       int err;
+
+       struct dpsw_vlan_cfg    vcfg = {
+               .fdb_id = 0,
+       };
+
+       if (ethsw->vlans[vid]) {
+               dev_err(ethsw->dev, "VLAN already configured\n");
+               return -EEXIST;
+       }
+
+       err = dpsw_vlan_add(ethsw->mc_io, 0,
+                           ethsw->dpsw_handle, vid, &vcfg);
+       if (err) {
+               dev_err(ethsw->dev, "dpsw_vlan_add err %d\n", err);
+               return err;
+       }
+       ethsw->vlans[vid] = ETHSW_VLAN_MEMBER;
+
+       return 0;
+}
+
+static int ethsw_port_set_tci(struct ethsw_port_priv *port_priv,
+                             struct dpsw_tci_cfg *tci_cfg)
+{
+       struct ethsw_core *ethsw = port_priv->ethsw_data;
+       struct net_device *netdev = port_priv->netdev;
+       bool is_oper;
+       int err, ret;
+
+       /* Interface needs to be down to change PVID */
+       is_oper = netif_oper_up(netdev);
+       if (is_oper) {
+               err = dpsw_if_disable(ethsw->mc_io, 0,
+                                     ethsw->dpsw_handle,
+                                     port_priv->idx);
+               if (err) {
+                       netdev_err(netdev, "dpsw_if_disable err %d\n", err);
+                       return err;
+               }
+       }
+
+       err = dpsw_if_set_tci(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                             port_priv->idx, tci_cfg);
+       if (err) {
+               netdev_err(netdev, "dpsw_if_set_tci err %d\n", err);
+               goto set_tci_error;
+       }
+
+       /* Delete previous PVID info and mark the new one */
+       if (port_priv->pvid)
+               port_priv->vlans[port_priv->pvid] &= ~ETHSW_VLAN_PVID;
+       port_priv->vlans[tci_cfg->vlan_id] |= ETHSW_VLAN_PVID;
+       port_priv->pvid = tci_cfg->vlan_id;
+
+set_tci_error:
+       if (is_oper) {
+               ret = dpsw_if_enable(ethsw->mc_io, 0,
+                                    ethsw->dpsw_handle,
+                                    port_priv->idx);
+               if (ret) {
+                       netdev_err(netdev, "dpsw_if_enable err %d\n", ret);
+                       return ret;
+               }
+       }
+
+       return err;
+}
+
+static int ethsw_port_add_vlan(struct ethsw_port_priv *port_priv,
+                              u16 vid, u16 flags)
+{
+       struct ethsw_core *ethsw = port_priv->ethsw_data;
+       struct net_device *netdev = port_priv->netdev;
+       struct dpsw_vlan_if_cfg vcfg;
+       int err;
+
+       if (port_priv->vlans[vid]) {
+               netdev_warn(netdev, "VLAN %d already configured\n", vid);
+               return -EEXIST;
+       }
+
+       vcfg.num_ifs = 1;
+       vcfg.if_id[0] = port_priv->idx;
+       err = dpsw_vlan_add_if(ethsw->mc_io, 0, ethsw->dpsw_handle, vid, &vcfg);
+       if (err) {
+               netdev_err(netdev, "dpsw_vlan_add_if err %d\n", err);
+               return err;
+       }
+
+       port_priv->vlans[vid] = ETHSW_VLAN_MEMBER;
+
+       if (flags & BRIDGE_VLAN_INFO_UNTAGGED) {
+               err = dpsw_vlan_add_if_untagged(ethsw->mc_io, 0,
+                                               ethsw->dpsw_handle,
+                                               vid, &vcfg);
+               if (err) {
+                       netdev_err(netdev,
+                                  "dpsw_vlan_add_if_untagged err %d\n", err);
+                       return err;
+               }
+               port_priv->vlans[vid] |= ETHSW_VLAN_UNTAGGED;
+       }
+
+       if (flags & BRIDGE_VLAN_INFO_PVID) {
+               struct dpsw_tci_cfg tci_cfg = {
+                       .pcp = 0,
+                       .dei = 0,
+                       .vlan_id = vid,
+               };
+
+               err = ethsw_port_set_tci(port_priv, &tci_cfg);
+               if (err)
+                       return err;
+       }
+
+       return 0;
+}
+
+static int ethsw_set_learning(struct ethsw_core *ethsw, u8 flag)
+{
+       enum dpsw_fdb_learning_mode learn_mode;
+       int err;
+
+       if (flag)
+               learn_mode = DPSW_FDB_LEARNING_MODE_HW;
+       else
+               learn_mode = DPSW_FDB_LEARNING_MODE_DIS;
+
+       err = dpsw_fdb_set_learning_mode(ethsw->mc_io, 0, ethsw->dpsw_handle, 0,
+                                        learn_mode);
+       if (err) {
+               dev_err(ethsw->dev, "dpsw_fdb_set_learning_mode err %d\n", err);
+               return err;
+       }
+       ethsw->learning = !!flag;
+
+       return 0;
+}
+
+static int ethsw_port_set_flood(struct ethsw_port_priv *port_priv, u8 flag)
+{
+       int err;
+
+       err = dpsw_if_set_flooding(port_priv->ethsw_data->mc_io, 0,
+                                  port_priv->ethsw_data->dpsw_handle,
+                                  port_priv->idx, flag);
+       if (err) {
+               netdev_err(port_priv->netdev,
+                          "dpsw_fdb_set_learning_mode err %d\n", err);
+               return err;
+       }
+       port_priv->flood = !!flag;
+
+       return 0;
+}
+
+static int ethsw_port_set_stp_state(struct ethsw_port_priv *port_priv, u8 state)
+{
+       struct dpsw_stp_cfg stp_cfg = {
+               .vlan_id = DEFAULT_VLAN_ID,
+               .state = state,
+       };
+       int err;
+
+       if (!netif_oper_up(port_priv->netdev) || state == port_priv->stp_state)
+               return 0;       /* Nothing to do */
+
+       err = dpsw_if_set_stp(port_priv->ethsw_data->mc_io, 0,
+                             port_priv->ethsw_data->dpsw_handle,
+                             port_priv->idx, &stp_cfg);
+       if (err) {
+               netdev_err(port_priv->netdev,
+                          "dpsw_if_set_stp err %d\n", err);
+               return err;
+       }
+
+       port_priv->stp_state = state;
+
+       return 0;
+}
+
+static int ethsw_dellink_switch(struct ethsw_core *ethsw, u16 vid)
+{
+       struct ethsw_port_priv *ppriv_local = NULL;
+       int i, err;
+
+       if (!ethsw->vlans[vid])
+               return -ENOENT;
+
+       err = dpsw_vlan_remove(ethsw->mc_io, 0, ethsw->dpsw_handle, vid);
+       if (err) {
+               dev_err(ethsw->dev, "dpsw_vlan_remove err %d\n", err);
+               return err;
+       }
+       ethsw->vlans[vid] = 0;
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++) {
+               ppriv_local = ethsw->ports[i];
+               ppriv_local->vlans[vid] = 0;
+       }
+
+       return 0;
+}
+
+static int ethsw_port_fdb_add_uc(struct ethsw_port_priv *port_priv,
+                                const unsigned char *addr)
+{
+       struct dpsw_fdb_unicast_cfg entry = {0};
+       int err;
+
+       entry.if_egress = port_priv->idx;
+       entry.type = DPSW_FDB_ENTRY_STATIC;
+       ether_addr_copy(entry.mac_addr, addr);
+
+       err = dpsw_fdb_add_unicast(port_priv->ethsw_data->mc_io, 0,
+                                  port_priv->ethsw_data->dpsw_handle,
+                                  0, &entry);
+       if (err)
+               netdev_err(port_priv->netdev,
+                          "dpsw_fdb_add_unicast err %d\n", err);
+       return err;
+}
+
+static int ethsw_port_fdb_del_uc(struct ethsw_port_priv *port_priv,
+                                const unsigned char *addr)
+{
+       struct dpsw_fdb_unicast_cfg entry = {0};
+       int err;
+
+       entry.if_egress = port_priv->idx;
+       entry.type = DPSW_FDB_ENTRY_STATIC;
+       ether_addr_copy(entry.mac_addr, addr);
+
+       err = dpsw_fdb_remove_unicast(port_priv->ethsw_data->mc_io, 0,
+                                     port_priv->ethsw_data->dpsw_handle,
+                                     0, &entry);
+       /* Silently discard error for calling multiple times the del command */
+       if (err && err != -ENXIO)
+               netdev_err(port_priv->netdev,
+                          "dpsw_fdb_remove_unicast err %d\n", err);
+       return err;
+}
+
+static int ethsw_port_fdb_add_mc(struct ethsw_port_priv *port_priv,
+                                const unsigned char *addr)
+{
+       struct dpsw_fdb_multicast_cfg entry = {0};
+       int err;
+
+       ether_addr_copy(entry.mac_addr, addr);
+       entry.type = DPSW_FDB_ENTRY_STATIC;
+       entry.num_ifs = 1;
+       entry.if_id[0] = port_priv->idx;
+
+       err = dpsw_fdb_add_multicast(port_priv->ethsw_data->mc_io, 0,
+                                    port_priv->ethsw_data->dpsw_handle,
+                                    0, &entry);
+       /* Silently discard error for calling multiple times the add command */
+       if (err && err != -ENXIO)
+               netdev_err(port_priv->netdev, "dpsw_fdb_add_multicast err %d\n",
+                          err);
+       return err;
+}
+
+static int ethsw_port_fdb_del_mc(struct ethsw_port_priv *port_priv,
+                                const unsigned char *addr)
+{
+       struct dpsw_fdb_multicast_cfg entry = {0};
+       int err;
+
+       ether_addr_copy(entry.mac_addr, addr);
+       entry.type = DPSW_FDB_ENTRY_STATIC;
+       entry.num_ifs = 1;
+       entry.if_id[0] = port_priv->idx;
+
+       err = dpsw_fdb_remove_multicast(port_priv->ethsw_data->mc_io, 0,
+                                       port_priv->ethsw_data->dpsw_handle,
+                                       0, &entry);
+       /* Silently discard error for calling multiple times the del command */
+       if (err && err != -ENAVAIL)
+               netdev_err(port_priv->netdev,
+                          "dpsw_fdb_remove_multicast err %d\n", err);
+       return err;
+}
+
+static void port_get_stats(struct net_device *netdev,
+                          struct rtnl_link_stats64 *stats)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       u64 tmp;
+       int err;
+
+       err = dpsw_if_get_counter(port_priv->ethsw_data->mc_io, 0,
+                                 port_priv->ethsw_data->dpsw_handle,
+                                 port_priv->idx,
+                                 DPSW_CNT_ING_FRAME, &stats->rx_packets);
+       if (err)
+               goto error;
+
+       err = dpsw_if_get_counter(port_priv->ethsw_data->mc_io, 0,
+                                 port_priv->ethsw_data->dpsw_handle,
+                                 port_priv->idx,
+                                 DPSW_CNT_EGR_FRAME, &stats->tx_packets);
+       if (err)
+               goto error;
+
+       err = dpsw_if_get_counter(port_priv->ethsw_data->mc_io, 0,
+                                 port_priv->ethsw_data->dpsw_handle,
+                                 port_priv->idx,
+                                 DPSW_CNT_ING_BYTE, &stats->rx_bytes);
+       if (err)
+               goto error;
+
+       err = dpsw_if_get_counter(port_priv->ethsw_data->mc_io, 0,
+                                 port_priv->ethsw_data->dpsw_handle,
+                                 port_priv->idx,
+                                 DPSW_CNT_EGR_BYTE, &stats->tx_bytes);
+       if (err)
+               goto error;
+
+       err = dpsw_if_get_counter(port_priv->ethsw_data->mc_io, 0,
+                                 port_priv->ethsw_data->dpsw_handle,
+                                 port_priv->idx,
+                                 DPSW_CNT_ING_FRAME_DISCARD,
+                                 &stats->rx_dropped);
+       if (err)
+               goto error;
+
+       err = dpsw_if_get_counter(port_priv->ethsw_data->mc_io, 0,
+                                 port_priv->ethsw_data->dpsw_handle,
+                                 port_priv->idx,
+                                 DPSW_CNT_ING_FLTR_FRAME,
+                                 &tmp);
+       if (err)
+               goto error;
+       stats->rx_dropped += tmp;
+
+       err = dpsw_if_get_counter(port_priv->ethsw_data->mc_io, 0,
+                                 port_priv->ethsw_data->dpsw_handle,
+                                 port_priv->idx,
+                                 DPSW_CNT_EGR_FRAME_DISCARD,
+                                 &stats->tx_dropped);
+       if (err)
+               goto error;
+
+       return;
+
+error:
+       netdev_err(netdev, "dpsw_if_get_counter err %d\n", err);
+}
+
+static bool port_has_offload_stats(const struct net_device *netdev,
+                                  int attr_id)
+{
+       return (attr_id == IFLA_OFFLOAD_XSTATS_CPU_HIT);
+}
+
+static int port_get_offload_stats(int attr_id,
+                                 const struct net_device *netdev,
+                                 void *sp)
+{
+       switch (attr_id) {
+       case IFLA_OFFLOAD_XSTATS_CPU_HIT:
+               port_get_stats((struct net_device *)netdev, sp);
+               return 0;
+       }
+
+       return -EINVAL;
+}
+
+static int port_change_mtu(struct net_device *netdev, int mtu)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int err;
+
+       err = dpsw_if_set_max_frame_length(port_priv->ethsw_data->mc_io,
+                                          0,
+                                          port_priv->ethsw_data->dpsw_handle,
+                                          port_priv->idx,
+                                          (u16)ETHSW_L2_MAX_FRM(mtu));
+       if (err) {
+               netdev_err(netdev,
+                          "dpsw_if_set_max_frame_length() err %d\n", err);
+               return err;
+       }
+
+       netdev->mtu = mtu;
+       return 0;
+}
+
+static int port_carrier_state_sync(struct net_device *netdev)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       struct dpsw_link_state state;
+       int err;
+
+       err = dpsw_if_get_link_state(port_priv->ethsw_data->mc_io, 0,
+                                    port_priv->ethsw_data->dpsw_handle,
+                                    port_priv->idx, &state);
+       if (err) {
+               netdev_err(netdev, "dpsw_if_get_link_state() err %d\n", err);
+               return err;
+       }
+
+       WARN_ONCE(state.up > 1, "Garbage read into link_state");
+
+       if (state.up != port_priv->link_state) {
+               if (state.up)
+                       netif_carrier_on(netdev);
+               else
+                       netif_carrier_off(netdev);
+               port_priv->link_state = state.up;
+       }
+       return 0;
+}
+
+static int port_open(struct net_device *netdev)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int err;
+
+       /* No need to allow Tx as control interface is disabled */
+       netif_tx_stop_all_queues(netdev);
+
+       err = dpsw_if_enable(port_priv->ethsw_data->mc_io, 0,
+                            port_priv->ethsw_data->dpsw_handle,
+                            port_priv->idx);
+       if (err) {
+               netdev_err(netdev, "dpsw_if_enable err %d\n", err);
+               return err;
+       }
+
+       /* sync carrier state */
+       err = port_carrier_state_sync(netdev);
+       if (err) {
+               netdev_err(netdev,
+                          "port_carrier_state_sync err %d\n", err);
+               goto err_carrier_sync;
+       }
+
+       return 0;
+
+err_carrier_sync:
+       dpsw_if_disable(port_priv->ethsw_data->mc_io, 0,
+                       port_priv->ethsw_data->dpsw_handle,
+                       port_priv->idx);
+       return err;
+}
+
+static int port_stop(struct net_device *netdev)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int err;
+
+       err = dpsw_if_disable(port_priv->ethsw_data->mc_io, 0,
+                             port_priv->ethsw_data->dpsw_handle,
+                             port_priv->idx);
+       if (err) {
+               netdev_err(netdev, "dpsw_if_disable err %d\n", err);
+               return err;
+       }
+
+       return 0;
+}
+
+static netdev_tx_t port_dropframe(struct sk_buff *skb,
+                                 struct net_device *netdev)
+{
+       /* we don't support I/O for now, drop the frame */
+       dev_kfree_skb_any(skb);
+
+       return NETDEV_TX_OK;
+}
+
+static const struct net_device_ops ethsw_port_ops = {
+       .ndo_open               = port_open,
+       .ndo_stop               = port_stop,
+
+       .ndo_set_mac_address    = eth_mac_addr,
+       .ndo_change_mtu         = port_change_mtu,
+       .ndo_has_offload_stats  = port_has_offload_stats,
+       .ndo_get_offload_stats  = port_get_offload_stats,
+
+       .ndo_start_xmit         = port_dropframe,
+};
+
+static void ethsw_links_state_update(struct ethsw_core *ethsw)
+{
+       int i;
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++)
+               port_carrier_state_sync(ethsw->ports[i]->netdev);
+}
+
+static irqreturn_t ethsw_irq0_handler_thread(int irq_num, void *arg)
+{
+       struct device *dev = (struct device *)arg;
+       struct ethsw_core *ethsw = dev_get_drvdata(dev);
+
+       /* Mask the events and the if_id reserved bits to be cleared on read */
+       u32 status = DPSW_IRQ_EVENT_LINK_CHANGED | 0xFFFF0000;
+       int err;
+
+       err = dpsw_get_irq_status(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                 DPSW_IRQ_INDEX_IF, &status);
+       if (err) {
+               dev_err(dev, "Can't get irq status (err %d)", err);
+
+               err = dpsw_clear_irq_status(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                           DPSW_IRQ_INDEX_IF, 0xFFFFFFFF);
+               if (err)
+                       dev_err(dev, "Can't clear irq status (err %d)", err);
+               goto out;
+       }
+
+       if (status & DPSW_IRQ_EVENT_LINK_CHANGED)
+               ethsw_links_state_update(ethsw);
+
+out:
+       return IRQ_HANDLED;
+}
+
+static int ethsw_setup_irqs(struct fsl_mc_device *sw_dev)
+{
+       struct device *dev = &sw_dev->dev;
+       struct ethsw_core *ethsw = dev_get_drvdata(dev);
+       u32 mask = DPSW_IRQ_EVENT_LINK_CHANGED;
+       struct fsl_mc_device_irq *irq;
+       int err;
+
+       err = fsl_mc_allocate_irqs(sw_dev);
+       if (err) {
+               dev_err(dev, "MC irqs allocation failed\n");
+               return err;
+       }
+
+       if (WARN_ON(sw_dev->obj_desc.irq_count != DPSW_IRQ_NUM)) {
+               err = -EINVAL;
+               goto free_irq;
+       }
+
+       err = dpsw_set_irq_enable(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                 DPSW_IRQ_INDEX_IF, 0);
+       if (err) {
+               dev_err(dev, "dpsw_set_irq_enable err %d\n", err);
+               goto free_irq;
+       }
+
+       irq = sw_dev->irqs[DPSW_IRQ_INDEX_IF];
+
+       err = devm_request_threaded_irq(dev, irq->msi_desc->irq,
+                                       NULL,
+                                       ethsw_irq0_handler_thread,
+                                       IRQF_NO_SUSPEND | IRQF_ONESHOT,
+                                       dev_name(dev), dev);
+       if (err) {
+               dev_err(dev, "devm_request_threaded_irq(): %d", err);
+               goto free_irq;
+       }
+
+       err = dpsw_set_irq_mask(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                               DPSW_IRQ_INDEX_IF, mask);
+       if (err) {
+               dev_err(dev, "dpsw_set_irq_mask(): %d", err);
+               goto free_devm_irq;
+       }
+
+       err = dpsw_set_irq_enable(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                 DPSW_IRQ_INDEX_IF, 1);
+       if (err) {
+               dev_err(dev, "dpsw_set_irq_enable(): %d", err);
+               goto free_devm_irq;
+       }
+
+       return 0;
+
+free_devm_irq:
+       devm_free_irq(dev, irq->msi_desc->irq, dev);
+free_irq:
+       fsl_mc_free_irqs(sw_dev);
+       return err;
+}
+
+static void ethsw_teardown_irqs(struct fsl_mc_device *sw_dev)
+{
+       struct device *dev = &sw_dev->dev;
+       struct ethsw_core *ethsw = dev_get_drvdata(dev);
+       struct fsl_mc_device_irq *irq;
+       int err;
+
+       irq = sw_dev->irqs[DPSW_IRQ_INDEX_IF];
+       err = dpsw_set_irq_enable(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                 DPSW_IRQ_INDEX_IF, 0);
+       if (err)
+               dev_err(dev, "dpsw_set_irq_enable err %d\n", err);
+
+       fsl_mc_free_irqs(sw_dev);
+}
+
+static int swdev_port_attr_get(struct net_device *netdev,
+                              struct switchdev_attr *attr)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+
+       switch (attr->id) {
+       case SWITCHDEV_ATTR_ID_PORT_PARENT_ID:
+               attr->u.ppid.id_len = 1;
+               attr->u.ppid.id[0] = port_priv->ethsw_data->dev_id;
+               break;
+       case SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS:
+               attr->u.brport_flags =
+                       (port_priv->ethsw_data->learning ? BR_LEARNING : 0) |
+                       (port_priv->flood ? BR_FLOOD : 0);
+               break;
+       case SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS_SUPPORT:
+               attr->u.brport_flags_support = BR_LEARNING | BR_FLOOD;
+               break;
+       default:
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
+static int port_attr_stp_state_set(struct net_device *netdev,
+                                  struct switchdev_trans *trans,
+                                  u8 state)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       return ethsw_port_set_stp_state(port_priv, state);
+}
+
+static int port_attr_br_flags_set(struct net_device *netdev,
+                                 struct switchdev_trans *trans,
+                                 unsigned long flags)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int err = 0;
+
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       /* Learning is enabled per switch */
+       err = ethsw_set_learning(port_priv->ethsw_data, flags & BR_LEARNING);
+       if (err)
+               goto exit;
+
+       err = ethsw_port_set_flood(port_priv, flags & BR_FLOOD);
+
+exit:
+       return err;
+}
+
+static int swdev_port_attr_set(struct net_device *netdev,
+                              const struct switchdev_attr *attr,
+                              struct switchdev_trans *trans)
+{
+       int err = 0;
+
+       switch (attr->id) {
+       case SWITCHDEV_ATTR_ID_PORT_STP_STATE:
+               err = port_attr_stp_state_set(netdev, trans,
+                                             attr->u.stp_state);
+               break;
+       case SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS:
+               err = port_attr_br_flags_set(netdev, trans,
+                                            attr->u.brport_flags);
+               break;
+       case SWITCHDEV_ATTR_ID_BRIDGE_VLAN_FILTERING:
+               /* VLANs are supported by default  */
+               break;
+       default:
+               err = -EOPNOTSUPP;
+               break;
+       }
+
+       return err;
+}
+
+static int port_vlans_add(struct net_device *netdev,
+                         const struct switchdev_obj_port_vlan *vlan,
+                         struct switchdev_trans *trans)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int vid, err;
+
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       for (vid = vlan->vid_begin; vid <= vlan->vid_end; vid++) {
+               if (!port_priv->ethsw_data->vlans[vid]) {
+                       /* this is a new VLAN */
+                       err = ethsw_add_vlan(port_priv->ethsw_data, vid);
+                       if (err)
+                               return err;
+
+                       port_priv->ethsw_data->vlans[vid] |= ETHSW_VLAN_GLOBAL;
+               }
+               err = ethsw_port_add_vlan(port_priv, vid, vlan->flags);
+               if (err)
+                       break;
+       }
+
+       return err;
+}
+
+static int port_lookup_address(struct net_device *netdev, int is_uc,
+                              const unsigned char *addr)
+{
+       struct netdev_hw_addr_list *list = (is_uc) ? &netdev->uc : &netdev->mc;
+       struct netdev_hw_addr *ha;
+
+       netif_addr_lock_bh(netdev);
+       list_for_each_entry(ha, &list->list, list) {
+               if (ether_addr_equal(ha->addr, addr)) {
+                       netif_addr_unlock_bh(netdev);
+                       return 1;
+               }
+       }
+       netif_addr_unlock_bh(netdev);
+       return 0;
+}
+
+static int port_mdb_add(struct net_device *netdev,
+                       const struct switchdev_obj_port_mdb *mdb,
+                       struct switchdev_trans *trans)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int err;
+
+       if (switchdev_trans_ph_prepare(trans))
+               return 0;
+
+       /* Check if address is already set on this port */
+       if (port_lookup_address(netdev, 0, mdb->addr))
+               return -EEXIST;
+
+       err = ethsw_port_fdb_add_mc(port_priv, mdb->addr);
+       if (err)
+               return err;
+
+       err = dev_mc_add(netdev, mdb->addr);
+       if (err) {
+               netdev_err(netdev, "dev_mc_add err %d\n", err);
+               ethsw_port_fdb_del_mc(port_priv, mdb->addr);
+       }
+
+       return err;
+}
+
+static int swdev_port_obj_add(struct net_device *netdev,
+                             const struct switchdev_obj *obj,
+                             struct switchdev_trans *trans)
+{
+       int err;
+
+       switch (obj->id) {
+       case SWITCHDEV_OBJ_ID_PORT_VLAN:
+               err = port_vlans_add(netdev,
+                                    SWITCHDEV_OBJ_PORT_VLAN(obj),
+                                    trans);
+               break;
+       case SWITCHDEV_OBJ_ID_PORT_MDB:
+               err = port_mdb_add(netdev,
+                                  SWITCHDEV_OBJ_PORT_MDB(obj),
+                                  trans);
+               break;
+       default:
+               err = -EOPNOTSUPP;
+               break;
+       }
+
+       return err;
+}
+
+static int ethsw_port_del_vlan(struct ethsw_port_priv *port_priv, u16 vid)
+{
+       struct ethsw_core *ethsw = port_priv->ethsw_data;
+       struct net_device *netdev = port_priv->netdev;
+       struct dpsw_vlan_if_cfg vcfg;
+       int i, err;
+
+       if (!port_priv->vlans[vid])
+               return -ENOENT;
+
+       if (port_priv->vlans[vid] & ETHSW_VLAN_PVID) {
+               struct dpsw_tci_cfg tci_cfg = { 0 };
+
+               err = ethsw_port_set_tci(port_priv, &tci_cfg);
+               if (err)
+                       return err;
+       }
+
+       vcfg.num_ifs = 1;
+       vcfg.if_id[0] = port_priv->idx;
+       if (port_priv->vlans[vid] & ETHSW_VLAN_UNTAGGED) {
+               err = dpsw_vlan_remove_if_untagged(ethsw->mc_io, 0,
+                                                  ethsw->dpsw_handle,
+                                                  vid, &vcfg);
+               if (err) {
+                       netdev_err(netdev,
+                                  "dpsw_vlan_remove_if_untagged err %d\n",
+                                  err);
+               }
+               port_priv->vlans[vid] &= ~ETHSW_VLAN_UNTAGGED;
+       }
+
+       if (port_priv->vlans[vid] & ETHSW_VLAN_MEMBER) {
+               err = dpsw_vlan_remove_if(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                         vid, &vcfg);
+               if (err) {
+                       netdev_err(netdev,
+                                  "dpsw_vlan_remove_if err %d\n", err);
+                       return err;
+               }
+               port_priv->vlans[vid] &= ~ETHSW_VLAN_MEMBER;
+
+               /* Delete VLAN from switch if it is no longer configured on
+                * any port
+                */
+               for (i = 0; i < ethsw->sw_attr.num_ifs; i++)
+                       if (ethsw->ports[i]->vlans[vid] & ETHSW_VLAN_MEMBER)
+                               return 0; /* Found a port member in VID */
+
+               ethsw->vlans[vid] &= ~ETHSW_VLAN_GLOBAL;
+
+               err = ethsw_dellink_switch(ethsw, vid);
+               if (err)
+                       return err;
+       }
+
+       return 0;
+}
+
+static int port_vlans_del(struct net_device *netdev,
+                         const struct switchdev_obj_port_vlan *vlan)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int vid, err;
+
+       for (vid = vlan->vid_begin; vid <= vlan->vid_end; vid++) {
+               err = ethsw_port_del_vlan(port_priv, vid);
+               if (err)
+                       break;
+       }
+
+       return err;
+}
+
+static int port_mdb_del(struct net_device *netdev,
+                       const struct switchdev_obj_port_mdb *mdb)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int err;
+
+       if (!port_lookup_address(netdev, 0, mdb->addr))
+               return -ENOENT;
+
+       err = ethsw_port_fdb_del_mc(port_priv, mdb->addr);
+       if (err)
+               return err;
+
+       err = dev_mc_del(netdev, mdb->addr);
+       if (err) {
+               netdev_err(netdev, "dev_mc_del err %d\n", err);
+               return err;
+       }
+
+       return err;
+}
+
+static int swdev_port_obj_del(struct net_device *netdev,
+                             const struct switchdev_obj *obj)
+{
+       int err;
+
+       switch (obj->id) {
+       case SWITCHDEV_OBJ_ID_PORT_VLAN:
+               err = port_vlans_del(netdev, SWITCHDEV_OBJ_PORT_VLAN(obj));
+               break;
+       case SWITCHDEV_OBJ_ID_PORT_MDB:
+               err = port_mdb_del(netdev, SWITCHDEV_OBJ_PORT_MDB(obj));
+               break;
+       default:
+               err = -EOPNOTSUPP;
+               break;
+       }
+       return err;
+}
+
+static const struct switchdev_ops ethsw_port_switchdev_ops = {
+       .switchdev_port_attr_get        = swdev_port_attr_get,
+       .switchdev_port_attr_set        = swdev_port_attr_set,
+       .switchdev_port_obj_add         = swdev_port_obj_add,
+       .switchdev_port_obj_del         = swdev_port_obj_del,
+};
+
+/* For the moment, only flood setting needs to be updated */
+static int port_bridge_join(struct net_device *netdev,
+                           struct net_device *upper_dev)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       struct ethsw_core *ethsw = port_priv->ethsw_data;
+       int i, err;
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++)
+               if (ethsw->ports[i]->bridge_dev &&
+                   (ethsw->ports[i]->bridge_dev != upper_dev)) {
+                       netdev_err(netdev,
+                                  "Another switch port is connected to %s\n",
+                                  ethsw->ports[i]->bridge_dev->name);
+                       return -EINVAL;
+               }
+
+       /* Enable flooding */
+       err = ethsw_port_set_flood(port_priv, 1);
+       if (!err)
+               port_priv->bridge_dev = upper_dev;
+
+       return err;
+}
+
+static int port_bridge_leave(struct net_device *netdev)
+{
+       struct ethsw_port_priv *port_priv = netdev_priv(netdev);
+       int err;
+
+       /* Disable flooding */
+       err = ethsw_port_set_flood(port_priv, 0);
+       if (!err)
+               port_priv->bridge_dev = NULL;
+
+       return err;
+}
+
+static int port_netdevice_event(struct notifier_block *unused,
+                               unsigned long event, void *ptr)
+{
+       struct net_device *netdev = netdev_notifier_info_to_dev(ptr);
+       struct netdev_notifier_changeupper_info *info = ptr;
+       struct net_device *upper_dev;
+       int err = 0;
+
+       if (netdev->netdev_ops != &ethsw_port_ops)
+               return NOTIFY_DONE;
+
+       /* Handle just upper dev link/unlink for the moment */
+       if (event == NETDEV_CHANGEUPPER) {
+               upper_dev = info->upper_dev;
+               if (netif_is_bridge_master(upper_dev)) {
+                       if (info->linking)
+                               err = port_bridge_join(netdev, upper_dev);
+                       else
+                               err = port_bridge_leave(netdev);
+               }
+       }
+
+       return notifier_from_errno(err);
+}
+
+static struct notifier_block port_nb __read_mostly = {
+       .notifier_call = port_netdevice_event,
+};
+
+struct ethsw_switchdev_event_work {
+       struct work_struct work;
+       struct switchdev_notifier_fdb_info fdb_info;
+       struct net_device *dev;
+       unsigned long event;
+};
+
+static void ethsw_switchdev_event_work(struct work_struct *work)
+{
+       struct ethsw_switchdev_event_work *switchdev_work =
+               container_of(work, struct ethsw_switchdev_event_work, work);
+       struct net_device *dev = switchdev_work->dev;
+       struct switchdev_notifier_fdb_info *fdb_info;
+       struct ethsw_port_priv *port_priv;
+
+       rtnl_lock();
+       port_priv = netdev_priv(dev);
+       fdb_info = &switchdev_work->fdb_info;
+
+       switch (switchdev_work->event) {
+       case SWITCHDEV_FDB_ADD_TO_DEVICE:
+               if (is_unicast_ether_addr(fdb_info->addr))
+                       ethsw_port_fdb_add_uc(netdev_priv(dev), fdb_info->addr);
+               else
+                       ethsw_port_fdb_add_mc(netdev_priv(dev), fdb_info->addr);
+               break;
+       case SWITCHDEV_FDB_DEL_TO_DEVICE:
+               if (is_unicast_ether_addr(fdb_info->addr))
+                       ethsw_port_fdb_del_uc(netdev_priv(dev), fdb_info->addr);
+               else
+                       ethsw_port_fdb_del_mc(netdev_priv(dev), fdb_info->addr);
+               break;
+       }
+
+       rtnl_unlock();
+       kfree(switchdev_work->fdb_info.addr);
+       kfree(switchdev_work);
+       dev_put(dev);
+}
+
+/* Called under rcu_read_lock() */
+static int port_switchdev_event(struct notifier_block *unused,
+                               unsigned long event, void *ptr)
+{
+       struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
+       struct ethsw_switchdev_event_work *switchdev_work;
+       struct switchdev_notifier_fdb_info *fdb_info = ptr;
+
+       switchdev_work = kzalloc(sizeof(*switchdev_work), GFP_ATOMIC);
+       if (!switchdev_work)
+               return NOTIFY_BAD;
+
+       INIT_WORK(&switchdev_work->work, ethsw_switchdev_event_work);
+       switchdev_work->dev = dev;
+       switchdev_work->event = event;
+
+       switch (event) {
+       case SWITCHDEV_FDB_ADD_TO_DEVICE:
+       case SWITCHDEV_FDB_DEL_TO_DEVICE:
+               memcpy(&switchdev_work->fdb_info, ptr,
+                      sizeof(switchdev_work->fdb_info));
+               switchdev_work->fdb_info.addr = kzalloc(ETH_ALEN, GFP_ATOMIC);
+               if (!switchdev_work->fdb_info.addr)
+                       goto err_addr_alloc;
+
+               ether_addr_copy((u8 *)switchdev_work->fdb_info.addr,
+                               fdb_info->addr);
+
+               /* Take a reference on the device to avoid being freed. */
+               dev_hold(dev);
+               break;
+       default:
+               return NOTIFY_DONE;
+       }
+
+       queue_work(ethsw_owq, &switchdev_work->work);
+
+       return NOTIFY_DONE;
+
+err_addr_alloc:
+       kfree(switchdev_work);
+       return NOTIFY_BAD;
+}
+
+static struct notifier_block port_switchdev_nb = {
+       .notifier_call = port_switchdev_event,
+};
+
+static int ethsw_register_notifier(struct device *dev)
+{
+       int err;
+
+       err = register_netdevice_notifier(&port_nb);
+       if (err) {
+               dev_err(dev, "Failed to register netdev notifier\n");
+               return err;
+       }
+
+       err = register_switchdev_notifier(&port_switchdev_nb);
+       if (err) {
+               dev_err(dev, "Failed to register switchdev notifier\n");
+               goto err_switchdev_nb;
+       }
+
+       return 0;
+
+err_switchdev_nb:
+       unregister_netdevice_notifier(&port_nb);
+       return err;
+}
+
+static int ethsw_open(struct ethsw_core *ethsw)
+{
+       struct ethsw_port_priv *port_priv = NULL;
+       int i, err;
+
+       err = dpsw_enable(ethsw->mc_io, 0, ethsw->dpsw_handle);
+       if (err) {
+               dev_err(ethsw->dev, "dpsw_enable err %d\n", err);
+               return err;
+       }
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++) {
+               port_priv = ethsw->ports[i];
+               err = dev_open(port_priv->netdev);
+               if (err) {
+                       netdev_err(port_priv->netdev, "dev_open err %d\n", err);
+                       return err;
+               }
+       }
+
+       return 0;
+}
+
+static int ethsw_stop(struct ethsw_core *ethsw)
+{
+       struct ethsw_port_priv *port_priv = NULL;
+       int i, err;
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++) {
+               port_priv = ethsw->ports[i];
+               dev_close(port_priv->netdev);
+       }
+
+       err = dpsw_disable(ethsw->mc_io, 0, ethsw->dpsw_handle);
+       if (err) {
+               dev_err(ethsw->dev, "dpsw_disable err %d\n", err);
+               return err;
+       }
+
+       return 0;
+}
+
+static int ethsw_init(struct fsl_mc_device *sw_dev)
+{
+       struct device *dev = &sw_dev->dev;
+       struct ethsw_core *ethsw = dev_get_drvdata(dev);
+       u16 version_major, version_minor, i;
+       struct dpsw_stp_cfg stp_cfg;
+       int err;
+
+       ethsw->dev_id = sw_dev->obj_desc.id;
+
+       err = dpsw_open(ethsw->mc_io, 0, ethsw->dev_id, &ethsw->dpsw_handle);
+       if (err) {
+               dev_err(dev, "dpsw_open err %d\n", err);
+               return err;
+       }
+
+       err = dpsw_get_attributes(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                 &ethsw->sw_attr);
+       if (err) {
+               dev_err(dev, "dpsw_get_attributes err %d\n", err);
+               goto err_close;
+       }
+
+       err = dpsw_get_api_version(ethsw->mc_io, 0,
+                                  &version_major,
+                                  &version_minor);
+       if (err) {
+               dev_err(dev, "dpsw_get_api_version err %d\n", err);
+               goto err_close;
+       }
+
+       /* Minimum supported DPSW version check */
+       if (version_major < DPSW_MIN_VER_MAJOR ||
+           (version_major == DPSW_MIN_VER_MAJOR &&
+            version_minor < DPSW_MIN_VER_MINOR)) {
+               dev_err(dev, "DPSW version %d:%d not supported. Use %d.%d or greater.\n",
+                       version_major,
+                       version_minor,
+                       DPSW_MIN_VER_MAJOR, DPSW_MIN_VER_MINOR);
+               err = -ENOTSUPP;
+               goto err_close;
+       }
+
+       err = dpsw_reset(ethsw->mc_io, 0, ethsw->dpsw_handle);
+       if (err) {
+               dev_err(dev, "dpsw_reset err %d\n", err);
+               goto err_close;
+       }
+
+       err = dpsw_fdb_set_learning_mode(ethsw->mc_io, 0, ethsw->dpsw_handle, 0,
+                                        DPSW_FDB_LEARNING_MODE_HW);
+       if (err) {
+               dev_err(dev, "dpsw_fdb_set_learning_mode err %d\n", err);
+               goto err_close;
+       }
+
+       stp_cfg.vlan_id = DEFAULT_VLAN_ID;
+       stp_cfg.state = DPSW_STP_STATE_FORWARDING;
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++) {
+               err = dpsw_if_set_stp(ethsw->mc_io, 0, ethsw->dpsw_handle, i,
+                                     &stp_cfg);
+               if (err) {
+                       dev_err(dev, "dpsw_if_set_stp err %d for port %d\n",
+                               err, i);
+                       goto err_close;
+               }
+
+               err = dpsw_if_set_broadcast(ethsw->mc_io, 0,
+                                           ethsw->dpsw_handle, i, 1);
+               if (err) {
+                       dev_err(dev,
+                               "dpsw_if_set_broadcast err %d for port %d\n",
+                               err, i);
+                       goto err_close;
+               }
+       }
+
+       ethsw_owq = alloc_ordered_workqueue("%s_ordered", WQ_MEM_RECLAIM,
+                                           "ethsw");
+       if (!ethsw_owq) {
+               err = -ENOMEM;
+               goto err_close;
+       }
+
+       err = ethsw_register_notifier(dev);
+       if (err)
+               goto err_destroy_ordered_workqueue;
+
+       return 0;
+
+err_destroy_ordered_workqueue:
+       destroy_workqueue(ethsw_owq);
+
+err_close:
+       dpsw_close(ethsw->mc_io, 0, ethsw->dpsw_handle);
+       return err;
+}
+
+static int ethsw_port_init(struct ethsw_port_priv *port_priv, u16 port)
+{
+       const char def_mcast[ETH_ALEN] = {0x01, 0x00, 0x5e, 0x00, 0x00, 0x01};
+       struct net_device *netdev = port_priv->netdev;
+       struct ethsw_core *ethsw = port_priv->ethsw_data;
+       struct dpsw_tci_cfg tci_cfg = {0};
+       struct dpsw_vlan_if_cfg vcfg;
+       int err;
+
+       /* Switch starts with all ports configured to VLAN 1. Need to
+        * remove this setting to allow configuration at bridge join
+        */
+       vcfg.num_ifs = 1;
+       vcfg.if_id[0] = port_priv->idx;
+
+       err = dpsw_vlan_remove_if_untagged(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                          DEFAULT_VLAN_ID, &vcfg);
+       if (err) {
+               netdev_err(netdev, "dpsw_vlan_remove_if_untagged err %d\n",
+                          err);
+               return err;
+       }
+
+       err = ethsw_port_set_tci(port_priv, &tci_cfg);
+       if (err)
+               return err;
+
+       err = dpsw_vlan_remove_if(ethsw->mc_io, 0, ethsw->dpsw_handle,
+                                 DEFAULT_VLAN_ID, &vcfg);
+       if (err) {
+               netdev_err(netdev, "dpsw_vlan_remove_if err %d\n", err);
+               return err;
+       }
+
+       err = ethsw_port_fdb_add_mc(port_priv, def_mcast);
+
+       return err;
+}
+
+static void ethsw_unregister_notifier(struct device *dev)
+{
+       int err;
+
+       err = unregister_switchdev_notifier(&port_switchdev_nb);
+       if (err)
+               dev_err(dev,
+                       "Failed to unregister switchdev notifier (%d)\n", err);
+
+       err = unregister_netdevice_notifier(&port_nb);
+       if (err)
+               dev_err(dev,
+                       "Failed to unregister netdev notifier (%d)\n", err);
+}
+
+static void ethsw_takedown(struct fsl_mc_device *sw_dev)
+{
+       struct device *dev = &sw_dev->dev;
+       struct ethsw_core *ethsw = dev_get_drvdata(dev);
+       int err;
+
+       ethsw_unregister_notifier(dev);
+
+       err = dpsw_close(ethsw->mc_io, 0, ethsw->dpsw_handle);
+       if (err)
+               dev_warn(dev, "dpsw_close err %d\n", err);
+}
+
+static int ethsw_remove(struct fsl_mc_device *sw_dev)
+{
+       struct ethsw_port_priv *port_priv;
+       struct ethsw_core *ethsw;
+       struct device *dev;
+       int i;
+
+       dev = &sw_dev->dev;
+       ethsw = dev_get_drvdata(dev);
+
+       ethsw_teardown_irqs(sw_dev);
+
+       destroy_workqueue(ethsw_owq);
+
+       rtnl_lock();
+       ethsw_stop(ethsw);
+       rtnl_unlock();
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++) {
+               port_priv = ethsw->ports[i];
+               unregister_netdev(port_priv->netdev);
+               free_netdev(port_priv->netdev);
+       }
+       kfree(ethsw->ports);
+
+       ethsw_takedown(sw_dev);
+       fsl_mc_portal_free(ethsw->mc_io);
+
+       kfree(ethsw);
+
+       dev_set_drvdata(dev, NULL);
+
+       return 0;
+}
+
+static int ethsw_probe_port(struct ethsw_core *ethsw, u16 port_idx)
+{
+       struct ethsw_port_priv *port_priv;
+       struct device *dev = ethsw->dev;
+       struct net_device *port_netdev;
+       int err;
+
+       port_netdev = alloc_etherdev(sizeof(struct ethsw_port_priv));
+       if (!port_netdev) {
+               dev_err(dev, "alloc_etherdev error\n");
+               return -ENOMEM;
+       }
+
+       port_priv = netdev_priv(port_netdev);
+       port_priv->netdev = port_netdev;
+       port_priv->ethsw_data = ethsw;
+
+       port_priv->idx = port_idx;
+       port_priv->stp_state = BR_STATE_FORWARDING;
+
+       /* Flooding is implicitly enabled */
+       port_priv->flood = true;
+
+       SET_NETDEV_DEV(port_netdev, dev);
+       port_netdev->netdev_ops = &ethsw_port_ops;
+       port_netdev->switchdev_ops = &ethsw_port_switchdev_ops;
+
+       /* Set MTU limits */
+       port_netdev->min_mtu = ETH_MIN_MTU;
+       port_netdev->max_mtu = ETHSW_MAX_FRAME_LENGTH;
+
+       err = register_netdev(port_netdev);
+       if (err < 0) {
+               dev_err(dev, "register_netdev error %d\n", err);
+               free_netdev(port_netdev);
+               return err;
+       }
+
+       ethsw->ports[port_idx] = port_priv;
+
+       return ethsw_port_init(port_priv, port_idx);
+}
+
+static int ethsw_probe(struct fsl_mc_device *sw_dev)
+{
+       struct device *dev = &sw_dev->dev;
+       struct ethsw_core *ethsw;
+       int i, err;
+
+       /* Allocate switch core*/
+       ethsw = kzalloc(sizeof(*ethsw), GFP_KERNEL);
+
+       if (!ethsw)
+               return -ENOMEM;
+
+       ethsw->dev = dev;
+       dev_set_drvdata(dev, ethsw);
+
+       err = fsl_mc_portal_allocate(sw_dev, 0, &ethsw->mc_io);
+       if (err) {
+               if (err == -ENXIO)
+                       err = -EPROBE_DEFER;
+               else
+                       dev_err(dev, "fsl_mc_portal_allocate err %d\n", err);
+               goto err_free_drvdata;
+       }
+
+       err = ethsw_init(sw_dev);
+       if (err)
+               goto err_free_cmdport;
+
+       /* DEFAULT_VLAN_ID is implicitly configured on the switch */
+       ethsw->vlans[DEFAULT_VLAN_ID] = ETHSW_VLAN_MEMBER;
+
+       /* Learning is implicitly enabled */
+       ethsw->learning = true;
+
+       ethsw->ports = kcalloc(ethsw->sw_attr.num_ifs, sizeof(*ethsw->ports),
+                              GFP_KERNEL);
+       if (!(ethsw->ports)) {
+               err = -ENOMEM;
+               goto err_takedown;
+       }
+
+       for (i = 0; i < ethsw->sw_attr.num_ifs; i++) {
+               err = ethsw_probe_port(ethsw, i);
+               if (err)
+                       goto err_free_ports;
+       }
+
+       /* Switch starts up enabled */
+       rtnl_lock();
+       err = ethsw_open(ethsw);
+       rtnl_unlock();
+       if (err)
+               goto err_free_ports;
+
+       /* Setup IRQs */
+       err = ethsw_setup_irqs(sw_dev);
+       if (err)
+               goto err_stop;
+
+       dev_info(dev, "probed %d port switch\n", ethsw->sw_attr.num_ifs);
+       return 0;
+
+err_stop:
+       rtnl_lock();
+       ethsw_stop(ethsw);
+       rtnl_unlock();
+
+err_free_ports:
+       /* Cleanup registered ports only */
+       for (i--; i >= 0; i--) {
+               unregister_netdev(ethsw->ports[i]->netdev);
+               free_netdev(ethsw->ports[i]->netdev);
+       }
+       kfree(ethsw->ports);
+
+err_takedown:
+       ethsw_takedown(sw_dev);
+
+err_free_cmdport:
+       fsl_mc_portal_free(ethsw->mc_io);
+
+err_free_drvdata:
+       kfree(ethsw);
+       dev_set_drvdata(dev, NULL);
+
+       return err;
+}
+
+static const struct fsl_mc_device_id ethsw_match_id_table[] = {
+       {
+               .vendor = FSL_MC_VENDOR_FREESCALE,
+               .obj_type = "dpsw",
+       },
+       { .vendor = 0x0 }
+};
+MODULE_DEVICE_TABLE(fslmc, ethsw_match_id_table);
+
+static struct fsl_mc_driver eth_sw_drv = {
+       .driver = {
+               .name = KBUILD_MODNAME,
+               .owner = THIS_MODULE,
+       },
+       .probe = ethsw_probe,
+       .remove = ethsw_remove,
+       .match_id_table = ethsw_match_id_table
+};
+
+module_fsl_mc_driver(eth_sw_drv);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("DPAA2 Ethernet Switch Driver");
diff --git a/drivers/staging/fsl-dpaa2/ethsw/ethsw.h b/drivers/staging/fsl-dpaa2/ethsw/ethsw.h
new file mode 100644 (file)
index 0000000..da607b4
--- /dev/null
@@ -0,0 +1,65 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * DPAA2 Ethernet Switch declarations
+ *
+ * Copyright 2014-2016 Freescale Semiconductor Inc.
+ * Copyright 2017-2018 NXP
+ *
+ */
+
+#ifndef __ETHSW_H
+#define __ETHSW_H
+
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/rtnetlink.h>
+#include <linux/if_vlan.h>
+#include <uapi/linux/if_bridge.h>
+#include <net/switchdev.h>
+#include <linux/if_bridge.h>
+
+#include "dpsw.h"
+
+/* Number of IRQs supported */
+#define DPSW_IRQ_NUM   2
+
+#define ETHSW_VLAN_MEMBER      1
+#define ETHSW_VLAN_UNTAGGED    2
+#define ETHSW_VLAN_PVID                4
+#define ETHSW_VLAN_GLOBAL      8
+
+/* Maximum Frame Length supported by HW (currently 10k) */
+#define DPAA2_MFL              (10 * 1024)
+#define ETHSW_MAX_FRAME_LENGTH (DPAA2_MFL - VLAN_ETH_HLEN - ETH_FCS_LEN)
+#define ETHSW_L2_MAX_FRM(mtu)  ((mtu) + VLAN_ETH_HLEN + ETH_FCS_LEN)
+
+struct ethsw_core;
+
+/* Per port private data */
+struct ethsw_port_priv {
+       struct net_device       *netdev;
+       u16                     idx;
+       struct ethsw_core       *ethsw_data;
+       u8                      link_state;
+       u8                      stp_state;
+       bool                    flood;
+
+       u8                      vlans[VLAN_VID_MASK + 1];
+       u16                     pvid;
+       struct net_device       *bridge_dev;
+};
+
+/* Switch data */
+struct ethsw_core {
+       struct device                   *dev;
+       struct fsl_mc_io                *mc_io;
+       u16                             dpsw_handle;
+       struct dpsw_attr                sw_attr;
+       int                             dev_id;
+       struct ethsw_port_priv          **ports;
+
+       u8                              vlans[VLAN_VID_MASK + 1];
+       bool                            learning;
+};
+
+#endif /* __ETHSW_H */