--- /dev/null
+// SPDX-License-Identifier: GPL-2.0
+/* Texas Instruments K3 AM65 Ethernet Switch SubSystem Driver ethtool ops
+ *
+ * Copyright (C) 2020 Texas Instruments Incorporated - http://www.ti.com/
+ *
+ */
+
+#include <linux/net_tstamp.h>
+#include <linux/phy.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+
+#include "am65-cpsw-nuss.h"
+#include "cpsw_ale.h"
+
+#define AM65_CPSW_REGDUMP_VER 0x1
+
+enum {
+       AM65_CPSW_REGDUMP_MOD_NUSS = 1,
+       AM65_CPSW_REGDUMP_MOD_RGMII_STATUS = 2,
+       AM65_CPSW_REGDUMP_MOD_MDIO = 3,
+       AM65_CPSW_REGDUMP_MOD_CPSW = 4,
+       AM65_CPSW_REGDUMP_MOD_CPSW_P0 = 5,
+       AM65_CPSW_REGDUMP_MOD_CPSW_P1 = 6,
+       AM65_CPSW_REGDUMP_MOD_CPSW_CPTS = 7,
+       AM65_CPSW_REGDUMP_MOD_CPSW_ALE = 8,
+       AM65_CPSW_REGDUMP_MOD_CPSW_ALE_TBL = 9,
+       AM65_CPSW_REGDUMP_MOD_LAST,
+};
+
+/**
+ * struct am65_cpsw_regdump_hdr - regdump record header
+ *
+ * @module_id: CPSW module ID
+ * @len: CPSW module registers space length in u32
+ */
+
+struct am65_cpsw_regdump_hdr {
+       u32 module_id;
+       u32 len;
+};
+
+/**
+ * struct am65_cpsw_regdump_item - regdump module description
+ *
+ * @hdr: CPSW module header
+ * @start_ofs: CPSW module registers start addr
+ * @end_ofs: CPSW module registers end addr
+ *
+ * Registers dump provided in the format:
+ *  u32 : module ID
+ *  u32 : dump length
+ *  u32[..len]: registers values
+ */
+struct am65_cpsw_regdump_item {
+       struct am65_cpsw_regdump_hdr hdr;
+       u32 start_ofs;
+       u32 end_ofs;
+};
+
+#define AM65_CPSW_REGDUMP_REC(mod, start, end) { \
+       .hdr.module_id = (mod), \
+       .hdr.len = (((u32 *)(end)) - ((u32 *)(start)) + 1) * sizeof(u32) * 2 + \
+                  sizeof(struct am65_cpsw_regdump_hdr), \
+       .start_ofs = (start), \
+       .end_ofs = end, \
+}
+
+static const struct am65_cpsw_regdump_item am65_cpsw_regdump[] = {
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_NUSS, 0x0, 0x1c),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_RGMII_STATUS, 0x30, 0x4c),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_MDIO, 0xf00, 0xffc),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_CPSW, 0x20000, 0x2011c),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_CPSW_P0, 0x21000, 0x21320),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_CPSW_P1, 0x22000, 0x223a4),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_CPSW_CPTS,
+                             0x3d000, 0x3d048),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_CPSW_ALE, 0x3e000, 0x3e13c),
+       AM65_CPSW_REGDUMP_REC(AM65_CPSW_REGDUMP_MOD_CPSW_ALE_TBL, 0, 0),
+};
+
+struct am65_cpsw_stats_regs {
+       u32     rx_good_frames;
+       u32     rx_broadcast_frames;
+       u32     rx_multicast_frames;
+       u32     rx_pause_frames;                /* slave */
+       u32     rx_crc_errors;
+       u32     rx_align_code_errors;           /* slave */
+       u32     rx_oversized_frames;
+       u32     rx_jabber_frames;               /* slave */
+       u32     rx_undersized_frames;
+       u32     rx_fragments;                   /* slave */
+       u32     ale_drop;
+       u32     ale_overrun_drop;
+       u32     rx_octets;
+       u32     tx_good_frames;
+       u32     tx_broadcast_frames;
+       u32     tx_multicast_frames;
+       u32     tx_pause_frames;                /* slave */
+       u32     tx_deferred_frames;             /* slave */
+       u32     tx_collision_frames;            /* slave */
+       u32     tx_single_coll_frames;          /* slave */
+       u32     tx_mult_coll_frames;            /* slave */
+       u32     tx_excessive_collisions;        /* slave */
+       u32     tx_late_collisions;             /* slave */
+       u32     rx_ipg_error;                   /* slave 10G only */
+       u32     tx_carrier_sense_errors;        /* slave */
+       u32     tx_octets;
+       u32     tx_64B_frames;
+       u32     tx_65_to_127B_frames;
+       u32     tx_128_to_255B_frames;
+       u32     tx_256_to_511B_frames;
+       u32     tx_512_to_1023B_frames;
+       u32     tx_1024B_frames;
+       u32     net_octets;
+       u32     rx_bottom_fifo_drop;
+       u32     rx_port_mask_drop;
+       u32     rx_top_fifo_drop;
+       u32     ale_rate_limit_drop;
+       u32     ale_vid_ingress_drop;
+       u32     ale_da_eq_sa_drop;
+       u32     ale_block_drop;                 /* K3 */
+       u32     ale_secure_drop;                /* K3 */
+       u32     ale_auth_drop;                  /* K3 */
+       u32     ale_unknown_ucast;
+       u32     ale_unknown_ucast_bytes;
+       u32     ale_unknown_mcast;
+       u32     ale_unknown_mcast_bytes;
+       u32     ale_unknown_bcast;
+       u32     ale_unknown_bcast_bytes;
+       u32     ale_pol_match;
+       u32     ale_pol_match_red;
+       u32     ale_pol_match_yellow;
+       u32     ale_mcast_sa_drop;              /* K3 */
+       u32     ale_dual_vlan_drop;             /* K3 */
+       u32     ale_len_err_drop;               /* K3 */
+       u32     ale_ip_next_hdr_drop;           /* K3 */
+       u32     ale_ipv4_frag_drop;             /* K3 */
+       u32     __rsvd_1[24];
+       u32     iet_rx_assembly_err;            /* K3 slave */
+       u32     iet_rx_assembly_ok;             /* K3 slave */
+       u32     iet_rx_smd_err;                 /* K3 slave */
+       u32     iet_rx_frag;                    /* K3 slave */
+       u32     iet_tx_hold;                    /* K3 slave */
+       u32     iet_tx_frag;                    /* K3 slave */
+       u32     __rsvd_2[9];
+       u32     tx_mem_protect_err;
+       /* following NU only */
+       u32     tx_pri0;
+       u32     tx_pri1;
+       u32     tx_pri2;
+       u32     tx_pri3;
+       u32     tx_pri4;
+       u32     tx_pri5;
+       u32     tx_pri6;
+       u32     tx_pri7;
+       u32     tx_pri0_bcnt;
+       u32     tx_pri1_bcnt;
+       u32     tx_pri2_bcnt;
+       u32     tx_pri3_bcnt;
+       u32     tx_pri4_bcnt;
+       u32     tx_pri5_bcnt;
+       u32     tx_pri6_bcnt;
+       u32     tx_pri7_bcnt;
+       u32     tx_pri0_drop;
+       u32     tx_pri1_drop;
+       u32     tx_pri2_drop;
+       u32     tx_pri3_drop;
+       u32     tx_pri4_drop;
+       u32     tx_pri5_drop;
+       u32     tx_pri6_drop;
+       u32     tx_pri7_drop;
+       u32     tx_pri0_drop_bcnt;
+       u32     tx_pri1_drop_bcnt;
+       u32     tx_pri2_drop_bcnt;
+       u32     tx_pri3_drop_bcnt;
+       u32     tx_pri4_drop_bcnt;
+       u32     tx_pri5_drop_bcnt;
+       u32     tx_pri6_drop_bcnt;
+       u32     tx_pri7_drop_bcnt;
+};
+
+struct am65_cpsw_ethtool_stat {
+       char desc[ETH_GSTRING_LEN];
+       int offset;
+};
+
+#define AM65_CPSW_STATS(prefix, field)                 \
+{                                                      \
+       #prefix#field,                                  \
+       offsetof(struct am65_cpsw_stats_regs, field)    \
+}
+
+static const struct am65_cpsw_ethtool_stat am65_host_stats[] = {
+       AM65_CPSW_STATS(p0_, rx_good_frames),
+       AM65_CPSW_STATS(p0_, rx_broadcast_frames),
+       AM65_CPSW_STATS(p0_, rx_multicast_frames),
+       AM65_CPSW_STATS(p0_, rx_crc_errors),
+       AM65_CPSW_STATS(p0_, rx_oversized_frames),
+       AM65_CPSW_STATS(p0_, rx_undersized_frames),
+       AM65_CPSW_STATS(p0_, ale_drop),
+       AM65_CPSW_STATS(p0_, ale_overrun_drop),
+       AM65_CPSW_STATS(p0_, rx_octets),
+       AM65_CPSW_STATS(p0_, tx_good_frames),
+       AM65_CPSW_STATS(p0_, tx_broadcast_frames),
+       AM65_CPSW_STATS(p0_, tx_multicast_frames),
+       AM65_CPSW_STATS(p0_, tx_octets),
+       AM65_CPSW_STATS(p0_, tx_64B_frames),
+       AM65_CPSW_STATS(p0_, tx_65_to_127B_frames),
+       AM65_CPSW_STATS(p0_, tx_128_to_255B_frames),
+       AM65_CPSW_STATS(p0_, tx_256_to_511B_frames),
+       AM65_CPSW_STATS(p0_, tx_512_to_1023B_frames),
+       AM65_CPSW_STATS(p0_, tx_1024B_frames),
+       AM65_CPSW_STATS(p0_, net_octets),
+       AM65_CPSW_STATS(p0_, rx_bottom_fifo_drop),
+       AM65_CPSW_STATS(p0_, rx_port_mask_drop),
+       AM65_CPSW_STATS(p0_, rx_top_fifo_drop),
+       AM65_CPSW_STATS(p0_, ale_rate_limit_drop),
+       AM65_CPSW_STATS(p0_, ale_vid_ingress_drop),
+       AM65_CPSW_STATS(p0_, ale_da_eq_sa_drop),
+       AM65_CPSW_STATS(p0_, ale_block_drop),
+       AM65_CPSW_STATS(p0_, ale_secure_drop),
+       AM65_CPSW_STATS(p0_, ale_auth_drop),
+       AM65_CPSW_STATS(p0_, ale_unknown_ucast),
+       AM65_CPSW_STATS(p0_, ale_unknown_ucast_bytes),
+       AM65_CPSW_STATS(p0_, ale_unknown_mcast),
+       AM65_CPSW_STATS(p0_, ale_unknown_mcast_bytes),
+       AM65_CPSW_STATS(p0_, ale_unknown_bcast),
+       AM65_CPSW_STATS(p0_, ale_unknown_bcast_bytes),
+       AM65_CPSW_STATS(p0_, ale_pol_match),
+       AM65_CPSW_STATS(p0_, ale_pol_match_red),
+       AM65_CPSW_STATS(p0_, ale_pol_match_yellow),
+       AM65_CPSW_STATS(p0_, ale_mcast_sa_drop),
+       AM65_CPSW_STATS(p0_, ale_dual_vlan_drop),
+       AM65_CPSW_STATS(p0_, ale_len_err_drop),
+       AM65_CPSW_STATS(p0_, ale_ip_next_hdr_drop),
+       AM65_CPSW_STATS(p0_, ale_ipv4_frag_drop),
+       AM65_CPSW_STATS(p0_, tx_mem_protect_err),
+       AM65_CPSW_STATS(p0_, tx_pri0),
+       AM65_CPSW_STATS(p0_, tx_pri1),
+       AM65_CPSW_STATS(p0_, tx_pri2),
+       AM65_CPSW_STATS(p0_, tx_pri3),
+       AM65_CPSW_STATS(p0_, tx_pri4),
+       AM65_CPSW_STATS(p0_, tx_pri5),
+       AM65_CPSW_STATS(p0_, tx_pri6),
+       AM65_CPSW_STATS(p0_, tx_pri7),
+       AM65_CPSW_STATS(p0_, tx_pri0_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri1_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri2_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri3_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri4_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri5_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri6_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri7_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri0_drop),
+       AM65_CPSW_STATS(p0_, tx_pri1_drop),
+       AM65_CPSW_STATS(p0_, tx_pri2_drop),
+       AM65_CPSW_STATS(p0_, tx_pri3_drop),
+       AM65_CPSW_STATS(p0_, tx_pri4_drop),
+       AM65_CPSW_STATS(p0_, tx_pri5_drop),
+       AM65_CPSW_STATS(p0_, tx_pri6_drop),
+       AM65_CPSW_STATS(p0_, tx_pri7_drop),
+       AM65_CPSW_STATS(p0_, tx_pri0_drop_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri1_drop_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri2_drop_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri3_drop_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri4_drop_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri5_drop_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri6_drop_bcnt),
+       AM65_CPSW_STATS(p0_, tx_pri7_drop_bcnt),
+};
+
+static const struct am65_cpsw_ethtool_stat am65_slave_stats[] = {
+       AM65_CPSW_STATS(, rx_good_frames),
+       AM65_CPSW_STATS(, rx_broadcast_frames),
+       AM65_CPSW_STATS(, rx_multicast_frames),
+       AM65_CPSW_STATS(, rx_pause_frames),
+       AM65_CPSW_STATS(, rx_crc_errors),
+       AM65_CPSW_STATS(, rx_align_code_errors),
+       AM65_CPSW_STATS(, rx_oversized_frames),
+       AM65_CPSW_STATS(, rx_jabber_frames),
+       AM65_CPSW_STATS(, rx_undersized_frames),
+       AM65_CPSW_STATS(, rx_fragments),
+       AM65_CPSW_STATS(, ale_drop),
+       AM65_CPSW_STATS(, ale_overrun_drop),
+       AM65_CPSW_STATS(, rx_octets),
+       AM65_CPSW_STATS(, tx_good_frames),
+       AM65_CPSW_STATS(, tx_broadcast_frames),
+       AM65_CPSW_STATS(, tx_multicast_frames),
+       AM65_CPSW_STATS(, tx_pause_frames),
+       AM65_CPSW_STATS(, tx_deferred_frames),
+       AM65_CPSW_STATS(, tx_collision_frames),
+       AM65_CPSW_STATS(, tx_single_coll_frames),
+       AM65_CPSW_STATS(, tx_mult_coll_frames),
+       AM65_CPSW_STATS(, tx_excessive_collisions),
+       AM65_CPSW_STATS(, tx_late_collisions),
+       AM65_CPSW_STATS(, rx_ipg_error),
+       AM65_CPSW_STATS(, tx_carrier_sense_errors),
+       AM65_CPSW_STATS(, tx_octets),
+       AM65_CPSW_STATS(, tx_64B_frames),
+       AM65_CPSW_STATS(, tx_65_to_127B_frames),
+       AM65_CPSW_STATS(, tx_128_to_255B_frames),
+       AM65_CPSW_STATS(, tx_256_to_511B_frames),
+       AM65_CPSW_STATS(, tx_512_to_1023B_frames),
+       AM65_CPSW_STATS(, tx_1024B_frames),
+       AM65_CPSW_STATS(, net_octets),
+       AM65_CPSW_STATS(, rx_bottom_fifo_drop),
+       AM65_CPSW_STATS(, rx_port_mask_drop),
+       AM65_CPSW_STATS(, rx_top_fifo_drop),
+       AM65_CPSW_STATS(, ale_rate_limit_drop),
+       AM65_CPSW_STATS(, ale_vid_ingress_drop),
+       AM65_CPSW_STATS(, ale_da_eq_sa_drop),
+       AM65_CPSW_STATS(, ale_block_drop),
+       AM65_CPSW_STATS(, ale_secure_drop),
+       AM65_CPSW_STATS(, ale_auth_drop),
+       AM65_CPSW_STATS(, ale_unknown_ucast),
+       AM65_CPSW_STATS(, ale_unknown_ucast_bytes),
+       AM65_CPSW_STATS(, ale_unknown_mcast),
+       AM65_CPSW_STATS(, ale_unknown_mcast_bytes),
+       AM65_CPSW_STATS(, ale_unknown_bcast),
+       AM65_CPSW_STATS(, ale_unknown_bcast_bytes),
+       AM65_CPSW_STATS(, ale_pol_match),
+       AM65_CPSW_STATS(, ale_pol_match_red),
+       AM65_CPSW_STATS(, ale_pol_match_yellow),
+       AM65_CPSW_STATS(, ale_mcast_sa_drop),
+       AM65_CPSW_STATS(, ale_dual_vlan_drop),
+       AM65_CPSW_STATS(, ale_len_err_drop),
+       AM65_CPSW_STATS(, ale_ip_next_hdr_drop),
+       AM65_CPSW_STATS(, ale_ipv4_frag_drop),
+       AM65_CPSW_STATS(, iet_rx_assembly_err),
+       AM65_CPSW_STATS(, iet_rx_assembly_ok),
+       AM65_CPSW_STATS(, iet_rx_smd_err),
+       AM65_CPSW_STATS(, iet_rx_frag),
+       AM65_CPSW_STATS(, iet_tx_hold),
+       AM65_CPSW_STATS(, iet_tx_frag),
+       AM65_CPSW_STATS(, tx_mem_protect_err),
+       AM65_CPSW_STATS(, tx_pri0),
+       AM65_CPSW_STATS(, tx_pri1),
+       AM65_CPSW_STATS(, tx_pri2),
+       AM65_CPSW_STATS(, tx_pri3),
+       AM65_CPSW_STATS(, tx_pri4),
+       AM65_CPSW_STATS(, tx_pri5),
+       AM65_CPSW_STATS(, tx_pri6),
+       AM65_CPSW_STATS(, tx_pri7),
+       AM65_CPSW_STATS(, tx_pri0_bcnt),
+       AM65_CPSW_STATS(, tx_pri1_bcnt),
+       AM65_CPSW_STATS(, tx_pri2_bcnt),
+       AM65_CPSW_STATS(, tx_pri3_bcnt),
+       AM65_CPSW_STATS(, tx_pri4_bcnt),
+       AM65_CPSW_STATS(, tx_pri5_bcnt),
+       AM65_CPSW_STATS(, tx_pri6_bcnt),
+       AM65_CPSW_STATS(, tx_pri7_bcnt),
+       AM65_CPSW_STATS(, tx_pri0_drop),
+       AM65_CPSW_STATS(, tx_pri1_drop),
+       AM65_CPSW_STATS(, tx_pri2_drop),
+       AM65_CPSW_STATS(, tx_pri3_drop),
+       AM65_CPSW_STATS(, tx_pri4_drop),
+       AM65_CPSW_STATS(, tx_pri5_drop),
+       AM65_CPSW_STATS(, tx_pri6_drop),
+       AM65_CPSW_STATS(, tx_pri7_drop),
+       AM65_CPSW_STATS(, tx_pri0_drop_bcnt),
+       AM65_CPSW_STATS(, tx_pri1_drop_bcnt),
+       AM65_CPSW_STATS(, tx_pri2_drop_bcnt),
+       AM65_CPSW_STATS(, tx_pri3_drop_bcnt),
+       AM65_CPSW_STATS(, tx_pri4_drop_bcnt),
+       AM65_CPSW_STATS(, tx_pri5_drop_bcnt),
+       AM65_CPSW_STATS(, tx_pri6_drop_bcnt),
+       AM65_CPSW_STATS(, tx_pri7_drop_bcnt),
+};
+
+/* Ethtool priv_flags */
+static const char am65_cpsw_ethtool_priv_flags[][ETH_GSTRING_LEN] = {
+#define        AM65_CPSW_PRIV_P0_RX_PTYPE_RROBIN       BIT(0)
+       "p0-rx-ptype-rrobin",
+};
+
+static int am65_cpsw_ethtool_op_begin(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       int ret;
+
+       ret = pm_runtime_get_sync(common->dev);
+       if (ret < 0) {
+               dev_err(common->dev, "ethtool begin failed %d\n", ret);
+               pm_runtime_put_noidle(common->dev);
+       }
+
+       return ret;
+}
+
+static void am65_cpsw_ethtool_op_complete(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       int ret;
+
+       ret = pm_runtime_put(common->dev);
+       if (ret < 0 && ret != -EBUSY)
+               dev_err(common->dev, "ethtool complete failed %d\n", ret);
+}
+
+static void am65_cpsw_get_drvinfo(struct net_device *ndev,
+                                 struct ethtool_drvinfo *info)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+
+       strlcpy(info->driver, dev_driver_string(common->dev),
+               sizeof(info->driver));
+       strlcpy(info->bus_info, dev_name(common->dev), sizeof(info->bus_info));
+}
+
+static u32 am65_cpsw_get_msglevel(struct net_device *ndev)
+{
+       struct am65_cpsw_ndev_priv *priv = am65_ndev_to_priv(ndev);
+
+       return priv->msg_enable;
+}
+
+static void am65_cpsw_set_msglevel(struct net_device *ndev, u32 value)
+{
+       struct am65_cpsw_ndev_priv *priv = am65_ndev_to_priv(ndev);
+
+       priv->msg_enable = value;
+}
+
+static void am65_cpsw_get_channels(struct net_device *ndev,
+                                  struct ethtool_channels *ch)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+
+       ch->max_rx = AM65_CPSW_MAX_RX_QUEUES;
+       ch->max_tx = AM65_CPSW_MAX_TX_QUEUES;
+       ch->rx_count = AM65_CPSW_MAX_RX_QUEUES;
+       ch->tx_count = common->tx_ch_num;
+}
+
+static int am65_cpsw_set_channels(struct net_device *ndev,
+                                 struct ethtool_channels *chs)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+
+       if (!chs->rx_count || !chs->tx_count)
+               return -EINVAL;
+
+       /* Check if interface is up. Can change the num queues when
+        * the interface is down.
+        */
+       if (netif_running(ndev))
+               return -EBUSY;
+
+       am65_cpsw_nuss_remove_tx_chns(common);
+
+       return am65_cpsw_nuss_update_tx_chns(common, chs->tx_count);
+}
+
+static void am65_cpsw_get_ringparam(struct net_device *ndev,
+                                   struct ethtool_ringparam *ering)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+
+       /* not supported */
+       ering->tx_pending = common->tx_chns[0].descs_num;
+       ering->rx_pending = common->rx_chns.descs_num;
+}
+
+static void am65_cpsw_get_pauseparam(struct net_device *ndev,
+                                    struct ethtool_pauseparam *pause)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       pause->autoneg = AUTONEG_DISABLE;
+       pause->rx_pause = salve->rx_pause ? true : false;
+       pause->tx_pause = salve->tx_pause ? true : false;
+}
+
+static int am65_cpsw_set_pauseparam(struct net_device *ndev,
+                                   struct ethtool_pauseparam *pause)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       if (!salve->phy)
+               return -EINVAL;
+
+       if (!phy_validate_pause(salve->phy, pause))
+               return -EINVAL;
+
+       salve->rx_pause = pause->rx_pause ? true : false;
+       salve->tx_pause = pause->tx_pause ? true : false;
+
+       phy_set_asym_pause(salve->phy, salve->rx_pause, salve->tx_pause);
+
+       return 0;
+}
+
+static void am65_cpsw_get_wol(struct net_device *ndev,
+                             struct ethtool_wolinfo *wol)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       wol->supported = 0;
+       wol->wolopts = 0;
+
+       if (salve->phy)
+               phy_ethtool_get_wol(salve->phy, wol);
+}
+
+static int am65_cpsw_set_wol(struct net_device *ndev,
+                            struct ethtool_wolinfo *wol)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       if (!salve->phy)
+               return -EOPNOTSUPP;
+
+       return phy_ethtool_set_wol(salve->phy, wol);
+}
+
+static int am65_cpsw_get_link_ksettings(struct net_device *ndev,
+                                       struct ethtool_link_ksettings *ecmd)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       if (!salve->phy)
+               return -EOPNOTSUPP;
+
+       phy_ethtool_ksettings_get(salve->phy, ecmd);
+       return 0;
+}
+
+static int
+am65_cpsw_set_link_ksettings(struct net_device *ndev,
+                            const struct ethtool_link_ksettings *ecmd)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       if (!salve->phy || phy_is_pseudo_fixed_link(salve->phy))
+               return -EOPNOTSUPP;
+
+       return phy_ethtool_ksettings_set(salve->phy, ecmd);
+}
+
+static int am65_cpsw_get_eee(struct net_device *ndev, struct ethtool_eee *edata)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       if (!salve->phy || phy_is_pseudo_fixed_link(salve->phy))
+               return -EOPNOTSUPP;
+
+       return phy_ethtool_get_eee(salve->phy, edata);
+}
+
+static int am65_cpsw_set_eee(struct net_device *ndev, struct ethtool_eee *edata)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       if (!salve->phy || phy_is_pseudo_fixed_link(salve->phy))
+               return -EOPNOTSUPP;
+
+       return phy_ethtool_set_eee(salve->phy, edata);
+}
+
+static int am65_cpsw_nway_reset(struct net_device *ndev)
+{
+       struct am65_cpsw_slave_data *salve = am65_ndev_to_slave(ndev);
+
+       if (!salve->phy || phy_is_pseudo_fixed_link(salve->phy))
+               return -EOPNOTSUPP;
+
+       return phy_restart_aneg(salve->phy);
+}
+
+static int am65_cpsw_get_regs_len(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       u32 i, regdump_len = 0;
+
+       for (i = 0; i < ARRAY_SIZE(am65_cpsw_regdump); i++) {
+               if (am65_cpsw_regdump[i].hdr.module_id ==
+                   AM65_CPSW_REGDUMP_MOD_CPSW_ALE_TBL) {
+                       regdump_len += sizeof(struct am65_cpsw_regdump_hdr);
+                       regdump_len += common->ale->params.ale_entries *
+                                      ALE_ENTRY_WORDS * sizeof(u32);
+                       continue;
+               }
+               regdump_len += am65_cpsw_regdump[i].hdr.len;
+       }
+
+       return regdump_len;
+}
+
+static void am65_cpsw_get_regs(struct net_device *ndev,
+                              struct ethtool_regs *regs, void *p)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       u32 i, j, pos, *reg = p;
+
+       /* update CPSW IP version */
+       regs->version = AM65_CPSW_REGDUMP_VER;
+
+       pos = 0;
+       for (i = 0; i < ARRAY_SIZE(am65_cpsw_regdump); i++) {
+               reg[pos++] = am65_cpsw_regdump[i].hdr.module_id;
+
+               if (am65_cpsw_regdump[i].hdr.module_id ==
+                   AM65_CPSW_REGDUMP_MOD_CPSW_ALE_TBL) {
+                       u32 ale_tbl_len = common->ale->params.ale_entries *
+                                         ALE_ENTRY_WORDS * sizeof(u32) +
+                                         sizeof(struct am65_cpsw_regdump_hdr);
+                       reg[pos++] = ale_tbl_len;
+                       cpsw_ale_dump(common->ale, ®[pos]);
+                       pos += ale_tbl_len;
+                       continue;
+               }
+
+               reg[pos++] = am65_cpsw_regdump[i].hdr.len;
+
+               j = am65_cpsw_regdump[i].start_ofs;
+               do {
+                       reg[pos++] = j;
+                       reg[pos++] = readl_relaxed(common->ss_base + j);
+                       j += sizeof(u32);
+               } while (j <= am65_cpsw_regdump[i].end_ofs);
+       }
+}
+
+static int am65_cpsw_get_sset_count(struct net_device *ndev, int sset)
+{
+       switch (sset) {
+       case ETH_SS_STATS:
+               return ARRAY_SIZE(am65_host_stats) +
+                      ARRAY_SIZE(am65_slave_stats);
+       case ETH_SS_PRIV_FLAGS:
+               return ARRAY_SIZE(am65_cpsw_ethtool_priv_flags);
+       default:
+               return -EOPNOTSUPP;
+       }
+}
+
+static void am65_cpsw_get_strings(struct net_device *ndev,
+                                 u32 stringset, u8 *data)
+{
+       const struct am65_cpsw_ethtool_stat *hw_stats;
+       u32 i, num_stats;
+       u8 *p = data;
+
+       switch (stringset) {
+       case ETH_SS_STATS:
+               num_stats = ARRAY_SIZE(am65_host_stats);
+               hw_stats = am65_host_stats;
+               for (i = 0; i < num_stats; i++) {
+                       memcpy(p, hw_stats[i].desc, ETH_GSTRING_LEN);
+                       p += ETH_GSTRING_LEN;
+               }
+
+               num_stats = ARRAY_SIZE(am65_slave_stats);
+               hw_stats = am65_slave_stats;
+               for (i = 0; i < num_stats; i++) {
+                       memcpy(p, hw_stats[i].desc, ETH_GSTRING_LEN);
+                       p += ETH_GSTRING_LEN;
+               }
+               break;
+       case ETH_SS_PRIV_FLAGS:
+               num_stats = ARRAY_SIZE(am65_cpsw_ethtool_priv_flags);
+
+               for (i = 0; i < num_stats; i++) {
+                       memcpy(p, am65_cpsw_ethtool_priv_flags[i],
+                              ETH_GSTRING_LEN);
+                       p += ETH_GSTRING_LEN;
+               }
+               break;
+       }
+}
+
+static void am65_cpsw_get_ethtool_stats(struct net_device *ndev,
+                                       struct ethtool_stats *stats, u64 *data)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       const struct am65_cpsw_ethtool_stat *hw_stats;
+       struct am65_cpsw_host *host_p;
+       struct am65_cpsw_port *port;
+       u32 i, num_stats;
+
+       host_p = am65_common_get_host(common);
+       port = am65_ndev_to_port(ndev);
+       num_stats = ARRAY_SIZE(am65_host_stats);
+       hw_stats = am65_host_stats;
+       for (i = 0; i < num_stats; i++)
+               *data++ = readl_relaxed(host_p->stat_base +
+                                       hw_stats[i].offset);
+
+       num_stats = ARRAY_SIZE(am65_slave_stats);
+       hw_stats = am65_slave_stats;
+       for (i = 0; i < num_stats; i++)
+               *data++ = readl_relaxed(port->stat_base +
+                                       hw_stats[i].offset);
+}
+
+static u32 am65_cpsw_get_ethtool_priv_flags(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       u32 priv_flags = 0;
+
+       if (common->pf_p0_rx_ptype_rrobin)
+               priv_flags |= AM65_CPSW_PRIV_P0_RX_PTYPE_RROBIN;
+
+       return priv_flags;
+}
+
+static int am65_cpsw_set_ethtool_priv_flags(struct net_device *ndev, u32 flags)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+
+       common->pf_p0_rx_ptype_rrobin =
+                       !!(flags & AM65_CPSW_PRIV_P0_RX_PTYPE_RROBIN);
+       am65_cpsw_nuss_set_p0_ptype(common);
+
+       return 0;
+}
+
+const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
+       .begin                  = am65_cpsw_ethtool_op_begin,
+       .complete               = am65_cpsw_ethtool_op_complete,
+       .get_drvinfo            = am65_cpsw_get_drvinfo,
+       .get_msglevel           = am65_cpsw_get_msglevel,
+       .set_msglevel           = am65_cpsw_set_msglevel,
+       .get_channels           = am65_cpsw_get_channels,
+       .set_channels           = am65_cpsw_set_channels,
+       .get_ringparam          = am65_cpsw_get_ringparam,
+       .get_regs_len           = am65_cpsw_get_regs_len,
+       .get_regs               = am65_cpsw_get_regs,
+       .get_sset_count         = am65_cpsw_get_sset_count,
+       .get_strings            = am65_cpsw_get_strings,
+       .get_ethtool_stats      = am65_cpsw_get_ethtool_stats,
+       .get_ts_info            = ethtool_op_get_ts_info,
+       .get_priv_flags         = am65_cpsw_get_ethtool_priv_flags,
+       .set_priv_flags         = am65_cpsw_set_ethtool_priv_flags,
+
+       .get_link               = ethtool_op_get_link,
+       .get_link_ksettings     = am65_cpsw_get_link_ksettings,
+       .set_link_ksettings     = am65_cpsw_set_link_ksettings,
+       .get_pauseparam         = am65_cpsw_get_pauseparam,
+       .set_pauseparam         = am65_cpsw_set_pauseparam,
+       .get_wol                = am65_cpsw_get_wol,
+       .set_wol                = am65_cpsw_set_wol,
+       .get_eee                = am65_cpsw_get_eee,
+       .set_eee                = am65_cpsw_set_eee,
+       .nway_reset             = am65_cpsw_nway_reset,
+};
 
--- /dev/null
+// SPDX-License-Identifier: GPL-2.0
+/* Texas Instruments K3 AM65 Ethernet Switch SubSystem Driver
+ *
+ * Copyright (C) 2020 Texas Instruments Incorporated - http://www.ti.com/
+ *
+ */
+
+#include <linux/etherdevice.h>
+#include <linux/if_vlan.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/kmemleak.h>
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/net_tstamp.h>
+#include <linux/of.h>
+#include <linux/of_mdio.h>
+#include <linux/of_net.h>
+#include <linux/of_device.h>
+#include <linux/phy.h>
+#include <linux/phy/phy.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+#include <linux/regmap.h>
+#include <linux/mfd/syscon.h>
+#include <linux/dma/ti-cppi5.h>
+#include <linux/dma/k3-udma-glue.h>
+
+#include "cpsw_ale.h"
+#include "cpsw_sl.h"
+#include "am65-cpsw-nuss.h"
+#include "k3-cppi-desc-pool.h"
+
+#define AM65_CPSW_SS_BASE      0x0
+#define AM65_CPSW_SGMII_BASE   0x100
+#define AM65_CPSW_XGMII_BASE   0x2100
+#define AM65_CPSW_CPSW_NU_BASE 0x20000
+#define AM65_CPSW_NU_PORTS_BASE        0x1000
+#define AM65_CPSW_NU_STATS_BASE        0x1a000
+#define AM65_CPSW_NU_ALE_BASE  0x1e000
+#define AM65_CPSW_NU_CPTS_BASE 0x1d000
+
+#define AM65_CPSW_NU_PORTS_OFFSET      0x1000
+#define AM65_CPSW_NU_STATS_PORT_OFFSET 0x200
+
+#define AM65_CPSW_MAX_PORTS    8
+
+#define AM65_CPSW_MIN_PACKET_SIZE      VLAN_ETH_ZLEN
+#define AM65_CPSW_MAX_PACKET_SIZE      (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)
+
+#define AM65_CPSW_REG_CTL              0x004
+#define AM65_CPSW_REG_STAT_PORT_EN     0x014
+#define AM65_CPSW_REG_PTYPE            0x018
+
+#define AM65_CPSW_P0_REG_CTL                   0x004
+#define AM65_CPSW_PORT0_REG_FLOW_ID_OFFSET     0x008
+
+#define AM65_CPSW_PORT_REG_PRI_CTL             0x01c
+#define AM65_CPSW_PORT_REG_RX_PRI_MAP          0x020
+#define AM65_CPSW_PORT_REG_RX_MAXLEN           0x024
+
+#define AM65_CPSW_PORTN_REG_SA_L               0x308
+#define AM65_CPSW_PORTN_REG_SA_H               0x30c
+#define AM65_CPSW_PORTN_REG_TS_CTL              0x310
+#define AM65_CPSW_PORTN_REG_TS_SEQ_LTYPE_REG   0x314
+#define AM65_CPSW_PORTN_REG_TS_VLAN_LTYPE_REG  0x318
+#define AM65_CPSW_PORTN_REG_TS_CTL_LTYPE2       0x31C
+
+#define AM65_CPSW_CTL_VLAN_AWARE               BIT(1)
+#define AM65_CPSW_CTL_P0_ENABLE                        BIT(2)
+#define AM65_CPSW_CTL_P0_TX_CRC_REMOVE         BIT(13)
+#define AM65_CPSW_CTL_P0_RX_PAD                        BIT(14)
+
+/* AM65_CPSW_P0_REG_CTL */
+#define AM65_CPSW_P0_REG_CTL_RX_CHECKSUM_EN    BIT(0)
+
+/* AM65_CPSW_PORT_REG_PRI_CTL */
+#define AM65_CPSW_PORT_REG_PRI_CTL_RX_PTYPE_RROBIN     BIT(8)
+
+/* AM65_CPSW_PN_TS_CTL register fields */
+#define AM65_CPSW_PN_TS_CTL_TX_ANX_F_EN                BIT(4)
+#define AM65_CPSW_PN_TS_CTL_TX_VLAN_LT1_EN     BIT(5)
+#define AM65_CPSW_PN_TS_CTL_TX_VLAN_LT2_EN     BIT(6)
+#define AM65_CPSW_PN_TS_CTL_TX_ANX_D_EN                BIT(7)
+#define AM65_CPSW_PN_TS_CTL_TX_ANX_E_EN                BIT(10)
+#define AM65_CPSW_PN_TS_CTL_TX_HOST_TS_EN      BIT(11)
+#define AM65_CPSW_PN_TS_CTL_MSG_TYPE_EN_SHIFT  16
+
+/* AM65_CPSW_PORTN_REG_TS_SEQ_LTYPE_REG register fields */
+#define AM65_CPSW_PN_TS_SEQ_ID_OFFSET_SHIFT    16
+
+/* AM65_CPSW_PORTN_REG_TS_CTL_LTYPE2 */
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_107      BIT(16)
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_129      BIT(17)
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_130      BIT(18)
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_131      BIT(19)
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_132      BIT(20)
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_319      BIT(21)
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_320      BIT(22)
+#define AM65_CPSW_PN_TS_CTL_LTYPE2_TS_TTL_NONZERO BIT(23)
+
+/* The PTP event messages - Sync, Delay_Req, Pdelay_Req, and Pdelay_Resp. */
+#define AM65_CPSW_TS_EVENT_MSG_TYPE_BITS (BIT(0) | BIT(1) | BIT(2) | BIT(3))
+
+#define AM65_CPSW_TS_SEQ_ID_OFFSET (0x1e)
+
+#define AM65_CPSW_TS_TX_ANX_ALL_EN             \
+       (AM65_CPSW_PN_TS_CTL_TX_ANX_D_EN |      \
+        AM65_CPSW_PN_TS_CTL_TX_ANX_E_EN |      \
+        AM65_CPSW_PN_TS_CTL_TX_ANX_F_EN)
+
+#define AM65_CPSW_ALE_AGEOUT_DEFAULT   30
+/* Number of TX/RX descriptors */
+#define AM65_CPSW_MAX_TX_DESC  500
+#define AM65_CPSW_MAX_RX_DESC  500
+
+#define AM65_CPSW_NAV_PS_DATA_SIZE 16
+#define AM65_CPSW_NAV_SW_DATA_SIZE 16
+
+#define AM65_CPSW_DEBUG        (NETIF_MSG_HW | NETIF_MSG_DRV | NETIF_MSG_LINK | \
+                        NETIF_MSG_IFUP | NETIF_MSG_PROBE | NETIF_MSG_IFDOWN | \
+                        NETIF_MSG_RX_ERR | NETIF_MSG_TX_ERR)
+
+static void am65_cpsw_port_set_sl_mac(struct am65_cpsw_port *slave,
+                                     const u8 *dev_addr)
+{
+       u32 mac_hi = (dev_addr[0] << 0) | (dev_addr[1] << 8) |
+                    (dev_addr[2] << 16) | (dev_addr[3] << 24);
+       u32 mac_lo = (dev_addr[4] << 0) | (dev_addr[5] << 8);
+
+       writel(mac_hi, slave->port_base + AM65_CPSW_PORTN_REG_SA_H);
+       writel(mac_lo, slave->port_base + AM65_CPSW_PORTN_REG_SA_L);
+}
+
+static void am65_cpsw_sl_ctl_reset(struct am65_cpsw_port *port)
+{
+       cpsw_sl_reset(port->slave.mac_sl, 100);
+       /* Max length register has to be restored after MAC SL reset */
+       writel(AM65_CPSW_MAX_PACKET_SIZE,
+              port->port_base + AM65_CPSW_PORT_REG_RX_MAXLEN);
+}
+
+static void am65_cpsw_nuss_get_ver(struct am65_cpsw_common *common)
+{
+       common->nuss_ver = readl(common->ss_base);
+       common->cpsw_ver = readl(common->cpsw_base);
+       dev_info(common->dev,
+                "initializing am65 cpsw nuss version 0x%08X, cpsw version 0x%08X Ports: %u\n",
+               common->nuss_ver,
+               common->cpsw_ver,
+               common->port_num + 1);
+}
+
+void am65_cpsw_nuss_adjust_link(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+       struct phy_device *phy = port->slave.phy;
+       u32 mac_control = 0;
+
+       if (!phy)
+               return;
+
+       if (phy->link) {
+               mac_control = CPSW_SL_CTL_GMII_EN;
+
+               if (phy->speed == 1000)
+                       mac_control |= CPSW_SL_CTL_GIG;
+               if (phy->speed == 10 && phy_interface_is_rgmii(phy))
+                       /* Can be used with in band mode only */
+                       mac_control |= CPSW_SL_CTL_EXT_EN;
+               if (phy->duplex)
+                       mac_control |= CPSW_SL_CTL_FULLDUPLEX;
+
+               /* RGMII speed is 100M if !CPSW_SL_CTL_GIG*/
+
+               /* rx_pause/tx_pause */
+               if (port->slave.rx_pause)
+                       mac_control |= CPSW_SL_CTL_RX_FLOW_EN;
+
+               if (port->slave.tx_pause)
+                       mac_control |= CPSW_SL_CTL_TX_FLOW_EN;
+
+               cpsw_sl_ctl_set(port->slave.mac_sl, mac_control);
+
+               /* enable forwarding */
+               cpsw_ale_control_set(common->ale, port->port_id,
+                                    ALE_PORT_STATE, ALE_PORT_STATE_FORWARD);
+
+               netif_tx_wake_all_queues(ndev);
+       } else {
+               int tmo;
+               /* disable forwarding */
+               cpsw_ale_control_set(common->ale, port->port_id,
+                                    ALE_PORT_STATE, ALE_PORT_STATE_DISABLE);
+
+               cpsw_sl_ctl_set(port->slave.mac_sl, CPSW_SL_CTL_CMD_IDLE);
+
+               tmo = cpsw_sl_wait_for_idle(port->slave.mac_sl, 100);
+               dev_dbg(common->dev, "donw msc_sl %08x tmo %d\n",
+                       cpsw_sl_reg_read(port->slave.mac_sl, CPSW_SL_MACSTATUS),
+                       tmo);
+
+               cpsw_sl_ctl_reset(port->slave.mac_sl);
+
+               netif_tx_stop_all_queues(ndev);
+       }
+
+       phy_print_status(phy);
+}
+
+static int am65_cpsw_nuss_ndo_slave_add_vid(struct net_device *ndev,
+                                           __be16 proto, u16 vid)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+       u32 port_mask, unreg_mcast = 0;
+       int ret;
+
+       ret = pm_runtime_get_sync(common->dev);
+       if (ret < 0) {
+               pm_runtime_put_noidle(common->dev);
+               return ret;
+       }
+
+       port_mask = BIT(port->port_id) | ALE_PORT_HOST;
+       if (!vid)
+               unreg_mcast = port_mask;
+       dev_info(common->dev, "Adding vlan %d to vlan filter\n", vid);
+       ret = cpsw_ale_add_vlan(common->ale, vid, port_mask,
+                               unreg_mcast, port_mask, 0);
+
+       pm_runtime_put(common->dev);
+       return ret;
+}
+
+static int am65_cpsw_nuss_ndo_slave_kill_vid(struct net_device *ndev,
+                                            __be16 proto, u16 vid)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       int ret;
+
+       ret = pm_runtime_get_sync(common->dev);
+       if (ret < 0) {
+               pm_runtime_put_noidle(common->dev);
+               return ret;
+       }
+
+       dev_info(common->dev, "Removing vlan %d from vlan filter\n", vid);
+       ret = cpsw_ale_del_vlan(common->ale, vid, 0);
+
+       pm_runtime_put(common->dev);
+       return ret;
+}
+
+static void am65_cpsw_slave_set_promisc_2g(struct am65_cpsw_port *port,
+                                          bool promisc)
+{
+       struct am65_cpsw_common *common = port->common;
+
+       if (promisc) {
+               /* Enable promiscuous mode */
+               cpsw_ale_control_set(common->ale, port->port_id,
+                                    ALE_PORT_MACONLY_CAF, 1);
+               dev_dbg(common->dev, "promisc enabled\n");
+       } else {
+               /* Disable promiscuous mode */
+               cpsw_ale_control_set(common->ale, port->port_id,
+                                    ALE_PORT_MACONLY_CAF, 0);
+               dev_dbg(common->dev, "promisc disabled\n");
+       }
+}
+
+static void am65_cpsw_nuss_ndo_slave_set_rx_mode(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+       u32 port_mask;
+       bool promisc;
+
+       promisc = !!(ndev->flags & IFF_PROMISC);
+       am65_cpsw_slave_set_promisc_2g(port, promisc);
+
+       if (promisc)
+               return;
+
+       /* Restore allmulti on vlans if necessary */
+       cpsw_ale_set_allmulti(common->ale,
+                             ndev->flags & IFF_ALLMULTI, port->port_id);
+
+       port_mask = ALE_PORT_HOST;
+       /* Clear all mcast from ALE */
+       cpsw_ale_flush_multicast(common->ale, port_mask, -1);
+
+       if (!netdev_mc_empty(ndev)) {
+               struct netdev_hw_addr *ha;
+
+               /* program multicast address list into ALE register */
+               netdev_for_each_mc_addr(ha, ndev) {
+                       cpsw_ale_add_mcast(common->ale, ha->addr,
+                                          port_mask, 0, 0, 0);
+               }
+       }
+}
+
+static void am65_cpsw_nuss_ndo_host_tx_timeout(struct net_device *ndev,
+                                              unsigned int txqueue)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct am65_cpsw_tx_chn *tx_chn;
+       struct netdev_queue *netif_txq;
+       unsigned long trans_start;
+
+       netif_txq = netdev_get_tx_queue(ndev, txqueue);
+       tx_chn = &common->tx_chns[txqueue];
+       trans_start = netif_txq->trans_start;
+
+       netdev_err(ndev, "txq:%d DRV_XOFF:%d tmo:%u dql_avail:%d free_desc:%zu\n",
+                  txqueue,
+                  netif_tx_queue_stopped(netif_txq),
+                  jiffies_to_msecs(jiffies - trans_start),
+                  dql_avail(&netif_txq->dql),
+                  k3_cppi_desc_pool_avail(tx_chn->desc_pool));
+
+       if (netif_tx_queue_stopped(netif_txq)) {
+               /* try recover if stopped by us */
+               txq_trans_update(netif_txq);
+               netif_tx_wake_queue(netif_txq);
+       }
+}
+
+static int am65_cpsw_nuss_rx_push(struct am65_cpsw_common *common,
+                                 struct sk_buff *skb)
+{
+       struct am65_cpsw_rx_chn *rx_chn = &common->rx_chns;
+       struct cppi5_host_desc_t *desc_rx;
+       struct device *dev = common->dev;
+       u32 pkt_len = skb_tailroom(skb);
+       dma_addr_t desc_dma;
+       dma_addr_t buf_dma;
+       void *swdata;
+
+       desc_rx = k3_cppi_desc_pool_alloc(rx_chn->desc_pool);
+       if (!desc_rx) {
+               dev_err(dev, "Failed to allocate RXFDQ descriptor\n");
+               return -ENOMEM;
+       }
+       desc_dma = k3_cppi_desc_pool_virt2dma(rx_chn->desc_pool, desc_rx);
+
+       buf_dma = dma_map_single(dev, skb->data, pkt_len, DMA_FROM_DEVICE);
+       if (unlikely(dma_mapping_error(dev, buf_dma))) {
+               k3_cppi_desc_pool_free(rx_chn->desc_pool, desc_rx);
+               dev_err(dev, "Failed to map rx skb buffer\n");
+               return -EINVAL;
+       }
+
+       cppi5_hdesc_init(desc_rx, CPPI5_INFO0_HDESC_EPIB_PRESENT,
+                        AM65_CPSW_NAV_PS_DATA_SIZE);
+       cppi5_hdesc_attach_buf(desc_rx, 0, 0, buf_dma, skb_tailroom(skb));
+       swdata = cppi5_hdesc_get_swdata(desc_rx);
+       *((void **)swdata) = skb;
+
+       return k3_udma_glue_push_rx_chn(rx_chn->rx_chn, 0, desc_rx, desc_dma);
+}
+
+void am65_cpsw_nuss_set_p0_ptype(struct am65_cpsw_common *common)
+{
+       struct am65_cpsw_host *host_p = am65_common_get_host(common);
+       u32 val, pri_map;
+
+       /* P0 set Receive Priority Type */
+       val = readl(host_p->port_base + AM65_CPSW_PORT_REG_PRI_CTL);
+
+       if (common->pf_p0_rx_ptype_rrobin) {
+               val |= AM65_CPSW_PORT_REG_PRI_CTL_RX_PTYPE_RROBIN;
+               /* Enet Ports fifos works in fixed priority mode only, so
+                * reset P0_Rx_Pri_Map so all packet will go in Enet fifo 0
+                */
+               pri_map = 0x0;
+       } else {
+               val &= ~AM65_CPSW_PORT_REG_PRI_CTL_RX_PTYPE_RROBIN;
+               /* restore P0_Rx_Pri_Map */
+               pri_map = 0x76543210;
+       }
+
+       writel(pri_map, host_p->port_base + AM65_CPSW_PORT_REG_RX_PRI_MAP);
+       writel(val, host_p->port_base + AM65_CPSW_PORT_REG_PRI_CTL);
+}
+
+static int am65_cpsw_nuss_common_open(struct am65_cpsw_common *common,
+                                     netdev_features_t features)
+{
+       struct am65_cpsw_host *host_p = am65_common_get_host(common);
+       int port_idx, i, ret;
+       struct sk_buff *skb;
+       u32 val, port_mask;
+
+       if (common->usage_count)
+               return 0;
+
+       /* Control register */
+       writel(AM65_CPSW_CTL_P0_ENABLE | AM65_CPSW_CTL_P0_TX_CRC_REMOVE |
+              AM65_CPSW_CTL_VLAN_AWARE | AM65_CPSW_CTL_P0_RX_PAD,
+              common->cpsw_base + AM65_CPSW_REG_CTL);
+       /* Max length register */
+       writel(AM65_CPSW_MAX_PACKET_SIZE,
+              host_p->port_base + AM65_CPSW_PORT_REG_RX_MAXLEN);
+       /* set base flow_id */
+       writel(common->rx_flow_id_base,
+              host_p->port_base + AM65_CPSW_PORT0_REG_FLOW_ID_OFFSET);
+       /* en tx crc offload */
+       if (features & NETIF_F_HW_CSUM)
+               writel(AM65_CPSW_P0_REG_CTL_RX_CHECKSUM_EN,
+                      host_p->port_base + AM65_CPSW_P0_REG_CTL);
+
+       am65_cpsw_nuss_set_p0_ptype(common);
+
+       /* enable statistic */
+       val = BIT(HOST_PORT_NUM);
+       for (port_idx = 0; port_idx < common->port_num; port_idx++) {
+               struct am65_cpsw_port *port = &common->ports[port_idx];
+
+               if (!port->disabled)
+                       val |=  BIT(port->port_id);
+       }
+       writel(val, common->cpsw_base + AM65_CPSW_REG_STAT_PORT_EN);
+
+       /* disable priority elevation */
+       writel(0, common->cpsw_base + AM65_CPSW_REG_PTYPE);
+
+       cpsw_ale_start(common->ale);
+
+       /* limit to one RX flow only */
+       cpsw_ale_control_set(common->ale, HOST_PORT_NUM,
+                            ALE_DEFAULT_THREAD_ID, 0);
+       cpsw_ale_control_set(common->ale, HOST_PORT_NUM,
+                            ALE_DEFAULT_THREAD_ENABLE, 1);
+       if (AM65_CPSW_IS_CPSW2G(common))
+               cpsw_ale_control_set(common->ale, HOST_PORT_NUM,
+                                    ALE_PORT_NOLEARN, 1);
+       /* switch to vlan unaware mode */
+       cpsw_ale_control_set(common->ale, HOST_PORT_NUM, ALE_VLAN_AWARE, 1);
+       cpsw_ale_control_set(common->ale, HOST_PORT_NUM,
+                            ALE_PORT_STATE, ALE_PORT_STATE_FORWARD);
+
+       /* default vlan cfg: create mask based on enabled ports */
+       port_mask = GENMASK(common->port_num, 0) &
+                   ~common->disabled_ports_mask;
+
+       cpsw_ale_add_vlan(common->ale, 0, port_mask,
+                         port_mask, port_mask,
+                         port_mask & ~ALE_PORT_HOST);
+
+       for (i = 0; i < common->rx_chns.descs_num; i++) {
+               skb = __netdev_alloc_skb_ip_align(NULL,
+                                                 AM65_CPSW_MAX_PACKET_SIZE,
+                                                 GFP_KERNEL);
+               if (!skb) {
+                       dev_err(common->dev, "cannot allocate skb\n");
+                       return -ENOMEM;
+               }
+
+               ret = am65_cpsw_nuss_rx_push(common, skb);
+               if (ret < 0) {
+                       dev_err(common->dev,
+                               "cannot submit skb to channel rx, error %d\n",
+                               ret);
+                       kfree_skb(skb);
+                       return ret;
+               }
+               kmemleak_not_leak(skb);
+       }
+       k3_udma_glue_enable_rx_chn(common->rx_chns.rx_chn);
+
+       for (i = 0; i < common->tx_ch_num; i++) {
+               ret = k3_udma_glue_enable_tx_chn(common->tx_chns[i].tx_chn);
+               if (ret)
+                       return ret;
+               napi_enable(&common->tx_chns[i].napi_tx);
+       }
+
+       napi_enable(&common->napi_rx);
+
+       dev_dbg(common->dev, "cpsw_nuss started\n");
+       return 0;
+}
+
+static void am65_cpsw_nuss_tx_cleanup(void *data, dma_addr_t desc_dma);
+static void am65_cpsw_nuss_rx_cleanup(void *data, dma_addr_t desc_dma);
+
+static int am65_cpsw_nuss_common_stop(struct am65_cpsw_common *common)
+{
+       int i;
+
+       if (common->usage_count != 1)
+               return 0;
+
+       cpsw_ale_control_set(common->ale, HOST_PORT_NUM,
+                            ALE_PORT_STATE, ALE_PORT_STATE_DISABLE);
+
+       /* shutdown tx channels */
+       atomic_set(&common->tdown_cnt, common->tx_ch_num);
+       /* ensure new tdown_cnt value is visible */
+       smp_mb__after_atomic();
+       reinit_completion(&common->tdown_complete);
+
+       for (i = 0; i < common->tx_ch_num; i++)
+               k3_udma_glue_tdown_tx_chn(common->tx_chns[i].tx_chn, false);
+
+       i = wait_for_completion_timeout(&common->tdown_complete,
+                                       msecs_to_jiffies(1000));
+       if (!i)
+               dev_err(common->dev, "tx timeout\n");
+       for (i = 0; i < common->tx_ch_num; i++)
+               napi_disable(&common->tx_chns[i].napi_tx);
+
+       for (i = 0; i < common->tx_ch_num; i++) {
+               k3_udma_glue_reset_tx_chn(common->tx_chns[i].tx_chn,
+                                         &common->tx_chns[i],
+                                         am65_cpsw_nuss_tx_cleanup);
+               k3_udma_glue_disable_tx_chn(common->tx_chns[i].tx_chn);
+       }
+
+       k3_udma_glue_tdown_rx_chn(common->rx_chns.rx_chn, true);
+       napi_disable(&common->napi_rx);
+
+       for (i = 0; i < AM65_CPSW_MAX_RX_FLOWS; i++)
+               k3_udma_glue_reset_rx_chn(common->rx_chns.rx_chn, i,
+                                         &common->rx_chns,
+                                         am65_cpsw_nuss_rx_cleanup, !!i);
+
+       k3_udma_glue_disable_rx_chn(common->rx_chns.rx_chn);
+
+       cpsw_ale_stop(common->ale);
+
+       writel(0, common->cpsw_base + AM65_CPSW_REG_CTL);
+       writel(0, common->cpsw_base + AM65_CPSW_REG_STAT_PORT_EN);
+
+       dev_dbg(common->dev, "cpsw_nuss stopped\n");
+       return 0;
+}
+
+static int am65_cpsw_nuss_ndo_slave_stop(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+       int ret;
+
+       if (port->slave.phy)
+               phy_stop(port->slave.phy);
+
+       netif_tx_stop_all_queues(ndev);
+
+       if (port->slave.phy) {
+               phy_disconnect(port->slave.phy);
+               port->slave.phy = NULL;
+       }
+
+       ret = am65_cpsw_nuss_common_stop(common);
+       if (ret)
+               return ret;
+
+       common->usage_count--;
+       pm_runtime_put(common->dev);
+       return 0;
+}
+
+static int am65_cpsw_nuss_ndo_slave_open(struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+       u32 port_mask;
+       int ret, i;
+
+       ret = pm_runtime_get_sync(common->dev);
+       if (ret < 0) {
+               pm_runtime_put_noidle(common->dev);
+               return ret;
+       }
+
+       /* Notify the stack of the actual queue counts. */
+       ret = netif_set_real_num_tx_queues(ndev, common->tx_ch_num);
+       if (ret) {
+               dev_err(common->dev, "cannot set real number of tx queues\n");
+               return ret;
+       }
+
+       ret = netif_set_real_num_rx_queues(ndev, AM65_CPSW_MAX_RX_QUEUES);
+       if (ret) {
+               dev_err(common->dev, "cannot set real number of rx queues\n");
+               return ret;
+       }
+
+       for (i = 0; i < common->tx_ch_num; i++)
+               netdev_tx_reset_queue(netdev_get_tx_queue(ndev, i));
+
+       ret = am65_cpsw_nuss_common_open(common, ndev->features);
+       if (ret)
+               return ret;
+
+       common->usage_count++;
+
+       am65_cpsw_port_set_sl_mac(port, ndev->dev_addr);
+
+       if (port->slave.mac_only)
+               /* enable mac-only mode on port */
+               cpsw_ale_control_set(common->ale, port->port_id,
+                                    ALE_PORT_MACONLY, 1);
+       if (AM65_CPSW_IS_CPSW2G(common))
+               cpsw_ale_control_set(common->ale, port->port_id,
+                                    ALE_PORT_NOLEARN, 1);
+
+       port_mask = BIT(port->port_id) | ALE_PORT_HOST;
+       cpsw_ale_add_ucast(common->ale, ndev->dev_addr,
+                          HOST_PORT_NUM, ALE_SECURE, 0);
+       cpsw_ale_add_mcast(common->ale, ndev->broadcast,
+                          port_mask, 0, 0, ALE_MCAST_FWD_2);
+
+       /* mac_sl should be configured via phy-link interface */
+       am65_cpsw_sl_ctl_reset(port);
+
+       ret = phy_set_mode_ext(port->slave.ifphy, PHY_MODE_ETHERNET,
+                              port->slave.phy_if);
+       if (ret)
+               goto error_cleanup;
+
+       if (port->slave.phy_node) {
+               port->slave.phy = of_phy_connect(ndev,
+                                                port->slave.phy_node,
+                                                &am65_cpsw_nuss_adjust_link,
+                                                0, port->slave.phy_if);
+               if (!port->slave.phy) {
+                       dev_err(common->dev, "phy %pOF not found on slave %d\n",
+                               port->slave.phy_node,
+                               port->port_id);
+                       ret = -ENODEV;
+                       goto error_cleanup;
+               }
+       }
+
+       phy_attached_info(port->slave.phy);
+       phy_start(port->slave.phy);
+
+       return 0;
+
+error_cleanup:
+       am65_cpsw_nuss_ndo_slave_stop(ndev);
+       return ret;
+}
+
+static void am65_cpsw_nuss_rx_cleanup(void *data, dma_addr_t desc_dma)
+{
+       struct am65_cpsw_rx_chn *rx_chn = data;
+       struct cppi5_host_desc_t *desc_rx;
+       struct sk_buff *skb;
+       dma_addr_t buf_dma;
+       u32 buf_dma_len;
+       void **swdata;
+
+       desc_rx = k3_cppi_desc_pool_dma2virt(rx_chn->desc_pool, desc_dma);
+       swdata = cppi5_hdesc_get_swdata(desc_rx);
+       skb = *swdata;
+       cppi5_hdesc_get_obuf(desc_rx, &buf_dma, &buf_dma_len);
+
+       dma_unmap_single(rx_chn->dev, buf_dma, buf_dma_len, DMA_FROM_DEVICE);
+       k3_cppi_desc_pool_free(rx_chn->desc_pool, desc_rx);
+
+       dev_kfree_skb_any(skb);
+}
+
+/* RX psdata[2] word format - checksum information */
+#define AM65_CPSW_RX_PSD_CSUM_ADD      GENMASK(15, 0)
+#define AM65_CPSW_RX_PSD_CSUM_ERR      BIT(16)
+#define AM65_CPSW_RX_PSD_IS_FRAGMENT   BIT(17)
+#define AM65_CPSW_RX_PSD_IS_TCP                BIT(18)
+#define AM65_CPSW_RX_PSD_IPV6_VALID    BIT(19)
+#define AM65_CPSW_RX_PSD_IPV4_VALID    BIT(20)
+
+static void am65_cpsw_nuss_rx_csum(struct sk_buff *skb, u32 csum_info)
+{
+       /* HW can verify IPv4/IPv6 TCP/UDP packets checksum
+        * csum information provides in psdata[2] word:
+        * AM65_CPSW_RX_PSD_CSUM_ERR bit - indicates csum error
+        * AM65_CPSW_RX_PSD_IPV6_VALID and AM65_CPSW_RX_PSD_IPV4_VALID
+        * bits - indicates IPv4/IPv6 packet
+        * AM65_CPSW_RX_PSD_IS_FRAGMENT bit - indicates fragmented packet
+        * AM65_CPSW_RX_PSD_CSUM_ADD has value 0xFFFF for non fragmented packets
+        * or csum value for fragmented packets if !AM65_CPSW_RX_PSD_CSUM_ERR
+        */
+       skb_checksum_none_assert(skb);
+
+       if (unlikely(!(skb->dev->features & NETIF_F_RXCSUM)))
+               return;
+
+       if ((csum_info & (AM65_CPSW_RX_PSD_IPV6_VALID |
+                         AM65_CPSW_RX_PSD_IPV4_VALID)) &&
+                         !(csum_info & AM65_CPSW_RX_PSD_CSUM_ERR)) {
+               /* csum for fragmented packets is unsupported */
+               if (!(csum_info & AM65_CPSW_RX_PSD_IS_FRAGMENT))
+                       skb->ip_summed = CHECKSUM_UNNECESSARY;
+       }
+}
+
+static int am65_cpsw_nuss_rx_packets(struct am65_cpsw_common *common,
+                                    u32 flow_idx)
+{
+       struct am65_cpsw_rx_chn *rx_chn = &common->rx_chns;
+       u32 buf_dma_len, pkt_len, port_id = 0, csum_info;
+       struct am65_cpsw_ndev_priv *ndev_priv;
+       struct am65_cpsw_ndev_stats *stats;
+       struct cppi5_host_desc_t *desc_rx;
+       struct device *dev = common->dev;
+       struct sk_buff *skb, *new_skb;
+       dma_addr_t desc_dma, buf_dma;
+       struct am65_cpsw_port *port;
+       struct net_device *ndev;
+       void **swdata;
+       u32 *psdata;
+       int ret = 0;
+
+       ret = k3_udma_glue_pop_rx_chn(rx_chn->rx_chn, flow_idx, &desc_dma);
+       if (ret) {
+               if (ret != -ENODATA)
+                       dev_err(dev, "RX: pop chn fail %d\n", ret);
+               return ret;
+       }
+
+       if (desc_dma & 0x1) {
+               dev_dbg(dev, "%s RX tdown flow: %u\n", __func__, flow_idx);
+               return 0;
+       }
+
+       desc_rx = k3_cppi_desc_pool_dma2virt(rx_chn->desc_pool, desc_dma);
+       dev_dbg(dev, "%s flow_idx: %u desc %pad\n",
+               __func__, flow_idx, &desc_dma);
+
+       swdata = cppi5_hdesc_get_swdata(desc_rx);
+       skb = *swdata;
+       cppi5_hdesc_get_obuf(desc_rx, &buf_dma, &buf_dma_len);
+       pkt_len = cppi5_hdesc_get_pktlen(desc_rx);
+       cppi5_desc_get_tags_ids(&desc_rx->hdr, &port_id, NULL);
+       dev_dbg(dev, "%s rx port_id:%d\n", __func__, port_id);
+       port = am65_common_get_port(common, port_id);
+       ndev = port->ndev;
+       skb->dev = ndev;
+
+       psdata = cppi5_hdesc_get_psdata(desc_rx);
+       csum_info = psdata[2];
+       dev_dbg(dev, "%s rx csum_info:%#x\n", __func__, csum_info);
+
+       dma_unmap_single(dev, buf_dma, buf_dma_len, DMA_FROM_DEVICE);
+
+       k3_cppi_desc_pool_free(rx_chn->desc_pool, desc_rx);
+
+       new_skb = netdev_alloc_skb_ip_align(ndev, AM65_CPSW_MAX_PACKET_SIZE);
+       if (new_skb) {
+               skb_put(skb, pkt_len);
+               skb->protocol = eth_type_trans(skb, ndev);
+               am65_cpsw_nuss_rx_csum(skb, csum_info);
+               napi_gro_receive(&common->napi_rx, skb);
+
+               ndev_priv = netdev_priv(ndev);
+               stats = this_cpu_ptr(ndev_priv->stats);
+
+               u64_stats_update_begin(&stats->syncp);
+               stats->rx_packets++;
+               stats->rx_bytes += pkt_len;
+               u64_stats_update_end(&stats->syncp);
+               kmemleak_not_leak(new_skb);
+       } else {
+               ndev->stats.rx_dropped++;
+               new_skb = skb;
+       }
+
+       if (netif_dormant(ndev)) {
+               dev_kfree_skb_any(new_skb);
+               ndev->stats.rx_dropped++;
+               return 0;
+       }
+
+       ret = am65_cpsw_nuss_rx_push(common, new_skb);
+       if (WARN_ON(ret < 0)) {
+               dev_kfree_skb_any(new_skb);
+               ndev->stats.rx_errors++;
+               ndev->stats.rx_dropped++;
+       }
+
+       return ret;
+}
+
+static int am65_cpsw_nuss_rx_poll(struct napi_struct *napi_rx, int budget)
+{
+       struct am65_cpsw_common *common = am65_cpsw_napi_to_common(napi_rx);
+       int flow = AM65_CPSW_MAX_RX_FLOWS;
+       int cur_budget, ret;
+       int num_rx = 0;
+
+       /* process every flow */
+       while (flow--) {
+               cur_budget = budget - num_rx;
+
+               while (cur_budget--) {
+                       ret = am65_cpsw_nuss_rx_packets(common, flow);
+                       if (ret)
+                               break;
+                       num_rx++;
+               }
+
+               if (num_rx >= budget)
+                       break;
+       }
+
+       dev_dbg(common->dev, "%s num_rx:%d %d\n", __func__, num_rx, budget);
+
+       if (num_rx < budget && napi_complete_done(napi_rx, num_rx))
+               enable_irq(common->rx_chns.irq);
+
+       return num_rx;
+}
+
+static void am65_cpsw_nuss_xmit_free(struct am65_cpsw_tx_chn *tx_chn,
+                                    struct device *dev,
+                                    struct cppi5_host_desc_t *desc)
+{
+       struct cppi5_host_desc_t *first_desc, *next_desc;
+       dma_addr_t buf_dma, next_desc_dma;
+       u32 buf_dma_len;
+
+       first_desc = desc;
+       next_desc = first_desc;
+
+       cppi5_hdesc_get_obuf(first_desc, &buf_dma, &buf_dma_len);
+
+       dma_unmap_single(dev, buf_dma, buf_dma_len,
+                        DMA_TO_DEVICE);
+
+       next_desc_dma = cppi5_hdesc_get_next_hbdesc(first_desc);
+       while (next_desc_dma) {
+               next_desc = k3_cppi_desc_pool_dma2virt(tx_chn->desc_pool,
+                                                      next_desc_dma);
+               cppi5_hdesc_get_obuf(next_desc, &buf_dma, &buf_dma_len);
+
+               dma_unmap_page(dev, buf_dma, buf_dma_len,
+                              DMA_TO_DEVICE);
+
+               next_desc_dma = cppi5_hdesc_get_next_hbdesc(next_desc);
+
+               k3_cppi_desc_pool_free(tx_chn->desc_pool, next_desc);
+       }
+
+       k3_cppi_desc_pool_free(tx_chn->desc_pool, first_desc);
+}
+
+static void am65_cpsw_nuss_tx_cleanup(void *data, dma_addr_t desc_dma)
+{
+       struct am65_cpsw_tx_chn *tx_chn = data;
+       struct cppi5_host_desc_t *desc_tx;
+       struct sk_buff *skb;
+       void **swdata;
+
+       desc_tx = k3_cppi_desc_pool_dma2virt(tx_chn->desc_pool, desc_dma);
+       swdata = cppi5_hdesc_get_swdata(desc_tx);
+       skb = *(swdata);
+       am65_cpsw_nuss_xmit_free(tx_chn, tx_chn->common->dev, desc_tx);
+
+       dev_kfree_skb_any(skb);
+}
+
+static int am65_cpsw_nuss_tx_compl_packets(struct am65_cpsw_common *common,
+                                          int chn, unsigned int budget)
+{
+       struct cppi5_host_desc_t *desc_tx;
+       struct device *dev = common->dev;
+       struct am65_cpsw_tx_chn *tx_chn;
+       struct netdev_queue *netif_txq;
+       unsigned int total_bytes = 0;
+       struct net_device *ndev;
+       struct sk_buff *skb;
+       dma_addr_t desc_dma;
+       int res, num_tx = 0;
+       void **swdata;
+
+       tx_chn = &common->tx_chns[chn];
+
+       while (true) {
+               struct am65_cpsw_ndev_priv *ndev_priv;
+               struct am65_cpsw_ndev_stats *stats;
+
+               res = k3_udma_glue_pop_tx_chn(tx_chn->tx_chn, &desc_dma);
+               if (res == -ENODATA)
+                       break;
+
+               if (desc_dma & 0x1) {
+                       if (atomic_dec_and_test(&common->tdown_cnt))
+                               complete(&common->tdown_complete);
+                       break;
+               }
+
+               desc_tx = k3_cppi_desc_pool_dma2virt(tx_chn->desc_pool,
+                                                    desc_dma);
+               swdata = cppi5_hdesc_get_swdata(desc_tx);
+               skb = *(swdata);
+               am65_cpsw_nuss_xmit_free(tx_chn, dev, desc_tx);
+
+               ndev = skb->dev;
+
+               ndev_priv = netdev_priv(ndev);
+               stats = this_cpu_ptr(ndev_priv->stats);
+               u64_stats_update_begin(&stats->syncp);
+               stats->tx_packets++;
+               stats->tx_bytes += skb->len;
+               u64_stats_update_end(&stats->syncp);
+
+               total_bytes += skb->len;
+               napi_consume_skb(skb, budget);
+               num_tx++;
+       }
+
+       if (!num_tx)
+               return 0;
+
+       netif_txq = netdev_get_tx_queue(ndev, chn);
+
+       netdev_tx_completed_queue(netif_txq, num_tx, total_bytes);
+
+       if (netif_tx_queue_stopped(netif_txq)) {
+               /* Check whether the queue is stopped due to stalled tx dma,
+                * if the queue is stopped then wake the queue as
+                * we have free desc for tx
+                */
+               __netif_tx_lock(netif_txq, smp_processor_id());
+               if (netif_running(ndev) &&
+                   (k3_cppi_desc_pool_avail(tx_chn->desc_pool) >=
+                    MAX_SKB_FRAGS))
+                       netif_tx_wake_queue(netif_txq);
+
+               __netif_tx_unlock(netif_txq);
+       }
+       dev_dbg(dev, "%s:%u pkt:%d\n", __func__, chn, num_tx);
+
+       return num_tx;
+}
+
+static int am65_cpsw_nuss_tx_poll(struct napi_struct *napi_tx, int budget)
+{
+       struct am65_cpsw_tx_chn *tx_chn = am65_cpsw_napi_to_tx_chn(napi_tx);
+       int num_tx;
+
+       num_tx = am65_cpsw_nuss_tx_compl_packets(tx_chn->common, tx_chn->id,
+                                                budget);
+       num_tx = min(num_tx, budget);
+       if (num_tx < budget) {
+               napi_complete(napi_tx);
+               enable_irq(tx_chn->irq);
+       }
+
+       return num_tx;
+}
+
+static irqreturn_t am65_cpsw_nuss_rx_irq(int irq, void *dev_id)
+{
+       struct am65_cpsw_common *common = dev_id;
+
+       disable_irq_nosync(irq);
+       napi_schedule(&common->napi_rx);
+
+       return IRQ_HANDLED;
+}
+
+static irqreturn_t am65_cpsw_nuss_tx_irq(int irq, void *dev_id)
+{
+       struct am65_cpsw_tx_chn *tx_chn = dev_id;
+
+       disable_irq_nosync(irq);
+       napi_schedule(&tx_chn->napi_tx);
+
+       return IRQ_HANDLED;
+}
+
+static netdev_tx_t am65_cpsw_nuss_ndo_slave_xmit(struct sk_buff *skb,
+                                                struct net_device *ndev)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct cppi5_host_desc_t *first_desc, *next_desc, *cur_desc;
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+       struct device *dev = common->dev;
+       struct am65_cpsw_tx_chn *tx_chn;
+       struct netdev_queue *netif_txq;
+       dma_addr_t desc_dma, buf_dma;
+       int ret, q_idx, i;
+       void **swdata;
+       u32 *psdata;
+       u32 pkt_len;
+
+       /* padding enabled in hw */
+       pkt_len = skb_headlen(skb);
+
+       q_idx = skb_get_queue_mapping(skb);
+       dev_dbg(dev, "%s skb_queue:%d\n", __func__, q_idx);
+
+       tx_chn = &common->tx_chns[q_idx];
+       netif_txq = netdev_get_tx_queue(ndev, q_idx);
+
+       /* Map the linear buffer */
+       buf_dma = dma_map_single(dev, skb->data, pkt_len,
+                                DMA_TO_DEVICE);
+       if (unlikely(dma_mapping_error(dev, buf_dma))) {
+               dev_err(dev, "Failed to map tx skb buffer\n");
+               ndev->stats.tx_errors++;
+               goto err_free_skb;
+       }
+
+       first_desc = k3_cppi_desc_pool_alloc(tx_chn->desc_pool);
+       if (!first_desc) {
+               dev_dbg(dev, "Failed to allocate descriptor\n");
+               dma_unmap_single(dev, buf_dma, pkt_len, DMA_TO_DEVICE);
+               goto busy_stop_q;
+       }
+
+       cppi5_hdesc_init(first_desc, CPPI5_INFO0_HDESC_EPIB_PRESENT,
+                        AM65_CPSW_NAV_PS_DATA_SIZE);
+       cppi5_desc_set_pktids(&first_desc->hdr, 0, 0x3FFF);
+       cppi5_hdesc_set_pkttype(first_desc, 0x7);
+       cppi5_desc_set_tags_ids(&first_desc->hdr, 0, port->port_id);
+
+       cppi5_hdesc_attach_buf(first_desc, buf_dma, pkt_len, buf_dma, pkt_len);
+       swdata = cppi5_hdesc_get_swdata(first_desc);
+       *(swdata) = skb;
+       psdata = cppi5_hdesc_get_psdata(first_desc);
+
+       /* HW csum offload if enabled */
+       psdata[2] = 0;
+       if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
+               unsigned int cs_start, cs_offset;
+
+               cs_start = skb_transport_offset(skb);
+               cs_offset = cs_start + skb->csum_offset;
+               /* HW numerates bytes starting from 1 */
+               psdata[2] = ((cs_offset + 1) << 24) |
+                           ((cs_start + 1) << 16) | (skb->len - cs_start);
+               dev_dbg(dev, "%s tx psdata:%#x\n", __func__, psdata[2]);
+       }
+
+       if (!skb_is_nonlinear(skb))
+               goto done_tx;
+
+       dev_dbg(dev, "fragmented SKB\n");
+
+       /* Handle the case where skb is fragmented in pages */
+       cur_desc = first_desc;
+       for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+               skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+               u32 frag_size = skb_frag_size(frag);
+
+               next_desc = k3_cppi_desc_pool_alloc(tx_chn->desc_pool);
+               if (!next_desc) {
+                       dev_err(dev, "Failed to allocate descriptor\n");
+                       goto busy_free_descs;
+               }
+
+               buf_dma = skb_frag_dma_map(dev, frag, 0, frag_size,
+                                          DMA_TO_DEVICE);
+               if (unlikely(dma_mapping_error(dev, buf_dma))) {
+                       dev_err(dev, "Failed to map tx skb page\n");
+                       k3_cppi_desc_pool_free(tx_chn->desc_pool, next_desc);
+                       ndev->stats.tx_errors++;
+                       goto err_free_descs;
+               }
+
+               cppi5_hdesc_reset_hbdesc(next_desc);
+               cppi5_hdesc_attach_buf(next_desc,
+                                      buf_dma, frag_size, buf_dma, frag_size);
+
+               desc_dma = k3_cppi_desc_pool_virt2dma(tx_chn->desc_pool,
+                                                     next_desc);
+               cppi5_hdesc_link_hbdesc(cur_desc, desc_dma);
+
+               pkt_len += frag_size;
+               cur_desc = next_desc;
+       }
+       WARN_ON(pkt_len != skb->len);
+
+done_tx:
+       skb_tx_timestamp(skb);
+
+       /* report bql before sending packet */
+       netdev_tx_sent_queue(netif_txq, pkt_len);
+
+       cppi5_hdesc_set_pktlen(first_desc, pkt_len);
+       desc_dma = k3_cppi_desc_pool_virt2dma(tx_chn->desc_pool, first_desc);
+       ret = k3_udma_glue_push_tx_chn(tx_chn->tx_chn, first_desc, desc_dma);
+       if (ret) {
+               dev_err(dev, "can't push desc %d\n", ret);
+               /* inform bql */
+               netdev_tx_completed_queue(netif_txq, 1, pkt_len);
+               ndev->stats.tx_errors++;
+               goto err_free_descs;
+       }
+
+       if (k3_cppi_desc_pool_avail(tx_chn->desc_pool) < MAX_SKB_FRAGS) {
+               netif_tx_stop_queue(netif_txq);
+               /* Barrier, so that stop_queue visible to other cpus */
+               smp_mb__after_atomic();
+               dev_dbg(dev, "netif_tx_stop_queue %d\n", q_idx);
+
+               /* re-check for smp */
+               if (k3_cppi_desc_pool_avail(tx_chn->desc_pool) >=
+                   MAX_SKB_FRAGS) {
+                       netif_tx_wake_queue(netif_txq);
+                       dev_dbg(dev, "netif_tx_wake_queue %d\n", q_idx);
+               }
+       }
+
+       return NETDEV_TX_OK;
+
+err_free_descs:
+       am65_cpsw_nuss_xmit_free(tx_chn, dev, first_desc);
+err_free_skb:
+       ndev->stats.tx_dropped++;
+       dev_kfree_skb_any(skb);
+       return NETDEV_TX_OK;
+
+busy_free_descs:
+       am65_cpsw_nuss_xmit_free(tx_chn, dev, first_desc);
+busy_stop_q:
+       netif_tx_stop_queue(netif_txq);
+       return NETDEV_TX_BUSY;
+}
+
+static int am65_cpsw_nuss_ndo_slave_set_mac_address(struct net_device *ndev,
+                                                   void *addr)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+       struct sockaddr *sockaddr = (struct sockaddr *)addr;
+       int ret;
+
+       ret = eth_prepare_mac_addr_change(ndev, addr);
+       if (ret < 0)
+               return ret;
+
+       ret = pm_runtime_get_sync(common->dev);
+       if (ret < 0) {
+               pm_runtime_put_noidle(common->dev);
+               return ret;
+       }
+
+       cpsw_ale_del_ucast(common->ale, ndev->dev_addr,
+                          HOST_PORT_NUM, 0, 0);
+       cpsw_ale_add_ucast(common->ale, sockaddr->sa_data,
+                          HOST_PORT_NUM, ALE_SECURE, 0);
+
+       am65_cpsw_port_set_sl_mac(port, addr);
+       eth_commit_mac_addr_change(ndev, sockaddr);
+
+       pm_runtime_put(common->dev);
+
+       return 0;
+}
+
+static int am65_cpsw_nuss_ndo_slave_ioctl(struct net_device *ndev,
+                                         struct ifreq *req, int cmd)
+{
+       struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+
+       if (!netif_running(ndev))
+               return -EINVAL;
+
+       if (!port->slave.phy)
+               return -EOPNOTSUPP;
+
+       return phy_mii_ioctl(port->slave.phy, req, cmd);
+}
+
+static void am65_cpsw_nuss_ndo_get_stats(struct net_device *dev,
+                                        struct rtnl_link_stats64 *stats)
+{
+       struct am65_cpsw_ndev_priv *ndev_priv = netdev_priv(dev);
+       unsigned int start;
+       int cpu;
+
+       for_each_possible_cpu(cpu) {
+               struct am65_cpsw_ndev_stats *cpu_stats;
+               u64 rx_packets;
+               u64 rx_bytes;
+               u64 tx_packets;
+               u64 tx_bytes;
+
+               cpu_stats = per_cpu_ptr(ndev_priv->stats, cpu);
+               do {
+                       start = u64_stats_fetch_begin_irq(&cpu_stats->syncp);
+                       rx_packets = cpu_stats->rx_packets;
+                       rx_bytes   = cpu_stats->rx_bytes;
+                       tx_packets = cpu_stats->tx_packets;
+                       tx_bytes   = cpu_stats->tx_bytes;
+               } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, start));
+
+               stats->rx_packets += rx_packets;
+               stats->rx_bytes   += rx_bytes;
+               stats->tx_packets += tx_packets;
+               stats->tx_bytes   += tx_bytes;
+       }
+
+       stats->rx_errors        = dev->stats.rx_errors;
+       stats->rx_dropped       = dev->stats.rx_dropped;
+       stats->tx_dropped       = dev->stats.tx_dropped;
+}
+
+static int am65_cpsw_nuss_ndo_slave_set_features(struct net_device *ndev,
+                                                netdev_features_t features)
+{
+       struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+       netdev_features_t changes = features ^ ndev->features;
+       struct am65_cpsw_host *host_p;
+
+       host_p = am65_common_get_host(common);
+
+       if (changes & NETIF_F_HW_CSUM) {
+               bool enable = !!(features & NETIF_F_HW_CSUM);
+
+               dev_info(common->dev, "Turn %s tx-checksum-ip-generic\n",
+                        enable ? "ON" : "OFF");
+               if (enable)
+                       writel(AM65_CPSW_P0_REG_CTL_RX_CHECKSUM_EN,
+                              host_p->port_base + AM65_CPSW_P0_REG_CTL);
+               else
+                       writel(0,
+                              host_p->port_base + AM65_CPSW_P0_REG_CTL);
+       }
+
+       return 0;
+}
+
+static const struct net_device_ops am65_cpsw_nuss_netdev_ops_2g = {
+       .ndo_open               = am65_cpsw_nuss_ndo_slave_open,
+       .ndo_stop               = am65_cpsw_nuss_ndo_slave_stop,
+       .ndo_start_xmit         = am65_cpsw_nuss_ndo_slave_xmit,
+       .ndo_set_rx_mode        = am65_cpsw_nuss_ndo_slave_set_rx_mode,
+       .ndo_get_stats64        = am65_cpsw_nuss_ndo_get_stats,
+       .ndo_validate_addr      = eth_validate_addr,
+       .ndo_set_mac_address    = am65_cpsw_nuss_ndo_slave_set_mac_address,
+       .ndo_tx_timeout         = am65_cpsw_nuss_ndo_host_tx_timeout,
+       .ndo_vlan_rx_add_vid    = am65_cpsw_nuss_ndo_slave_add_vid,
+       .ndo_vlan_rx_kill_vid   = am65_cpsw_nuss_ndo_slave_kill_vid,
+       .ndo_do_ioctl           = am65_cpsw_nuss_ndo_slave_ioctl,
+       .ndo_set_features       = am65_cpsw_nuss_ndo_slave_set_features,
+};
+
+static void am65_cpsw_nuss_slave_disable_unused(struct am65_cpsw_port *port)
+{
+       struct am65_cpsw_common *common = port->common;
+
+       if (!port->disabled)
+               return;
+
+       common->disabled_ports_mask |= BIT(port->port_id);
+       cpsw_ale_control_set(common->ale, port->port_id,
+                            ALE_PORT_STATE, ALE_PORT_STATE_DISABLE);
+
+       cpsw_sl_reset(port->slave.mac_sl, 100);
+       cpsw_sl_ctl_reset(port->slave.mac_sl);
+}
+
+static void am65_cpsw_nuss_free_tx_chns(void *data)
+{
+       struct am65_cpsw_common *common = data;
+       int i;
+
+       for (i = 0; i < common->tx_ch_num; i++) {
+               struct am65_cpsw_tx_chn *tx_chn = &common->tx_chns[i];
+
+               if (!IS_ERR_OR_NULL(tx_chn->tx_chn))
+                       k3_udma_glue_release_tx_chn(tx_chn->tx_chn);
+
+               if (!IS_ERR_OR_NULL(tx_chn->desc_pool))
+                       k3_cppi_desc_pool_destroy(tx_chn->desc_pool);
+
+               memset(tx_chn, 0, sizeof(*tx_chn));
+       }
+}
+
+void am65_cpsw_nuss_remove_tx_chns(struct am65_cpsw_common *common)
+{
+       struct device *dev = common->dev;
+       int i;
+
+       devm_remove_action(dev, am65_cpsw_nuss_free_tx_chns, common);
+
+       for (i = 0; i < common->tx_ch_num; i++) {
+               struct am65_cpsw_tx_chn *tx_chn = &common->tx_chns[i];
+
+               if (tx_chn->irq)
+                       devm_free_irq(dev, tx_chn->irq, tx_chn);
+
+               netif_napi_del(&tx_chn->napi_tx);
+
+               if (!IS_ERR_OR_NULL(tx_chn->tx_chn))
+                       k3_udma_glue_release_tx_chn(tx_chn->tx_chn);
+
+               if (!IS_ERR_OR_NULL(tx_chn->desc_pool))
+                       k3_cppi_desc_pool_destroy(tx_chn->desc_pool);
+
+               memset(tx_chn, 0, sizeof(*tx_chn));
+       }
+}
+
+static int am65_cpsw_nuss_init_tx_chns(struct am65_cpsw_common *common)
+{
+       u32  max_desc_num = ALIGN(AM65_CPSW_MAX_TX_DESC, MAX_SKB_FRAGS);
+       struct k3_udma_glue_tx_channel_cfg tx_cfg = { 0 };
+       struct device *dev = common->dev;
+       struct k3_ring_cfg ring_cfg = {
+               .elm_size = K3_RINGACC_RING_ELSIZE_8,
+               .mode = K3_RINGACC_RING_MODE_RING,
+               .flags = 0
+       };
+       u32 hdesc_size;
+       int i, ret = 0;
+
+       hdesc_size = cppi5_hdesc_calc_size(true, AM65_CPSW_NAV_PS_DATA_SIZE,
+                                          AM65_CPSW_NAV_SW_DATA_SIZE);
+
+       tx_cfg.swdata_size = AM65_CPSW_NAV_SW_DATA_SIZE;
+       tx_cfg.tx_cfg = ring_cfg;
+       tx_cfg.txcq_cfg = ring_cfg;
+       tx_cfg.tx_cfg.size = max_desc_num;
+       tx_cfg.txcq_cfg.size = max_desc_num;
+
+       for (i = 0; i < common->tx_ch_num; i++) {
+               struct am65_cpsw_tx_chn *tx_chn = &common->tx_chns[i];
+
+               snprintf(tx_chn->tx_chn_name,
+                        sizeof(tx_chn->tx_chn_name), "tx%d", i);
+
+               tx_chn->common = common;
+               tx_chn->id = i;
+               tx_chn->descs_num = max_desc_num;
+               tx_chn->desc_pool =
+                       k3_cppi_desc_pool_create_name(dev,
+                                                     tx_chn->descs_num,
+                                                     hdesc_size,
+                                                     tx_chn->tx_chn_name);
+               if (IS_ERR(tx_chn->desc_pool)) {
+                       ret = PTR_ERR(tx_chn->desc_pool);
+                       dev_err(dev, "Failed to create poll %d\n", ret);
+                       goto err;
+               }
+
+               tx_chn->tx_chn =
+                       k3_udma_glue_request_tx_chn(dev,
+                                                   tx_chn->tx_chn_name,
+                                                   &tx_cfg);
+               if (IS_ERR(tx_chn->tx_chn)) {
+                       ret = PTR_ERR(tx_chn->tx_chn);
+                       dev_err(dev, "Failed to request tx dma channel %d\n",
+                               ret);
+                       goto err;
+               }
+
+               tx_chn->irq = k3_udma_glue_tx_get_irq(tx_chn->tx_chn);
+               if (tx_chn->irq <= 0) {
+                       dev_err(dev, "Failed to get tx dma irq %d\n",
+                               tx_chn->irq);
+                       goto err;
+               }
+
+               snprintf(tx_chn->tx_chn_name,
+                        sizeof(tx_chn->tx_chn_name), "%s-tx%d",
+                        dev_name(dev), tx_chn->id);
+       }
+
+err:
+       i = devm_add_action(dev, am65_cpsw_nuss_free_tx_chns, common);
+       if (i) {
+               dev_err(dev, "failed to add free_tx_chns action %d", i);
+               return i;
+       }
+
+       return ret;
+}
+
+static void am65_cpsw_nuss_free_rx_chns(void *data)
+{
+       struct am65_cpsw_common *common = data;
+       struct am65_cpsw_rx_chn *rx_chn;
+
+       rx_chn = &common->rx_chns;
+
+       if (!IS_ERR_OR_NULL(rx_chn->rx_chn))
+               k3_udma_glue_release_rx_chn(rx_chn->rx_chn);
+
+       if (!IS_ERR_OR_NULL(rx_chn->desc_pool))
+               k3_cppi_desc_pool_destroy(rx_chn->desc_pool);
+}
+
+static int am65_cpsw_nuss_init_rx_chns(struct am65_cpsw_common *common)
+{
+       struct am65_cpsw_rx_chn *rx_chn = &common->rx_chns;
+       struct k3_udma_glue_rx_channel_cfg rx_cfg = { 0 };
+       u32  max_desc_num = AM65_CPSW_MAX_RX_DESC;
+       struct device *dev = common->dev;
+       u32 hdesc_size;
+       u32 fdqring_id;
+       int i, ret = 0;
+
+       hdesc_size = cppi5_hdesc_calc_size(true, AM65_CPSW_NAV_PS_DATA_SIZE,
+                                          AM65_CPSW_NAV_SW_DATA_SIZE);
+
+       rx_cfg.swdata_size = AM65_CPSW_NAV_SW_DATA_SIZE;
+       rx_cfg.flow_id_num = AM65_CPSW_MAX_RX_FLOWS;
+       rx_cfg.flow_id_base = common->rx_flow_id_base;
+
+       /* init all flows */
+       rx_chn->dev = dev;
+       rx_chn->descs_num = max_desc_num;
+       rx_chn->desc_pool = k3_cppi_desc_pool_create_name(dev,
+                                                         rx_chn->descs_num,
+                                                         hdesc_size, "rx");
+       if (IS_ERR(rx_chn->desc_pool)) {
+               ret = PTR_ERR(rx_chn->desc_pool);
+               dev_err(dev, "Failed to create rx poll %d\n", ret);
+               goto err;
+       }
+
+       rx_chn->rx_chn = k3_udma_glue_request_rx_chn(dev, "rx", &rx_cfg);
+       if (IS_ERR(rx_chn->rx_chn)) {
+               ret = PTR_ERR(rx_chn->rx_chn);
+               dev_err(dev, "Failed to request rx dma channel %d\n", ret);
+               goto err;
+       }
+
+       common->rx_flow_id_base =
+                       k3_udma_glue_rx_get_flow_id_base(rx_chn->rx_chn);
+       dev_info(dev, "set new flow-id-base %u\n", common->rx_flow_id_base);
+
+       fdqring_id = K3_RINGACC_RING_ID_ANY;
+       for (i = 0; i < rx_cfg.flow_id_num; i++) {
+               struct k3_ring_cfg rxring_cfg = {
+                       .elm_size = K3_RINGACC_RING_ELSIZE_8,
+                       .mode = K3_RINGACC_RING_MODE_RING,
+                       .flags = 0,
+               };
+               struct k3_ring_cfg fdqring_cfg = {
+                       .elm_size = K3_RINGACC_RING_ELSIZE_8,
+                       .mode = K3_RINGACC_RING_MODE_MESSAGE,
+                       .flags = K3_RINGACC_RING_SHARED,
+               };
+               struct k3_udma_glue_rx_flow_cfg rx_flow_cfg = {
+                       .rx_cfg = rxring_cfg,
+                       .rxfdq_cfg = fdqring_cfg,
+                       .ring_rxq_id = K3_RINGACC_RING_ID_ANY,
+                       .src_tag_lo_sel =
+                               K3_UDMA_GLUE_SRC_TAG_LO_USE_REMOTE_SRC_TAG,
+               };
+
+               rx_flow_cfg.ring_rxfdq0_id = fdqring_id;
+               rx_flow_cfg.rx_cfg.size = max_desc_num;
+               rx_flow_cfg.rxfdq_cfg.size = max_desc_num;
+
+               ret = k3_udma_glue_rx_flow_init(rx_chn->rx_chn,
+                                               i, &rx_flow_cfg);
+               if (ret) {
+                       dev_err(dev, "Failed to init rx flow%d %d\n", i, ret);
+                       goto err;
+               }
+               if (!i)
+                       fdqring_id =
+                               k3_udma_glue_rx_flow_get_fdq_id(rx_chn->rx_chn,
+                                                               i);
+
+               rx_chn->irq = k3_udma_glue_rx_get_irq(rx_chn->rx_chn, i);
+
+               if (rx_chn->irq <= 0) {
+                       dev_err(dev, "Failed to get rx dma irq %d\n",
+                               rx_chn->irq);
+                       ret = -ENXIO;
+                       goto err;
+               }
+       }
+
+err:
+       i = devm_add_action(dev, am65_cpsw_nuss_free_rx_chns, common);
+       if (i) {
+               dev_err(dev, "failed to add free_rx_chns action %d", i);
+               return i;
+       }
+
+       return ret;
+}
+
+static int am65_cpsw_nuss_init_host_p(struct am65_cpsw_common *common)
+{
+       struct am65_cpsw_host *host_p = am65_common_get_host(common);
+
+       host_p->common = common;
+       host_p->port_base = common->cpsw_base + AM65_CPSW_NU_PORTS_BASE;
+       host_p->stat_base = common->cpsw_base + AM65_CPSW_NU_STATS_BASE;
+
+       return 0;
+}
+
+static int am65_cpsw_am654_get_efuse_macid(struct device_node *of_node,
+                                          int slave, u8 *mac_addr)
+{
+       u32 mac_lo, mac_hi, offset;
+       struct regmap *syscon;
+       int ret;
+
+       syscon = syscon_regmap_lookup_by_phandle(of_node, "ti,syscon-efuse");
+       if (IS_ERR(syscon)) {
+               if (PTR_ERR(syscon) == -ENODEV)
+                       return 0;
+               return PTR_ERR(syscon);
+       }
+
+       ret = of_property_read_u32_index(of_node, "ti,syscon-efuse", 1,
+                                        &offset);
+       if (ret)
+               return ret;
+
+       regmap_read(syscon, offset, &mac_lo);
+       regmap_read(syscon, offset + 4, &mac_hi);
+
+       mac_addr[0] = (mac_hi >> 8) & 0xff;
+       mac_addr[1] = mac_hi & 0xff;
+       mac_addr[2] = (mac_lo >> 24) & 0xff;
+       mac_addr[3] = (mac_lo >> 16) & 0xff;
+       mac_addr[4] = (mac_lo >> 8) & 0xff;
+       mac_addr[5] = mac_lo & 0xff;
+
+       return 0;
+}
+
+static int am65_cpsw_nuss_init_slave_ports(struct am65_cpsw_common *common)
+{
+       struct device_node *node, *port_np;
+       struct device *dev = common->dev;
+       int ret;
+
+       node = of_get_child_by_name(dev->of_node, "ethernet-ports");
+       if (!node)
+               return -ENOENT;
+
+       for_each_child_of_node(node, port_np) {
+               struct am65_cpsw_port *port;
+               const void *mac_addr;
+               u32 port_id;
+
+               /* it is not a slave port node, continue */
+               if (strcmp(port_np->name, "port"))
+                       continue;
+
+               ret = of_property_read_u32(port_np, "reg", &port_id);
+               if (ret < 0) {
+                       dev_err(dev, "%pOF error reading port_id %d\n",
+                               port_np, ret);
+                       return ret;
+               }
+
+               if (!port_id || port_id > common->port_num) {
+                       dev_err(dev, "%pOF has invalid port_id %u %s\n",
+                               port_np, port_id, port_np->name);
+                       return -EINVAL;
+               }
+
+               port = am65_common_get_port(common, port_id);
+               port->port_id = port_id;
+               port->common = common;
+               port->port_base = common->cpsw_base + AM65_CPSW_NU_PORTS_BASE +
+                                 AM65_CPSW_NU_PORTS_OFFSET * (port_id);
+               port->stat_base = common->cpsw_base + AM65_CPSW_NU_STATS_BASE +
+                                 (AM65_CPSW_NU_STATS_PORT_OFFSET * port_id);
+               port->name = of_get_property(port_np, "label", NULL);
+
+               port->disabled = !of_device_is_available(port_np);
+               if (port->disabled)
+                       continue;
+
+               port->slave.ifphy = devm_of_phy_get(dev, port_np, NULL);
+               if (IS_ERR(port->slave.ifphy)) {
+                       ret = PTR_ERR(port->slave.ifphy);
+                       dev_err(dev, "%pOF error retrieving port phy: %d\n",
+                               port_np, ret);
+                       return ret;
+               }
+
+               port->slave.mac_only =
+                               of_property_read_bool(port_np, "ti,mac-only");
+
+               /* get phy/link info */
+               if (of_phy_is_fixed_link(port_np)) {
+                       ret = of_phy_register_fixed_link(port_np);
+                       if (ret) {
+                               if (ret != -EPROBE_DEFER)
+                                       dev_err(dev, "%pOF failed to register fixed-link phy: %d\n",
+                                               port_np, ret);
+                               return ret;
+                       }
+                       port->slave.phy_node = of_node_get(port_np);
+               } else {
+                       port->slave.phy_node =
+                               of_parse_phandle(port_np, "phy-handle", 0);
+               }
+
+               if (!port->slave.phy_node) {
+                       dev_err(dev,
+                               "slave[%d] no phy found\n", port_id);
+                       return -ENODEV;
+               }
+
+               ret = of_get_phy_mode(port_np, &port->slave.phy_if);
+               if (ret) {
+                       dev_err(dev, "%pOF read phy-mode err %d\n",
+                               port_np, ret);
+                       return ret;
+               }
+
+               port->slave.mac_sl = cpsw_sl_get("am65", dev, port->port_base);
+               if (IS_ERR(port->slave.mac_sl))
+                       return PTR_ERR(port->slave.mac_sl);
+
+               mac_addr = of_get_mac_address(port_np);
+               if (!IS_ERR(mac_addr)) {
+                       ether_addr_copy(port->slave.mac_addr, mac_addr);
+               } else if (am65_cpsw_am654_get_efuse_macid(port_np,
+                                                          port->port_id,
+                                                          port->slave.mac_addr) ||
+                          !is_valid_ether_addr(port->slave.mac_addr)) {
+                       random_ether_addr(port->slave.mac_addr);
+                       dev_err(dev, "Use rundom MAC address\n");
+               }
+       }
+       of_node_put(node);
+
+       return 0;
+}
+
+static void am65_cpsw_pcpu_stats_free(void *data)
+{
+       struct am65_cpsw_ndev_stats __percpu *stats = data;
+
+       free_percpu(stats);
+}
+
+static int am65_cpsw_nuss_init_ndev_2g(struct am65_cpsw_common *common)
+{
+       struct am65_cpsw_ndev_priv *ndev_priv;
+       struct device *dev = common->dev;
+       struct am65_cpsw_port *port;
+       int ret;
+
+       port = am65_common_get_port(common, 1);
+
+       /* alloc netdev */
+       port->ndev = devm_alloc_etherdev_mqs(common->dev,
+                                            sizeof(struct am65_cpsw_ndev_priv),
+                                            AM65_CPSW_MAX_TX_QUEUES,
+                                            AM65_CPSW_MAX_RX_QUEUES);
+       if (!port->ndev) {
+               dev_err(dev, "error allocating slave net_device %u\n",
+                       port->port_id);
+               return -ENOMEM;
+       }
+
+       ndev_priv = netdev_priv(port->ndev);
+       ndev_priv->port = port;
+       ndev_priv->msg_enable = AM65_CPSW_DEBUG;
+       SET_NETDEV_DEV(port->ndev, dev);
+
+       ether_addr_copy(port->ndev->dev_addr, port->slave.mac_addr);
+
+       port->ndev->min_mtu = AM65_CPSW_MIN_PACKET_SIZE;
+       port->ndev->max_mtu = AM65_CPSW_MAX_PACKET_SIZE;
+       port->ndev->hw_features = NETIF_F_SG |
+                                 NETIF_F_RXCSUM |
+                                 NETIF_F_HW_CSUM;
+       port->ndev->features = port->ndev->hw_features |
+                              NETIF_F_HW_VLAN_CTAG_FILTER;
+       port->ndev->vlan_features |=  NETIF_F_SG;
+       port->ndev->netdev_ops = &am65_cpsw_nuss_netdev_ops_2g;
+       port->ndev->ethtool_ops = &am65_cpsw_ethtool_ops_slave;
+
+       /* Disable TX checksum offload by default due to HW bug */
+       if (common->pdata->quirks & AM65_CPSW_QUIRK_I2027_NO_TX_CSUM)
+               port->ndev->features &= ~NETIF_F_HW_CSUM;
+
+       ndev_priv->stats = netdev_alloc_pcpu_stats(struct am65_cpsw_ndev_stats);
+       if (!ndev_priv->stats)
+               return -ENOMEM;
+
+       ret = devm_add_action_or_reset(dev, am65_cpsw_pcpu_stats_free,
+                                      ndev_priv->stats);
+       if (ret) {
+               dev_err(dev, "failed to add percpu stat free action %d", ret);
+               return ret;
+       }
+
+       netif_napi_add(port->ndev, &common->napi_rx,
+                      am65_cpsw_nuss_rx_poll, NAPI_POLL_WEIGHT);
+
+       common->pf_p0_rx_ptype_rrobin = false;
+
+       return ret;
+}
+
+static int am65_cpsw_nuss_ndev_add_napi_2g(struct am65_cpsw_common *common)
+{
+       struct device *dev = common->dev;
+       struct am65_cpsw_port *port;
+       int i, ret = 0;
+
+       port = am65_common_get_port(common, 1);
+
+       for (i = 0; i < common->tx_ch_num; i++) {
+               struct am65_cpsw_tx_chn *tx_chn = &common->tx_chns[i];
+
+               netif_tx_napi_add(port->ndev, &tx_chn->napi_tx,
+                                 am65_cpsw_nuss_tx_poll, NAPI_POLL_WEIGHT);
+
+               ret = devm_request_irq(dev, tx_chn->irq,
+                                      am65_cpsw_nuss_tx_irq,
+                                      0, tx_chn->tx_chn_name, tx_chn);
+               if (ret) {
+                       dev_err(dev, "failure requesting tx%u irq %u, %d\n",
+                               tx_chn->id, tx_chn->irq, ret);
+                       goto err;
+               }
+       }
+
+err:
+       return ret;
+}
+
+static int am65_cpsw_nuss_ndev_reg_2g(struct am65_cpsw_common *common)
+{
+       struct device *dev = common->dev;
+       struct am65_cpsw_port *port;
+       int ret = 0;
+
+       port = am65_common_get_port(common, 1);
+       ret = am65_cpsw_nuss_ndev_add_napi_2g(common);
+       if (ret)
+               goto err;
+
+       ret = devm_request_irq(dev, common->rx_chns.irq,
+                              am65_cpsw_nuss_rx_irq,
+                              0, dev_name(dev), common);
+       if (ret) {
+               dev_err(dev, "failure requesting rx irq %u, %d\n",
+                       common->rx_chns.irq, ret);
+               goto err;
+       }
+
+       ret = register_netdev(port->ndev);
+       if (ret)
+               dev_err(dev, "error registering slave net device %d\n", ret);
+
+       /* can't auto unregister ndev using devm_add_action() due to
+        * devres release sequence in DD core for DMA
+        */
+err:
+       return ret;
+}
+
+int am65_cpsw_nuss_update_tx_chns(struct am65_cpsw_common *common, int num_tx)
+{
+       int ret;
+
+       common->tx_ch_num = num_tx;
+       ret = am65_cpsw_nuss_init_tx_chns(common);
+       if (ret)
+               return ret;
+
+       return am65_cpsw_nuss_ndev_add_napi_2g(common);
+}
+
+static void am65_cpsw_nuss_cleanup_ndev(struct am65_cpsw_common *common)
+{
+       struct am65_cpsw_port *port;
+       int i;
+
+       for (i = 0; i < common->port_num; i++) {
+               port = &common->ports[i];
+               if (port->ndev)
+                       unregister_netdev(port->ndev);
+       }
+}
+
+static const struct am65_cpsw_pdata am65x_sr1_0 = {
+       .quirks = AM65_CPSW_QUIRK_I2027_NO_TX_CSUM,
+};
+
+static const struct am65_cpsw_pdata j721e_sr1_0 = {
+       .quirks = 0,
+};
+
+static const struct of_device_id am65_cpsw_nuss_of_mtable[] = {
+       { .compatible = "ti,am654-cpsw-nuss", .data = &am65x_sr1_0 },
+       { .compatible = "ti,j721e-cpsw-nuss", .data = &j721e_sr1_0 },
+       { /* sentinel */ },
+};
+MODULE_DEVICE_TABLE(of, am65_cpsw_nuss_of_mtable);
+
+static int am65_cpsw_nuss_probe(struct platform_device *pdev)
+{
+       struct cpsw_ale_params ale_params;
+       const struct of_device_id *of_id;
+       struct device *dev = &pdev->dev;
+       struct am65_cpsw_common *common;
+       struct device_node *node;
+       struct resource *res;
+       int ret, i;
+
+       common = devm_kzalloc(dev, sizeof(struct am65_cpsw_common), GFP_KERNEL);
+       if (!common)
+               return -ENOMEM;
+       common->dev = dev;
+
+       of_id = of_match_device(am65_cpsw_nuss_of_mtable, dev);
+       if (!of_id)
+               return -EINVAL;
+       common->pdata = of_id->data;
+
+       res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "cpsw_nuss");
+       common->ss_base = devm_ioremap_resource(&pdev->dev, res);
+       if (IS_ERR(common->ss_base))
+               return PTR_ERR(common->ss_base);
+       common->cpsw_base = common->ss_base + AM65_CPSW_CPSW_NU_BASE;
+
+       node = of_get_child_by_name(dev->of_node, "ethernet-ports");
+       if (!node)
+               return -ENOENT;
+       common->port_num = of_get_child_count(node);
+       if (common->port_num < 1 || common->port_num > AM65_CPSW_MAX_PORTS)
+               return -ENOENT;
+       of_node_put(node);
+
+       if (common->port_num != 1)
+               return -EOPNOTSUPP;
+
+       common->rx_flow_id_base = -1;
+       init_completion(&common->tdown_complete);
+       common->tx_ch_num = 1;
+
+       ret = dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(48));
+       if (ret) {
+               dev_err(dev, "error setting dma mask: %d\n", ret);
+               return ret;
+       }
+
+       common->ports = devm_kcalloc(dev, common->port_num,
+                                    sizeof(*common->ports),
+                                    GFP_KERNEL);
+       if (!common->ports)
+               return -ENOMEM;
+
+       pm_runtime_enable(dev);
+       ret = pm_runtime_get_sync(dev);
+       if (ret < 0) {
+               pm_runtime_put_noidle(dev);
+               pm_runtime_disable(dev);
+               return ret;
+       }
+
+       ret = of_platform_populate(dev->of_node, NULL, NULL, dev);
+       /* We do not want to force this, as in some cases may not have child */
+       if (ret)
+               dev_warn(dev, "populating child nodes err:%d\n", ret);
+
+       am65_cpsw_nuss_get_ver(common);
+
+       /* init tx channels */
+       ret = am65_cpsw_nuss_init_tx_chns(common);
+       if (ret)
+               goto err_of_clear;
+       ret = am65_cpsw_nuss_init_rx_chns(common);
+       if (ret)
+               goto err_of_clear;
+
+       ret = am65_cpsw_nuss_init_host_p(common);
+       if (ret)
+               goto err_of_clear;
+
+       ret = am65_cpsw_nuss_init_slave_ports(common);
+       if (ret)
+               goto err_of_clear;
+
+       /* init common data */
+       ale_params.dev = dev;
+       ale_params.ale_ageout = AM65_CPSW_ALE_AGEOUT_DEFAULT;
+       ale_params.ale_entries = 0;
+       ale_params.ale_ports = common->port_num + 1;
+       ale_params.ale_regs = common->cpsw_base + AM65_CPSW_NU_ALE_BASE;
+       ale_params.nu_switch_ale = true;
+
+       common->ale = cpsw_ale_create(&ale_params);
+       if (!common->ale) {
+               dev_err(dev, "error initializing ale engine\n");
+               goto err_of_clear;
+       }
+
+       /* init ports */
+       for (i = 0; i < common->port_num; i++)
+               am65_cpsw_nuss_slave_disable_unused(&common->ports[i]);
+
+       dev_set_drvdata(dev, common);
+
+       ret = am65_cpsw_nuss_init_ndev_2g(common);
+       if (ret)
+               goto err_of_clear;
+
+       ret = am65_cpsw_nuss_ndev_reg_2g(common);
+       if (ret)
+               goto err_of_clear;
+
+       pm_runtime_put(dev);
+       return 0;
+
+err_of_clear:
+       of_platform_depopulate(dev);
+       pm_runtime_put_sync(dev);
+       pm_runtime_disable(dev);
+       return ret;
+}
+
+static int am65_cpsw_nuss_remove(struct platform_device *pdev)
+{
+       struct device *dev = &pdev->dev;
+       struct am65_cpsw_common *common;
+       int ret;
+
+       common = dev_get_drvdata(dev);
+
+       ret = pm_runtime_get_sync(&pdev->dev);
+       if (ret < 0) {
+               pm_runtime_put_noidle(&pdev->dev);
+               return ret;
+       }
+
+       /* must unregister ndevs here because DD release_driver routine calls
+        * dma_deconfigure(dev) before devres_release_all(dev)
+        */
+       am65_cpsw_nuss_cleanup_ndev(common);
+
+       of_platform_depopulate(dev);
+
+       pm_runtime_put_sync(&pdev->dev);
+       pm_runtime_disable(&pdev->dev);
+       return 0;
+}
+
+static struct platform_driver am65_cpsw_nuss_driver = {
+       .driver = {
+               .name    = AM65_CPSW_DRV_NAME,
+               .of_match_table = am65_cpsw_nuss_of_mtable,
+       },
+       .probe = am65_cpsw_nuss_probe,
+       .remove = am65_cpsw_nuss_remove,
+};
+
+module_platform_driver(am65_cpsw_nuss_driver);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Grygorii Strashko <grygorii.strashko@ti.com>");
+MODULE_DESCRIPTION("TI AM65 CPSW Ethernet driver");