--- /dev/null
+/*
+ * net/dsa/tag_ksz.c - Microchip KSZ Switch tag format handling
+ * Copyright (c) 2017 Microchip Technology
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/etherdevice.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <net/dsa.h>
+#include "dsa_priv.h"
+
+/* For Ingress (Host -> KSZ), 2 bytes are added before FCS.
+ * ---------------------------------------------------------------------------
+ * DA(6bytes)|SA(6bytes)|....|Data(nbytes)|tag0(1byte)|tag1(1byte)|FCS(4bytes)
+ * ---------------------------------------------------------------------------
+ * tag0 : Prioritization (not used now)
+ * tag1 : each bit represents port (eg, 0x01=port1, 0x02=port2, 0x10=port5)
+ *
+ * For Egress (KSZ -> Host), 1 byte is added before FCS.
+ * ---------------------------------------------------------------------------
+ * DA(6bytes)|SA(6bytes)|....|Data(nbytes)|tag0(1byte)|FCS(4bytes)
+ * ---------------------------------------------------------------------------
+ * tag0 : zero-based value represents port
+ *       (eg, 0x00=port1, 0x02=port3, 0x06=port7)
+ */
+
+#define        KSZ_INGRESS_TAG_LEN     2
+#define        KSZ_EGRESS_TAG_LEN      1
+
+static struct sk_buff *ksz_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       struct dsa_slave_priv *p = netdev_priv(dev);
+       struct sk_buff *nskb;
+       int padlen;
+       u8 *tag;
+
+       padlen = (skb->len >= ETH_ZLEN) ? 0 : ETH_ZLEN - skb->len;
+
+       if (skb_tailroom(skb) >= padlen + KSZ_INGRESS_TAG_LEN) {
+               nskb = skb;
+       } else {
+               nskb = alloc_skb(NET_IP_ALIGN + skb->len +
+                                padlen + KSZ_INGRESS_TAG_LEN, GFP_ATOMIC);
+               if (!nskb) {
+                       kfree_skb(skb);
+                       return NULL;
+               }
+               skb_reserve(nskb, NET_IP_ALIGN);
+
+               skb_reset_mac_header(nskb);
+               skb_set_network_header(nskb,
+                                      skb_network_header(skb) - skb->head);
+               skb_set_transport_header(nskb,
+                                        skb_transport_header(skb) - skb->head);
+               skb_copy_and_csum_dev(skb, skb_put(nskb, skb->len));
+               kfree_skb(skb);
+       }
+
+       /* skb is freed when it fails */
+       if (skb_put_padto(nskb, nskb->len + padlen))
+               return NULL;
+
+       tag = skb_put(nskb, KSZ_INGRESS_TAG_LEN);
+       tag[0] = 0;
+       tag[1] = 1 << p->dp->index; /* destination port */
+
+       return nskb;
+}
+
+struct sk_buff *ksz_rcv(struct sk_buff *skb, struct net_device *dev,
+                       struct packet_type *pt, struct net_device *orig_dev)
+{
+       struct dsa_switch_tree *dst = dev->dsa_ptr;
+       struct dsa_switch *ds;
+       u8 *tag;
+       int source_port;
+
+       ds = dst->cpu_dp->ds;
+
+       tag = skb_tail_pointer(skb) - KSZ_EGRESS_TAG_LEN;
+
+       source_port = tag[0] & 7;
+       if (source_port >= ds->num_ports || !ds->ports[source_port].netdev)
+               return NULL;
+
+       pskb_trim_rcsum(skb, skb->len - KSZ_EGRESS_TAG_LEN);
+
+       skb->dev = ds->ports[source_port].netdev;
+
+       return skb;
+}
+
+const struct dsa_device_ops ksz_netdev_ops = {
+       .xmit   = ksz_xmit,
+       .rcv    = ksz_rcv,
+};