From: Vivek Pernamitta <quic_vpernami@quicinc.com>

Currently, we only have support for the NET driver. This update allows a
new client to be configured as an Ethernet type over MHI by setting
"mhi_device_info.ethernet_if = true". A new interface for Ethernet will
be created with <mhi_device_name>_<mhi_channel_name>.

Signed-off-by: Vivek Pernamitta <quic_vpernami@quicinc.com>
---
 drivers/net/mhi_net.c | 84 +++++++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 71 insertions(+), 13 deletions(-)

diff --git a/drivers/net/mhi_net.c b/drivers/net/mhi_net.c
index 08ab67f9a769ec64e1007853e47743003a197ec4..ba65abcee32a253fc1eb9d75f1693734f4d53ee3 100644
--- a/drivers/net/mhi_net.c
+++ b/drivers/net/mhi_net.c
@@ -11,6 +11,7 @@
 #include <linux/netdevice.h>
 #include <linux/skbuff.h>
 #include <linux/u64_stats_sync.h>
+#include <linux/etherdevice.h>
 
 #define MHI_NET_MIN_MTU		ETH_MIN_MTU
 #define MHI_NET_MAX_MTU		0xffff
@@ -38,10 +39,12 @@ struct mhi_net_dev {
 	u32 rx_queue_sz;
 	int msg_enable;
 	unsigned int mru;
+	bool ethernet_if;
 };
 
 struct mhi_device_info {
 	const char *netname;
+	bool ethernet_if;
 };
 
 static int mhi_ndo_open(struct net_device *ndev)
@@ -119,11 +122,37 @@ static void mhi_ndo_get_stats64(struct net_device *ndev,
 	} while (u64_stats_fetch_retry(&mhi_netdev->stats.tx_syncp, start));
 }
 
+static int mhi_mac_address(struct net_device *dev, void *p)
+{
+	int ret;
+
+	if (dev->type == ARPHRD_ETHER) {
+		ret = eth_mac_addr(dev, p);
+	return ret;
+	}
+
+	return 0;
+}
+
+static int mhi_validate_address(struct net_device *dev)
+{
+	int ret;
+
+	if (dev->type == ARPHRD_ETHER) {
+		ret = eth_validate_addr(dev);
+		return ret;
+	}
+
+	return 0;
+}
+
 static const struct net_device_ops mhi_netdev_ops = {
 	.ndo_open               = mhi_ndo_open,
 	.ndo_stop               = mhi_ndo_stop,
 	.ndo_start_xmit         = mhi_ndo_xmit,
 	.ndo_get_stats64	= mhi_ndo_get_stats64,
+	.ndo_set_mac_address    = mhi_mac_address,
+	.ndo_validate_addr      = mhi_validate_address,
 };
 
 static void mhi_net_setup(struct net_device *ndev)
@@ -140,6 +169,14 @@ static void mhi_net_setup(struct net_device *ndev)
 	ndev->tx_queue_len = 1000;
 }
 
+static void mhi_ethernet_setup(struct net_device *ndev)
+{
+	ndev->netdev_ops = &mhi_netdev_ops;
+	ether_setup(ndev);
+	ndev->min_mtu = ETH_MIN_MTU;
+	ndev->max_mtu = ETH_MAX_MTU;
+}
+
 static struct sk_buff *mhi_net_skb_agg(struct mhi_net_dev *mhi_netdev,
 				       struct sk_buff *skb)
 {
@@ -209,16 +246,22 @@ static void mhi_net_dl_callback(struct mhi_device *mhi_dev,
 			mhi_netdev->skbagg_head = NULL;
 		}
 
-		switch (skb->data[0] & 0xf0) {
-		case 0x40:
-			skb->protocol = htons(ETH_P_IP);
-			break;
-		case 0x60:
-			skb->protocol = htons(ETH_P_IPV6);
-			break;
-		default:
-			skb->protocol = htons(ETH_P_MAP);
-			break;
+		if (mhi_netdev->ethernet_if) {
+			skb_copy_to_linear_data(skb, skb->data,
+						mhi_res->bytes_xferd);
+			skb->protocol = eth_type_trans(skb, mhi_netdev->ndev);
+		} else {
+			switch (skb->data[0] & 0xf0) {
+			case 0x40:
+				skb->protocol = htons(ETH_P_IP);
+				break;
+			case 0x60:
+				skb->protocol = htons(ETH_P_IPV6);
+				break;
+			default:
+				skb->protocol = htons(ETH_P_MAP);
+				break;
+			}
 		}
 
 		u64_stats_update_begin(&mhi_netdev->stats.rx_syncp);
@@ -301,11 +344,17 @@ static void mhi_net_rx_refill_work(struct work_struct *work)
 		schedule_delayed_work(&mhi_netdev->rx_refill, HZ / 2);
 }
 
-static int mhi_net_newlink(struct mhi_device *mhi_dev, struct net_device *ndev)
+static int mhi_net_newlink(struct mhi_device *mhi_dev, struct net_device *ndev, bool eth_dev)
 {
 	struct mhi_net_dev *mhi_netdev;
 	int err;
 
+	if (eth_dev) {
+		eth_hw_addr_random(ndev);
+		if (!is_valid_ether_addr(ndev->dev_addr))
+			return -EADDRNOTAVAIL;
+	}
+
 	mhi_netdev = netdev_priv(ndev);
 
 	dev_set_drvdata(&mhi_dev->dev, mhi_netdev);
@@ -313,6 +362,7 @@ static int mhi_net_newlink(struct mhi_device *mhi_dev, struct net_device *ndev)
 	mhi_netdev->mdev = mhi_dev;
 	mhi_netdev->skbagg_head = NULL;
 	mhi_netdev->mru = mhi_dev->mhi_cntrl->mru;
+	mhi_netdev->ethernet_if = eth_dev;
 
 	INIT_DELAYED_WORK(&mhi_netdev->rx_refill, mhi_net_rx_refill_work);
 	u64_stats_init(&mhi_netdev->stats.rx_syncp);
@@ -365,13 +415,14 @@ static int mhi_net_probe(struct mhi_device *mhi_dev,
 	}
 
 	ndev = alloc_netdev(sizeof(struct mhi_net_dev), netname,
-			    NET_NAME_PREDICTABLE, mhi_net_setup);
+			    NET_NAME_PREDICTABLE, info->ethernet_if ?
+			    mhi_ethernet_setup : mhi_net_setup);
 	if (!ndev)
 		return -ENOMEM;
 
 	SET_NETDEV_DEV(ndev, &mhi_dev->dev);
 
-	err = mhi_net_newlink(mhi_dev, ndev);
+	err = mhi_net_newlink(mhi_dev, ndev, info->ethernet_if);
 	if (err) {
 		free_netdev(ndev);
 		return err;
@@ -389,10 +440,17 @@ static void mhi_net_remove(struct mhi_device *mhi_dev)
 
 static const struct mhi_device_info mhi_hwip0 = {
 	.netname = "mhi_hwip%d",
+	.ethernet_if = false,
 };
 
 static const struct mhi_device_info mhi_swip0 = {
 	.netname = "mhi_swip%d",
+	.ethernet_if = false,
+};
+
+static const struct mhi_device_info mhi_eth0 = {
+	.netname = "mhi_eth%d",
+	.ethernet_if = true,
 };
 
 static const struct mhi_device_id mhi_net_id_table[] = {

-- 
2.34.1

On Thu, 24 Jul 2025 19:21:18 +0530 Vivek.Pernamitta@quicinc.com wrote:
>  	ndev = alloc_netdev(sizeof(struct mhi_net_dev), netname,
> -			    NET_NAME_PREDICTABLE, mhi_net_setup);
> +			    NET_NAME_PREDICTABLE, info->ethernet_if ?
> +			    mhi_ethernet_setup : mhi_net_setup);

Similar story, if it's an Ethernet device is should be called eth%d
and user space can name it whatever it wants based on the full sysfs
attribute set.

For n-th time in our history we discover that having the kernel create
"predictable" names always falls short, sigh.