[PATCH v10 0/2] net: mhi: Add support to enable ethernet interface

[PATCH v10 0/2] net: mhi: Add support to enable ethernet interface

[Vivek Pernamitta]

Add support to configure a new client as Ethernet type over MHI
by setting "mhi_device_info.ethernet_if = true". Create a new
Ethernet interface named eth%d. This complements existing NET
driver support.

Enable IP_SW1 (ch:48/49), IP_ETH0 (ch:50,51) and IP_ETH1 (ch:52,
53) channels over MHI for M-plane, NETCONF and S-plane interface
for QDU100.

IP_ETH channels as representing the Ethernet interface of the MHI
device, and create a corresponding netdev instance to expose this
interface to the host networking stack.

M-plane:
Implement DU M-Plane software for non-real-time O-RAN
management between O-DU and O-RU using NETCONF/YANG and
O-RAN WG4 M-Plane YANG models. Provide capability exchange,
configuration management, performance monitoring, and fault
management per O-RAN.WG4.TS.MP.0-R004-v18.00.
YANG model based interface aligned with O-RAN WG4 M-Plane
specifications over TCP between the OAM application on the
host and the DU M-Plane software running on the X100 platform.

Netconf:
Use NETCONF protocol for configuration operations such as
fetching, modifying, and deleting network device
configurations.
This interface is used for IETF Netconf communication,
enabling a Netconf server on the ORU to interact with a
Netconf client running on the host.

S-plane:
Support frequency and time synchronization between O-DUs and
O-RUs using Synchronous Ethernet and IEEE 1588. Assume PTP
transport over L2 Ethernet (ITU-T G.8275.1) for full timing
support; allow PTP over UDP/IP (ITU-T G.8275.2) with reduced
reliability, as per ORAN spec O-RAN.WG4.CUS.0-R003-v12.00.
To support accurate phase and time synchronization between
the host (L2) and device (L1-High), the system must exchange
PTP messages as raw Layer-2 Ethernet frames, because the
ITU-T G.8275.1 profile operates strictly over Ethernet
multicast and not over IP networks. This means the device’s
PTP stack can only send and receive PTP Announce, Sync,
Follow-Up, and Delay messages in native Ethernet format, not
as IPv4/IPv6 packets. However, the host and device communicate
only through MHI/PCIe, which provides no native Ethernet
interface. Therefore, the system must implement a virtual
Ethernet interface over MHI on both sides. This virtual
Layer-2 link enables true Ethernet-frame transport, ensuring
the device’s PTP implementation remains fully compliant with
the G.8275.1 Ethernet-based timing model.

The actual link between the device (QDU100) and the host is
PCIe/MHI. The device has the Ethernet interface and is exposed
as the MHI channel to the host. So this patch creates the
Ethernet interface on the host based on the 'IP_ETH' channel
so that the host can use this interface for exchanging the
NETCONF packets.

The patch primarily addresses host-to-DU(QDU100) communication.
However, the NETCONF/M-Plane packets originating from the host
will eventually be transmitted from the DU to the RU over the
fronthaul, which uses Ethernet. For additional details on this
architecture and data flow, refer to the O-RAN Management
Plane Specification:

O-RAN.WG4.MP.0-v07.00
O-RAN Alliance Working Group 4 – Management Plane
Specification
Chapter 4: O-RU to O-DU Interface Management

MHI_CHANNEL_CONFIG defines channel attributes for the host
controller to set up channel rings. These entries are part of
the MHI controller’s configuration so that client drivers, such
as the MHI network driver, can attach to them. Each interface is
mapped to an MHI channel (for example, eth0 → IP_ETH0 channels
50/51), which is why this configuration resides in the bus code.

Allocate MHI netdev's using NET_NAME_ENUM to reflect kernel-enumerated
naming. This updates the reported name_assign_type for MHI net
interfaces created by this driver, aligning naming semantics across
existing and new devices. No functional or interface naming changes
are introduced.

Signed-off-by: Vivek Pernamitta <vivek.pernamitta@oss.qualcomm.com>
---
patchset link for V9 : https://lore.kernel.org/all/20260324-vdev_eth_next-20260323-v9-0-da6e4e486437@oss.qualcomm.com/
patchset link for V8 : https://lore.kernel.org/all/872d3520-f749-4d4d-a8c3-25dba30e947c@oss.qualcomm.com/#r
patchset link for V7 : https://lore.kernel.org/all/20260205-eth_vdev_next-20260204_eth-v7-0-f85645210f81@qti.qualcomm.com/
patchset link for V6 : https://lore.kernel.org/all/20251209-vdev_next-20251208_eth_v6-v6-0-80898204f5d8@quicinc.com/
patchset link for V1 (first post) : https://lore.kernel.org/all/20250724-b4-eth_us-v1-0-4dff04a9a128@quicinc.com/

changes to v10:
- Updated patchset as per comments provided by Loic Poulain
  Return -EOPNOTSUPP for non-ethernet device type which changing MAC address.
  Removed double negation in mhi_net_link for if check.

changes to v9:
- Updated patchset as per comments provided by Mani to  check
  for ARPHRD_ETHER type while assigining ethernet protocol and also
  removed explictly setting ethernet_if to false for IP_SW/IP_HW
  interfaces.
- Moved adding IP_SW1 channel to different patch in same series.
- Added more description in comments.
- As MHI(pci_generic) got merged, dropping the patch from series.
 
changes to v8:
- Removed skb_copy_to_linear_data call as it is not needed, updated
  as per Loic Poulain and Paolo Abeni.
- Removed ethernet_if member in struct mhi_net_dev instead used
  !!ndev->header_ops check for ethernet protocol as per paolo Abeni.
- Updated more infromation in commit text for change from
  NET_NAME_PREDICTABLE to NET_NAME_ENUM.

changes to v7:
- Updated to NET_NAME_ENUM while allocating netdev as per Andrew
- Updated more information as per comments from Jakub and Mani

changes to v6:
- Removed interm variable useage as per comments from Simon and Dmirty.
- Squashed gerrits 1 and 2 in single gerrit.
- Added more description for M-plane, Netconf and S-plane.

changes to v5:
- change in email ID from "quic_vpernami@quicinc.com" to "vivek.pernamitta@oss.qualcomm.com"
- Renamed to patch v5 as per comments from Manivannan
- Restored to original name as per comments from Jakub
- Renamed the ethernet interfce to eth%d as per Jakub
---

---
Vivek Pernamitta (2):
      net: mhi: Enable Ethernet interface support
      net: mhi: Add IP_SW1 interface for M-plane support over MHI

 drivers/net/mhi_net.c | 66 +++++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 54 insertions(+), 12 deletions(-)
---
base-commit: db7efce4ae23ad5e42f5f55428f529ff62b86fab
change-id: 20260409-vdev_b1_eth_b1_next-20260408-d1fb6698cae3

Best regards,
-- 
Vivek Pernamitta <vivek.pernamitta@oss.qualcomm.com>

[PATCH v10 1/2] net: mhi: Enable Ethernet interface support

[Vivek Pernamitta]

Add support to configure a new client as Ethernet type over MHI by
setting "mhi_device_info.ethernet_if = true". Create a new Ethernet
interface named eth%d. This complements existing NET driver support.

Allocate MHI netdevs using NET_NAME_ENUM to reflect kernel-enumerated
naming. This updates the reported name_assign_type for MHI net
interfaces created by this driver, aligning naming semantics across
existing and new devices. No functional or interface naming changes
are introduced

Introduce ETH0 and ETH1 network interfaces required for NETCONF
and S-plane components.

IP_ETH channels represent the Ethernet interface exposed by the
MHI device. The driver creates a corresponding netdev instance,
allowing the host to access and operate the device’s Ethernet
interface.

NETCONF:
Use NETCONF protocol for configuration operations such as fetching,
modifying, and deleting network device configurations.

S-plane:
Support frequency and time synchronization between O-DUs and O-RUs
using Synchronous Ethernet and IEEE 1588. Assume PTP transport over
L2 Ethernet (ITU-T G.8275.1) for full timing support; allow PTP over
UDP/IP (ITU-T G.8275.2) with reduced reliability. as per ORAN spec
O-RAN.WG4.CUS.0-R003-v12.00.

Signed-off-by: Vivek Pernamitta <vivek.pernamitta@oss.qualcomm.com>
---
 drivers/net/mhi_net.c | 65 +++++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 53 insertions(+), 12 deletions(-)

diff --git a/drivers/net/mhi_net.c b/drivers/net/mhi_net.c
index ae169929a9d8e449b5a427993abf68e8d032fae2..47b8617de027980a69c57261a9b4bcefc828dc96 100644
--- a/drivers/net/mhi_net.c
+++ b/drivers/net/mhi_net.c
@@ -4,6 +4,7 @@
  * Copyright (C) 2020 Linaro Ltd <loic.poulain@linaro.org>
  */
 
+#include <linux/etherdevice.h>
 #include <linux/if_arp.h>
 #include <linux/mhi.h>
 #include <linux/mod_devicetable.h>
@@ -42,6 +43,7 @@ struct mhi_net_dev {
 
 struct mhi_device_info {
 	const char *netname;
+	bool ethernet_if;
 };
 
 static int mhi_ndo_open(struct net_device *ndev)
@@ -119,11 +121,29 @@ 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)
+{
+	if (dev->type == ARPHRD_ETHER)
+		return eth_mac_addr(dev, p);
+
+	return -EOPNOTSUPP;
+}
+
+static int mhi_validate_address(struct net_device *dev)
+{
+	if (dev->type == ARPHRD_ETHER)
+		return eth_validate_addr(dev);
+
+	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 +160,13 @@ 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->max_mtu = ETH_MAX_MTU;
+}
+
 static struct sk_buff *mhi_net_skb_agg(struct mhi_net_dev *mhi_netdev,
 				       struct sk_buff *skb)
 {
@@ -208,17 +235,20 @@ static void mhi_net_dl_callback(struct mhi_device *mhi_dev,
 			skb = mhi_net_skb_agg(mhi_netdev, skb);
 			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->ndev->type == ARPHRD_ETHER) {
+			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);
@@ -306,6 +336,9 @@ static int mhi_net_newlink(struct mhi_device *mhi_dev, struct net_device *ndev)
 	struct mhi_net_dev *mhi_netdev;
 	int err;
 
+	if (ndev->header_ops)
+		eth_hw_addr_random(ndev);
+
 	mhi_netdev = netdev_priv(ndev);
 
 	dev_set_drvdata(&mhi_dev->dev, mhi_netdev);
@@ -356,7 +389,8 @@ static int mhi_net_probe(struct mhi_device *mhi_dev,
 	int err;
 
 	ndev = alloc_netdev(sizeof(struct mhi_net_dev), info->netname,
-			    NET_NAME_PREDICTABLE, mhi_net_setup);
+			    NET_NAME_ENUM, info->ethernet_if ?
+			    mhi_ethernet_setup : mhi_net_setup);
 	if (!ndev)
 		return -ENOMEM;
 
@@ -386,11 +420,18 @@ static const struct mhi_device_info mhi_swip0 = {
 	.netname = "mhi_swip%d",
 };
 
+static const struct mhi_device_info mhi_eth0 = {
+	.netname = "eth%d",
+	.ethernet_if = true,
+};
+
 static const struct mhi_device_id mhi_net_id_table[] = {
 	/* Hardware accelerated data PATH (to modem IPA), protocol agnostic */
 	{ .chan = "IP_HW0", .driver_data = (kernel_ulong_t)&mhi_hwip0 },
 	/* Software data PATH (to modem CPU) */
 	{ .chan = "IP_SW0", .driver_data = (kernel_ulong_t)&mhi_swip0 },
+	{ .chan = "IP_ETH0", .driver_data = (kernel_ulong_t)&mhi_eth0 },
+	{ .chan = "IP_ETH1", .driver_data = (kernel_ulong_t)&mhi_eth0 },
 	{}
 };
 MODULE_DEVICE_TABLE(mhi, mhi_net_id_table);

-- 
2.34.1

Re: [PATCH v10 1/2] net: mhi: Enable Ethernet interface support

[Paolo Abeni]

On 4/9/26 8:08 AM, Vivek Pernamitta wrote:
> @@ -208,17 +235,20 @@ static void mhi_net_dl_callback(struct mhi_device *mhi_dev,
>  			skb = mhi_net_skb_agg(mhi_netdev, skb);
>  			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->ndev->type == ARPHRD_ETHER) {
> +			skb->protocol = eth_type_trans(skb, mhi_netdev->ndev);

Sashiko says:

Is there a risk of an out-of-bounds read or kernel panic here if a malformed
or fragmented packet is received?
eth_type_trans() assumes the SKB has at least a 14-byte MAC header and calls
skb_pull_inline(). If the linear portion of the SKB is smaller than 14
bytes,
__skb_pull() will trigger a BUG_ON(skb->len < skb->data_len).
Should we call pskb_may_pull(skb, ETH_HLEN) before parsing the Ethernet
header?

> +		} 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);
> @@ -306,6 +336,9 @@ static int mhi_net_newlink(struct mhi_device *mhi_dev, struct net_device *ndev)
>  	struct mhi_net_dev *mhi_netdev;
>  	int err;
>  
> +	if (ndev->header_ops)
> +		eth_hw_addr_random(ndev);
> +
>  	mhi_netdev = netdev_priv(ndev);
>  
>  	dev_set_drvdata(&mhi_dev->dev, mhi_netdev);
> @@ -356,7 +389,8 @@ static int mhi_net_probe(struct mhi_device *mhi_dev,
>  	int err;
>  
>  	ndev = alloc_netdev(sizeof(struct mhi_net_dev), info->netname,
> -			    NET_NAME_PREDICTABLE, mhi_net_setup);
> +			    NET_NAME_ENUM, info->ethernet_if ?
> +			    mhi_ethernet_setup : mhi_net_setup);

Sashiko says:

Does changing the name assignment type from NET_NAME_PREDICTABLE to
NET_NAME_ENUM break backwards compatibility for existing legacy interfaces?
NET_NAME_PREDICTABLE instructs userspace to leave the kernel-provided
name alone, while NET_NAME_ENUM signals that the interface is a generic
enumeration and should be renamed. Applying this to existing interfaces
like mhi_hwip0 and mhi_swip0 might cause them to be unexpectedly renamed
on boot, potentially breaking existing userspace network configurations.

please have a look at the full report:

https://sashiko.dev/#/patchset/20260409-vdev_b1_eth_b1_next-20260408-v10-0-6d44ca48f189%40oss.qualcomm.com

/P

[PATCH v10 2/2] net: mhi: Add IP_SW1 interface for M-plane support over MHI

[Vivek Pernamitta]

Introduce the IP_SW1 channel to represent the network interface
exposed by the MHI device for M-plane communication. The driver
creates a corresponding netdev, allowing the host to access the
device network interface over MHI.

This interface enables support for O-RAN M-plane management
between O-DU and O-RU, including capability exchange, configuration
management, performance monitoring, and fault management using
NETCONF/YANG, as defined by O-RAN WG4 M-plane specifications.

Signed-off-by: Vivek Pernamitta <vivek.pernamitta@oss.qualcomm.com>
---
 drivers/net/mhi_net.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/mhi_net.c b/drivers/net/mhi_net.c
index 47b8617de027980a69c57261a9b4bcefc828dc96..e285aa289cb945cd1afd2f3d581214f2f2f15145 100644
--- a/drivers/net/mhi_net.c
+++ b/drivers/net/mhi_net.c
@@ -430,6 +430,7 @@ static const struct mhi_device_id mhi_net_id_table[] = {
 	{ .chan = "IP_HW0", .driver_data = (kernel_ulong_t)&mhi_hwip0 },
 	/* Software data PATH (to modem CPU) */
 	{ .chan = "IP_SW0", .driver_data = (kernel_ulong_t)&mhi_swip0 },
+	{ .chan = "IP_SW1", .driver_data = (kernel_ulong_t)&mhi_swip0 },
 	{ .chan = "IP_ETH0", .driver_data = (kernel_ulong_t)&mhi_eth0 },
 	{ .chan = "IP_ETH1", .driver_data = (kernel_ulong_t)&mhi_eth0 },
 	{}

-- 
2.34.1