bonding: LAG with outgoing traffic distribution based on h/w

Linux bonding driver provides a method for aggregating multiple network
interface controllers (NICs) into a single logical bonded interface of
two or more so called (NIC) slaves. Slave selection for outgoing traffic
is done according to the transmit hash policy, which may be changed from
the default simple XOR policy via the xmit_hash_policy option. This
selection algorithm in Linux bonding driver is based on software.

The QorIQ Data Path Acceleration Architecture (DPAA) is a comprehensive
architecture which integrates all aspects of packet processing in the
SoC, addressing issues and requirements resulting from the multicore
nature of QorIQ SoCs. The DPAA includes Cores, Network and packet I/O,
Hardware offload accelerators. Hardware offload accelerators include
FMan/BMan/QMan and etc.

Offline port is one of FMan elements, which supports (Parse, Classify,
Distribute) PCD function on frames extracted frame descriptor (FD).
Offline port also can inspect traffic, split it into FQs on ingress, and
send traffic from the FQs to the interface on egress by the PCD function.

These patches are enhancing Linux kernel LAG (Link Aggregation) with
Freescale DPAA value added. The main idea is to utilize offline ports
with PCD function to help to distribute outgoing traffics, including
outgoing slaves device searching and selection. In another world,
patches are using CRC-64 based hashing of Keygen/scheme and the parser
result of outgoing frames header information to distribute outgoing
frames.

Beside of above, after integration this HW based LAG with Freescale
CEETMQos, these two features can support hardware based Qos for bundles
links rather than individual links.

These patches mainly include 2 parts:
'glue logic' and 'kernel space PCD'.

The glue logic first probes all available offline ports information via
reading dts, including tx fqid/default fqid/errors fqid, pcd fqs, other
private data pointer of offline ports for future reusing. The glue logic
also creates frames from skb and then sends these frames to offline port
directly, this offline port will continue to distribution frames from
the PCD FQs to the slave interface on egress by the PCD function, rather
than select slave device by CPU, neither make slave device driver create
frame from skb, nor make slave devices driver send frames.

These patches are supporting the mapping among offline ports and
available bundles at run-time. PCD based outgoing traffic distribution
can be enabled or disabled at run-time by sysfs interface in patches.

To do:
1. PCD policy L23/L34 have not been veryfied.
2. offline port buffer pool/buffer layout will be enhanced.
3. software based L4 csum for now, offline port based L4 csum need be fixed.

To test this HW based LAG after booting up Linux:
	cd /sys/class/net/bond0/bonding/
	echo 4 >mode
	cat offline_ports
	echo fman0-oh@1 > oh_needed_for_hw_distribution
	cat oh_needed_for_hw_distribution
	cat oh_en
	echo 1 >oh_en
	cat oh_en
	echo +fm1-gb0 >slaves
	echo +fm1-gb1 >slaves
	ifconfig bond0 192.168.10.2/24 up
	ping 192.168.10.1 -c 5

Signed-off-by: Jianhua Xie <jianhua.xie@freescale.com>
Change-Id: I3a6664bfcc9ec9ca3f86a5e36381220c5fcb07cf
Reviewed-on: http://git.am.freescale.net:8181/10413
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Jiafei Pan <Jiafei.Pan@freescale.com>
Reviewed-by: Jose Rivera <German.Rivera@freescale.com>

7 files changed, 2501 insertions, 3 deletions

diff --git a/drivers/net/bonding/Makefile b/drivers/net/bonding/Makefile
index 5a5d720..0024f07 100644
--- a/drivers/net/bonding/Makefile
+++ b/drivers/net/bonding/Makefile
@@ -2,6 +2,13 @@
 # Makefile for the Ethernet Bonding driver
 #
 
+include $(srctree)/drivers/net/ethernet/freescale/fman/ncsw_config.mk
+ccflags-y += \
+	-I$(srctree)/drivers/net/ethernet/freescale/dpa \
+	-I$(srctree)/drivers/net/ethernet/freescale/fman/src/wrapper \
+	-I$(srctree)/drivers/net/ethernet/freescale/fman/Peripherals/FM/Pcd \
+	-I$(srctree)/drivers/net/ethernet/freescale/fman/Peripherals/FM/inc
+
 obj-$(CONFIG_BONDING) += bonding.o
 
 bonding-objs := bond_main.o bond_3ad.o bond_alb.o bond_sysfs.o bond_debugfs.o bond_netlink.o bond_options.o
@@ -9,3 +16,7 @@ bonding-objs := bond_main.o bond_3ad.o bond_alb.o bond_sysfs.o bond_debugfs.o bo
 proc-$(CONFIG_PROC_FS) += bond_procfs.o
 bonding-objs += $(proc-y)
 
+hash_pcd_based_xmit_frames_distribution-$(CONFIG_HW_DISTRIBUTION_WITH_OH) += \
+	hw_distribution.o
+
+bonding-objs += $(hash_pcd_based_xmit_frames_distribution-y)
diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c
index 187b1b7..698ac84 100644
--- a/drivers/net/bonding/bond_3ad.c
+++ b/drivers/net/bonding/bond_3ad.c
@@ -2402,6 +2402,15 @@ int bond_3ad_xmit_xor(struct sk_buff *skb, struct net_device *dev)
 		goto out;
 	}
 
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	hw_lag_dbg("skb->protocol:0x%0x\n", skb->protocol);
+	/* exclude ARP/LACP */
+	if ((bond->slave_cnt == SLAVES_PER_BOND) &&
+			are_all_slaves_linkup(bond) &&
+			(bond->params.ohp) && (bond->params.ohp->oh_en == 1))
+		return enqueue_pkt_to_oh(bond, skb, NULL);
+#endif
+
 	slave_agg_no = bond_xmit_hash(bond, skb, slaves_in_agg);
 	first_ok_slave = NULL;
 
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 36eab0c..77619a6 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1607,6 +1607,10 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
 		new_slave->link != BOND_LINK_DOWN ? "n up" : " down");
 
 	/* enslave is successful */
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	fill_oh_pcd_fqs_with_slave_info(bond, new_slave);
+	apply_pcd(bond, NO_POLICY);
+#endif
 	return 0;
 
 /* Undo stages on error */
@@ -1848,6 +1852,9 @@ static int __bond_release_one(struct net_device *bond_dev,
 
 	slave_dev->priv_flags &= ~IFF_BONDING;
 
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	del_oh_pcd_fqs_with_slave_info(bond, slave);
+#endif
 	kfree(slave);
 
 	return 0;  /* deletion OK */
@@ -3164,6 +3171,10 @@ static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
 		stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors;
 		stats->tx_window_errors += sstats->tx_window_errors;
 	}
+
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	add_statistics(bond, stats);
+#endif
 	read_unlock_bh(&bond->lock);
 
 	return stats;
@@ -3890,6 +3901,9 @@ static const struct device_type bond_type = {
 static void bond_destructor(struct net_device *bond_dev)
 {
 	struct bonding *bond = netdev_priv(bond_dev);
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	release_pcd_mem(bond);
+#endif
 	if (bond->wq)
 		destroy_workqueue(bond->wq);
 	free_netdev(bond_dev);
@@ -4436,7 +4450,10 @@ int bond_create(struct net *net, const char *name)
 	res = register_netdevice(bond_dev);
 
 	netif_carrier_off(bond_dev);
-
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	if (res == 0)
+		init_status(bond_dev);
+#endif
 	rtnl_unlock();
 	if (res < 0)
 		bond_destructor(bond_dev);
@@ -4508,6 +4525,13 @@ static int __init bonding_init(void)
 	}
 
 	register_netdevice_notifier(&bond_netdev_notifier);
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	if (get_oh_info())
+		pr_err("oh ports probe error, use software distribution\n");
+	else
+		pr_info("get offline ports information ok.\n");
+#endif
+
 out:
 	return res;
 err:
@@ -4526,6 +4550,10 @@ static void __exit bonding_exit(void)
 
 	bond_netlink_fini();
 	unregister_pernet_subsys(&bond_net_ops);
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	kfree(poh);
+	hw_lag_dbg("released offline port resources\n");
+#endif
 
 #ifdef CONFIG_NET_POLL_CONTROLLER
 	/*
@@ -4534,8 +4562,13 @@ static void __exit bonding_exit(void)
 	WARN_ON(atomic_read(&netpoll_block_tx));
 #endif
 }
-
-module_init(bonding_init);
+/**
+ * late init to wait till oh port initilization ready,
+ * oh port can help distribute outgoing traffics based
+ * on hardware (FSL DPAA Offline port and PCD).
+ * module_init(bonding_init);
+ */
+late_initcall(bonding_init);
 module_exit(bonding_exit);
 MODULE_LICENSE("GPL");
 MODULE_VERSION(DRV_VERSION);
diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index abf5e10..1e591c4 100644
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -329,6 +329,9 @@ static ssize_t bonding_store_xmit_hash(struct device *d,
 		       (int)strlen(buf) - 1, buf);
 		ret = -EINVAL;
 	} else {
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+		apply_pcd(bond, new_value);
+#endif
 		bond->params.xmit_policy = new_value;
 		pr_info("%s: setting xmit hash policy to %s (%d).\n",
 			bond->dev->name,
@@ -883,6 +886,21 @@ static ssize_t bonding_store_min_links(struct device *d,
 }
 static DEVICE_ATTR(min_links, S_IRUGO | S_IWUSR,
 		   bonding_show_min_links, bonding_store_min_links);
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+static DEVICE_ATTR(offline_port_xmit_statistics, S_IRUGO,
+		bonding_show_offline_port_xmit_statistics, NULL);
+
+static DEVICE_ATTR(offline_ports, S_IRUGO,
+		bonding_show_offline_ports, NULL);
+
+static DEVICE_ATTR(oh_needed_for_hw_distribution, S_IRUGO | S_IWUSR,
+		bonding_show_oh_needed_for_hw_distribution,
+		bonding_store_oh_needed_for_hw_distribution);
+
+static DEVICE_ATTR(oh_en, S_IRUGO | S_IWUSR,
+		bonding_show_oh_enable,
+		bonding_store_oh_enable);
+#endif
 
 static ssize_t bonding_show_ad_select(struct device *d,
 				      struct device_attribute *attr,
@@ -1709,6 +1727,12 @@ static struct attribute *per_bond_attrs[] = {
 	&dev_attr_min_links.attr,
 	&dev_attr_lp_interval.attr,
 	&dev_attr_packets_per_slave.attr,
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	&dev_attr_offline_ports.attr,
+	&dev_attr_oh_needed_for_hw_distribution.attr,
+	&dev_attr_oh_en.attr,
+	&dev_attr_offline_port_xmit_statistics.attr,
+#endif
 	NULL,
 };
 
diff --git a/drivers/net/bonding/bonding.h b/drivers/net/bonding/bonding.h
index a9f4f9f..a606ea0 100644
--- a/drivers/net/bonding/bonding.h
+++ b/drivers/net/bonding/bonding.h
@@ -25,6 +25,9 @@
 #include <linux/etherdevice.h>
 #include "bond_3ad.h"
 #include "bond_alb.h"
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+#include "hw_distribution.h"
+#endif
 
 #define DRV_VERSION	"3.7.1"
 #define DRV_RELDATE	"April 27, 2011"
@@ -167,6 +170,10 @@ struct bond_params {
 	int resend_igmp;
 	int lp_interval;
 	int packets_per_slave;
+#ifdef CONFIG_HW_DISTRIBUTION_WITH_OH
+	struct oh_port_priv *ohp;
+	struct rtnl_link_stats64 oh_stats;
+#endif
 };
 
 struct bond_parm_tbl {
diff --git a/drivers/net/bonding/hw_distribution.c b/drivers/net/bonding/hw_distribution.c
new file mode 100644
index 0000000..dce0ea1
--- /dev/null
+++ b/drivers/net/bonding/hw_distribution.c
@@ -0,0 +1,2276 @@
+/* Copyright 2014 Freescale Semiconductor Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *     * Redistributions of source code must retain the above copyright
+ *	 notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *	 notice, this list of conditions and the following disclaimer in the
+ *	 documentation and/or other materials provided with the distribution.
+ *     * Neither the name of Freescale Semiconductor nor the
+ *	 names of its contributors may be used to endorse or promote products
+ *	 derived from this software without specific prior written permission.
+ *
+ *
+ * ALTERNATIVELY, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") as published by the Free Software
+ * Foundation, either version 2 of that License or (at your option) any
+ * later version.
+ *
+ * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY
+ * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY
+ * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include <linux/kthread.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/udp.h>
+#include <linux/ipv6.h>
+#include <linux/if_vlan.h>
+#include <net/ip.h>
+#include "hw_distribution.h"
+#include "mac.h"
+#define to_dev(obj)     container_of(obj, struct device, kobj)
+#define to_bond(cd)     ((struct bonding *)(netdev_priv(to_net_dev(cd))))
+#define master_to_bond(net_dev)     ((struct bonding *)(netdev_priv(net_dev)))
+/**
+ * This includes L4 checksum errors, but also other errors that
+ * the Hard Parser can detect, such as invalid combinations of
+ * TCP control flags, or bad UDP lengths.
+ */
+#define FM_L4_PARSE_ERROR      0x10
+/* Check if the hardware parser has run */
+#define FM_L4_HXS_RUN          0xE0
+/**
+ * Check if the FMan Hardware Parser has run for L4 protocols.
+ * @parse_result_ptr must be of type (fm_prs_result_t *).
+ */
+#define fm_l4_hxs_has_run(parse_result_ptr) \
+	((parse_result_ptr)->l4r & FM_L4_HXS_RUN)
+/**
+ * If the FMan Hardware Parser has run for L4 protocols, check
+ * error status.
+ * @parse_result_ptr must be of type fm_prs_result_t *).
+ */
+#define fm_l4_hxs_error(parse_result_ptr) \
+	       ((parse_result_ptr)->l4r & FM_L4_PARSE_ERROR)
+
+#define DPA_WRITE_SKB_PTR(skb, skbh, addr, off) \
+	{ \
+		skbh = (struct sk_buff **)addr; \
+		*(skbh + (off)) = skb; \
+	}
+
+#define DPA_READ_SKB_PTR(skb, skbh, addr, off) \
+	{ \
+		skbh = (struct sk_buff **)addr; \
+		skb = *(skbh + (off)); \
+	}
+
+static const struct of_device_id oh_port_match_table[] = {
+	{
+		.compatible	= "fsl,dpa-oh"
+	},
+	{}
+};
+
+struct oh_port_priv *poh; /* Offline port information pointer */
+int available_num_of_oh_ports;
+/**
+ * Sysfs interfaces:
+ * Show the statistics information by offline port xmit.
+ */
+ssize_t bonding_show_offline_port_xmit_statistics(struct device *d,
+		struct device_attribute *attr, char *buf)
+{
+	int res = 0;
+	struct bonding *bond = to_bond(d);
+
+	if (bond->params.mode != BOND_MODE_8023AD) {
+		pr_err("%s: This command only support 802.3ad mode.\n",
+				bond->dev->name);
+		return -EPERM;
+	}
+
+	if (!bond->params.ohp) {
+		pr_err("error, have not bind an offline port\n");
+
+		return -EPERM;
+	}
+
+	if (!bond->params.ohp->oh_en) {
+		pr_err("error, binded offline port is not enabled.\n");
+
+		return -EPERM;
+	}
+
+	res += sprintf(buf + res, "offline port TX packets: %llu\n",
+			bond->params.oh_stats.tx_packets);
+	res += sprintf(buf + res, "offline port TX bytes: %llu\n",
+			bond->params.oh_stats.tx_bytes);
+	res += sprintf(buf + res, "offline port TX errors: %llu\n",
+			bond->params.oh_stats.tx_errors);
+	res += sprintf(buf + res, "offline port TX dropped: %llu\n",
+			bond->params.oh_stats.tx_dropped);
+
+	if (res)
+		buf[res-1] = '\n'; /* eat the leftover space */
+	return res;
+}
+/**
+ * Sysfs interfaces:
+ * Show all available offline ports can be binded to a bond.
+ */
+ssize_t bonding_show_offline_ports(struct device *d,
+		struct device_attribute *attr, char *buf)
+{
+	int i, res = 0;
+	struct bonding *bond = to_bond(d);
+
+	if (bond->params.mode != BOND_MODE_8023AD) {
+		pr_err("%s: This command only support 802.3ad mode.\n",
+				bond->dev->name);
+		return -EPERM;
+	}
+
+	for (i = 0; i < available_num_of_oh_ports; i++) {
+		if (poh[i].oh_dev)
+			res += sprintf(buf + res, "%s\n", poh[i].friendname);
+	}
+	if (res)
+		buf[res-1] = '\n'; /* eat the leftover space */
+	return res;
+}
+/**
+ * Sysfs interfaces:
+ * Show the offline_port has already attached to the current bond,
+ * which can help bond to do hardware based slave selection.
+ */
+ssize_t bonding_show_oh_needed_for_hw_distribution(struct device *d,
+		struct device_attribute *attr, char *buf)
+{
+	int res = 0;
+	struct bonding *bond = to_bond(d);
+
+	if (bond->params.mode != BOND_MODE_8023AD) {
+		pr_err("%s: This command only support 802.3ad mode.\n",
+				bond->dev->name);
+		return -EPERM;
+	}
+
+	if (!bond->params.ohp) {
+		pr_err("error, have not bind an offline port\n");
+
+		return -EPERM;
+	}
+
+	res += sprintf(buf + res, "%s\n", bond->params.ohp->friendname);
+	if (res)
+		buf[res-1] = '\n'; /* eat the leftover space */
+
+	return res;
+}
+/**
+ * System interface:
+ * Add one Offline port into the current bond for utilizing PCD to
+ * do TX traffic distribution based on hard ware.
+ * This codes firt verify the input Offline port name validation,
+ * then store the Offline port to the current bond->params.
+ */
+ssize_t bonding_store_oh_needed_for_hw_distribution(struct device *d,
+		struct device_attribute *attr, const char *buffer, size_t count)
+{
+	char command[OHFRIENDNAMSIZ + 1] = { 0, };
+	int ret = count, i, errno;
+	struct bonding *bond = to_bond(d);
+	struct oh_port_priv *tmp = poh;
+	bool find = false;
+
+	if (bond->params.mode != BOND_MODE_8023AD) {
+		pr_err("%s: This command only support 802.3ad mode.\n",
+				bond->dev->name);
+		return -EPERM;
+	}
+	if (bond->slave_cnt > 0) {
+		pr_err("%s: Detach slaves before change oh binding.\n",
+				bond->dev->name);
+		return -EPERM;
+	}
+
+	if (!rtnl_trylock())
+		return restart_syscall();
+
+	/* OHFRIENDNAMSIZ = 10, there is 10 chars in a command. */
+	errno = sscanf(buffer, "%10s", command);
+	if ((strlen(command) <= 1) || (errno != 1))
+		goto err_no_cmd;
+
+	if ((bond->params.ohp) && (bond->params.ohp->friendname[0]) &&
+		strncasecmp(command, bond->params.ohp->friendname,
+			OHFRIENDNAMSIZ) == 0) {
+				pr_err("%s: has already used %s.\n",
+					bond->dev->name, command);
+		ret = -EPERM;
+		goto out;
+	} else
+		for (i = 0; i < available_num_of_oh_ports; i++) {
+			if (tmp->oh_dev) {
+				if (strncasecmp(command, tmp->friendname,
+						OHFRIENDNAMSIZ) == 0) {
+					find = true;
+					bond->params.ohp = tmp;
+					break;
+				} else
+					tmp++;
+			}
+		}
+
+	if (!find)
+		goto err_no_cmd;
+
+	pr_info("bind OH port oh_needed_for_hw_distribution: %s to %s\n",
+			bond->params.ohp->friendname, bond->dev->name);
+
+	goto out;
+
+err_no_cmd:
+	pr_err("%s:bad command or no such OH port,\n"
+			"please try other OH ports.\n"
+			"Eg: echo OH0 > oh_needed_for_hw_distribution.\n",
+			bond->dev->name);
+	ret = -EPERM;
+
+out:
+	rtnl_unlock();
+	return ret;
+}
+/**
+ * Sysfs interfaces:
+ * Show whether current offline port binding to the bond is active or not.
+ */
+ssize_t bonding_show_oh_enable(struct device *d,
+		struct device_attribute *attr, char *buf)
+{
+
+	int res = 0;
+	struct bonding *bond = to_bond(d);
+	uint16_t channel;
+	unsigned long fman_dcpid, oh_offset, cell_index;
+
+	if (bond->params.mode != BOND_MODE_8023AD) {
+		pr_err("%s: This command only support 802.3ad mode.\n",
+				bond->dev->name);
+		return -EPERM;
+	}
+
+	if (!bond->params.ohp) {
+		pr_err("error, have not bind a offline port\n");
+
+		return -EPERM;
+	}
+
+	res += sprintf(buf + res, "%d\n", bond->params.ohp->oh_en);
+	if (res)
+		buf[res-1] = '\n'; /* eat the leftover space */
+
+	if ((bond->params.ohp->oh_en) &&
+		(!export_oh_port_info_to_ceetm(bond, &channel,
+			&fman_dcpid, &oh_offset, &cell_index)))
+		hw_lag_dbg("offline port channel:%d\n", channel);
+
+	return res;
+}
+/**
+ * Sysfs interfaces:
+ * Set current offline port which is binding to the bond active or not,
+ * this interface can disable or enable the offline port which is binding
+ * to a bond at run-time.
+ */
+ssize_t bonding_store_oh_enable(struct device *d,
+		struct device_attribute *attr, const char *buffer,
+		size_t count)
+{
+	int new_value, ret;
+	struct bonding *bond = to_bond(d);
+
+	if (bond->params.mode != BOND_MODE_8023AD) {
+		pr_err("%s: This command only support 802.3ad mode.\n",
+				bond->dev->name);
+		return -EPERM;
+	}
+
+	ret = sscanf(buffer, "%d", &new_value);
+	pr_info("new_value:%d, ret: %d\n", new_value, ret);
+	if (ret != 1) {
+		pr_err("%s: Bad command, use echo [1|0] > oh_en.\n",
+				bond->dev->name);
+		return -EINVAL;
+	}
+
+	if (!bond->params.ohp) {
+		pr_err("error, have not bind a offline port\n");
+		return -EPERM;
+	}
+
+	if ((new_value == 0) || (new_value == 1)) {
+		bond->params.ohp->oh_en = new_value;
+		return count;
+	} else {
+		pr_err("%s: Bad value, only is 1 or 0.\n",
+				bond->dev->name);
+		return -EINVAL;
+	}
+}
+
+/**
+ * Judge a slave net device is a dpa-eth NIC,
+ * return true if it is a dpa-eth NIC,
+ * otherwise return false.
+ */
+static bool is_dpa_eth_port(struct net_device *netdev)
+{
+	struct device *dev = (struct device *) &(netdev->dev);
+
+	if ((strlen(dev_driver_string(dev->parent)) >= 7) &&
+		strncmp(dev_driver_string(dev->parent), "fsl_dpa", 7) == 0)
+		return true;
+	else
+		return false;
+}
+
+bool are_all_slaves_linkup(struct bonding *bond)
+{
+	struct slave *slave;
+	struct list_head *iter;
+
+	read_lock(&bond->lock);
+	bond_for_each_slave(bond, slave, iter)
+		if (!(SLAVE_IS_OK(slave))) {
+			read_unlock(&bond->lock);
+			return false;
+		}
+
+	read_unlock(&bond->lock);
+	return true;
+}
+
+unsigned int to_which_oh_i_attached(struct oh_port_priv *current_poh)
+{
+	struct oh_port_priv *org = poh;
+	int i = 0;
+	while (current_poh - org) {
+		i++;
+		org++;
+	}
+
+	return i;
+}
+/* Borrowed from dpa_fd_release, removed netdev params. */
+static void __attribute__((nonnull))
+dpa_oh_fd_release(const struct qm_fd *fd)
+{
+	struct qm_sg_entry *sgt;
+	struct dpa_bp *dpa_bp;
+	struct bm_buffer bmb;
+
+	bmb.hi = fd->addr_hi;
+	bmb.lo = fd->addr_lo;
+
+	dpa_bp = dpa_bpid2pool(fd->bpid);
+	DPA_BUG_ON(!dpa_bp);
+
+	if (fd->format == qm_fd_sg) {
+		sgt = (phys_to_virt(bm_buf_addr(&bmb)) + dpa_fd_offset(fd));
+		dpa_release_sgt(sgt);
+	}
+
+	while (bman_release(dpa_bp->pool, &bmb, 1, 0))
+		cpu_relax();
+}
+
+static void dpa_oh_drain_bp(struct dpa_bp *bp)
+{
+	int i, num;
+	struct bm_buffer bmb[8];
+	dma_addr_t addr;
+	int *countptr = __this_cpu_ptr(bp->percpu_count);
+	int count = *countptr;
+	struct sk_buff **skbh;
+
+	while (count >= 8) {
+		num = bman_acquire(bp->pool, bmb, 8, 0);
+		/* There may still be up to 7 buffers in the pool;
+		 * just leave them there until more arrive
+		 */
+		if (num < 0)
+			break;
+		for (i = 0; i < num; i++) {
+			addr = bm_buf_addr(&bmb[i]);
+			/* bp->free_buf_cb(phys_to_virt(addr)); */
+			skbh = (struct sk_buff **)phys_to_virt(addr);
+			dma_unmap_single(bp->dev, addr, bp->size,
+					 DMA_TO_DEVICE);
+			dev_kfree_skb(*skbh);
+		}
+		count -= num;
+	}
+	*countptr = count;
+}
+static int dpa_oh_tx_bp_probe(struct device *dev,
+			    struct fm_port *tx_port,
+			    struct dpa_bp **draining_tx_bp,
+			    struct dpa_buffer_layout_s **tx_buf_layout)
+{
+	struct fm_port_params params;
+	struct dpa_bp *bp = NULL;
+	struct dpa_buffer_layout_s *buf_layout = NULL;
+
+	buf_layout = devm_kzalloc(dev, sizeof(*buf_layout), GFP_KERNEL);
+	if (!buf_layout) {
+		dev_err(dev, "devm_kzalloc() failed\n");
+		return -ENOMEM;
+	}
+
+	buf_layout->priv_data_size = DPA_TX_PRIV_DATA_SIZE;
+	buf_layout->parse_results = true;
+	buf_layout->hash_results = true;
+	buf_layout->time_stamp = false;
+
+	fm_port_get_buff_layout_ext_params(tx_port, &params);
+	buf_layout->manip_extra_space = params.manip_extra_space;
+	buf_layout->data_align = params.data_align ? : DPA_FD_DATA_ALIGNMENT;
+
+	bp = devm_kzalloc(dev, sizeof(*bp), GFP_KERNEL);
+	if (unlikely(bp == NULL)) {
+		dev_err(dev, "devm_kzalloc() failed\n");
+		return -ENOMEM;
+	}
+
+	bp->size = dpa_bp_size(buf_layout);
+	bp->percpu_count = alloc_percpu(*bp->percpu_count);
+	bp->target_count = CONFIG_FSL_DPAA_ETH_MAX_BUF_COUNT;
+
+	*draining_tx_bp = bp;
+	*tx_buf_layout = buf_layout;
+
+	return 0;
+}
+static int dpa_oh_bp_create(struct oh_port_priv *ohp)
+{
+	int err = 0;
+	struct dpa_bp *draining_tx_bp;
+	struct dpa_buffer_layout_s *tx_buf_layout;
+
+	err = dpa_oh_tx_bp_probe(ohp->dpa_oh_dev, ohp->oh_config->oh_port,
+			&draining_tx_bp, &tx_buf_layout);
+	if (err) {
+		pr_err("errors on dpa_oh_tx_bp_probe()\n");
+		return err;
+	}
+
+	ohp->tx_bp = draining_tx_bp;
+	ohp->tx_buf_layout = tx_buf_layout;
+
+	err = dpa_bp_alloc(ohp->tx_bp);
+	if (err < 0) {
+		/* _dpa_bp_free(ohp->tx_bp); */
+		pr_err("error on dpa_bp_alloc()\n");
+		ohp->tx_bp = NULL;
+		return err;
+	}
+	hw_lag_dbg("created bp, bpid(ohp->tx_bp):%d\n", ohp->tx_bp->bpid);
+
+	return 0;
+}
+/**
+ * Copied from DPA-Eth driver (since they have different params type):
+ * Cleanup function for outgoing frame descriptors that were built on Tx path,
+ * either contiguous frames or scatter/gather ones.
+ * Skb freeing is not handled here.
+ *
+ * This function may be called on error paths in the Tx function, so guard
+ * against cases when not all fd relevant fields were filled in.
+ *
+ * Return the skb backpointer, since for S/G frames the buffer containing it
+ * gets freed here.
+ */
+struct sk_buff *oh_cleanup_tx_fd(const struct qm_fd *fd)
+{
+	int i, nr_frags;
+	const struct qm_sg_entry *sgt;
+	struct sk_buff **skbh;
+	struct sk_buff *skb = NULL;
+	dma_addr_t addr = qm_fd_addr(fd);
+	struct dpa_bp *dpa_bp = dpa_bpid2pool(fd->bpid);
+	const enum dma_data_direction dma_dir = DMA_TO_DEVICE;
+
+	DPA_BUG_ON(fd->cmd & FM_FD_CMD_FCO);
+	dma_unmap_single(dpa_bp->dev, addr, dpa_bp->size, dma_dir);
+
+	/* retrieve skb back pointer */
+	DPA_READ_SKB_PTR(skb, skbh, phys_to_virt(addr), 0);
+	nr_frags = skb_shinfo(skb)->nr_frags;
+
+	if (fd->format == qm_fd_sg) {
+		/* The sgt buffer has been allocated with netdev_alloc_frag(),
+		 * it's from lowmem.
+		 */
+		sgt = phys_to_virt(addr + dpa_fd_offset(fd));
+
+		/* sgt[0] is from lowmem, was dma_map_single()-ed */
+		dma_unmap_single(dpa_bp->dev, sgt[0].addr,
+				sgt[0].length, dma_dir);
+
+		/* remaining pages were mapped with dma_map_page() */
+		for (i = 1; i < nr_frags; i++) {
+			DPA_BUG_ON(sgt[i].extension);
+
+			dma_unmap_page(dpa_bp->dev, sgt[i].addr,
+					sgt[i].length, dma_dir);
+		}
+
+		/* Free the page frag that we allocated on Tx */
+		put_page(virt_to_head_page(sgt));
+	}
+
+	return skb;
+}
+
+static void dump_parser_result(const struct qm_fd *fd)
+{
+#ifdef CONFIG_HW_LAG_DEBUG
+	dma_addr_t addr = qm_fd_addr(fd);
+	void *vaddr;
+	const fm_prs_result_t *parse_results;
+
+	vaddr = phys_to_virt(addr);
+	DPA_BUG_ON(!IS_ALIGNED((unsigned long)vaddr, SMP_CACHE_BYTES));
+
+	parse_results = (const fm_prs_result_t *)(vaddr +
+				DPA_TX_PRIV_DATA_SIZE);
+
+	hw_lag_dbg("parse_results->l2r:0x%08x\n", parse_results->l2r);
+
+	hw_lag_dbg("FM_L3_PARSE_RESULT_IPV4:0x%0x\n"
+			"FM_L3_PARSE_RESULT_IPV6:0x%0x\n"
+			"parse_results->l3r:0x%08x\n",
+			parse_results->l3r & FM_L3_PARSE_RESULT_IPV4,
+			parse_results->l3r & FM_L3_PARSE_RESULT_IPV6,
+			parse_results->l3r);
+
+	hw_lag_dbg("fm_l4_hxs_has_run(parse_results):0x%0x\n"
+			"fm_l4_hxs_error(parse_results):0x%0x\n",
+			fm_l4_hxs_has_run(parse_results),
+			fm_l4_hxs_error(parse_results));
+
+	hw_lag_dbg("fd->status & FM_FD_STAT_L4CV:0x%x\n"
+			"parse_results->l4r:0x%08x\n"
+			"fm_l4_frame_is_tcp(parse_results):0x%0x\n",
+			fd->status & FM_FD_STAT_L4CV,
+			parse_results->l4r,
+			fm_l4_frame_is_tcp(parse_results));
+#endif
+}
+
+static void show_dbg_info(const struct qm_fd *fd, const char *func_name,
+		struct sk_buff *skb)
+{
+#ifdef CONFIG_HW_LAG_DEBUG
+	u32 pad, fd_status;
+	dma_addr_t addr;
+	struct ethhdr *eth;
+	struct iphdr  *iph;
+	struct tcphdr *tcph;
+	struct udphdr *udph;
+	unsigned int data_start;
+	unsigned long skb_addr;
+
+	fd_status = fd->status;
+	addr = qm_fd_addr(fd);
+
+	/* find out the pad */
+	skb_addr = virt_to_phys(skb->head);
+	pad = addr - skb_addr;
+
+	/* The skb is currently pointed at head + headroom. The packet
+	 * starts at skb->head + pad + fd offset.
+	 */
+	data_start = pad + dpa_fd_offset(fd) - skb_headroom(skb);
+
+	skb_pull(skb, data_start);
+
+	pr_info("[%s]:fd->status:0x%08x\n", func_name, fd_status);
+	pr_info("[%s]:fd tx status:0x%08x. fd rx status:0x%08x\n",
+			func_name,
+			fd_status & FM_FD_STAT_TX_ERRORS,
+			fd_status & FM_FD_STAT_RX_ERRORS);
+
+	if (likely(fd_status & FM_FD_STAT_ERR_PHYSICAL))
+		pr_err("FM_FD_STAT_ERR_PHYSICAL\n");
+	if (fd_status & FM_PORT_FRM_ERR_DMA)
+		pr_err("FM_PORT_FRM_ERR_DMA\n");
+	if (fd_status & FM_PORT_FRM_ERR_PHYSICAL)
+		pr_err("FM_PORT_FRM_ERR_PHYSICAL\n");
+	if (fd_status & FM_PORT_FRM_ERR_SIZE)
+		pr_err("FM_PORT_FRM_ERR_SIZE\n");
+	if (fd_status & FM_PORT_FRM_ERR_PRS_HDR_ERR)
+		pr_err("oh_pcd_confq FM_PORT_FRM_ERR_PRS_HDR_ERR\n");
+
+	pr_info("[%s]:fd->format - qm_fd_contig:%d\n", func_name,
+			fd->format - qm_fd_contig);
+	pr_info("[%s]:fd->bpid:%d\n", func_name, fd->bpid);
+
+	/* get L2 info */
+	skb->protocol = htons(ETH_P_802_3);
+	skb_reset_mac_header(skb);
+	skb_pull_inline(skb, ETH_HLEN);
+
+	eth = eth_hdr(skb);
+
+	pr_info("\n[%s]:dmac:%02x:%02x:%02x:%02x:%02x:%02x\n"
+			"smac:%02x:%02x:%02x:%02x:%02x:%02x\n"
+			"h_proto:0x%04x\n", func_name,
+			eth->h_dest[0], eth->h_dest[1], eth->h_dest[2],
+			eth->h_dest[3], eth->h_dest[4], eth->h_dest[5],
+			eth->h_source[0], eth->h_source[1], eth->h_source[2],
+			eth->h_source[3], eth->h_source[4], eth->h_source[5],
+			eth->h_proto);
+
+	if (fd_status & FM_FD_STAT_L4CV) {
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
+		pr_info("[%s]:skb->ip_summed = CHECKSUM_UNNECESSARY\n",
+				func_name);
+	} else {
+		skb->ip_summed = CHECKSUM_NONE;
+		pr_info("[%s]:skb->ip_summed = CHECKSUM_NONE\n", func_name);
+	}
+
+	/* get L3 and part of L4 info */
+	skb_reset_network_header(skb);
+	skb_reset_transport_header(skb);
+	skb_reset_mac_len(skb);
+
+	if (eth->h_proto == ETH_P_IP) {
+		iph = ip_hdr(skb);
+		pr_info("[%s]:L3_pro:0x%0x, dip:0x%0x, sip:0x%0x\n", func_name,
+			iph->protocol, iph->daddr, iph->saddr);
+
+		skb_pull_inline(skb, ip_hdrlen(skb));
+		skb_reset_transport_header(skb);
+
+		if (iph->protocol == IPPROTO_TCP) {
+			tcph = tcp_hdr(skb);
+			pr_info("[%s]:tcp csum:0x%0x\n",
+					func_name, tcph->check);
+		} else if (iph->protocol == IPPROTO_UDP) {
+			udph = udp_hdr(skb);
+			pr_info("[%s]:udp csum:0x%0x\n",
+					func_name, udph->check);
+		}
+
+	} else if (eth->h_proto == ETH_P_ARP) {
+		pr_info("[%s]:ARP.\n", func_name);
+	} else if (eth->h_proto == ETH_P_IPV6) {
+		pr_info("[%s]:IPv6.\n", func_name);
+	} else if (eth->h_proto == ETH_P_SLOW) {
+		pr_info("[%s]:802.3ad pkt.\n", func_name);
+	} else {
+		pr_info("[%s]:other pkt.\n", func_name);
+	}
+
+	return;
+#endif
+}
+/**
+ * When enqueue an frame from kernel module to offline port,
+ * once errors happeds, this callback will be entered.
+ */
+static enum qman_cb_dqrr_result
+oh_ingress_tx_error_dqrr(struct qman_portal		*portal,
+		      struct qman_fq			*fq,
+		      const struct qm_dqrr_entry	*dq)
+{
+	struct sk_buff *skb;
+	const struct qm_fd *fd = &dq->fd;
+
+	skb = oh_cleanup_tx_fd(fd);
+	dump_parser_result(fd);
+	show_dbg_info(fd, __func__, skb);
+	dev_kfree_skb_any(skb);
+
+	return qman_cb_dqrr_consume;
+}
+/**
+ * This subroutine is copied from oNIC, it should not be call
+ * in normal case, only for debugging outgoing traffics to oh
+ * tx port while no PCD applied for oh port. such as debugging
+ * oh port tx L4 csum.
+ */
+static enum qman_cb_dqrr_result __hot
+oh_ingress_tx_default_dqrr(struct qman_portal		*portal,
+			struct qman_fq			*fq,
+			const struct qm_dqrr_entry	*dq)
+{
+	struct net_device *netdev;
+	struct dpa_priv_s *priv;
+	struct dpa_bp *bp;
+	struct dpa_percpu_priv_s *percpu_priv;
+	struct sk_buff **skbh;
+	struct sk_buff *skb;
+	struct iphdr  *iph;
+	const struct qm_fd *fd = &dq->fd;
+	u32 fd_status = fd->status;
+	u32 pad;
+	dma_addr_t addr = qm_fd_addr(fd);
+	unsigned int data_start;
+	unsigned long skb_addr;
+	int *countptr;
+	struct ethhdr *eth;
+
+	hw_lag_dbg("fd->status:0x%08x\n", fd_status);
+
+	hw_lag_dbg("fd tx status:0x%08x. fd rx status:0x%08x\n",
+			fd_status & FM_FD_STAT_TX_ERRORS,
+			fd_status & FM_FD_STAT_RX_ERRORS);
+
+	if (likely(fd_status & FM_FD_STAT_ERR_PHYSICAL))
+		pr_err("FM_FD_STAT_ERR_PHYSICAL\n");
+
+	if (fd_status & FM_PORT_FRM_ERR_DMA)
+		pr_err("FM_PORT_FRM_ERR_DMA\n");
+	if (fd_status & FM_PORT_FRM_ERR_PHYSICAL)
+		pr_err("FM_PORT_FRM_ERR_PHYSICAL\n");
+	if (fd_status & FM_PORT_FRM_ERR_SIZE)
+		pr_err("FM_PORT_FRM_ERR_SIZE\n");
+	if (fd_status & FM_PORT_FRM_ERR_PRS_HDR_ERR)
+		pr_err("oh_tx_defq FM_PORT_FRM_ERR_PRS_HDR_ERR\n");
+
+	netdev = ((struct dpa_fq *)fq)->net_dev;
+	if (!netdev) {
+		pr_err("error netdev == NULL.\n");
+		skbh = (struct sk_buff **)phys_to_virt(addr);
+		dev_kfree_skb(*skbh);
+		return qman_cb_dqrr_consume;
+	}
+	priv = netdev_priv(netdev);
+	dump_parser_result(fd);
+
+	percpu_priv = __this_cpu_ptr(priv->percpu_priv);
+	countptr = __this_cpu_ptr(priv->dpa_bp->percpu_count);
+
+	skbh = (struct sk_buff **)phys_to_virt(addr);
+	/* according to the last common code (bp refill) the skb pointer is set
+	 * to another address shifted by sizeof(struct sk_buff) to the left
+	 */
+	skb = *(skbh - 1);
+
+	if (unlikely(fd_status & FM_FD_STAT_RX_ERRORS) != 0) {
+		hw_lag_dbg("FD status = 0x%08x\n",
+				fd_status & FM_FD_STAT_RX_ERRORS);
+
+		percpu_priv->stats.rx_errors++;
+		oh_cleanup_tx_fd(fd);
+		goto qman_consume;
+	}
+	if (unlikely(fd->format != qm_fd_contig)) {
+		percpu_priv->stats.rx_dropped++;
+		hw_lag_dbg("Dropping a SG frame\n");
+		oh_cleanup_tx_fd(fd);
+		goto qman_consume;
+	}
+
+	hw_lag_dbg("fd->bpid:%d\n", fd->bpid);
+	bp = dpa_bpid2pool(fd->bpid);
+	dma_unmap_single(bp->dev, addr, bp->size, DMA_TO_DEVICE);
+
+	/* find out the pad */
+	skb_addr = virt_to_phys(skb->head);
+	pad = addr - skb_addr;
+
+	countptr = __this_cpu_ptr(bp->percpu_count);
+	(*countptr)--;
+
+	/* The skb is currently pointed at head + headroom. The packet
+	 * starts at skb->head + pad + fd offset.
+	 */
+	data_start = pad + dpa_fd_offset(fd) - skb_headroom(skb);
+	skb_pull(skb, data_start);
+
+	/* get L2 info */
+	skb->protocol = eth_type_trans(skb, netdev);
+	eth = eth_hdr(skb);
+
+	hw_lag_dbg("dmac:%02x:%02x:%02x:%02x:%02x:%02x\n"
+			"smac:%02x:%02x:%02x:%02x:%02x:%02x\n"
+			"h_proto:0x%04x\n",
+			eth->h_dest[0], eth->h_dest[1], eth->h_dest[2],
+			eth->h_dest[3], eth->h_dest[4], eth->h_dest[5],
+			eth->h_source[0], eth->h_source[1], eth->h_source[2],
+			eth->h_source[3], eth->h_source[4], eth->h_source[5],
+			eth->h_proto);
+
+	if (unlikely(dpa_check_rx_mtu(skb, netdev->mtu))) {
+		percpu_priv->stats.rx_dropped++;
+		goto qman_consume;
+	}
+
+	if (fd_status & FM_FD_STAT_L4CV) {
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
+		hw_lag_dbg("skb->ip_summed = CHECKSUM_UNNECESSARY\n");
+	} else {
+		skb->ip_summed = CHECKSUM_NONE;
+		hw_lag_dbg("skb->ip_summed = CHECKSUM_NONE\n");
+	}
+
+	/* get L3 and part of L4 info */
+	skb_reset_network_header(skb);
+	skb_reset_transport_header(skb);
+	skb_reset_mac_len(skb);
+
+	if (eth->h_proto == ETH_P_IP) {
+		iph = ip_hdr(skb);
+		hw_lag_dbg("L3_pro:0x%0x, dip:0x%0x, sip:0x%0x\n",
+			iph->protocol, iph->daddr, iph->saddr);
+	} else if (eth->h_proto == ETH_P_ARP) {
+		hw_lag_dbg("ARP.\n");
+	} else if (eth->h_proto == ETH_P_IPV6) {
+		hw_lag_dbg("IPv6.\n");
+	} else if (eth->h_proto == ETH_P_SLOW) {
+		hw_lag_dbg("802.3ad pkt.\n");
+	} else {
+		hw_lag_dbg("other pkt.\n");
+	}
+
+qman_consume:
+	dev_kfree_skb_any(skb);
+
+	return qman_cb_dqrr_consume;
+}
+/**
+ * When frame leave from PCD fqs then goes final terminated physical
+ * ports(MAC ports),once errors happend, this callback will be entered.
+ * dump debugging information when HW_LAG_DEBUG enabled .
+ */
+static enum qman_cb_dqrr_result
+oh_pcd_err_dqrr(struct qman_portal *portal, struct qman_fq *fq,
+		const struct qm_dqrr_entry *dq)
+{
+	struct sk_buff *skb;
+	const struct qm_fd *fd = &dq->fd;
+
+	skb = oh_cleanup_tx_fd(fd);
+	dump_parser_result(fd);
+	show_dbg_info(fd, __func__, skb);
+	dev_kfree_skb_any(skb);
+
+	return qman_cb_dqrr_consume;
+
+}
+/**
+ * When frame leave from offline port tx fqs then goes into offline tx
+ * ports(MAC ports), it will be into confirm fq, this callback will be
+ * entered.
+ * dump debugging information when HW_LAG_DEBUG enabled.
+ * don't free skb, since offline port is not the final consumer.
+ */
+static enum qman_cb_dqrr_result __hot
+oh_tx_conf_dqrr(struct qman_portal *portal, struct qman_fq *fq,
+		const struct qm_dqrr_entry *dq)
+{
+	struct sk_buff *skb;
+	const struct qm_fd *fd = &dq->fd;
+
+	skb = oh_cleanup_tx_fd(fd);
+	dump_parser_result(fd);
+	show_dbg_info(fd, __func__, skb);
+
+	return qman_cb_dqrr_consume;
+}
+
+static void lag_public_egress_ern(struct qman_portal *portal,
+		struct qman_fq *fq, const struct qm_mr_entry *msg)
+{
+	/* will add ERN statistics in the future version. */
+	const struct qm_fd *fd = &msg->ern.fd;
+	struct sk_buff *skb;
+
+	if (msg->ern.fd.cmd & FM_FD_CMD_FCO) {
+		dpa_oh_fd_release(fd);
+		return;
+	}
+
+	skb = oh_cleanup_tx_fd(fd);
+	dump_parser_result(fd);
+	show_dbg_info(fd, __func__, skb);
+	dev_kfree_skb_any(skb);
+}
+
+/**
+ * This subroutine will be called when frame out of oh pcd fqs and
+ * consumed by (MAC ports).
+ * Display debugging information if HW_LAG_DEBUG on.
+ */
+static enum qman_cb_dqrr_result __hot
+oh_pcd_conf_dqrr(struct qman_portal *portal, struct qman_fq *fq,
+		const struct qm_dqrr_entry *dq)
+{
+	struct sk_buff *skb;
+	const struct qm_fd *fd = &dq->fd;
+
+	skb = oh_cleanup_tx_fd(fd);
+	show_dbg_info(fd, __func__, skb);
+	dev_kfree_skb_any(skb);
+
+	return qman_cb_dqrr_consume;
+}
+
+static const struct qman_fq oh_tx_defq = {
+	.cb = { .dqrr = oh_ingress_tx_default_dqrr}
+};
+/* for OH ports Rx Error queues = Tx Error queues */
+static const struct qman_fq oh_tx_errq = {
+	.cb = { .dqrr = oh_ingress_tx_error_dqrr}
+};
+
+static const struct qman_fq oh_pcd_confq = {
+	.cb = { .dqrr = oh_pcd_conf_dqrr}
+};
+static const struct qman_fq oh_pcd_errq = {
+	.cb = { .dqrr = oh_pcd_err_dqrr}
+};
+static const struct qman_fq oh_tx_confq = {
+	.cb = { .dqrr = oh_tx_conf_dqrr}
+};
+static const struct qman_fq oh_pcd_egress_ernq = {
+	.cb = { .ern = lag_public_egress_ern}
+};
+static const struct qman_fq oh_egress_ernq = {
+	.cb = { .ern = lag_public_egress_ern}
+};
+
+static int oh_add_channel(void *__arg)
+{
+	int cpu;
+	struct qman_portal *portal;
+	const cpumask_t *cpus = qman_affine_cpus();
+	u32 pool = QM_SDQCR_CHANNELS_POOL_CONV((u32)(unsigned long)__arg);
+
+	for_each_cpu(cpu, cpus) {
+		portal = (struct qman_portal *)qman_get_affine_portal(cpu);
+		qman_p_static_dequeue_add(portal, pool);
+	}
+
+	return 0;
+}
+
+static int init_oh_errq_defq(struct device *dev,
+		uint32_t fqid_err, uint32_t fqid_def,
+		struct dpa_fq **oh_errq, struct dpa_fq **oh_defq,
+		uint16_t *priv_channel)
+{
+	int errno;
+	struct dpa_fq *errq, *defq;
+	/* These two vaules come from DPA-Eth driver */
+	uint8_t wq_errq = 2, wq_defq = 1;
+	u32 channel;
+	struct qm_mcc_initfq initfq;
+	struct qm_fqd fqd;
+	struct task_struct *kth;
+
+	/* Get a channel */
+	errno = qman_alloc_pool(&channel);
+	if (errno) {
+		pr_err("error on getting pool channel.\n");
+		return errno;
+	}
+
+	if (channel < 0) {
+		errno = channel;
+		pr_err("error on dpa_get_channel().\n");
+		return errno;
+	}
+
+	/* Start a thread that will walk the cpus with affine portals
+	 * and add this pool channel to each's dequeue mask.
+	 */
+
+	kth = kthread_run(oh_add_channel, (void *)(unsigned long)channel,
+			  "oh_add_channel");
+	if (!kth) {
+		pr_warn("run kthread faild...\n");
+		return -ENOMEM;
+	}
+
+	/* Allocate memories for Tx ErrQ and Tx DefQ of oh port */
+	errq = devm_kzalloc(dev, sizeof(struct dpa_fq), GFP_KERNEL);
+	if (!errq) {
+		pr_err("devm_kzalloc() for OH errq failed\n");
+		return -ENOMEM;
+	}
+	defq = devm_kzalloc(dev, sizeof(struct dpa_fq), GFP_KERNEL);
+	if (!defq) {
+		pr_err("devm_kzalloc() for OH defq failed.\n");
+		return -ENOMEM;
+	}
+
+	/* Set Tx ErrQ callbacks of oh port */
+	errq->fq_base = oh_tx_errq;
+
+	/* Set the flags of the oh port Tx ErrQ/Tx DefQ and create the FQs */
+	errq->fq_base.flags = QMAN_FQ_FLAG_NO_ENQUEUE;
+	errno = qman_create_fq(fqid_err, errq->fq_base.flags, &errq->fq_base);
+	if (errno) {
+		pr_err("error on create OH errq.\n");
+		return errno;
+	}
+
+	defq->fq_base = oh_tx_defq;
+	defq->fq_base.flags = QMAN_FQ_FLAG_NO_ENQUEUE;
+	errno = qman_create_fq(fqid_def, defq->fq_base.flags, &defq->fq_base);
+	if (errno) {
+		pr_err("error on create OH defq.\n");
+		return errno;
+	}
+
+	*priv_channel = (uint16_t)channel;
+	/* Set the FQs init options then init the FQs */
+	initfq.we_mask = QM_INITFQ_WE_DESTWQ;
+	initfq.fqd.dest.channel = (uint16_t)channel;
+	initfq.fqd.dest.wq = wq_errq;
+	initfq.we_mask |= QM_INITFQ_WE_FQCTRL;
+	initfq.fqd.fq_ctrl = QM_FQCTRL_PREFERINCACHE;
+	initfq.we_mask |= QM_INITFQ_WE_CONTEXTA;
+	initfq.fqd.fq_ctrl |= QM_FQCTRL_CTXASTASHING | QM_FQCTRL_AVOIDBLOCK;
+	initfq.fqd.context_a.stashing.exclusive = QM_STASHING_EXCL_DATA |
+		QM_STASHING_EXCL_CTX | QM_STASHING_EXCL_ANNOTATION;
+	initfq.fqd.context_a.stashing.data_cl = 2;
+	initfq.fqd.context_a.stashing.annotation_cl = 1;
+	initfq.fqd.context_a.stashing.context_cl =
+		DIV_ROUND_UP(sizeof(struct qman_fq), 64);
+
+	/* init oh ports errors fq */
+	errno = qman_init_fq(&errq->fq_base, QMAN_INITFQ_FLAG_SCHED, &initfq);
+	if (errno < 0) {
+		pr_err("error on qman_init_fq %u = %d\n", fqid_err, errno);
+		qman_destroy_fq(&errq->fq_base, 0);
+		devm_kfree(dev, errq);
+		return errno;
+	}
+
+	errno = qman_query_fq(&errq->fq_base, &fqd);
+	hw_lag_dbg("errno of qman_query_fq:0x%08x\n", errno);
+	if (fqd.fq_ctrl != initfq.fqd.fq_ctrl) {
+		pr_err("queried fq_ctrl=%x, should be=%x\n", fqd.fq_ctrl,
+			initfq.fqd.fq_ctrl);
+		panic("fail");
+	}
+	if (memcmp(&fqd.td, &initfq.fqd.td, sizeof(fqd.td))) {
+		pr_err("queried td_thresh=%x:%x, should be=%x:%x\n",
+			fqd.td.exp, fqd.td.mant,
+			initfq.fqd.td.exp, initfq.fqd.td.mant);
+		panic("fail");
+	}
+
+	/* init oh ports default fq */
+	initfq.fqd.dest.wq = wq_defq;
+	errno = qman_init_fq(&defq->fq_base, QMAN_INITFQ_FLAG_SCHED, &initfq);
+	if (errno < 0) {
+		pr_err("error on qman_init_fq %u = %d\n", fqid_def, errno);
+		qman_destroy_fq(&defq->fq_base, 0);
+		devm_kfree(dev, defq);
+		return errno;
+	}
+
+	*oh_errq = errq;
+	*oh_defq = defq;
+
+	hw_lag_dbg("oh port defq and oh port errq initialize OK\n");
+
+	return BOND_OH_SUCCESS;
+}
+/**
+ * Initialize pcd err fqs and pcd confirmation fqs.
+ * HW LAG uses this method rather than reuse DPA-Eth private rx err/
+ * rx def/tx err/tx confirm FQs and callbacks, since HW LAG uses
+ * different data structure from DPA-Eth private driver.
+ */
+static int init_oh_pcderrq_pcdconfq(struct device *dev,
+		uint32_t *fqid_pcderr, uint32_t *fqid_pcdconf,
+		struct dpa_fq **oh_pcderrq, struct dpa_fq **oh_pcdconfq,
+		uint16_t priv_channel)
+{
+	int errno;
+	struct dpa_fq *pcderrq, *pcdconfq;
+	/* These two vaules come from DPA-Eth driver */
+	uint8_t wq_errq = 2, wq_confq = 1;
+	struct qm_mcc_initfq initfq;
+
+	/* Allocate memories for PCD ErrQ and PCD confirm Q of oh port */
+	pcderrq = devm_kzalloc(dev, sizeof(struct dpa_fq), GFP_KERNEL);
+	if (!pcderrq) {
+		pr_err("devm_kzalloc() for OH pcderrq failed\n");
+		return -ENOMEM;
+	}
+
+	pcdconfq = devm_kzalloc(dev, sizeof(struct dpa_fq), GFP_KERNEL);
+	if (!pcdconfq) {
+		pr_err("devm_kzalloc() for OH pcdconfq failed.\n");
+		return -ENOMEM;
+	}
+
+	/* Set PCD ErrQ callbacks of oh port */
+	pcderrq->fq_base = oh_pcd_errq;
+
+	/* Set the flags of the oh port PCD ErrQ, create the FQs */
+	pcderrq->fq_base.flags = QMAN_FQ_FLAG_NO_ENQUEUE |
+		QMAN_FQ_FLAG_DYNAMIC_FQID;
+	errno = qman_create_fq(0, pcderrq->fq_base.flags, &pcderrq->fq_base);
+	if (errno) {
+		pr_err("error on create OH pcderrq.\n");
+		return errno;
+	}
+	*fqid_pcderr = pcderrq->fq_base.fqid;
+	hw_lag_dbg("*fqid_pcderr:%d\n", *fqid_pcderr);
+
+	/* Set PCD confirm Q callbacks of oh port */
+	pcdconfq->fq_base = oh_pcd_confq;
+	/* Set the flags of the oh port PCD confQ, create the FQs */
+	pcdconfq->fq_base.flags = QMAN_FQ_FLAG_NO_ENQUEUE|
+		QMAN_FQ_FLAG_DYNAMIC_FQID;
+	errno = qman_create_fq(0, pcdconfq->fq_base.flags, &pcdconfq->fq_base);
+	if (errno) {
+		pr_err("error on create OH pcdconfq.\n");
+		return errno;
+	}
+	*fqid_pcdconf = pcdconfq->fq_base.fqid;
+	hw_lag_dbg("*fqid_pcdconf:%d\n", *fqid_pcdconf);
+
+	/* Set the FQs init options then init the FQs */
+	initfq.we_mask = QM_INITFQ_WE_DESTWQ;
+	initfq.fqd.dest.channel = priv_channel;
+	initfq.fqd.dest.wq = wq_errq;
+	initfq.we_mask |= QM_INITFQ_WE_FQCTRL;
+	initfq.fqd.fq_ctrl = QM_FQCTRL_PREFERINCACHE;
+	initfq.we_mask |= QM_INITFQ_WE_CONTEXTA;
+	initfq.fqd.fq_ctrl |= QM_FQCTRL_CTXASTASHING | QM_FQCTRL_AVOIDBLOCK;
+	initfq.fqd.context_a.stashing.exclusive = QM_STASHING_EXCL_DATA |
+		QM_STASHING_EXCL_CTX | QM_STASHING_EXCL_ANNOTATION;
+	initfq.fqd.context_a.stashing.data_cl = 2;
+	initfq.fqd.context_a.stashing.annotation_cl = 1;
+	initfq.fqd.context_a.stashing.context_cl =
+		DIV_ROUND_UP(sizeof(struct qman_fq), 64);
+
+	/* init pcd errors fq */
+	errno = qman_init_fq(&pcderrq->fq_base,
+			QMAN_INITFQ_FLAG_SCHED,	&initfq);
+	if (errno < 0) {
+		pr_err("error on qman_init_fq pcderrq:%u = %d\n",
+				*fqid_pcderr, errno);
+		qman_destroy_fq(&pcderrq->fq_base, 0);
+		devm_kfree(dev, pcderrq);
+
+		return errno;
+	}
+
+	/* init pcd confirm fq */
+	initfq.fqd.dest.wq = wq_confq;
+	errno = qman_init_fq(&pcdconfq->fq_base,
+			QMAN_INITFQ_FLAG_SCHED, &initfq);
+	if (errno < 0) {
+		pr_err("error on qman_init_fq pcdconfq:%u = %d\n",
+				*fqid_pcdconf, errno);
+		qman_destroy_fq(&pcdconfq->fq_base, 0);
+		devm_kfree(dev, pcdconfq);
+
+		return errno;
+	}
+
+	*oh_pcderrq = pcderrq;
+	*oh_pcdconfq = pcdconfq;
+
+	hw_lag_dbg("oh pcd confq and pcd errq initialize OK\n");
+
+	return BOND_OH_SUCCESS;
+}
+/**
+ * Initialize confirmation fq for offline port tx fqs.
+ * This confirmation call back is enabled in case of buffer is released
+ * by BM after frame entered into tx port of offline port.
+ */
+static int init_oh_txconfq(struct device *dev, uint32_t *fqid_ohtxconf,
+		struct dpa_fq **oh_txconfq, uint16_t priv_channel)
+{
+	int errno;
+	struct dpa_fq *txconfq;
+	/* This vaule comes from DPA-Eth driver */
+	uint8_t wq_confq = 1;
+	struct qm_mcc_initfq initfq;
+
+	/* Allocate memories for PCD ErrQ and PCD confirm Q of oh port */
+	txconfq = devm_kzalloc(dev, sizeof(struct dpa_fq), GFP_KERNEL);
+	if (!txconfq) {
+		pr_err("devm_kzalloc() for OH tx confq failed.\n");
+		return -ENOMEM;
+	}
+
+	/* Set tx confirm Q callbacks of oh port */
+	txconfq->fq_base = oh_tx_confq;
+	/* Set the flags of the oh port PCD confQ, create the FQs */
+	txconfq->fq_base.flags = QMAN_FQ_FLAG_NO_ENQUEUE |
+		QMAN_FQ_FLAG_DYNAMIC_FQID;
+	errno = qman_create_fq(0, txconfq->fq_base.flags, &txconfq->fq_base);
+	if (errno) {
+		pr_err("error on create OH tx confq.\n");
+		return errno;
+	}
+	*fqid_ohtxconf = txconfq->fq_base.fqid;
+	hw_lag_dbg("dynamic *fqid_ohtxconf:%d\n", *fqid_ohtxconf);
+
+	/* Set the FQs init options then init the FQs */
+	initfq.we_mask = QM_INITFQ_WE_DESTWQ;
+	initfq.fqd.dest.channel = priv_channel;
+	initfq.fqd.dest.wq = wq_confq;
+	initfq.we_mask |= QM_INITFQ_WE_FQCTRL;
+	initfq.fqd.fq_ctrl = QM_FQCTRL_PREFERINCACHE;
+	initfq.we_mask |= QM_INITFQ_WE_CONTEXTA;
+	initfq.fqd.fq_ctrl |= QM_FQCTRL_CTXASTASHING | QM_FQCTRL_AVOIDBLOCK;
+	initfq.fqd.context_a.stashing.exclusive = QM_STASHING_EXCL_DATA |
+		QM_STASHING_EXCL_CTX | QM_STASHING_EXCL_ANNOTATION;
+	initfq.fqd.context_a.stashing.data_cl = 2;
+	initfq.fqd.context_a.stashing.annotation_cl = 1;
+	initfq.fqd.context_a.stashing.context_cl =
+		DIV_ROUND_UP(sizeof(struct qman_fq), 64);
+
+	/* init oh tx confirm fq */
+	initfq.fqd.dest.wq = wq_confq;
+	errno = qman_init_fq(&txconfq->fq_base,
+			QMAN_INITFQ_FLAG_SCHED, &initfq);
+	if (errno < 0) {
+		pr_err("error on qman_init_fq oh tx confq:%u = %d\n",
+				*fqid_ohtxconf, errno);
+		qman_destroy_fq(&txconfq->fq_base, 0);
+		devm_kfree(dev, txconfq);
+
+		return errno;
+	}
+
+	*oh_txconfq = txconfq;
+
+	hw_lag_dbg("oh tx confq initialize OK\n");
+
+	return BOND_OH_SUCCESS;
+}
+/**
+ * Initialize dynamic particular tx fqs of offline port for LAG xmit,
+ * does not reuse tx fqs initialized by offline port driver. This method
+ * can avoid to modify offline port driver even if the confirmation fq
+ * need to be enabled.
+ */
+static int init_oh_tx_lag_fqs(struct device *dev,
+		struct dpa_fq **oh_tx_lag_fqs, uint32_t fqid_ohtxconf,
+		uint16_t oh_tx_channel)
+{
+	int errno = BOND_OH_SUCCESS, i, tx_fqs_count;
+	uint16_t wq_id;
+	struct dpa_fq *lag_fqs;
+	struct qm_mcc_initfq fq_opts;
+	uint32_t create_flags, init_flags;
+
+	tx_fqs_count = num_possible_cpus();
+	/* Allocate particular tx queues of offline port for LAG xmit. */
+	lag_fqs = devm_kzalloc(dev, sizeof(struct dpa_fq) * tx_fqs_count,
+			GFP_KERNEL);
+	if (!lag_fqs) {
+		pr_err("Can't allocate tx fqs for LAG xmit.\n");
+		errno = -ENOMEM;
+		goto return_kfree;
+	}
+
+	/* Set flags for particular tx fqs, especially for dynamic fqid. */
+	create_flags = QMAN_FQ_FLAG_TO_DCPORTAL | QMAN_FQ_FLAG_DYNAMIC_FQID;
+
+	/* Create particular tx fqs of offline port for LAG xmit */
+	for (i = 0; i < tx_fqs_count; i++) {
+		/* set egress_ern callback for offline port tx fq */
+		lag_fqs[i].fq_base = oh_egress_ernq;
+		errno = qman_create_fq(0, create_flags, &lag_fqs[i].fq_base);
+		if (errno) {
+			pr_err("Error on creating tx fqs for LAG xmit.\n");
+			goto return_kfree;
+		}
+	}
+
+	/* Set init flags for tx fqs of oh port */
+	init_flags = QMAN_INITFQ_FLAG_SCHED;
+
+	/* Set fq init options. Specify destination wq id and channel */
+	memset(&fq_opts, 0, sizeof(fq_opts));
+	fq_opts.we_mask = QM_INITFQ_WE_DESTWQ;
+	/* wq info from DPA-Eth driver */
+	wq_id = 3;
+	fq_opts.fqd.dest.wq = wq_id;
+	fq_opts.fqd.dest.channel = oh_tx_channel;
+
+	fq_opts.we_mask |= QM_INITFQ_WE_FQCTRL;
+	fq_opts.fqd.fq_ctrl = QM_FQCTRL_PREFERINCACHE;
+	fq_opts.fqd.fq_ctrl |= QM_FQCTRL_CTXASTASHING | QM_FQCTRL_AVOIDBLOCK;
+	fq_opts.fqd.context_a.stashing.exclusive = QM_STASHING_EXCL_DATA |
+	QM_STASHING_EXCL_CTX | QM_STASHING_EXCL_ANNOTATION;
+	fq_opts.fqd.context_a.stashing.data_cl = 2;
+	fq_opts.fqd.context_a.stashing.annotation_cl = 1;
+	fq_opts.fqd.context_a.stashing.context_cl =
+		DIV_ROUND_UP(sizeof(struct qman_fq), 64);
+
+#ifdef CONFIG_HW_LAG_DEBUG
+	fq_opts.we_mask |= QM_INITFQ_WE_CONTEXTA | QM_INITFQ_WE_CONTEXTB;
+	/**
+	 * CTXA[OVFQ] = 1
+	 * we set particular tx own confirmation fq and their own callback
+	 * in case of interrupt DPA-Eth private conf callback/err callback
+	 * /def callback.
+	 */
+
+	fq_opts.fqd.context_a.hi = 0x80000000;
+	fq_opts.fqd.context_a.lo = 0x0;
+	fq_opts.fqd.context_b = fqid_ohtxconf;
+#endif
+	/* Initialize particular tx frame queue of offline port for LAG xmit */
+	for (i = 0; i < tx_fqs_count; i++) {
+		errno = qman_init_fq(&lag_fqs[i].fq_base, init_flags, &fq_opts);
+		if (errno)
+			goto init_error;
+	}
+
+	for (i = 0; i < tx_fqs_count; i++) {
+		hw_lag_dbg("ok, created lag_fqs: fqid:%d\n",
+				lag_fqs[i].fq_base.fqid);
+	}
+
+	*oh_tx_lag_fqs = lag_fqs;
+
+	return BOND_OH_SUCCESS;
+init_error:
+	while (i-- < 0) {
+		hw_lag_dbg("errors on initializing tx fqs, No.:%d tx fq.\n", i);
+		qman_destroy_fq(&lag_fqs[i].fq_base, 0);
+	}
+
+return_kfree:
+	if (lag_fqs)
+		devm_kfree(dev, lag_fqs);
+
+	return errno;
+}
+/**
+ * This subroutine has been copied from offline_port driver
+ * to get all information of all offline ports by parse DTS
+ * return BOND_OH_SUCCESS when get information successfully.
+ */
+int get_oh_info(void)
+{
+	struct platform_device	*oh_of_dev, *of_dev;
+	struct device	*dpa_oh_dev, *oh_dev;
+	struct device_node	*dpa_oh_node = NULL, *oh_node;
+	int	lenp, errno = BOND_OH_SUCCESS, i = 0;
+	const phandle	*p_oh_port_handle;
+	const unsigned int	*p_port_id;
+	const unsigned int	*p_channel_id;
+	struct fm_port		*oh_port;
+	unsigned long port_handle_cnt;
+	struct fm_port_params params;
+
+	available_num_of_oh_ports = 0;
+
+	/* probe offline ports and alloc memory, these codes need refining
+	 * to save memory and need to get rid of the global variable.
+	 */
+	poh = kzalloc(sizeof(struct oh_port_priv) * FM_MAX_NUM_OF_OH_PORTS,
+			GFP_KERNEL);
+	if (!poh)
+		return -ENOMEM;
+
+	for_each_matching_node(dpa_oh_node, oh_port_match_table) {
+		if (dpa_oh_node) {
+			p_oh_port_handle = of_get_property(dpa_oh_node,
+					"fsl,fman-oh-port", &lenp);
+			if (!p_oh_port_handle) {
+				pr_err("No OH port handle in node %s\n",
+						dpa_oh_node->full_name);
+				return -EINVAL;
+			}
+			hw_lag_dbg("dpa_oh_node->name:%s\n",
+					dpa_oh_node->full_name);
+			BUG_ON(lenp % sizeof(*p_oh_port_handle));
+			if (lenp != sizeof(*p_oh_port_handle)) {
+				port_handle_cnt =
+					lenp / sizeof(*p_oh_port_handle);
+
+				pr_err("Found %lu oh port in node %s\n"
+					"only 1 phandle is allowed.\n",
+					port_handle_cnt,
+					dpa_oh_node->full_name);
+				return -EINVAL;
+			}
+
+			oh_node = of_find_node_by_phandle(*p_oh_port_handle);
+			if (!oh_node) {
+				pr_err("no oh node referenced from %s\n",
+						dpa_oh_node->full_name);
+				return -EINVAL;
+			}
+			hw_lag_dbg("Found oh_node->full_name  %s.\n",
+					oh_node->full_name);
+			p_port_id = of_get_property(oh_node,
+							"cell-index", &lenp);
+
+			if (!p_port_id) {
+				pr_err("No port id found in node %s\n",
+						dpa_oh_node->full_name);
+				return -EINVAL;
+			}
+
+			hw_lag_dbg("Found port id %ud, in node %s\n",
+					*p_port_id, dpa_oh_node->full_name);
+			BUG_ON(lenp % sizeof(*p_port_id));
+
+			/* Read channel id for the queues */
+			p_channel_id =
+				of_get_property(oh_node,
+						"fsl,qman-channel-id", &lenp);
+			if (!p_channel_id) {
+				pr_err("No channel id found in node %s\n",
+						dpa_oh_node->full_name);
+				return -EINVAL;
+			}
+
+			BUG_ON(lenp % sizeof(*p_channel_id));
+
+			oh_of_dev = of_find_device_by_node(oh_node);
+			BUG_ON(!oh_of_dev);
+			oh_dev = &oh_of_dev->dev;
+			of_dev = of_find_device_by_node(dpa_oh_node);
+			BUG_ON(!oh_of_dev);
+			dpa_oh_dev = &of_dev->dev;
+			poh[i].of_dev = of_dev;
+			poh[i].oh_of_dev = oh_of_dev;
+			poh[i].dpa_oh_dev = dpa_oh_dev;
+			poh[i].oh_dev = oh_dev;
+			poh[i].dpa_oh_node = dpa_oh_node;
+			poh[i].oh_node = oh_node;
+			poh[i].cell_index = *p_port_id;
+			poh[i].oh_config = dev_get_drvdata(dpa_oh_dev);
+			poh[i].p_oh_port_handle = p_oh_port_handle;
+			poh[i].oh_channel_id = *p_channel_id;
+			oh_port = poh[i].oh_config->oh_port;
+			fm_port_get_buff_layout_ext_params(oh_port, &params);
+			poh[i].bpid = params.pool_param[0].id;
+			poh[i].bp_size = params.pool_param[0].size;
+			/* give a friend name like "fman0-oh@1"
+			 * rather than "/fsl,dpaa/dpa-fman0-oh@1".
+			 * fill friendname array with dpa_oh_node->full_name,
+			 * please don't use oh0 since documentatin says oh0
+			 * has bad performance.
+			 */
+			memcpy(poh[i].friendname,
+					dpa_oh_node->full_name + 14, 10);
+
+			if (qman_alloc_fqid_range(&poh[i].pcd_fqids_base,
+						FM_MAX_NUM_OF_MACS, true, 0)
+					!= FM_MAX_NUM_OF_MACS) {
+				pr_err("error on alloc continuous pcd fqid\n");
+				return -EINVAL;
+			}
+
+			errno = init_oh_errq_defq(poh[i].dpa_oh_dev,
+					poh[i].oh_config->error_fqid,
+					poh[i].oh_config->default_fqid,
+					&poh[i].oh_errq,
+					&poh[i].oh_defq,
+					&poh[i].p_oh_rcv_channel);
+			if (errno != BOND_OH_SUCCESS) {
+				pr_err("error when probe errq or defq.\n");
+				return errno;
+			}
+
+			errno = init_oh_pcderrq_pcdconfq(poh[i].dpa_oh_dev,
+					&poh[i].fqid_pcderr,
+					&poh[i].fqid_pcdconf,
+					&poh[i].oh_pcderrq,
+					&poh[i].oh_pcdconfq,
+					poh[i].p_oh_rcv_channel);
+			if (errno != BOND_OH_SUCCESS) {
+				pr_err("error on probe pcderrq or pcdconfq\n");
+				return errno;
+			}
+
+			errno = init_oh_txconfq(poh[i].dpa_oh_dev,
+					&poh[i].fqid_ohtxconf,
+					&poh[i].oh_txconfq,
+					poh[i].oh_channel_id);
+			if (errno != BOND_OH_SUCCESS) {
+				pr_err("error on init offline port tx confq\n");
+				return errno;
+			}
+
+			errno = init_oh_tx_lag_fqs(poh[i].dpa_oh_dev,
+					&poh[i].oh_tx_lag_fqs,
+					poh[i].fqid_ohtxconf,
+					poh[i].oh_channel_id);
+			if (errno != BOND_OH_SUCCESS) {
+				pr_err("error on init offline port tx confq\n");
+				return errno;
+			}
+
+			errno = dpa_oh_bp_create(&poh[i]);
+			if (errno != BOND_OH_SUCCESS) {
+				pr_err("error on init offline tx bp.\n");
+				return errno;
+			}
+
+			available_num_of_oh_ports = ++i;
+		}
+	}
+
+	return errno;
+}
+/**
+ * Get the FM_MAC_RES_ID from a dpa-eth NIC, return 0 if it is not a dpa-eth，
+ * otherwise return FM_MAC_RES_ID
+ * this function does not process macless, LAG does not need a macless IF.
+ */
+static unsigned long long get_fm_mac_res_id(struct net_device *netdev)
+{
+	struct dpa_priv_s *priv = netdev_priv(netdev);
+	if (!is_dpa_eth_port(netdev))
+		return 0;
+
+	return (unsigned long long)priv->mac_dev->res->start;
+}
+/**
+ * Get the DCP_ID from a dpa-eth NIC, return 0 if it is not a dpa-eth，
+ * return 1 if it's fm0, return 2 if it's fm1, since there are only 2
+ * FMAN in current DPA SOC.
+ * this function does not process macless, LAG does not need a macless IF.
+ */
+int get_dcp_id_from_dpa_eth_port(struct net_device *netdev)
+{
+	unsigned long long mac_res_start = get_fm_mac_res_id(netdev);
+
+	if ((mac_res_start >= FM1_GB0) && (mac_res_start <= FM1_10G))
+		return 1;
+	else if ((mac_res_start >= FM2_GB0) && (mac_res_start <= FM2_10G))
+		return 2;
+	else
+		return 0;
+}
+/**
+ * Get all information of the offline port which is being used
+ * by a bundle, such as fman_dcpid, offline port offset, cell index,
+ * offline port channel. This API is required by CEETM Qos.
+ * Regarding fman dpcid, till sdk1.6, there is one fman in p1023, the
+ * offset is 0x1000000, for other dpaa socs, the offset of fman0 is
+ * 0x400000, the offset of fman1 is 0x500000, hence for current socs,
+ * the offset of fman0 <=0x4000000, 0x400000 < fman1 <=0x500000.
+ * return BOND_OH_SUCCESS when got all information, otherwise return
+ * Non-Zero.
+ */
+#define FMAN0_MAX_OFFSET 0x400000
+#define FMAN1_MAX_OFFSET 0x500000
+int export_oh_port_info_to_ceetm(struct bonding *bond, uint16_t *channel,
+		unsigned long *fman_dcpid, unsigned long *oh_offset,
+		unsigned long *cell_index)
+{
+	/**
+	 * split str: "/soc@ffe000000/fman@400000/port@84000", then get
+	 * the fman@ part and port@ part from them. regex is good enough
+	 * as below:
+	 * ret = sscanf((char *) p, "%*[^@]@%*[^@]@%[^/]/port@%s", s1, s2);
+	 * but the kernel version does not support the method.
+	 */
+	int errno;
+	char s1[16] = {0}, s2[16] = {0};
+	char *p, *p1;
+
+	if (!bond->params.ohp) {
+		pr_err("The bundle has not binded an offline port.\n");
+		return 1;
+	}
+
+	if (!bond->params.ohp->oh_en) {
+		pr_err("The offline is disabled, to enable it, use sysfs.\n");
+		return 2;
+	}
+
+	if (!bond->params.ohp->oh_node) {
+		pr_err("The offline node error.\n");
+		return 3;
+	}
+
+	p = strstr(bond->params.ohp->oh_node->full_name, "fman@");
+	p += strlen("fman@");
+	p1 = strstr(p, "/port@");
+
+	memcpy(s1, p, p1 - p);
+
+	p = strstr(p, "/port@");
+	p += strlen("/port@");
+
+	errno = sscanf((const char *) p, "%s", s2);
+	if (errno != 1) {
+		pr_err("parser error while process offline port node\n");
+		return 4;
+	}
+
+	errno = kstrtoul(s1, 16, fman_dcpid) | kstrtoul(s2, 16, oh_offset);
+	if (errno) {
+		pr_err("error on kstrtoul fman_dcpid, of_offset\n");
+		return 5;
+	}
+	if (*fman_dcpid <= FMAN0_MAX_OFFSET) {
+		*fman_dcpid = 0;
+	} else if ((*fman_dcpid > FMAN0_MAX_OFFSET) &&
+			(*fman_dcpid <= FMAN1_MAX_OFFSET)) {
+		*fman_dcpid = 1;
+	} else {
+		pr_err("error on calculating fman dcpid, new soc appears.\n");
+		return 6;
+	}
+
+	*channel = bond->params.ohp->oh_channel_id;
+	*cell_index = bond->params.ohp->cell_index;
+
+	hw_lag_dbg("This oh port mapped to bond has channel:0x%0x\n", *channel);
+	hw_lag_dbg("fman_dcpid:0x%0lx, oh_offset:0x%0lx, cell-index:%0lx\n",
+			*fman_dcpid, *oh_offset, *cell_index);
+
+	return BOND_OH_SUCCESS;
+}
+EXPORT_SYMBOL(export_oh_port_info_to_ceetm);
+/**
+ * Public APIs which can use for Link Aggregation and CEETM Qos
+ * show bond info and slave device info when they are available
+ */
+int show_dpa_slave_info(struct bonding *bond, struct slave *slave)
+{
+	struct dpa_priv_s *priv = netdev_priv(slave->dev);
+	if (bond)
+		pr_info("bond->dev->name:%s, slave_cnt:%d\n",
+			bond->dev->name, bond->slave_cnt);
+	if (slave)
+		pr_info("new_slave:%s\n", slave->dev->name);
+
+	if (is_dpa_eth_port(slave->dev)) {
+		pr_info("priv->mac_dev->res->start:%llx\n",
+				(unsigned long long)priv->mac_dev->res->start);
+		pr_info("get_dcp_id_from_dpa_eth_port(netdev):0x%0x\n",
+				get_dcp_id_from_dpa_eth_port(slave->dev));
+	} else
+		pr_info("the slave device %s is not a DPAA-Eth NIC\n",
+				slave->dev->name);
+
+	return 0;
+}
+
+int init_status(struct net_device *netdev)
+{
+	struct bonding *bond = master_to_bond(netdev);
+	memset(&bond->params.oh_stats, 0, sizeof(struct rtnl_link_stats64));
+
+	return BOND_OH_SUCCESS;
+}
+
+void add_statistics(struct bonding *bond, struct rtnl_link_stats64 *stats)
+{
+	stats->tx_packets += bond->params.oh_stats.tx_packets;
+	stats->tx_bytes += bond->params.oh_stats.tx_bytes;
+	stats->tx_errors += bond->params.oh_stats.tx_errors;
+	stats->tx_dropped += bond->params.oh_stats.tx_dropped;
+}
+/**
+ * Copied from oNIC (removed priv)
+ * Turn on HW checksum computation for this outgoing frame.
+ * If the current protocol is not something we support in this regard
+ * (or if the stack has already computed the SW checksum), we do nothing.
+ *
+ * Returns 0 if all goes well (or HW csum doesn't apply), and a negative value
+ * otherwise.
+ *
+ * Note that this function may modify the fd->cmd field and the skb data buffer
+ * (the Parse Results area).
+ */
+int oh_tx_csum_enable(struct sk_buff *skb,
+			struct qm_fd *fd,
+			char *parse_results)
+{
+	fm_prs_result_t *parse_result;
+	struct iphdr *iph;
+	struct ipv6hdr *ipv6h = NULL;
+	struct tcphdr *tcph;
+	struct udphdr *udph;
+	int l4_proto;
+	int ethertype = ntohs(skb->protocol);
+	int retval = 0, i;
+	unsigned char *p;
+
+	if (skb->ip_summed != CHECKSUM_PARTIAL)
+		return 0;
+
+	/* Fill in some fields of the Parse Results array, so the FMan
+	 * can find them as if they came from the FMan Parser.
+	 */
+	parse_result = (fm_prs_result_t *)parse_results;
+	/* If we're dealing with VLAN, get the real Ethernet type */
+	if (ethertype == ETH_P_8021Q) {
+		/* We can't always assume the MAC header is set correctly
+		 * by the stack, so reset to beginning of skb->data
+		 */
+		skb_reset_mac_header(skb);
+		ethertype = ntohs(vlan_eth_hdr(skb)->h_vlan_encapsulated_proto);
+		/* below l2r need look up FMAN RM to verify */
+		parse_result->l2r = FM_PR_L2_VLAN | FM_PR_L2_VLAN_STACK;
+	} else {
+		parse_result->l2r = FM_PR_L2_ETHERNET;
+	}
+
+	/* Fill in the relevant L3 parse result fields
+	 * and read the L4 protocol type
+	 */
+	switch (ethertype) {
+	case ETH_P_IP:
+		parse_result->l3r = FM_L3_PARSE_RESULT_IPV4;
+		iph = ip_hdr(skb);
+		BUG_ON(iph == NULL);
+		l4_proto = ntohs(iph->protocol);
+		break;
+	case ETH_P_IPV6:
+		parse_result->l3r = FM_L3_PARSE_RESULT_IPV6;
+		ipv6h = ipv6_hdr(skb);
+		BUG_ON(ipv6h == NULL);
+		l4_proto = ntohs(ipv6h->nexthdr);
+		break;
+	default:
+		/* We shouldn't even be here */
+		hw_lag_dbg("Can't compute HW csum for L3 proto 0x%x\n",
+			     ntohs(skb->protocol));
+		retval = -EIO;
+		goto return_error;
+	}
+
+	hw_lag_dbg("skb->protocol(L3):0x%04x, ethertype:%x\n",
+			ntohs(skb->protocol), ethertype);
+
+	/* Fill in the relevant L4 parse result fields */
+	switch (l4_proto) {
+	case IPPROTO_UDP:
+		parse_result->l4r = FM_L4_PARSE_RESULT_UDP;
+		udph = (struct udphdr *)(skb->data + skb_transport_offset(skb));
+		hw_lag_dbg("udp org csum:0x%0x\n", udph->check);
+		skb_set_transport_header(skb, skb_checksum_start_offset(skb));
+		skb_checksum_help(skb);
+		hw_lag_dbg("udp software csum:0x%0x\n", udph->check);
+		break;
+	case IPPROTO_TCP:
+		parse_result->l4r = FM_L4_PARSE_RESULT_TCP;
+		tcph = (struct tcphdr *)(skb->data + skb_transport_offset(skb));
+		p = skb->data;
+		hw_lag_dbg("\ndmac:%02x:%02x:%02x:%02x:%02x:%02x\n"
+			"smac:%02x:%02x:%02x:%02x:%02x:%02x\n"
+			"h_proto:0x%04x\n", p[0], p[1], p[2], p[3], p[4], p[5],
+			p[6], p[7], p[8], p[9], p[10], p[11],
+			*((short *)(p + 12)));
+
+		/* dump skb data info for manual calculate L4CSUM,
+		 * jump over net header first
+		 */
+		p += skb_network_offset(skb);
+		for (i = 0; i < skb->len - skb_network_offset(skb) - 4; i += 4)
+			hw_lag_dbg("%08x\n", *((unsigned int *) (p + i)));
+
+		for (; i < skb->len - skb_network_offset(skb); i++)
+			hw_lag_dbg("%02x\n", *(p + i));
+
+		hw_lag_dbg("tcp org csum:0x%0x.\n", tcph->check);
+		skb_set_transport_header(skb, skb_checksum_start_offset(skb));
+		skb_checksum_help(skb);
+		hw_lag_dbg("tcp software csum:0x%0x,\n", tcph->check);
+
+		break;
+	default:
+		/* This can as well be a BUG() */
+		pr_err("Can't compute HW csum for L4 proto 0x%x\n",
+				l4_proto);
+		retval = -EIO;
+		goto return_error;
+	}
+
+	hw_lag_dbg("l4_proto:0x%04x, result->l2r:0x%04x\n",
+			l4_proto, parse_result->l2r);
+	hw_lag_dbg("result->l3r:0x%04x, result->l4r:0x%02x.\n",
+			parse_result->l3r, parse_result->l4r);
+
+	/* At index 0 is IPOffset_1 as defined in the Parse Results */
+	parse_result->ip_off[0] = skb_network_offset(skb);
+	parse_result->l4_off = skb_transport_offset(skb);
+
+	/* Enable L3 (and L4, if TCP or UDP) HW checksum. */
+	fd->cmd |= FM_FD_CMD_RPD | FM_FD_CMD_DTC;
+
+	/* On P1023 and similar platforms fd->cmd interpretation could
+	 * be disabled by setting CONTEXT_A bit ICMD; currently this bit
+	 * is not set so we do not need to check; in the future, if/when
+	 * using context_a we need to check this bit
+	 */
+
+return_error:
+	return retval;
+}
+
+static int __hot dpa_oh_xmit(struct qm_fd *fd, struct qman_fq *tx_fq)
+{
+	int err, i;
+
+	for (i = 0; i < 100000; i++) {
+		err = qman_enqueue(tx_fq, fd, 0);
+		if (err != -EBUSY)
+			break;
+	}
+
+	if (unlikely(err < 0)) {
+		/* TODO differentiate b/w -EBUSY (EQCR full) and other codes? */
+		pr_err("qman_enqueue() error.\n");
+		return err;
+	}
+
+	return 0;
+}
+
+
+int __hot dpa_oh_tx(struct sk_buff *skb, struct bonding *bond,
+		struct net_device *net_dev, struct dpa_fq *tx_fq)
+{
+	struct dpa_priv_s	*priv;
+	struct dpa_bp *bp = bond->params.ohp->tx_bp;
+
+	struct sk_buff **skbh = NULL;
+	dma_addr_t addr;
+	struct qm_fd fd;
+	int err = 0;
+	int *countptr;
+	struct rtnl_link_stats64 *percpu_stats;
+
+	tx_fq->net_dev = bond->params.ohp->slave[0]->dev;
+	priv = netdev_priv(bond->params.ohp->slave[0]->dev);
+	percpu_stats = &bond->params.oh_stats;
+	countptr = __this_cpu_ptr(bond->params.ohp->tx_bp->percpu_count);
+
+	if (unlikely(skb_headroom(skb) < priv->tx_headroom)) {
+		struct sk_buff *skb_new;
+
+		skb_new = skb_realloc_headroom(skb, priv->tx_headroom);
+		if (unlikely(!skb_new)) {
+			percpu_stats->tx_errors++;
+			kfree_skb(skb);
+			goto done;
+		}
+		kfree_skb(skb);
+		skb = skb_new;
+	}
+
+	clear_fd(&fd);
+
+	/* store skb backpointer to release the skb later */
+	skbh = (struct sk_buff **)(skb->data - priv->tx_headroom);
+	*skbh = skb;
+
+	/* TODO check if skb->len + priv->tx_headroom < bp->size */
+
+	/* Enable L3/L4 hardware checksum computation.
+	 *
+	 * We must do this before dma_map_single(), because we may
+	 * need to write into the skb.
+	 */
+	err = oh_tx_csum_enable(skb, &fd,
+			((char *)skbh) + DPA_TX_PRIV_DATA_SIZE);
+
+	if (unlikely(err < 0)) {
+		pr_err("HW csum error: %d\n", err);
+
+		return err;
+	}
+
+	addr = dma_map_single(bp->dev, skbh,
+			skb->len + priv->tx_headroom, DMA_TO_DEVICE);
+	if (unlikely(dma_mapping_error(bp->dev, addr))) {
+			pr_err("dma_map_single() failed\n");
+		goto dma_mapping_failed;
+	}
+
+	fd.format = qm_fd_contig;
+	fd.length20 = skb->len;
+	fd.offset = priv->tx_headroom;
+	fd.addr_hi = upper_32_bits(addr);
+	fd.addr_lo = lower_32_bits(addr);
+	/* fd.cmd |= FM_FD_CMD_FCO; */
+	fd.bpid = bp->bpid;
+
+	/* (Partially) drain the Draining Buffer Pool pool; each core
+	 * acquires at most the number of buffers it put there; since
+	 * BMan allows for up to 8 buffers to be acquired at one time,
+	 * work in batches of 8 for efficiency reasons
+	 */
+	dpa_oh_drain_bp(bp);
+
+	if (unlikely(dpa_oh_xmit(&fd, &tx_fq->fq_base) < 0)) {
+		/* oh tx error, add statistics */
+		bond->params.oh_stats.tx_packets++;
+		bond->params.oh_stats.tx_errors++;
+		hw_lag_dbg("3ad enqueue_pkt error...txerr_pkt:%llu\n",
+				bond->params.oh_stats.tx_packets);
+		goto xmit_failed;
+	} else {
+		/* oh tx OK, add statistics */
+		bond->params.oh_stats.tx_packets++;
+		bond->params.oh_stats.tx_bytes += skb->len;
+		hw_lag_dbg("3ad enqueue_pkt OK...tx_pkt:%llu\n",
+				bond->params.oh_stats.tx_packets);
+		return NETDEV_TX_OK;
+	}
+
+	countptr = __this_cpu_ptr(bp->percpu_count);
+	(*countptr)++;
+
+	goto done;
+
+xmit_failed:
+	dma_unmap_single(bp->dev, addr, fd.offset + fd.length20, DMA_TO_DEVICE);
+dma_mapping_failed:
+	percpu_stats->tx_errors++;
+	dev_kfree_skb(skb);
+done:
+	return NETDEV_TX_OK;
+}
+/**
+ * Enqueue one skb pkt to offline port which attached to a bond.
+ * bond: current bond's pointer
+ * skb:  pkt which will be enqueued to the offline port
+ * ceetm_fq: pkt will use this fq for xmit. if this ceetm_fq is
+ * pointing to NULL, will use default tx_fq for xmit.
+ * return BOND_OH_SUCCESS if enqueued, otherwise return errors.
+ */
+int enqueue_pkt_to_oh(struct bonding *bond, struct sk_buff *skb,
+		struct dpa_fq *ceetm_fq)
+{
+	struct oh_port_priv *p_oh = bond->params.ohp;
+	struct net_device *slave_netdev = NULL;
+	struct dpa_fq *tx_fq = p_oh->oh_tx_lag_fqs;
+
+	slave_netdev = p_oh->slave[0]->dev;
+
+	p_oh->oh_errq->net_dev = slave_netdev;
+	p_oh->oh_defq->net_dev = slave_netdev;
+
+	if (!is_dpa_eth_port(slave_netdev)) {
+		pr_err("is not dpaa NIC or NULL pointer.\n");
+		return -EINVAL;
+	}
+
+	if (ceetm_fq)
+		return dpa_oh_tx(skb, bond, slave_netdev, ceetm_fq);
+	else
+		return dpa_oh_tx(skb, bond, slave_netdev, tx_fq);
+}
+EXPORT_SYMBOL(enqueue_pkt_to_oh);
+
+static int get_dpa_slave_info(struct slave *slave, uint16_t *tx_channel)
+{
+	struct dpa_priv_s *priv = netdev_priv(slave->dev);
+
+	if (!is_dpa_eth_port(slave->dev) || !(priv->mac_dev))
+		return BOND_OH_ERROR;
+
+	*tx_channel = fm_get_tx_port_channel(priv->mac_dev->port_dev[TX]);
+
+	return BOND_OH_SUCCESS;
+}
+
+int get_dpa_slave_info_ex(struct slave *slave, uint16_t *tx_channel,
+		struct qman_fq **egress_fq, u32 *first_fqid)
+{
+	struct dpa_priv_s *priv = netdev_priv(slave->dev);
+
+	if (!is_dpa_eth_port(slave->dev) || !(priv->mac_dev))
+		return BOND_OH_ERROR;
+
+	*tx_channel	= fm_get_tx_port_channel(priv->mac_dev->port_dev[TX]);
+	*egress_fq	= priv->egress_fqs[0];
+	*first_fqid	= priv->egress_fqs[0]->fqid;
+
+	return BOND_OH_SUCCESS;
+}
+
+/* Creates Frame Queues, these 2 good subroutines are completely copied from
+ * Bogdan Purcareata's good patch "Offline port queues initialization", HW_LAG
+ * need to initialize FQs for an offline port PCD usage with tx_channel/wq of
+ * slave devices which have already attached to a bond, HW_LAG OH port dequeue,
+ * then enqueue PCD FQs to DPA-Eth via these PCD FQs.
+ */
+static int create_oh_pcd_fq(struct qman_fq *fq, u32 fqid_pcdconf,
+	uint32_t fq_id, uint16_t tx_channel, uint16_t wq_id)
+{
+	struct qm_mcc_initfq fq_opts;
+	uint32_t create_flags, init_flags;
+	uint32_t ret = 0;
+
+	if (!fq)
+		return BOND_OH_ERROR;
+
+	/* Set flags for FQ create */
+	create_flags = QMAN_FQ_FLAG_TO_DCPORTAL;
+
+	/* set egress_ern callback for pcd fqs */
+	*fq = oh_pcd_egress_ernq;
+
+	/* Create frame queue */
+	ret = qman_create_fq(fq_id, create_flags, fq);
+	if (ret != 0)
+		return BOND_OH_ERROR;
+
+	/* Set flags for FQ init */
+	init_flags = QMAN_INITFQ_FLAG_SCHED;
+
+	/* Set FQ init options. Specify destination WQ ID and channel */
+	memset(&fq_opts, 0, sizeof(fq_opts));
+	fq_opts.we_mask = QM_INITFQ_WE_DESTWQ;
+	fq_opts.fqd.dest.wq = wq_id;
+	fq_opts.fqd.dest.channel = tx_channel;
+
+	fq_opts.we_mask |= QM_INITFQ_WE_FQCTRL;
+	fq_opts.fqd.fq_ctrl = QM_FQCTRL_PREFERINCACHE;
+	fq_opts.fqd.fq_ctrl |= QM_FQCTRL_CTXASTASHING | QM_FQCTRL_AVOIDBLOCK;
+	fq_opts.fqd.context_a.stashing.exclusive = QM_STASHING_EXCL_DATA |
+		QM_STASHING_EXCL_CTX | QM_STASHING_EXCL_ANNOTATION;
+	fq_opts.fqd.context_a.stashing.data_cl = 2;
+	fq_opts.fqd.context_a.stashing.annotation_cl = 1;
+	fq_opts.fqd.context_a.stashing.context_cl =
+		DIV_ROUND_UP(sizeof(struct qman_fq), 64);
+
+	fq_opts.we_mask |= QM_INITFQ_WE_CONTEXTA | QM_INITFQ_WE_CONTEXTB;
+	/**
+	 * CTXA[OVFQ] = 1
+	 * we set PCD own confirmation Q and their own callback in case of
+	 * interrupt DPA-Eth private conf callback/err callback/def callback.
+	 */
+	fq_opts.fqd.context_a.hi = 0x80000000;
+	fq_opts.fqd.context_a.lo = 0x0;
+	fq_opts.fqd.context_b = fqid_pcdconf;
+
+	/* Initialize frame queue */
+	ret = qman_init_fq(fq, init_flags, &fq_opts);
+	if (ret != 0) {
+		qman_destroy_fq(fq, 0);
+		return BOND_OH_ERROR;
+	}
+	hw_lag_dbg("FQ create_flags:0X%0x, init_flags:0X%0x\n",
+			create_flags, init_flags);
+
+	return BOND_OH_SUCCESS;
+}
+
+static int hw_lag_allocate_pcd_queues(struct device *dev,
+		struct dpa_fq **p_pcd_fq, u32 fqid_pcdconf, u32 fqid,
+		uint16_t tx_channel, uint16_t wq)
+{
+	/* Allocate pcd queues */
+	int errno = BOND_OH_SUCCESS;
+	struct dpa_fq *pcd_fq;
+	hw_lag_dbg("Allocating PCD queues...p_pcd_fq:%p, fqid:%d\n",
+			*p_pcd_fq, fqid);
+	pcd_fq = devm_kzalloc(dev, sizeof(struct dpa_fq), GFP_KERNEL);
+	if (!pcd_fq) {
+		pr_err("can't allocate slave PCD FQ!\n");
+		errno = -ENOMEM;
+		goto return_kfree;
+	}
+
+	hw_lag_dbg("Allocated pcd_fq:%p, fqid:%d\n", pcd_fq, fqid);
+	/* Create pcd queues */
+	errno = create_oh_pcd_fq(&pcd_fq->fq_base, fqid_pcdconf,
+			fqid, tx_channel, wq);
+	if (errno != BOND_OH_SUCCESS) {
+		pr_err("can't create lag PCD FQ:%u\n", fqid);
+		errno = -EINVAL;
+		goto return_kfree;
+	}
+
+	*p_pcd_fq = pcd_fq;
+	hw_lag_dbg("created pcd_fq:%p, fqid:%d, *p_pcd_fq:%p\n",
+			pcd_fq, fqid, *p_pcd_fq);
+	return BOND_OH_SUCCESS;
+
+return_kfree:
+	if (pcd_fq)
+		devm_kfree(dev, pcd_fq);
+	return errno;
+}
+
+/* Destroys Frame Queues */
+static void hw_lag_fq_destroy(struct qman_fq *fq)
+{
+	int errno = BOND_OH_SUCCESS;
+
+	errno = qman_retire_fq(fq, NULL);
+	if (unlikely(errno < 0))
+		pr_err("qman_retire_fq(%u)=%d\n", qman_fq_fqid(fq), errno);
+
+	errno = qman_oos_fq(fq);
+	if (unlikely(errno < 0))
+		pr_err("qman_oos_fq(%u)=%d\n", qman_fq_fqid(fq), errno);
+
+	qman_destroy_fq(fq, 0);
+}
+/* release fq memory */
+static int hw_lag_release_fq(struct device *dev, struct dpa_fq *fq)
+{
+
+	if (!fq)
+		return BOND_OH_ERROR;
+
+	hw_lag_fq_destroy(&fq->fq_base);
+	if (!dev)
+		return BOND_OH_ERROR;
+
+	devm_kfree(dev, fq);
+
+	return BOND_OH_SUCCESS;
+}
+/**
+ * Get DPA slave device information: wq/channel_id, allocate FQID/FQ memory,
+ * then set FQ flags, record the slave pointer in case of remove these
+ * information when detaching slave in the future.
+ */
+int fill_oh_pcd_fqs_with_slave_info(struct bonding *bond, struct slave *slave)
+{
+	uint16_t tx_channel;
+	struct dpa_fq *pcd_fq = NULL;
+	struct oh_port_priv *cur;
+	u32 fqid;
+	uint16_t wq_id = 3; /* the default value in DPA-Eth private driver */
+
+	if (bond->params.mode != BOND_MODE_8023AD) {
+		pr_err("error, only support 802.3ad when fill OH FQs.\n");
+		return BOND_OH_ERROR;
+	}
+
+	if (is_dpa_eth_port(slave->dev) == false) {
+		pr_err("error, only support dpa eth nic.\n");
+		return BOND_OH_ERROR;
+	}
+
+	if (bond->slave_cnt > SLAVES_PER_BOND) {
+		pr_err("error, only support 2 dpa nic per bond.\n");
+		return BOND_OH_ERROR;
+	}
+
+	if (get_dpa_slave_info(slave, &tx_channel) == BOND_OH_ERROR) {
+		pr_err("error on getting dpa info when fill OH FQs.\n");
+		return BOND_OH_ERROR;
+	}
+
+	cur = bond->params.ohp;
+	if (!cur) {
+		pr_err("have not bind an OH port,\n");
+		pr_err("will use software tx traffic distribution.\n");
+		return BOND_OH_ERROR;
+	}
+
+	hw_lag_dbg("cur->pcd_fqs[0]:%p, cur->pcd_fqs[1]:%p\n",
+			cur->pcd_fqs[0], cur->pcd_fqs[1]);
+	if (!cur->pcd_fqs[0])
+		fqid = cur->pcd_fqids_base;
+	else
+		fqid = cur->pcd_fqids_base + 1;
+
+	hw_lag_dbg("pcd_fq:%p, fqid:%d Before alloc.\n", pcd_fq, fqid);
+
+	if (hw_lag_allocate_pcd_queues(cur->dpa_oh_dev, &pcd_fq,
+				cur->fqid_pcdconf, fqid, tx_channel,
+				wq_id) == BOND_OH_ERROR) {
+		pr_err("error on create pcd fqs\n");
+		return BOND_OH_ERROR;
+	}
+
+	hw_lag_dbg("pcd_fq:%p, fqid:%d, tx_channel:%d, wq_id:%d After alloc.\n",
+			pcd_fq, fqid, tx_channel, wq_id);
+	hw_lag_dbg("fqid:0x%0x, tx_channel:0x%0x, wq_id:0x%0x After alloc.\n",
+			fqid, tx_channel, wq_id);
+
+	if (!cur->pcd_fqs[0]) {
+		cur->pcd_fqs[0] = pcd_fq;
+		cur->slave[0] = slave;
+	} else if (!cur->pcd_fqs[1]) {
+		cur->pcd_fqs[1] = pcd_fq;
+		cur->slave[1] = slave;
+	}
+
+	return BOND_OH_SUCCESS;
+}
+
+/* forget offline port pcd information according to slave pointer,
+ * then destroy fq and release the fq memory.
+ */
+int del_oh_pcd_fqs_with_slave_info(struct bonding *bond, struct slave *slave)
+{
+	struct oh_port_priv *cur;
+	struct dpa_fq *pcd_fq;
+
+	if (is_dpa_eth_port(slave->dev) == false) {
+		pr_err("error, only support dpa eth nic.\n");
+		return BOND_OH_ERROR;
+	}
+	cur = bond->params.ohp;
+	if (!cur) {
+		pr_err("have not bind an OH port,\n");
+		pr_err("will use software tx traffic distribution.\n");
+		return BOND_OH_ERROR;
+	}
+	if (slave == cur->slave[0]) {
+		pcd_fq = cur->pcd_fqs[0];
+		cur->pcd_fqs[0] = NULL;
+		cur->slave[0] = NULL;
+	} else if (slave == cur->slave[1]) {
+		pcd_fq = cur->pcd_fqs[1];
+		cur->pcd_fqs[1] = NULL;
+		cur->slave[1] = NULL;
+	} else
+		pcd_fq = NULL;
+
+	return hw_lag_release_fq(cur->dpa_oh_dev, pcd_fq);
+}
+
+int release_pcd_mem(struct bonding *bond)
+{
+	return BOND_OH_SUCCESS;
+}
+
+/* get all offline port information from bond, including
+ * dev,oh handler, PCD FQid base and PCD FQ count, then
+ * get the new xmit policy, copy schemes needed from the
+ * cached_scheme pointer, config PCD params, init PCD dev,
+ * set PCD Net Env Characteristics, then set Keygen Scheme
+ * params to the PCD dev, disable offline port, set PCD
+ * params to the offline port dev, at last enable the offline
+ * port.
+ * this subroutine return true when it can apply PCD to
+ * the offline port, otherwise return false.
+ */
+bool apply_pcd(struct bonding *bond, int new_xmit_policy)
+{
+	return true;
+}
diff --git a/drivers/net/bonding/hw_distribution.h b/drivers/net/bonding/hw_distribution.h
new file mode 100644
index 0000000..55e0420
--- /dev/null
+++ b/drivers/net/bonding/hw_distribution.h
@@ -0,0 +1,138 @@
+/**
+ * Copyright 2014 Freescale Semiconductor Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *     * Redistributions of source code must retain the above copyright
+ *	 notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *	 notice, this list of conditions and the following disclaimer in the
+ *	 documentation and/or other materials provided with the distribution.
+ *     * Neither the name of Freescale Semiconductor nor the
+ *	 names of its contributors may be used to endorse or promote products
+ *	 derived from this software without specific prior written permission.
+ *
+ *
+ * ALTERNATIVELY, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") as published by the Free Software
+ * Foundation, either version 2 of that License or (at your option) any
+ * later version.
+ *
+ * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY
+ * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY
+ * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __HARDWARE_DISTRIBUTION_H
+#define __HARDWARE_DISTRIBUTION_H
+
+#include <linux/of_platform.h>
+
+#include "bonding.h"
+#include "bond_3ad.h"
+#include "bond_alb.h"
+#include "offline_port.h"
+#include "dpaa_eth.h"
+#include "dpaa_eth_common.h"
+
+#define OHFRIENDNAMSIZ	10	/* fman0-oh@1, ...  fman1-oh@6 */
+#define OHNODENAMSIZ	24	/* /fsl,dpaa/dpa-fman0-oh@1 */
+#define BOND_OH_SUCCESS	0
+#define BOND_OH_ERROR	-1
+#define NO_POLICY	0xFF	/* this is a magic number */
+
+#define FM1_GB0 0xffe4e0000
+#define FM1_10G 0xffe4f0000
+#define FM2_GB0 0xffe5e0000
+#define FM2_10G 0xffe5f0000
+
+#define DPA_FQ_TD 0x200000
+
+/* There are 4 FMAN Ethernet Ports per T1040, 2 of them are for the
+ * Link Aggregation for the L2Swith trunk link, thus there are at
+ * most 2 ports left for the other Link Aggregation, this implies
+ * 2 MAX_BOND_CNT * SLAVES_PER_BOND = 4 FMAN Ethernet Ports.
+ * In fact,we only need numbers of offline port in a DTS:
+ * offline port count = min(FM_MAX_NUM_OF_OH_PORTS, MAX_BOND_CNT)
+ */
+#define MAX_BOND_CNT	2
+#define SLAVES_PER_BOND	2
+
+#ifdef CONFIG_HW_LAG_DEBUG
+#define hw_lag_dbg(fmt, arg...)  \
+	pr_info("LAG:[CPU %d ln %d fn %s] - " fmt, smp_processor_id(), \
+			__LINE__, __func__, ##arg)
+#else
+#define hw_lag_dbg(fmt, arg...)
+#endif
+
+struct oh_port_priv {
+	unsigned int	oh_channel_id;
+	struct	dpa_oh_config_s	*oh_config;
+	struct	dpa_fq	*pcd_fqs[SLAVES_PER_BOND];
+	struct	dpa_fq *oh_defq, *oh_errq;
+	uint16_t p_oh_rcv_channel;
+	struct	slave *slave[SLAVES_PER_BOND];
+	u32	pcd_fqids_base;
+	uint32_t fqid_pcderr, fqid_pcdconf, fqid_ohtxconf;
+	struct	dpa_fq *oh_pcderrq, *oh_pcdconfq, *oh_txconfq;
+	/* init dynamic particular tx fqs of offline port for LAG xmit,
+	 * does not reuse tx fqs initialized by offline port driver.
+	 */
+	struct	dpa_fq *oh_tx_lag_fqs;
+	const	phandle	*p_oh_port_handle;
+	struct	platform_device	*oh_of_dev, *of_dev;
+	struct	device	*dpa_oh_dev, *oh_dev;
+	struct	device_node	*dpa_oh_node, *oh_node;
+	struct	dpa_bp *tx_bp;
+	struct	dpa_buffer_layout_s *tx_buf_layout;
+	uint8_t bpid;	/**< External buffer pool id */
+	uint16_t bp_size;	/**< External buffer pool buffer size */
+	int	oh_en;	/* enable or disable offline port's help at run-time */
+	unsigned char friendname[OHFRIENDNAMSIZ];
+	unsigned long cell_index;
+};
+
+extern struct oh_port_priv *poh;
+extern int available_num_of_oh_ports;
+
+int get_oh_info(void);
+unsigned int to_which_oh_i_attached(struct oh_port_priv *current_poh);
+bool are_all_slaves_linkup(struct bonding *bond);
+int get_dcp_id_from_dpa_eth_port(struct net_device *netdev);
+int export_oh_port_info_to_ceetm(struct bonding *bond, uint16_t *channel,
+		unsigned long *fman_dcpid, unsigned long *oh_offset,
+		unsigned long *cell_index);
+int show_dpa_slave_info(struct bonding *bond, struct slave *slave);
+int get_dpa_slave_info_ex(struct slave *slave, uint16_t *tx_channel,
+		struct qman_fq **egress_fq, u32 *first_fqid);
+int enqueue_pkt_to_oh(struct bonding *bond, struct sk_buff *skb,
+		struct dpa_fq *ceetm_fq);
+ssize_t bonding_show_offline_port_xmit_statistics(struct device *d,
+		struct device_attribute *attr, char *buf);
+ssize_t bonding_show_offline_ports(struct device *d,
+		struct device_attribute *attr, char *buf);
+ssize_t bonding_show_oh_needed_for_hw_distribution(struct device *d,
+		struct device_attribute *attr, char *buf);
+ssize_t bonding_store_oh_needed_for_hw_distribution(struct device *d,
+		struct device_attribute *attr, const char *buffer,
+		size_t count);
+ssize_t bonding_show_oh_enable(struct device *d,
+		struct device_attribute *attr, char *buf);
+ssize_t bonding_store_oh_enable(struct device *d,
+		struct device_attribute *attr, const char *buffer,
+		size_t count);
+int fill_oh_pcd_fqs_with_slave_info(struct bonding *bond, struct slave *slave);
+int del_oh_pcd_fqs_with_slave_info(struct bonding *bond, struct slave *slave);
+bool apply_pcd(struct bonding *bond, int new_xmit_policy);
+int release_pcd_mem(struct bonding *bond);
+int init_status(struct net_device *netdev);
+void add_statistics(struct bonding *bond, struct rtnl_link_stats64 *stats);
+#endif /* __HARDWARE_DISTRIBUTION_H */