staging: fsl_ppfe/eth: introduce pfe driver

[pfe part]

	This patch introduces Linux support for NXP's LS1012A Packet
Forwarding Engine (pfe_eth). LS1012A uses hardware packet forwarding
engine to provide high performance Ethernet interfaces. The device
includes two Ethernet ports.

Signed-off-by: Calvin Johnson <calvin.johnson@nxp.com>
Signed-off-by: Anjaneyulu Jagarlmudi <anji.jagarlmudi@nxp.com>
Integrated-by: Zhao Qiang <qiang.zhao@nxp.com>

14 files changed, 7915 insertions, 0 deletions

diff --git a/drivers/staging/fsl_ppfe/Kconfig b/drivers/staging/fsl_ppfe/Kconfig
new file mode 100644
index 0000000..e409643
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/Kconfig
@@ -0,0 +1,20 @@
+#
+# Freescale Programmable Packet Forwarding Engine driver
+#
+config FSL_PPFE
+	bool "Freescale PPFE Driver"
+	default n
+	---help---
+	Freescale LS1012A SoC has a Programmable Packet Forwarding Engine.
+	It provides two high performance ethernet interfaces.
+	This driver initializes, programs and controls the PPFE.
+	Use this driver to enable network connectivity on LS1012A platforms.
+
+if FSL_PPFE
+
+config FSL_PPFE_UTIL_DISABLED
+	bool "Disable PPFE UTIL Processor Engine"
+	---help---
+	UTIL PE has to be enabled only if required.
+
+endif # FSL_PPFE
diff --git a/drivers/staging/fsl_ppfe/Makefile b/drivers/staging/fsl_ppfe/Makefile
new file mode 100644
index 0000000..07cd351
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/Makefile
@@ -0,0 +1,19 @@
+#
+# Makefile for Freesecale PPFE driver
+#
+
+ccflags-y +=  -I$(src)/include  -I$(src)
+
+obj-m += pfe.o
+
+pfe-y += pfe_mod.o \
+	pfe_hw.o \
+	pfe_firmware.o \
+	pfe_ctrl.o \
+	pfe_hif.o \
+	pfe_hif_lib.o\
+	pfe_eth.o \
+	pfe_sysfs.o \
+	pfe_debugfs.o \
+	pfe_ls1012a_platform.o \
+	pfe_hal.o
diff --git a/drivers/staging/fsl_ppfe/TODO b/drivers/staging/fsl_ppfe/TODO
new file mode 100644
index 0000000..43c48cc
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/TODO
@@ -0,0 +1,2 @@
+TODO:
+	- provide pfe pe monitoring support
diff --git a/drivers/staging/fsl_ppfe/pfe_ctrl.c b/drivers/staging/fsl_ppfe/pfe_ctrl.c
new file mode 100644
index 0000000..dfa8547
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_ctrl.c
@@ -0,0 +1,238 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/kthread.h>
+
+#include "pfe_mod.h"
+#include "pfe_ctrl.h"
+
+#define TIMEOUT_MS	1000
+
+int relax(unsigned long end)
+{
+	if (time_after(jiffies, end)) {
+		if (time_after(jiffies, end + (TIMEOUT_MS * HZ) / 1000))
+			return -1;
+
+		if (need_resched())
+			schedule();
+	}
+
+	return 0;
+}
+
+void pfe_ctrl_suspend(struct pfe_ctrl *ctrl)
+{
+	int id;
+
+	mutex_lock(&ctrl->mutex);
+
+	for (id = CLASS0_ID; id <= CLASS_MAX_ID; id++)
+		pe_dmem_write(id, cpu_to_be32(0x1), CLASS_DM_RESUME, 4);
+
+	for (id = TMU0_ID; id <= TMU_MAX_ID; id++) {
+		if (id == TMU2_ID)
+			continue;
+		pe_dmem_write(id, cpu_to_be32(0x1), TMU_DM_RESUME, 4);
+	}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	pe_dmem_write(UTIL_ID, cpu_to_be32(0x1), UTIL_DM_RESUME, 4);
+#endif
+	mutex_unlock(&ctrl->mutex);
+}
+
+void pfe_ctrl_resume(struct pfe_ctrl *ctrl)
+{
+	int pe_mask = CLASS_MASK | TMU_MASK;
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	pe_mask |= UTIL_MASK;
+#endif
+	mutex_lock(&ctrl->mutex);
+	pe_start(&pfe->ctrl, pe_mask);
+	mutex_unlock(&ctrl->mutex);
+}
+
+/* PE sync stop.
+ * Stops packet processing for a list of PE's (specified using a bitmask).
+ * The caller must hold ctrl->mutex.
+ *
+ * @param ctrl		Control context
+ * @param pe_mask	Mask of PE id's to stop
+ *
+ */
+int pe_sync_stop(struct pfe_ctrl *ctrl, int pe_mask)
+{
+	struct pe_sync_mailbox *mbox;
+	int pe_stopped = 0;
+	unsigned long end = jiffies + 2;
+	int i;
+
+	pe_mask &= 0x2FF;  /*Exclude Util + TMU2 */
+
+	for (i = 0; i < MAX_PE; i++)
+		if (pe_mask & (1 << i)) {
+			mbox = (void *)ctrl->sync_mailbox_baseaddr[i];
+
+			pe_dmem_write(i, cpu_to_be32(0x1), (unsigned
+					long)&mbox->stop, 4);
+		}
+
+	while (pe_stopped != pe_mask) {
+		for (i = 0; i < MAX_PE; i++)
+			if ((pe_mask & (1 << i)) && !(pe_stopped & (1 << i))) {
+				mbox = (void *)ctrl->sync_mailbox_baseaddr[i];
+
+				if (pe_dmem_read(i, (unsigned
+					long)&mbox->stopped, 4) &
+					cpu_to_be32(0x1))
+					pe_stopped |= (1 << i);
+			}
+
+		if (relax(end) < 0)
+			goto err;
+	}
+
+	return 0;
+
+err:
+	pr_err("%s: timeout, %x %x\n", __func__, pe_mask, pe_stopped);
+
+	for (i = 0; i < MAX_PE; i++)
+		if (pe_mask & (1 << i)) {
+			mbox = (void *)ctrl->sync_mailbox_baseaddr[i];
+
+			pe_dmem_write(i, cpu_to_be32(0x0), (unsigned
+					long)&mbox->stop, 4);
+	}
+
+	return -EIO;
+}
+
+/* PE start.
+ * Starts packet processing for a list of PE's (specified using a bitmask).
+ * The caller must hold ctrl->mutex.
+ *
+ * @param ctrl		Control context
+ * @param pe_mask	Mask of PE id's to start
+ *
+ */
+void pe_start(struct pfe_ctrl *ctrl, int pe_mask)
+{
+	struct pe_sync_mailbox *mbox;
+	int i;
+
+	for (i = 0; i < MAX_PE; i++)
+		if (pe_mask & (1 << i)) {
+			mbox = (void *)ctrl->sync_mailbox_baseaddr[i];
+
+			pe_dmem_write(i, cpu_to_be32(0x0), (unsigned
+					long)&mbox->stop, 4);
+		}
+}
+
+/* This function will ensure all PEs are put in to idle state */
+int pe_reset_all(struct pfe_ctrl *ctrl)
+{
+	struct pe_sync_mailbox *mbox;
+	int pe_stopped = 0;
+	unsigned long end = jiffies + 2;
+	int i;
+	int pe_mask  = CLASS_MASK | TMU_MASK;
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	pe_mask |= UTIL_MASK;
+#endif
+
+	for (i = 0; i < MAX_PE; i++)
+		if (pe_mask & (1 << i)) {
+			mbox = (void *)ctrl->sync_mailbox_baseaddr[i];
+
+			pe_dmem_write(i, cpu_to_be32(0x2), (unsigned
+					long)&mbox->stop, 4);
+		}
+
+	while (pe_stopped != pe_mask) {
+		for (i = 0; i < MAX_PE; i++)
+			if ((pe_mask & (1 << i)) && !(pe_stopped & (1 << i))) {
+				mbox = (void *)ctrl->sync_mailbox_baseaddr[i];
+
+				if (pe_dmem_read(i, (unsigned long)
+							&mbox->stopped, 4) &
+						cpu_to_be32(0x1))
+					pe_stopped |= (1 << i);
+			}
+
+		if (relax(end) < 0)
+			goto err;
+	}
+
+	return 0;
+
+err:
+	pr_err("%s: timeout, %x %x\n", __func__, pe_mask, pe_stopped);
+	return -EIO;
+}
+
+int pfe_ctrl_init(struct pfe *pfe)
+{
+	struct pfe_ctrl *ctrl = &pfe->ctrl;
+	int id;
+
+	pr_info("%s\n", __func__);
+
+	mutex_init(&ctrl->mutex);
+	spin_lock_init(&ctrl->lock);
+
+	for (id = CLASS0_ID; id <= CLASS_MAX_ID; id++) {
+		ctrl->sync_mailbox_baseaddr[id] = CLASS_DM_SYNC_MBOX;
+		ctrl->msg_mailbox_baseaddr[id] = CLASS_DM_MSG_MBOX;
+	}
+
+	for (id = TMU0_ID; id <= TMU_MAX_ID; id++) {
+		if (id == TMU2_ID)
+			continue;
+		ctrl->sync_mailbox_baseaddr[id] = TMU_DM_SYNC_MBOX;
+		ctrl->msg_mailbox_baseaddr[id] = TMU_DM_MSG_MBOX;
+	}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	ctrl->sync_mailbox_baseaddr[UTIL_ID] = UTIL_DM_SYNC_MBOX;
+	ctrl->msg_mailbox_baseaddr[UTIL_ID] = UTIL_DM_MSG_MBOX;
+#endif
+
+	ctrl->hash_array_baseaddr = pfe->ddr_baseaddr + ROUTE_TABLE_BASEADDR;
+	ctrl->hash_array_phys_baseaddr = pfe->ddr_phys_baseaddr +
+						ROUTE_TABLE_BASEADDR;
+
+	ctrl->dev = pfe->dev;
+
+	pr_info("%s finished\n", __func__);
+
+	return 0;
+}
+
+void pfe_ctrl_exit(struct pfe *pfe)
+{
+	pr_info("%s\n", __func__);
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_debugfs.c b/drivers/staging/fsl_ppfe/pfe_debugfs.c
new file mode 100644
index 0000000..4156610
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_debugfs.c
@@ -0,0 +1,111 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include <linux/platform_device.h>
+
+#include "pfe_mod.h"
+
+static int dmem_show(struct seq_file *s, void *unused)
+{
+	u32 dmem_addr, val;
+	int id = (long int)s->private;
+	int i;
+
+	for (dmem_addr = 0; dmem_addr < CLASS_DMEM_SIZE; dmem_addr += 8 * 4) {
+		seq_printf(s, "%04x:", dmem_addr);
+
+		for (i = 0; i < 8; i++) {
+			val = pe_dmem_read(id, dmem_addr + i * 4, 4);
+			seq_printf(s, " %02x %02x %02x %02x", val & 0xff,
+				   (val >> 8) & 0xff, (val >> 16) & 0xff,
+				   (val >> 24) & 0xff);
+		}
+
+		seq_puts(s, "\n");
+	}
+
+	return 0;
+}
+
+static int dmem_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, dmem_show, inode->i_private);
+}
+
+static const struct file_operations dmem_fops = {
+	.open		= dmem_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+int pfe_debugfs_init(struct pfe *pfe)
+{
+	struct dentry *d;
+
+	pr_info("%s\n", __func__);
+
+	pfe->dentry = debugfs_create_dir("pfe", NULL);
+	if (IS_ERR_OR_NULL(pfe->dentry))
+		goto err_dir;
+
+	d = debugfs_create_file("pe0_dmem", 0444, pfe->dentry, (void *)0,
+				&dmem_fops);
+	if (IS_ERR_OR_NULL(d))
+		goto err_pe;
+
+	d = debugfs_create_file("pe1_dmem", 0444, pfe->dentry, (void *)1,
+				&dmem_fops);
+	if (IS_ERR_OR_NULL(d))
+		goto err_pe;
+
+	d = debugfs_create_file("pe2_dmem", 0444, pfe->dentry, (void *)2,
+				&dmem_fops);
+	if (IS_ERR_OR_NULL(d))
+		goto err_pe;
+
+	d = debugfs_create_file("pe3_dmem", 0444, pfe->dentry, (void *)3,
+				&dmem_fops);
+	if (IS_ERR_OR_NULL(d))
+		goto err_pe;
+
+	d = debugfs_create_file("pe4_dmem", 0444, pfe->dentry, (void *)4,
+				&dmem_fops);
+	if (IS_ERR_OR_NULL(d))
+		goto err_pe;
+
+	d = debugfs_create_file("pe5_dmem", 0444, pfe->dentry, (void *)5,
+				&dmem_fops);
+	if (IS_ERR_OR_NULL(d))
+		goto err_pe;
+
+	return 0;
+
+err_pe:
+	debugfs_remove_recursive(pfe->dentry);
+
+err_dir:
+	return -1;
+}
+
+void pfe_debugfs_exit(struct pfe *pfe)
+{
+	debugfs_remove_recursive(pfe->dentry);
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_eth.c b/drivers/staging/fsl_ppfe/pfe_eth.c
new file mode 100644
index 0000000..5aa718b
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_eth.c
@@ -0,0 +1,2434 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/* @pfe_eth.c.
+ *  Ethernet driver for to handle exception path for PFE.
+ *  - uses HIF functions to send/receive packets.
+ *  - uses ctrl function to start/stop interfaces.
+ *  - uses direct register accesses to control phy operation.
+ */
+#include <linux/version.h>
+#include <linux/kernel.h>
+#include <linux/interrupt.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmapool.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+#include <linux/phy.h>
+#include <linux/timer.h>
+#include <linux/hrtimer.h>
+#include <linux/platform_device.h>
+
+#include <net/ip.h>
+#include <net/sock.h>
+
+#include <linux/io.h>
+#include <asm/irq.h>
+#include <linux/delay.h>
+#include <linux/regmap.h>
+#include <linux/i2c.h>
+
+#if defined(CONFIG_NF_CONNTRACK_MARK)
+#include <net/netfilter/nf_conntrack.h>
+#endif
+
+#include "pfe_mod.h"
+#include "pfe_eth.h"
+
+static void *cbus_emac_base[3];
+static void *cbus_gpi_base[3];
+
+/* Forward Declaration */
+static void pfe_eth_exit_one(struct pfe_eth_priv_s *priv);
+static void pfe_eth_flush_tx(struct pfe_eth_priv_s *priv);
+static void pfe_eth_flush_txQ(struct pfe_eth_priv_s *priv, int tx_q_num, int
+				from_tx, int n_desc);
+
+unsigned int gemac_regs[] = {
+	0x0004, /* Interrupt event */
+	0x0008, /* Interrupt mask */
+	0x0024, /* Ethernet control */
+	0x0064, /* MIB Control/Status */
+	0x0084, /* Receive control/status */
+	0x00C4, /* Transmit control */
+	0x00E4, /* Physical address low */
+	0x00E8, /* Physical address high */
+	0x0144, /* Transmit FIFO Watermark and Store and Forward Control*/
+	0x0190, /* Receive FIFO Section Full Threshold */
+	0x01A0, /* Transmit FIFO Section Empty Threshold */
+	0x01B0, /* Frame Truncation Length */
+};
+
+/********************************************************************/
+/*                   SYSFS INTERFACE				    */
+/********************************************************************/
+
+#ifdef PFE_ETH_NAPI_STATS
+/*
+ * pfe_eth_show_napi_stats
+ */
+static ssize_t pfe_eth_show_napi_stats(struct device *dev,
+				       struct device_attribute *attr,
+				       char *buf)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(to_net_dev(dev));
+	ssize_t len = 0;
+
+	len += sprintf(buf + len, "sched:  %u\n",
+			priv->napi_counters[NAPI_SCHED_COUNT]);
+	len += sprintf(buf + len, "poll:   %u\n",
+			priv->napi_counters[NAPI_POLL_COUNT]);
+	len += sprintf(buf + len, "packet: %u\n",
+			priv->napi_counters[NAPI_PACKET_COUNT]);
+	len += sprintf(buf + len, "budget: %u\n",
+			priv->napi_counters[NAPI_FULL_BUDGET_COUNT]);
+	len += sprintf(buf + len, "desc:   %u\n",
+			priv->napi_counters[NAPI_DESC_COUNT]);
+
+	return len;
+}
+
+/*
+ * pfe_eth_set_napi_stats
+ */
+static ssize_t pfe_eth_set_napi_stats(struct device *dev,
+				      struct device_attribute *attr,
+				      const char *buf, size_t count)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(to_net_dev(dev));
+
+	memset(priv->napi_counters, 0, sizeof(priv->napi_counters));
+
+	return count;
+}
+#endif
+#ifdef PFE_ETH_TX_STATS
+/* pfe_eth_show_tx_stats
+ *
+ */
+static ssize_t pfe_eth_show_tx_stats(struct device *dev,
+				     struct device_attribute *attr,
+				     char *buf)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(to_net_dev(dev));
+	ssize_t len = 0;
+	int i;
+
+	len += sprintf(buf + len, "TX queues stats:\n");
+
+	for (i = 0; i < emac_txq_cnt; i++) {
+		struct netdev_queue *tx_queue = netdev_get_tx_queue(priv->ndev,
+									i);
+
+		len += sprintf(buf + len, "\n");
+		__netif_tx_lock_bh(tx_queue);
+
+		hif_tx_lock(&pfe->hif);
+		len += sprintf(buf + len,
+				"Queue %2d :  credits               = %10d\n"
+				, i, hif_lib_tx_credit_avail(pfe, priv->id, i));
+		len += sprintf(buf + len,
+				 "            tx packets            = %10d\n"
+				,  pfe->tmu_credit.tx_packets[priv->id][i]);
+		hif_tx_unlock(&pfe->hif);
+
+		/* Don't output additionnal stats if queue never used */
+		if (!pfe->tmu_credit.tx_packets[priv->id][i])
+			goto skip;
+
+		len += sprintf(buf + len,
+				 "            clean_fail            = %10d\n"
+				, priv->clean_fail[i]);
+		len += sprintf(buf + len,
+				 "            stop_queue            = %10d\n"
+				, priv->stop_queue_total[i]);
+		len += sprintf(buf + len,
+				 "            stop_queue_hif        = %10d\n"
+				, priv->stop_queue_hif[i]);
+		len += sprintf(buf + len,
+				"            stop_queue_hif_client = %10d\n"
+				, priv->stop_queue_hif_client[i]);
+		len += sprintf(buf + len,
+				 "            stop_queue_credit     = %10d\n"
+				, priv->stop_queue_credit[i]);
+skip:
+		__netif_tx_unlock_bh(tx_queue);
+	}
+	return len;
+}
+
+/* pfe_eth_set_tx_stats
+ *
+ */
+static ssize_t pfe_eth_set_tx_stats(struct device *dev,
+				    struct device_attribute *attr,
+				    const char *buf, size_t count)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(to_net_dev(dev));
+	int i;
+
+	for (i = 0; i < emac_txq_cnt; i++) {
+		struct netdev_queue *tx_queue = netdev_get_tx_queue(priv->ndev,
+									i);
+
+		__netif_tx_lock_bh(tx_queue);
+		priv->clean_fail[i] = 0;
+		priv->stop_queue_total[i] = 0;
+		priv->stop_queue_hif[i] = 0;
+		priv->stop_queue_hif_client[i] = 0;
+		priv->stop_queue_credit[i] = 0;
+		__netif_tx_unlock_bh(tx_queue);
+	}
+
+	return count;
+}
+#endif
+/* pfe_eth_show_txavail
+ *
+ */
+static ssize_t pfe_eth_show_txavail(struct device *dev,
+				    struct device_attribute *attr,
+				    char *buf)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(to_net_dev(dev));
+	ssize_t len = 0;
+	int i;
+
+	for (i = 0; i < emac_txq_cnt; i++) {
+		struct netdev_queue *tx_queue = netdev_get_tx_queue(priv->ndev,
+									i);
+
+		__netif_tx_lock_bh(tx_queue);
+
+		len += sprintf(buf + len, "%d",
+				hif_lib_tx_avail(&priv->client, i));
+
+		__netif_tx_unlock_bh(tx_queue);
+
+		if (i == (emac_txq_cnt - 1))
+			len += sprintf(buf + len, "\n");
+		else
+			len += sprintf(buf + len, " ");
+	}
+
+	return len;
+}
+
+/* pfe_eth_show_default_priority
+ *
+ */
+static ssize_t pfe_eth_show_default_priority(struct device *dev,
+					     struct device_attribute *attr,
+						char *buf)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(to_net_dev(dev));
+	unsigned long flags;
+	int rc;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	rc = sprintf(buf, "%d\n", priv->default_priority);
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	return rc;
+}
+
+/* pfe_eth_set_default_priority
+ *
+ */
+
+static ssize_t pfe_eth_set_default_priority(struct device *dev,
+					    struct device_attribute *attr,
+					    const char *buf, size_t count)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(to_net_dev(dev));
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->lock, flags);
+	priv->default_priority = kstrtoul(buf, 0, 0);
+	spin_unlock_irqrestore(&priv->lock, flags);
+
+	return count;
+}
+
+static DEVICE_ATTR(txavail, 0444, pfe_eth_show_txavail, NULL);
+static DEVICE_ATTR(default_priority, 0644, pfe_eth_show_default_priority,
+			pfe_eth_set_default_priority);
+
+#ifdef PFE_ETH_NAPI_STATS
+static DEVICE_ATTR(napi_stats, 0644, pfe_eth_show_napi_stats,
+			pfe_eth_set_napi_stats);
+#endif
+
+#ifdef PFE_ETH_TX_STATS
+static DEVICE_ATTR(tx_stats, 0644, pfe_eth_show_tx_stats,
+			pfe_eth_set_tx_stats);
+#endif
+
+/*
+ * pfe_eth_sysfs_init
+ *
+ */
+static int pfe_eth_sysfs_init(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	int err;
+
+	/* Initialize the default values */
+
+	/*
+	 * By default, packets without conntrack will use this default high
+	 * priority queue
+	 */
+	priv->default_priority = 15;
+
+	/* Create our sysfs files */
+	err = device_create_file(&ndev->dev, &dev_attr_default_priority);
+	if (err) {
+		netdev_err(ndev,
+			   "failed to create default_priority sysfs files\n");
+		goto err_priority;
+	}
+
+	err = device_create_file(&ndev->dev, &dev_attr_txavail);
+	if (err) {
+		netdev_err(ndev,
+			   "failed to create default_priority sysfs files\n");
+		goto err_txavail;
+	}
+
+#ifdef PFE_ETH_NAPI_STATS
+	err = device_create_file(&ndev->dev, &dev_attr_napi_stats);
+	if (err) {
+		netdev_err(ndev, "failed to create napi stats sysfs files\n");
+		goto err_napi;
+	}
+#endif
+
+#ifdef PFE_ETH_TX_STATS
+	err = device_create_file(&ndev->dev, &dev_attr_tx_stats);
+	if (err) {
+		netdev_err(ndev, "failed to create tx stats sysfs files\n");
+		goto err_tx;
+	}
+#endif
+
+	return 0;
+
+#ifdef PFE_ETH_TX_STATS
+err_tx:
+#endif
+#ifdef PFE_ETH_NAPI_STATS
+	device_remove_file(&ndev->dev, &dev_attr_napi_stats);
+
+err_napi:
+#endif
+	device_remove_file(&ndev->dev, &dev_attr_txavail);
+
+err_txavail:
+	device_remove_file(&ndev->dev, &dev_attr_default_priority);
+
+err_priority:
+	return -1;
+}
+
+/* pfe_eth_sysfs_exit
+ *
+ */
+void pfe_eth_sysfs_exit(struct net_device *ndev)
+{
+#ifdef PFE_ETH_TX_STATS
+	device_remove_file(&ndev->dev, &dev_attr_tx_stats);
+#endif
+
+#ifdef PFE_ETH_NAPI_STATS
+	device_remove_file(&ndev->dev, &dev_attr_napi_stats);
+#endif
+	device_remove_file(&ndev->dev, &dev_attr_txavail);
+	device_remove_file(&ndev->dev, &dev_attr_default_priority);
+}
+
+/*************************************************************************/
+/*		ETHTOOL INTERCAE					 */
+/*************************************************************************/
+
+/*MTIP GEMAC */
+static const struct fec_stat {
+	char name[ETH_GSTRING_LEN];
+	u16 offset;
+} fec_stats[] = {
+	/* RMON TX */
+	{ "tx_dropped", RMON_T_DROP },
+	{ "tx_packets", RMON_T_PACKETS },
+	{ "tx_broadcast", RMON_T_BC_PKT },
+	{ "tx_multicast", RMON_T_MC_PKT },
+	{ "tx_crc_errors", RMON_T_CRC_ALIGN },
+	{ "tx_undersize", RMON_T_UNDERSIZE },
+	{ "tx_oversize", RMON_T_OVERSIZE },
+	{ "tx_fragment", RMON_T_FRAG },
+	{ "tx_jabber", RMON_T_JAB },
+	{ "tx_collision", RMON_T_COL },
+	{ "tx_64byte", RMON_T_P64 },
+	{ "tx_65to127byte", RMON_T_P65TO127 },
+	{ "tx_128to255byte", RMON_T_P128TO255 },
+	{ "tx_256to511byte", RMON_T_P256TO511 },
+	{ "tx_512to1023byte", RMON_T_P512TO1023 },
+	{ "tx_1024to2047byte", RMON_T_P1024TO2047 },
+	{ "tx_GTE2048byte", RMON_T_P_GTE2048 },
+	{ "tx_octets", RMON_T_OCTETS },
+
+	/* IEEE TX */
+	{ "IEEE_tx_drop", IEEE_T_DROP },
+	{ "IEEE_tx_frame_ok", IEEE_T_FRAME_OK },
+	{ "IEEE_tx_1col", IEEE_T_1COL },
+	{ "IEEE_tx_mcol", IEEE_T_MCOL },
+	{ "IEEE_tx_def", IEEE_T_DEF },
+	{ "IEEE_tx_lcol", IEEE_T_LCOL },
+	{ "IEEE_tx_excol", IEEE_T_EXCOL },
+	{ "IEEE_tx_macerr", IEEE_T_MACERR },
+	{ "IEEE_tx_cserr", IEEE_T_CSERR },
+	{ "IEEE_tx_sqe", IEEE_T_SQE },
+	{ "IEEE_tx_fdxfc", IEEE_T_FDXFC },
+	{ "IEEE_tx_octets_ok", IEEE_T_OCTETS_OK },
+
+	/* RMON RX */
+	{ "rx_packets", RMON_R_PACKETS },
+	{ "rx_broadcast", RMON_R_BC_PKT },
+	{ "rx_multicast", RMON_R_MC_PKT },
+	{ "rx_crc_errors", RMON_R_CRC_ALIGN },
+	{ "rx_undersize", RMON_R_UNDERSIZE },
+	{ "rx_oversize", RMON_R_OVERSIZE },
+	{ "rx_fragment", RMON_R_FRAG },
+	{ "rx_jabber", RMON_R_JAB },
+	{ "rx_64byte", RMON_R_P64 },
+	{ "rx_65to127byte", RMON_R_P65TO127 },
+	{ "rx_128to255byte", RMON_R_P128TO255 },
+	{ "rx_256to511byte", RMON_R_P256TO511 },
+	{ "rx_512to1023byte", RMON_R_P512TO1023 },
+	{ "rx_1024to2047byte", RMON_R_P1024TO2047 },
+	{ "rx_GTE2048byte", RMON_R_P_GTE2048 },
+	{ "rx_octets", RMON_R_OCTETS },
+
+	/* IEEE RX */
+	{ "IEEE_rx_drop", IEEE_R_DROP },
+	{ "IEEE_rx_frame_ok", IEEE_R_FRAME_OK },
+	{ "IEEE_rx_crc", IEEE_R_CRC },
+	{ "IEEE_rx_align", IEEE_R_ALIGN },
+	{ "IEEE_rx_macerr", IEEE_R_MACERR },
+	{ "IEEE_rx_fdxfc", IEEE_R_FDXFC },
+	{ "IEEE_rx_octets_ok", IEEE_R_OCTETS_OK },
+};
+
+static void pfe_eth_fill_stats(struct net_device *ndev, struct ethtool_stats
+				*stats, u64 *data)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(fec_stats); i++)
+		data[i] = readl(priv->EMAC_baseaddr + fec_stats[i].offset);
+}
+
+static void pfe_eth_gstrings(struct net_device *netdev,
+			     u32 stringset, u8 *data)
+{
+	int i;
+
+	switch (stringset) {
+	case ETH_SS_STATS:
+		for (i = 0; i < ARRAY_SIZE(fec_stats); i++)
+			memcpy(data + i * ETH_GSTRING_LEN,
+			       fec_stats[i].name, ETH_GSTRING_LEN);
+		break;
+	}
+}
+
+static int pfe_eth_stats_count(struct net_device *ndev, int sset)
+{
+	switch (sset) {
+	case ETH_SS_STATS:
+		return ARRAY_SIZE(fec_stats);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+/*
+ * pfe_eth_gemac_reglen - Return the length of the register structure.
+ *
+ */
+static int pfe_eth_gemac_reglen(struct net_device *ndev)
+{
+	pr_info("%s()\n", __func__);
+	return (sizeof(gemac_regs) / sizeof(u32));
+}
+
+/*
+ * pfe_eth_gemac_get_regs - Return the gemac register structure.
+ *
+ */
+static void  pfe_eth_gemac_get_regs(struct net_device *ndev, struct ethtool_regs
+					*regs, void *regbuf)
+{
+	int i;
+
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	u32 *buf = (u32 *)regbuf;
+
+	pr_info("%s()\n", __func__);
+	for (i = 0; i < sizeof(gemac_regs) / sizeof(u32); i++)
+		buf[i] = readl(priv->EMAC_baseaddr + gemac_regs[i]);
+}
+
+/*
+ * pfe_eth_set_wol - Set the magic packet option, in WoL register.
+ *
+ */
+static int pfe_eth_set_wol(struct net_device *ndev, struct ethtool_wolinfo *wol)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	if (wol->wolopts & ~WAKE_MAGIC)
+		return -EOPNOTSUPP;
+
+	/* for MTIP we store wol->wolopts */
+	priv->wol = wol->wolopts;
+
+	device_set_wakeup_enable(&ndev->dev, wol->wolopts & WAKE_MAGIC);
+
+	return 0;
+}
+
+/*
+ *
+ * pfe_eth_get_wol - Get the WoL options.
+ *
+ */
+static void pfe_eth_get_wol(struct net_device *ndev, struct ethtool_wolinfo
+				*wol)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	wol->supported = WAKE_MAGIC;
+	wol->wolopts = 0;
+
+	if (priv->wol & WAKE_MAGIC)
+		wol->wolopts = WAKE_MAGIC;
+
+	memset(&wol->sopass, 0, sizeof(wol->sopass));
+}
+
+/*
+ * pfe_eth_get_drvinfo -  Fills in the drvinfo structure with some basic info
+ *
+ */
+static void pfe_eth_get_drvinfo(struct net_device *ndev, struct ethtool_drvinfo
+				*drvinfo)
+{
+	strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
+	strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
+	strlcpy(drvinfo->fw_version, "N/A", sizeof(drvinfo->fw_version));
+	strlcpy(drvinfo->bus_info, "N/A", sizeof(drvinfo->bus_info));
+}
+
+/*
+ * pfe_eth_set_settings - Used to send commands to PHY.
+ *
+ */
+static int pfe_eth_set_settings(struct net_device *ndev,
+				const struct ethtool_link_ksettings *cmd)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	struct phy_device *phydev = priv->phydev;
+
+	if (!phydev)
+		return -ENODEV;
+
+	return phy_ethtool_ksettings_set(phydev, cmd);
+}
+
+/*
+ * pfe_eth_getsettings - Return the current settings in the ethtool_cmd
+ * structure.
+ *
+ */
+static int pfe_eth_get_settings(struct net_device *ndev,
+				struct ethtool_link_ksettings *cmd)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	struct phy_device *phydev = priv->phydev;
+
+	if (!phydev)
+		return -ENODEV;
+
+	return phy_ethtool_ksettings_get(phydev, cmd);
+}
+
+/*
+ * pfe_eth_get_msglevel - Gets the debug message mask.
+ *
+ */
+static uint32_t pfe_eth_get_msglevel(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	return priv->msg_enable;
+}
+
+/*
+ * pfe_eth_set_msglevel - Sets the debug message mask.
+ *
+ */
+static void pfe_eth_set_msglevel(struct net_device *ndev, uint32_t data)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	priv->msg_enable = data;
+}
+
+#define HIF_RX_COAL_MAX_CLKS		(~(1 << 31))
+#define HIF_RX_COAL_CLKS_PER_USEC	(pfe->ctrl.sys_clk / 1000)
+#define HIF_RX_COAL_MAX_USECS		(HIF_RX_COAL_MAX_CLKS	/ \
+						HIF_RX_COAL_CLKS_PER_USEC)
+
+/*
+ * pfe_eth_set_coalesce - Sets rx interrupt coalescing timer.
+ *
+ */
+static int pfe_eth_set_coalesce(struct net_device *ndev,
+				struct ethtool_coalesce *ec)
+{
+	if (ec->rx_coalesce_usecs > HIF_RX_COAL_MAX_USECS)
+		return -EINVAL;
+
+	if (!ec->rx_coalesce_usecs) {
+		writel(0, HIF_INT_COAL);
+		return 0;
+	}
+
+	writel((ec->rx_coalesce_usecs * HIF_RX_COAL_CLKS_PER_USEC) |
+			HIF_INT_COAL_ENABLE, HIF_INT_COAL);
+
+	return 0;
+}
+
+/*
+ * pfe_eth_get_coalesce - Gets rx interrupt coalescing timer value.
+ *
+ */
+static int pfe_eth_get_coalesce(struct net_device *ndev,
+				struct ethtool_coalesce *ec)
+{
+	int reg_val = readl(HIF_INT_COAL);
+
+	if (reg_val & HIF_INT_COAL_ENABLE)
+		ec->rx_coalesce_usecs = (reg_val & HIF_RX_COAL_MAX_CLKS) /
+						HIF_RX_COAL_CLKS_PER_USEC;
+	else
+		ec->rx_coalesce_usecs = 0;
+
+	return 0;
+}
+
+/*
+ * pfe_eth_set_pauseparam - Sets pause parameters
+ *
+ */
+static int pfe_eth_set_pauseparam(struct net_device *ndev,
+				  struct ethtool_pauseparam *epause)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	if (epause->tx_pause != epause->rx_pause) {
+		netdev_info(ndev,
+			    "hardware only support enable/disable both tx and rx\n");
+		return -EINVAL;
+	}
+
+	priv->pause_flag = 0;
+	priv->pause_flag |= epause->rx_pause ? PFE_PAUSE_FLAG_ENABLE : 0;
+	priv->pause_flag |= epause->autoneg ? PFE_PAUSE_FLAG_AUTONEG : 0;
+
+	if (epause->rx_pause || epause->autoneg) {
+		gemac_enable_pause_rx(priv->EMAC_baseaddr);
+		writel((readl(priv->GPI_baseaddr + GPI_TX_PAUSE_TIME) |
+					EGPI_PAUSE_ENABLE),
+				priv->GPI_baseaddr + GPI_TX_PAUSE_TIME);
+		if (priv->phydev) {
+			priv->phydev->supported |= ADVERTISED_Pause |
+							ADVERTISED_Asym_Pause;
+			priv->phydev->advertising |= ADVERTISED_Pause |
+							ADVERTISED_Asym_Pause;
+		}
+	} else {
+		gemac_disable_pause_rx(priv->EMAC_baseaddr);
+		writel((readl(priv->GPI_baseaddr + GPI_TX_PAUSE_TIME) &
+					~EGPI_PAUSE_ENABLE),
+				priv->GPI_baseaddr + GPI_TX_PAUSE_TIME);
+		if (priv->phydev) {
+			priv->phydev->supported &= ~(ADVERTISED_Pause |
+							ADVERTISED_Asym_Pause);
+			priv->phydev->advertising &= ~(ADVERTISED_Pause |
+							ADVERTISED_Asym_Pause);
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * pfe_eth_get_pauseparam - Gets pause parameters
+ *
+ */
+static void pfe_eth_get_pauseparam(struct net_device *ndev,
+				   struct ethtool_pauseparam *epause)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	epause->autoneg = (priv->pause_flag & PFE_PAUSE_FLAG_AUTONEG) != 0;
+	epause->tx_pause = (priv->pause_flag & PFE_PAUSE_FLAG_ENABLE) != 0;
+	epause->rx_pause = epause->tx_pause;
+}
+
+/*
+ * pfe_eth_get_hash
+ */
+#define PFE_HASH_BITS	6		/* #bits in hash */
+#define CRC32_POLY	0xEDB88320
+
+static int pfe_eth_get_hash(u8 *addr)
+{
+	unsigned int i, bit, data, crc, hash;
+
+	/* calculate crc32 value of mac address */
+	crc = 0xffffffff;
+
+	for (i = 0; i < 6; i++) {
+		data = addr[i];
+		for (bit = 0; bit < 8; bit++, data >>= 1) {
+			crc = (crc >> 1) ^
+				(((crc ^ data) & 1) ? CRC32_POLY : 0);
+		}
+	}
+
+	/*
+	 * only upper 6 bits (PFE_HASH_BITS) are used
+	 * which point to specific bit in the hash registers
+	 */
+	hash = (crc >> (32 - PFE_HASH_BITS)) & 0x3f;
+
+	return hash;
+}
+
+const struct ethtool_ops pfe_ethtool_ops = {
+	.get_drvinfo = pfe_eth_get_drvinfo,
+	.get_regs_len = pfe_eth_gemac_reglen,
+	.get_regs = pfe_eth_gemac_get_regs,
+	.get_link = ethtool_op_get_link,
+	.get_wol  = pfe_eth_get_wol,
+	.set_wol  = pfe_eth_set_wol,
+	.set_pauseparam = pfe_eth_set_pauseparam,
+	.get_pauseparam = pfe_eth_get_pauseparam,
+	.get_strings = pfe_eth_gstrings,
+	.get_sset_count = pfe_eth_stats_count,
+	.get_ethtool_stats = pfe_eth_fill_stats,
+	.get_msglevel = pfe_eth_get_msglevel,
+	.set_msglevel = pfe_eth_set_msglevel,
+	.set_coalesce = pfe_eth_set_coalesce,
+	.get_coalesce = pfe_eth_get_coalesce,
+	.get_link_ksettings = pfe_eth_get_settings,
+	.set_link_ksettings = pfe_eth_set_settings,
+};
+
+/* pfe_eth_mdio_reset
+ */
+int pfe_eth_mdio_reset(struct mii_bus *bus)
+{
+	struct pfe_eth_priv_s *priv = (struct pfe_eth_priv_s *)bus->priv;
+	u32 phy_speed;
+
+	netif_info(priv, hw, priv->ndev, "%s\n", __func__);
+
+	mutex_lock(&bus->mdio_lock);
+
+	/*
+	 * Set MII speed to 2.5 MHz (= clk_get_rate() / 2 * phy_speed)
+	 *
+	 * The formula for FEC MDC is 'ref_freq / (MII_SPEED x 2)' while
+	 * for ENET-MAC is 'ref_freq / ((MII_SPEED + 1) x 2)'.
+	 */
+	phy_speed = (DIV_ROUND_UP((pfe->ctrl.sys_clk * 1000), 4000000)
+		     << EMAC_MII_SPEED_SHIFT);
+	phy_speed |= EMAC_HOLDTIME(0x5);
+	__raw_writel(phy_speed, priv->PHY_baseaddr + EMAC_MII_CTRL_REG);
+
+	mutex_unlock(&bus->mdio_lock);
+
+	return 0;
+}
+
+/* pfe_eth_gemac_phy_timeout
+ *
+ */
+static int pfe_eth_gemac_phy_timeout(struct pfe_eth_priv_s *priv, int timeout)
+{
+	while (!(__raw_readl(priv->PHY_baseaddr + EMAC_IEVENT_REG) &
+			EMAC_IEVENT_MII)) {
+		if (timeout-- <= 0)
+			return -1;
+		usleep_range(10, 20);
+	}
+	__raw_writel(EMAC_IEVENT_MII, priv->PHY_baseaddr + EMAC_IEVENT_REG);
+	return 0;
+}
+
+static int pfe_eth_mdio_mux(u8 muxval)
+{
+	struct i2c_adapter *a;
+	struct i2c_msg msg;
+	unsigned char buf[2];
+	int ret;
+
+	a = i2c_get_adapter(0);
+	if (!a)
+		return -ENODEV;
+
+	/* set bit 1 (the second bit) of chip at 0x09, register 0x13 */
+	buf[0] = 0x54; /* reg number */
+	buf[1] = (muxval << 6) | 0x3; /* data */
+	msg.addr = 0x66;
+	msg.buf = buf;
+	msg.len = 2;
+	msg.flags = 0;
+	ret = i2c_transfer(a, &msg, 1);
+	i2c_put_adapter(a);
+	if (ret != 1)
+		return -ENODEV;
+	return 0;
+}
+
+static int pfe_eth_mdio_write_addr(struct mii_bus *bus, int mii_id,
+				   int dev_addr, int regnum)
+{
+	struct pfe_eth_priv_s *priv = (struct pfe_eth_priv_s *)bus->priv;
+
+	__raw_writel(EMAC_MII_DATA_PA(mii_id) |
+		     EMAC_MII_DATA_RA(dev_addr) |
+		     EMAC_MII_DATA_TA | EMAC_MII_DATA(regnum),
+		     priv->PHY_baseaddr + EMAC_MII_DATA_REG);
+
+	if (pfe_eth_gemac_phy_timeout(priv, EMAC_MDIO_TIMEOUT)) {
+		netdev_err(priv->ndev, "%s: phy MDIO address write timeout\n",
+			   __func__);
+		return -1;
+	}
+
+	return 0;
+}
+
+static int pfe_eth_mdio_write(struct mii_bus *bus, int mii_id, int regnum,
+			      u16 value)
+{
+	struct pfe_eth_priv_s *priv = (struct pfe_eth_priv_s *)bus->priv;
+
+	/*To access external PHYs on QDS board mux needs to be configured*/
+	if ((mii_id) && (pfe->mdio_muxval[mii_id]))
+		pfe_eth_mdio_mux(pfe->mdio_muxval[mii_id]);
+
+	if (regnum & MII_ADDR_C45) {
+		pfe_eth_mdio_write_addr(bus, mii_id, (regnum >> 16) & 0x1f,
+					regnum & 0xffff);
+		__raw_writel(EMAC_MII_DATA_OP_CL45_WR |
+			     EMAC_MII_DATA_PA(mii_id) |
+			     EMAC_MII_DATA_RA((regnum >> 16) & 0x1f) |
+			     EMAC_MII_DATA_TA | EMAC_MII_DATA(value),
+			     priv->PHY_baseaddr + EMAC_MII_DATA_REG);
+	} else {
+		/* start a write op */
+		__raw_writel(EMAC_MII_DATA_ST | EMAC_MII_DATA_OP_WR |
+			     EMAC_MII_DATA_PA(mii_id) |
+			     EMAC_MII_DATA_RA(regnum) |
+			     EMAC_MII_DATA_TA | EMAC_MII_DATA(value),
+			     priv->PHY_baseaddr + EMAC_MII_DATA_REG);
+	}
+
+	if (pfe_eth_gemac_phy_timeout(priv, EMAC_MDIO_TIMEOUT)) {
+		netdev_err(priv->ndev, "%s: phy MDIO write timeout\n",
+			   __func__);
+		return -1;
+	}
+	netif_info(priv, hw, priv->ndev, "%s: phy %x reg %x val %x\n", __func__,
+		   mii_id, regnum, value);
+
+	return 0;
+}
+
+static int pfe_eth_mdio_read(struct mii_bus *bus, int mii_id, int regnum)
+{
+	struct pfe_eth_priv_s *priv = (struct pfe_eth_priv_s *)bus->priv;
+	u16 value = 0;
+
+	/*To access external PHYs on QDS board mux needs to be configured*/
+	if ((mii_id) && (pfe->mdio_muxval[mii_id]))
+		pfe_eth_mdio_mux(pfe->mdio_muxval[mii_id]);
+
+	if (regnum & MII_ADDR_C45) {
+		pfe_eth_mdio_write_addr(bus, mii_id, (regnum >> 16) & 0x1f,
+					regnum & 0xffff);
+		__raw_writel(EMAC_MII_DATA_OP_CL45_RD |
+			     EMAC_MII_DATA_PA(mii_id) |
+			     EMAC_MII_DATA_RA((regnum >> 16) & 0x1f) |
+			     EMAC_MII_DATA_TA,
+			     priv->PHY_baseaddr + EMAC_MII_DATA_REG);
+	} else {
+		/* start a read op */
+		__raw_writel(EMAC_MII_DATA_ST | EMAC_MII_DATA_OP_RD |
+			     EMAC_MII_DATA_PA(mii_id) |
+			     EMAC_MII_DATA_RA(regnum) |
+			     EMAC_MII_DATA_TA, priv->PHY_baseaddr +
+			     EMAC_MII_DATA_REG);
+	}
+
+	if (pfe_eth_gemac_phy_timeout(priv, EMAC_MDIO_TIMEOUT)) {
+		netdev_err(priv->ndev, "%s: phy MDIO read timeout\n", __func__);
+		return -1;
+	}
+
+	value = EMAC_MII_DATA(__raw_readl(priv->PHY_baseaddr +
+						EMAC_MII_DATA_REG));
+	netif_info(priv, hw, priv->ndev, "%s: phy %x reg %x val %x\n", __func__,
+		   mii_id, regnum, value);
+	return value;
+}
+
+static int pfe_eth_mdio_init(struct pfe_eth_priv_s *priv,
+			     struct ls1012a_mdio_platform_data *minfo)
+{
+	struct mii_bus *bus;
+	int rc;
+
+	netif_info(priv, drv, priv->ndev, "%s\n", __func__);
+	pr_info("%s\n", __func__);
+
+	bus = mdiobus_alloc();
+	if (!bus) {
+		netdev_err(priv->ndev, "mdiobus_alloc() failed\n");
+		rc = -ENOMEM;
+		goto err0;
+	}
+
+	bus->name = "ls1012a MDIO Bus";
+	bus->read = &pfe_eth_mdio_read;
+	bus->write = &pfe_eth_mdio_write;
+	bus->reset = &pfe_eth_mdio_reset;
+	snprintf(bus->id, MII_BUS_ID_SIZE, "ls1012a-%x", priv->id);
+	bus->priv = priv;
+
+	bus->phy_mask = minfo->phy_mask;
+	priv->mdc_div = minfo->mdc_div;
+
+	if (!priv->mdc_div)
+		priv->mdc_div = 64;
+
+	bus->irq[0] = minfo->irq[0];
+
+	bus->parent = priv->pfe->dev;
+
+	netif_info(priv, drv, priv->ndev, "%s: mdc_div: %d, phy_mask: %x\n",
+		   __func__, priv->mdc_div, bus->phy_mask);
+	rc = mdiobus_register(bus);
+	if (rc) {
+		netdev_err(priv->ndev, "mdiobus_register(%s) failed\n",
+			   bus->name);
+		goto err1;
+	}
+
+	priv->mii_bus = bus;
+	pfe_eth_mdio_reset(bus);
+
+	return 0;
+
+err1:
+	mdiobus_free(bus);
+err0:
+	return rc;
+}
+
+/* pfe_eth_mdio_exit
+ */
+static void pfe_eth_mdio_exit(struct mii_bus *bus)
+{
+	if (!bus)
+		return;
+
+	netif_info((struct pfe_eth_priv_s *)bus->priv, drv, ((struct
+			pfe_eth_priv_s *)(bus->priv))->ndev, "%s\n", __func__);
+
+	mdiobus_unregister(bus);
+	mdiobus_free(bus);
+}
+
+/* pfe_get_phydev_speed
+ */
+static int pfe_get_phydev_speed(struct phy_device *phydev)
+{
+	switch (phydev->speed) {
+	case 10:
+			return SPEED_10M;
+	case 100:
+			return SPEED_100M;
+	case 1000:
+	default:
+			return SPEED_1000M;
+	}
+}
+
+/* pfe_set_rgmii_speed
+ */
+#define RGMIIPCR	0x434
+/* RGMIIPCR bit definitions*/
+#define SCFG_RGMIIPCR_EN_AUTO           (0x00000008)
+#define SCFG_RGMIIPCR_SETSP_1000M       (0x00000004)
+#define SCFG_RGMIIPCR_SETSP_100M        (0x00000000)
+#define SCFG_RGMIIPCR_SETSP_10M         (0x00000002)
+#define SCFG_RGMIIPCR_SETFD             (0x00000001)
+
+static void pfe_set_rgmii_speed(struct phy_device *phydev)
+{
+	u32 rgmii_pcr;
+
+	regmap_read(pfe->scfg, RGMIIPCR, &rgmii_pcr);
+	rgmii_pcr  &= ~(SCFG_RGMIIPCR_SETSP_1000M | SCFG_RGMIIPCR_SETSP_10M);
+
+	switch (phydev->speed) {
+	case 10:
+			rgmii_pcr |= SCFG_RGMIIPCR_SETSP_10M;
+			break;
+	case 1000:
+			rgmii_pcr |= SCFG_RGMIIPCR_SETSP_1000M;
+			break;
+	case 100:
+	default:
+			/* Default is 100M */
+			break;
+	}
+	regmap_write(pfe->scfg, RGMIIPCR, rgmii_pcr);
+}
+
+/* pfe_get_phydev_duplex
+ */
+static int pfe_get_phydev_duplex(struct phy_device *phydev)
+{
+	/*return (phydev->duplex == DUPLEX_HALF) ? DUP_HALF:DUP_FULL ; */
+	return DUPLEX_FULL;
+}
+
+/* pfe_eth_adjust_link
+ */
+static void pfe_eth_adjust_link(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	unsigned long flags;
+	struct phy_device *phydev = priv->phydev;
+	int new_state = 0;
+
+	netif_info(priv, drv, ndev, "%s\n", __func__);
+
+	spin_lock_irqsave(&priv->lock, flags);
+
+	if (phydev->link) {
+		/*
+		 * Now we make sure that we can be in full duplex mode.
+		 * If not, we operate in half-duplex mode.
+		 */
+		if (phydev->duplex != priv->oldduplex) {
+			new_state = 1;
+			gemac_set_duplex(priv->EMAC_baseaddr,
+					 pfe_get_phydev_duplex(phydev));
+			priv->oldduplex = phydev->duplex;
+		}
+
+		if (phydev->speed != priv->oldspeed) {
+			new_state = 1;
+			gemac_set_speed(priv->EMAC_baseaddr,
+					pfe_get_phydev_speed(phydev));
+			if (priv->einfo->mii_config == PHY_INTERFACE_MODE_RGMII)
+				pfe_set_rgmii_speed(phydev);
+			priv->oldspeed = phydev->speed;
+		}
+
+		if (!priv->oldlink) {
+			new_state = 1;
+			priv->oldlink = 1;
+		}
+
+	} else if (priv->oldlink) {
+		new_state = 1;
+		priv->oldlink = 0;
+		priv->oldspeed = 0;
+		priv->oldduplex = -1;
+	}
+
+	if (new_state && netif_msg_link(priv))
+		phy_print_status(phydev);
+
+	spin_unlock_irqrestore(&priv->lock, flags);
+}
+
+/* pfe_phy_exit
+ */
+static void pfe_phy_exit(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	netif_info(priv, drv, ndev, "%s\n", __func__);
+
+	phy_disconnect(priv->phydev);
+	priv->phydev = NULL;
+}
+
+/* pfe_eth_stop
+ */
+static void pfe_eth_stop(struct net_device *ndev, int wake)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	netif_info(priv, drv, ndev, "%s\n", __func__);
+
+	if (wake) {
+		gemac_tx_disable(priv->EMAC_baseaddr);
+	} else {
+		gemac_disable(priv->EMAC_baseaddr);
+		gpi_disable(priv->GPI_baseaddr);
+
+		if (priv->phydev)
+			phy_stop(priv->phydev);
+	}
+}
+
+/* pfe_eth_start
+ */
+static int pfe_eth_start(struct pfe_eth_priv_s *priv)
+{
+	netif_info(priv, drv, priv->ndev, "%s\n", __func__);
+
+	if (priv->phydev)
+		phy_start(priv->phydev);
+
+	gpi_enable(priv->GPI_baseaddr);
+	gemac_enable(priv->EMAC_baseaddr);
+
+	return 0;
+}
+
+/*
+ * Configure on chip serdes through mdio
+ */
+static void ls1012a_configure_serdes(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = pfe->eth.eth_priv[0];
+	int sgmii_2500 = 0;
+	struct mii_bus *bus = priv->mii_bus;
+
+	if (priv->einfo->mii_config == PHY_INTERFACE_MODE_SGMII_2500)
+		sgmii_2500 = 1;
+
+	netif_info(priv, drv, ndev, "%s\n", __func__);
+	/* PCS configuration done with corresponding GEMAC */
+
+	pfe_eth_mdio_read(bus, 0, 0);
+	pfe_eth_mdio_read(bus, 0, 1);
+
+       /*These settings taken from validtion team */
+	pfe_eth_mdio_write(bus, 0, 0x0, 0x8000);
+	if (sgmii_2500) {
+		pfe_eth_mdio_write(bus, 0, 0x14, 0x9);
+		pfe_eth_mdio_write(bus, 0, 0x4, 0x4001);
+		pfe_eth_mdio_write(bus, 0, 0x12, 0xa120);
+		pfe_eth_mdio_write(bus, 0, 0x13, 0x7);
+	} else {
+		pfe_eth_mdio_write(bus, 0, 0x14, 0xb);
+		pfe_eth_mdio_write(bus, 0, 0x4, 0x1a1);
+		pfe_eth_mdio_write(bus, 0, 0x12, 0x400);
+		pfe_eth_mdio_write(bus, 0, 0x13, 0x0);
+	}
+
+	pfe_eth_mdio_write(bus, 0, 0x0, 0x1140);
+}
+
+/*
+ * pfe_phy_init
+ *
+ */
+static int pfe_phy_init(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	struct phy_device *phydev;
+	char phy_id[MII_BUS_ID_SIZE + 3];
+	char bus_id[MII_BUS_ID_SIZE];
+	phy_interface_t interface;
+
+	priv->oldlink = 0;
+	priv->oldspeed = 0;
+	priv->oldduplex = -1;
+
+	snprintf(bus_id, MII_BUS_ID_SIZE, "ls1012a-%d", 0);
+	snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT, bus_id,
+		 priv->einfo->phy_id);
+
+	netif_info(priv, drv, ndev, "%s: %s\n", __func__, phy_id);
+	interface = priv->einfo->mii_config;
+	if ((interface == PHY_INTERFACE_MODE_SGMII) ||
+	    (interface == PHY_INTERFACE_MODE_SGMII_2500)) {
+		/*Configure SGMII PCS */
+		if (pfe->scfg) {
+			/*Config MDIO from serdes */
+			regmap_write(pfe->scfg, 0x484, 0x00000000);
+		}
+		ls1012a_configure_serdes(ndev);
+	}
+
+	if (pfe->scfg) {
+		/*Config MDIO from PAD */
+		regmap_write(pfe->scfg, 0x484, 0x80000000);
+	}
+
+	priv->oldlink = 0;
+	priv->oldspeed = 0;
+	priv->oldduplex = -1;
+	pr_info("%s interface %x\n", __func__, interface);
+	phydev = phy_connect(ndev, phy_id, &pfe_eth_adjust_link, interface);
+
+	if (IS_ERR(phydev)) {
+		netdev_err(ndev, "phy_connect() failed\n");
+		return PTR_ERR(phydev);
+	}
+
+	priv->phydev = phydev;
+	phydev->irq = PHY_POLL;
+
+	return 0;
+}
+
+/* pfe_gemac_init
+ */
+static int pfe_gemac_init(struct pfe_eth_priv_s *priv)
+{
+	struct gemac_cfg cfg;
+
+	netif_info(priv, ifup, priv->ndev, "%s\n", __func__);
+
+	cfg.speed = SPEED_1000M;
+	cfg.duplex = DUPLEX_FULL;
+
+	gemac_set_config(priv->EMAC_baseaddr, &cfg);
+	gemac_allow_broadcast(priv->EMAC_baseaddr);
+	gemac_enable_1536_rx(priv->EMAC_baseaddr);
+	gemac_enable_rx_jmb(priv->EMAC_baseaddr);
+	gemac_enable_stacked_vlan(priv->EMAC_baseaddr);
+	gemac_enable_pause_rx(priv->EMAC_baseaddr);
+	gemac_set_bus_width(priv->EMAC_baseaddr, 64);
+
+	/*GEM will perform checksum verifications*/
+	if (priv->ndev->features & NETIF_F_RXCSUM)
+		gemac_enable_rx_checksum_offload(priv->EMAC_baseaddr);
+	else
+		gemac_disable_rx_checksum_offload(priv->EMAC_baseaddr);
+
+	return 0;
+}
+
+/* pfe_eth_event_handler
+ */
+static int pfe_eth_event_handler(void *data, int event, int qno)
+{
+	struct pfe_eth_priv_s *priv = data;
+
+	switch (event) {
+	case EVENT_RX_PKT_IND:
+
+		if (qno == 0) {
+			if (napi_schedule_prep(&priv->high_napi)) {
+				netif_info(priv, intr, priv->ndev,
+					   "%s: schedule high prio poll\n"
+					   , __func__);
+
+#ifdef PFE_ETH_NAPI_STATS
+				priv->napi_counters[NAPI_SCHED_COUNT]++;
+#endif
+
+				__napi_schedule(&priv->high_napi);
+			}
+		} else if (qno == 1) {
+			if (napi_schedule_prep(&priv->low_napi)) {
+				netif_info(priv, intr, priv->ndev,
+					   "%s: schedule low prio poll\n"
+					   , __func__);
+
+#ifdef PFE_ETH_NAPI_STATS
+				priv->napi_counters[NAPI_SCHED_COUNT]++;
+#endif
+				__napi_schedule(&priv->low_napi);
+			}
+		} else if (qno == 2) {
+			if (napi_schedule_prep(&priv->lro_napi)) {
+				netif_info(priv, intr, priv->ndev,
+					   "%s: schedule lro prio poll\n"
+					   , __func__);
+
+#ifdef PFE_ETH_NAPI_STATS
+				priv->napi_counters[NAPI_SCHED_COUNT]++;
+#endif
+				__napi_schedule(&priv->lro_napi);
+			}
+		}
+
+		break;
+
+	case EVENT_TXDONE_IND:
+		pfe_eth_flush_tx(priv);
+		hif_lib_event_handler_start(&priv->client, EVENT_TXDONE_IND, 0);
+		break;
+	case EVENT_HIGH_RX_WM:
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+/* pfe_eth_open
+ */
+static int pfe_eth_open(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	struct hif_client_s *client;
+	int rc;
+
+	netif_info(priv, ifup, ndev, "%s\n", __func__);
+
+	/* Register client driver with HIF */
+	client = &priv->client;
+	memset(client, 0, sizeof(*client));
+	client->id = PFE_CL_GEM0 + priv->id;
+	client->tx_qn = emac_txq_cnt;
+	client->rx_qn = EMAC_RXQ_CNT;
+	client->priv = priv;
+	client->pfe = priv->pfe;
+	client->event_handler = pfe_eth_event_handler;
+
+	client->tx_qsize = EMAC_TXQ_DEPTH;
+	client->rx_qsize = EMAC_RXQ_DEPTH;
+
+	rc = hif_lib_client_register(client);
+	if (rc) {
+		netdev_err(ndev, "%s: hif_lib_client_register(%d) failed\n",
+			   __func__, client->id);
+		goto err0;
+	}
+
+	netif_info(priv, drv, ndev, "%s: registered client: %p\n", __func__,
+		   client);
+
+	pfe_gemac_init(priv);
+
+	if (!is_valid_ether_addr(ndev->dev_addr)) {
+		netdev_err(ndev, "%s: invalid MAC address\n", __func__);
+		rc = -EADDRNOTAVAIL;
+		goto err1;
+	}
+
+	gemac_set_laddrN(priv->EMAC_baseaddr,
+			 (struct pfe_mac_addr *)ndev->dev_addr, 1);
+
+	napi_enable(&priv->high_napi);
+	napi_enable(&priv->low_napi);
+	napi_enable(&priv->lro_napi);
+
+	rc = pfe_eth_start(priv);
+
+	netif_tx_wake_all_queues(ndev);
+
+	return rc;
+
+err1:
+	hif_lib_client_unregister(&priv->client);
+
+err0:
+	return rc;
+}
+
+/*
+ *  pfe_eth_shutdown
+ */
+int pfe_eth_shutdown(struct net_device *ndev, int wake)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	int i, qstatus;
+	unsigned long next_poll = jiffies + 1, end = jiffies +
+				(TX_POLL_TIMEOUT_MS * HZ) / 1000;
+	int tx_pkts, prv_tx_pkts;
+
+	netif_info(priv, ifdown, ndev, "%s\n", __func__);
+
+	for (i = 0; i < emac_txq_cnt; i++)
+		hrtimer_cancel(&priv->fast_tx_timeout[i].timer);
+
+	netif_tx_stop_all_queues(ndev);
+
+	do {
+		tx_pkts = 0;
+		pfe_eth_flush_tx(priv);
+
+		for (i = 0; i < emac_txq_cnt; i++)
+			tx_pkts += hif_lib_tx_pending(&priv->client, i);
+
+		if (tx_pkts) {
+			/*Don't wait forever, break if we cross max timeout */
+			if (time_after(jiffies, end)) {
+				pr_err(
+					"(%s)Tx is not complete after %dmsec\n",
+					ndev->name, TX_POLL_TIMEOUT_MS);
+				break;
+			}
+
+			pr_info("%s : (%s) Waiting for tx packets to free. Pending tx pkts = %d.\n"
+				, __func__, ndev->name, tx_pkts);
+			if (need_resched())
+				schedule();
+		}
+
+	} while (tx_pkts);
+
+	end = jiffies + (TX_POLL_TIMEOUT_MS * HZ) / 1000;
+
+	prv_tx_pkts = tmu_pkts_processed(priv->id);
+	/*
+	 * Wait till TMU transmits all pending packets
+	 * poll tmu_qstatus and pkts processed by TMU for every 10ms
+	 * Consider TMU is busy, If we see TMU qeueu pending or any packets
+	 * processed by TMU
+	 */
+	while (1) {
+		if (time_after(jiffies, next_poll)) {
+			tx_pkts = tmu_pkts_processed(priv->id);
+			qstatus = tmu_qstatus(priv->id) & 0x7ffff;
+
+			if (!qstatus && (tx_pkts == prv_tx_pkts))
+				break;
+			/* Don't wait forever, break if we cross max
+			 * timeout(TX_POLL_TIMEOUT_MS)
+			 */
+			if (time_after(jiffies, end)) {
+				pr_err("TMU%d is busy after %dmsec\n",
+				       priv->id, TX_POLL_TIMEOUT_MS);
+				break;
+			}
+			prv_tx_pkts = tx_pkts;
+			next_poll++;
+		}
+		if (need_resched())
+			schedule();
+	}
+	/* Wait for some more time to complete transmitting packet if any */
+	next_poll = jiffies + 1;
+	while (1) {
+		if (time_after(jiffies, next_poll))
+			break;
+		if (need_resched())
+			schedule();
+	}
+
+	pfe_eth_stop(ndev, wake);
+
+	napi_disable(&priv->lro_napi);
+	napi_disable(&priv->low_napi);
+	napi_disable(&priv->high_napi);
+
+	hif_lib_client_unregister(&priv->client);
+
+	return 0;
+}
+
+/* pfe_eth_close
+ *
+ */
+static int pfe_eth_close(struct net_device *ndev)
+{
+	pfe_eth_shutdown(ndev, 0);
+
+	return 0;
+}
+
+/* pfe_eth_suspend
+ *
+ * return value : 1 if netdevice is configured to wakeup system
+ *                0 otherwise
+ */
+int pfe_eth_suspend(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	int retval = 0;
+
+	if (priv->wol) {
+		gemac_set_wol(priv->EMAC_baseaddr, priv->wol);
+		retval = 1;
+	}
+	pfe_eth_shutdown(ndev, priv->wol);
+
+	return retval;
+}
+
+/* pfe_eth_resume
+ *
+ */
+int pfe_eth_resume(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	if (priv->wol)
+		gemac_set_wol(priv->EMAC_baseaddr, 0);
+	gemac_tx_enable(priv->EMAC_baseaddr);
+
+	return pfe_eth_open(ndev);
+}
+
+/* pfe_eth_get_queuenum
+ */
+static int pfe_eth_get_queuenum(struct pfe_eth_priv_s *priv, struct sk_buff
+					*skb)
+{
+	int queuenum = 0;
+	unsigned long flags;
+
+	/* Get the Fast Path queue number */
+	/*
+	 * Use conntrack mark (if conntrack exists), then packet mark (if any),
+	 * then fallback to default
+	 */
+#if defined(CONFIG_IP_NF_CONNTRACK_MARK) || defined(CONFIG_NF_CONNTRACK_MARK)
+	if (skb->_nfct) {
+		enum ip_conntrack_info cinfo;
+		struct nf_conn *ct;
+
+		ct = nf_ct_get(skb, &cinfo);
+
+		if (ct) {
+			u32 connmark;
+
+			connmark = ct->mark;
+
+			if ((connmark & 0x80000000) && priv->id != 0)
+				connmark >>= 16;
+
+			queuenum = connmark & EMAC_QUEUENUM_MASK;
+		}
+	} else  {/* continued after #endif ... */
+#endif
+		if (skb->mark) {
+			queuenum = skb->mark & EMAC_QUEUENUM_MASK;
+		} else {
+			spin_lock_irqsave(&priv->lock, flags);
+			queuenum = priv->default_priority & EMAC_QUEUENUM_MASK;
+			spin_unlock_irqrestore(&priv->lock, flags);
+		}
+#if defined(CONFIG_IP_NF_CONNTRACK_MARK) || defined(CONFIG_NF_CONNTRACK_MARK)
+	}
+#endif
+	return queuenum;
+}
+
+/* pfe_eth_might_stop_tx
+ *
+ */
+static int pfe_eth_might_stop_tx(struct pfe_eth_priv_s *priv, int queuenum,
+				 struct netdev_queue *tx_queue,
+				 unsigned int n_desc,
+				 unsigned int n_segs)
+{
+	ktime_t kt;
+
+	if (unlikely((__hif_tx_avail(&pfe->hif) < n_desc) ||
+		     (hif_lib_tx_avail(&priv->client, queuenum) < n_desc) ||
+	(hif_lib_tx_credit_avail(pfe, priv->id, queuenum) < n_segs))) {
+#ifdef PFE_ETH_TX_STATS
+		if (__hif_tx_avail(&pfe->hif) < n_desc) {
+			priv->stop_queue_hif[queuenum]++;
+		} else if (hif_lib_tx_avail(&priv->client, queuenum) < n_desc) {
+			priv->stop_queue_hif_client[queuenum]++;
+		} else if (hif_lib_tx_credit_avail(pfe, priv->id, queuenum) <
+			n_segs) {
+			priv->stop_queue_credit[queuenum]++;
+		}
+		priv->stop_queue_total[queuenum]++;
+#endif
+		netif_tx_stop_queue(tx_queue);
+
+		kt = ktime_set(0, LS1012A_TX_FAST_RECOVERY_TIMEOUT_MS *
+				NSEC_PER_MSEC);
+		hrtimer_start(&priv->fast_tx_timeout[queuenum].timer, kt,
+			      HRTIMER_MODE_REL);
+		return -1;
+	} else {
+		return 0;
+	}
+}
+
+#define SA_MAX_OP 2
+/* pfe_hif_send_packet
+ *
+ * At this level if TX fails we drop the packet
+ */
+static void pfe_hif_send_packet(struct sk_buff *skb, struct  pfe_eth_priv_s
+					*priv, int queuenum)
+{
+	struct skb_shared_info *sh = skb_shinfo(skb);
+	unsigned int nr_frags;
+	u32 ctrl = 0;
+
+	netif_info(priv, tx_queued, priv->ndev, "%s\n", __func__);
+
+	if (skb_is_gso(skb)) {
+		priv->stats.tx_dropped++;
+		return;
+	}
+
+	if (skb->ip_summed == CHECKSUM_PARTIAL)
+		ctrl = HIF_CTRL_TX_CHECKSUM;
+
+	nr_frags = sh->nr_frags;
+
+	if (nr_frags) {
+		skb_frag_t *f;
+		int i;
+
+		__hif_lib_xmit_pkt(&priv->client, queuenum, skb->data,
+				   skb_headlen(skb), ctrl, HIF_FIRST_BUFFER,
+				   skb);
+
+		for (i = 0; i < nr_frags - 1; i++) {
+			f = &sh->frags[i];
+			__hif_lib_xmit_pkt(&priv->client, queuenum,
+					   skb_frag_address(f),
+					   skb_frag_size(f),
+					   0x0, 0x0, skb);
+		}
+
+		f = &sh->frags[i];
+
+		__hif_lib_xmit_pkt(&priv->client, queuenum,
+				   skb_frag_address(f), skb_frag_size(f),
+				   0x0, HIF_LAST_BUFFER | HIF_DATA_VALID,
+				   skb);
+
+		netif_info(priv, tx_queued, priv->ndev,
+			   "%s: pkt sent successfully skb:%p nr_frags:%d len:%d\n",
+			   __func__, skb, nr_frags, skb->len);
+	} else {
+		__hif_lib_xmit_pkt(&priv->client, queuenum, skb->data,
+				   skb->len, ctrl, HIF_FIRST_BUFFER |
+				   HIF_LAST_BUFFER | HIF_DATA_VALID,
+				   skb);
+		netif_info(priv, tx_queued, priv->ndev,
+			   "%s: pkt sent successfully skb:%p len:%d\n",
+			   __func__, skb, skb->len);
+	}
+	hif_tx_dma_start();
+	priv->stats.tx_packets++;
+	priv->stats.tx_bytes += skb->len;
+	hif_lib_tx_credit_use(pfe, priv->id, queuenum, 1);
+}
+
+/* pfe_eth_flush_txQ
+ */
+static void pfe_eth_flush_txQ(struct pfe_eth_priv_s *priv, int tx_q_num, int
+				from_tx, int n_desc)
+{
+	struct sk_buff *skb;
+	struct netdev_queue *tx_queue = netdev_get_tx_queue(priv->ndev,
+								tx_q_num);
+	unsigned int flags;
+
+	netif_info(priv, tx_done, priv->ndev, "%s\n", __func__);
+
+	if (!from_tx)
+		__netif_tx_lock_bh(tx_queue);
+
+	/* Clean HIF and client queue */
+	while ((skb = hif_lib_tx_get_next_complete(&priv->client,
+						   tx_q_num, &flags,
+						   HIF_TX_DESC_NT))) {
+		if (flags & HIF_DATA_VALID)
+			dev_kfree_skb_any(skb);
+	}
+	if (!from_tx)
+		__netif_tx_unlock_bh(tx_queue);
+}
+
+/* pfe_eth_flush_tx
+ */
+static void pfe_eth_flush_tx(struct pfe_eth_priv_s *priv)
+{
+	int ii;
+
+	netif_info(priv, tx_done, priv->ndev, "%s\n", __func__);
+
+	for (ii = 0; ii < emac_txq_cnt; ii++)
+		pfe_eth_flush_txQ(priv, ii, 0, 0);
+}
+
+void pfe_tx_get_req_desc(struct sk_buff *skb, unsigned int *n_desc, unsigned int
+				*n_segs)
+{
+	struct skb_shared_info *sh = skb_shinfo(skb);
+
+	/* Scattered data */
+	if (sh->nr_frags) {
+		*n_desc = sh->nr_frags + 1;
+		*n_segs = 1;
+	/* Regular case */
+	} else {
+		*n_desc = 1;
+		*n_segs = 1;
+	}
+}
+
+/* pfe_eth_send_packet
+ */
+static int pfe_eth_send_packet(struct sk_buff *skb, struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	int tx_q_num = skb_get_queue_mapping(skb);
+	int n_desc, n_segs;
+	struct netdev_queue *tx_queue = netdev_get_tx_queue(priv->ndev,
+								tx_q_num);
+
+	netif_info(priv, tx_queued, ndev, "%s\n", __func__);
+
+	if ((!skb_is_gso(skb)) && (skb_headroom(skb) < (PFE_PKT_HEADER_SZ +
+			sizeof(unsigned long)))) {
+		netif_warn(priv, tx_err, priv->ndev, "%s: copying skb\n",
+			   __func__);
+
+		if (pskb_expand_head(skb, (PFE_PKT_HEADER_SZ + sizeof(unsigned
+					long)), 0, GFP_ATOMIC)) {
+			/* No need to re-transmit, no way to recover*/
+			kfree_skb(skb);
+			priv->stats.tx_dropped++;
+			return NETDEV_TX_OK;
+		}
+	}
+
+	pfe_tx_get_req_desc(skb, &n_desc, &n_segs);
+
+	hif_tx_lock(&pfe->hif);
+	if (unlikely(pfe_eth_might_stop_tx(priv, tx_q_num, tx_queue, n_desc,
+					   n_segs))) {
+#ifdef PFE_ETH_TX_STATS
+		if (priv->was_stopped[tx_q_num]) {
+			priv->clean_fail[tx_q_num]++;
+			priv->was_stopped[tx_q_num] = 0;
+		}
+#endif
+		hif_tx_unlock(&pfe->hif);
+		return NETDEV_TX_BUSY;
+	}
+
+	pfe_hif_send_packet(skb, priv, tx_q_num);
+
+	hif_tx_unlock(&pfe->hif);
+
+	tx_queue->trans_start = jiffies;
+
+#ifdef PFE_ETH_TX_STATS
+	priv->was_stopped[tx_q_num] = 0;
+#endif
+
+	return NETDEV_TX_OK;
+}
+
+/* pfe_eth_select_queue
+ *
+ */
+static u16 pfe_eth_select_queue(struct net_device *ndev, struct sk_buff *skb,
+				void *accel_priv,
+				select_queue_fallback_t fallback)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	return pfe_eth_get_queuenum(priv, skb);
+}
+
+/* pfe_eth_get_stats
+ */
+static struct net_device_stats *pfe_eth_get_stats(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+
+	netif_info(priv, drv, ndev, "%s\n", __func__);
+
+	return &priv->stats;
+}
+
+/* pfe_eth_set_mac_address
+ */
+static int pfe_eth_set_mac_address(struct net_device *ndev, void *addr)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	struct sockaddr *sa = addr;
+
+	netif_info(priv, drv, ndev, "%s\n", __func__);
+
+	if (!is_valid_ether_addr(sa->sa_data))
+		return -EADDRNOTAVAIL;
+
+	memcpy(ndev->dev_addr, sa->sa_data, ETH_ALEN);
+
+	gemac_set_laddrN(priv->EMAC_baseaddr,
+			 (struct pfe_mac_addr *)ndev->dev_addr, 1);
+
+	return 0;
+}
+
+/* pfe_eth_enet_addr_byte_mac
+ */
+int pfe_eth_enet_addr_byte_mac(u8 *enet_byte_addr,
+			       struct pfe_mac_addr *enet_addr)
+{
+	if (!enet_byte_addr || !enet_addr) {
+		return -1;
+
+	} else {
+		enet_addr->bottom = enet_byte_addr[0] |
+			(enet_byte_addr[1] << 8) |
+			(enet_byte_addr[2] << 16) |
+			(enet_byte_addr[3] << 24);
+		enet_addr->top = enet_byte_addr[4] |
+			(enet_byte_addr[5] << 8);
+		return 0;
+	}
+}
+
+/* pfe_eth_set_multi
+ */
+static void pfe_eth_set_multi(struct net_device *ndev)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	struct pfe_mac_addr    hash_addr; /* hash register structure */
+	/* specific mac address	register structure */
+	struct pfe_mac_addr    spec_addr;
+	int		result; /* index into hash register to set.. */
+	int		uc_count = 0;
+	struct netdev_hw_addr *ha;
+
+	if (ndev->flags & IFF_PROMISC) {
+		netif_info(priv, drv, ndev, "entering promiscuous mode\n");
+
+		priv->promisc = 1;
+		gemac_enable_copy_all(priv->EMAC_baseaddr);
+	} else {
+		priv->promisc = 0;
+		gemac_disable_copy_all(priv->EMAC_baseaddr);
+	}
+
+	/* Enable broadcast frame reception if required. */
+	if (ndev->flags & IFF_BROADCAST) {
+		gemac_allow_broadcast(priv->EMAC_baseaddr);
+	} else {
+		netif_info(priv, drv, ndev,
+			   "disabling broadcast frame reception\n");
+
+		gemac_no_broadcast(priv->EMAC_baseaddr);
+	}
+
+	if (ndev->flags & IFF_ALLMULTI) {
+		/* Set the hash to rx all multicast frames */
+		hash_addr.bottom = 0xFFFFFFFF;
+		hash_addr.top = 0xFFFFFFFF;
+		gemac_set_hash(priv->EMAC_baseaddr, &hash_addr);
+		netdev_for_each_uc_addr(ha, ndev) {
+			if (uc_count >= MAX_UC_SPEC_ADDR_REG)
+				break;
+			pfe_eth_enet_addr_byte_mac(ha->addr, &spec_addr);
+			gemac_set_laddrN(priv->EMAC_baseaddr, &spec_addr,
+					 uc_count + 2);
+			uc_count++;
+		}
+	} else if ((netdev_mc_count(ndev) > 0)  || (netdev_uc_count(ndev))) {
+		u8 *addr;
+
+		hash_addr.bottom = 0;
+		hash_addr.top = 0;
+
+		netdev_for_each_mc_addr(ha, ndev) {
+			addr = ha->addr;
+
+			netif_info(priv, drv, ndev,
+				   "adding multicast address %X:%X:%X:%X:%X:%X to gem filter\n",
+				addr[0], addr[1], addr[2],
+				addr[3], addr[4], addr[5]);
+
+			result = pfe_eth_get_hash(addr);
+
+			if (result < EMAC_HASH_REG_BITS) {
+				if (result < 32)
+					hash_addr.bottom |= (1 << result);
+				else
+					hash_addr.top |= (1 << (result - 32));
+			} else {
+				break;
+			}
+		}
+
+		uc_count = -1;
+		netdev_for_each_uc_addr(ha, ndev) {
+			addr = ha->addr;
+
+			if (++uc_count < MAX_UC_SPEC_ADDR_REG)   {
+				netdev_info(ndev,
+					    "adding unicast address %02x:%02x:%02x:%02x:%02x:%02x to gem filter\n",
+					    addr[0], addr[1], addr[2],
+					    addr[3], addr[4], addr[5]);
+				pfe_eth_enet_addr_byte_mac(addr, &spec_addr);
+				gemac_set_laddrN(priv->EMAC_baseaddr,
+						 &spec_addr, uc_count + 2);
+			} else {
+				netif_info(priv, drv, ndev,
+					   "adding unicast address %02x:%02x:%02x:%02x:%02x:%02x to gem hash\n",
+					   addr[0], addr[1], addr[2],
+					   addr[3], addr[4], addr[5]);
+
+				result = pfe_eth_get_hash(addr);
+				if (result >= EMAC_HASH_REG_BITS) {
+					break;
+
+				} else {
+					if (result < 32)
+						hash_addr.bottom |= (1 <<
+								result);
+					else
+						hash_addr.top |= (1 <<
+								(result - 32));
+				}
+			}
+		}
+
+		gemac_set_hash(priv->EMAC_baseaddr, &hash_addr);
+	}
+
+	if (!(netdev_uc_count(ndev) >= MAX_UC_SPEC_ADDR_REG)) {
+		/*
+		 *  Check if there are any specific address HW registers that
+		 * need to be flushed
+		 */
+		for (uc_count = netdev_uc_count(ndev); uc_count <
+			MAX_UC_SPEC_ADDR_REG; uc_count++)
+			gemac_clear_laddrN(priv->EMAC_baseaddr, uc_count + 2);
+	}
+
+	if (ndev->flags & IFF_LOOPBACK)
+		gemac_set_loop(priv->EMAC_baseaddr, LB_LOCAL);
+}
+
+/* pfe_eth_set_features
+ */
+static int pfe_eth_set_features(struct net_device *ndev, netdev_features_t
+					features)
+{
+	struct pfe_eth_priv_s *priv = netdev_priv(ndev);
+	int rc = 0;
+
+	if (features & NETIF_F_RXCSUM)
+		gemac_enable_rx_checksum_offload(priv->EMAC_baseaddr);
+	else
+		gemac_disable_rx_checksum_offload(priv->EMAC_baseaddr);
+	return rc;
+}
+
+/* pfe_eth_fast_tx_timeout
+ */
+static enum hrtimer_restart pfe_eth_fast_tx_timeout(struct hrtimer *timer)
+{
+	struct pfe_eth_fast_timer *fast_tx_timeout = container_of(timer, struct
+							pfe_eth_fast_timer,
+							timer);
+	struct pfe_eth_priv_s *priv =  container_of(fast_tx_timeout->base,
+							struct pfe_eth_priv_s,
+							fast_tx_timeout);
+	struct netdev_queue *tx_queue = netdev_get_tx_queue(priv->ndev,
+						fast_tx_timeout->queuenum);
+
+	if (netif_tx_queue_stopped(tx_queue)) {
+#ifdef PFE_ETH_TX_STATS
+		priv->was_stopped[fast_tx_timeout->queuenum] = 1;
+#endif
+		netif_tx_wake_queue(tx_queue);
+	}
+
+	return HRTIMER_NORESTART;
+}
+
+/* pfe_eth_fast_tx_timeout_init
+ */
+static void pfe_eth_fast_tx_timeout_init(struct pfe_eth_priv_s *priv)
+{
+	int i;
+
+	for (i = 0; i < emac_txq_cnt; i++) {
+		priv->fast_tx_timeout[i].queuenum = i;
+		hrtimer_init(&priv->fast_tx_timeout[i].timer, CLOCK_MONOTONIC,
+			     HRTIMER_MODE_REL);
+		priv->fast_tx_timeout[i].timer.function =
+				pfe_eth_fast_tx_timeout;
+		priv->fast_tx_timeout[i].base = priv->fast_tx_timeout;
+	}
+}
+
+static struct sk_buff *pfe_eth_rx_skb(struct net_device *ndev,
+				      struct	pfe_eth_priv_s *priv,
+				      unsigned int qno)
+{
+	void *buf_addr;
+	unsigned int rx_ctrl;
+	unsigned int desc_ctrl = 0;
+	struct hif_ipsec_hdr *ipsec_hdr = NULL;
+	struct sk_buff *skb;
+	struct sk_buff *skb_frag, *skb_frag_last = NULL;
+	int length = 0, offset;
+
+	skb = priv->skb_inflight[qno];
+
+	if (skb) {
+		skb_frag_last = skb_shinfo(skb)->frag_list;
+		if (skb_frag_last) {
+			while (skb_frag_last->next)
+				skb_frag_last = skb_frag_last->next;
+		}
+	}
+
+	while (!(desc_ctrl & CL_DESC_LAST)) {
+		buf_addr = hif_lib_receive_pkt(&priv->client, qno, &length,
+					       &offset, &rx_ctrl, &desc_ctrl,
+					       (void **)&ipsec_hdr);
+		if (!buf_addr)
+			goto incomplete;
+
+#ifdef PFE_ETH_NAPI_STATS
+		priv->napi_counters[NAPI_DESC_COUNT]++;
+#endif
+
+		/* First frag */
+		if (desc_ctrl & CL_DESC_FIRST) {
+			skb = build_skb(buf_addr, 0);
+			if (unlikely(!skb))
+				goto pkt_drop;
+
+			skb_reserve(skb, offset);
+			skb_put(skb, length);
+			skb->dev = ndev;
+
+			if ((ndev->features & NETIF_F_RXCSUM) && (rx_ctrl &
+					HIF_CTRL_RX_CHECKSUMMED))
+				skb->ip_summed = CHECKSUM_UNNECESSARY;
+			else
+				skb_checksum_none_assert(skb);
+
+		} else {
+			/* Next frags */
+			if (unlikely(!skb)) {
+				pr_err("%s: NULL skb_inflight\n",
+				       __func__);
+				goto pkt_drop;
+			}
+
+			skb_frag = build_skb(buf_addr, 0);
+
+			if (unlikely(!skb_frag)) {
+				kfree(buf_addr);
+				goto pkt_drop;
+			}
+
+			skb_reserve(skb_frag, offset);
+			skb_put(skb_frag, length);
+
+			skb_frag->dev = ndev;
+
+			if (skb_shinfo(skb)->frag_list)
+				skb_frag_last->next = skb_frag;
+			else
+				skb_shinfo(skb)->frag_list = skb_frag;
+
+			skb->truesize += skb_frag->truesize;
+			skb->data_len += length;
+			skb->len += length;
+			skb_frag_last = skb_frag;
+		}
+	}
+
+	priv->skb_inflight[qno] = NULL;
+	return skb;
+
+incomplete:
+	priv->skb_inflight[qno] = skb;
+	return NULL;
+
+pkt_drop:
+	priv->skb_inflight[qno] = NULL;
+
+	if (skb)
+		kfree_skb(skb);
+	else
+		kfree(buf_addr);
+
+	priv->stats.rx_errors++;
+
+	return NULL;
+}
+
+/* pfe_eth_poll
+ */
+static int pfe_eth_poll(struct pfe_eth_priv_s *priv, struct napi_struct *napi,
+			unsigned int qno, int budget)
+{
+	struct net_device *ndev = priv->ndev;
+	struct sk_buff *skb;
+	int work_done = 0;
+	unsigned int len;
+
+	netif_info(priv, intr, priv->ndev, "%s\n", __func__);
+
+#ifdef PFE_ETH_NAPI_STATS
+	priv->napi_counters[NAPI_POLL_COUNT]++;
+#endif
+
+	do {
+		skb = pfe_eth_rx_skb(ndev, priv, qno);
+
+		if (!skb)
+			break;
+
+		len = skb->len;
+
+		/* Packet will be processed */
+		skb->protocol = eth_type_trans(skb, ndev);
+
+		netif_receive_skb(skb);
+
+		priv->stats.rx_packets++;
+		priv->stats.rx_bytes += len;
+
+		work_done++;
+
+#ifdef PFE_ETH_NAPI_STATS
+		priv->napi_counters[NAPI_PACKET_COUNT]++;
+#endif
+
+	} while (work_done < budget);
+
+	/*
+	 * If no Rx receive nor cleanup work was done, exit polling mode.
+	 * No more netif_running(dev) check is required here , as this is
+	 * checked in net/core/dev.c (2.6.33.5 kernel specific).
+	 */
+	if (work_done < budget) {
+		napi_complete(napi);
+
+		hif_lib_event_handler_start(&priv->client, EVENT_RX_PKT_IND,
+					    qno);
+	}
+#ifdef PFE_ETH_NAPI_STATS
+	else
+		priv->napi_counters[NAPI_FULL_BUDGET_COUNT]++;
+#endif
+
+	return work_done;
+}
+
+/*
+ * pfe_eth_lro_poll
+ */
+static int pfe_eth_lro_poll(struct napi_struct *napi, int budget)
+{
+	struct pfe_eth_priv_s *priv = container_of(napi, struct pfe_eth_priv_s,
+							lro_napi);
+
+	netif_info(priv, intr, priv->ndev, "%s\n", __func__);
+
+	return pfe_eth_poll(priv, napi, 2, budget);
+}
+
+/* pfe_eth_low_poll
+ */
+static int pfe_eth_low_poll(struct napi_struct *napi, int budget)
+{
+	struct pfe_eth_priv_s *priv = container_of(napi, struct pfe_eth_priv_s,
+							low_napi);
+
+	netif_info(priv, intr, priv->ndev, "%s\n", __func__);
+
+	return pfe_eth_poll(priv, napi, 1, budget);
+}
+
+/* pfe_eth_high_poll
+ */
+static int pfe_eth_high_poll(struct napi_struct *napi, int budget)
+{
+	struct pfe_eth_priv_s *priv = container_of(napi, struct pfe_eth_priv_s,
+							high_napi);
+
+	netif_info(priv, intr, priv->ndev, "%s\n", __func__);
+
+	return pfe_eth_poll(priv, napi, 0, budget);
+}
+
+static const struct net_device_ops pfe_netdev_ops = {
+	.ndo_open = pfe_eth_open,
+	.ndo_stop = pfe_eth_close,
+	.ndo_start_xmit = pfe_eth_send_packet,
+	.ndo_select_queue = pfe_eth_select_queue,
+	.ndo_get_stats = pfe_eth_get_stats,
+	.ndo_set_mac_address = pfe_eth_set_mac_address,
+	.ndo_set_rx_mode = pfe_eth_set_multi,
+	.ndo_set_features = pfe_eth_set_features,
+	.ndo_validate_addr = eth_validate_addr,
+};
+
+/* pfe_eth_init_one
+ */
+static int pfe_eth_init_one(struct pfe *pfe, int id)
+{
+	struct net_device *ndev = NULL;
+	struct pfe_eth_priv_s *priv = NULL;
+	struct ls1012a_eth_platform_data *einfo;
+	struct ls1012a_mdio_platform_data *minfo;
+	struct ls1012a_pfe_platform_data *pfe_info;
+	int err;
+
+	/* Extract pltform data */
+	pfe_info = (struct ls1012a_pfe_platform_data *)
+					pfe->dev->platform_data;
+	if (!pfe_info) {
+		pr_err(
+			"%s: pfe missing additional platform data\n"
+			, __func__);
+		err = -ENODEV;
+		goto err0;
+	}
+
+	einfo = (struct ls1012a_eth_platform_data *)
+				pfe_info->ls1012a_eth_pdata;
+
+	/* einfo never be NULL, but no harm in having this check */
+	if (!einfo) {
+		pr_err(
+			"%s: pfe missing additional gemacs platform data\n"
+			, __func__);
+		err = -ENODEV;
+		goto err0;
+	}
+
+	minfo = (struct ls1012a_mdio_platform_data *)
+				pfe_info->ls1012a_mdio_pdata;
+
+	/* einfo never be NULL, but no harm in having this check */
+	if (!minfo) {
+		pr_err(
+			"%s: pfe missing additional mdios platform data\n",
+			 __func__);
+		err = -ENODEV;
+		goto err0;
+	}
+
+	/* Create an ethernet device instance */
+	ndev = alloc_etherdev_mq(sizeof(*priv), emac_txq_cnt);
+
+	if (!ndev) {
+		pr_err("%s: gemac %d device allocation failed\n",
+		       __func__, einfo[id].gem_id);
+		err = -ENOMEM;
+		goto err0;
+	}
+
+	priv = netdev_priv(ndev);
+	priv->ndev = ndev;
+	priv->id = einfo[id].gem_id;
+	priv->pfe = pfe;
+
+	SET_NETDEV_DEV(priv->ndev, priv->pfe->dev);
+
+	pfe->eth.eth_priv[id] = priv;
+
+	/* Set the info in the priv to the current info */
+	priv->einfo = &einfo[id];
+	priv->EMAC_baseaddr = cbus_emac_base[id];
+	priv->PHY_baseaddr = cbus_emac_base[0];
+	priv->GPI_baseaddr = cbus_gpi_base[id];
+
+#define HIF_GEMAC_TMUQ_BASE	6
+	priv->low_tmu_q	=  HIF_GEMAC_TMUQ_BASE + (id * 2);
+	priv->high_tmu_q	=  priv->low_tmu_q + 1;
+
+	spin_lock_init(&priv->lock);
+
+	pfe_eth_fast_tx_timeout_init(priv);
+
+	/* Copy the station address into the dev structure, */
+	memcpy(ndev->dev_addr, einfo[id].mac_addr, ETH_ALEN);
+
+	/* Initialize mdio */
+	if (minfo[id].enabled) {
+		err = pfe_eth_mdio_init(priv, &minfo[id]);
+		if (err) {
+			netdev_err(ndev, "%s: pfe_eth_mdio_init() failed\n",
+				   __func__);
+			goto err2;
+		}
+	}
+
+	ndev->mtu = 1500;
+
+	/* Set MTU limits */
+	ndev->min_mtu = ETH_MIN_MTU;
+	ndev->max_mtu = JUMBO_FRAME_SIZE;
+
+	/* supported features */
+	ndev->hw_features = NETIF_F_SG;
+
+	/*Enable after checksum offload is validated */
+	ndev->hw_features = NETIF_F_RXCSUM | NETIF_F_IP_CSUM |
+		NETIF_F_IPV6_CSUM | NETIF_F_SG;
+
+	/* enabled by default */
+	ndev->features = ndev->hw_features;
+
+	priv->usr_features = ndev->features;
+
+	ndev->netdev_ops = &pfe_netdev_ops;
+
+	ndev->ethtool_ops = &pfe_ethtool_ops;
+
+	/* Enable basic messages by default */
+	priv->msg_enable = NETIF_MSG_IFUP | NETIF_MSG_IFDOWN | NETIF_MSG_LINK |
+				NETIF_MSG_PROBE;
+
+	netif_napi_add(ndev, &priv->low_napi, pfe_eth_low_poll,
+		       HIF_RX_POLL_WEIGHT - 16);
+	netif_napi_add(ndev, &priv->high_napi, pfe_eth_high_poll,
+		       HIF_RX_POLL_WEIGHT - 16);
+	netif_napi_add(ndev, &priv->lro_napi, pfe_eth_lro_poll,
+		       HIF_RX_POLL_WEIGHT - 16);
+
+	err = register_netdev(ndev);
+
+	if (err) {
+		netdev_err(ndev, "register_netdev() failed\n");
+		goto err3;
+	}
+	device_init_wakeup(&ndev->dev, WAKE_MAGIC);
+
+	if (!(priv->einfo->phy_flags & GEMAC_NO_PHY)) {
+		err = pfe_phy_init(ndev);
+		if (err) {
+			netdev_err(ndev, "%s: pfe_phy_init() failed\n",
+				   __func__);
+			goto err4;
+		}
+	}
+
+	netif_carrier_on(ndev);
+
+	/* Create all the sysfs files */
+	if (pfe_eth_sysfs_init(ndev))
+		goto err4;
+
+	netif_info(priv, probe, ndev, "%s: created interface, baseaddr: %p\n",
+		   __func__, priv->EMAC_baseaddr);
+
+	return 0;
+err4:
+	unregister_netdev(ndev);
+err3:
+	pfe_eth_mdio_exit(priv->mii_bus);
+err2:
+	free_netdev(priv->ndev);
+err0:
+	return err;
+}
+
+/* pfe_eth_init
+ */
+int pfe_eth_init(struct pfe *pfe)
+{
+	int ii = 0;
+	int err;
+
+	pr_info("%s\n", __func__);
+
+	cbus_emac_base[0] = EMAC1_BASE_ADDR;
+	cbus_emac_base[1] = EMAC2_BASE_ADDR;
+
+	cbus_gpi_base[0] = EGPI1_BASE_ADDR;
+	cbus_gpi_base[1] = EGPI2_BASE_ADDR;
+
+	for (ii = 0; ii < NUM_GEMAC_SUPPORT; ii++) {
+		err = pfe_eth_init_one(pfe, ii);
+		if (err)
+			goto err0;
+	}
+
+	return 0;
+
+err0:
+	while (ii--)
+		pfe_eth_exit_one(pfe->eth.eth_priv[ii]);
+
+	/* Register three network devices in the kernel */
+	return err;
+}
+
+/* pfe_eth_exit_one
+ */
+static void pfe_eth_exit_one(struct pfe_eth_priv_s *priv)
+{
+	netif_info(priv, probe, priv->ndev, "%s\n", __func__);
+
+	pfe_eth_sysfs_exit(priv->ndev);
+
+	unregister_netdev(priv->ndev);
+
+	if (!(priv->einfo->phy_flags & GEMAC_NO_PHY))
+		pfe_phy_exit(priv->ndev);
+
+	if (priv->mii_bus)
+		pfe_eth_mdio_exit(priv->mii_bus);
+
+	free_netdev(priv->ndev);
+}
+
+/* pfe_eth_exit
+ */
+void pfe_eth_exit(struct pfe *pfe)
+{
+	int ii;
+
+	pr_info("%s\n", __func__);
+
+	for (ii = NUM_GEMAC_SUPPORT - 1; ii >= 0; ii--)
+		pfe_eth_exit_one(pfe->eth.eth_priv[ii]);
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_firmware.c b/drivers/staging/fsl_ppfe/pfe_firmware.c
new file mode 100644
index 0000000..47462b9
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_firmware.c
@@ -0,0 +1,314 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * @file
+ * Contains all the functions to handle parsing and loading of PE firmware
+ * files.
+ */
+#include <linux/firmware.h>
+
+#include "pfe_mod.h"
+#include "pfe_firmware.h"
+#include "pfe/pfe.h"
+
+static struct elf32_shdr *get_elf_section_header(const struct firmware *fw,
+						 const char *section)
+{
+	struct elf32_hdr *elf_hdr = (struct elf32_hdr *)fw->data;
+	struct elf32_shdr *shdr;
+	struct elf32_shdr *shdr_shstr;
+	Elf32_Off e_shoff = be32_to_cpu(elf_hdr->e_shoff);
+	Elf32_Half e_shentsize = be16_to_cpu(elf_hdr->e_shentsize);
+	Elf32_Half e_shnum = be16_to_cpu(elf_hdr->e_shnum);
+	Elf32_Half e_shstrndx = be16_to_cpu(elf_hdr->e_shstrndx);
+	Elf32_Off shstr_offset;
+	Elf32_Word sh_name;
+	const char *name;
+	int i;
+
+	/* Section header strings */
+	shdr_shstr = (struct elf32_shdr *)(fw->data + e_shoff + e_shstrndx *
+					e_shentsize);
+	shstr_offset = be32_to_cpu(shdr_shstr->sh_offset);
+
+	for (i = 0; i < e_shnum; i++) {
+		shdr = (struct elf32_shdr *)(fw->data + e_shoff
+					     + i * e_shentsize);
+
+		sh_name = be32_to_cpu(shdr->sh_name);
+
+		name = (const char *)(fw->data + shstr_offset + sh_name);
+
+		if (!strcmp(name, section))
+			return shdr;
+	}
+
+	pr_err("%s: didn't find section %s\n", __func__, section);
+
+	return NULL;
+}
+
+#if defined(CFG_DIAGS)
+static int pfe_get_diags_info(const struct firmware *fw, struct pfe_diags_info
+				*diags_info)
+{
+	struct elf32_shdr *shdr;
+	unsigned long offset, size;
+
+	shdr = get_elf_section_header(fw, ".pfe_diags_str");
+	if (shdr) {
+		offset = be32_to_cpu(shdr->sh_offset);
+		size = be32_to_cpu(shdr->sh_size);
+		diags_info->diags_str_base = be32_to_cpu(shdr->sh_addr);
+		diags_info->diags_str_size = size;
+		diags_info->diags_str_array = kmalloc(size, GFP_KERNEL);
+		memcpy(diags_info->diags_str_array, fw->data + offset, size);
+
+		return 0;
+	} else {
+		return -1;
+	}
+}
+#endif
+
+static void pfe_check_version_info(const struct firmware *fw)
+{
+	/*static char *version = NULL;*/
+	static char *version;
+
+	struct elf32_shdr *shdr = get_elf_section_header(fw, ".version");
+
+	if (shdr) {
+		if (!version) {
+			/*
+			 * this is the first fw we load, use its version
+			 * string as reference (whatever it is)
+			 */
+			version = (char *)(fw->data +
+					be32_to_cpu(shdr->sh_offset));
+
+			pr_info("PFE binary version: %s\n", version);
+		} else {
+			/*
+			 * already have loaded at least one firmware, check
+			 * sequence can start now
+			 */
+			if (strcmp(version, (char *)(fw->data +
+				be32_to_cpu(shdr->sh_offset)))) {
+				pr_info(
+				"WARNING: PFE firmware binaries from incompatible version\n");
+			}
+		}
+	} else {
+		/*
+		 * version cannot be verified, a potential issue that should
+		 * be reported
+		 */
+		pr_info(
+			 "WARNING: PFE firmware binaries from incompatible version\n");
+	}
+}
+
+/* PFE elf firmware loader.
+ * Loads an elf firmware image into a list of PE's (specified using a bitmask)
+ *
+ * @param pe_mask	Mask of PE id's to load firmware to
+ * @param fw		Pointer to the firmware image
+ *
+ * @return		0 on success, a negative value on error
+ *
+ */
+int pfe_load_elf(int pe_mask, const struct firmware *fw, struct pfe *pfe)
+{
+	struct elf32_hdr *elf_hdr = (struct elf32_hdr *)fw->data;
+	Elf32_Half sections = be16_to_cpu(elf_hdr->e_shnum);
+	struct elf32_shdr *shdr = (struct elf32_shdr *)(fw->data +
+					be32_to_cpu(elf_hdr->e_shoff));
+	int id, section;
+	int rc;
+
+	pr_info("%s\n", __func__);
+
+	/* Some sanity checks */
+	if (strncmp(&elf_hdr->e_ident[EI_MAG0], ELFMAG, SELFMAG)) {
+		pr_err("%s: incorrect elf magic number\n", __func__);
+		return -EINVAL;
+	}
+
+	if (elf_hdr->e_ident[EI_CLASS] != ELFCLASS32) {
+		pr_err("%s: incorrect elf class(%x)\n", __func__,
+		       elf_hdr->e_ident[EI_CLASS]);
+		return -EINVAL;
+	}
+
+	if (elf_hdr->e_ident[EI_DATA] != ELFDATA2MSB) {
+		pr_err("%s: incorrect elf data(%x)\n", __func__,
+		       elf_hdr->e_ident[EI_DATA]);
+		return -EINVAL;
+	}
+
+	if (be16_to_cpu(elf_hdr->e_type) != ET_EXEC) {
+		pr_err("%s: incorrect elf file type(%x)\n", __func__,
+		       be16_to_cpu(elf_hdr->e_type));
+		return -EINVAL;
+	}
+
+	for (section = 0; section < sections; section++, shdr++) {
+		if (!(be32_to_cpu(shdr->sh_flags) & (SHF_WRITE | SHF_ALLOC |
+			SHF_EXECINSTR)))
+			continue;
+
+		for (id = 0; id < MAX_PE; id++)
+			if (pe_mask & (1 << id)) {
+				rc = pe_load_elf_section(id, fw->data, shdr,
+							 pfe->dev);
+				if (rc < 0)
+					goto err;
+			}
+	}
+
+	pfe_check_version_info(fw);
+
+	return 0;
+
+err:
+	return rc;
+}
+
+/* PFE firmware initialization.
+ * Loads different firmware files from filesystem.
+ * Initializes PE IMEM/DMEM and UTIL-PE DDR
+ * Initializes control path symbol addresses (by looking them up in the elf
+ * firmware files
+ * Takes PE's out of reset
+ *
+ * @return	0 on success, a negative value on error
+ *
+ */
+int pfe_firmware_init(struct pfe *pfe)
+{
+	const struct firmware *class_fw, *tmu_fw;
+	int rc = 0;
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	const char *util_fw_name;
+	const struct firmware *util_fw;
+#endif
+
+	pr_info("%s\n", __func__);
+
+	if (request_firmware(&class_fw, CLASS_FIRMWARE_FILENAME, pfe->dev)) {
+		pr_err("%s: request firmware %s failed\n", __func__,
+		       CLASS_FIRMWARE_FILENAME);
+		rc = -ETIMEDOUT;
+		goto err0;
+	}
+
+	if (request_firmware(&tmu_fw, TMU_FIRMWARE_FILENAME, pfe->dev)) {
+		pr_err("%s: request firmware %s failed\n", __func__,
+		       TMU_FIRMWARE_FILENAME);
+		rc = -ETIMEDOUT;
+		goto err1;
+}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	util_fw_name = UTIL_FIRMWARE_FILENAME;
+
+	if (request_firmware(&util_fw, util_fw_name, pfe->dev)) {
+		pr_err("%s: request firmware %s failed\n", __func__,
+		       util_fw_name);
+		rc = -ETIMEDOUT;
+		goto err2;
+	}
+#endif
+	rc = pfe_load_elf(CLASS_MASK, class_fw, pfe);
+	if (rc < 0) {
+		pr_err("%s: class firmware load failed\n", __func__);
+		goto err3;
+	}
+
+#if defined(CFG_DIAGS)
+	rc = pfe_get_diags_info(class_fw, &pfe->diags.class_diags_info);
+	if (rc < 0) {
+		pr_warn(
+			"PFE diags won't be available for class PEs\n");
+		rc = 0;
+	}
+#endif
+
+	rc = pfe_load_elf(TMU_MASK, tmu_fw, pfe);
+	if (rc < 0) {
+		pr_err("%s: tmu firmware load failed\n", __func__);
+		goto err3;
+	}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	rc = pfe_load_elf(UTIL_MASK, util_fw, pfe);
+	if (rc < 0) {
+		pr_err("%s: util firmware load failed\n", __func__);
+		goto err3;
+	}
+
+#if defined(CFG_DIAGS)
+	rc = pfe_get_diags_info(util_fw, &pfe->diags.util_diags_info);
+	if (rc < 0) {
+		pr_warn(
+			"PFE diags won't be available for util PE\n");
+		rc = 0;
+	}
+#endif
+
+	util_enable();
+#endif
+
+	tmu_enable(0xf);
+	class_enable();
+
+err3:
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	release_firmware(util_fw);
+
+err2:
+#endif
+	release_firmware(tmu_fw);
+
+err1:
+	release_firmware(class_fw);
+
+err0:
+	return rc;
+}
+
+/* PFE firmware cleanup
+ * Puts PE's in reset
+ *
+ *
+ */
+void pfe_firmware_exit(struct pfe *pfe)
+{
+	pr_info("%s\n", __func__);
+
+	if (pe_reset_all(&pfe->ctrl) != 0)
+		pr_err("Error: Failed to stop PEs, PFE reload may not work correctly\n");
+
+	class_disable();
+	tmu_disable(0xf);
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	util_disable();
+#endif
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_hal.c b/drivers/staging/fsl_ppfe/pfe_hal.c
new file mode 100644
index 0000000..0915034
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_hal.c
@@ -0,0 +1,1516 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "pfe_mod.h"
+#include "pfe/pfe.h"
+
+void *cbus_base_addr;
+void *ddr_base_addr;
+unsigned long ddr_phys_base_addr;
+unsigned int ddr_size;
+
+static struct pe_info pe[MAX_PE];
+
+/* Initializes the PFE library.
+ * Must be called before using any of the library functions.
+ *
+ * @param[in] cbus_base		CBUS virtual base address (as mapped in
+ * the host CPU address space)
+ * @param[in] ddr_base		PFE DDR range virtual base address (as
+ * mapped in the host CPU address space)
+ * @param[in] ddr_phys_base	PFE DDR range physical base address (as
+ * mapped in platform)
+ * @param[in] size		PFE DDR range size (as defined by the host
+ * software)
+ */
+void pfe_lib_init(void *cbus_base, void *ddr_base, unsigned long ddr_phys_base,
+		  unsigned int size)
+{
+	cbus_base_addr = cbus_base;
+	ddr_base_addr = ddr_base;
+	ddr_phys_base_addr = ddr_phys_base;
+	ddr_size = size;
+
+	pe[CLASS0_ID].dmem_base_addr = CLASS_DMEM_BASE_ADDR(0);
+	pe[CLASS0_ID].pmem_base_addr = CLASS_IMEM_BASE_ADDR(0);
+	pe[CLASS0_ID].pmem_size = CLASS_IMEM_SIZE;
+	pe[CLASS0_ID].mem_access_wdata = CLASS_MEM_ACCESS_WDATA;
+	pe[CLASS0_ID].mem_access_addr = CLASS_MEM_ACCESS_ADDR;
+	pe[CLASS0_ID].mem_access_rdata = CLASS_MEM_ACCESS_RDATA;
+
+	pe[CLASS1_ID].dmem_base_addr = CLASS_DMEM_BASE_ADDR(1);
+	pe[CLASS1_ID].pmem_base_addr = CLASS_IMEM_BASE_ADDR(1);
+	pe[CLASS1_ID].pmem_size = CLASS_IMEM_SIZE;
+	pe[CLASS1_ID].mem_access_wdata = CLASS_MEM_ACCESS_WDATA;
+	pe[CLASS1_ID].mem_access_addr = CLASS_MEM_ACCESS_ADDR;
+	pe[CLASS1_ID].mem_access_rdata = CLASS_MEM_ACCESS_RDATA;
+
+	pe[CLASS2_ID].dmem_base_addr = CLASS_DMEM_BASE_ADDR(2);
+	pe[CLASS2_ID].pmem_base_addr = CLASS_IMEM_BASE_ADDR(2);
+	pe[CLASS2_ID].pmem_size = CLASS_IMEM_SIZE;
+	pe[CLASS2_ID].mem_access_wdata = CLASS_MEM_ACCESS_WDATA;
+	pe[CLASS2_ID].mem_access_addr = CLASS_MEM_ACCESS_ADDR;
+	pe[CLASS2_ID].mem_access_rdata = CLASS_MEM_ACCESS_RDATA;
+
+	pe[CLASS3_ID].dmem_base_addr = CLASS_DMEM_BASE_ADDR(3);
+	pe[CLASS3_ID].pmem_base_addr = CLASS_IMEM_BASE_ADDR(3);
+	pe[CLASS3_ID].pmem_size = CLASS_IMEM_SIZE;
+	pe[CLASS3_ID].mem_access_wdata = CLASS_MEM_ACCESS_WDATA;
+	pe[CLASS3_ID].mem_access_addr = CLASS_MEM_ACCESS_ADDR;
+	pe[CLASS3_ID].mem_access_rdata = CLASS_MEM_ACCESS_RDATA;
+
+	pe[CLASS4_ID].dmem_base_addr = CLASS_DMEM_BASE_ADDR(4);
+	pe[CLASS4_ID].pmem_base_addr = CLASS_IMEM_BASE_ADDR(4);
+	pe[CLASS4_ID].pmem_size = CLASS_IMEM_SIZE;
+	pe[CLASS4_ID].mem_access_wdata = CLASS_MEM_ACCESS_WDATA;
+	pe[CLASS4_ID].mem_access_addr = CLASS_MEM_ACCESS_ADDR;
+	pe[CLASS4_ID].mem_access_rdata = CLASS_MEM_ACCESS_RDATA;
+
+	pe[CLASS5_ID].dmem_base_addr = CLASS_DMEM_BASE_ADDR(5);
+	pe[CLASS5_ID].pmem_base_addr = CLASS_IMEM_BASE_ADDR(5);
+	pe[CLASS5_ID].pmem_size = CLASS_IMEM_SIZE;
+	pe[CLASS5_ID].mem_access_wdata = CLASS_MEM_ACCESS_WDATA;
+	pe[CLASS5_ID].mem_access_addr = CLASS_MEM_ACCESS_ADDR;
+	pe[CLASS5_ID].mem_access_rdata = CLASS_MEM_ACCESS_RDATA;
+
+	pe[TMU0_ID].dmem_base_addr = TMU_DMEM_BASE_ADDR(0);
+	pe[TMU0_ID].pmem_base_addr = TMU_IMEM_BASE_ADDR(0);
+	pe[TMU0_ID].pmem_size = TMU_IMEM_SIZE;
+	pe[TMU0_ID].mem_access_wdata = TMU_MEM_ACCESS_WDATA;
+	pe[TMU0_ID].mem_access_addr = TMU_MEM_ACCESS_ADDR;
+	pe[TMU0_ID].mem_access_rdata = TMU_MEM_ACCESS_RDATA;
+
+	pe[TMU1_ID].dmem_base_addr = TMU_DMEM_BASE_ADDR(1);
+	pe[TMU1_ID].pmem_base_addr = TMU_IMEM_BASE_ADDR(1);
+	pe[TMU1_ID].pmem_size = TMU_IMEM_SIZE;
+	pe[TMU1_ID].mem_access_wdata = TMU_MEM_ACCESS_WDATA;
+	pe[TMU1_ID].mem_access_addr = TMU_MEM_ACCESS_ADDR;
+	pe[TMU1_ID].mem_access_rdata = TMU_MEM_ACCESS_RDATA;
+
+	pe[TMU3_ID].dmem_base_addr = TMU_DMEM_BASE_ADDR(3);
+	pe[TMU3_ID].pmem_base_addr = TMU_IMEM_BASE_ADDR(3);
+	pe[TMU3_ID].pmem_size = TMU_IMEM_SIZE;
+	pe[TMU3_ID].mem_access_wdata = TMU_MEM_ACCESS_WDATA;
+	pe[TMU3_ID].mem_access_addr = TMU_MEM_ACCESS_ADDR;
+	pe[TMU3_ID].mem_access_rdata = TMU_MEM_ACCESS_RDATA;
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	pe[UTIL_ID].dmem_base_addr = UTIL_DMEM_BASE_ADDR;
+	pe[UTIL_ID].mem_access_wdata = UTIL_MEM_ACCESS_WDATA;
+	pe[UTIL_ID].mem_access_addr = UTIL_MEM_ACCESS_ADDR;
+	pe[UTIL_ID].mem_access_rdata = UTIL_MEM_ACCESS_RDATA;
+#endif
+}
+
+/* Writes a buffer to PE internal memory from the host
+ * through indirect access registers.
+ *
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., UTIL_ID)
+ * @param[in] src		Buffer source address
+ * @param[in] mem_access_addr	DMEM destination address (must be 32bit
+ * aligned)
+ * @param[in] len		Number of bytes to copy
+ */
+void pe_mem_memcpy_to32(int id, u32 mem_access_addr, const void *src, unsigned
+int len)
+{
+	u32 offset = 0, val, addr;
+	unsigned int len32 = len >> 2;
+	int i;
+
+	addr = mem_access_addr | PE_MEM_ACCESS_WRITE |
+		PE_MEM_ACCESS_BYTE_ENABLE(0, 4);
+
+	for (i = 0; i < len32; i++, offset += 4, src += 4) {
+		val = *(u32 *)src;
+		writel(cpu_to_be32(val), pe[id].mem_access_wdata);
+		writel(addr + offset, pe[id].mem_access_addr);
+	}
+
+	len = (len & 0x3);
+	if (len) {
+		val = 0;
+
+		addr = (mem_access_addr | PE_MEM_ACCESS_WRITE |
+			PE_MEM_ACCESS_BYTE_ENABLE(0, len)) + offset;
+
+		for (i = 0; i < len; i++, src++)
+			val |= (*(u8 *)src) << (8 * i);
+
+		writel(cpu_to_be32(val), pe[id].mem_access_wdata);
+		writel(addr, pe[id].mem_access_addr);
+	}
+}
+
+/* Writes a buffer to PE internal data memory (DMEM) from the host
+ * through indirect access registers.
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., UTIL_ID)
+ * @param[in] src		Buffer source address
+ * @param[in] dst		DMEM destination address (must be 32bit
+ * aligned)
+ * @param[in] len		Number of bytes to copy
+ */
+void pe_dmem_memcpy_to32(int id, u32 dst, const void *src, unsigned int len)
+{
+	pe_mem_memcpy_to32(id, pe[id].dmem_base_addr | dst |
+				PE_MEM_ACCESS_DMEM, src, len);
+}
+
+/* Writes a buffer to PE internal program memory (PMEM) from the host
+ * through indirect access registers.
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., TMU3_ID)
+ * @param[in] src		Buffer source address
+ * @param[in] dst		PMEM destination address (must be 32bit
+ * aligned)
+ * @param[in] len		Number of bytes to copy
+ */
+void pe_pmem_memcpy_to32(int id, u32 dst, const void *src, unsigned int len)
+{
+	pe_mem_memcpy_to32(id, pe[id].pmem_base_addr | (dst & (pe[id].pmem_size
+				- 1)) | PE_MEM_ACCESS_IMEM, src, len);
+}
+
+/* Reads PE internal program memory (IMEM) from the host
+ * through indirect access registers.
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., TMU3_ID)
+ * @param[in] addr		PMEM read address (must be aligned on size)
+ * @param[in] size		Number of bytes to read (maximum 4, must not
+ * cross 32bit boundaries)
+ * @return			the data read (in PE endianness, i.e BE).
+ */
+u32 pe_pmem_read(int id, u32 addr, u8 size)
+{
+	u32 offset = addr & 0x3;
+	u32 mask = 0xffffffff >> ((4 - size) << 3);
+	u32 val;
+
+	addr = pe[id].pmem_base_addr | ((addr & ~0x3) & (pe[id].pmem_size - 1))
+		| PE_MEM_ACCESS_IMEM | PE_MEM_ACCESS_BYTE_ENABLE(offset, size);
+
+	writel(addr, pe[id].mem_access_addr);
+	val = be32_to_cpu(readl(pe[id].mem_access_rdata));
+
+	return (val >> (offset << 3)) & mask;
+}
+
+/* Writes PE internal data memory (DMEM) from the host
+ * through indirect access registers.
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., UTIL_ID)
+ * @param[in] addr		DMEM write address (must be aligned on size)
+ * @param[in] val		Value to write (in PE endianness, i.e BE)
+ * @param[in] size		Number of bytes to write (maximum 4, must not
+ * cross 32bit boundaries)
+ */
+void pe_dmem_write(int id, u32 val, u32 addr, u8 size)
+{
+	u32 offset = addr & 0x3;
+
+	addr = pe[id].dmem_base_addr | (addr & ~0x3) | PE_MEM_ACCESS_WRITE |
+		PE_MEM_ACCESS_DMEM | PE_MEM_ACCESS_BYTE_ENABLE(offset, size);
+
+	/* Indirect access interface is byte swapping data being written */
+	writel(cpu_to_be32(val << (offset << 3)), pe[id].mem_access_wdata);
+	writel(addr, pe[id].mem_access_addr);
+}
+
+/* Reads PE internal data memory (DMEM) from the host
+ * through indirect access registers.
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., UTIL_ID)
+ * @param[in] addr		DMEM read address (must be aligned on size)
+ * @param[in] size		Number of bytes to read (maximum 4, must not
+ * cross 32bit boundaries)
+ * @return			the data read (in PE endianness, i.e BE).
+ */
+u32 pe_dmem_read(int id, u32 addr, u8 size)
+{
+	u32 offset = addr & 0x3;
+	u32 mask = 0xffffffff >> ((4 - size) << 3);
+	u32 val;
+
+	addr = pe[id].dmem_base_addr | (addr & ~0x3) | PE_MEM_ACCESS_DMEM |
+			PE_MEM_ACCESS_BYTE_ENABLE(offset, size);
+
+	writel(addr, pe[id].mem_access_addr);
+
+	/* Indirect access interface is byte swapping data being read */
+	val = be32_to_cpu(readl(pe[id].mem_access_rdata));
+
+	return (val >> (offset << 3)) & mask;
+}
+
+/* This function is used to write to CLASS internal bus peripherals (ccu,
+ * pe-lem) from the host
+ * through indirect access registers.
+ * @param[in]	val	value to write
+ * @param[in]	addr	Address to write to (must be aligned on size)
+ * @param[in]	size	Number of bytes to write (1, 2 or 4)
+ *
+ */
+void class_bus_write(u32 val, u32 addr, u8 size)
+{
+	u32 offset = addr & 0x3;
+
+	writel((addr & CLASS_BUS_ACCESS_BASE_MASK), CLASS_BUS_ACCESS_BASE);
+
+	addr = (addr & ~CLASS_BUS_ACCESS_BASE_MASK) | PE_MEM_ACCESS_WRITE |
+			(size << 24);
+
+	writel(cpu_to_be32(val << (offset << 3)), CLASS_BUS_ACCESS_WDATA);
+	writel(addr, CLASS_BUS_ACCESS_ADDR);
+}
+
+/* Reads from CLASS internal bus peripherals (ccu, pe-lem) from the host
+ * through indirect access registers.
+ * @param[in] addr	Address to read from (must be aligned on size)
+ * @param[in] size	Number of bytes to read (1, 2 or 4)
+ * @return		the read data
+ *
+ */
+u32 class_bus_read(u32 addr, u8 size)
+{
+	u32 offset = addr & 0x3;
+	u32 mask = 0xffffffff >> ((4 - size) << 3);
+	u32 val;
+
+	writel((addr & CLASS_BUS_ACCESS_BASE_MASK), CLASS_BUS_ACCESS_BASE);
+
+	addr = (addr & ~CLASS_BUS_ACCESS_BASE_MASK) | (size << 24);
+
+	writel(addr, CLASS_BUS_ACCESS_ADDR);
+	val = be32_to_cpu(readl(CLASS_BUS_ACCESS_RDATA));
+
+	return (val >> (offset << 3)) & mask;
+}
+
+/* Writes data to the cluster memory (PE_LMEM)
+ * @param[in] dst	PE LMEM destination address (must be 32bit aligned)
+ * @param[in] src	Buffer source address
+ * @param[in] len	Number of bytes to copy
+ */
+void class_pe_lmem_memcpy_to32(u32 dst, const void *src, unsigned int len)
+{
+	u32 len32 = len >> 2;
+	int i;
+
+	for (i = 0; i < len32; i++, src += 4, dst += 4)
+		class_bus_write(*(u32 *)src, dst, 4);
+
+	if (len & 0x2) {
+		class_bus_write(*(u16 *)src, dst, 2);
+		src += 2;
+		dst += 2;
+	}
+
+	if (len & 0x1) {
+		class_bus_write(*(u8 *)src, dst, 1);
+		src++;
+		dst++;
+	}
+}
+
+/* Writes value to the cluster memory (PE_LMEM)
+ * @param[in] dst	PE LMEM destination address (must be 32bit aligned)
+ * @param[in] val	Value to write
+ * @param[in] len	Number of bytes to write
+ */
+void class_pe_lmem_memset(u32 dst, int val, unsigned int len)
+{
+	u32 len32 = len >> 2;
+	int i;
+
+	val = val | (val << 8) | (val << 16) | (val << 24);
+
+	for (i = 0; i < len32; i++, dst += 4)
+		class_bus_write(val, dst, 4);
+
+	if (len & 0x2) {
+		class_bus_write(val, dst, 2);
+		dst += 2;
+	}
+
+	if (len & 0x1) {
+		class_bus_write(val, dst, 1);
+		dst++;
+	}
+}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+
+/* Writes UTIL program memory (DDR) from the host.
+ *
+ * @param[in] addr	Address to write (virtual, must be aligned on size)
+ * @param[in] val		Value to write (in PE endianness, i.e BE)
+ * @param[in] size		Number of bytes to write (2 or 4)
+ */
+static void util_pmem_write(u32 val, void *addr, u8 size)
+{
+	void *addr64 = (void *)((unsigned long)addr & ~0x7);
+	unsigned long off = 8 - ((unsigned long)addr & 0x7) - size;
+
+	/*
+	 * IMEM should  be loaded as a 64bit swapped value in a 64bit aligned
+	 * location
+	 */
+	if (size == 4)
+		writel(be32_to_cpu(val), addr64 + off);
+	else
+		writew(be16_to_cpu((u16)val), addr64 + off);
+}
+
+/* Writes a buffer to UTIL program memory (DDR) from the host.
+ *
+ * @param[in] dst	Address to write (virtual, must be at least 16bit
+ * aligned)
+ * @param[in] src	Buffer to write (in PE endianness, i.e BE, must have
+ * same alignment as dst)
+ * @param[in] len	Number of bytes to write (must be at least 16bit
+ * aligned)
+ */
+static void util_pmem_memcpy(void *dst, const void *src, unsigned int len)
+{
+	unsigned int len32;
+	int i;
+
+	if ((unsigned long)src & 0x2) {
+		util_pmem_write(*(u16 *)src, dst, 2);
+		src += 2;
+		dst += 2;
+		len -= 2;
+	}
+
+	len32 = len >> 2;
+
+	for (i = 0; i < len32; i++, dst += 4, src += 4)
+		util_pmem_write(*(u32 *)src, dst, 4);
+
+	if (len & 0x2)
+		util_pmem_write(*(u16 *)src, dst, len & 0x2);
+}
+#endif
+
+/* Loads an elf section into pmem
+ * Code needs to be at least 16bit aligned and only PROGBITS sections are
+ * supported
+ *
+ * @param[in] id	PE identification (CLASS0_ID, ..., TMU0_ID, ...,
+ * TMU3_ID)
+ * @param[in] data	pointer to the elf firmware
+ * @param[in] shdr	pointer to the elf section header
+ *
+ */
+static int pe_load_pmem_section(int id, const void *data,
+				struct elf32_shdr *shdr)
+{
+	u32 offset = be32_to_cpu(shdr->sh_offset);
+	u32 addr = be32_to_cpu(shdr->sh_addr);
+	u32 size = be32_to_cpu(shdr->sh_size);
+	u32 type = be32_to_cpu(shdr->sh_type);
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	if (id == UTIL_ID) {
+		pr_err("%s: unsupported pmem section for UTIL\n",
+		       __func__);
+		return -EINVAL;
+	}
+#endif
+
+	if (((unsigned long)(data + offset) & 0x3) != (addr & 0x3)) {
+		pr_err(
+			"%s: load address(%x) and elf file address(%lx) don't have the same alignment\n"
+			, __func__, addr, (unsigned long)data + offset);
+
+		return -EINVAL;
+	}
+
+	if (addr & 0x1) {
+		pr_err("%s: load address(%x) is not 16bit aligned\n",
+		       __func__, addr);
+		return -EINVAL;
+	}
+
+	if (size & 0x1) {
+		pr_err("%s: load size(%x) is not 16bit aligned\n",
+		       __func__, size);
+		return -EINVAL;
+	}
+
+	switch (type) {
+	case SHT_PROGBITS:
+		pe_pmem_memcpy_to32(id, addr, data + offset, size);
+
+		break;
+
+	default:
+		pr_err("%s: unsupported section type(%x)\n", __func__,
+		       type);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* Loads an elf section into dmem
+ * Data needs to be at least 32bit aligned, NOBITS sections are correctly
+ * initialized to 0
+ *
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., UTIL_ID)
+ * @param[in] data		pointer to the elf firmware
+ * @param[in] shdr		pointer to the elf section header
+ *
+ */
+static int pe_load_dmem_section(int id, const void *data,
+				struct elf32_shdr *shdr)
+{
+	u32 offset = be32_to_cpu(shdr->sh_offset);
+	u32 addr = be32_to_cpu(shdr->sh_addr);
+	u32 size = be32_to_cpu(shdr->sh_size);
+	u32 type = be32_to_cpu(shdr->sh_type);
+	u32 size32 = size >> 2;
+	int i;
+
+	if (((unsigned long)(data + offset) & 0x3) != (addr & 0x3)) {
+		pr_err(
+			"%s: load address(%x) and elf file address(%lx) don't have the same alignment\n",
+			__func__, addr, (unsigned long)data + offset);
+
+		return -EINVAL;
+	}
+
+	if (addr & 0x3) {
+		pr_err("%s: load address(%x) is not 32bit aligned\n",
+		       __func__, addr);
+		return -EINVAL;
+	}
+
+	switch (type) {
+	case SHT_PROGBITS:
+		pe_dmem_memcpy_to32(id, addr, data + offset, size);
+		break;
+
+	case SHT_NOBITS:
+		for (i = 0; i < size32; i++, addr += 4)
+			pe_dmem_write(id, 0, addr, 4);
+
+		if (size & 0x3)
+			pe_dmem_write(id, 0, addr, size & 0x3);
+
+		break;
+
+	default:
+		pr_err("%s: unsupported section type(%x)\n", __func__,
+		       type);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* Loads an elf section into DDR
+ * Data needs to be at least 32bit aligned, NOBITS sections are correctly
+ * initialized to 0
+ *
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., UTIL_ID)
+ * @param[in] data		pointer to the elf firmware
+ * @param[in] shdr		pointer to the elf section header
+ *
+ */
+static int pe_load_ddr_section(int id, const void *data,
+			       struct elf32_shdr *shdr,
+			       struct device *dev) {
+	u32 offset = be32_to_cpu(shdr->sh_offset);
+	u32 addr = be32_to_cpu(shdr->sh_addr);
+	u32 size = be32_to_cpu(shdr->sh_size);
+	u32 type = be32_to_cpu(shdr->sh_type);
+	u32 flags = be32_to_cpu(shdr->sh_flags);
+
+	switch (type) {
+	case SHT_PROGBITS:
+		if (flags & SHF_EXECINSTR) {
+			if (id <= CLASS_MAX_ID) {
+				/* DO the loading only once in DDR */
+				if (id == CLASS0_ID) {
+					pr_err(
+						"%s: load address(%x) and elf file address(%lx) rcvd\n",
+						__func__, addr,
+						(unsigned long)data + offset);
+					if (((unsigned long)(data + offset)
+						& 0x3) != (addr & 0x3)) {
+						pr_err(
+							"%s: load address(%x) and elf file address(%lx) don't have the same alignment\n"
+							, __func__, addr,
+						(unsigned long)data + offset);
+
+						return -EINVAL;
+					}
+
+					if (addr & 0x1) {
+						pr_err(
+							"%s: load address(%x) is not 16bit aligned\n"
+							, __func__, addr);
+						return -EINVAL;
+					}
+
+					if (size & 0x1) {
+						pr_err(
+							"%s: load length(%x) is not 16bit aligned\n"
+							, __func__, size);
+						return -EINVAL;
+					}
+					memcpy(DDR_PHYS_TO_VIRT(
+						DDR_PFE_TO_PHYS(addr)),
+						data + offset, size);
+				}
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+			} else if (id == UTIL_ID) {
+				if (((unsigned long)(data + offset) & 0x3)
+					!= (addr & 0x3)) {
+					pr_err(
+						"%s: load address(%x) and elf file address(%lx) don't have the same alignment\n"
+						, __func__, addr,
+						(unsigned long)data + offset);
+
+					return -EINVAL;
+				}
+
+				if (addr & 0x1) {
+					pr_err(
+						"%s: load address(%x) is not 16bit aligned\n"
+						, __func__, addr);
+					return -EINVAL;
+				}
+
+				if (size & 0x1) {
+					pr_err(
+						"%s: load length(%x) is not 16bit aligned\n"
+						, __func__, size);
+					return -EINVAL;
+				}
+
+				util_pmem_memcpy(DDR_PHYS_TO_VIRT(
+							DDR_PFE_TO_PHYS(addr)),
+							data + offset, size);
+			}
+#endif
+			} else {
+				pr_err(
+					"%s: unsupported ddr section type(%x) for PE(%d)\n"
+						, __func__, type, id);
+				return -EINVAL;
+			}
+
+		} else {
+			memcpy(DDR_PHYS_TO_VIRT(DDR_PFE_TO_PHYS(addr)), data
+				+ offset, size);
+		}
+
+		break;
+
+	case SHT_NOBITS:
+		memset(DDR_PHYS_TO_VIRT(DDR_PFE_TO_PHYS(addr)), 0, size);
+
+		break;
+
+	default:
+		pr_err("%s: unsupported section type(%x)\n", __func__,
+		       type);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* Loads an elf section into pe lmem
+ * Data needs to be at least 32bit aligned, NOBITS sections are correctly
+ * initialized to 0
+ *
+ * @param[in] id		PE identification (CLASS0_ID,..., CLASS5_ID)
+ * @param[in] data		pointer to the elf firmware
+ * @param[in] shdr		pointer to the elf section header
+ *
+ */
+static int pe_load_pe_lmem_section(int id, const void *data,
+				   struct elf32_shdr *shdr)
+{
+	u32 offset = be32_to_cpu(shdr->sh_offset);
+	u32 addr = be32_to_cpu(shdr->sh_addr);
+	u32 size = be32_to_cpu(shdr->sh_size);
+	u32 type = be32_to_cpu(shdr->sh_type);
+
+	if (id > CLASS_MAX_ID) {
+		pr_err(
+			"%s: unsupported pe-lmem section type(%x) for PE(%d)\n",
+			 __func__, type, id);
+		return -EINVAL;
+	}
+
+	if (((unsigned long)(data + offset) & 0x3) != (addr & 0x3)) {
+		pr_err(
+			"%s: load address(%x) and elf file address(%lx) don't have the same alignment\n",
+			__func__, addr, (unsigned long)data + offset);
+
+		return -EINVAL;
+	}
+
+	if (addr & 0x3) {
+		pr_err("%s: load address(%x) is not 32bit aligned\n",
+		       __func__, addr);
+		return -EINVAL;
+	}
+
+	switch (type) {
+	case SHT_PROGBITS:
+		class_pe_lmem_memcpy_to32(addr, data + offset, size);
+		break;
+
+	case SHT_NOBITS:
+		class_pe_lmem_memset(addr, 0, size);
+		break;
+
+	default:
+		pr_err("%s: unsupported section type(%x)\n", __func__,
+		       type);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* Loads an elf section into a PE
+ * For now only supports loading a section to dmem (all PE's), pmem (class and
+ * tmu PE's),
+ * DDDR (util PE code)
+ *
+ * @param[in] id		PE identification (CLASS0_ID, ..., TMU0_ID,
+ * ..., UTIL_ID)
+ * @param[in] data		pointer to the elf firmware
+ * @param[in] shdr		pointer to the elf section header
+ *
+ */
+int pe_load_elf_section(int id, const void *data, struct elf32_shdr *shdr,
+			struct device *dev) {
+	u32 addr = be32_to_cpu(shdr->sh_addr);
+	u32 size = be32_to_cpu(shdr->sh_size);
+
+	if (IS_DMEM(addr, size))
+		return pe_load_dmem_section(id, data, shdr);
+	else if (IS_PMEM(addr, size))
+		return pe_load_pmem_section(id, data, shdr);
+	else if (IS_PFE_LMEM(addr, size))
+		return 0;
+	else if (IS_PHYS_DDR(addr, size))
+		return pe_load_ddr_section(id, data, shdr, dev);
+	else if (IS_PE_LMEM(addr, size))
+		return pe_load_pe_lmem_section(id, data, shdr);
+
+	pr_err("%s: unsupported memory range(%x)\n", __func__,
+	       addr);
+	return 0;
+}
+
+/**************************** BMU ***************************/
+
+/* Initializes a BMU block.
+ * @param[in] base	BMU block base address
+ * @param[in] cfg	BMU configuration
+ */
+void bmu_init(void *base, struct BMU_CFG *cfg)
+{
+	bmu_disable(base);
+
+	bmu_set_config(base, cfg);
+
+	bmu_reset(base);
+}
+
+/* Resets a BMU block.
+ * @param[in] base	BMU block base address
+ */
+void bmu_reset(void *base)
+{
+	writel(CORE_SW_RESET, base + BMU_CTRL);
+
+	/* Wait for self clear */
+	while (readl(base + BMU_CTRL) & CORE_SW_RESET)
+		;
+}
+
+/* Enabled a BMU block.
+ * @param[in] base	BMU block base address
+ */
+void bmu_enable(void *base)
+{
+	writel(CORE_ENABLE, base + BMU_CTRL);
+}
+
+/* Disables a BMU block.
+ * @param[in] base	BMU block base address
+ */
+void bmu_disable(void *base)
+{
+	writel(CORE_DISABLE, base + BMU_CTRL);
+}
+
+/* Sets the configuration of a BMU block.
+ * @param[in] base	BMU block base address
+ * @param[in] cfg	BMU configuration
+ */
+void bmu_set_config(void *base, struct BMU_CFG *cfg)
+{
+	writel(cfg->baseaddr, base + BMU_UCAST_BASE_ADDR);
+	writel(cfg->count & 0xffff, base + BMU_UCAST_CONFIG);
+	writel(cfg->size & 0xffff, base + BMU_BUF_SIZE);
+
+	/* Interrupts are never used */
+	writel(cfg->low_watermark, base + BMU_LOW_WATERMARK);
+	writel(cfg->high_watermark, base + BMU_HIGH_WATERMARK);
+	writel(0x0, base + BMU_INT_ENABLE);
+}
+
+/**************************** MTIP GEMAC ***************************/
+
+/* Enable Rx Checksum Engine. With this enabled, Frame with bad IP,
+ *   TCP or UDP checksums are discarded
+ *
+ * @param[in] base	GEMAC base address.
+ */
+void gemac_enable_rx_checksum_offload(void *base)
+{
+	/*Do not find configuration to do this */
+}
+
+/* Disable Rx Checksum Engine.
+ *
+ * @param[in] base	GEMAC base address.
+ */
+void gemac_disable_rx_checksum_offload(void *base)
+{
+	/*Do not find configuration to do this */
+}
+
+/* GEMAC set speed.
+ * @param[in] base	GEMAC base address
+ * @param[in] speed	GEMAC speed (10, 100 or 1000 Mbps)
+ */
+void gemac_set_speed(void *base, enum mac_speed gem_speed)
+{
+	u32 ecr = readl(base + EMAC_ECNTRL_REG) & ~EMAC_ECNTRL_SPEED;
+	u32 rcr = readl(base + EMAC_RCNTRL_REG) & ~EMAC_RCNTRL_RMII_10T;
+
+	switch (gem_speed) {
+	case SPEED_10M:
+			rcr |= EMAC_RCNTRL_RMII_10T;
+			break;
+
+	case SPEED_1000M:
+			ecr |= EMAC_ECNTRL_SPEED;
+			break;
+
+	case SPEED_100M:
+	default:
+			/*It is in 100M mode */
+			break;
+	}
+	writel(ecr, (base + EMAC_ECNTRL_REG));
+	writel(rcr, (base + EMAC_RCNTRL_REG));
+}
+
+/* GEMAC set duplex.
+ * @param[in] base	GEMAC base address
+ * @param[in] duplex	GEMAC duplex mode (Full, Half)
+ */
+void gemac_set_duplex(void *base, int duplex)
+{
+	if (duplex == DUPLEX_HALF) {
+		writel(readl(base + EMAC_TCNTRL_REG) & ~EMAC_TCNTRL_FDEN, base
+			+ EMAC_TCNTRL_REG);
+		writel(readl(base + EMAC_RCNTRL_REG) | EMAC_RCNTRL_DRT, (base
+			+ EMAC_RCNTRL_REG));
+	} else{
+		writel(readl(base + EMAC_TCNTRL_REG) | EMAC_TCNTRL_FDEN, base
+			+ EMAC_TCNTRL_REG);
+		writel(readl(base + EMAC_RCNTRL_REG) & ~EMAC_RCNTRL_DRT, (base
+			+ EMAC_RCNTRL_REG));
+	}
+}
+
+/* GEMAC set mode.
+ * @param[in] base	GEMAC base address
+ * @param[in] mode	GEMAC operation mode (MII, RMII, RGMII, SGMII)
+ */
+void gemac_set_mode(void *base, int mode)
+{
+	u32 val = readl(base + EMAC_RCNTRL_REG);
+
+	/*Remove loopbank*/
+	val &= ~EMAC_RCNTRL_LOOP;
+
+	/*Enable flow control and MII mode*/
+	val |= (EMAC_RCNTRL_FCE | EMAC_RCNTRL_MII_MODE);
+
+	writel(val, base + EMAC_RCNTRL_REG);
+}
+
+/* GEMAC enable function.
+ * @param[in] base	GEMAC base address
+ */
+void gemac_enable(void *base)
+{
+	writel(readl(base + EMAC_ECNTRL_REG) | EMAC_ECNTRL_ETHER_EN, base +
+		EMAC_ECNTRL_REG);
+}
+
+/* GEMAC disable function.
+ * @param[in] base	GEMAC base address
+ */
+void gemac_disable(void *base)
+{
+	writel(readl(base + EMAC_ECNTRL_REG) & ~EMAC_ECNTRL_ETHER_EN, base +
+		EMAC_ECNTRL_REG);
+}
+
+/* GEMAC TX disable function.
+ * @param[in] base	GEMAC base address
+ */
+void gemac_tx_disable(void *base)
+{
+	writel(readl(base + EMAC_TCNTRL_REG) | EMAC_TCNTRL_GTS, base +
+		EMAC_TCNTRL_REG);
+}
+
+void gemac_tx_enable(void *base)
+{
+	writel(readl(base + EMAC_TCNTRL_REG) & ~EMAC_TCNTRL_GTS, base +
+			EMAC_TCNTRL_REG);
+}
+
+/* Sets the hash register of the MAC.
+ * This register is used for matching unicast and multicast frames.
+ *
+ * @param[in] base	GEMAC base address.
+ * @param[in] hash	64-bit hash to be configured.
+ */
+void gemac_set_hash(void *base, struct pfe_mac_addr *hash)
+{
+	writel(hash->bottom,  base + EMAC_GALR);
+	writel(hash->top, base + EMAC_GAUR);
+}
+
+void gemac_set_laddrN(void *base, struct pfe_mac_addr *address,
+		      unsigned int entry_index)
+{
+	if ((entry_index < 1) || (entry_index > EMAC_SPEC_ADDR_MAX))
+		return;
+
+	entry_index = entry_index - 1;
+	if (entry_index < 1) {
+		writel(htonl(address->bottom),  base + EMAC_PHY_ADDR_LOW);
+		writel((htonl(address->top) | 0x8808), base +
+			EMAC_PHY_ADDR_HIGH);
+	} else {
+		writel(htonl(address->bottom),  base + ((entry_index - 1) * 8)
+			+ EMAC_SMAC_0_0);
+		writel((htonl(address->top) | 0x8808), base + ((entry_index -
+			1) * 8) + EMAC_SMAC_0_1);
+	}
+}
+
+void gemac_clear_laddrN(void *base, unsigned int entry_index)
+{
+	if ((entry_index < 1) || (entry_index > EMAC_SPEC_ADDR_MAX))
+		return;
+
+	entry_index = entry_index - 1;
+	if (entry_index < 1) {
+		writel(0, base + EMAC_PHY_ADDR_LOW);
+		writel(0, base + EMAC_PHY_ADDR_HIGH);
+	} else {
+		writel(0,  base + ((entry_index - 1) * 8) + EMAC_SMAC_0_0);
+		writel(0, base + ((entry_index - 1) * 8) + EMAC_SMAC_0_1);
+	}
+}
+
+/* Set the loopback mode of the MAC.  This can be either no loopback for
+ * normal operation, local loopback through MAC internal loopback module or PHY
+ *   loopback for external loopback through a PHY.  This asserts the external
+ * loop pin.
+ *
+ * @param[in] base	GEMAC base address.
+ * @param[in] gem_loop	Loopback mode to be enabled. LB_LOCAL - MAC
+ * Loopback,
+ *			LB_EXT - PHY Loopback.
+ */
+void gemac_set_loop(void *base, enum mac_loop gem_loop)
+{
+	pr_info("%s()\n", __func__);
+	writel(readl(base + EMAC_RCNTRL_REG) | EMAC_RCNTRL_LOOP, (base +
+		EMAC_RCNTRL_REG));
+}
+
+/* GEMAC allow frames
+ * @param[in] base	GEMAC base address
+ */
+void gemac_enable_copy_all(void *base)
+{
+	writel(readl(base + EMAC_RCNTRL_REG) | EMAC_RCNTRL_PROM, (base +
+		EMAC_RCNTRL_REG));
+}
+
+/* GEMAC do not allow frames
+ * @param[in] base	GEMAC base address
+ */
+void gemac_disable_copy_all(void *base)
+{
+	writel(readl(base + EMAC_RCNTRL_REG) & ~EMAC_RCNTRL_PROM, (base +
+		EMAC_RCNTRL_REG));
+}
+
+/* GEMAC allow broadcast function.
+ * @param[in] base	GEMAC base address
+ */
+void gemac_allow_broadcast(void *base)
+{
+	writel(readl(base + EMAC_RCNTRL_REG) & ~EMAC_RCNTRL_BC_REJ, base +
+		EMAC_RCNTRL_REG);
+}
+
+/* GEMAC no broadcast function.
+ * @param[in] base	GEMAC base address
+ */
+void gemac_no_broadcast(void *base)
+{
+	writel(readl(base + EMAC_RCNTRL_REG) | EMAC_RCNTRL_BC_REJ, base +
+		EMAC_RCNTRL_REG);
+}
+
+/* GEMAC enable 1536 rx function.
+ * @param[in]	base	GEMAC base address
+ */
+void gemac_enable_1536_rx(void *base)
+{
+	/* Set 1536 as Maximum frame length */
+	writel(readl(base + EMAC_RCNTRL_REG) | (1536 << 16), base +
+		EMAC_RCNTRL_REG);
+}
+
+/* GEMAC enable jumbo function.
+ * @param[in]	base	GEMAC base address
+ */
+void gemac_enable_rx_jmb(void *base)
+{
+	writel(readl(base + EMAC_RCNTRL_REG) | (JUMBO_FRAME_SIZE << 16), base
+		+ EMAC_RCNTRL_REG);
+}
+
+/* GEMAC enable stacked vlan function.
+ * @param[in]	base	GEMAC base address
+ */
+void gemac_enable_stacked_vlan(void *base)
+{
+	/* MTIP doesn't support stacked vlan */
+}
+
+/* GEMAC enable pause rx function.
+ * @param[in] base	GEMAC base address
+ */
+void gemac_enable_pause_rx(void *base)
+{
+	writel(readl(base + EMAC_RCNTRL_REG) | EMAC_RCNTRL_FCE,
+	       base + EMAC_RCNTRL_REG);
+}
+
+/* GEMAC disable pause rx function.
+ * @param[in] base	GEMAC base address
+ */
+void gemac_disable_pause_rx(void *base)
+{
+	writel(readl(base + EMAC_RCNTRL_REG) & ~EMAC_RCNTRL_FCE,
+	       base + EMAC_RCNTRL_REG);
+}
+
+/* GEMAC enable pause tx function.
+ * @param[in] base GEMAC base address
+ */
+void gemac_enable_pause_tx(void *base)
+{
+	writel(EMAC_RX_SECTION_EMPTY_V, base + EMAC_RX_SECTION_EMPTY);
+}
+
+/* GEMAC disable pause tx function.
+ * @param[in] base GEMAC base address
+ */
+void gemac_disable_pause_tx(void *base)
+{
+	writel(0x0, base + EMAC_RX_SECTION_EMPTY);
+}
+
+/* GEMAC wol configuration
+ * @param[in] base	GEMAC base address
+ * @param[in] wol_conf	WoL register configuration
+ */
+void gemac_set_wol(void *base, u32 wol_conf)
+{
+	u32  val = readl(base + EMAC_ECNTRL_REG);
+
+	if (wol_conf)
+		val |= (EMAC_ECNTRL_MAGIC_ENA | EMAC_ECNTRL_SLEEP);
+	else
+		val &= ~(EMAC_ECNTRL_MAGIC_ENA | EMAC_ECNTRL_SLEEP);
+	writel(val, base + EMAC_ECNTRL_REG);
+}
+
+/* Sets Gemac bus width to 64bit
+ * @param[in] base       GEMAC base address
+ * @param[in] width     gemac bus width to be set possible values are 32/64/128
+ */
+void gemac_set_bus_width(void *base, int width)
+{
+}
+
+/* Sets Gemac configuration.
+ * @param[in] base	GEMAC base address
+ * @param[in] cfg	GEMAC configuration
+ */
+void gemac_set_config(void *base, struct gemac_cfg *cfg)
+{
+	/*GEMAC config taken from VLSI */
+	writel(0x00000004, base + EMAC_TFWR_STR_FWD);
+	writel(0x00000005, base + EMAC_RX_SECTION_FULL);
+	writel(0x00003fff, base + EMAC_TRUNC_FL);
+	writel(0x00000030, base + EMAC_TX_SECTION_EMPTY);
+	writel(0x00000000, base + EMAC_MIB_CTRL_STS_REG);
+
+	gemac_set_mode(base, cfg->mode);
+
+	gemac_set_speed(base, cfg->speed);
+
+	gemac_set_duplex(base, cfg->duplex);
+}
+
+/**************************** GPI ***************************/
+
+/* Initializes a GPI block.
+ * @param[in] base	GPI base address
+ * @param[in] cfg	GPI configuration
+ */
+void gpi_init(void *base, struct gpi_cfg *cfg)
+{
+	gpi_reset(base);
+
+	gpi_disable(base);
+
+	gpi_set_config(base, cfg);
+}
+
+/* Resets a GPI block.
+ * @param[in] base	GPI base address
+ */
+void gpi_reset(void *base)
+{
+	writel(CORE_SW_RESET, base + GPI_CTRL);
+}
+
+/* Enables a GPI block.
+ * @param[in] base	GPI base address
+ */
+void gpi_enable(void *base)
+{
+	writel(CORE_ENABLE, base + GPI_CTRL);
+}
+
+/* Disables a GPI block.
+ * @param[in] base	GPI base address
+ */
+void gpi_disable(void *base)
+{
+	writel(CORE_DISABLE, base + GPI_CTRL);
+}
+
+/* Sets the configuration of a GPI block.
+ * @param[in] base	GPI base address
+ * @param[in] cfg	GPI configuration
+ */
+void gpi_set_config(void *base, struct gpi_cfg *cfg)
+{
+	writel(CBUS_VIRT_TO_PFE(BMU1_BASE_ADDR + BMU_ALLOC_CTRL),	base
+		+ GPI_LMEM_ALLOC_ADDR);
+	writel(CBUS_VIRT_TO_PFE(BMU1_BASE_ADDR + BMU_FREE_CTRL),	base
+		+ GPI_LMEM_FREE_ADDR);
+	writel(CBUS_VIRT_TO_PFE(BMU2_BASE_ADDR + BMU_ALLOC_CTRL),	base
+		+ GPI_DDR_ALLOC_ADDR);
+	writel(CBUS_VIRT_TO_PFE(BMU2_BASE_ADDR + BMU_FREE_CTRL),	base
+		+ GPI_DDR_FREE_ADDR);
+	writel(CBUS_VIRT_TO_PFE(CLASS_INQ_PKTPTR), base + GPI_CLASS_ADDR);
+	writel(DDR_HDR_SIZE, base + GPI_DDR_DATA_OFFSET);
+	writel(LMEM_HDR_SIZE, base + GPI_LMEM_DATA_OFFSET);
+	writel(0, base + GPI_LMEM_SEC_BUF_DATA_OFFSET);
+	writel(0, base + GPI_DDR_SEC_BUF_DATA_OFFSET);
+	writel((DDR_HDR_SIZE << 16) |	LMEM_HDR_SIZE,	base + GPI_HDR_SIZE);
+	writel((DDR_BUF_SIZE << 16) |	LMEM_BUF_SIZE,	base + GPI_BUF_SIZE);
+
+	writel(((cfg->lmem_rtry_cnt << 16) | (GPI_DDR_BUF_EN << 1) |
+		GPI_LMEM_BUF_EN), base + GPI_RX_CONFIG);
+	writel(cfg->tmlf_txthres, base + GPI_TMLF_TX);
+	writel(cfg->aseq_len,	base + GPI_DTX_ASEQ);
+	writel(1, base + GPI_TOE_CHKSUM_EN);
+
+	if (cfg->mtip_pause_reg) {
+		writel(cfg->mtip_pause_reg, base + GPI_CSR_MTIP_PAUSE_REG);
+		writel(EGPI_PAUSE_TIME, base + GPI_TX_PAUSE_TIME);
+	}
+}
+
+/**************************** CLASSIFIER ***************************/
+
+/* Initializes CLASSIFIER block.
+ * @param[in] cfg	CLASSIFIER configuration
+ */
+void class_init(struct class_cfg *cfg)
+{
+	class_reset();
+
+	class_disable();
+
+	class_set_config(cfg);
+}
+
+/* Resets CLASSIFIER block.
+ *
+ */
+void class_reset(void)
+{
+	writel(CORE_SW_RESET, CLASS_TX_CTRL);
+}
+
+/* Enables all CLASS-PE's cores.
+ *
+ */
+void class_enable(void)
+{
+	writel(CORE_ENABLE, CLASS_TX_CTRL);
+}
+
+/* Disables all CLASS-PE's cores.
+ *
+ */
+void class_disable(void)
+{
+	writel(CORE_DISABLE, CLASS_TX_CTRL);
+}
+
+/*
+ * Sets the configuration of the CLASSIFIER block.
+ * @param[in] cfg	CLASSIFIER configuration
+ */
+void class_set_config(struct class_cfg *cfg)
+{
+	u32 val;
+
+	/* Initialize route table */
+	if (!cfg->resume)
+		memset(DDR_PHYS_TO_VIRT(cfg->route_table_baseaddr), 0, (1 <<
+		cfg->route_table_hash_bits) * CLASS_ROUTE_SIZE);
+
+#if !defined(LS1012A_PFE_RESET_WA)
+	writel(cfg->pe_sys_clk_ratio,	CLASS_PE_SYS_CLK_RATIO);
+#endif
+
+	writel((DDR_HDR_SIZE << 16) | LMEM_HDR_SIZE,	CLASS_HDR_SIZE);
+	writel(LMEM_BUF_SIZE,				CLASS_LMEM_BUF_SIZE);
+	writel(CLASS_ROUTE_ENTRY_SIZE(CLASS_ROUTE_SIZE) |
+		CLASS_ROUTE_HASH_SIZE(cfg->route_table_hash_bits),
+		CLASS_ROUTE_HASH_ENTRY_SIZE);
+	writel(HIF_PKT_CLASS_EN | HIF_PKT_OFFSET(sizeof(struct hif_hdr)),
+	       CLASS_HIF_PARSE);
+
+	val = HASH_CRC_PORT_IP | QB2BUS_LE;
+
+#if defined(CONFIG_IP_ALIGNED)
+	val |= IP_ALIGNED;
+#endif
+
+	/*
+	 *  Class PE packet steering will only work if TOE mode, bridge fetch or
+	 * route fetch are enabled (see class/qb_fet.v). Route fetch would
+	 * trigger additional memory copies (likely from DDR because of hash
+	 * table size, which cannot be reduced because PE software still
+	 * relies on hash value computed in HW), so when not in TOE mode we
+	 * simply enable HW bridge fetch even though we don't use it.
+	 */
+	if (cfg->toe_mode)
+		val |= CLASS_TOE;
+	else
+		val |= HW_BRIDGE_FETCH;
+
+	writel(val, CLASS_ROUTE_MULTI);
+
+	writel(DDR_PHYS_TO_PFE(cfg->route_table_baseaddr),
+	       CLASS_ROUTE_TABLE_BASE);
+	writel(CLASS_PE0_RO_DM_ADDR0_VAL,		CLASS_PE0_RO_DM_ADDR0);
+	writel(CLASS_PE0_RO_DM_ADDR1_VAL,		CLASS_PE0_RO_DM_ADDR1);
+	writel(CLASS_PE0_QB_DM_ADDR0_VAL,		CLASS_PE0_QB_DM_ADDR0);
+	writel(CLASS_PE0_QB_DM_ADDR1_VAL,		CLASS_PE0_QB_DM_ADDR1);
+	writel(CBUS_VIRT_TO_PFE(TMU_PHY_INQ_PKTPTR),	CLASS_TM_INQ_ADDR);
+
+	writel(23, CLASS_AFULL_THRES);
+	writel(23, CLASS_TSQ_FIFO_THRES);
+
+	writel(24, CLASS_MAX_BUF_CNT);
+	writel(24, CLASS_TSQ_MAX_CNT);
+}
+
+/**************************** TMU ***************************/
+
+void tmu_reset(void)
+{
+	writel(SW_RESET, TMU_CTRL);
+}
+
+/* Initializes TMU block.
+ * @param[in] cfg	TMU configuration
+ */
+void tmu_init(struct tmu_cfg *cfg)
+{
+	int q, phyno;
+
+	tmu_disable(0xF);
+	mdelay(10);
+
+#if !defined(LS1012A_PFE_RESET_WA)
+	/* keep in soft reset */
+	writel(SW_RESET, TMU_CTRL);
+#endif
+	writel(0x3, TMU_SYS_GENERIC_CONTROL);
+	writel(750, TMU_INQ_WATERMARK);
+	writel(CBUS_VIRT_TO_PFE(EGPI1_BASE_ADDR +
+		GPI_INQ_PKTPTR),	TMU_PHY0_INQ_ADDR);
+	writel(CBUS_VIRT_TO_PFE(EGPI2_BASE_ADDR +
+		GPI_INQ_PKTPTR),	TMU_PHY1_INQ_ADDR);
+	writel(CBUS_VIRT_TO_PFE(HGPI_BASE_ADDR +
+		GPI_INQ_PKTPTR),	TMU_PHY3_INQ_ADDR);
+	writel(CBUS_VIRT_TO_PFE(HIF_NOCPY_RX_INQ0_PKTPTR), TMU_PHY4_INQ_ADDR);
+	writel(CBUS_VIRT_TO_PFE(UTIL_INQ_PKTPTR), TMU_PHY5_INQ_ADDR);
+	writel(CBUS_VIRT_TO_PFE(BMU2_BASE_ADDR + BMU_FREE_CTRL),
+	       TMU_BMU_INQ_ADDR);
+
+	writel(0x3FF,	TMU_TDQ0_SCH_CTRL);	/*
+						 * enabling all 10
+						 * schedulers [9:0] of each TDQ
+						 */
+	writel(0x3FF,	TMU_TDQ1_SCH_CTRL);
+	writel(0x3FF,	TMU_TDQ3_SCH_CTRL);
+
+#if !defined(LS1012A_PFE_RESET_WA)
+	writel(cfg->pe_sys_clk_ratio,	TMU_PE_SYS_CLK_RATIO);
+#endif
+
+#if !defined(LS1012A_PFE_RESET_WA)
+	writel(DDR_PHYS_TO_PFE(cfg->llm_base_addr),	TMU_LLM_BASE_ADDR);
+	/* Extra packet pointers will be stored from this address onwards */
+
+	writel(cfg->llm_queue_len,	TMU_LLM_QUE_LEN);
+	writel(5,			TMU_TDQ_IIFG_CFG);
+	writel(DDR_BUF_SIZE,		TMU_BMU_BUF_SIZE);
+
+	writel(0x0,			TMU_CTRL);
+
+	/* MEM init */
+	pr_info("%s: mem init\n", __func__);
+	writel(MEM_INIT,	TMU_CTRL);
+
+	while (!(readl(TMU_CTRL) & MEM_INIT_DONE))
+		;
+
+	/* LLM init */
+	pr_info("%s: lmem init\n", __func__);
+	writel(LLM_INIT,	TMU_CTRL);
+
+	while (!(readl(TMU_CTRL) & LLM_INIT_DONE))
+		;
+#endif
+	/* set up each queue for tail drop */
+	for (phyno = 0; phyno < 4; phyno++) {
+		if (phyno == 2)
+			continue;
+		for (q = 0; q < 16; q++) {
+			u32 qdepth;
+
+			writel((phyno << 8) | q, TMU_TEQ_CTRL);
+			writel(1 << 22, TMU_TEQ_QCFG); /*Enable tail drop */
+
+			if (phyno == 3)
+				qdepth = DEFAULT_TMU3_QDEPTH;
+			else
+				qdepth = (q == 0) ? DEFAULT_Q0_QDEPTH :
+						DEFAULT_MAX_QDEPTH;
+
+			/* LOG: 68855 */
+			/*
+			 * The following is a workaround for the reordered
+			 * packet and BMU2 buffer leakage issue.
+			 */
+			if (CHIP_REVISION() == 0)
+				qdepth = 31;
+
+			writel(qdepth << 18, TMU_TEQ_HW_PROB_CFG2);
+			writel(qdepth >> 14, TMU_TEQ_HW_PROB_CFG3);
+		}
+	}
+
+#ifdef CFG_LRO
+	/* Set TMU-3 queue 5 (LRO) in no-drop mode */
+	writel((3 << 8) | TMU_QUEUE_LRO, TMU_TEQ_CTRL);
+	writel(0, TMU_TEQ_QCFG);
+#endif
+
+	writel(0x05, TMU_TEQ_DISABLE_DROPCHK);
+
+	writel(0x0, TMU_CTRL);
+}
+
+/* Enables TMU-PE cores.
+ * @param[in] pe_mask	TMU PE mask
+ */
+void tmu_enable(u32 pe_mask)
+{
+	writel(readl(TMU_TX_CTRL) | (pe_mask & 0xF), TMU_TX_CTRL);
+}
+
+/* Disables TMU cores.
+ * @param[in] pe_mask	TMU PE mask
+ */
+void tmu_disable(u32 pe_mask)
+{
+	writel(readl(TMU_TX_CTRL) & ~(pe_mask & 0xF), TMU_TX_CTRL);
+}
+
+/* This will return the tmu queue status
+ * @param[in] if_id	gem interface id or TMU index
+ * @return		returns the bit mask of busy queues, zero means all
+ * queues are empty
+ */
+u32 tmu_qstatus(u32 if_id)
+{
+	return cpu_to_be32(pe_dmem_read(TMU0_ID + if_id, TMU_DM_PESTATUS +
+		offsetof(struct pe_status, tmu_qstatus), 4));
+}
+
+u32 tmu_pkts_processed(u32 if_id)
+{
+	return cpu_to_be32(pe_dmem_read(TMU0_ID + if_id, TMU_DM_PESTATUS +
+		offsetof(struct pe_status, rx), 4));
+}
+
+/**************************** UTIL ***************************/
+
+/* Resets UTIL block.
+ */
+void util_reset(void)
+{
+	writel(CORE_SW_RESET, UTIL_TX_CTRL);
+}
+
+/* Initializes UTIL block.
+ * @param[in] cfg	UTIL configuration
+ */
+void util_init(struct util_cfg *cfg)
+{
+	writel(cfg->pe_sys_clk_ratio,   UTIL_PE_SYS_CLK_RATIO);
+}
+
+/* Enables UTIL-PE core.
+ *
+ */
+void util_enable(void)
+{
+	writel(CORE_ENABLE, UTIL_TX_CTRL);
+}
+
+/* Disables UTIL-PE core.
+ *
+ */
+void util_disable(void)
+{
+	writel(CORE_DISABLE, UTIL_TX_CTRL);
+}
+
+/**************************** HIF ***************************/
+/* Initializes HIF copy block.
+ *
+ */
+void hif_init(void)
+{
+	/*Initialize HIF registers*/
+	writel((HIF_RX_POLL_CTRL_CYCLE << 16) | HIF_TX_POLL_CTRL_CYCLE,
+	       HIF_POLL_CTRL);
+}
+
+/* Enable hif tx DMA and interrupt
+ *
+ */
+void hif_tx_enable(void)
+{
+	writel(HIF_CTRL_DMA_EN, HIF_TX_CTRL);
+	writel((readl(HIF_INT_ENABLE) | HIF_INT_EN | HIF_TXPKT_INT_EN),
+	       HIF_INT_ENABLE);
+}
+
+/* Disable hif tx DMA and interrupt
+ *
+ */
+void hif_tx_disable(void)
+{
+	u32	hif_int;
+
+	writel(0, HIF_TX_CTRL);
+
+	hif_int = readl(HIF_INT_ENABLE);
+	hif_int &= HIF_TXPKT_INT_EN;
+	writel(hif_int, HIF_INT_ENABLE);
+}
+
+/* Enable hif rx DMA and interrupt
+ *
+ */
+void hif_rx_enable(void)
+{
+	hif_rx_dma_start();
+	writel((readl(HIF_INT_ENABLE) | HIF_INT_EN | HIF_RXPKT_INT_EN),
+	       HIF_INT_ENABLE);
+}
+
+/* Disable hif rx DMA and interrupt
+ *
+ */
+void hif_rx_disable(void)
+{
+	u32	hif_int;
+
+	writel(0, HIF_RX_CTRL);
+
+	hif_int = readl(HIF_INT_ENABLE);
+	hif_int &= HIF_RXPKT_INT_EN;
+	writel(hif_int, HIF_INT_ENABLE);
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_hif.c b/drivers/staging/fsl_ppfe/pfe_hif.c
new file mode 100644
index 0000000..6844595
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_hif.c
@@ -0,0 +1,1094 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/kernel.h>
+#include <linux/interrupt.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmapool.h>
+#include <linux/sched.h>
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/kthread.h>
+#include <linux/slab.h>
+
+#include <linux/io.h>
+#include <asm/irq.h>
+
+#include "pfe_mod.h"
+
+#define HIF_INT_MASK	(HIF_INT | HIF_RXPKT_INT | HIF_TXPKT_INT)
+
+unsigned char napi_first_batch;
+
+static void pfe_tx_do_cleanup(unsigned long data);
+
+static int pfe_hif_alloc_descr(struct pfe_hif *hif)
+{
+	void *addr;
+	dma_addr_t dma_addr;
+	int err = 0;
+
+	pr_info("%s\n", __func__);
+	addr = dma_alloc_coherent(pfe->dev,
+				  HIF_RX_DESC_NT * sizeof(struct hif_desc) +
+				  HIF_TX_DESC_NT * sizeof(struct hif_desc),
+				  &dma_addr, GFP_KERNEL);
+
+	if (!addr) {
+		pr_err("%s: Could not allocate buffer descriptors!\n"
+			, __func__);
+		err = -ENOMEM;
+		goto err0;
+	}
+
+	hif->descr_baseaddr_p = dma_addr;
+	hif->descr_baseaddr_v = addr;
+	hif->rx_ring_size = HIF_RX_DESC_NT;
+	hif->tx_ring_size = HIF_TX_DESC_NT;
+
+	return 0;
+
+err0:
+	return err;
+}
+
+#if defined(LS1012A_PFE_RESET_WA)
+static void pfe_hif_disable_rx_desc(struct pfe_hif *hif)
+{
+	int ii;
+	struct hif_desc	*desc = hif->rx_base;
+
+	/*Mark all descriptors as LAST_BD */
+	for (ii = 0; ii < hif->rx_ring_size; ii++) {
+		desc->ctrl |= BD_CTRL_LAST_BD;
+		desc++;
+	}
+}
+
+struct class_rx_hdr_t {
+	u32     next_ptr;       /* ptr to the start of the first DDR buffer */
+	u16     length;         /* total packet length */
+	u16     phyno;          /* input physical port number */
+	u32     status;         /* gemac status bits */
+	u32     status2;            /* reserved for software usage */
+};
+
+/* STATUS_BAD_FRAME_ERR is set for all errors (including checksums if enabled)
+ * except overflow
+ */
+#define STATUS_BAD_FRAME_ERR            BIT(16)
+#define STATUS_LENGTH_ERR               BIT(17)
+#define STATUS_CRC_ERR                  BIT(18)
+#define STATUS_TOO_SHORT_ERR            BIT(19)
+#define STATUS_TOO_LONG_ERR             BIT(20)
+#define STATUS_CODE_ERR                 BIT(21)
+#define STATUS_MC_HASH_MATCH            BIT(22)
+#define STATUS_CUMULATIVE_ARC_HIT       BIT(23)
+#define STATUS_UNICAST_HASH_MATCH       BIT(24)
+#define STATUS_IP_CHECKSUM_CORRECT      BIT(25)
+#define STATUS_TCP_CHECKSUM_CORRECT     BIT(26)
+#define STATUS_UDP_CHECKSUM_CORRECT     BIT(27)
+#define STATUS_OVERFLOW_ERR             BIT(28) /* GPI error */
+#define MIN_PKT_SIZE			64
+
+static inline void copy_to_lmem(u32 *dst, u32 *src, int len)
+{
+	int i;
+
+	for (i = 0; i < len; i += sizeof(u32))	{
+		*dst = htonl(*src);
+		dst++; src++;
+	}
+}
+
+static void send_dummy_pkt_to_hif(void)
+{
+	void *lmem_ptr, *ddr_ptr, *lmem_virt_addr;
+	u32 physaddr;
+	struct class_rx_hdr_t local_hdr;
+	static u32 dummy_pkt[] =  {
+		0x33221100, 0x2b785544, 0xd73093cb, 0x01000608,
+		0x04060008, 0x2b780200, 0xd73093cb, 0x0a01a8c0,
+		0x33221100, 0xa8c05544, 0x00000301, 0x00000000,
+		0x00000000, 0x00000000, 0x00000000, 0xbe86c51f };
+
+	ddr_ptr = (void *)((u64)readl(BMU2_BASE_ADDR + BMU_ALLOC_CTRL));
+	if (!ddr_ptr)
+		return;
+
+	lmem_ptr = (void *)((u64)readl(BMU1_BASE_ADDR + BMU_ALLOC_CTRL));
+	if (!lmem_ptr)
+		return;
+
+	pr_info("Sending a dummy pkt to HIF %p %p\n", ddr_ptr, lmem_ptr);
+	physaddr = (u32)DDR_VIRT_TO_PFE(ddr_ptr);
+
+	lmem_virt_addr = (void *)CBUS_PFE_TO_VIRT((unsigned long int)lmem_ptr);
+
+	local_hdr.phyno = htons(0); /* RX_PHY_0 */
+	local_hdr.length = htons(MIN_PKT_SIZE);
+
+	local_hdr.next_ptr = htonl((u32)physaddr);
+	/*Mark checksum is correct */
+	local_hdr.status = htonl((STATUS_IP_CHECKSUM_CORRECT |
+				STATUS_UDP_CHECKSUM_CORRECT |
+				STATUS_TCP_CHECKSUM_CORRECT |
+				STATUS_UNICAST_HASH_MATCH |
+				STATUS_CUMULATIVE_ARC_HIT));
+	copy_to_lmem((u32 *)lmem_virt_addr, (u32 *)&local_hdr,
+		     sizeof(local_hdr));
+
+	copy_to_lmem((u32 *)(lmem_virt_addr + LMEM_HDR_SIZE), (u32 *)dummy_pkt,
+		     0x40);
+
+	writel((unsigned long int)lmem_ptr, CLASS_INQ_PKTPTR);
+}
+
+void pfe_hif_rx_idle(struct pfe_hif *hif)
+{
+	int hif_stop_loop = 10;
+	u32 rx_status;
+
+	pfe_hif_disable_rx_desc(hif);
+	pr_info("Bringing hif to idle state...");
+	writel(0, HIF_INT_ENABLE);
+	/*If HIF Rx BDP is busy send a dummy packet */
+	do {
+		rx_status = readl(HIF_RX_STATUS);
+		if (rx_status & BDP_CSR_RX_DMA_ACTV)
+			send_dummy_pkt_to_hif();
+
+		usleep_range(100, 150);
+	} while (--hif_stop_loop);
+
+	if (readl(HIF_RX_STATUS) & BDP_CSR_RX_DMA_ACTV)
+		pr_info("Failed\n");
+	else
+		pr_info("Done\n");
+}
+#endif
+
+static void pfe_hif_free_descr(struct pfe_hif *hif)
+{
+	pr_info("%s\n", __func__);
+
+	dma_free_coherent(pfe->dev,
+			  hif->rx_ring_size * sizeof(struct hif_desc) +
+			  hif->tx_ring_size * sizeof(struct hif_desc),
+			  hif->descr_baseaddr_v, hif->descr_baseaddr_p);
+}
+
+void pfe_hif_desc_dump(struct pfe_hif *hif)
+{
+	struct hif_desc	*desc;
+	unsigned long desc_p;
+	int ii = 0;
+
+	pr_info("%s\n", __func__);
+
+	desc = hif->rx_base;
+	desc_p = (u32)((u64)desc - (u64)hif->descr_baseaddr_v +
+			hif->descr_baseaddr_p);
+
+	pr_info("HIF Rx desc base %p physical %x\n", desc, (u32)desc_p);
+	for (ii = 0; ii < hif->rx_ring_size; ii++) {
+		pr_info("status: %08x, ctrl: %08x, data: %08x, next: %x\n",
+			readl(&desc->status), readl(&desc->ctrl),
+			readl(&desc->data), readl(&desc->next));
+			desc++;
+	}
+
+	desc = hif->tx_base;
+	desc_p = ((u64)desc - (u64)hif->descr_baseaddr_v +
+			hif->descr_baseaddr_p);
+
+	pr_info("HIF Tx desc base %p physical %x\n", desc, (u32)desc_p);
+	for (ii = 0; ii < hif->tx_ring_size; ii++) {
+		pr_info("status: %08x, ctrl: %08x, data: %08x, next: %x\n",
+			readl(&desc->status), readl(&desc->ctrl),
+			readl(&desc->data), readl(&desc->next));
+		desc++;
+	}
+}
+
+/* pfe_hif_release_buffers */
+static void pfe_hif_release_buffers(struct pfe_hif *hif)
+{
+	struct hif_desc	*desc;
+	int i = 0;
+
+	hif->rx_base = hif->descr_baseaddr_v;
+
+	pr_info("%s\n", __func__);
+
+	/*Free Rx buffers */
+	desc = hif->rx_base;
+	for (i = 0; i < hif->rx_ring_size; i++) {
+		if (readl(&desc->data)) {
+			if ((i < hif->shm->rx_buf_pool_cnt) &&
+			    (!hif->shm->rx_buf_pool[i])) {
+				/*
+				 * dma_unmap_single(hif->dev, desc->data,
+				 * hif->rx_buf_len[i], DMA_FROM_DEVICE);
+				 */
+				dma_unmap_single(hif->dev,
+						 DDR_PFE_TO_PHYS(
+						 readl(&desc->data)),
+						 hif->rx_buf_len[i],
+						 DMA_FROM_DEVICE);
+				hif->shm->rx_buf_pool[i] = hif->rx_buf_addr[i];
+			} else {
+				pr_err("%s: buffer pool already full\n"
+					, __func__);
+			}
+		}
+
+		writel(0, &desc->data);
+		writel(0, &desc->status);
+		writel(0, &desc->ctrl);
+		desc++;
+	}
+}
+
+/*
+ * pfe_hif_init_buffers
+ * This function initializes the HIF Rx/Tx ring descriptors and
+ * initialize Rx queue with buffers.
+ */
+static int pfe_hif_init_buffers(struct pfe_hif *hif)
+{
+	struct hif_desc	*desc, *first_desc_p;
+	u32 data;
+	int i = 0;
+
+	pr_info("%s\n", __func__);
+
+	/* Check enough Rx buffers available in the shared memory */
+	if (hif->shm->rx_buf_pool_cnt < hif->rx_ring_size)
+		return -ENOMEM;
+
+	hif->rx_base = hif->descr_baseaddr_v;
+	memset(hif->rx_base, 0, hif->rx_ring_size * sizeof(struct hif_desc));
+
+	/*Initialize Rx descriptors */
+	desc = hif->rx_base;
+	first_desc_p = (struct hif_desc *)hif->descr_baseaddr_p;
+
+	for (i = 0; i < hif->rx_ring_size; i++) {
+		/* Initialize Rx buffers from the shared memory */
+
+		data = (u32)dma_map_single(hif->dev, hif->shm->rx_buf_pool[i],
+				pfe_pkt_size, DMA_FROM_DEVICE);
+		hif->rx_buf_addr[i] = hif->shm->rx_buf_pool[i];
+		hif->rx_buf_len[i] = pfe_pkt_size;
+		hif->shm->rx_buf_pool[i] = NULL;
+
+		if (likely(dma_mapping_error(hif->dev, data) == 0)) {
+			writel(DDR_PHYS_TO_PFE(data), &desc->data);
+		} else {
+			pr_err("%s : low on mem\n",  __func__);
+
+			goto err;
+		}
+
+		writel(0, &desc->status);
+
+		/*
+		 * Ensure everything else is written to DDR before
+		 * writing bd->ctrl
+		 */
+		wmb();
+
+		writel((BD_CTRL_PKT_INT_EN | BD_CTRL_LIFM
+			| BD_CTRL_DIR | BD_CTRL_DESC_EN
+			| BD_BUF_LEN(pfe_pkt_size)), &desc->ctrl);
+
+		/* Chain descriptors */
+		writel((u32)DDR_PHYS_TO_PFE(first_desc_p + i + 1), &desc->next);
+		desc++;
+	}
+
+	/* Overwrite last descriptor to chain it to first one*/
+	desc--;
+	writel((u32)DDR_PHYS_TO_PFE(first_desc_p), &desc->next);
+
+	hif->rxtoclean_index = 0;
+
+	/*Initialize Rx buffer descriptor ring base address */
+	writel(DDR_PHYS_TO_PFE(hif->descr_baseaddr_p), HIF_RX_BDP_ADDR);
+
+	hif->tx_base = hif->rx_base + hif->rx_ring_size;
+	first_desc_p = (struct hif_desc *)hif->descr_baseaddr_p +
+				hif->rx_ring_size;
+	memset(hif->tx_base, 0, hif->tx_ring_size * sizeof(struct hif_desc));
+
+	/*Initialize tx descriptors */
+	desc = hif->tx_base;
+
+	for (i = 0; i < hif->tx_ring_size; i++) {
+		/* Chain descriptors */
+		writel((u32)DDR_PHYS_TO_PFE(first_desc_p + i + 1), &desc->next);
+		writel(0, &desc->ctrl);
+		desc++;
+	}
+
+	/* Overwrite last descriptor to chain it to first one */
+	desc--;
+	writel((u32)DDR_PHYS_TO_PFE(first_desc_p), &desc->next);
+	hif->txavail = hif->tx_ring_size;
+	hif->txtosend = 0;
+	hif->txtoclean = 0;
+	hif->txtoflush = 0;
+
+	/*Initialize Tx buffer descriptor ring base address */
+	writel((u32)DDR_PHYS_TO_PFE(first_desc_p), HIF_TX_BDP_ADDR);
+
+	return 0;
+
+err:
+	pfe_hif_release_buffers(hif);
+	return -ENOMEM;
+}
+
+/*
+ * pfe_hif_client_register
+ *
+ * This function used to register a client driver with the HIF driver.
+ *
+ * Return value:
+ * 0 - on Successful registration
+ */
+static int pfe_hif_client_register(struct pfe_hif *hif, u32 client_id,
+				   struct hif_client_shm *client_shm)
+{
+	struct hif_client *client = &hif->client[client_id];
+	u32 i, cnt;
+	struct rx_queue_desc *rx_qbase;
+	struct tx_queue_desc *tx_qbase;
+	struct hif_rx_queue *rx_queue;
+	struct hif_tx_queue *tx_queue;
+	int err = 0;
+
+	pr_info("%s\n", __func__);
+
+	spin_lock_bh(&hif->tx_lock);
+
+	if (test_bit(client_id, &hif->shm->g_client_status[0])) {
+		pr_err("%s: client %d already registered\n",
+		       __func__, client_id);
+		err = -1;
+		goto unlock;
+	}
+
+	memset(client, 0, sizeof(struct hif_client));
+
+	/* Initialize client Rx queues baseaddr, size */
+
+	cnt = CLIENT_CTRL_RX_Q_CNT(client_shm->ctrl);
+	/* Check if client is requesting for more queues than supported */
+	if (cnt > HIF_CLIENT_QUEUES_MAX)
+		cnt = HIF_CLIENT_QUEUES_MAX;
+
+	client->rx_qn = cnt;
+	rx_qbase = (struct rx_queue_desc *)client_shm->rx_qbase;
+	for (i = 0; i < cnt; i++) {
+		rx_queue = &client->rx_q[i];
+		rx_queue->base = rx_qbase + i * client_shm->rx_qsize;
+		rx_queue->size = client_shm->rx_qsize;
+		rx_queue->write_idx = 0;
+	}
+
+	/* Initialize client Tx queues baseaddr, size */
+	cnt = CLIENT_CTRL_TX_Q_CNT(client_shm->ctrl);
+
+	/* Check if client is requesting for more queues than supported */
+	if (cnt > HIF_CLIENT_QUEUES_MAX)
+		cnt = HIF_CLIENT_QUEUES_MAX;
+
+	client->tx_qn = cnt;
+	tx_qbase = (struct tx_queue_desc *)client_shm->tx_qbase;
+	for (i = 0; i < cnt; i++) {
+		tx_queue = &client->tx_q[i];
+		tx_queue->base = tx_qbase + i * client_shm->tx_qsize;
+		tx_queue->size = client_shm->tx_qsize;
+		tx_queue->ack_idx = 0;
+	}
+
+	set_bit(client_id, &hif->shm->g_client_status[0]);
+
+unlock:
+	spin_unlock_bh(&hif->tx_lock);
+
+	return err;
+}
+
+/*
+ * pfe_hif_client_unregister
+ *
+ * This function used to unregister a client  from the HIF driver.
+ *
+ */
+static void pfe_hif_client_unregister(struct pfe_hif *hif, u32 client_id)
+{
+	pr_info("%s\n", __func__);
+
+	/*
+	 * Mark client as no longer available (which prevents further packet
+	 * receive for this client)
+	 */
+	spin_lock_bh(&hif->tx_lock);
+
+	if (!test_bit(client_id, &hif->shm->g_client_status[0])) {
+		pr_err("%s: client %d not registered\n", __func__,
+		       client_id);
+
+		spin_unlock_bh(&hif->tx_lock);
+		return;
+	}
+
+	clear_bit(client_id, &hif->shm->g_client_status[0]);
+
+	spin_unlock_bh(&hif->tx_lock);
+}
+
+/*
+ * client_put_rxpacket-
+ * This functions puts the Rx pkt  in the given client Rx queue.
+ * It actually swap the Rx pkt in the client Rx descriptor buffer
+ * and returns the free buffer from it.
+ *
+ * If the function returns NULL means client Rx queue is full and
+ * packet couldn't send to client queue.
+ */
+static void *client_put_rxpacket(struct hif_rx_queue *queue, void *pkt, u32 len,
+				 u32 flags, u32 client_ctrl, u32 *rem_len)
+{
+	void *free_pkt = NULL;
+	struct rx_queue_desc *desc = queue->base + queue->write_idx;
+
+	if (readl(&desc->ctrl) & CL_DESC_OWN) {
+		if (page_mode) {
+			int rem_page_size = PAGE_SIZE -
+					PRESENT_OFST_IN_PAGE(pkt);
+			int cur_pkt_size = ROUND_MIN_RX_SIZE(len +
+					pfe_pkt_headroom);
+			*rem_len = (rem_page_size - cur_pkt_size);
+			if (*rem_len) {
+				free_pkt = pkt + cur_pkt_size;
+				get_page(virt_to_page(free_pkt));
+			} else {
+				free_pkt = (void
+				*)__get_free_page(GFP_ATOMIC | GFP_DMA_PFE);
+				*rem_len = pfe_pkt_size;
+			}
+		} else {
+			free_pkt = kmalloc(PFE_BUF_SIZE, GFP_ATOMIC |
+					GFP_DMA_PFE);
+			*rem_len = PFE_BUF_SIZE - pfe_pkt_headroom;
+		}
+
+		if (free_pkt) {
+			desc->data = pkt;
+			desc->client_ctrl = client_ctrl;
+			/*
+			 * Ensure everything else is written to DDR before
+			 * writing bd->ctrl
+			 */
+			smp_wmb();
+			writel(CL_DESC_BUF_LEN(len) | flags, &desc->ctrl);
+			/* queue->write_idx = (queue->write_idx + 1)
+			 *		       & (queue->size - 1);
+			 */
+			free_pkt += pfe_pkt_headroom;
+		}
+	}
+
+	return free_pkt;
+}
+
+/*
+ * pfe_hif_rx_process-
+ * This function does pfe hif rx queue processing.
+ * Dequeue packet from Rx queue and send it to corresponding client queue
+ */
+static int pfe_hif_rx_process(struct pfe_hif *hif, int budget)
+{
+	struct hif_desc	*desc;
+	struct hif_hdr *pkt_hdr;
+	struct __hif_hdr hif_hdr;
+	void *free_buf;
+	int rtc, len, rx_processed = 0;
+	struct __hif_desc local_desc;
+	int flags;
+	unsigned int desc_p;
+	unsigned int buf_size = 0;
+
+	spin_lock_bh(&hif->lock);
+
+	rtc = hif->rxtoclean_index;
+
+	while (rx_processed < budget) {
+		desc = hif->rx_base + rtc;
+
+		__memcpy12(&local_desc, desc);
+
+		/* ACK pending Rx interrupt */
+		if (local_desc.ctrl & BD_CTRL_DESC_EN) {
+			writel(HIF_INT | HIF_RXPKT_INT, HIF_INT_SRC);
+
+			if (rx_processed == 0) {
+				if (napi_first_batch == 1) {
+					desc_p = hif->descr_baseaddr_p +
+					((unsigned long int)(desc) -
+					(unsigned long
+					int)hif->descr_baseaddr_v);
+					napi_first_batch = 0;
+				}
+			}
+
+			__memcpy12(&local_desc, desc);
+
+			if (local_desc.ctrl & BD_CTRL_DESC_EN)
+				break;
+		}
+
+		napi_first_batch = 0;
+
+#ifdef HIF_NAPI_STATS
+		hif->napi_counters[NAPI_DESC_COUNT]++;
+#endif
+		len = BD_BUF_LEN(local_desc.ctrl);
+		/*
+		 * dma_unmap_single(hif->dev, DDR_PFE_TO_PHYS(local_desc.data),
+		 * hif->rx_buf_len[rtc], DMA_FROM_DEVICE);
+		 */
+		dma_unmap_single(hif->dev, DDR_PFE_TO_PHYS(local_desc.data),
+				 hif->rx_buf_len[rtc], DMA_FROM_DEVICE);
+
+		pkt_hdr = (struct hif_hdr *)hif->rx_buf_addr[rtc];
+
+		/* Track last HIF header received */
+		if (!hif->started) {
+			hif->started = 1;
+
+			__memcpy8(&hif_hdr, pkt_hdr);
+
+			hif->qno = hif_hdr.hdr.q_num;
+			hif->client_id = hif_hdr.hdr.client_id;
+			hif->client_ctrl = (hif_hdr.hdr.client_ctrl1 << 16) |
+						hif_hdr.hdr.client_ctrl;
+			flags = CL_DESC_FIRST;
+
+		} else {
+			flags = 0;
+		}
+
+		if (local_desc.ctrl & BD_CTRL_LIFM)
+			flags |= CL_DESC_LAST;
+
+		/* Check for valid client id and still registered */
+		if ((hif->client_id >= HIF_CLIENTS_MAX) ||
+		    !(test_bit(hif->client_id,
+			&hif->shm->g_client_status[0]))) {
+			printk_ratelimited("%s: packet with invalid client id %d q_num %d\n",
+					   __func__,
+					   hif->client_id,
+					   hif->qno);
+
+			free_buf = pkt_hdr;
+
+			goto pkt_drop;
+		}
+
+		/* Check to valid queue number */
+		if (hif->client[hif->client_id].rx_qn <= hif->qno) {
+			pr_info("%s: packet with invalid queue: %d\n"
+				, __func__, hif->qno);
+			hif->qno = 0;
+		}
+
+		free_buf =
+		client_put_rxpacket(&hif->client[hif->client_id].rx_q[hif->qno],
+				    (void *)pkt_hdr, len, flags,
+			hif->client_ctrl, &buf_size);
+
+		hif_lib_indicate_client(hif->client_id, EVENT_RX_PKT_IND,
+					hif->qno);
+
+		if (unlikely(!free_buf)) {
+#ifdef HIF_NAPI_STATS
+			hif->napi_counters[NAPI_CLIENT_FULL_COUNT]++;
+#endif
+			/*
+			 * If we want to keep in polling mode to retry later,
+			 * we need to tell napi that we consumed
+			 * the full budget or we will hit a livelock scenario.
+			 * The core code keeps this napi instance
+			 * at the head of the list and none of the other
+			 * instances get to run
+			 */
+			rx_processed = budget;
+
+			if (flags & CL_DESC_FIRST)
+				hif->started = 0;
+
+			break;
+		}
+
+pkt_drop:
+		/*Fill free buffer in the descriptor */
+		hif->rx_buf_addr[rtc] = free_buf;
+		hif->rx_buf_len[rtc] = min(pfe_pkt_size, buf_size);
+		writel((DDR_PHYS_TO_PFE
+			((u32)dma_map_single(hif->dev,
+			free_buf, hif->rx_buf_len[rtc], DMA_FROM_DEVICE))),
+			&desc->data);
+		/*
+		 * Ensure everything else is written to DDR before
+		 * writing bd->ctrl
+		 */
+		wmb();
+		writel((BD_CTRL_PKT_INT_EN | BD_CTRL_LIFM | BD_CTRL_DIR |
+			BD_CTRL_DESC_EN | BD_BUF_LEN(hif->rx_buf_len[rtc])),
+			&desc->ctrl);
+
+		rtc = (rtc + 1) & (hif->rx_ring_size - 1);
+
+		if (local_desc.ctrl & BD_CTRL_LIFM) {
+			if (!(hif->client_ctrl & HIF_CTRL_RX_CONTINUED)) {
+				rx_processed++;
+
+#ifdef HIF_NAPI_STATS
+				hif->napi_counters[NAPI_PACKET_COUNT]++;
+#endif
+			}
+			hif->started = 0;
+		}
+	}
+
+	hif->rxtoclean_index = rtc;
+	spin_unlock_bh(&hif->lock);
+
+	/* we made some progress, re-start rx dma in case it stopped */
+	hif_rx_dma_start();
+
+	return rx_processed;
+}
+
+/*
+ * client_ack_txpacket-
+ * This function ack the Tx packet in the give client Tx queue by resetting
+ * ownership bit in the descriptor.
+ */
+static int client_ack_txpacket(struct pfe_hif *hif, unsigned int client_id,
+			       unsigned int q_no)
+{
+	struct hif_tx_queue *queue = &hif->client[client_id].tx_q[q_no];
+	struct tx_queue_desc *desc = queue->base + queue->ack_idx;
+
+	if (readl(&desc->ctrl) & CL_DESC_OWN) {
+		writel((readl(&desc->ctrl) & ~CL_DESC_OWN), &desc->ctrl);
+		/* queue->ack_idx = (queue->ack_idx + 1) & (queue->size - 1); */
+
+		return 0;
+
+	} else {
+		/*This should not happen */
+		pr_err("%s: %d %d %d %d %d %p %d\n", __func__,
+		       hif->txtosend, hif->txtoclean, hif->txavail,
+			client_id, q_no, queue, queue->ack_idx);
+		WARN(1, "%s: doesn't own this descriptor", __func__);
+		return 1;
+	}
+}
+
+void __hif_tx_done_process(struct pfe_hif *hif, int count)
+{
+	struct hif_desc *desc;
+	struct hif_desc_sw *desc_sw;
+	int ttc, tx_avl;
+	int pkts_done[HIF_CLIENTS_MAX] = {0, 0};
+
+	ttc = hif->txtoclean;
+	tx_avl = hif->txavail;
+
+	while ((tx_avl < hif->tx_ring_size) && count--) {
+		desc = hif->tx_base + ttc;
+
+		if (readl(&desc->ctrl) & BD_CTRL_DESC_EN)
+			break;
+
+		desc_sw = &hif->tx_sw_queue[ttc];
+
+		if (desc_sw->data) {
+			/*
+			 * dmap_unmap_single(hif->dev, desc_sw->data,
+			 * desc_sw->len, DMA_TO_DEVICE);
+			 */
+			dma_unmap_single(hif->dev, desc_sw->data,
+					 desc_sw->len, DMA_TO_DEVICE);
+		}
+
+		if (desc_sw->client_id > HIF_CLIENTS_MAX)
+			pr_err("Invalid cl id %d\n", desc_sw->client_id);
+
+		pkts_done[desc_sw->client_id]++;
+
+		client_ack_txpacket(hif, desc_sw->client_id, desc_sw->q_no);
+
+		ttc = (ttc + 1) & (hif->tx_ring_size - 1);
+		tx_avl++;
+	}
+
+	if (pkts_done[0])
+		hif_lib_indicate_client(0, EVENT_TXDONE_IND, 0);
+	if (pkts_done[1])
+		hif_lib_indicate_client(1, EVENT_TXDONE_IND, 0);
+
+	hif->txtoclean = ttc;
+	hif->txavail = tx_avl;
+
+	if (!count) {
+		tasklet_schedule(&hif->tx_cleanup_tasklet);
+	} else {
+		/*Enable Tx done interrupt */
+		writel(readl_relaxed(HIF_INT_ENABLE) | HIF_TXPKT_INT,
+		       HIF_INT_ENABLE);
+	}
+}
+
+static void pfe_tx_do_cleanup(unsigned long data)
+{
+	struct pfe_hif *hif = (struct pfe_hif *)data;
+
+	writel(HIF_INT | HIF_TXPKT_INT, HIF_INT_SRC);
+
+	hif_tx_done_process(hif, 64);
+}
+
+/*
+ * __hif_xmit_pkt -
+ * This function puts one packet in the HIF Tx queue
+ */
+void __hif_xmit_pkt(struct pfe_hif *hif, unsigned int client_id, unsigned int
+			q_no, void *data, u32 len, unsigned int flags)
+{
+	struct hif_desc	*desc;
+	struct hif_desc_sw *desc_sw;
+
+	desc = hif->tx_base + hif->txtosend;
+	desc_sw = &hif->tx_sw_queue[hif->txtosend];
+
+	desc_sw->len = len;
+	desc_sw->client_id = client_id;
+	desc_sw->q_no = q_no;
+	desc_sw->flags = flags;
+
+	if (flags & HIF_DONT_DMA_MAP) {
+		desc_sw->data = 0;
+		writel((u32)DDR_PHYS_TO_PFE(data), &desc->data);
+	} else {
+		desc_sw->data = dma_map_single(hif->dev, data, len,
+						DMA_TO_DEVICE);
+		writel((u32)DDR_PHYS_TO_PFE(desc_sw->data), &desc->data);
+	}
+
+	hif->txtosend = (hif->txtosend + 1) & (hif->tx_ring_size - 1);
+	hif->txavail--;
+
+	if ((!((flags & HIF_DATA_VALID) && (flags &
+				HIF_LAST_BUFFER))))
+		goto skip_tx;
+
+	/*
+	 * Ensure everything else is written to DDR before
+	 * writing bd->ctrl
+	 */
+	wmb();
+
+	do {
+		desc_sw = &hif->tx_sw_queue[hif->txtoflush];
+		desc = hif->tx_base + hif->txtoflush;
+
+		if (desc_sw->flags & HIF_LAST_BUFFER) {
+			writel((BD_CTRL_LIFM |
+			       BD_CTRL_BRFETCH_DISABLE | BD_CTRL_RTFETCH_DISABLE
+			       | BD_CTRL_PARSE_DISABLE | BD_CTRL_DESC_EN |
+				BD_CTRL_PKT_INT_EN | BD_BUF_LEN(desc_sw->len)),
+				&desc->ctrl);
+		} else {
+			writel((BD_CTRL_DESC_EN |
+				BD_BUF_LEN(desc_sw->len)), &desc->ctrl);
+		}
+		hif->txtoflush = (hif->txtoflush + 1) & (hif->tx_ring_size - 1);
+	}
+	while (hif->txtoflush != hif->txtosend)
+		;
+
+skip_tx:
+	return;
+}
+
+int hif_xmit_pkt(struct pfe_hif *hif, unsigned int client_id, unsigned int q_no,
+		 void *data, unsigned int len)
+{
+	int rc = 0;
+
+	spin_lock_bh(&hif->tx_lock);
+
+	if (!hif->txavail) {
+		rc = 1;
+	} else {
+		__hif_xmit_pkt(hif, client_id, q_no, data, len,
+			       HIF_FIRST_BUFFER | HIF_LAST_BUFFER);
+		hif_tx_dma_start();
+	}
+
+	if (hif->txavail < (hif->tx_ring_size >> 1))
+		__hif_tx_done_process(hif, TX_FREE_MAX_COUNT);
+
+	spin_unlock_bh(&hif->tx_lock);
+
+	return rc;
+}
+
+static irqreturn_t wol_isr(int irq, void *dev_id)
+{
+	pr_info("WoL\n");
+	gemac_set_wol(EMAC1_BASE_ADDR, 0);
+	gemac_set_wol(EMAC2_BASE_ADDR, 0);
+	return IRQ_HANDLED;
+}
+
+/*
+ * hif_isr-
+ * This ISR routine processes Rx/Tx done interrupts from the HIF hardware block
+ */
+static irqreturn_t hif_isr(int irq, void *dev_id)
+{
+	struct pfe_hif *hif = (struct pfe_hif *)dev_id;
+	int int_status;
+	int int_enable_mask;
+
+	/*Read hif interrupt source register */
+	int_status = readl_relaxed(HIF_INT_SRC);
+	int_enable_mask = readl_relaxed(HIF_INT_ENABLE);
+
+	if ((int_status & HIF_INT) == 0)
+		return IRQ_NONE;
+
+	int_status &= ~(HIF_INT);
+
+	if (int_status & HIF_RXPKT_INT) {
+		int_status &= ~(HIF_RXPKT_INT);
+		int_enable_mask &= ~(HIF_RXPKT_INT);
+
+		napi_first_batch = 1;
+
+		if (napi_schedule_prep(&hif->napi)) {
+#ifdef HIF_NAPI_STATS
+			hif->napi_counters[NAPI_SCHED_COUNT]++;
+#endif
+			__napi_schedule(&hif->napi);
+		}
+	}
+	if (int_status & HIF_TXPKT_INT) {
+		int_status &= ~(HIF_TXPKT_INT);
+		int_enable_mask &= ~(HIF_TXPKT_INT);
+		/*Schedule tx cleanup tassklet */
+		tasklet_schedule(&hif->tx_cleanup_tasklet);
+	}
+
+	/*Disable interrupts, they will be enabled after they are serviced */
+	writel_relaxed(int_enable_mask, HIF_INT_ENABLE);
+
+	if (int_status) {
+		pr_info("%s : Invalid interrupt : %d\n", __func__,
+			int_status);
+		writel(int_status, HIF_INT_SRC);
+	}
+
+	return IRQ_HANDLED;
+}
+
+void hif_process_client_req(struct pfe_hif *hif, int req, int data1, int data2)
+{
+	unsigned int client_id = data1;
+
+	if (client_id >= HIF_CLIENTS_MAX) {
+		pr_err("%s: client id %d out of bounds\n", __func__,
+		       client_id);
+		return;
+	}
+
+	switch (req) {
+	case REQUEST_CL_REGISTER:
+			/* Request for register a client */
+			pr_info("%s: register client_id %d\n",
+				__func__, client_id);
+			pfe_hif_client_register(hif, client_id, (struct
+				hif_client_shm *)&hif->shm->client[client_id]);
+			break;
+
+	case REQUEST_CL_UNREGISTER:
+			pr_info("%s: unregister client_id %d\n",
+				__func__, client_id);
+
+			/* Request for unregister a client */
+			pfe_hif_client_unregister(hif, client_id);
+
+			break;
+
+	default:
+			pr_err("%s: unsupported request %d\n",
+			       __func__, req);
+			break;
+	}
+
+	/*
+	 * Process client Tx queues
+	 * Currently we don't have checking for tx pending
+	 */
+}
+
+/*
+ * pfe_hif_rx_poll
+ *  This function is NAPI poll function to process HIF Rx queue.
+ */
+static int pfe_hif_rx_poll(struct napi_struct *napi, int budget)
+{
+	struct pfe_hif *hif = container_of(napi, struct pfe_hif, napi);
+	int work_done;
+
+#ifdef HIF_NAPI_STATS
+	hif->napi_counters[NAPI_POLL_COUNT]++;
+#endif
+
+	work_done = pfe_hif_rx_process(hif, budget);
+
+	if (work_done < budget) {
+		napi_complete(napi);
+		writel(readl_relaxed(HIF_INT_ENABLE) | HIF_RXPKT_INT,
+		       HIF_INT_ENABLE);
+	}
+#ifdef HIF_NAPI_STATS
+	else
+		hif->napi_counters[NAPI_FULL_BUDGET_COUNT]++;
+#endif
+
+	return work_done;
+}
+
+/*
+ * pfe_hif_init
+ * This function initializes the baseaddresses and irq, etc.
+ */
+int pfe_hif_init(struct pfe *pfe)
+{
+	struct pfe_hif *hif = &pfe->hif;
+	int err;
+
+	pr_info("%s\n", __func__);
+
+	hif->dev = pfe->dev;
+	hif->irq = pfe->hif_irq;
+
+	err = pfe_hif_alloc_descr(hif);
+	if (err)
+		goto err0;
+
+	if (pfe_hif_init_buffers(hif)) {
+		pr_err("%s: Could not initialize buffer descriptors\n"
+			, __func__);
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	/* Initialize NAPI for Rx processing */
+	init_dummy_netdev(&hif->dummy_dev);
+	netif_napi_add(&hif->dummy_dev, &hif->napi, pfe_hif_rx_poll,
+		       HIF_RX_POLL_WEIGHT);
+	napi_enable(&hif->napi);
+
+	spin_lock_init(&hif->tx_lock);
+	spin_lock_init(&hif->lock);
+
+	hif_init();
+	hif_rx_enable();
+	hif_tx_enable();
+
+	/* Disable tx done interrupt */
+	writel(HIF_INT_MASK, HIF_INT_ENABLE);
+
+	gpi_enable(HGPI_BASE_ADDR);
+
+	err = request_irq(hif->irq, hif_isr, 0, "pfe_hif", hif);
+	if (err) {
+		pr_err("%s: failed to get the hif IRQ = %d\n",
+		       __func__, hif->irq);
+		goto err1;
+	}
+
+	err = request_irq(pfe->wol_irq, wol_isr, 0, "pfe_wol", pfe);
+	if (err) {
+		pr_err("%s: failed to get the wol IRQ = %d\n",
+		       __func__, pfe->wol_irq);
+		goto err1;
+	}
+
+	tasklet_init(&hif->tx_cleanup_tasklet,
+		     (void(*)(unsigned long))pfe_tx_do_cleanup,
+		     (unsigned long)hif);
+
+	return 0;
+err1:
+	pfe_hif_free_descr(hif);
+err0:
+	return err;
+}
+
+/* pfe_hif_exit- */
+void pfe_hif_exit(struct pfe *pfe)
+{
+	struct pfe_hif *hif = &pfe->hif;
+
+	pr_info("%s\n", __func__);
+
+	tasklet_kill(&hif->tx_cleanup_tasklet);
+
+	spin_lock_bh(&hif->lock);
+	hif->shm->g_client_status[0] = 0;
+	/* Make sure all clients are disabled*/
+	hif->shm->g_client_status[1] = 0;
+
+	spin_unlock_bh(&hif->lock);
+
+	/*Disable Rx/Tx */
+	gpi_disable(HGPI_BASE_ADDR);
+	hif_rx_disable();
+	hif_tx_disable();
+
+	napi_disable(&hif->napi);
+	netif_napi_del(&hif->napi);
+
+	free_irq(pfe->wol_irq, pfe);
+	free_irq(hif->irq, hif);
+
+	pfe_hif_release_buffers(hif);
+	pfe_hif_free_descr(hif);
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_hif_lib.c b/drivers/staging/fsl_ppfe/pfe_hif_lib.c
new file mode 100644
index 0000000..4274a59
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_hif_lib.c
@@ -0,0 +1,638 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/version.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/interrupt.h>
+#include <linux/workqueue.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmapool.h>
+#include <linux/sched.h>
+#include <linux/skbuff.h>
+#include <linux/moduleparam.h>
+#include <linux/cpu.h>
+
+#include "pfe_mod.h"
+#include "pfe_hif.h"
+#include "pfe_hif_lib.h"
+
+unsigned int lro_mode;
+unsigned int page_mode;
+unsigned int tx_qos;
+unsigned int pfe_pkt_size;
+unsigned int pfe_pkt_headroom;
+unsigned int emac_txq_cnt;
+
+/*
+ * @pfe_hal_lib.c.
+ * Common functions used by HIF client drivers
+ */
+
+/*HIF shared memory Global variable */
+struct hif_shm ghif_shm;
+
+/* Cleanup the HIF shared memory, release HIF rx_buffer_pool.
+ * This function should be called after pfe_hif_exit
+ *
+ * @param[in] hif_shm		Shared memory address location in DDR
+ */
+static void pfe_hif_shm_clean(struct hif_shm *hif_shm)
+{
+	int i;
+	void *pkt;
+
+	for (i = 0; i < hif_shm->rx_buf_pool_cnt; i++) {
+		pkt = hif_shm->rx_buf_pool[i];
+		if (pkt) {
+			hif_shm->rx_buf_pool[i] = NULL;
+			pkt -= pfe_pkt_headroom;
+
+			if (page_mode)
+				put_page(virt_to_page(pkt));
+			else
+				kfree(pkt);
+		}
+	}
+}
+
+/* Initialize shared memory used between HIF driver and clients,
+ * allocate rx_buffer_pool required for HIF Rx descriptors.
+ * This function should be called before initializing HIF driver.
+ *
+ * @param[in] hif_shm		Shared memory address location in DDR
+ * @rerurn			0 - on succes, <0 on fail to initialize
+ */
+static int pfe_hif_shm_init(struct hif_shm *hif_shm)
+{
+	int i;
+	void *pkt;
+
+	memset(hif_shm, 0, sizeof(struct hif_shm));
+	hif_shm->rx_buf_pool_cnt = HIF_RX_DESC_NT;
+
+	for (i = 0; i < hif_shm->rx_buf_pool_cnt; i++) {
+		if (page_mode) {
+			pkt = (void *)__get_free_page(GFP_KERNEL |
+				GFP_DMA_PFE);
+		} else {
+			pkt = kmalloc(PFE_BUF_SIZE, GFP_KERNEL | GFP_DMA_PFE);
+		}
+
+		if (pkt)
+			hif_shm->rx_buf_pool[i] = pkt + pfe_pkt_headroom;
+		else
+			goto err0;
+	}
+
+	return 0;
+
+err0:
+	pr_err("%s Low memory\n", __func__);
+	pfe_hif_shm_clean(hif_shm);
+	return -ENOMEM;
+}
+
+/*This function sends indication to HIF driver
+ *
+ * @param[in] hif	hif context
+ */
+static void hif_lib_indicate_hif(struct pfe_hif *hif, int req, int data1, int
+					data2)
+{
+	hif_process_client_req(hif, req, data1, data2);
+}
+
+void hif_lib_indicate_client(int client_id, int event_type, int qno)
+{
+	struct hif_client_s *client = pfe->hif_client[client_id];
+
+	if (!client || (event_type >= HIF_EVENT_MAX) || (qno >=
+		HIF_CLIENT_QUEUES_MAX))
+		return;
+
+	if (!test_and_set_bit(qno, &client->queue_mask[event_type]))
+		client->event_handler(client->priv, event_type, qno);
+}
+
+/*This function releases Rx queue descriptors memory and pre-filled buffers
+ *
+ * @param[in] client	hif_client context
+ */
+static void hif_lib_client_release_rx_buffers(struct hif_client_s *client)
+{
+	struct rx_queue_desc *desc;
+	int qno, ii;
+	void *buf;
+
+	for (qno = 0; qno < client->rx_qn; qno++) {
+		desc = client->rx_q[qno].base;
+
+		for (ii = 0; ii < client->rx_q[qno].size; ii++) {
+			buf = (void *)desc->data;
+			if (buf) {
+				buf -= pfe_pkt_headroom;
+
+				if (page_mode)
+					free_page((unsigned long)buf);
+				else
+					kfree(buf);
+
+				desc->ctrl = 0;
+			}
+
+			desc++;
+		}
+	}
+
+	kfree(client->rx_qbase);
+}
+
+/*This function allocates memory for the rxq descriptors and pre-fill rx queues
+ * with buffers.
+ * @param[in] client	client context
+ * @param[in] q_size	size of the rxQ, all queues are of same size
+ */
+static int hif_lib_client_init_rx_buffers(struct hif_client_s *client, int
+						q_size)
+{
+	struct rx_queue_desc *desc;
+	struct hif_client_rx_queue *queue;
+	int ii, qno;
+
+	/*Allocate memory for the client queues */
+	client->rx_qbase = kzalloc(client->rx_qn * q_size * sizeof(struct
+				rx_queue_desc), GFP_KERNEL);
+	if (!client->rx_qbase)
+		goto err;
+
+	for (qno = 0; qno < client->rx_qn; qno++) {
+		queue = &client->rx_q[qno];
+
+		queue->base = client->rx_qbase + qno * q_size * sizeof(struct
+				rx_queue_desc);
+		queue->size = q_size;
+		queue->read_idx = 0;
+		queue->write_idx = 0;
+
+		pr_debug("rx queue: %d, base: %p, size: %d\n", qno,
+			 queue->base, queue->size);
+	}
+
+	for (qno = 0; qno < client->rx_qn; qno++) {
+		queue = &client->rx_q[qno];
+		desc = queue->base;
+
+		for (ii = 0; ii < queue->size; ii++) {
+			desc->ctrl = CL_DESC_BUF_LEN(pfe_pkt_size) |
+					CL_DESC_OWN;
+			desc++;
+		}
+	}
+
+	return 0;
+
+err:
+	return 1;
+}
+
+#define inc_cl_idx(idxname)					\
+	({ typeof(idxname) idxname_ = (idxname);		\
+	((idxname_) = (idxname_ + 1) & (queue->size - 1)); })
+
+static void hif_lib_client_cleanup_tx_queue(struct hif_client_tx_queue *queue)
+{
+	pr_debug("%s\n", __func__);
+
+	/*
+	 * Check if there are any pending packets. Client must flush the tx
+	 * queues before unregistering, by calling by calling
+	 * hif_lib_tx_get_next_complete()
+	 *
+	 * Hif no longer calls since we are no longer registered
+	 */
+	if (queue->tx_pending)
+		pr_err("%s: pending transmit packets\n", __func__);
+}
+
+static void hif_lib_client_release_tx_buffers(struct hif_client_s *client)
+{
+	int qno;
+
+	pr_debug("%s\n", __func__);
+
+	for (qno = 0; qno < client->tx_qn; qno++)
+		hif_lib_client_cleanup_tx_queue(&client->tx_q[qno]);
+
+	kfree(client->tx_qbase);
+}
+
+static int hif_lib_client_init_tx_buffers(struct hif_client_s *client, int
+						q_size)
+{
+	struct hif_client_tx_queue *queue;
+	int qno;
+
+	client->tx_qbase = kzalloc(client->tx_qn * q_size * sizeof(struct
+					tx_queue_desc), GFP_KERNEL);
+	if (!client->tx_qbase)
+		return 1;
+
+	for (qno = 0; qno < client->tx_qn; qno++) {
+		queue = &client->tx_q[qno];
+
+		queue->base = client->tx_qbase + qno * q_size * sizeof(struct
+				tx_queue_desc);
+		queue->size = q_size;
+		queue->read_idx = 0;
+		queue->write_idx = 0;
+		queue->tx_pending = 0;
+		queue->nocpy_flag = 0;
+		queue->prev_tmu_tx_pkts = 0;
+		queue->done_tmu_tx_pkts = 0;
+
+		pr_debug("tx queue: %d, base: %p, size: %d\n", qno,
+			 queue->base, queue->size);
+	}
+
+	return 0;
+}
+
+static int hif_lib_event_dummy(void *priv, int event_type, int qno)
+{
+	return 0;
+}
+
+int hif_lib_client_register(struct hif_client_s *client)
+{
+	struct hif_shm *hif_shm;
+	struct hif_client_shm *client_shm;
+	int err, i;
+	/* int loop_cnt = 0; */
+
+	pr_debug("%s\n", __func__);
+
+	/*Allocate memory before spin_lock*/
+	if (hif_lib_client_init_rx_buffers(client, client->rx_qsize)) {
+		err = -ENOMEM;
+		goto err_rx;
+	}
+
+	if (hif_lib_client_init_tx_buffers(client, client->tx_qsize)) {
+		err = -ENOMEM;
+		goto err_tx;
+	}
+
+	spin_lock_bh(&pfe->hif.lock);
+	if (!(client->pfe) || (client->id >= HIF_CLIENTS_MAX) ||
+	    (pfe->hif_client[client->id])) {
+		err = -EINVAL;
+		goto err;
+	}
+
+	hif_shm = client->pfe->hif.shm;
+
+	if (!client->event_handler)
+		client->event_handler = hif_lib_event_dummy;
+
+	/*Initialize client specific shared memory */
+	client_shm = (struct hif_client_shm *)&hif_shm->client[client->id];
+	client_shm->rx_qbase = (unsigned long int)client->rx_qbase;
+	client_shm->rx_qsize = client->rx_qsize;
+	client_shm->tx_qbase = (unsigned long int)client->tx_qbase;
+	client_shm->tx_qsize = client->tx_qsize;
+	client_shm->ctrl = (client->tx_qn << CLIENT_CTRL_TX_Q_CNT_OFST) |
+				(client->rx_qn << CLIENT_CTRL_RX_Q_CNT_OFST);
+	/* spin_lock_init(&client->rx_lock); */
+
+	for (i = 0; i < HIF_EVENT_MAX; i++) {
+		client->queue_mask[i] = 0;  /*
+					     * By default all events are
+					     * unmasked
+					     */
+	}
+
+	/*Indicate to HIF driver*/
+	hif_lib_indicate_hif(&pfe->hif, REQUEST_CL_REGISTER, client->id, 0);
+
+	pr_debug("%s: client: %p, client_id: %d, tx_qsize: %d, rx_qsize: %d\n",
+		 __func__, client, client->id, client->tx_qsize,
+		 client->rx_qsize);
+
+	client->cpu_id = -1;
+
+	pfe->hif_client[client->id] = client;
+	spin_unlock_bh(&pfe->hif.lock);
+
+	return 0;
+
+err:
+	spin_unlock_bh(&pfe->hif.lock);
+	hif_lib_client_release_tx_buffers(client);
+
+err_tx:
+	hif_lib_client_release_rx_buffers(client);
+
+err_rx:
+	return err;
+}
+
+int hif_lib_client_unregister(struct hif_client_s *client)
+{
+	struct pfe *pfe = client->pfe;
+	u32 client_id = client->id;
+
+	pr_info(
+		"%s : client: %p, client_id: %d, txQ_depth: %d, rxQ_depth: %d\n"
+		, __func__, client, client->id, client->tx_qsize,
+		client->rx_qsize);
+
+	spin_lock_bh(&pfe->hif.lock);
+	hif_lib_indicate_hif(&pfe->hif, REQUEST_CL_UNREGISTER, client->id, 0);
+
+	hif_lib_client_release_tx_buffers(client);
+	hif_lib_client_release_rx_buffers(client);
+	pfe->hif_client[client_id] = NULL;
+	spin_unlock_bh(&pfe->hif.lock);
+
+	return 0;
+}
+
+int hif_lib_event_handler_start(struct hif_client_s *client, int event,
+				int qno)
+{
+	struct hif_client_rx_queue *queue = &client->rx_q[qno];
+	struct rx_queue_desc *desc = queue->base + queue->read_idx;
+
+	if ((event >= HIF_EVENT_MAX) || (qno >= HIF_CLIENT_QUEUES_MAX)) {
+		pr_debug("%s: Unsupported event : %d  queue number : %d\n",
+			 __func__, event, qno);
+		return -1;
+	}
+
+	test_and_clear_bit(qno, &client->queue_mask[event]);
+
+	switch (event) {
+	case EVENT_RX_PKT_IND:
+		if (!(desc->ctrl & CL_DESC_OWN))
+			hif_lib_indicate_client(client->id,
+						EVENT_RX_PKT_IND, qno);
+		break;
+
+	case EVENT_HIGH_RX_WM:
+	case EVENT_TXDONE_IND:
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+/*
+ * This function gets one packet from the specified client queue
+ * It also refill the rx buffer
+ */
+void *hif_lib_receive_pkt(struct hif_client_s *client, int qno, int *len, int
+				*ofst, unsigned int *rx_ctrl,
+				unsigned int *desc_ctrl, void **priv_data)
+{
+	struct hif_client_rx_queue *queue = &client->rx_q[qno];
+	struct rx_queue_desc *desc;
+	void *pkt = NULL;
+
+	/*
+	 * Following lock is to protect rx queue access from,
+	 * hif_lib_event_handler_start.
+	 * In general below lock is not required, because hif_lib_xmit_pkt and
+	 * hif_lib_event_handler_start are called from napi poll and which is
+	 * not re-entrant. But if some client use in different way this lock is
+	 * required.
+	 */
+	/*spin_lock_irqsave(&client->rx_lock, flags); */
+	desc = queue->base + queue->read_idx;
+	if (!(desc->ctrl & CL_DESC_OWN)) {
+		pkt = desc->data - pfe_pkt_headroom;
+
+		*rx_ctrl = desc->client_ctrl;
+		*desc_ctrl = desc->ctrl;
+
+		if (desc->ctrl & CL_DESC_FIRST) {
+			u16 size = *rx_ctrl >> HIF_CTRL_RX_OFFSET_OFST;
+
+			if (size) {
+				*len = CL_DESC_BUF_LEN(desc->ctrl) -
+						PFE_PKT_HEADER_SZ - size;
+				*ofst = pfe_pkt_headroom + PFE_PKT_HEADER_SZ
+								+ size;
+				*priv_data = desc->data + PFE_PKT_HEADER_SZ;
+			} else {
+				*len = CL_DESC_BUF_LEN(desc->ctrl) -
+						PFE_PKT_HEADER_SZ;
+				*ofst = pfe_pkt_headroom + PFE_PKT_HEADER_SZ;
+				*priv_data = NULL;
+			}
+
+		} else {
+			*len = CL_DESC_BUF_LEN(desc->ctrl);
+			*ofst = pfe_pkt_headroom;
+		}
+
+		/*
+		 * Needed so we don't free a buffer/page
+		 * twice on module_exit
+		 */
+		desc->data = NULL;
+
+		/*
+		 * Ensure everything else is written to DDR before
+		 * writing bd->ctrl
+		 */
+		smp_wmb();
+
+		desc->ctrl = CL_DESC_BUF_LEN(pfe_pkt_size) | CL_DESC_OWN;
+		inc_cl_idx(queue->read_idx);
+	}
+
+	/*spin_unlock_irqrestore(&client->rx_lock, flags); */
+	return pkt;
+}
+
+static inline void hif_hdr_write(struct hif_hdr *pkt_hdr, unsigned int
+					client_id, unsigned int qno,
+					u32 client_ctrl)
+{
+	/* Optimize the write since the destinaton may be non-cacheable */
+	if (!((unsigned long)pkt_hdr & 0x3)) {
+		((u32 *)pkt_hdr)[0] = (client_ctrl << 16) | (qno << 8) |
+					client_id;
+	} else {
+		((u16 *)pkt_hdr)[0] = (qno << 8) | (client_id & 0xFF);
+		((u16 *)pkt_hdr)[1] = (client_ctrl & 0xFFFF);
+	}
+}
+
+/*This function puts the given packet in the specific client queue */
+void __hif_lib_xmit_pkt(struct hif_client_s *client, unsigned int qno, void
+				*data, unsigned int len, u32 client_ctrl,
+				unsigned int flags, void *client_data)
+{
+	struct hif_client_tx_queue *queue = &client->tx_q[qno];
+	struct tx_queue_desc *desc = queue->base + queue->write_idx;
+
+	/* First buffer */
+	if (flags & HIF_FIRST_BUFFER) {
+		data -= sizeof(struct hif_hdr);
+		len += sizeof(struct hif_hdr);
+
+		hif_hdr_write(data, client->id, qno, client_ctrl);
+	}
+
+	desc->data = client_data;
+	desc->ctrl = CL_DESC_OWN | CL_DESC_FLAGS(flags);
+
+	__hif_xmit_pkt(&pfe->hif, client->id, qno, data, len, flags);
+
+	inc_cl_idx(queue->write_idx);
+	queue->tx_pending++;
+	queue->jiffies_last_packet = jiffies;
+}
+
+/*This function puts the given packet in the specific client queue */
+int hif_lib_xmit_pkt(struct hif_client_s *client, unsigned int qno, void *data,
+		     unsigned int len, u32 client_ctrl, void *client_data)
+{
+	struct hif_client_tx_queue *queue = &client->tx_q[qno];
+	struct tx_queue_desc *desc = queue->base + queue->write_idx;
+
+	if (queue->tx_pending < queue->size) {
+		/*Construct pkt header */
+
+		data -= sizeof(struct hif_hdr);
+		len += sizeof(struct hif_hdr);
+
+		hif_hdr_write(data, client->id, qno, client_ctrl);
+
+		desc->data = client_data;
+		desc->ctrl = CL_DESC_OWN | CL_DESC_FLAGS(HIF_FIRST_BUFFER |
+				HIF_LAST_BUFFER | HIF_DATA_VALID);
+
+		if (hif_xmit_pkt(&pfe->hif, client->id, qno, data, len))
+			return 1;
+
+		inc_cl_idx(queue->write_idx);
+		queue->tx_pending++;
+		queue->jiffies_last_packet = jiffies;
+
+		return 0;
+	}
+
+	pr_debug("%s Tx client %d qno %d is full\n", __func__, client->id,
+		 qno);
+	return 1;
+}
+
+void *hif_lib_tx_get_next_complete(struct hif_client_s *client, int qno,
+				   unsigned int *flags, int count)
+{
+	struct hif_client_tx_queue *queue = &client->tx_q[qno];
+	struct tx_queue_desc *desc = queue->base + queue->read_idx;
+
+	pr_debug("%s: qno : %d rd_indx: %d pending:%d\n", __func__, qno,
+		 queue->read_idx, queue->tx_pending);
+
+	if (!queue->tx_pending)
+		return NULL;
+
+	if (queue->nocpy_flag && !queue->done_tmu_tx_pkts) {
+		u32 tmu_tx_pkts = be32_to_cpu(pe_dmem_read(TMU0_ID +
+			client->id, TMU_DM_TX_TRANS, 4));
+
+		if (queue->prev_tmu_tx_pkts > tmu_tx_pkts)
+			queue->done_tmu_tx_pkts = UINT_MAX -
+				queue->prev_tmu_tx_pkts + tmu_tx_pkts;
+		else
+			queue->done_tmu_tx_pkts = tmu_tx_pkts -
+						queue->prev_tmu_tx_pkts;
+
+		queue->prev_tmu_tx_pkts  = tmu_tx_pkts;
+
+		if (!queue->done_tmu_tx_pkts)
+			return NULL;
+	}
+
+	if (desc->ctrl & CL_DESC_OWN)
+		return NULL;
+
+	inc_cl_idx(queue->read_idx);
+	queue->tx_pending--;
+
+	*flags = CL_DESC_GET_FLAGS(desc->ctrl);
+
+	if (queue->done_tmu_tx_pkts && (*flags & HIF_LAST_BUFFER))
+		queue->done_tmu_tx_pkts--;
+
+	return desc->data;
+}
+
+static void hif_lib_tmu_credit_init(struct pfe *pfe)
+{
+	int i, q;
+
+	for (i = 0; i < NUM_GEMAC_SUPPORT; i++)
+		for (q = 0; q < emac_txq_cnt; q++) {
+			pfe->tmu_credit.tx_credit_max[i][q] = (q == 0) ?
+					DEFAULT_Q0_QDEPTH : DEFAULT_MAX_QDEPTH;
+			pfe->tmu_credit.tx_credit[i][q] =
+					pfe->tmu_credit.tx_credit_max[i][q];
+		}
+}
+
+int pfe_hif_lib_init(struct pfe *pfe)
+{
+	int rc;
+
+	pr_info("%s\n", __func__);
+
+	if (lro_mode) {
+		page_mode = 1;
+		pfe_pkt_size = min(PAGE_SIZE, MAX_PFE_PKT_SIZE);
+		pfe_pkt_headroom = 0;
+	} else {
+		page_mode = 0;
+		pfe_pkt_size = PFE_PKT_SIZE;
+		pfe_pkt_headroom = PFE_PKT_HEADROOM;
+	}
+
+	if (tx_qos)
+		emac_txq_cnt = EMAC_TXQ_CNT / 2;
+	else
+		emac_txq_cnt = EMAC_TXQ_CNT;
+
+	hif_lib_tmu_credit_init(pfe);
+	pfe->hif.shm = &ghif_shm;
+	rc = pfe_hif_shm_init(pfe->hif.shm);
+
+	return rc;
+}
+
+void pfe_hif_lib_exit(struct pfe *pfe)
+{
+	pr_info("%s\n", __func__);
+
+	pfe_hif_shm_clean(pfe->hif.shm);
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_hw.c b/drivers/staging/fsl_ppfe/pfe_hw.c
new file mode 100644
index 0000000..16ea2c6
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_hw.c
@@ -0,0 +1,176 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "pfe_mod.h"
+#include "pfe_hw.h"
+
+/* Functions to handle most of pfe hw register initialization */
+int pfe_hw_init(struct pfe *pfe, int resume)
+{
+	struct class_cfg class_cfg = {
+		.pe_sys_clk_ratio = PE_SYS_CLK_RATIO,
+		.route_table_baseaddr = pfe->ddr_phys_baseaddr +
+					ROUTE_TABLE_BASEADDR,
+		.route_table_hash_bits = ROUTE_TABLE_HASH_BITS,
+	};
+
+	struct tmu_cfg tmu_cfg = {
+		.pe_sys_clk_ratio = PE_SYS_CLK_RATIO,
+		.llm_base_addr = pfe->ddr_phys_baseaddr + TMU_LLM_BASEADDR,
+		.llm_queue_len = TMU_LLM_QUEUE_LEN,
+	};
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	struct util_cfg util_cfg = {
+		.pe_sys_clk_ratio = PE_SYS_CLK_RATIO,
+	};
+#endif
+
+	struct BMU_CFG bmu1_cfg = {
+		.baseaddr = CBUS_VIRT_TO_PFE(LMEM_BASE_ADDR +
+						BMU1_LMEM_BASEADDR),
+		.count = BMU1_BUF_COUNT,
+		.size = BMU1_BUF_SIZE,
+		.low_watermark = 10,
+		.high_watermark = 15,
+	};
+
+	struct BMU_CFG bmu2_cfg = {
+		.baseaddr = DDR_PHYS_TO_PFE(pfe->ddr_phys_baseaddr +
+						BMU2_DDR_BASEADDR),
+		.count = BMU2_BUF_COUNT,
+		.size = BMU2_BUF_SIZE,
+		.low_watermark = 250,
+		.high_watermark = 253,
+	};
+
+	struct gpi_cfg egpi1_cfg = {
+		.lmem_rtry_cnt = EGPI1_LMEM_RTRY_CNT,
+		.tmlf_txthres = EGPI1_TMLF_TXTHRES,
+		.aseq_len = EGPI1_ASEQ_LEN,
+		.mtip_pause_reg = CBUS_VIRT_TO_PFE(EMAC1_BASE_ADDR +
+						EMAC_TCNTRL_REG),
+	};
+
+	struct gpi_cfg egpi2_cfg = {
+		.lmem_rtry_cnt = EGPI2_LMEM_RTRY_CNT,
+		.tmlf_txthres = EGPI2_TMLF_TXTHRES,
+		.aseq_len = EGPI2_ASEQ_LEN,
+		.mtip_pause_reg = CBUS_VIRT_TO_PFE(EMAC2_BASE_ADDR +
+						EMAC_TCNTRL_REG),
+	};
+
+	struct gpi_cfg hgpi_cfg = {
+		.lmem_rtry_cnt = HGPI_LMEM_RTRY_CNT,
+		.tmlf_txthres = HGPI_TMLF_TXTHRES,
+		.aseq_len = HGPI_ASEQ_LEN,
+		.mtip_pause_reg = 0,
+	};
+
+	pr_info("%s\n", __func__);
+
+#if !defined(LS1012A_PFE_RESET_WA)
+	/* LS1012A needs this to make PE work correctly */
+	writel(0x3,     CLASS_PE_SYS_CLK_RATIO);
+	writel(0x3,     TMU_PE_SYS_CLK_RATIO);
+	writel(0x3,     UTIL_PE_SYS_CLK_RATIO);
+	usleep_range(10, 20);
+#endif
+
+	pr_info("CLASS version: %x\n", readl(CLASS_VERSION));
+	pr_info("TMU version: %x\n", readl(TMU_VERSION));
+
+	pr_info("BMU1 version: %x\n", readl(BMU1_BASE_ADDR +
+		BMU_VERSION));
+	pr_info("BMU2 version: %x\n", readl(BMU2_BASE_ADDR +
+		BMU_VERSION));
+
+	pr_info("EGPI1 version: %x\n", readl(EGPI1_BASE_ADDR +
+		GPI_VERSION));
+	pr_info("EGPI2 version: %x\n", readl(EGPI2_BASE_ADDR +
+		GPI_VERSION));
+	pr_info("HGPI version: %x\n", readl(HGPI_BASE_ADDR +
+		GPI_VERSION));
+
+	pr_info("HIF version: %x\n", readl(HIF_VERSION));
+	pr_info("HIF NOPCY version: %x\n", readl(HIF_NOCPY_VERSION));
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	pr_info("UTIL version: %x\n", readl(UTIL_VERSION));
+#endif
+	while (!(readl(TMU_CTRL) & ECC_MEM_INIT_DONE))
+		;
+
+	hif_rx_disable();
+	hif_tx_disable();
+
+	bmu_init(BMU1_BASE_ADDR, &bmu1_cfg);
+
+	pr_info("bmu_init(1) done\n");
+
+	bmu_init(BMU2_BASE_ADDR, &bmu2_cfg);
+
+	pr_info("bmu_init(2) done\n");
+
+	class_cfg.resume = resume ? 1 : 0;
+
+	class_init(&class_cfg);
+
+	pr_info("class_init() done\n");
+
+	tmu_init(&tmu_cfg);
+
+	pr_info("tmu_init() done\n");
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	util_init(&util_cfg);
+
+	pr_info("util_init() done\n");
+#endif
+	gpi_init(EGPI1_BASE_ADDR, &egpi1_cfg);
+
+	pr_info("gpi_init(1) done\n");
+
+	gpi_init(EGPI2_BASE_ADDR, &egpi2_cfg);
+
+	pr_info("gpi_init(2) done\n");
+
+	gpi_init(HGPI_BASE_ADDR, &hgpi_cfg);
+
+	pr_info("gpi_init(hif) done\n");
+
+	bmu_enable(BMU1_BASE_ADDR);
+
+	pr_info("bmu_enable(1) done\n");
+
+	bmu_enable(BMU2_BASE_ADDR);
+
+	pr_info("bmu_enable(2) done\n");
+
+	return 0;
+}
+
+void pfe_hw_exit(struct pfe *pfe)
+{
+	pr_info("%s\n", __func__);
+
+	bmu_disable(BMU1_BASE_ADDR);
+	bmu_reset(BMU1_BASE_ADDR);
+
+	bmu_disable(BMU2_BASE_ADDR);
+	bmu_reset(BMU2_BASE_ADDR);
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_ls1012a_platform.c b/drivers/staging/fsl_ppfe/pfe_ls1012a_platform.c
new file mode 100644
index 0000000..c579eb5
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_ls1012a_platform.c
@@ -0,0 +1,394 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/module.h>
+#include <linux/device.h>
+#include <linux/of_net.h>
+#include <linux/of_address.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/clk.h>
+#include <linux/mfd/syscon.h>
+#include <linux/regmap.h>
+
+#include "pfe_mod.h"
+
+struct ls1012a_pfe_platform_data pfe_platform_data;
+
+static int pfe_get_gemac_if_proprties(struct device_node *parent, int port, int
+					if_cnt,
+					struct ls1012a_pfe_platform_data
+					*pdata)
+{
+	struct device_node *gem = NULL, *phy = NULL;
+	int size;
+	int ii = 0, phy_id = 0;
+	const u32 *addr;
+	const void *mac_addr;
+
+	for (ii = 0; ii < if_cnt; ii++) {
+		gem = of_get_next_child(parent, gem);
+		if (!gem)
+			goto err;
+		addr = of_get_property(gem, "reg", &size);
+		if (addr && (be32_to_cpup(addr) == port))
+			break;
+	}
+
+	if (ii >= if_cnt) {
+		pr_err("%s:%d Failed to find interface = %d\n",
+		       __func__, __LINE__, if_cnt);
+		goto err;
+	}
+
+	pdata->ls1012a_eth_pdata[port].gem_id = port;
+
+	mac_addr = of_get_mac_address(gem);
+
+	if (mac_addr) {
+		memcpy(pdata->ls1012a_eth_pdata[port].mac_addr, mac_addr,
+		       ETH_ALEN);
+	}
+
+	pdata->ls1012a_eth_pdata[port].mii_config = of_get_phy_mode(gem);
+
+	if ((pdata->ls1012a_eth_pdata[port].mii_config) < 0)
+		pr_err("%s:%d Incorrect Phy mode....\n", __func__,
+		       __LINE__);
+
+	addr = of_get_property(gem, "fsl,gemac-bus-id", &size);
+	if (!addr)
+		pr_err("%s:%d Invalid gemac-bus-id....\n", __func__,
+		       __LINE__);
+	else
+		pdata->ls1012a_eth_pdata[port].bus_id = be32_to_cpup(addr);
+
+	addr = of_get_property(gem, "fsl,gemac-phy-id", &size);
+	if (!addr) {
+		pr_err("%s:%d Invalid gemac-phy-id....\n", __func__,
+		       __LINE__);
+	} else {
+		phy_id = be32_to_cpup(addr);
+		pdata->ls1012a_eth_pdata[port].phy_id = phy_id;
+		pdata->ls1012a_mdio_pdata[0].phy_mask &= ~(1 << phy_id);
+	}
+
+	addr = of_get_property(gem, "fsl,mdio-mux-val", &size);
+	if (!addr)
+		pr_err("%s: Invalid mdio-mux-val....\n", __func__);
+	else
+		phy_id = be32_to_cpup(addr);
+		pdata->ls1012a_eth_pdata[port].mdio_muxval = phy_id;
+
+	if (pdata->ls1012a_eth_pdata[port].phy_id < 32)
+		pfe->mdio_muxval[pdata->ls1012a_eth_pdata[port].phy_id] =
+			 pdata->ls1012a_eth_pdata[port].mdio_muxval;
+
+	addr = of_get_property(gem, "fsl,pfe-phy-if-flags", &size);
+	if (!addr)
+		pr_err("%s:%d Invalid pfe-phy-if-flags....\n",
+		       __func__, __LINE__);
+	else
+		pdata->ls1012a_eth_pdata[port].phy_flags = be32_to_cpup(addr);
+
+	/* If PHY is enabled, read mdio properties */
+	if (pdata->ls1012a_eth_pdata[port].phy_flags & GEMAC_NO_PHY)
+		goto done;
+
+	phy = of_get_next_child(gem, NULL);
+
+	addr = of_get_property(phy, "reg", &size);
+
+	if (!addr)
+		pr_err("%s:%d Invalid phy enable flag....\n",
+		       __func__, __LINE__);
+	else
+		pdata->ls1012a_mdio_pdata[port].enabled = be32_to_cpup(addr);
+
+	pdata->ls1012a_mdio_pdata[port].irq[0] = PHY_POLL;
+
+done:
+
+	return 0;
+
+err:
+	return -1;
+}
+
+/*
+ *
+ * pfe_platform_probe -
+ *
+ *
+ */
+static int pfe_platform_probe(struct platform_device *pdev)
+{
+	struct resource res;
+	int ii, rc, interface_count = 0, size = 0;
+	const u32 *prop;
+	struct device_node  *np;
+	struct clk *pfe_clk;
+
+	np = pdev->dev.of_node;
+
+	if (!np) {
+		pr_err("Invalid device node\n");
+		return -EINVAL;
+	}
+
+	pfe = kzalloc(sizeof(*pfe), GFP_KERNEL);
+	if (!pfe) {
+		rc = -ENOMEM;
+		goto err_alloc;
+	}
+
+	platform_set_drvdata(pdev, pfe);
+
+	dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
+
+	if (of_address_to_resource(np, 1, &res)) {
+		rc = -ENOMEM;
+		pr_err("failed to get ddr resource\n");
+		goto err_ddr;
+	}
+
+	pfe->ddr_phys_baseaddr = res.start;
+	pfe->ddr_size = resource_size(&res);
+
+	pfe->ddr_baseaddr = phys_to_virt(res.start);
+	if (!pfe->ddr_baseaddr) {
+		pr_err("ioremap() ddr failed\n");
+		rc = -ENOMEM;
+		goto err_ddr;
+	}
+
+	pfe->scfg =
+		syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+						"fsl,pfe-scfg");
+	if (IS_ERR(pfe->scfg)) {
+		dev_err(&pdev->dev, "No syscfg phandle specified\n");
+		return PTR_ERR(pfe->scfg);
+	}
+
+	pfe->cbus_baseaddr = of_iomap(np, 0);
+	if (!pfe->cbus_baseaddr) {
+		rc = -ENOMEM;
+		pr_err("failed to get axi resource\n");
+		goto err_axi;
+	}
+
+	pfe->hif_irq = platform_get_irq(pdev, 0);
+	if (pfe->hif_irq < 0) {
+		pr_err("platform_get_irq for hif failed\n");
+		rc = pfe->hif_irq;
+		goto err_hif_irq;
+	}
+
+	pfe->wol_irq = platform_get_irq(pdev, 2);
+	if (pfe->wol_irq < 0) {
+		pr_err("platform_get_irq for WoL failed\n");
+		rc = pfe->wol_irq;
+		goto err_hif_irq;
+	}
+
+	/* Read interface count */
+	prop = of_get_property(np, "fsl,pfe-num-interfaces", &size);
+	if (!prop) {
+		pr_err("Failed to read number of interfaces\n");
+		rc = -ENXIO;
+		goto err_prop;
+	}
+
+	interface_count = be32_to_cpup(prop);
+	if (interface_count <= 0) {
+		pr_err("No ethernet interface count : %d\n",
+		       interface_count);
+		rc = -ENXIO;
+		goto err_prop;
+	}
+
+	pfe_platform_data.ls1012a_mdio_pdata[0].phy_mask = 0xffffffff;
+
+	for (ii = 0; ii < interface_count; ii++) {
+		pfe_get_gemac_if_proprties(np, ii, interface_count,
+					   &pfe_platform_data);
+	}
+
+	pfe->dev = &pdev->dev;
+
+	pfe->dev->platform_data = &pfe_platform_data;
+
+	/* declare WoL capabilities */
+	device_init_wakeup(&pdev->dev, true);
+
+	/* find the clocks */
+	pfe_clk = devm_clk_get(pfe->dev, "pfe");
+	if (IS_ERR(pfe_clk))
+		return PTR_ERR(pfe_clk);
+
+	/* PFE clock is (platform clock / 2) */
+	/* save sys_clk value as KHz */
+	pfe->ctrl.sys_clk = clk_get_rate(pfe_clk) / (2 * 1000);
+
+	rc = pfe_probe(pfe);
+	if (rc < 0)
+		goto err_probe;
+
+	return 0;
+
+err_probe:
+err_prop:
+err_hif_irq:
+	iounmap(pfe->cbus_baseaddr);
+
+err_axi:
+	iounmap(pfe->ddr_baseaddr);
+
+err_ddr:
+	platform_set_drvdata(pdev, NULL);
+
+	kfree(pfe);
+
+err_alloc:
+	return rc;
+}
+
+/*
+ * pfe_platform_remove -
+ */
+static int pfe_platform_remove(struct platform_device *pdev)
+{
+	struct pfe *pfe = platform_get_drvdata(pdev);
+	int rc;
+
+	pr_info("%s\n", __func__);
+
+	rc = pfe_remove(pfe);
+
+	iounmap(pfe->cbus_baseaddr);
+	iounmap(pfe->ddr_baseaddr);
+
+	platform_set_drvdata(pdev, NULL);
+
+	kfree(pfe);
+
+	return rc;
+}
+
+#ifdef CONFIG_PM
+#ifdef CONFIG_PM_SLEEP
+int pfe_platform_suspend(struct device *dev)
+{
+	struct pfe *pfe = platform_get_drvdata(to_platform_device(dev));
+	struct net_device *netdev;
+	int i;
+
+	pfe->wake = 0;
+
+	for (i = 0; i < (NUM_GEMAC_SUPPORT); i++) {
+		netdev = pfe->eth.eth_priv[i]->ndev;
+
+		netif_device_detach(netdev);
+
+		if (netif_running(netdev))
+			if (pfe_eth_suspend(netdev))
+				pfe->wake = 1;
+	}
+
+	/* Shutdown PFE only if we're not waking up the system */
+	if (!pfe->wake) {
+#if defined(LS1012A_PFE_RESET_WA)
+		pfe_hif_rx_idle(&pfe->hif);
+#endif
+		pfe_ctrl_suspend(&pfe->ctrl);
+		pfe_firmware_exit(pfe);
+
+		pfe_hif_exit(pfe);
+		pfe_hif_lib_exit(pfe);
+
+		pfe_hw_exit(pfe);
+	}
+
+	return 0;
+}
+
+static int pfe_platform_resume(struct device *dev)
+{
+	struct pfe *pfe = platform_get_drvdata(to_platform_device(dev));
+	struct net_device *netdev;
+	int i;
+
+	if (!pfe->wake) {
+		pfe_hw_init(pfe, 1);
+		pfe_hif_lib_init(pfe);
+		pfe_hif_init(pfe);
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+		util_enable();
+#endif
+		tmu_enable(0xf);
+		class_enable();
+		pfe_ctrl_resume(&pfe->ctrl);
+	}
+
+	for (i = 0; i < (NUM_GEMAC_SUPPORT); i++) {
+		netdev = pfe->eth.eth_priv[i]->ndev;
+
+		if (pfe->eth.eth_priv[i]->mii_bus)
+			pfe_eth_mdio_reset(pfe->eth.eth_priv[i]->mii_bus);
+
+		if (netif_running(netdev))
+			pfe_eth_resume(netdev);
+
+		netif_device_attach(netdev);
+	}
+	return 0;
+}
+#else
+#define pfe_platform_suspend NULL
+#define pfe_platform_resume NULL
+#endif
+
+static const struct dev_pm_ops pfe_platform_pm_ops = {
+	SET_SYSTEM_SLEEP_PM_OPS(pfe_platform_suspend, pfe_platform_resume)
+};
+#endif
+
+static const struct of_device_id pfe_match[] = {
+	{
+		.compatible = "fsl,pfe",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(of, pfe_match);
+
+static struct platform_driver pfe_platform_driver = {
+	.probe = pfe_platform_probe,
+	.remove = pfe_platform_remove,
+	.driver = {
+		.name = "pfe",
+		.of_match_table = pfe_match,
+#ifdef CONFIG_PM
+		.pm = &pfe_platform_pm_ops,
+#endif
+	},
+};
+
+module_platform_driver(pfe_platform_driver);
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("PFE Ethernet driver");
+MODULE_AUTHOR("NXP DNCPE");
diff --git a/drivers/staging/fsl_ppfe/pfe_mod.c b/drivers/staging/fsl_ppfe/pfe_mod.c
new file mode 100644
index 0000000..d5ba56a
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_mod.c
@@ -0,0 +1,141 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/dma-mapping.h>
+#include "pfe_mod.h"
+
+struct pfe *pfe;
+
+/*
+ * pfe_probe -
+ */
+int pfe_probe(struct pfe *pfe)
+{
+	int rc;
+
+	if (pfe->ddr_size < DDR_MAX_SIZE) {
+		pr_err("%s: required DDR memory (%x) above platform ddr memory (%x)\n",
+		       __func__, (unsigned int)DDR_MAX_SIZE, pfe->ddr_size);
+		rc = -ENOMEM;
+		goto err_hw;
+	}
+
+	if (((int)(pfe->ddr_phys_baseaddr + BMU2_DDR_BASEADDR) &
+			(8 * SZ_1M - 1)) != 0) {
+		pr_err("%s: BMU2 base address (0x%x) must be aligned on 8MB boundary\n",
+		       __func__, (int)pfe->ddr_phys_baseaddr +
+			BMU2_DDR_BASEADDR);
+		rc = -ENOMEM;
+		goto err_hw;
+	}
+
+	pr_info("cbus_baseaddr: %lx, ddr_baseaddr: %lx, ddr_phys_baseaddr: %lx, ddr_size: %x\n",
+		(unsigned long)pfe->cbus_baseaddr,
+		(unsigned long)pfe->ddr_baseaddr,
+		pfe->ddr_phys_baseaddr, pfe->ddr_size);
+
+	pfe_lib_init(pfe->cbus_baseaddr, pfe->ddr_baseaddr,
+		     pfe->ddr_phys_baseaddr, pfe->ddr_size);
+
+	rc = pfe_hw_init(pfe, 0);
+	if (rc < 0)
+		goto err_hw;
+
+	rc = pfe_hif_lib_init(pfe);
+	if (rc < 0)
+		goto err_hif_lib;
+
+	rc = pfe_hif_init(pfe);
+	if (rc < 0)
+		goto err_hif;
+
+	rc = pfe_firmware_init(pfe);
+	if (rc < 0)
+		goto err_firmware;
+
+	rc = pfe_ctrl_init(pfe);
+	if (rc < 0)
+		goto err_ctrl;
+
+	rc = pfe_eth_init(pfe);
+	if (rc < 0)
+		goto err_eth;
+
+	rc = pfe_sysfs_init(pfe);
+	if (rc < 0)
+		goto err_sysfs;
+
+	rc = pfe_debugfs_init(pfe);
+	if (rc < 0)
+		goto err_debugfs;
+
+	return 0;
+
+err_debugfs:
+	pfe_sysfs_exit(pfe);
+
+err_sysfs:
+	pfe_eth_exit(pfe);
+
+err_eth:
+	pfe_ctrl_exit(pfe);
+
+err_ctrl:
+	pfe_firmware_exit(pfe);
+
+err_firmware:
+	pfe_hif_exit(pfe);
+
+err_hif:
+	pfe_hif_lib_exit(pfe);
+
+err_hif_lib:
+	pfe_hw_exit(pfe);
+
+err_hw:
+	return rc;
+}
+
+/*
+ * pfe_remove -
+ */
+int pfe_remove(struct pfe *pfe)
+{
+	pr_info("%s\n", __func__);
+
+	pfe_debugfs_exit(pfe);
+
+	pfe_sysfs_exit(pfe);
+
+	pfe_eth_exit(pfe);
+
+	pfe_ctrl_exit(pfe);
+
+#if defined(LS1012A_PFE_RESET_WA)
+	pfe_hif_rx_idle(&pfe->hif);
+#endif
+	pfe_firmware_exit(pfe);
+
+	pfe_hif_exit(pfe);
+
+	pfe_hif_lib_exit(pfe);
+
+	pfe_hw_exit(pfe);
+
+	return 0;
+}
diff --git a/drivers/staging/fsl_ppfe/pfe_sysfs.c b/drivers/staging/fsl_ppfe/pfe_sysfs.c
new file mode 100644
index 0000000..2a76384
--- /dev/null
+++ b/drivers/staging/fsl_ppfe/pfe_sysfs.c
@@ -0,0 +1,818 @@
+/*
+ * Copyright 2015-2016 Freescale Semiconductor, Inc.
+ * Copyright 2017 NXP
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/module.h>
+#include <linux/platform_device.h>
+
+#include "pfe_mod.h"
+
+#define PE_EXCEPTION_DUMP_ADDRESS 0x1fa8
+#define NUM_QUEUES		16
+
+static char register_name[20][5] = {
+	"EPC", "ECAS", "EID", "ED",
+	"r0", "r1", "r2", "r3",
+	"r4", "r5", "r6", "r7",
+	"r8", "r9", "r10", "r11",
+	"r12", "r13", "r14", "r15",
+};
+
+static char exception_name[14][20] = {
+	"Reset",
+	"HardwareFailure",
+	"NMI",
+	"InstBreakpoint",
+	"DataBreakpoint",
+	"Unsupported",
+	"PrivilegeViolation",
+	"InstBusError",
+	"DataBusError",
+	"AlignmentError",
+	"ArithmeticError",
+	"SystemCall",
+	"MemoryManagement",
+	"Interrupt",
+};
+
+static unsigned long class_do_clear;
+static unsigned long tmu_do_clear;
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+static unsigned long util_do_clear;
+#endif
+
+static ssize_t display_pe_status(char *buf, int id, u32 dmem_addr, unsigned long
+					do_clear)
+{
+	ssize_t len = 0;
+	u32 val;
+	char statebuf[5];
+	struct pfe_cpumon *cpumon = &pfe->cpumon;
+	u32 debug_indicator;
+	u32 debug[20];
+
+	*(u32 *)statebuf = pe_dmem_read(id, dmem_addr, 4);
+	dmem_addr += 4;
+
+	statebuf[4] = '\0';
+	len += sprintf(buf + len, "state=%4s ", statebuf);
+
+	val = pe_dmem_read(id, dmem_addr, 4);
+	dmem_addr += 4;
+	len += sprintf(buf + len, "ctr=%08x ", cpu_to_be32(val));
+
+	val = pe_dmem_read(id, dmem_addr, 4);
+	if (do_clear && val)
+		pe_dmem_write(id, 0, dmem_addr, 4);
+	dmem_addr += 4;
+	len += sprintf(buf + len, "rx=%u ", cpu_to_be32(val));
+
+	val = pe_dmem_read(id, dmem_addr, 4);
+	if (do_clear && val)
+		pe_dmem_write(id, 0, dmem_addr, 4);
+	dmem_addr += 4;
+	if (id >= TMU0_ID && id <= TMU_MAX_ID)
+		len += sprintf(buf + len, "qstatus=%x", cpu_to_be32(val));
+	else
+		len += sprintf(buf + len, "tx=%u", cpu_to_be32(val));
+
+	val = pe_dmem_read(id, dmem_addr, 4);
+	if (do_clear && val)
+		pe_dmem_write(id, 0, dmem_addr, 4);
+	dmem_addr += 4;
+	if (val)
+		len += sprintf(buf + len, " drop=%u", cpu_to_be32(val));
+
+	len += sprintf(buf + len, " load=%d%%", cpumon->cpu_usage_pct[id]);
+
+	len += sprintf(buf + len, "\n");
+
+	debug_indicator = pe_dmem_read(id, dmem_addr, 4);
+	dmem_addr += 4;
+	if (!strncmp((char *)&debug_indicator, "DBUG", 4)) {
+		int j, last = 0;
+
+		for (j = 0; j < 16; j++) {
+			debug[j] = pe_dmem_read(id, dmem_addr, 4);
+			if (debug[j]) {
+				if (do_clear)
+					pe_dmem_write(id, 0, dmem_addr, 4);
+				last = j + 1;
+			}
+			dmem_addr += 4;
+		}
+		for (j = 0; j < last; j++) {
+			len += sprintf(buf + len, "%08x%s",
+			cpu_to_be32(debug[j]),
+			(j & 0x7) == 0x7 || j == last - 1 ? "\n" : " ");
+		}
+	}
+
+	if (!strncmp(statebuf, "DEAD", 4)) {
+		u32 i, dump = PE_EXCEPTION_DUMP_ADDRESS;
+
+		len += sprintf(buf + len, "Exception details:\n");
+		for (i = 0; i < 20; i++) {
+			debug[i] = pe_dmem_read(id, dump, 4);
+			dump += 4;
+			if (i == 2)
+				len += sprintf(buf + len, "%4s = %08x (=%s) ",
+				register_name[i], cpu_to_be32(debug[i]),
+				exception_name[min((u32)
+				cpu_to_be32(debug[i]), (u32)13)]);
+			else
+				len += sprintf(buf + len, "%4s = %08x%s",
+				register_name[i], cpu_to_be32(debug[i]),
+				(i & 0x3) == 0x3 || i == 19 ? "\n" : " ");
+		}
+	}
+
+	return len;
+}
+
+static ssize_t class_phy_stats(char *buf, int phy)
+{
+	ssize_t len = 0;
+	int off1 = phy * 0x28;
+	int off2 = phy * 0x10;
+
+	if (phy == 3)
+		off1 = CLASS_PHY4_RX_PKTS - CLASS_PHY1_RX_PKTS;
+
+	len += sprintf(buf + len, "phy: %d\n", phy);
+	len += sprintf(buf + len,
+			"  rx:   %10u, tx:   %10u, intf:  %10u, ipv4:    %10u, ipv6: %10u\n",
+			readl(CLASS_PHY1_RX_PKTS + off1),
+			readl(CLASS_PHY1_TX_PKTS + off1),
+			readl(CLASS_PHY1_INTF_MATCH_PKTS + off1),
+			readl(CLASS_PHY1_V4_PKTS + off1),
+			readl(CLASS_PHY1_V6_PKTS + off1));
+
+	len += sprintf(buf + len,
+			"  icmp: %10u, igmp: %10u, tcp:   %10u, udp:     %10u\n",
+			readl(CLASS_PHY1_ICMP_PKTS + off2),
+			readl(CLASS_PHY1_IGMP_PKTS + off2),
+			readl(CLASS_PHY1_TCP_PKTS + off2),
+			readl(CLASS_PHY1_UDP_PKTS + off2));
+
+	len += sprintf(buf + len, "  err\n");
+	len += sprintf(buf + len,
+			"  lp:   %10u, intf: %10u, l3:    %10u, chcksum: %10u, ttl:  %10u\n",
+			readl(CLASS_PHY1_LP_FAIL_PKTS + off1),
+			readl(CLASS_PHY1_INTF_FAIL_PKTS + off1),
+			readl(CLASS_PHY1_L3_FAIL_PKTS + off1),
+			readl(CLASS_PHY1_CHKSUM_ERR_PKTS + off1),
+			readl(CLASS_PHY1_TTL_ERR_PKTS + off1));
+
+	return len;
+}
+
+/* qm_read_drop_stat
+ * This function is used to read the drop statistics from the TMU
+ * hw drop counter.  Since the hw counter is always cleared afer
+ * reading, this function maintains the previous drop count, and
+ * adds the new value to it.  That value can be retrieved by
+ * passing a pointer to it with the total_drops arg.
+ *
+ * @param tmu		TMU number (0 - 3)
+ * @param queue		queue number (0 - 15)
+ * @param total_drops	pointer to location to store total drops (or NULL)
+ * @param do_reset	if TRUE, clear total drops after updating
+ */
+u32 qm_read_drop_stat(u32 tmu, u32 queue, u32 *total_drops, int do_reset)
+{
+	static u32 qtotal[TMU_MAX_ID + 1][NUM_QUEUES];
+	u32 val;
+
+	writel((tmu << 8) | queue, TMU_TEQ_CTRL);
+	writel((tmu << 8) | queue, TMU_LLM_CTRL);
+	val = readl(TMU_TEQ_DROP_STAT);
+	qtotal[tmu][queue] += val;
+	if (total_drops)
+		*total_drops = qtotal[tmu][queue];
+	if (do_reset)
+		qtotal[tmu][queue] = 0;
+	return val;
+}
+
+static ssize_t tmu_queue_stats(char *buf, int tmu, int queue)
+{
+	ssize_t len = 0;
+	u32 drops;
+
+	len += sprintf(buf + len, "%d-%02d, ", tmu, queue);
+
+	drops = qm_read_drop_stat(tmu, queue, NULL, 0);
+
+	/* Select queue */
+	writel((tmu << 8) | queue, TMU_TEQ_CTRL);
+	writel((tmu << 8) | queue, TMU_LLM_CTRL);
+
+	len += sprintf(buf + len,
+			"(teq) drop: %10u, tx: %10u (llm) head: %08x, tail: %08x, drop: %10u\n",
+		drops, readl(TMU_TEQ_TRANS_STAT),
+		readl(TMU_LLM_QUE_HEADPTR), readl(TMU_LLM_QUE_TAILPTR),
+		readl(TMU_LLM_QUE_DROPCNT));
+
+	return len;
+}
+
+static ssize_t tmu_queues(char *buf, int tmu)
+{
+	ssize_t len = 0;
+	int queue;
+
+	for (queue = 0; queue < 16; queue++)
+		len += tmu_queue_stats(buf + len, tmu, queue);
+
+	return len;
+}
+
+static ssize_t block_version(char *buf, void *addr)
+{
+	ssize_t len = 0;
+	u32 val;
+
+	val = readl(addr);
+	len += sprintf(buf + len, "revision: %x, version: %x, id: %x\n",
+		(val >> 24) & 0xff, (val >> 16) & 0xff, val & 0xffff);
+
+	return len;
+}
+
+static ssize_t bmu(char *buf, int id, void *base)
+{
+	ssize_t len = 0;
+
+	len += sprintf(buf + len, "%s: %d\n  ", __func__, id);
+
+	len += block_version(buf + len, base + BMU_VERSION);
+
+	len += sprintf(buf + len, "  buf size:  %x\n", (1 << readl(base +
+			BMU_BUF_SIZE)));
+	len += sprintf(buf + len, "  buf count: %x\n", readl(base +
+			BMU_BUF_CNT));
+	len += sprintf(buf + len, "  buf rem:   %x\n", readl(base +
+			BMU_REM_BUF_CNT));
+	len += sprintf(buf + len, "  buf curr:  %x\n", readl(base +
+			BMU_CURR_BUF_CNT));
+	len += sprintf(buf + len, "  free err:  %x\n", readl(base +
+			BMU_FREE_ERR_ADDR));
+
+	return len;
+}
+
+static ssize_t gpi(char *buf, int id, void *base)
+{
+	ssize_t len = 0;
+	u32 val;
+
+	len += sprintf(buf + len, "%s%d:\n  ", __func__, id);
+	len += block_version(buf + len, base + GPI_VERSION);
+
+	len += sprintf(buf + len, "  tx under stick: %x\n", readl(base +
+			GPI_FIFO_STATUS));
+	val = readl(base + GPI_FIFO_DEBUG);
+	len += sprintf(buf + len, "  tx pkts:        %x\n", (val >> 23) &
+			0x3f);
+	len += sprintf(buf + len, "  rx pkts:        %x\n", (val >> 18) &
+			0x3f);
+	len += sprintf(buf + len, "  tx bytes:       %x\n", (val >> 9) &
+			0x1ff);
+	len += sprintf(buf + len, "  rx bytes:       %x\n", (val >> 0) &
+			0x1ff);
+	len += sprintf(buf + len, "  overrun:        %x\n", readl(base +
+			GPI_OVERRUN_DROPCNT));
+
+	return len;
+}
+
+static ssize_t pfe_set_class(struct device *dev, struct device_attribute *attr,
+			     const char *buf, size_t count)
+{
+	class_do_clear = kstrtoul(buf, 0, 0);
+	return count;
+}
+
+static ssize_t pfe_show_class(struct device *dev, struct device_attribute *attr,
+			      char *buf)
+{
+	ssize_t len = 0;
+	int id;
+	u32 val;
+	struct pfe_cpumon *cpumon = &pfe->cpumon;
+
+	len += block_version(buf + len, CLASS_VERSION);
+
+	for (id = CLASS0_ID; id <= CLASS_MAX_ID; id++) {
+		len += sprintf(buf + len, "%d: ", id - CLASS0_ID);
+
+		val = readl(CLASS_PE0_DEBUG + id * 4);
+		len += sprintf(buf + len, "pc=1%04x ", val & 0xffff);
+
+		len += display_pe_status(buf + len, id, CLASS_DM_PESTATUS,
+						class_do_clear);
+	}
+	len += sprintf(buf + len, "aggregate load=%d%%\n\n",
+			cpumon->class_usage_pct);
+
+	len += sprintf(buf + len, "pe status:   0x%x\n",
+			readl(CLASS_PE_STATUS));
+	len += sprintf(buf + len, "max buf cnt: 0x%x   afull thres: 0x%x\n",
+			readl(CLASS_MAX_BUF_CNT), readl(CLASS_AFULL_THRES));
+	len += sprintf(buf + len, "tsq max cnt: 0x%x   tsq fifo thres: 0x%x\n",
+			readl(CLASS_TSQ_MAX_CNT), readl(CLASS_TSQ_FIFO_THRES));
+	len += sprintf(buf + len, "state:       0x%x\n", readl(CLASS_STATE));
+
+	len += class_phy_stats(buf + len, 0);
+	len += class_phy_stats(buf + len, 1);
+	len += class_phy_stats(buf + len, 2);
+	len += class_phy_stats(buf + len, 3);
+
+	return len;
+}
+
+static ssize_t pfe_set_tmu(struct device *dev, struct device_attribute *attr,
+			   const char *buf, size_t count)
+{
+	tmu_do_clear = kstrtoul(buf, 0, 0);
+	return count;
+}
+
+static ssize_t pfe_show_tmu(struct device *dev, struct device_attribute *attr,
+			    char *buf)
+{
+	ssize_t len = 0;
+	int id;
+	u32 val;
+
+	len += block_version(buf + len, TMU_VERSION);
+
+	for (id = TMU0_ID; id <= TMU_MAX_ID; id++) {
+		if (id == TMU2_ID)
+			continue;
+		len += sprintf(buf + len, "%d: ", id - TMU0_ID);
+
+		len += display_pe_status(buf + len, id, TMU_DM_PESTATUS,
+						tmu_do_clear);
+	}
+
+	len += sprintf(buf + len, "pe status:    %x\n", readl(TMU_PE_STATUS));
+	len += sprintf(buf + len, "inq fifo cnt: %x\n",
+			readl(TMU_PHY_INQ_FIFO_CNT));
+	val = readl(TMU_INQ_STAT);
+	len += sprintf(buf + len, "inq wr ptr:     %x\n", val & 0x3ff);
+	len += sprintf(buf + len, "inq rd ptr:     %x\n", val >> 10);
+
+	return len;
+}
+
+static unsigned long drops_do_clear;
+static u32 class_drop_counter[CLASS_NUM_DROP_COUNTERS];
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+static u32 util_drop_counter[UTIL_NUM_DROP_COUNTERS];
+#endif
+
+char *class_drop_description[CLASS_NUM_DROP_COUNTERS] = {
+	"ICC",
+	"Host Pkt Error",
+	"Rx Error",
+	"IPsec Outbound",
+	"IPsec Inbound",
+	"EXPT IPsec Error",
+	"Reassembly",
+	"Fragmenter",
+	"NAT-T",
+	"Socket",
+	"Multicast",
+	"NAT-PT",
+	"Tx Disabled",
+};
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+char *util_drop_description[UTIL_NUM_DROP_COUNTERS] = {
+	"IPsec Outbound",
+	"IPsec Inbound",
+	"IPsec Rate Limiter",
+	"Fragmenter",
+	"Socket",
+	"Tx Disabled",
+	"Rx Error",
+};
+#endif
+
+static ssize_t pfe_set_drops(struct device *dev, struct device_attribute *attr,
+			     const char *buf, size_t count)
+{
+	drops_do_clear = kstrtoul(buf, 0, 0);
+	return count;
+}
+
+static u32 tmu_drops[4][16];
+static ssize_t pfe_show_drops(struct device *dev, struct device_attribute *attr,
+			      char *buf)
+{
+	ssize_t len = 0;
+	int id, dropnum;
+	int tmu, queue;
+	u32 val;
+	u32 dmem_addr;
+	int num_class_drops = 0, num_tmu_drops = 0, num_util_drops = 0;
+	struct pfe_ctrl *ctrl = &pfe->ctrl;
+
+	memset(class_drop_counter, 0, sizeof(class_drop_counter));
+	for (id = CLASS0_ID; id <= CLASS_MAX_ID; id++) {
+		if (drops_do_clear)
+			pe_sync_stop(ctrl, (1 << id));
+		for (dropnum = 0; dropnum < CLASS_NUM_DROP_COUNTERS;
+			dropnum++) {
+			dmem_addr = CLASS_DM_DROP_CNTR;
+			val = be32_to_cpu(pe_dmem_read(id, dmem_addr, 4));
+			class_drop_counter[dropnum] += val;
+			num_class_drops += val;
+			if (drops_do_clear)
+				pe_dmem_write(id, 0, dmem_addr, 4);
+		}
+		if (drops_do_clear)
+			pe_start(ctrl, (1 << id));
+	}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	if (drops_do_clear)
+		pe_sync_stop(ctrl, (1 << UTIL_ID));
+	for (dropnum = 0; dropnum < UTIL_NUM_DROP_COUNTERS; dropnum++) {
+		dmem_addr = UTIL_DM_DROP_CNTR;
+		val = be32_to_cpu(pe_dmem_read(UTIL_ID, dmem_addr, 4));
+		util_drop_counter[dropnum] = val;
+		num_util_drops += val;
+		if (drops_do_clear)
+			pe_dmem_write(UTIL_ID, 0, dmem_addr, 4);
+	}
+	if (drops_do_clear)
+		pe_start(ctrl, (1 << UTIL_ID));
+#endif
+	for (tmu = 0; tmu < 4; tmu++) {
+		for (queue = 0; queue < 16; queue++) {
+			qm_read_drop_stat(tmu, queue, &tmu_drops[tmu][queue],
+					  drops_do_clear);
+			num_tmu_drops += tmu_drops[tmu][queue];
+		}
+	}
+
+	if (num_class_drops == 0 && num_util_drops == 0 && num_tmu_drops == 0)
+		len += sprintf(buf + len, "No PE drops\n\n");
+
+	if (num_class_drops > 0) {
+		len += sprintf(buf + len, "Class PE drops --\n");
+		for (dropnum = 0; dropnum < CLASS_NUM_DROP_COUNTERS;
+			dropnum++) {
+			if (class_drop_counter[dropnum] > 0)
+				len += sprintf(buf + len, "  %s: %d\n",
+					class_drop_description[dropnum],
+					class_drop_counter[dropnum]);
+		}
+		len += sprintf(buf + len, "\n");
+	}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	if (num_util_drops > 0) {
+		len += sprintf(buf + len, "Util PE drops --\n");
+		for (dropnum = 0; dropnum < UTIL_NUM_DROP_COUNTERS; dropnum++) {
+			if (util_drop_counter[dropnum] > 0)
+				len += sprintf(buf + len, "  %s: %d\n",
+					util_drop_description[dropnum],
+					util_drop_counter[dropnum]);
+		}
+		len += sprintf(buf + len, "\n");
+	}
+#endif
+	if (num_tmu_drops > 0) {
+		len += sprintf(buf + len, "TMU drops --\n");
+		for (tmu = 0; tmu < 4; tmu++) {
+			for (queue = 0; queue < 16; queue++) {
+				if (tmu_drops[tmu][queue] > 0)
+					len += sprintf(buf + len,
+						"  TMU%d-Q%d: %d\n"
+					, tmu, queue, tmu_drops[tmu][queue]);
+			}
+		}
+		len += sprintf(buf + len, "\n");
+	}
+
+	return len;
+}
+
+static ssize_t pfe_show_tmu0_queues(struct device *dev, struct device_attribute
+					*attr, char *buf)
+{
+	return tmu_queues(buf, 0);
+}
+
+static ssize_t pfe_show_tmu1_queues(struct device *dev, struct device_attribute
+					*attr, char *buf)
+{
+	return tmu_queues(buf, 1);
+}
+
+static ssize_t pfe_show_tmu2_queues(struct device *dev, struct device_attribute
+					*attr, char *buf)
+{
+	return tmu_queues(buf, 2);
+}
+
+static ssize_t pfe_show_tmu3_queues(struct device *dev, struct device_attribute
+					*attr, char *buf)
+{
+	return tmu_queues(buf, 3);
+}
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+static ssize_t pfe_set_util(struct device *dev, struct device_attribute *attr,
+			    const char *buf, size_t count)
+{
+	util_do_clear = kstrtoul(buf, NULL, 0);
+	return count;
+}
+
+static ssize_t pfe_show_util(struct device *dev, struct device_attribute *attr,
+			     char *buf)
+{
+	ssize_t len = 0;
+	struct pfe_ctrl *ctrl = &pfe->ctrl;
+
+	len += block_version(buf + len, UTIL_VERSION);
+
+	pe_sync_stop(ctrl, (1 << UTIL_ID));
+	len += display_pe_status(buf + len, UTIL_ID, UTIL_DM_PESTATUS,
+					util_do_clear);
+	pe_start(ctrl, (1 << UTIL_ID));
+
+	len += sprintf(buf + len, "pe status:   %x\n", readl(UTIL_PE_STATUS));
+	len += sprintf(buf + len, "max buf cnt: %x\n",
+			readl(UTIL_MAX_BUF_CNT));
+	len += sprintf(buf + len, "tsq max cnt: %x\n",
+			readl(UTIL_TSQ_MAX_CNT));
+
+	return len;
+}
+#endif
+
+static ssize_t pfe_show_bmu(struct device *dev, struct device_attribute *attr,
+			    char *buf)
+{
+	ssize_t len = 0;
+
+	len += bmu(buf + len, 1, BMU1_BASE_ADDR);
+	len += bmu(buf + len, 2, BMU2_BASE_ADDR);
+
+	return len;
+}
+
+static ssize_t pfe_show_hif(struct device *dev, struct device_attribute *attr,
+			    char *buf)
+{
+	ssize_t len = 0;
+
+	len += sprintf(buf + len, "hif:\n  ");
+	len += block_version(buf + len, HIF_VERSION);
+
+	len += sprintf(buf + len, "  tx curr bd:    %x\n",
+			readl(HIF_TX_CURR_BD_ADDR));
+	len += sprintf(buf + len, "  tx status:     %x\n",
+			readl(HIF_TX_STATUS));
+	len += sprintf(buf + len, "  tx dma status: %x\n",
+			readl(HIF_TX_DMA_STATUS));
+
+	len += sprintf(buf + len, "  rx curr bd:    %x\n",
+			readl(HIF_RX_CURR_BD_ADDR));
+	len += sprintf(buf + len, "  rx status:     %x\n",
+			readl(HIF_RX_STATUS));
+	len += sprintf(buf + len, "  rx dma status: %x\n",
+			readl(HIF_RX_DMA_STATUS));
+
+	len += sprintf(buf + len, "hif nocopy:\n  ");
+	len += block_version(buf + len, HIF_NOCPY_VERSION);
+
+	len += sprintf(buf + len, "  tx curr bd:    %x\n",
+			readl(HIF_NOCPY_TX_CURR_BD_ADDR));
+	len += sprintf(buf + len, "  tx status:     %x\n",
+			readl(HIF_NOCPY_TX_STATUS));
+	len += sprintf(buf + len, "  tx dma status: %x\n",
+			readl(HIF_NOCPY_TX_DMA_STATUS));
+
+	len += sprintf(buf + len, "  rx curr bd:    %x\n",
+			readl(HIF_NOCPY_RX_CURR_BD_ADDR));
+	len += sprintf(buf + len, "  rx status:     %x\n",
+			readl(HIF_NOCPY_RX_STATUS));
+	len += sprintf(buf + len, "  rx dma status: %x\n",
+			readl(HIF_NOCPY_RX_DMA_STATUS));
+
+	return len;
+}
+
+static ssize_t pfe_show_gpi(struct device *dev, struct device_attribute *attr,
+			    char *buf)
+{
+	ssize_t len = 0;
+
+	len += gpi(buf + len, 0, EGPI1_BASE_ADDR);
+	len += gpi(buf + len, 1, EGPI2_BASE_ADDR);
+	len += gpi(buf + len, 3, HGPI_BASE_ADDR);
+
+	return len;
+}
+
+static ssize_t pfe_show_pfemem(struct device *dev, struct device_attribute
+				*attr, char *buf)
+{
+	ssize_t len = 0;
+	struct pfe_memmon *memmon = &pfe->memmon;
+
+	len += sprintf(buf + len, "Kernel Memory: %d Bytes (%d KB)\n",
+		memmon->kernel_memory_allocated,
+		(memmon->kernel_memory_allocated + 1023) / 1024);
+
+	return len;
+}
+
+#ifdef HIF_NAPI_STATS
+static ssize_t pfe_show_hif_napi_stats(struct device *dev,
+				       struct device_attribute *attr,
+				       char *buf)
+{
+	struct platform_device *pdev = to_platform_device(dev);
+	struct pfe *pfe = platform_get_drvdata(pdev);
+	ssize_t len = 0;
+
+	len += sprintf(buf + len, "sched:  %u\n",
+			pfe->hif.napi_counters[NAPI_SCHED_COUNT]);
+	len += sprintf(buf + len, "poll:   %u\n",
+			pfe->hif.napi_counters[NAPI_POLL_COUNT]);
+	len += sprintf(buf + len, "packet: %u\n",
+			pfe->hif.napi_counters[NAPI_PACKET_COUNT]);
+	len += sprintf(buf + len, "budget: %u\n",
+			pfe->hif.napi_counters[NAPI_FULL_BUDGET_COUNT]);
+	len += sprintf(buf + len, "desc:   %u\n",
+			pfe->hif.napi_counters[NAPI_DESC_COUNT]);
+	len += sprintf(buf + len, "full:   %u\n",
+			pfe->hif.napi_counters[NAPI_CLIENT_FULL_COUNT]);
+
+	return len;
+}
+
+static ssize_t pfe_set_hif_napi_stats(struct device *dev,
+				      struct device_attribute *attr,
+					const char *buf, size_t count)
+{
+	struct platform_device *pdev = to_platform_device(dev);
+	struct pfe *pfe = platform_get_drvdata(pdev);
+
+	memset(pfe->hif.napi_counters, 0, sizeof(pfe->hif.napi_counters));
+
+	return count;
+}
+
+static DEVICE_ATTR(hif_napi_stats, 0644, pfe_show_hif_napi_stats,
+			pfe_set_hif_napi_stats);
+#endif
+
+static DEVICE_ATTR(class, 0644, pfe_show_class, pfe_set_class);
+static DEVICE_ATTR(tmu, 0644, pfe_show_tmu, pfe_set_tmu);
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+static DEVICE_ATTR(util, 0644, pfe_show_util, pfe_set_util);
+#endif
+static DEVICE_ATTR(bmu, 0444, pfe_show_bmu, NULL);
+static DEVICE_ATTR(hif, 0444, pfe_show_hif, NULL);
+static DEVICE_ATTR(gpi, 0444, pfe_show_gpi, NULL);
+static DEVICE_ATTR(drops, 0644, pfe_show_drops, pfe_set_drops);
+static DEVICE_ATTR(tmu0_queues, 0444, pfe_show_tmu0_queues, NULL);
+static DEVICE_ATTR(tmu1_queues, 0444, pfe_show_tmu1_queues, NULL);
+static DEVICE_ATTR(tmu2_queues, 0444, pfe_show_tmu2_queues, NULL);
+static DEVICE_ATTR(tmu3_queues, 0444, pfe_show_tmu3_queues, NULL);
+static DEVICE_ATTR(pfemem, 0444, pfe_show_pfemem, NULL);
+
+int pfe_sysfs_init(struct pfe *pfe)
+{
+	if (device_create_file(pfe->dev, &dev_attr_class))
+		goto err_class;
+
+	if (device_create_file(pfe->dev, &dev_attr_tmu))
+		goto err_tmu;
+
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	if (device_create_file(pfe->dev, &dev_attr_util))
+		goto err_util;
+#endif
+
+	if (device_create_file(pfe->dev, &dev_attr_bmu))
+		goto err_bmu;
+
+	if (device_create_file(pfe->dev, &dev_attr_hif))
+		goto err_hif;
+
+	if (device_create_file(pfe->dev, &dev_attr_gpi))
+		goto err_gpi;
+
+	if (device_create_file(pfe->dev, &dev_attr_drops))
+		goto err_drops;
+
+	if (device_create_file(pfe->dev, &dev_attr_tmu0_queues))
+		goto err_tmu0_queues;
+
+	if (device_create_file(pfe->dev, &dev_attr_tmu1_queues))
+		goto err_tmu1_queues;
+
+	if (device_create_file(pfe->dev, &dev_attr_tmu2_queues))
+		goto err_tmu2_queues;
+
+	if (device_create_file(pfe->dev, &dev_attr_tmu3_queues))
+		goto err_tmu3_queues;
+
+	if (device_create_file(pfe->dev, &dev_attr_pfemem))
+		goto err_pfemem;
+
+#ifdef HIF_NAPI_STATS
+	if (device_create_file(pfe->dev, &dev_attr_hif_napi_stats))
+		goto err_hif_napi_stats;
+#endif
+
+	return 0;
+
+#ifdef HIF_NAPI_STATS
+err_hif_napi_stats:
+	device_remove_file(pfe->dev, &dev_attr_pfemem);
+#endif
+
+err_pfemem:
+	device_remove_file(pfe->dev, &dev_attr_tmu3_queues);
+
+err_tmu3_queues:
+	device_remove_file(pfe->dev, &dev_attr_tmu2_queues);
+
+err_tmu2_queues:
+	device_remove_file(pfe->dev, &dev_attr_tmu1_queues);
+
+err_tmu1_queues:
+	device_remove_file(pfe->dev, &dev_attr_tmu0_queues);
+
+err_tmu0_queues:
+	device_remove_file(pfe->dev, &dev_attr_drops);
+
+err_drops:
+	device_remove_file(pfe->dev, &dev_attr_gpi);
+
+err_gpi:
+	device_remove_file(pfe->dev, &dev_attr_hif);
+
+err_hif:
+	device_remove_file(pfe->dev, &dev_attr_bmu);
+
+err_bmu:
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	device_remove_file(pfe->dev, &dev_attr_util);
+
+err_util:
+#endif
+	device_remove_file(pfe->dev, &dev_attr_tmu);
+
+err_tmu:
+	device_remove_file(pfe->dev, &dev_attr_class);
+
+err_class:
+	return -1;
+}
+
+void pfe_sysfs_exit(struct pfe *pfe)
+{
+#ifdef HIF_NAPI_STATS
+	device_remove_file(pfe->dev, &dev_attr_hif_napi_stats);
+#endif
+	device_remove_file(pfe->dev, &dev_attr_pfemem);
+	device_remove_file(pfe->dev, &dev_attr_tmu3_queues);
+	device_remove_file(pfe->dev, &dev_attr_tmu2_queues);
+	device_remove_file(pfe->dev, &dev_attr_tmu1_queues);
+	device_remove_file(pfe->dev, &dev_attr_tmu0_queues);
+	device_remove_file(pfe->dev, &dev_attr_drops);
+	device_remove_file(pfe->dev, &dev_attr_gpi);
+	device_remove_file(pfe->dev, &dev_attr_hif);
+	device_remove_file(pfe->dev, &dev_attr_bmu);
+#if !defined(CONFIG_FSL_PPFE_UTIL_DISABLED)
+	device_remove_file(pfe->dev, &dev_attr_util);
+#endif
+	device_remove_file(pfe->dev, &dev_attr_tmu);
+	device_remove_file(pfe->dev, &dev_attr_class);
+}