Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / net / ethernet / stmicro / stmmac / stmmac_main.c
diff --git a/kernel/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/kernel/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
new file mode 100644 (file)
index 0000000..2c5ce2b
--- /dev/null
@@ -0,0 +1,3184 @@
+/*******************************************************************************
+  This is the driver for the ST MAC 10/100/1000 on-chip Ethernet controllers.
+  ST Ethernet IPs are built around a Synopsys IP Core.
+
+       Copyright(C) 2007-2011 STMicroelectronics Ltd
+
+  This program is free software; you can redistribute it and/or modify it
+  under the terms and conditions of the GNU General Public License,
+  version 2, as published by the Free Software Foundation.
+
+  This program is distributed in the hope 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, write to the Free Software Foundation, Inc.,
+  51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+
+  The full GNU General Public License is included in this distribution in
+  the file called "COPYING".
+
+  Author: Giuseppe Cavallaro <peppe.cavallaro@st.com>
+
+  Documentation available at:
+       http://www.stlinux.com
+  Support available at:
+       https://bugzilla.stlinux.com/
+*******************************************************************************/
+
+#include <linux/clk.h>
+#include <linux/kernel.h>
+#include <linux/interrupt.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/skbuff.h>
+#include <linux/ethtool.h>
+#include <linux/if_ether.h>
+#include <linux/crc32.h>
+#include <linux/mii.h>
+#include <linux/if.h>
+#include <linux/if_vlan.h>
+#include <linux/dma-mapping.h>
+#include <linux/slab.h>
+#include <linux/prefetch.h>
+#include <linux/pinctrl/consumer.h>
+#ifdef CONFIG_DEBUG_FS
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#endif /* CONFIG_DEBUG_FS */
+#include <linux/net_tstamp.h>
+#include "stmmac_ptp.h"
+#include "stmmac.h"
+#include <linux/reset.h>
+
+#define STMMAC_ALIGN(x)        L1_CACHE_ALIGN(x)
+
+/* Module parameters */
+#define TX_TIMEO       5000
+static int watchdog = TX_TIMEO;
+module_param(watchdog, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(watchdog, "Transmit timeout in milliseconds (default 5s)");
+
+static int debug = -1;
+module_param(debug, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(debug, "Message Level (-1: default, 0: no output, 16: all)");
+
+static int phyaddr = -1;
+module_param(phyaddr, int, S_IRUGO);
+MODULE_PARM_DESC(phyaddr, "Physical device address");
+
+#define DMA_TX_SIZE 256
+static int dma_txsize = DMA_TX_SIZE;
+module_param(dma_txsize, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(dma_txsize, "Number of descriptors in the TX list");
+
+#define DMA_RX_SIZE 256
+static int dma_rxsize = DMA_RX_SIZE;
+module_param(dma_rxsize, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(dma_rxsize, "Number of descriptors in the RX list");
+
+static int flow_ctrl = FLOW_OFF;
+module_param(flow_ctrl, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(flow_ctrl, "Flow control ability [on/off]");
+
+static int pause = PAUSE_TIME;
+module_param(pause, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(pause, "Flow Control Pause Time");
+
+#define TC_DEFAULT 64
+static int tc = TC_DEFAULT;
+module_param(tc, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(tc, "DMA threshold control value");
+
+#define        DEFAULT_BUFSIZE 1536
+static int buf_sz = DEFAULT_BUFSIZE;
+module_param(buf_sz, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(buf_sz, "DMA buffer size");
+
+static const u32 default_msg_level = (NETIF_MSG_DRV | NETIF_MSG_PROBE |
+                                     NETIF_MSG_LINK | NETIF_MSG_IFUP |
+                                     NETIF_MSG_IFDOWN | NETIF_MSG_TIMER);
+
+#define STMMAC_DEFAULT_LPI_TIMER       1000
+static int eee_timer = STMMAC_DEFAULT_LPI_TIMER;
+module_param(eee_timer, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(eee_timer, "LPI tx expiration time in msec");
+#define STMMAC_LPI_T(x) (jiffies + msecs_to_jiffies(x))
+
+/* By default the driver will use the ring mode to manage tx and rx descriptors
+ * but passing this value so user can force to use the chain instead of the ring
+ */
+static unsigned int chain_mode;
+module_param(chain_mode, int, S_IRUGO);
+MODULE_PARM_DESC(chain_mode, "To use chain instead of ring mode");
+
+static irqreturn_t stmmac_interrupt(int irq, void *dev_id);
+
+#ifdef CONFIG_DEBUG_FS
+static int stmmac_init_fs(struct net_device *dev);
+static void stmmac_exit_fs(struct net_device *dev);
+#endif
+
+#define STMMAC_COAL_TIMER(x) (jiffies + usecs_to_jiffies(x))
+
+/**
+ * stmmac_verify_args - verify the driver parameters.
+ * Description: it checks the driver parameters and set a default in case of
+ * errors.
+ */
+static void stmmac_verify_args(void)
+{
+       if (unlikely(watchdog < 0))
+               watchdog = TX_TIMEO;
+       if (unlikely(dma_rxsize < 0))
+               dma_rxsize = DMA_RX_SIZE;
+       if (unlikely(dma_txsize < 0))
+               dma_txsize = DMA_TX_SIZE;
+       if (unlikely((buf_sz < DEFAULT_BUFSIZE) || (buf_sz > BUF_SIZE_16KiB)))
+               buf_sz = DEFAULT_BUFSIZE;
+       if (unlikely(flow_ctrl > 1))
+               flow_ctrl = FLOW_AUTO;
+       else if (likely(flow_ctrl < 0))
+               flow_ctrl = FLOW_OFF;
+       if (unlikely((pause < 0) || (pause > 0xffff)))
+               pause = PAUSE_TIME;
+       if (eee_timer < 0)
+               eee_timer = STMMAC_DEFAULT_LPI_TIMER;
+}
+
+/**
+ * stmmac_clk_csr_set - dynamically set the MDC clock
+ * @priv: driver private structure
+ * Description: this is to dynamically set the MDC clock according to the csr
+ * clock input.
+ * Note:
+ *     If a specific clk_csr value is passed from the platform
+ *     this means that the CSR Clock Range selection cannot be
+ *     changed at run-time and it is fixed (as reported in the driver
+ *     documentation). Viceversa the driver will try to set the MDC
+ *     clock dynamically according to the actual clock input.
+ */
+static void stmmac_clk_csr_set(struct stmmac_priv *priv)
+{
+       u32 clk_rate;
+
+       clk_rate = clk_get_rate(priv->stmmac_clk);
+
+       /* Platform provided default clk_csr would be assumed valid
+        * for all other cases except for the below mentioned ones.
+        * For values higher than the IEEE 802.3 specified frequency
+        * we can not estimate the proper divider as it is not known
+        * the frequency of clk_csr_i. So we do not change the default
+        * divider.
+        */
+       if (!(priv->clk_csr & MAC_CSR_H_FRQ_MASK)) {
+               if (clk_rate < CSR_F_35M)
+                       priv->clk_csr = STMMAC_CSR_20_35M;
+               else if ((clk_rate >= CSR_F_35M) && (clk_rate < CSR_F_60M))
+                       priv->clk_csr = STMMAC_CSR_35_60M;
+               else if ((clk_rate >= CSR_F_60M) && (clk_rate < CSR_F_100M))
+                       priv->clk_csr = STMMAC_CSR_60_100M;
+               else if ((clk_rate >= CSR_F_100M) && (clk_rate < CSR_F_150M))
+                       priv->clk_csr = STMMAC_CSR_100_150M;
+               else if ((clk_rate >= CSR_F_150M) && (clk_rate < CSR_F_250M))
+                       priv->clk_csr = STMMAC_CSR_150_250M;
+               else if ((clk_rate >= CSR_F_250M) && (clk_rate < CSR_F_300M))
+                       priv->clk_csr = STMMAC_CSR_250_300M;
+       }
+}
+
+static void print_pkt(unsigned char *buf, int len)
+{
+       pr_debug("len = %d byte, buf addr: 0x%p\n", len, buf);
+       print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, buf, len);
+}
+
+/* minimum number of free TX descriptors required to wake up TX process */
+#define STMMAC_TX_THRESH(x)    (x->dma_tx_size/4)
+
+static inline u32 stmmac_tx_avail(struct stmmac_priv *priv)
+{
+       return priv->dirty_tx + priv->dma_tx_size - priv->cur_tx - 1;
+}
+
+/**
+ * stmmac_hw_fix_mac_speed - callback for speed selection
+ * @priv: driver private structure
+ * Description: on some platforms (e.g. ST), some HW system configuraton
+ * registers have to be set according to the link speed negotiated.
+ */
+static inline void stmmac_hw_fix_mac_speed(struct stmmac_priv *priv)
+{
+       struct phy_device *phydev = priv->phydev;
+
+       if (likely(priv->plat->fix_mac_speed))
+               priv->plat->fix_mac_speed(priv->plat->bsp_priv, phydev->speed);
+}
+
+/**
+ * stmmac_enable_eee_mode - check and enter in LPI mode
+ * @priv: driver private structure
+ * Description: this function is to verify and enter in LPI mode in case of
+ * EEE.
+ */
+static void stmmac_enable_eee_mode(struct stmmac_priv *priv)
+{
+       /* Check and enter in LPI mode */
+       if ((priv->dirty_tx == priv->cur_tx) &&
+           (priv->tx_path_in_lpi_mode == false))
+               priv->hw->mac->set_eee_mode(priv->hw);
+}
+
+/**
+ * stmmac_disable_eee_mode - disable and exit from LPI mode
+ * @priv: driver private structure
+ * Description: this function is to exit and disable EEE in case of
+ * LPI state is true. This is called by the xmit.
+ */
+void stmmac_disable_eee_mode(struct stmmac_priv *priv)
+{
+       priv->hw->mac->reset_eee_mode(priv->hw);
+       del_timer_sync(&priv->eee_ctrl_timer);
+       priv->tx_path_in_lpi_mode = false;
+}
+
+/**
+ * stmmac_eee_ctrl_timer - EEE TX SW timer.
+ * @arg : data hook
+ * Description:
+ *  if there is no data transfer and if we are not in LPI state,
+ *  then MAC Transmitter can be moved to LPI state.
+ */
+static void stmmac_eee_ctrl_timer(unsigned long arg)
+{
+       struct stmmac_priv *priv = (struct stmmac_priv *)arg;
+
+       stmmac_enable_eee_mode(priv);
+       mod_timer(&priv->eee_ctrl_timer, STMMAC_LPI_T(eee_timer));
+}
+
+/**
+ * stmmac_eee_init - init EEE
+ * @priv: driver private structure
+ * Description:
+ *  if the GMAC supports the EEE (from the HW cap reg) and the phy device
+ *  can also manage EEE, this function enable the LPI state and start related
+ *  timer.
+ */
+bool stmmac_eee_init(struct stmmac_priv *priv)
+{
+       char *phy_bus_name = priv->plat->phy_bus_name;
+       unsigned long flags;
+       bool ret = false;
+
+       /* Using PCS we cannot dial with the phy registers at this stage
+        * so we do not support extra feature like EEE.
+        */
+       if ((priv->pcs == STMMAC_PCS_RGMII) || (priv->pcs == STMMAC_PCS_TBI) ||
+           (priv->pcs == STMMAC_PCS_RTBI))
+               goto out;
+
+       /* Never init EEE in case of a switch is attached */
+       if (phy_bus_name && (!strcmp(phy_bus_name, "fixed")))
+               goto out;
+
+       /* MAC core supports the EEE feature. */
+       if (priv->dma_cap.eee) {
+               int tx_lpi_timer = priv->tx_lpi_timer;
+
+               /* Check if the PHY supports EEE */
+               if (phy_init_eee(priv->phydev, 1)) {
+                       /* To manage at run-time if the EEE cannot be supported
+                        * anymore (for example because the lp caps have been
+                        * changed).
+                        * In that case the driver disable own timers.
+                        */
+                       spin_lock_irqsave(&priv->lock, flags);
+                       if (priv->eee_active) {
+                               pr_debug("stmmac: disable EEE\n");
+                               del_timer_sync(&priv->eee_ctrl_timer);
+                               priv->hw->mac->set_eee_timer(priv->hw, 0,
+                                                            tx_lpi_timer);
+                       }
+                       priv->eee_active = 0;
+                       spin_unlock_irqrestore(&priv->lock, flags);
+                       goto out;
+               }
+               /* Activate the EEE and start timers */
+               spin_lock_irqsave(&priv->lock, flags);
+               if (!priv->eee_active) {
+                       priv->eee_active = 1;
+                       setup_timer(&priv->eee_ctrl_timer,
+                                   stmmac_eee_ctrl_timer,
+                                   (unsigned long)priv);
+                       mod_timer(&priv->eee_ctrl_timer,
+                                 STMMAC_LPI_T(eee_timer));
+
+                       priv->hw->mac->set_eee_timer(priv->hw,
+                                                    STMMAC_DEFAULT_LIT_LS,
+                                                    tx_lpi_timer);
+               }
+               /* Set HW EEE according to the speed */
+               priv->hw->mac->set_eee_pls(priv->hw, priv->phydev->link);
+
+               ret = true;
+               spin_unlock_irqrestore(&priv->lock, flags);
+
+               pr_debug("stmmac: Energy-Efficient Ethernet initialized\n");
+       }
+out:
+       return ret;
+}
+
+/* stmmac_get_tx_hwtstamp - get HW TX timestamps
+ * @priv: driver private structure
+ * @entry : descriptor index to be used.
+ * @skb : the socket buffer
+ * Description :
+ * This function will read timestamp from the descriptor & pass it to stack.
+ * and also perform some sanity checks.
+ */
+static void stmmac_get_tx_hwtstamp(struct stmmac_priv *priv,
+                                  unsigned int entry, struct sk_buff *skb)
+{
+       struct skb_shared_hwtstamps shhwtstamp;
+       u64 ns;
+       void *desc = NULL;
+
+       if (!priv->hwts_tx_en)
+               return;
+
+       /* exit if skb doesn't support hw tstamp */
+       if (likely(!skb || !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)))
+               return;
+
+       if (priv->adv_ts)
+               desc = (priv->dma_etx + entry);
+       else
+               desc = (priv->dma_tx + entry);
+
+       /* check tx tstamp status */
+       if (!priv->hw->desc->get_tx_timestamp_status((struct dma_desc *)desc))
+               return;
+
+       /* get the valid tstamp */
+       ns = priv->hw->desc->get_timestamp(desc, priv->adv_ts);
+
+       memset(&shhwtstamp, 0, sizeof(struct skb_shared_hwtstamps));
+       shhwtstamp.hwtstamp = ns_to_ktime(ns);
+       /* pass tstamp to stack */
+       skb_tstamp_tx(skb, &shhwtstamp);
+
+       return;
+}
+
+/* stmmac_get_rx_hwtstamp - get HW RX timestamps
+ * @priv: driver private structure
+ * @entry : descriptor index to be used.
+ * @skb : the socket buffer
+ * Description :
+ * This function will read received packet's timestamp from the descriptor
+ * and pass it to stack. It also perform some sanity checks.
+ */
+static void stmmac_get_rx_hwtstamp(struct stmmac_priv *priv,
+                                  unsigned int entry, struct sk_buff *skb)
+{
+       struct skb_shared_hwtstamps *shhwtstamp = NULL;
+       u64 ns;
+       void *desc = NULL;
+
+       if (!priv->hwts_rx_en)
+               return;
+
+       if (priv->adv_ts)
+               desc = (priv->dma_erx + entry);
+       else
+               desc = (priv->dma_rx + entry);
+
+       /* exit if rx tstamp is not valid */
+       if (!priv->hw->desc->get_rx_timestamp_status(desc, priv->adv_ts))
+               return;
+
+       /* get valid tstamp */
+       ns = priv->hw->desc->get_timestamp(desc, priv->adv_ts);
+       shhwtstamp = skb_hwtstamps(skb);
+       memset(shhwtstamp, 0, sizeof(struct skb_shared_hwtstamps));
+       shhwtstamp->hwtstamp = ns_to_ktime(ns);
+}
+
+/**
+ *  stmmac_hwtstamp_ioctl - control hardware timestamping.
+ *  @dev: device pointer.
+ *  @ifr: An IOCTL specefic structure, that can contain a pointer to
+ *  a proprietary structure used to pass information to the driver.
+ *  Description:
+ *  This function configures the MAC to enable/disable both outgoing(TX)
+ *  and incoming(RX) packets time stamping based on user input.
+ *  Return Value:
+ *  0 on success and an appropriate -ve integer on failure.
+ */
+static int stmmac_hwtstamp_ioctl(struct net_device *dev, struct ifreq *ifr)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       struct hwtstamp_config config;
+       struct timespec now;
+       u64 temp = 0;
+       u32 ptp_v2 = 0;
+       u32 tstamp_all = 0;
+       u32 ptp_over_ipv4_udp = 0;
+       u32 ptp_over_ipv6_udp = 0;
+       u32 ptp_over_ethernet = 0;
+       u32 snap_type_sel = 0;
+       u32 ts_master_en = 0;
+       u32 ts_event_en = 0;
+       u32 value = 0;
+
+       if (!(priv->dma_cap.time_stamp || priv->adv_ts)) {
+               netdev_alert(priv->dev, "No support for HW time stamping\n");
+               priv->hwts_tx_en = 0;
+               priv->hwts_rx_en = 0;
+
+               return -EOPNOTSUPP;
+       }
+
+       if (copy_from_user(&config, ifr->ifr_data,
+                          sizeof(struct hwtstamp_config)))
+               return -EFAULT;
+
+       pr_debug("%s config flags:0x%x, tx_type:0x%x, rx_filter:0x%x\n",
+                __func__, config.flags, config.tx_type, config.rx_filter);
+
+       /* reserved for future extensions */
+       if (config.flags)
+               return -EINVAL;
+
+       if (config.tx_type != HWTSTAMP_TX_OFF &&
+           config.tx_type != HWTSTAMP_TX_ON)
+               return -ERANGE;
+
+       if (priv->adv_ts) {
+               switch (config.rx_filter) {
+               case HWTSTAMP_FILTER_NONE:
+                       /* time stamp no incoming packet at all */
+                       config.rx_filter = HWTSTAMP_FILTER_NONE;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
+                       /* PTP v1, UDP, any kind of event packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V1_L4_EVENT;
+                       /* take time stamp for all event messages */
+                       snap_type_sel = PTP_TCR_SNAPTYPSEL_1;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
+                       /* PTP v1, UDP, Sync packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V1_L4_SYNC;
+                       /* take time stamp for SYNC messages only */
+                       ts_event_en = PTP_TCR_TSEVNTENA;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
+                       /* PTP v1, UDP, Delay_req packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ;
+                       /* take time stamp for Delay_Req messages only */
+                       ts_master_en = PTP_TCR_TSMSTRENA;
+                       ts_event_en = PTP_TCR_TSEVNTENA;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
+                       /* PTP v2, UDP, any kind of event packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V2_L4_EVENT;
+                       ptp_v2 = PTP_TCR_TSVER2ENA;
+                       /* take time stamp for all event messages */
+                       snap_type_sel = PTP_TCR_SNAPTYPSEL_1;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
+                       /* PTP v2, UDP, Sync packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V2_L4_SYNC;
+                       ptp_v2 = PTP_TCR_TSVER2ENA;
+                       /* take time stamp for SYNC messages only */
+                       ts_event_en = PTP_TCR_TSEVNTENA;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
+                       /* PTP v2, UDP, Delay_req packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ;
+                       ptp_v2 = PTP_TCR_TSVER2ENA;
+                       /* take time stamp for Delay_Req messages only */
+                       ts_master_en = PTP_TCR_TSMSTRENA;
+                       ts_event_en = PTP_TCR_TSEVNTENA;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V2_EVENT:
+                       /* PTP v2/802.AS1 any layer, any kind of event packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
+                       ptp_v2 = PTP_TCR_TSVER2ENA;
+                       /* take time stamp for all event messages */
+                       snap_type_sel = PTP_TCR_SNAPTYPSEL_1;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       ptp_over_ethernet = PTP_TCR_TSIPENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V2_SYNC:
+                       /* PTP v2/802.AS1, any layer, Sync packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V2_SYNC;
+                       ptp_v2 = PTP_TCR_TSVER2ENA;
+                       /* take time stamp for SYNC messages only */
+                       ts_event_en = PTP_TCR_TSEVNTENA;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       ptp_over_ethernet = PTP_TCR_TSIPENA;
+                       break;
+
+               case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
+                       /* PTP v2/802.AS1, any layer, Delay_req packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V2_DELAY_REQ;
+                       ptp_v2 = PTP_TCR_TSVER2ENA;
+                       /* take time stamp for Delay_Req messages only */
+                       ts_master_en = PTP_TCR_TSMSTRENA;
+                       ts_event_en = PTP_TCR_TSEVNTENA;
+
+                       ptp_over_ipv4_udp = PTP_TCR_TSIPV4ENA;
+                       ptp_over_ipv6_udp = PTP_TCR_TSIPV6ENA;
+                       ptp_over_ethernet = PTP_TCR_TSIPENA;
+                       break;
+
+               case HWTSTAMP_FILTER_ALL:
+                       /* time stamp any incoming packet */
+                       config.rx_filter = HWTSTAMP_FILTER_ALL;
+                       tstamp_all = PTP_TCR_TSENALL;
+                       break;
+
+               default:
+                       return -ERANGE;
+               }
+       } else {
+               switch (config.rx_filter) {
+               case HWTSTAMP_FILTER_NONE:
+                       config.rx_filter = HWTSTAMP_FILTER_NONE;
+                       break;
+               default:
+                       /* PTP v1, UDP, any kind of event packet */
+                       config.rx_filter = HWTSTAMP_FILTER_PTP_V1_L4_EVENT;
+                       break;
+               }
+       }
+       priv->hwts_rx_en = ((config.rx_filter == HWTSTAMP_FILTER_NONE) ? 0 : 1);
+       priv->hwts_tx_en = config.tx_type == HWTSTAMP_TX_ON;
+
+       if (!priv->hwts_tx_en && !priv->hwts_rx_en)
+               priv->hw->ptp->config_hw_tstamping(priv->ioaddr, 0);
+       else {
+               value = (PTP_TCR_TSENA | PTP_TCR_TSCFUPDT | PTP_TCR_TSCTRLSSR |
+                        tstamp_all | ptp_v2 | ptp_over_ethernet |
+                        ptp_over_ipv6_udp | ptp_over_ipv4_udp | ts_event_en |
+                        ts_master_en | snap_type_sel);
+
+               priv->hw->ptp->config_hw_tstamping(priv->ioaddr, value);
+
+               /* program Sub Second Increment reg */
+               priv->hw->ptp->config_sub_second_increment(priv->ioaddr);
+
+               /* calculate default added value:
+                * formula is :
+                * addend = (2^32)/freq_div_ratio;
+                * where, freq_div_ratio = clk_ptp_ref_i/50MHz
+                * hence, addend = ((2^32) * 50MHz)/clk_ptp_ref_i;
+                * NOTE: clk_ptp_ref_i should be >= 50MHz to
+                *       achieve 20ns accuracy.
+                *
+                * 2^x * y == (y << x), hence
+                * 2^32 * 50000000 ==> (50000000 << 32)
+                */
+               temp = (u64) (50000000ULL << 32);
+               priv->default_addend = div_u64(temp, priv->clk_ptp_rate);
+               priv->hw->ptp->config_addend(priv->ioaddr,
+                                            priv->default_addend);
+
+               /* initialize system time */
+               getnstimeofday(&now);
+               priv->hw->ptp->init_systime(priv->ioaddr, now.tv_sec,
+                                           now.tv_nsec);
+       }
+
+       return copy_to_user(ifr->ifr_data, &config,
+                           sizeof(struct hwtstamp_config)) ? -EFAULT : 0;
+}
+
+/**
+ * stmmac_init_ptp - init PTP
+ * @priv: driver private structure
+ * Description: this is to verify if the HW supports the PTPv1 or PTPv2.
+ * This is done by looking at the HW cap. register.
+ * This function also registers the ptp driver.
+ */
+static int stmmac_init_ptp(struct stmmac_priv *priv)
+{
+       if (!(priv->dma_cap.time_stamp || priv->dma_cap.atime_stamp))
+               return -EOPNOTSUPP;
+
+       /* Fall-back to main clock in case of no PTP ref is passed */
+       priv->clk_ptp_ref = devm_clk_get(priv->device, "clk_ptp_ref");
+       if (IS_ERR(priv->clk_ptp_ref)) {
+               priv->clk_ptp_rate = clk_get_rate(priv->stmmac_clk);
+               priv->clk_ptp_ref = NULL;
+       } else {
+               clk_prepare_enable(priv->clk_ptp_ref);
+               priv->clk_ptp_rate = clk_get_rate(priv->clk_ptp_ref);
+       }
+
+       priv->adv_ts = 0;
+       if (priv->dma_cap.atime_stamp && priv->extend_desc)
+               priv->adv_ts = 1;
+
+       if (netif_msg_hw(priv) && priv->dma_cap.time_stamp)
+               pr_debug("IEEE 1588-2002 Time Stamp supported\n");
+
+       if (netif_msg_hw(priv) && priv->adv_ts)
+               pr_debug("IEEE 1588-2008 Advanced Time Stamp supported\n");
+
+       priv->hw->ptp = &stmmac_ptp;
+       priv->hwts_tx_en = 0;
+       priv->hwts_rx_en = 0;
+
+       return stmmac_ptp_register(priv);
+}
+
+static void stmmac_release_ptp(struct stmmac_priv *priv)
+{
+       if (priv->clk_ptp_ref)
+               clk_disable_unprepare(priv->clk_ptp_ref);
+       stmmac_ptp_unregister(priv);
+}
+
+/**
+ * stmmac_adjust_link - adjusts the link parameters
+ * @dev: net device structure
+ * Description: this is the helper called by the physical abstraction layer
+ * drivers to communicate the phy link status. According the speed and duplex
+ * this driver can invoke registered glue-logic as well.
+ * It also invoke the eee initialization because it could happen when switch
+ * on different networks (that are eee capable).
+ */
+static void stmmac_adjust_link(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       struct phy_device *phydev = priv->phydev;
+       unsigned long flags;
+       int new_state = 0;
+       unsigned int fc = priv->flow_ctrl, pause_time = priv->pause;
+
+       if (phydev == NULL)
+               return;
+
+       spin_lock_irqsave(&priv->lock, flags);
+
+       if (phydev->link) {
+               u32 ctrl = readl(priv->ioaddr + MAC_CTRL_REG);
+
+               /* 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;
+                       if (!(phydev->duplex))
+                               ctrl &= ~priv->hw->link.duplex;
+                       else
+                               ctrl |= priv->hw->link.duplex;
+                       priv->oldduplex = phydev->duplex;
+               }
+               /* Flow Control operation */
+               if (phydev->pause)
+                       priv->hw->mac->flow_ctrl(priv->hw, phydev->duplex,
+                                                fc, pause_time);
+
+               if (phydev->speed != priv->speed) {
+                       new_state = 1;
+                       switch (phydev->speed) {
+                       case 1000:
+                               if (likely(priv->plat->has_gmac))
+                                       ctrl &= ~priv->hw->link.port;
+                               stmmac_hw_fix_mac_speed(priv);
+                               break;
+                       case 100:
+                       case 10:
+                               if (priv->plat->has_gmac) {
+                                       ctrl |= priv->hw->link.port;
+                                       if (phydev->speed == SPEED_100) {
+                                               ctrl |= priv->hw->link.speed;
+                                       } else {
+                                               ctrl &= ~(priv->hw->link.speed);
+                                       }
+                               } else {
+                                       ctrl &= ~priv->hw->link.port;
+                               }
+                               stmmac_hw_fix_mac_speed(priv);
+                               break;
+                       default:
+                               if (netif_msg_link(priv))
+                                       pr_warn("%s: Speed (%d) not 10/100\n",
+                                               dev->name, phydev->speed);
+                               break;
+                       }
+
+                       priv->speed = phydev->speed;
+               }
+
+               writel(ctrl, priv->ioaddr + MAC_CTRL_REG);
+
+               if (!priv->oldlink) {
+                       new_state = 1;
+                       priv->oldlink = 1;
+               }
+       } else if (priv->oldlink) {
+               new_state = 1;
+               priv->oldlink = 0;
+               priv->speed = 0;
+               priv->oldduplex = -1;
+       }
+
+       if (new_state && netif_msg_link(priv))
+               phy_print_status(phydev);
+
+       spin_unlock_irqrestore(&priv->lock, flags);
+
+       /* At this stage, it could be needed to setup the EEE or adjust some
+        * MAC related HW registers.
+        */
+       priv->eee_enabled = stmmac_eee_init(priv);
+}
+
+/**
+ * stmmac_check_pcs_mode - verify if RGMII/SGMII is supported
+ * @priv: driver private structure
+ * Description: this is to verify if the HW supports the PCS.
+ * Physical Coding Sublayer (PCS) interface that can be used when the MAC is
+ * configured for the TBI, RTBI, or SGMII PHY interface.
+ */
+static void stmmac_check_pcs_mode(struct stmmac_priv *priv)
+{
+       int interface = priv->plat->interface;
+
+       if (priv->dma_cap.pcs) {
+               if ((interface == PHY_INTERFACE_MODE_RGMII) ||
+                   (interface == PHY_INTERFACE_MODE_RGMII_ID) ||
+                   (interface == PHY_INTERFACE_MODE_RGMII_RXID) ||
+                   (interface == PHY_INTERFACE_MODE_RGMII_TXID)) {
+                       pr_debug("STMMAC: PCS RGMII support enable\n");
+                       priv->pcs = STMMAC_PCS_RGMII;
+               } else if (interface == PHY_INTERFACE_MODE_SGMII) {
+                       pr_debug("STMMAC: PCS SGMII support enable\n");
+                       priv->pcs = STMMAC_PCS_SGMII;
+               }
+       }
+}
+
+/**
+ * stmmac_init_phy - PHY initialization
+ * @dev: net device structure
+ * Description: it initializes the driver's PHY state, and attaches the PHY
+ * to the mac driver.
+ *  Return value:
+ *  0 on success
+ */
+static int stmmac_init_phy(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       struct phy_device *phydev;
+       char phy_id_fmt[MII_BUS_ID_SIZE + 3];
+       char bus_id[MII_BUS_ID_SIZE];
+       int interface = priv->plat->interface;
+       int max_speed = priv->plat->max_speed;
+       priv->oldlink = 0;
+       priv->speed = 0;
+       priv->oldduplex = -1;
+
+       if (priv->plat->phy_bus_name)
+               snprintf(bus_id, MII_BUS_ID_SIZE, "%s-%x",
+                        priv->plat->phy_bus_name, priv->plat->bus_id);
+       else
+               snprintf(bus_id, MII_BUS_ID_SIZE, "stmmac-%x",
+                        priv->plat->bus_id);
+
+       snprintf(phy_id_fmt, MII_BUS_ID_SIZE + 3, PHY_ID_FMT, bus_id,
+                priv->plat->phy_addr);
+       pr_debug("stmmac_init_phy:  trying to attach to %s\n", phy_id_fmt);
+
+       phydev = phy_connect(dev, phy_id_fmt, &stmmac_adjust_link, interface);
+
+       if (IS_ERR(phydev)) {
+               pr_err("%s: Could not attach to PHY\n", dev->name);
+               return PTR_ERR(phydev);
+       }
+
+       /* Stop Advertising 1000BASE Capability if interface is not GMII */
+       if ((interface == PHY_INTERFACE_MODE_MII) ||
+           (interface == PHY_INTERFACE_MODE_RMII) ||
+               (max_speed < 1000 && max_speed > 0))
+               phydev->advertising &= ~(SUPPORTED_1000baseT_Half |
+                                        SUPPORTED_1000baseT_Full);
+
+       /*
+        * Broken HW is sometimes missing the pull-up resistor on the
+        * MDIO line, which results in reads to non-existent devices returning
+        * 0 rather than 0xffff. Catch this here and treat 0 as a non-existent
+        * device as well.
+        * Note: phydev->phy_id is the result of reading the UID PHY registers.
+        */
+       if (phydev->phy_id == 0) {
+               phy_disconnect(phydev);
+               return -ENODEV;
+       }
+       pr_debug("stmmac_init_phy:  %s: attached to PHY (UID 0x%x)"
+                " Link = %d\n", dev->name, phydev->phy_id, phydev->link);
+
+       priv->phydev = phydev;
+
+       return 0;
+}
+
+/**
+ * stmmac_display_ring - display ring
+ * @head: pointer to the head of the ring passed.
+ * @size: size of the ring.
+ * @extend_desc: to verify if extended descriptors are used.
+ * Description: display the control/status and buffer descriptors.
+ */
+static void stmmac_display_ring(void *head, int size, int extend_desc)
+{
+       int i;
+       struct dma_extended_desc *ep = (struct dma_extended_desc *)head;
+       struct dma_desc *p = (struct dma_desc *)head;
+
+       for (i = 0; i < size; i++) {
+               u64 x;
+               if (extend_desc) {
+                       x = *(u64 *) ep;
+                       pr_info("%d [0x%x]: 0x%x 0x%x 0x%x 0x%x\n",
+                               i, (unsigned int)virt_to_phys(ep),
+                               (unsigned int)x, (unsigned int)(x >> 32),
+                               ep->basic.des2, ep->basic.des3);
+                       ep++;
+               } else {
+                       x = *(u64 *) p;
+                       pr_info("%d [0x%x]: 0x%x 0x%x 0x%x 0x%x",
+                               i, (unsigned int)virt_to_phys(p),
+                               (unsigned int)x, (unsigned int)(x >> 32),
+                               p->des2, p->des3);
+                       p++;
+               }
+               pr_info("\n");
+       }
+}
+
+static void stmmac_display_rings(struct stmmac_priv *priv)
+{
+       unsigned int txsize = priv->dma_tx_size;
+       unsigned int rxsize = priv->dma_rx_size;
+
+       if (priv->extend_desc) {
+               pr_info("Extended RX descriptor ring:\n");
+               stmmac_display_ring((void *)priv->dma_erx, rxsize, 1);
+               pr_info("Extended TX descriptor ring:\n");
+               stmmac_display_ring((void *)priv->dma_etx, txsize, 1);
+       } else {
+               pr_info("RX descriptor ring:\n");
+               stmmac_display_ring((void *)priv->dma_rx, rxsize, 0);
+               pr_info("TX descriptor ring:\n");
+               stmmac_display_ring((void *)priv->dma_tx, txsize, 0);
+       }
+}
+
+static int stmmac_set_bfsize(int mtu, int bufsize)
+{
+       int ret = bufsize;
+
+       if (mtu >= BUF_SIZE_4KiB)
+               ret = BUF_SIZE_8KiB;
+       else if (mtu >= BUF_SIZE_2KiB)
+               ret = BUF_SIZE_4KiB;
+       else if (mtu > DEFAULT_BUFSIZE)
+               ret = BUF_SIZE_2KiB;
+       else
+               ret = DEFAULT_BUFSIZE;
+
+       return ret;
+}
+
+/**
+ * stmmac_clear_descriptors - clear descriptors
+ * @priv: driver private structure
+ * Description: this function is called to clear the tx and rx descriptors
+ * in case of both basic and extended descriptors are used.
+ */
+static void stmmac_clear_descriptors(struct stmmac_priv *priv)
+{
+       int i;
+       unsigned int txsize = priv->dma_tx_size;
+       unsigned int rxsize = priv->dma_rx_size;
+
+       /* Clear the Rx/Tx descriptors */
+       for (i = 0; i < rxsize; i++)
+               if (priv->extend_desc)
+                       priv->hw->desc->init_rx_desc(&priv->dma_erx[i].basic,
+                                                    priv->use_riwt, priv->mode,
+                                                    (i == rxsize - 1));
+               else
+                       priv->hw->desc->init_rx_desc(&priv->dma_rx[i],
+                                                    priv->use_riwt, priv->mode,
+                                                    (i == rxsize - 1));
+       for (i = 0; i < txsize; i++)
+               if (priv->extend_desc)
+                       priv->hw->desc->init_tx_desc(&priv->dma_etx[i].basic,
+                                                    priv->mode,
+                                                    (i == txsize - 1));
+               else
+                       priv->hw->desc->init_tx_desc(&priv->dma_tx[i],
+                                                    priv->mode,
+                                                    (i == txsize - 1));
+}
+
+/**
+ * stmmac_init_rx_buffers - init the RX descriptor buffer.
+ * @priv: driver private structure
+ * @p: descriptor pointer
+ * @i: descriptor index
+ * @flags: gfp flag.
+ * Description: this function is called to allocate a receive buffer, perform
+ * the DMA mapping and init the descriptor.
+ */
+static int stmmac_init_rx_buffers(struct stmmac_priv *priv, struct dma_desc *p,
+                                 int i, gfp_t flags)
+{
+       struct sk_buff *skb;
+
+       skb = __netdev_alloc_skb(priv->dev, priv->dma_buf_sz + NET_IP_ALIGN,
+                                flags);
+       if (!skb) {
+               pr_err("%s: Rx init fails; skb is NULL\n", __func__);
+               return -ENOMEM;
+       }
+       skb_reserve(skb, NET_IP_ALIGN);
+       priv->rx_skbuff[i] = skb;
+       priv->rx_skbuff_dma[i] = dma_map_single(priv->device, skb->data,
+                                               priv->dma_buf_sz,
+                                               DMA_FROM_DEVICE);
+       if (dma_mapping_error(priv->device, priv->rx_skbuff_dma[i])) {
+               pr_err("%s: DMA mapping error\n", __func__);
+               dev_kfree_skb_any(skb);
+               return -EINVAL;
+       }
+
+       p->des2 = priv->rx_skbuff_dma[i];
+
+       if ((priv->hw->mode->init_desc3) &&
+           (priv->dma_buf_sz == BUF_SIZE_16KiB))
+               priv->hw->mode->init_desc3(p);
+
+       return 0;
+}
+
+static void stmmac_free_rx_buffers(struct stmmac_priv *priv, int i)
+{
+       if (priv->rx_skbuff[i]) {
+               dma_unmap_single(priv->device, priv->rx_skbuff_dma[i],
+                                priv->dma_buf_sz, DMA_FROM_DEVICE);
+               dev_kfree_skb_any(priv->rx_skbuff[i]);
+       }
+       priv->rx_skbuff[i] = NULL;
+}
+
+/**
+ * init_dma_desc_rings - init the RX/TX descriptor rings
+ * @dev: net device structure
+ * @flags: gfp flag.
+ * Description: this function initializes the DMA RX/TX descriptors
+ * and allocates the socket buffers. It suppors the chained and ring
+ * modes.
+ */
+static int init_dma_desc_rings(struct net_device *dev, gfp_t flags)
+{
+       int i;
+       struct stmmac_priv *priv = netdev_priv(dev);
+       unsigned int txsize = priv->dma_tx_size;
+       unsigned int rxsize = priv->dma_rx_size;
+       unsigned int bfsize = 0;
+       int ret = -ENOMEM;
+
+       if (priv->hw->mode->set_16kib_bfsize)
+               bfsize = priv->hw->mode->set_16kib_bfsize(dev->mtu);
+
+       if (bfsize < BUF_SIZE_16KiB)
+               bfsize = stmmac_set_bfsize(dev->mtu, priv->dma_buf_sz);
+
+       priv->dma_buf_sz = bfsize;
+
+       if (netif_msg_probe(priv))
+               pr_debug("%s: txsize %d, rxsize %d, bfsize %d\n", __func__,
+                        txsize, rxsize, bfsize);
+
+       if (netif_msg_probe(priv)) {
+               pr_debug("(%s) dma_rx_phy=0x%08x dma_tx_phy=0x%08x\n", __func__,
+                        (u32) priv->dma_rx_phy, (u32) priv->dma_tx_phy);
+
+               /* RX INITIALIZATION */
+               pr_debug("\tSKB addresses:\nskb\t\tskb data\tdma data\n");
+       }
+       for (i = 0; i < rxsize; i++) {
+               struct dma_desc *p;
+               if (priv->extend_desc)
+                       p = &((priv->dma_erx + i)->basic);
+               else
+                       p = priv->dma_rx + i;
+
+               ret = stmmac_init_rx_buffers(priv, p, i, flags);
+               if (ret)
+                       goto err_init_rx_buffers;
+
+               if (netif_msg_probe(priv))
+                       pr_debug("[%p]\t[%p]\t[%x]\n", priv->rx_skbuff[i],
+                                priv->rx_skbuff[i]->data,
+                                (unsigned int)priv->rx_skbuff_dma[i]);
+       }
+       priv->cur_rx = 0;
+       priv->dirty_rx = (unsigned int)(i - rxsize);
+       buf_sz = bfsize;
+
+       /* Setup the chained descriptor addresses */
+       if (priv->mode == STMMAC_CHAIN_MODE) {
+               if (priv->extend_desc) {
+                       priv->hw->mode->init(priv->dma_erx, priv->dma_rx_phy,
+                                            rxsize, 1);
+                       priv->hw->mode->init(priv->dma_etx, priv->dma_tx_phy,
+                                            txsize, 1);
+               } else {
+                       priv->hw->mode->init(priv->dma_rx, priv->dma_rx_phy,
+                                            rxsize, 0);
+                       priv->hw->mode->init(priv->dma_tx, priv->dma_tx_phy,
+                                            txsize, 0);
+               }
+       }
+
+       /* TX INITIALIZATION */
+       for (i = 0; i < txsize; i++) {
+               struct dma_desc *p;
+               if (priv->extend_desc)
+                       p = &((priv->dma_etx + i)->basic);
+               else
+                       p = priv->dma_tx + i;
+               p->des2 = 0;
+               priv->tx_skbuff_dma[i].buf = 0;
+               priv->tx_skbuff_dma[i].map_as_page = false;
+               priv->tx_skbuff[i] = NULL;
+       }
+
+       priv->dirty_tx = 0;
+       priv->cur_tx = 0;
+       netdev_reset_queue(priv->dev);
+
+       stmmac_clear_descriptors(priv);
+
+       if (netif_msg_hw(priv))
+               stmmac_display_rings(priv);
+
+       return 0;
+err_init_rx_buffers:
+       while (--i >= 0)
+               stmmac_free_rx_buffers(priv, i);
+       return ret;
+}
+
+static void dma_free_rx_skbufs(struct stmmac_priv *priv)
+{
+       int i;
+
+       for (i = 0; i < priv->dma_rx_size; i++)
+               stmmac_free_rx_buffers(priv, i);
+}
+
+static void dma_free_tx_skbufs(struct stmmac_priv *priv)
+{
+       int i;
+
+       for (i = 0; i < priv->dma_tx_size; i++) {
+               struct dma_desc *p;
+
+               if (priv->extend_desc)
+                       p = &((priv->dma_etx + i)->basic);
+               else
+                       p = priv->dma_tx + i;
+
+               if (priv->tx_skbuff_dma[i].buf) {
+                       if (priv->tx_skbuff_dma[i].map_as_page)
+                               dma_unmap_page(priv->device,
+                                              priv->tx_skbuff_dma[i].buf,
+                                              priv->hw->desc->get_tx_len(p),
+                                              DMA_TO_DEVICE);
+                       else
+                               dma_unmap_single(priv->device,
+                                                priv->tx_skbuff_dma[i].buf,
+                                                priv->hw->desc->get_tx_len(p),
+                                                DMA_TO_DEVICE);
+               }
+
+               if (priv->tx_skbuff[i] != NULL) {
+                       dev_kfree_skb_any(priv->tx_skbuff[i]);
+                       priv->tx_skbuff[i] = NULL;
+                       priv->tx_skbuff_dma[i].buf = 0;
+                       priv->tx_skbuff_dma[i].map_as_page = false;
+               }
+       }
+}
+
+/**
+ * alloc_dma_desc_resources - alloc TX/RX resources.
+ * @priv: private structure
+ * Description: according to which descriptor can be used (extend or basic)
+ * this function allocates the resources for TX and RX paths. In case of
+ * reception, for example, it pre-allocated the RX socket buffer in order to
+ * allow zero-copy mechanism.
+ */
+static int alloc_dma_desc_resources(struct stmmac_priv *priv)
+{
+       unsigned int txsize = priv->dma_tx_size;
+       unsigned int rxsize = priv->dma_rx_size;
+       int ret = -ENOMEM;
+
+       priv->rx_skbuff_dma = kmalloc_array(rxsize, sizeof(dma_addr_t),
+                                           GFP_KERNEL);
+       if (!priv->rx_skbuff_dma)
+               return -ENOMEM;
+
+       priv->rx_skbuff = kmalloc_array(rxsize, sizeof(struct sk_buff *),
+                                       GFP_KERNEL);
+       if (!priv->rx_skbuff)
+               goto err_rx_skbuff;
+
+       priv->tx_skbuff_dma = kmalloc_array(txsize,
+                                           sizeof(*priv->tx_skbuff_dma),
+                                           GFP_KERNEL);
+       if (!priv->tx_skbuff_dma)
+               goto err_tx_skbuff_dma;
+
+       priv->tx_skbuff = kmalloc_array(txsize, sizeof(struct sk_buff *),
+                                       GFP_KERNEL);
+       if (!priv->tx_skbuff)
+               goto err_tx_skbuff;
+
+       if (priv->extend_desc) {
+               priv->dma_erx = dma_alloc_coherent(priv->device, rxsize *
+                                                  sizeof(struct
+                                                         dma_extended_desc),
+                                                  &priv->dma_rx_phy,
+                                                  GFP_KERNEL);
+               if (!priv->dma_erx)
+                       goto err_dma;
+
+               priv->dma_etx = dma_alloc_coherent(priv->device, txsize *
+                                                  sizeof(struct
+                                                         dma_extended_desc),
+                                                  &priv->dma_tx_phy,
+                                                  GFP_KERNEL);
+               if (!priv->dma_etx) {
+                       dma_free_coherent(priv->device, priv->dma_rx_size *
+                                       sizeof(struct dma_extended_desc),
+                                       priv->dma_erx, priv->dma_rx_phy);
+                       goto err_dma;
+               }
+       } else {
+               priv->dma_rx = dma_alloc_coherent(priv->device, rxsize *
+                                                 sizeof(struct dma_desc),
+                                                 &priv->dma_rx_phy,
+                                                 GFP_KERNEL);
+               if (!priv->dma_rx)
+                       goto err_dma;
+
+               priv->dma_tx = dma_alloc_coherent(priv->device, txsize *
+                                                 sizeof(struct dma_desc),
+                                                 &priv->dma_tx_phy,
+                                                 GFP_KERNEL);
+               if (!priv->dma_tx) {
+                       dma_free_coherent(priv->device, priv->dma_rx_size *
+                                       sizeof(struct dma_desc),
+                                       priv->dma_rx, priv->dma_rx_phy);
+                       goto err_dma;
+               }
+       }
+
+       return 0;
+
+err_dma:
+       kfree(priv->tx_skbuff);
+err_tx_skbuff:
+       kfree(priv->tx_skbuff_dma);
+err_tx_skbuff_dma:
+       kfree(priv->rx_skbuff);
+err_rx_skbuff:
+       kfree(priv->rx_skbuff_dma);
+       return ret;
+}
+
+static void free_dma_desc_resources(struct stmmac_priv *priv)
+{
+       /* Release the DMA TX/RX socket buffers */
+       dma_free_rx_skbufs(priv);
+       dma_free_tx_skbufs(priv);
+
+       /* Free DMA regions of consistent memory previously allocated */
+       if (!priv->extend_desc) {
+               dma_free_coherent(priv->device,
+                                 priv->dma_tx_size * sizeof(struct dma_desc),
+                                 priv->dma_tx, priv->dma_tx_phy);
+               dma_free_coherent(priv->device,
+                                 priv->dma_rx_size * sizeof(struct dma_desc),
+                                 priv->dma_rx, priv->dma_rx_phy);
+       } else {
+               dma_free_coherent(priv->device, priv->dma_tx_size *
+                                 sizeof(struct dma_extended_desc),
+                                 priv->dma_etx, priv->dma_tx_phy);
+               dma_free_coherent(priv->device, priv->dma_rx_size *
+                                 sizeof(struct dma_extended_desc),
+                                 priv->dma_erx, priv->dma_rx_phy);
+       }
+       kfree(priv->rx_skbuff_dma);
+       kfree(priv->rx_skbuff);
+       kfree(priv->tx_skbuff_dma);
+       kfree(priv->tx_skbuff);
+}
+
+/**
+ *  stmmac_dma_operation_mode - HW DMA operation mode
+ *  @priv: driver private structure
+ *  Description: it is used for configuring the DMA operation mode register in
+ *  order to program the tx/rx DMA thresholds or Store-And-Forward mode.
+ */
+static void stmmac_dma_operation_mode(struct stmmac_priv *priv)
+{
+       int rxfifosz = priv->plat->rx_fifo_size;
+
+       if (priv->plat->force_thresh_dma_mode)
+               priv->hw->dma->dma_mode(priv->ioaddr, tc, tc, rxfifosz);
+       else if (priv->plat->force_sf_dma_mode || priv->plat->tx_coe) {
+               /*
+                * In case of GMAC, SF mode can be enabled
+                * to perform the TX COE in HW. This depends on:
+                * 1) TX COE if actually supported
+                * 2) There is no bugged Jumbo frame support
+                *    that needs to not insert csum in the TDES.
+                */
+               priv->hw->dma->dma_mode(priv->ioaddr, SF_DMA_MODE, SF_DMA_MODE,
+                                       rxfifosz);
+               priv->xstats.threshold = SF_DMA_MODE;
+       } else
+               priv->hw->dma->dma_mode(priv->ioaddr, tc, SF_DMA_MODE,
+                                       rxfifosz);
+}
+
+/**
+ * stmmac_tx_clean - to manage the transmission completion
+ * @priv: driver private structure
+ * Description: it reclaims the transmit resources after transmission completes.
+ */
+static void stmmac_tx_clean(struct stmmac_priv *priv)
+{
+       unsigned int txsize = priv->dma_tx_size;
+       unsigned int bytes_compl = 0, pkts_compl = 0;
+
+       spin_lock(&priv->tx_lock);
+
+       priv->xstats.tx_clean++;
+
+       while (priv->dirty_tx != priv->cur_tx) {
+               int last;
+               unsigned int entry = priv->dirty_tx % txsize;
+               struct sk_buff *skb = priv->tx_skbuff[entry];
+               struct dma_desc *p;
+
+               if (priv->extend_desc)
+                       p = (struct dma_desc *)(priv->dma_etx + entry);
+               else
+                       p = priv->dma_tx + entry;
+
+               /* Check if the descriptor is owned by the DMA. */
+               if (priv->hw->desc->get_tx_owner(p))
+                       break;
+
+               /* Verify tx error by looking at the last segment. */
+               last = priv->hw->desc->get_tx_ls(p);
+               if (likely(last)) {
+                       int tx_error =
+                           priv->hw->desc->tx_status(&priv->dev->stats,
+                                                     &priv->xstats, p,
+                                                     priv->ioaddr);
+                       if (likely(tx_error == 0)) {
+                               priv->dev->stats.tx_packets++;
+                               priv->xstats.tx_pkt_n++;
+                       } else
+                               priv->dev->stats.tx_errors++;
+
+                       stmmac_get_tx_hwtstamp(priv, entry, skb);
+               }
+               if (netif_msg_tx_done(priv))
+                       pr_debug("%s: curr %d, dirty %d\n", __func__,
+                                priv->cur_tx, priv->dirty_tx);
+
+               if (likely(priv->tx_skbuff_dma[entry].buf)) {
+                       if (priv->tx_skbuff_dma[entry].map_as_page)
+                               dma_unmap_page(priv->device,
+                                              priv->tx_skbuff_dma[entry].buf,
+                                              priv->hw->desc->get_tx_len(p),
+                                              DMA_TO_DEVICE);
+                       else
+                               dma_unmap_single(priv->device,
+                                                priv->tx_skbuff_dma[entry].buf,
+                                                priv->hw->desc->get_tx_len(p),
+                                                DMA_TO_DEVICE);
+                       priv->tx_skbuff_dma[entry].buf = 0;
+                       priv->tx_skbuff_dma[entry].map_as_page = false;
+               }
+               priv->hw->mode->clean_desc3(priv, p);
+
+               if (likely(skb != NULL)) {
+                       pkts_compl++;
+                       bytes_compl += skb->len;
+                       dev_consume_skb_any(skb);
+                       priv->tx_skbuff[entry] = NULL;
+               }
+
+               priv->hw->desc->release_tx_desc(p, priv->mode);
+
+               priv->dirty_tx++;
+       }
+
+       netdev_completed_queue(priv->dev, pkts_compl, bytes_compl);
+
+       if (unlikely(netif_queue_stopped(priv->dev) &&
+                    stmmac_tx_avail(priv) > STMMAC_TX_THRESH(priv))) {
+               netif_tx_lock(priv->dev);
+               if (netif_queue_stopped(priv->dev) &&
+                   stmmac_tx_avail(priv) > STMMAC_TX_THRESH(priv)) {
+                       if (netif_msg_tx_done(priv))
+                               pr_debug("%s: restart transmit\n", __func__);
+                       netif_wake_queue(priv->dev);
+               }
+               netif_tx_unlock(priv->dev);
+       }
+
+       if ((priv->eee_enabled) && (!priv->tx_path_in_lpi_mode)) {
+               stmmac_enable_eee_mode(priv);
+               mod_timer(&priv->eee_ctrl_timer, STMMAC_LPI_T(eee_timer));
+       }
+       spin_unlock(&priv->tx_lock);
+}
+
+static inline void stmmac_enable_dma_irq(struct stmmac_priv *priv)
+{
+       priv->hw->dma->enable_dma_irq(priv->ioaddr);
+}
+
+static inline void stmmac_disable_dma_irq(struct stmmac_priv *priv)
+{
+       priv->hw->dma->disable_dma_irq(priv->ioaddr);
+}
+
+/**
+ * stmmac_tx_err - to manage the tx error
+ * @priv: driver private structure
+ * Description: it cleans the descriptors and restarts the transmission
+ * in case of transmission errors.
+ */
+static void stmmac_tx_err(struct stmmac_priv *priv)
+{
+       int i;
+       int txsize = priv->dma_tx_size;
+       netif_stop_queue(priv->dev);
+
+       priv->hw->dma->stop_tx(priv->ioaddr);
+       dma_free_tx_skbufs(priv);
+       for (i = 0; i < txsize; i++)
+               if (priv->extend_desc)
+                       priv->hw->desc->init_tx_desc(&priv->dma_etx[i].basic,
+                                                    priv->mode,
+                                                    (i == txsize - 1));
+               else
+                       priv->hw->desc->init_tx_desc(&priv->dma_tx[i],
+                                                    priv->mode,
+                                                    (i == txsize - 1));
+       priv->dirty_tx = 0;
+       priv->cur_tx = 0;
+       netdev_reset_queue(priv->dev);
+       priv->hw->dma->start_tx(priv->ioaddr);
+
+       priv->dev->stats.tx_errors++;
+       netif_wake_queue(priv->dev);
+}
+
+/**
+ * stmmac_dma_interrupt - DMA ISR
+ * @priv: driver private structure
+ * Description: this is the DMA ISR. It is called by the main ISR.
+ * It calls the dwmac dma routine and schedule poll method in case of some
+ * work can be done.
+ */
+static void stmmac_dma_interrupt(struct stmmac_priv *priv)
+{
+       int status;
+       int rxfifosz = priv->plat->rx_fifo_size;
+
+       status = priv->hw->dma->dma_interrupt(priv->ioaddr, &priv->xstats);
+       if (likely((status & handle_rx)) || (status & handle_tx)) {
+               if (likely(napi_schedule_prep(&priv->napi))) {
+                       stmmac_disable_dma_irq(priv);
+                       __napi_schedule(&priv->napi);
+               }
+       }
+       if (unlikely(status & tx_hard_error_bump_tc)) {
+               /* Try to bump up the dma threshold on this failure */
+               if (unlikely(priv->xstats.threshold != SF_DMA_MODE) &&
+                   (tc <= 256)) {
+                       tc += 64;
+                       if (priv->plat->force_thresh_dma_mode)
+                               priv->hw->dma->dma_mode(priv->ioaddr, tc, tc,
+                                                       rxfifosz);
+                       else
+                               priv->hw->dma->dma_mode(priv->ioaddr, tc,
+                                                       SF_DMA_MODE, rxfifosz);
+                       priv->xstats.threshold = tc;
+               }
+       } else if (unlikely(status == tx_hard_error))
+               stmmac_tx_err(priv);
+}
+
+/**
+ * stmmac_mmc_setup: setup the Mac Management Counters (MMC)
+ * @priv: driver private structure
+ * Description: this masks the MMC irq, in fact, the counters are managed in SW.
+ */
+static void stmmac_mmc_setup(struct stmmac_priv *priv)
+{
+       unsigned int mode = MMC_CNTRL_RESET_ON_READ | MMC_CNTRL_COUNTER_RESET |
+           MMC_CNTRL_PRESET | MMC_CNTRL_FULL_HALF_PRESET;
+
+       dwmac_mmc_intr_all_mask(priv->ioaddr);
+
+       if (priv->dma_cap.rmon) {
+               dwmac_mmc_ctrl(priv->ioaddr, mode);
+               memset(&priv->mmc, 0, sizeof(struct stmmac_counters));
+       } else
+               pr_info(" No MAC Management Counters available\n");
+}
+
+/**
+ * stmmac_get_synopsys_id - return the SYINID.
+ * @priv: driver private structure
+ * Description: this simple function is to decode and return the SYINID
+ * starting from the HW core register.
+ */
+static u32 stmmac_get_synopsys_id(struct stmmac_priv *priv)
+{
+       u32 hwid = priv->hw->synopsys_uid;
+
+       /* Check Synopsys Id (not available on old chips) */
+       if (likely(hwid)) {
+               u32 uid = ((hwid & 0x0000ff00) >> 8);
+               u32 synid = (hwid & 0x000000ff);
+
+               pr_info("stmmac - user ID: 0x%x, Synopsys ID: 0x%x\n",
+                       uid, synid);
+
+               return synid;
+       }
+       return 0;
+}
+
+/**
+ * stmmac_selec_desc_mode - to select among: normal/alternate/extend descriptors
+ * @priv: driver private structure
+ * Description: select the Enhanced/Alternate or Normal descriptors.
+ * In case of Enhanced/Alternate, it checks if the extended descriptors are
+ * supported by the HW capability register.
+ */
+static void stmmac_selec_desc_mode(struct stmmac_priv *priv)
+{
+       if (priv->plat->enh_desc) {
+               pr_info(" Enhanced/Alternate descriptors\n");
+
+               /* GMAC older than 3.50 has no extended descriptors */
+               if (priv->synopsys_id >= DWMAC_CORE_3_50) {
+                       pr_info("\tEnabled extended descriptors\n");
+                       priv->extend_desc = 1;
+               } else
+                       pr_warn("Extended descriptors not supported\n");
+
+               priv->hw->desc = &enh_desc_ops;
+       } else {
+               pr_info(" Normal descriptors\n");
+               priv->hw->desc = &ndesc_ops;
+       }
+}
+
+/**
+ * stmmac_get_hw_features - get MAC capabilities from the HW cap. register.
+ * @priv: driver private structure
+ * Description:
+ *  new GMAC chip generations have a new register to indicate the
+ *  presence of the optional feature/functions.
+ *  This can be also used to override the value passed through the
+ *  platform and necessary for old MAC10/100 and GMAC chips.
+ */
+static int stmmac_get_hw_features(struct stmmac_priv *priv)
+{
+       u32 hw_cap = 0;
+
+       if (priv->hw->dma->get_hw_feature) {
+               hw_cap = priv->hw->dma->get_hw_feature(priv->ioaddr);
+
+               priv->dma_cap.mbps_10_100 = (hw_cap & DMA_HW_FEAT_MIISEL);
+               priv->dma_cap.mbps_1000 = (hw_cap & DMA_HW_FEAT_GMIISEL) >> 1;
+               priv->dma_cap.half_duplex = (hw_cap & DMA_HW_FEAT_HDSEL) >> 2;
+               priv->dma_cap.hash_filter = (hw_cap & DMA_HW_FEAT_HASHSEL) >> 4;
+               priv->dma_cap.multi_addr = (hw_cap & DMA_HW_FEAT_ADDMAC) >> 5;
+               priv->dma_cap.pcs = (hw_cap & DMA_HW_FEAT_PCSSEL) >> 6;
+               priv->dma_cap.sma_mdio = (hw_cap & DMA_HW_FEAT_SMASEL) >> 8;
+               priv->dma_cap.pmt_remote_wake_up =
+                   (hw_cap & DMA_HW_FEAT_RWKSEL) >> 9;
+               priv->dma_cap.pmt_magic_frame =
+                   (hw_cap & DMA_HW_FEAT_MGKSEL) >> 10;
+               /* MMC */
+               priv->dma_cap.rmon = (hw_cap & DMA_HW_FEAT_MMCSEL) >> 11;
+               /* IEEE 1588-2002 */
+               priv->dma_cap.time_stamp =
+                   (hw_cap & DMA_HW_FEAT_TSVER1SEL) >> 12;
+               /* IEEE 1588-2008 */
+               priv->dma_cap.atime_stamp =
+                   (hw_cap & DMA_HW_FEAT_TSVER2SEL) >> 13;
+               /* 802.3az - Energy-Efficient Ethernet (EEE) */
+               priv->dma_cap.eee = (hw_cap & DMA_HW_FEAT_EEESEL) >> 14;
+               priv->dma_cap.av = (hw_cap & DMA_HW_FEAT_AVSEL) >> 15;
+               /* TX and RX csum */
+               priv->dma_cap.tx_coe = (hw_cap & DMA_HW_FEAT_TXCOESEL) >> 16;
+               priv->dma_cap.rx_coe_type1 =
+                   (hw_cap & DMA_HW_FEAT_RXTYP1COE) >> 17;
+               priv->dma_cap.rx_coe_type2 =
+                   (hw_cap & DMA_HW_FEAT_RXTYP2COE) >> 18;
+               priv->dma_cap.rxfifo_over_2048 =
+                   (hw_cap & DMA_HW_FEAT_RXFIFOSIZE) >> 19;
+               /* TX and RX number of channels */
+               priv->dma_cap.number_rx_channel =
+                   (hw_cap & DMA_HW_FEAT_RXCHCNT) >> 20;
+               priv->dma_cap.number_tx_channel =
+                   (hw_cap & DMA_HW_FEAT_TXCHCNT) >> 22;
+               /* Alternate (enhanced) DESC mode */
+               priv->dma_cap.enh_desc = (hw_cap & DMA_HW_FEAT_ENHDESSEL) >> 24;
+       }
+
+       return hw_cap;
+}
+
+/**
+ * stmmac_check_ether_addr - check if the MAC addr is valid
+ * @priv: driver private structure
+ * Description:
+ * it is to verify if the MAC address is valid, in case of failures it
+ * generates a random MAC address
+ */
+static void stmmac_check_ether_addr(struct stmmac_priv *priv)
+{
+       if (!is_valid_ether_addr(priv->dev->dev_addr)) {
+               priv->hw->mac->get_umac_addr(priv->hw,
+                                            priv->dev->dev_addr, 0);
+               if (!is_valid_ether_addr(priv->dev->dev_addr))
+                       eth_hw_addr_random(priv->dev);
+               pr_info("%s: device MAC address %pM\n", priv->dev->name,
+                       priv->dev->dev_addr);
+       }
+}
+
+/**
+ * stmmac_init_dma_engine - DMA init.
+ * @priv: driver private structure
+ * Description:
+ * It inits the DMA invoking the specific MAC/GMAC callback.
+ * Some DMA parameters can be passed from the platform;
+ * in case of these are not passed a default is kept for the MAC or GMAC.
+ */
+static int stmmac_init_dma_engine(struct stmmac_priv *priv)
+{
+       int pbl = DEFAULT_DMA_PBL, fixed_burst = 0, burst_len = 0;
+       int mixed_burst = 0;
+       int atds = 0;
+
+       if (priv->plat->dma_cfg) {
+               pbl = priv->plat->dma_cfg->pbl;
+               fixed_burst = priv->plat->dma_cfg->fixed_burst;
+               mixed_burst = priv->plat->dma_cfg->mixed_burst;
+               burst_len = priv->plat->dma_cfg->burst_len;
+       }
+
+       if (priv->extend_desc && (priv->mode == STMMAC_RING_MODE))
+               atds = 1;
+
+       return priv->hw->dma->init(priv->ioaddr, pbl, fixed_burst, mixed_burst,
+                                  burst_len, priv->dma_tx_phy,
+                                  priv->dma_rx_phy, atds);
+}
+
+/**
+ * stmmac_tx_timer - mitigation sw timer for tx.
+ * @data: data pointer
+ * Description:
+ * This is the timer handler to directly invoke the stmmac_tx_clean.
+ */
+static void stmmac_tx_timer(unsigned long data)
+{
+       struct stmmac_priv *priv = (struct stmmac_priv *)data;
+
+       stmmac_tx_clean(priv);
+}
+
+/**
+ * stmmac_init_tx_coalesce - init tx mitigation options.
+ * @priv: driver private structure
+ * Description:
+ * This inits the transmit coalesce parameters: i.e. timer rate,
+ * timer handler and default threshold used for enabling the
+ * interrupt on completion bit.
+ */
+static void stmmac_init_tx_coalesce(struct stmmac_priv *priv)
+{
+       priv->tx_coal_frames = STMMAC_TX_FRAMES;
+       priv->tx_coal_timer = STMMAC_COAL_TX_TIMER;
+       init_timer(&priv->txtimer);
+       priv->txtimer.expires = STMMAC_COAL_TIMER(priv->tx_coal_timer);
+       priv->txtimer.data = (unsigned long)priv;
+       priv->txtimer.function = stmmac_tx_timer;
+       add_timer(&priv->txtimer);
+}
+
+/**
+ * stmmac_hw_setup - setup mac in a usable state.
+ *  @dev : pointer to the device structure.
+ *  Description:
+ *  this is the main function to setup the HW in a usable state because the
+ *  dma engine is reset, the core registers are configured (e.g. AXI,
+ *  Checksum features, timers). The DMA is ready to start receiving and
+ *  transmitting.
+ *  Return value:
+ *  0 on success and an appropriate (-)ve integer as defined in errno.h
+ *  file on failure.
+ */
+static int stmmac_hw_setup(struct net_device *dev, bool init_ptp)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       int ret;
+
+       /* DMA initialization and SW reset */
+       ret = stmmac_init_dma_engine(priv);
+       if (ret < 0) {
+               pr_err("%s: DMA engine initialization failed\n", __func__);
+               return ret;
+       }
+
+       /* Copy the MAC addr into the HW  */
+       priv->hw->mac->set_umac_addr(priv->hw, dev->dev_addr, 0);
+
+       /* If required, perform hw setup of the bus. */
+       if (priv->plat->bus_setup)
+               priv->plat->bus_setup(priv->ioaddr);
+
+       /* Initialize the MAC Core */
+       priv->hw->mac->core_init(priv->hw, dev->mtu);
+
+       ret = priv->hw->mac->rx_ipc(priv->hw);
+       if (!ret) {
+               pr_warn(" RX IPC Checksum Offload disabled\n");
+               priv->plat->rx_coe = STMMAC_RX_COE_NONE;
+               priv->hw->rx_csum = 0;
+       }
+
+       /* Enable the MAC Rx/Tx */
+       stmmac_set_mac(priv->ioaddr, true);
+
+       /* Set the HW DMA mode and the COE */
+       stmmac_dma_operation_mode(priv);
+
+       stmmac_mmc_setup(priv);
+
+       if (init_ptp) {
+               ret = stmmac_init_ptp(priv);
+               if (ret && ret != -EOPNOTSUPP)
+                       pr_warn("%s: failed PTP initialisation\n", __func__);
+       }
+
+#ifdef CONFIG_DEBUG_FS
+       ret = stmmac_init_fs(dev);
+       if (ret < 0)
+               pr_warn("%s: failed debugFS registration\n", __func__);
+#endif
+       /* Start the ball rolling... */
+       pr_debug("%s: DMA RX/TX processes started...\n", dev->name);
+       priv->hw->dma->start_tx(priv->ioaddr);
+       priv->hw->dma->start_rx(priv->ioaddr);
+
+       /* Dump DMA/MAC registers */
+       if (netif_msg_hw(priv)) {
+               priv->hw->mac->dump_regs(priv->hw);
+               priv->hw->dma->dump_regs(priv->ioaddr);
+       }
+       priv->tx_lpi_timer = STMMAC_DEFAULT_TWT_LS;
+
+       if ((priv->use_riwt) && (priv->hw->dma->rx_watchdog)) {
+               priv->rx_riwt = MAX_DMA_RIWT;
+               priv->hw->dma->rx_watchdog(priv->ioaddr, MAX_DMA_RIWT);
+       }
+
+       if (priv->pcs && priv->hw->mac->ctrl_ane)
+               priv->hw->mac->ctrl_ane(priv->hw, 0);
+
+       return 0;
+}
+
+/**
+ *  stmmac_open - open entry point of the driver
+ *  @dev : pointer to the device structure.
+ *  Description:
+ *  This function is the open entry point of the driver.
+ *  Return value:
+ *  0 on success and an appropriate (-)ve integer as defined in errno.h
+ *  file on failure.
+ */
+static int stmmac_open(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       int ret;
+
+       stmmac_check_ether_addr(priv);
+
+       if (priv->pcs != STMMAC_PCS_RGMII && priv->pcs != STMMAC_PCS_TBI &&
+           priv->pcs != STMMAC_PCS_RTBI) {
+               ret = stmmac_init_phy(dev);
+               if (ret) {
+                       pr_err("%s: Cannot attach to PHY (error: %d)\n",
+                              __func__, ret);
+                       return ret;
+               }
+       }
+
+       /* Extra statistics */
+       memset(&priv->xstats, 0, sizeof(struct stmmac_extra_stats));
+       priv->xstats.threshold = tc;
+
+       /* Create and initialize the TX/RX descriptors chains. */
+       priv->dma_tx_size = STMMAC_ALIGN(dma_txsize);
+       priv->dma_rx_size = STMMAC_ALIGN(dma_rxsize);
+       priv->dma_buf_sz = STMMAC_ALIGN(buf_sz);
+
+       ret = alloc_dma_desc_resources(priv);
+       if (ret < 0) {
+               pr_err("%s: DMA descriptors allocation failed\n", __func__);
+               goto dma_desc_error;
+       }
+
+       ret = init_dma_desc_rings(dev, GFP_KERNEL);
+       if (ret < 0) {
+               pr_err("%s: DMA descriptors initialization failed\n", __func__);
+               goto init_error;
+       }
+
+       ret = stmmac_hw_setup(dev, true);
+       if (ret < 0) {
+               pr_err("%s: Hw setup failed\n", __func__);
+               goto init_error;
+       }
+
+       stmmac_init_tx_coalesce(priv);
+
+       if (priv->phydev)
+               phy_start(priv->phydev);
+
+       /* Request the IRQ lines */
+       ret = request_irq(dev->irq, stmmac_interrupt,
+                         IRQF_SHARED, dev->name, dev);
+       if (unlikely(ret < 0)) {
+               pr_err("%s: ERROR: allocating the IRQ %d (error: %d)\n",
+                      __func__, dev->irq, ret);
+               goto init_error;
+       }
+
+       /* Request the Wake IRQ in case of another line is used for WoL */
+       if (priv->wol_irq != dev->irq) {
+               ret = request_irq(priv->wol_irq, stmmac_interrupt,
+                                 IRQF_SHARED, dev->name, dev);
+               if (unlikely(ret < 0)) {
+                       pr_err("%s: ERROR: allocating the WoL IRQ %d (%d)\n",
+                              __func__, priv->wol_irq, ret);
+                       goto wolirq_error;
+               }
+       }
+
+       /* Request the IRQ lines */
+       if (priv->lpi_irq > 0) {
+               ret = request_irq(priv->lpi_irq, stmmac_interrupt, IRQF_SHARED,
+                                 dev->name, dev);
+               if (unlikely(ret < 0)) {
+                       pr_err("%s: ERROR: allocating the LPI IRQ %d (%d)\n",
+                              __func__, priv->lpi_irq, ret);
+                       goto lpiirq_error;
+               }
+       }
+
+       napi_enable(&priv->napi);
+       netif_start_queue(dev);
+
+       return 0;
+
+lpiirq_error:
+       if (priv->wol_irq != dev->irq)
+               free_irq(priv->wol_irq, dev);
+wolirq_error:
+       free_irq(dev->irq, dev);
+
+init_error:
+       free_dma_desc_resources(priv);
+dma_desc_error:
+       if (priv->phydev)
+               phy_disconnect(priv->phydev);
+
+       return ret;
+}
+
+/**
+ *  stmmac_release - close entry point of the driver
+ *  @dev : device pointer.
+ *  Description:
+ *  This is the stop entry point of the driver.
+ */
+static int stmmac_release(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       if (priv->eee_enabled)
+               del_timer_sync(&priv->eee_ctrl_timer);
+
+       /* Stop and disconnect the PHY */
+       if (priv->phydev) {
+               phy_stop(priv->phydev);
+               phy_disconnect(priv->phydev);
+               priv->phydev = NULL;
+       }
+
+       netif_stop_queue(dev);
+
+       napi_disable(&priv->napi);
+
+       del_timer_sync(&priv->txtimer);
+
+       /* Free the IRQ lines */
+       free_irq(dev->irq, dev);
+       if (priv->wol_irq != dev->irq)
+               free_irq(priv->wol_irq, dev);
+       if (priv->lpi_irq > 0)
+               free_irq(priv->lpi_irq, dev);
+
+       /* Stop TX/RX DMA and clear the descriptors */
+       priv->hw->dma->stop_tx(priv->ioaddr);
+       priv->hw->dma->stop_rx(priv->ioaddr);
+
+       /* Release and free the Rx/Tx resources */
+       free_dma_desc_resources(priv);
+
+       /* Disable the MAC Rx/Tx */
+       stmmac_set_mac(priv->ioaddr, false);
+
+       netif_carrier_off(dev);
+
+#ifdef CONFIG_DEBUG_FS
+       stmmac_exit_fs(dev);
+#endif
+
+       stmmac_release_ptp(priv);
+
+       return 0;
+}
+
+/**
+ *  stmmac_xmit - Tx entry point of the driver
+ *  @skb : the socket buffer
+ *  @dev : device pointer
+ *  Description : this is the tx entry point of the driver.
+ *  It programs the chain or the ring and supports oversized frames
+ *  and SG feature.
+ */
+static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       unsigned int txsize = priv->dma_tx_size;
+       unsigned int entry;
+       int i, csum_insertion = 0, is_jumbo = 0;
+       int nfrags = skb_shinfo(skb)->nr_frags;
+       struct dma_desc *desc, *first;
+       unsigned int nopaged_len = skb_headlen(skb);
+       unsigned int enh_desc = priv->plat->enh_desc;
+
+       spin_lock(&priv->tx_lock);
+
+       if (unlikely(stmmac_tx_avail(priv) < nfrags + 1)) {
+               spin_unlock(&priv->tx_lock);
+               if (!netif_queue_stopped(dev)) {
+                       netif_stop_queue(dev);
+                       /* This is a hard error, log it. */
+                       pr_err("%s: Tx Ring full when queue awake\n", __func__);
+               }
+               return NETDEV_TX_BUSY;
+       }
+
+       if (priv->tx_path_in_lpi_mode)
+               stmmac_disable_eee_mode(priv);
+
+       entry = priv->cur_tx % txsize;
+
+       csum_insertion = (skb->ip_summed == CHECKSUM_PARTIAL);
+
+       if (priv->extend_desc)
+               desc = (struct dma_desc *)(priv->dma_etx + entry);
+       else
+               desc = priv->dma_tx + entry;
+
+       first = desc;
+
+       /* To program the descriptors according to the size of the frame */
+       if (enh_desc)
+               is_jumbo = priv->hw->mode->is_jumbo_frm(skb->len, enh_desc);
+
+       if (likely(!is_jumbo)) {
+               desc->des2 = dma_map_single(priv->device, skb->data,
+                                           nopaged_len, DMA_TO_DEVICE);
+               if (dma_mapping_error(priv->device, desc->des2))
+                       goto dma_map_err;
+               priv->tx_skbuff_dma[entry].buf = desc->des2;
+               priv->hw->desc->prepare_tx_desc(desc, 1, nopaged_len,
+                                               csum_insertion, priv->mode);
+       } else {
+               desc = first;
+               entry = priv->hw->mode->jumbo_frm(priv, skb, csum_insertion);
+               if (unlikely(entry < 0))
+                       goto dma_map_err;
+       }
+
+       for (i = 0; i < nfrags; i++) {
+               const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+               int len = skb_frag_size(frag);
+
+               priv->tx_skbuff[entry] = NULL;
+               entry = (++priv->cur_tx) % txsize;
+               if (priv->extend_desc)
+                       desc = (struct dma_desc *)(priv->dma_etx + entry);
+               else
+                       desc = priv->dma_tx + entry;
+
+               desc->des2 = skb_frag_dma_map(priv->device, frag, 0, len,
+                                             DMA_TO_DEVICE);
+               if (dma_mapping_error(priv->device, desc->des2))
+                       goto dma_map_err; /* should reuse desc w/o issues */
+
+               priv->tx_skbuff_dma[entry].buf = desc->des2;
+               priv->tx_skbuff_dma[entry].map_as_page = true;
+               priv->hw->desc->prepare_tx_desc(desc, 0, len, csum_insertion,
+                                               priv->mode);
+               wmb();
+               priv->hw->desc->set_tx_owner(desc);
+               wmb();
+       }
+
+       priv->tx_skbuff[entry] = skb;
+
+       /* Finalize the latest segment. */
+       priv->hw->desc->close_tx_desc(desc);
+
+       wmb();
+       /* According to the coalesce parameter the IC bit for the latest
+        * segment could be reset and the timer re-started to invoke the
+        * stmmac_tx function. This approach takes care about the fragments.
+        */
+       priv->tx_count_frames += nfrags + 1;
+       if (priv->tx_coal_frames > priv->tx_count_frames) {
+               priv->hw->desc->clear_tx_ic(desc);
+               priv->xstats.tx_reset_ic_bit++;
+               mod_timer(&priv->txtimer,
+                         STMMAC_COAL_TIMER(priv->tx_coal_timer));
+       } else
+               priv->tx_count_frames = 0;
+
+       /* To avoid raise condition */
+       priv->hw->desc->set_tx_owner(first);
+       wmb();
+
+       priv->cur_tx++;
+
+       if (netif_msg_pktdata(priv)) {
+               pr_debug("%s: curr %d dirty=%d entry=%d, first=%p, nfrags=%d",
+                       __func__, (priv->cur_tx % txsize),
+                       (priv->dirty_tx % txsize), entry, first, nfrags);
+
+               if (priv->extend_desc)
+                       stmmac_display_ring((void *)priv->dma_etx, txsize, 1);
+               else
+                       stmmac_display_ring((void *)priv->dma_tx, txsize, 0);
+
+               pr_debug(">>> frame to be transmitted: ");
+               print_pkt(skb->data, skb->len);
+       }
+       if (unlikely(stmmac_tx_avail(priv) <= (MAX_SKB_FRAGS + 1))) {
+               if (netif_msg_hw(priv))
+                       pr_debug("%s: stop transmitted packets\n", __func__);
+               netif_stop_queue(dev);
+       }
+
+       dev->stats.tx_bytes += skb->len;
+
+       if (unlikely((skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
+                    priv->hwts_tx_en)) {
+               /* declare that device is doing timestamping */
+               skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
+               priv->hw->desc->enable_tx_timestamp(first);
+       }
+
+       if (!priv->hwts_tx_en)
+               skb_tx_timestamp(skb);
+
+       netdev_sent_queue(dev, skb->len);
+       priv->hw->dma->enable_dma_transmission(priv->ioaddr);
+
+       spin_unlock(&priv->tx_lock);
+       return NETDEV_TX_OK;
+
+dma_map_err:
+       spin_unlock(&priv->tx_lock);
+       dev_err(priv->device, "Tx dma map failed\n");
+       dev_kfree_skb(skb);
+       priv->dev->stats.tx_dropped++;
+       return NETDEV_TX_OK;
+}
+
+static void stmmac_rx_vlan(struct net_device *dev, struct sk_buff *skb)
+{
+       struct ethhdr *ehdr;
+       u16 vlanid;
+
+       if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) ==
+           NETIF_F_HW_VLAN_CTAG_RX &&
+           !__vlan_get_tag(skb, &vlanid)) {
+               /* pop the vlan tag */
+               ehdr = (struct ethhdr *)skb->data;
+               memmove(skb->data + VLAN_HLEN, ehdr, ETH_ALEN * 2);
+               skb_pull(skb, VLAN_HLEN);
+               __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlanid);
+       }
+}
+
+
+/**
+ * stmmac_rx_refill - refill used skb preallocated buffers
+ * @priv: driver private structure
+ * Description : this is to reallocate the skb for the reception process
+ * that is based on zero-copy.
+ */
+static inline void stmmac_rx_refill(struct stmmac_priv *priv)
+{
+       unsigned int rxsize = priv->dma_rx_size;
+       int bfsize = priv->dma_buf_sz;
+
+       for (; priv->cur_rx - priv->dirty_rx > 0; priv->dirty_rx++) {
+               unsigned int entry = priv->dirty_rx % rxsize;
+               struct dma_desc *p;
+
+               if (priv->extend_desc)
+                       p = (struct dma_desc *)(priv->dma_erx + entry);
+               else
+                       p = priv->dma_rx + entry;
+
+               if (likely(priv->rx_skbuff[entry] == NULL)) {
+                       struct sk_buff *skb;
+
+                       skb = netdev_alloc_skb_ip_align(priv->dev, bfsize);
+
+                       if (unlikely(skb == NULL))
+                               break;
+
+                       priv->rx_skbuff[entry] = skb;
+                       priv->rx_skbuff_dma[entry] =
+                           dma_map_single(priv->device, skb->data, bfsize,
+                                          DMA_FROM_DEVICE);
+                       if (dma_mapping_error(priv->device,
+                                             priv->rx_skbuff_dma[entry])) {
+                               dev_err(priv->device, "Rx dma map failed\n");
+                               dev_kfree_skb(skb);
+                               break;
+                       }
+                       p->des2 = priv->rx_skbuff_dma[entry];
+
+                       priv->hw->mode->refill_desc3(priv, p);
+
+                       if (netif_msg_rx_status(priv))
+                               pr_debug("\trefill entry #%d\n", entry);
+               }
+               wmb();
+               priv->hw->desc->set_rx_owner(p);
+               wmb();
+       }
+}
+
+/**
+ * stmmac_rx - manage the receive process
+ * @priv: driver private structure
+ * @limit: napi bugget.
+ * Description :  this the function called by the napi poll method.
+ * It gets all the frames inside the ring.
+ */
+static int stmmac_rx(struct stmmac_priv *priv, int limit)
+{
+       unsigned int rxsize = priv->dma_rx_size;
+       unsigned int entry = priv->cur_rx % rxsize;
+       unsigned int next_entry;
+       unsigned int count = 0;
+       int coe = priv->hw->rx_csum;
+
+       if (netif_msg_rx_status(priv)) {
+               pr_debug("%s: descriptor ring:\n", __func__);
+               if (priv->extend_desc)
+                       stmmac_display_ring((void *)priv->dma_erx, rxsize, 1);
+               else
+                       stmmac_display_ring((void *)priv->dma_rx, rxsize, 0);
+       }
+       while (count < limit) {
+               int status;
+               struct dma_desc *p;
+
+               if (priv->extend_desc)
+                       p = (struct dma_desc *)(priv->dma_erx + entry);
+               else
+                       p = priv->dma_rx + entry;
+
+               if (priv->hw->desc->get_rx_owner(p))
+                       break;
+
+               count++;
+
+               next_entry = (++priv->cur_rx) % rxsize;
+               if (priv->extend_desc)
+                       prefetch(priv->dma_erx + next_entry);
+               else
+                       prefetch(priv->dma_rx + next_entry);
+
+               /* read the status of the incoming frame */
+               status = priv->hw->desc->rx_status(&priv->dev->stats,
+                                                  &priv->xstats, p);
+               if ((priv->extend_desc) && (priv->hw->desc->rx_extended_status))
+                       priv->hw->desc->rx_extended_status(&priv->dev->stats,
+                                                          &priv->xstats,
+                                                          priv->dma_erx +
+                                                          entry);
+               if (unlikely(status == discard_frame)) {
+                       priv->dev->stats.rx_errors++;
+                       if (priv->hwts_rx_en && !priv->extend_desc) {
+                               /* DESC2 & DESC3 will be overwitten by device
+                                * with timestamp value, hence reinitialize
+                                * them in stmmac_rx_refill() function so that
+                                * device can reuse it.
+                                */
+                               priv->rx_skbuff[entry] = NULL;
+                               dma_unmap_single(priv->device,
+                                                priv->rx_skbuff_dma[entry],
+                                                priv->dma_buf_sz,
+                                                DMA_FROM_DEVICE);
+                       }
+               } else {
+                       struct sk_buff *skb;
+                       int frame_len;
+
+                       frame_len = priv->hw->desc->get_rx_frame_len(p, coe);
+
+                       /* ACS is set; GMAC core strips PAD/FCS for IEEE 802.3
+                        * Type frames (LLC/LLC-SNAP)
+                        */
+                       if (unlikely(status != llc_snap))
+                               frame_len -= ETH_FCS_LEN;
+
+                       if (netif_msg_rx_status(priv)) {
+                               pr_debug("\tdesc: %p [entry %d] buff=0x%x\n",
+                                        p, entry, p->des2);
+                               if (frame_len > ETH_FRAME_LEN)
+                                       pr_debug("\tframe size %d, COE: %d\n",
+                                                frame_len, status);
+                       }
+                       skb = priv->rx_skbuff[entry];
+                       if (unlikely(!skb)) {
+                               pr_err("%s: Inconsistent Rx descriptor chain\n",
+                                      priv->dev->name);
+                               priv->dev->stats.rx_dropped++;
+                               break;
+                       }
+                       prefetch(skb->data - NET_IP_ALIGN);
+                       priv->rx_skbuff[entry] = NULL;
+
+                       stmmac_get_rx_hwtstamp(priv, entry, skb);
+
+                       skb_put(skb, frame_len);
+                       dma_unmap_single(priv->device,
+                                        priv->rx_skbuff_dma[entry],
+                                        priv->dma_buf_sz, DMA_FROM_DEVICE);
+
+                       if (netif_msg_pktdata(priv)) {
+                               pr_debug("frame received (%dbytes)", frame_len);
+                               print_pkt(skb->data, frame_len);
+                       }
+
+                       stmmac_rx_vlan(priv->dev, skb);
+
+                       skb->protocol = eth_type_trans(skb, priv->dev);
+
+                       if (unlikely(!coe))
+                               skb_checksum_none_assert(skb);
+                       else
+                               skb->ip_summed = CHECKSUM_UNNECESSARY;
+
+                       napi_gro_receive(&priv->napi, skb);
+
+                       priv->dev->stats.rx_packets++;
+                       priv->dev->stats.rx_bytes += frame_len;
+               }
+               entry = next_entry;
+       }
+
+       stmmac_rx_refill(priv);
+
+       priv->xstats.rx_pkt_n += count;
+
+       return count;
+}
+
+/**
+ *  stmmac_poll - stmmac poll method (NAPI)
+ *  @napi : pointer to the napi structure.
+ *  @budget : maximum number of packets that the current CPU can receive from
+ *           all interfaces.
+ *  Description :
+ *  To look at the incoming frames and clear the tx resources.
+ */
+static int stmmac_poll(struct napi_struct *napi, int budget)
+{
+       struct stmmac_priv *priv = container_of(napi, struct stmmac_priv, napi);
+       int work_done = 0;
+
+       priv->xstats.napi_poll++;
+       stmmac_tx_clean(priv);
+
+       work_done = stmmac_rx(priv, budget);
+       if (work_done < budget) {
+               napi_complete(napi);
+               stmmac_enable_dma_irq(priv);
+       }
+       return work_done;
+}
+
+/**
+ *  stmmac_tx_timeout
+ *  @dev : Pointer to net device structure
+ *  Description: this function is called when a packet transmission fails to
+ *   complete within a reasonable time. The driver will mark the error in the
+ *   netdev structure and arrange for the device to be reset to a sane state
+ *   in order to transmit a new packet.
+ */
+static void stmmac_tx_timeout(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       /* Clear Tx resources and restart transmitting again */
+       stmmac_tx_err(priv);
+}
+
+/**
+ *  stmmac_set_rx_mode - entry point for multicast addressing
+ *  @dev : pointer to the device structure
+ *  Description:
+ *  This function is a driver entry point which gets called by the kernel
+ *  whenever multicast addresses must be enabled/disabled.
+ *  Return value:
+ *  void.
+ */
+static void stmmac_set_rx_mode(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       priv->hw->mac->set_filter(priv->hw, dev);
+}
+
+/**
+ *  stmmac_change_mtu - entry point to change MTU size for the device.
+ *  @dev : device pointer.
+ *  @new_mtu : the new MTU size for the device.
+ *  Description: the Maximum Transfer Unit (MTU) is used by the network layer
+ *  to drive packet transmission. Ethernet has an MTU of 1500 octets
+ *  (ETH_DATA_LEN). This value can be changed with ifconfig.
+ *  Return value:
+ *  0 on success and an appropriate (-)ve integer as defined in errno.h
+ *  file on failure.
+ */
+static int stmmac_change_mtu(struct net_device *dev, int new_mtu)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       int max_mtu;
+
+       if (netif_running(dev)) {
+               pr_err("%s: must be stopped to change its MTU\n", dev->name);
+               return -EBUSY;
+       }
+
+       if (priv->plat->enh_desc)
+               max_mtu = JUMBO_LEN;
+       else
+               max_mtu = SKB_MAX_HEAD(NET_SKB_PAD + NET_IP_ALIGN);
+
+       if (priv->plat->maxmtu < max_mtu)
+               max_mtu = priv->plat->maxmtu;
+
+       if ((new_mtu < 46) || (new_mtu > max_mtu)) {
+               pr_err("%s: invalid MTU, max MTU is: %d\n", dev->name, max_mtu);
+               return -EINVAL;
+       }
+
+       dev->mtu = new_mtu;
+       netdev_update_features(dev);
+
+       return 0;
+}
+
+static netdev_features_t stmmac_fix_features(struct net_device *dev,
+                                            netdev_features_t features)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       if (priv->plat->rx_coe == STMMAC_RX_COE_NONE)
+               features &= ~NETIF_F_RXCSUM;
+
+       if (!priv->plat->tx_coe)
+               features &= ~NETIF_F_ALL_CSUM;
+
+       /* Some GMAC devices have a bugged Jumbo frame support that
+        * needs to have the Tx COE disabled for oversized frames
+        * (due to limited buffer sizes). In this case we disable
+        * the TX csum insertionin the TDES and not use SF.
+        */
+       if (priv->plat->bugged_jumbo && (dev->mtu > ETH_DATA_LEN))
+               features &= ~NETIF_F_ALL_CSUM;
+
+       return features;
+}
+
+static int stmmac_set_features(struct net_device *netdev,
+                              netdev_features_t features)
+{
+       struct stmmac_priv *priv = netdev_priv(netdev);
+
+       /* Keep the COE Type in case of csum is supporting */
+       if (features & NETIF_F_RXCSUM)
+               priv->hw->rx_csum = priv->plat->rx_coe;
+       else
+               priv->hw->rx_csum = 0;
+       /* No check needed because rx_coe has been set before and it will be
+        * fixed in case of issue.
+        */
+       priv->hw->mac->rx_ipc(priv->hw);
+
+       return 0;
+}
+
+/**
+ *  stmmac_interrupt - main ISR
+ *  @irq: interrupt number.
+ *  @dev_id: to pass the net device pointer.
+ *  Description: this is the main driver interrupt service routine.
+ *  It can call:
+ *  o DMA service routine (to manage incoming frame reception and transmission
+ *    status)
+ *  o Core interrupts to manage: remote wake-up, management counter, LPI
+ *    interrupts.
+ */
+static irqreturn_t stmmac_interrupt(int irq, void *dev_id)
+{
+       struct net_device *dev = (struct net_device *)dev_id;
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       if (priv->irq_wake)
+               pm_wakeup_event(priv->device, 0);
+
+       if (unlikely(!dev)) {
+               pr_err("%s: invalid dev pointer\n", __func__);
+               return IRQ_NONE;
+       }
+
+       /* To handle GMAC own interrupts */
+       if (priv->plat->has_gmac) {
+               int status = priv->hw->mac->host_irq_status(priv->hw,
+                                                           &priv->xstats);
+               if (unlikely(status)) {
+                       /* For LPI we need to save the tx status */
+                       if (status & CORE_IRQ_TX_PATH_IN_LPI_MODE)
+                               priv->tx_path_in_lpi_mode = true;
+                       if (status & CORE_IRQ_TX_PATH_EXIT_LPI_MODE)
+                               priv->tx_path_in_lpi_mode = false;
+               }
+       }
+
+       /* To handle DMA interrupts */
+       stmmac_dma_interrupt(priv);
+
+       return IRQ_HANDLED;
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+/* Polling receive - used by NETCONSOLE and other diagnostic tools
+ * to allow network I/O with interrupts disabled.
+ */
+static void stmmac_poll_controller(struct net_device *dev)
+{
+       disable_irq(dev->irq);
+       stmmac_interrupt(dev->irq, dev);
+       enable_irq(dev->irq);
+}
+#endif
+
+/**
+ *  stmmac_ioctl - Entry point for the Ioctl
+ *  @dev: Device pointer.
+ *  @rq: An IOCTL specefic structure, that can contain a pointer to
+ *  a proprietary structure used to pass information to the driver.
+ *  @cmd: IOCTL command
+ *  Description:
+ *  Currently it supports the phy_mii_ioctl(...) and HW time stamping.
+ */
+static int stmmac_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+       int ret = -EOPNOTSUPP;
+
+       if (!netif_running(dev))
+               return -EINVAL;
+
+       switch (cmd) {
+       case SIOCGMIIPHY:
+       case SIOCGMIIREG:
+       case SIOCSMIIREG:
+               if (!priv->phydev)
+                       return -EINVAL;
+               ret = phy_mii_ioctl(priv->phydev, rq, cmd);
+               break;
+       case SIOCSHWTSTAMP:
+               ret = stmmac_hwtstamp_ioctl(dev, rq);
+               break;
+       default:
+               break;
+       }
+
+       return ret;
+}
+
+#ifdef CONFIG_DEBUG_FS
+static struct dentry *stmmac_fs_dir;
+
+static void sysfs_display_ring(void *head, int size, int extend_desc,
+                              struct seq_file *seq)
+{
+       int i;
+       struct dma_extended_desc *ep = (struct dma_extended_desc *)head;
+       struct dma_desc *p = (struct dma_desc *)head;
+
+       for (i = 0; i < size; i++) {
+               u64 x;
+               if (extend_desc) {
+                       x = *(u64 *) ep;
+                       seq_printf(seq, "%d [0x%x]: 0x%x 0x%x 0x%x 0x%x\n",
+                                  i, (unsigned int)virt_to_phys(ep),
+                                  (unsigned int)x, (unsigned int)(x >> 32),
+                                  ep->basic.des2, ep->basic.des3);
+                       ep++;
+               } else {
+                       x = *(u64 *) p;
+                       seq_printf(seq, "%d [0x%x]: 0x%x 0x%x 0x%x 0x%x\n",
+                                  i, (unsigned int)virt_to_phys(ep),
+                                  (unsigned int)x, (unsigned int)(x >> 32),
+                                  p->des2, p->des3);
+                       p++;
+               }
+               seq_printf(seq, "\n");
+       }
+}
+
+static int stmmac_sysfs_ring_read(struct seq_file *seq, void *v)
+{
+       struct net_device *dev = seq->private;
+       struct stmmac_priv *priv = netdev_priv(dev);
+       unsigned int txsize = priv->dma_tx_size;
+       unsigned int rxsize = priv->dma_rx_size;
+
+       if (priv->extend_desc) {
+               seq_printf(seq, "Extended RX descriptor ring:\n");
+               sysfs_display_ring((void *)priv->dma_erx, rxsize, 1, seq);
+               seq_printf(seq, "Extended TX descriptor ring:\n");
+               sysfs_display_ring((void *)priv->dma_etx, txsize, 1, seq);
+       } else {
+               seq_printf(seq, "RX descriptor ring:\n");
+               sysfs_display_ring((void *)priv->dma_rx, rxsize, 0, seq);
+               seq_printf(seq, "TX descriptor ring:\n");
+               sysfs_display_ring((void *)priv->dma_tx, txsize, 0, seq);
+       }
+
+       return 0;
+}
+
+static int stmmac_sysfs_ring_open(struct inode *inode, struct file *file)
+{
+       return single_open(file, stmmac_sysfs_ring_read, inode->i_private);
+}
+
+static const struct file_operations stmmac_rings_status_fops = {
+       .owner = THIS_MODULE,
+       .open = stmmac_sysfs_ring_open,
+       .read = seq_read,
+       .llseek = seq_lseek,
+       .release = single_release,
+};
+
+static int stmmac_sysfs_dma_cap_read(struct seq_file *seq, void *v)
+{
+       struct net_device *dev = seq->private;
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       if (!priv->hw_cap_support) {
+               seq_printf(seq, "DMA HW features not supported\n");
+               return 0;
+       }
+
+       seq_printf(seq, "==============================\n");
+       seq_printf(seq, "\tDMA HW features\n");
+       seq_printf(seq, "==============================\n");
+
+       seq_printf(seq, "\t10/100 Mbps %s\n",
+                  (priv->dma_cap.mbps_10_100) ? "Y" : "N");
+       seq_printf(seq, "\t1000 Mbps %s\n",
+                  (priv->dma_cap.mbps_1000) ? "Y" : "N");
+       seq_printf(seq, "\tHalf duple %s\n",
+                  (priv->dma_cap.half_duplex) ? "Y" : "N");
+       seq_printf(seq, "\tHash Filter: %s\n",
+                  (priv->dma_cap.hash_filter) ? "Y" : "N");
+       seq_printf(seq, "\tMultiple MAC address registers: %s\n",
+                  (priv->dma_cap.multi_addr) ? "Y" : "N");
+       seq_printf(seq, "\tPCS (TBI/SGMII/RTBI PHY interfatces): %s\n",
+                  (priv->dma_cap.pcs) ? "Y" : "N");
+       seq_printf(seq, "\tSMA (MDIO) Interface: %s\n",
+                  (priv->dma_cap.sma_mdio) ? "Y" : "N");
+       seq_printf(seq, "\tPMT Remote wake up: %s\n",
+                  (priv->dma_cap.pmt_remote_wake_up) ? "Y" : "N");
+       seq_printf(seq, "\tPMT Magic Frame: %s\n",
+                  (priv->dma_cap.pmt_magic_frame) ? "Y" : "N");
+       seq_printf(seq, "\tRMON module: %s\n",
+                  (priv->dma_cap.rmon) ? "Y" : "N");
+       seq_printf(seq, "\tIEEE 1588-2002 Time Stamp: %s\n",
+                  (priv->dma_cap.time_stamp) ? "Y" : "N");
+       seq_printf(seq, "\tIEEE 1588-2008 Advanced Time Stamp:%s\n",
+                  (priv->dma_cap.atime_stamp) ? "Y" : "N");
+       seq_printf(seq, "\t802.3az - Energy-Efficient Ethernet (EEE) %s\n",
+                  (priv->dma_cap.eee) ? "Y" : "N");
+       seq_printf(seq, "\tAV features: %s\n", (priv->dma_cap.av) ? "Y" : "N");
+       seq_printf(seq, "\tChecksum Offload in TX: %s\n",
+                  (priv->dma_cap.tx_coe) ? "Y" : "N");
+       seq_printf(seq, "\tIP Checksum Offload (type1) in RX: %s\n",
+                  (priv->dma_cap.rx_coe_type1) ? "Y" : "N");
+       seq_printf(seq, "\tIP Checksum Offload (type2) in RX: %s\n",
+                  (priv->dma_cap.rx_coe_type2) ? "Y" : "N");
+       seq_printf(seq, "\tRXFIFO > 2048bytes: %s\n",
+                  (priv->dma_cap.rxfifo_over_2048) ? "Y" : "N");
+       seq_printf(seq, "\tNumber of Additional RX channel: %d\n",
+                  priv->dma_cap.number_rx_channel);
+       seq_printf(seq, "\tNumber of Additional TX channel: %d\n",
+                  priv->dma_cap.number_tx_channel);
+       seq_printf(seq, "\tEnhanced descriptors: %s\n",
+                  (priv->dma_cap.enh_desc) ? "Y" : "N");
+
+       return 0;
+}
+
+static int stmmac_sysfs_dma_cap_open(struct inode *inode, struct file *file)
+{
+       return single_open(file, stmmac_sysfs_dma_cap_read, inode->i_private);
+}
+
+static const struct file_operations stmmac_dma_cap_fops = {
+       .owner = THIS_MODULE,
+       .open = stmmac_sysfs_dma_cap_open,
+       .read = seq_read,
+       .llseek = seq_lseek,
+       .release = single_release,
+};
+
+static int stmmac_init_fs(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       /* Create per netdev entries */
+       priv->dbgfs_dir = debugfs_create_dir(dev->name, stmmac_fs_dir);
+
+       if (!priv->dbgfs_dir || IS_ERR(priv->dbgfs_dir)) {
+               pr_err("ERROR %s/%s, debugfs create directory failed\n",
+                      STMMAC_RESOURCE_NAME, dev->name);
+
+               return -ENOMEM;
+       }
+
+       /* Entry to report DMA RX/TX rings */
+       priv->dbgfs_rings_status =
+               debugfs_create_file("descriptors_status", S_IRUGO,
+                                   priv->dbgfs_dir, dev,
+                                   &stmmac_rings_status_fops);
+
+       if (!priv->dbgfs_rings_status || IS_ERR(priv->dbgfs_rings_status)) {
+               pr_info("ERROR creating stmmac ring debugfs file\n");
+               debugfs_remove_recursive(priv->dbgfs_dir);
+
+               return -ENOMEM;
+       }
+
+       /* Entry to report the DMA HW features */
+       priv->dbgfs_dma_cap = debugfs_create_file("dma_cap", S_IRUGO,
+                                           priv->dbgfs_dir,
+                                           dev, &stmmac_dma_cap_fops);
+
+       if (!priv->dbgfs_dma_cap || IS_ERR(priv->dbgfs_dma_cap)) {
+               pr_info("ERROR creating stmmac MMC debugfs file\n");
+               debugfs_remove_recursive(priv->dbgfs_dir);
+
+               return -ENOMEM;
+       }
+
+       return 0;
+}
+
+static void stmmac_exit_fs(struct net_device *dev)
+{
+       struct stmmac_priv *priv = netdev_priv(dev);
+
+       debugfs_remove_recursive(priv->dbgfs_dir);
+}
+#endif /* CONFIG_DEBUG_FS */
+
+static const struct net_device_ops stmmac_netdev_ops = {
+       .ndo_open = stmmac_open,
+       .ndo_start_xmit = stmmac_xmit,
+       .ndo_stop = stmmac_release,
+       .ndo_change_mtu = stmmac_change_mtu,
+       .ndo_fix_features = stmmac_fix_features,
+       .ndo_set_features = stmmac_set_features,
+       .ndo_set_rx_mode = stmmac_set_rx_mode,
+       .ndo_tx_timeout = stmmac_tx_timeout,
+       .ndo_do_ioctl = stmmac_ioctl,
+#ifdef CONFIG_NET_POLL_CONTROLLER
+       .ndo_poll_controller = stmmac_poll_controller,
+#endif
+       .ndo_set_mac_address = eth_mac_addr,
+};
+
+/**
+ *  stmmac_hw_init - Init the MAC device
+ *  @priv: driver private structure
+ *  Description: this function is to configure the MAC device according to
+ *  some platform parameters or the HW capability register. It prepares the
+ *  driver to use either ring or chain modes and to setup either enhanced or
+ *  normal descriptors.
+ */
+static int stmmac_hw_init(struct stmmac_priv *priv)
+{
+       struct mac_device_info *mac;
+
+       /* Identify the MAC HW device */
+       if (priv->plat->has_gmac) {
+               priv->dev->priv_flags |= IFF_UNICAST_FLT;
+               mac = dwmac1000_setup(priv->ioaddr,
+                                     priv->plat->multicast_filter_bins,
+                                     priv->plat->unicast_filter_entries);
+       } else {
+               mac = dwmac100_setup(priv->ioaddr);
+       }
+       if (!mac)
+               return -ENOMEM;
+
+       priv->hw = mac;
+
+       /* Get and dump the chip ID */
+       priv->synopsys_id = stmmac_get_synopsys_id(priv);
+
+       /* To use the chained or ring mode */
+       if (chain_mode) {
+               priv->hw->mode = &chain_mode_ops;
+               pr_info(" Chain mode enabled\n");
+               priv->mode = STMMAC_CHAIN_MODE;
+       } else {
+               priv->hw->mode = &ring_mode_ops;
+               pr_info(" Ring mode enabled\n");
+               priv->mode = STMMAC_RING_MODE;
+       }
+
+       /* Get the HW capability (new GMAC newer than 3.50a) */
+       priv->hw_cap_support = stmmac_get_hw_features(priv);
+       if (priv->hw_cap_support) {
+               pr_info(" DMA HW capability register supported");
+
+               /* We can override some gmac/dma configuration fields: e.g.
+                * enh_desc, tx_coe (e.g. that are passed through the
+                * platform) with the values from the HW capability
+                * register (if supported).
+                */
+               priv->plat->enh_desc = priv->dma_cap.enh_desc;
+               priv->plat->pmt = priv->dma_cap.pmt_remote_wake_up;
+
+               /* TXCOE doesn't work in thresh DMA mode */
+               if (priv->plat->force_thresh_dma_mode)
+                       priv->plat->tx_coe = 0;
+               else
+                       priv->plat->tx_coe = priv->dma_cap.tx_coe;
+
+               if (priv->dma_cap.rx_coe_type2)
+                       priv->plat->rx_coe = STMMAC_RX_COE_TYPE2;
+               else if (priv->dma_cap.rx_coe_type1)
+                       priv->plat->rx_coe = STMMAC_RX_COE_TYPE1;
+
+       } else
+               pr_info(" No HW DMA feature register supported");
+
+       /* To use alternate (extended) or normal descriptor structures */
+       stmmac_selec_desc_mode(priv);
+
+       if (priv->plat->rx_coe) {
+               priv->hw->rx_csum = priv->plat->rx_coe;
+               pr_info(" RX Checksum Offload Engine supported (type %d)\n",
+                       priv->plat->rx_coe);
+       }
+       if (priv->plat->tx_coe)
+               pr_info(" TX Checksum insertion supported\n");
+
+       if (priv->plat->pmt) {
+               pr_info(" Wake-Up On Lan supported\n");
+               device_set_wakeup_capable(priv->device, 1);
+       }
+
+       return 0;
+}
+
+/**
+ * stmmac_dvr_probe
+ * @device: device pointer
+ * @plat_dat: platform data pointer
+ * @addr: iobase memory address
+ * Description: this is the main probe function used to
+ * call the alloc_etherdev, allocate the priv structure.
+ * Return:
+ * on success the new private structure is returned, otherwise the error
+ * pointer.
+ */
+struct stmmac_priv *stmmac_dvr_probe(struct device *device,
+                                    struct plat_stmmacenet_data *plat_dat,
+                                    void __iomem *addr)
+{
+       int ret = 0;
+       struct net_device *ndev = NULL;
+       struct stmmac_priv *priv;
+
+       ndev = alloc_etherdev(sizeof(struct stmmac_priv));
+       if (!ndev)
+               return ERR_PTR(-ENOMEM);
+
+       SET_NETDEV_DEV(ndev, device);
+
+       priv = netdev_priv(ndev);
+       priv->device = device;
+       priv->dev = ndev;
+
+       stmmac_set_ethtool_ops(ndev);
+       priv->pause = pause;
+       priv->plat = plat_dat;
+       priv->ioaddr = addr;
+       priv->dev->base_addr = (unsigned long)addr;
+
+       /* Verify driver arguments */
+       stmmac_verify_args();
+
+       /* Override with kernel parameters if supplied XXX CRS XXX
+        * this needs to have multiple instances
+        */
+       if ((phyaddr >= 0) && (phyaddr <= 31))
+               priv->plat->phy_addr = phyaddr;
+
+       priv->stmmac_clk = devm_clk_get(priv->device, STMMAC_RESOURCE_NAME);
+       if (IS_ERR(priv->stmmac_clk)) {
+               dev_warn(priv->device, "%s: warning: cannot get CSR clock\n",
+                        __func__);
+               /* If failed to obtain stmmac_clk and specific clk_csr value
+                * is NOT passed from the platform, probe fail.
+                */
+               if (!priv->plat->clk_csr) {
+                       ret = PTR_ERR(priv->stmmac_clk);
+                       goto error_clk_get;
+               } else {
+                       priv->stmmac_clk = NULL;
+               }
+       }
+       clk_prepare_enable(priv->stmmac_clk);
+
+       priv->pclk = devm_clk_get(priv->device, "pclk");
+       if (IS_ERR(priv->pclk)) {
+               if (PTR_ERR(priv->pclk) == -EPROBE_DEFER) {
+                       ret = -EPROBE_DEFER;
+                       goto error_pclk_get;
+               }
+               priv->pclk = NULL;
+       }
+       clk_prepare_enable(priv->pclk);
+
+       priv->stmmac_rst = devm_reset_control_get(priv->device,
+                                                 STMMAC_RESOURCE_NAME);
+       if (IS_ERR(priv->stmmac_rst)) {
+               if (PTR_ERR(priv->stmmac_rst) == -EPROBE_DEFER) {
+                       ret = -EPROBE_DEFER;
+                       goto error_hw_init;
+               }
+               dev_info(priv->device, "no reset control found\n");
+               priv->stmmac_rst = NULL;
+       }
+       if (priv->stmmac_rst)
+               reset_control_deassert(priv->stmmac_rst);
+
+       /* Init MAC and get the capabilities */
+       ret = stmmac_hw_init(priv);
+       if (ret)
+               goto error_hw_init;
+
+       ndev->netdev_ops = &stmmac_netdev_ops;
+
+       ndev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
+                           NETIF_F_RXCSUM;
+       ndev->features |= ndev->hw_features | NETIF_F_HIGHDMA;
+       ndev->watchdog_timeo = msecs_to_jiffies(watchdog);
+#ifdef STMMAC_VLAN_TAG_USED
+       /* Both mac100 and gmac support receive VLAN tag detection */
+       ndev->features |= NETIF_F_HW_VLAN_CTAG_RX;
+#endif
+       priv->msg_enable = netif_msg_init(debug, default_msg_level);
+
+       if (flow_ctrl)
+               priv->flow_ctrl = FLOW_AUTO;    /* RX/TX pause on */
+
+       /* Rx Watchdog is available in the COREs newer than the 3.40.
+        * In some case, for example on bugged HW this feature
+        * has to be disable and this can be done by passing the
+        * riwt_off field from the platform.
+        */
+       if ((priv->synopsys_id >= DWMAC_CORE_3_50) && (!priv->plat->riwt_off)) {
+               priv->use_riwt = 1;
+               pr_info(" Enable RX Mitigation via HW Watchdog Timer\n");
+       }
+
+       netif_napi_add(ndev, &priv->napi, stmmac_poll, 64);
+
+       spin_lock_init(&priv->lock);
+       spin_lock_init(&priv->tx_lock);
+
+       ret = register_netdev(ndev);
+       if (ret) {
+               pr_err("%s: ERROR %i registering the device\n", __func__, ret);
+               goto error_netdev_register;
+       }
+
+       /* If a specific clk_csr value is passed from the platform
+        * this means that the CSR Clock Range selection cannot be
+        * changed at run-time and it is fixed. Viceversa the driver'll try to
+        * set the MDC clock dynamically according to the csr actual
+        * clock input.
+        */
+       if (!priv->plat->clk_csr)
+               stmmac_clk_csr_set(priv);
+       else
+               priv->clk_csr = priv->plat->clk_csr;
+
+       stmmac_check_pcs_mode(priv);
+
+       if (priv->pcs != STMMAC_PCS_RGMII && priv->pcs != STMMAC_PCS_TBI &&
+           priv->pcs != STMMAC_PCS_RTBI) {
+               /* MDIO bus Registration */
+               ret = stmmac_mdio_register(ndev);
+               if (ret < 0) {
+                       pr_debug("%s: MDIO bus (id: %d) registration failed",
+                                __func__, priv->plat->bus_id);
+                       goto error_mdio_register;
+               }
+       }
+
+       return priv;
+
+error_mdio_register:
+       unregister_netdev(ndev);
+error_netdev_register:
+       netif_napi_del(&priv->napi);
+error_hw_init:
+       clk_disable_unprepare(priv->pclk);
+error_pclk_get:
+       clk_disable_unprepare(priv->stmmac_clk);
+error_clk_get:
+       free_netdev(ndev);
+
+       return ERR_PTR(ret);
+}
+EXPORT_SYMBOL_GPL(stmmac_dvr_probe);
+
+/**
+ * stmmac_dvr_remove
+ * @ndev: net device pointer
+ * Description: this function resets the TX/RX processes, disables the MAC RX/TX
+ * changes the link status, releases the DMA descriptor rings.
+ */
+int stmmac_dvr_remove(struct net_device *ndev)
+{
+       struct stmmac_priv *priv = netdev_priv(ndev);
+
+       pr_info("%s:\n\tremoving driver", __func__);
+
+       priv->hw->dma->stop_rx(priv->ioaddr);
+       priv->hw->dma->stop_tx(priv->ioaddr);
+
+       stmmac_set_mac(priv->ioaddr, false);
+       netif_carrier_off(ndev);
+       unregister_netdev(ndev);
+       if (priv->stmmac_rst)
+               reset_control_assert(priv->stmmac_rst);
+       clk_disable_unprepare(priv->pclk);
+       clk_disable_unprepare(priv->stmmac_clk);
+       if (priv->pcs != STMMAC_PCS_RGMII && priv->pcs != STMMAC_PCS_TBI &&
+           priv->pcs != STMMAC_PCS_RTBI)
+               stmmac_mdio_unregister(ndev);
+       free_netdev(ndev);
+
+       return 0;
+}
+EXPORT_SYMBOL_GPL(stmmac_dvr_remove);
+
+/**
+ * stmmac_suspend - suspend callback
+ * @ndev: net device pointer
+ * Description: this is the function to suspend the device and it is called
+ * by the platform driver to stop the network queue, release the resources,
+ * program the PMT register (for WoL), clean and release driver resources.
+ */
+int stmmac_suspend(struct net_device *ndev)
+{
+       struct stmmac_priv *priv = netdev_priv(ndev);
+       unsigned long flags;
+
+       if (!ndev || !netif_running(ndev))
+               return 0;
+
+       if (priv->phydev)
+               phy_stop(priv->phydev);
+
+       spin_lock_irqsave(&priv->lock, flags);
+
+       netif_device_detach(ndev);
+       netif_stop_queue(ndev);
+
+       napi_disable(&priv->napi);
+
+       /* Stop TX/RX DMA */
+       priv->hw->dma->stop_tx(priv->ioaddr);
+       priv->hw->dma->stop_rx(priv->ioaddr);
+
+       stmmac_clear_descriptors(priv);
+
+       /* Enable Power down mode by programming the PMT regs */
+       if (device_may_wakeup(priv->device)) {
+               priv->hw->mac->pmt(priv->hw, priv->wolopts);
+               priv->irq_wake = 1;
+       } else {
+               stmmac_set_mac(priv->ioaddr, false);
+               pinctrl_pm_select_sleep_state(priv->device);
+               /* Disable clock in case of PWM is off */
+               clk_disable(priv->pclk);
+               clk_disable(priv->stmmac_clk);
+       }
+       spin_unlock_irqrestore(&priv->lock, flags);
+
+       priv->oldlink = 0;
+       priv->speed = 0;
+       priv->oldduplex = -1;
+       return 0;
+}
+EXPORT_SYMBOL_GPL(stmmac_suspend);
+
+/**
+ * stmmac_resume - resume callback
+ * @ndev: net device pointer
+ * Description: when resume this function is invoked to setup the DMA and CORE
+ * in a usable state.
+ */
+int stmmac_resume(struct net_device *ndev)
+{
+       struct stmmac_priv *priv = netdev_priv(ndev);
+       unsigned long flags;
+
+       if (!netif_running(ndev))
+               return 0;
+
+       spin_lock_irqsave(&priv->lock, flags);
+
+       /* Power Down bit, into the PM register, is cleared
+        * automatically as soon as a magic packet or a Wake-up frame
+        * is received. Anyway, it's better to manually clear
+        * this bit because it can generate problems while resuming
+        * from another devices (e.g. serial console).
+        */
+       if (device_may_wakeup(priv->device)) {
+               priv->hw->mac->pmt(priv->hw, 0);
+               priv->irq_wake = 0;
+       } else {
+               pinctrl_pm_select_default_state(priv->device);
+               /* enable the clk prevously disabled */
+               clk_enable(priv->stmmac_clk);
+               clk_enable(priv->pclk);
+               /* reset the phy so that it's ready */
+               if (priv->mii)
+                       stmmac_mdio_reset(priv->mii);
+       }
+
+       netif_device_attach(ndev);
+
+       init_dma_desc_rings(ndev, GFP_ATOMIC);
+       stmmac_hw_setup(ndev, false);
+       stmmac_init_tx_coalesce(priv);
+
+       napi_enable(&priv->napi);
+
+       netif_start_queue(ndev);
+
+       spin_unlock_irqrestore(&priv->lock, flags);
+
+       if (priv->phydev)
+               phy_start(priv->phydev);
+
+       return 0;
+}
+EXPORT_SYMBOL_GPL(stmmac_resume);
+
+#ifndef MODULE
+static int __init stmmac_cmdline_opt(char *str)
+{
+       char *opt;
+
+       if (!str || !*str)
+               return -EINVAL;
+       while ((opt = strsep(&str, ",")) != NULL) {
+               if (!strncmp(opt, "debug:", 6)) {
+                       if (kstrtoint(opt + 6, 0, &debug))
+                               goto err;
+               } else if (!strncmp(opt, "phyaddr:", 8)) {
+                       if (kstrtoint(opt + 8, 0, &phyaddr))
+                               goto err;
+               } else if (!strncmp(opt, "dma_txsize:", 11)) {
+                       if (kstrtoint(opt + 11, 0, &dma_txsize))
+                               goto err;
+               } else if (!strncmp(opt, "dma_rxsize:", 11)) {
+                       if (kstrtoint(opt + 11, 0, &dma_rxsize))
+                               goto err;
+               } else if (!strncmp(opt, "buf_sz:", 7)) {
+                       if (kstrtoint(opt + 7, 0, &buf_sz))
+                               goto err;
+               } else if (!strncmp(opt, "tc:", 3)) {
+                       if (kstrtoint(opt + 3, 0, &tc))
+                               goto err;
+               } else if (!strncmp(opt, "watchdog:", 9)) {
+                       if (kstrtoint(opt + 9, 0, &watchdog))
+                               goto err;
+               } else if (!strncmp(opt, "flow_ctrl:", 10)) {
+                       if (kstrtoint(opt + 10, 0, &flow_ctrl))
+                               goto err;
+               } else if (!strncmp(opt, "pause:", 6)) {
+                       if (kstrtoint(opt + 6, 0, &pause))
+                               goto err;
+               } else if (!strncmp(opt, "eee_timer:", 10)) {
+                       if (kstrtoint(opt + 10, 0, &eee_timer))
+                               goto err;
+               } else if (!strncmp(opt, "chain_mode:", 11)) {
+                       if (kstrtoint(opt + 11, 0, &chain_mode))
+                               goto err;
+               }
+       }
+       return 0;
+
+err:
+       pr_err("%s: ERROR broken module parameter conversion", __func__);
+       return -EINVAL;
+}
+
+__setup("stmmaceth=", stmmac_cmdline_opt);
+#endif /* MODULE */
+
+static int __init stmmac_init(void)
+{
+#ifdef CONFIG_DEBUG_FS
+       /* Create debugfs main directory if it doesn't exist yet */
+       if (!stmmac_fs_dir) {
+               stmmac_fs_dir = debugfs_create_dir(STMMAC_RESOURCE_NAME, NULL);
+
+               if (!stmmac_fs_dir || IS_ERR(stmmac_fs_dir)) {
+                       pr_err("ERROR %s, debugfs create directory failed\n",
+                              STMMAC_RESOURCE_NAME);
+
+                       return -ENOMEM;
+               }
+       }
+#endif
+
+       return 0;
+}
+
+static void __exit stmmac_exit(void)
+{
+#ifdef CONFIG_DEBUG_FS
+       debugfs_remove_recursive(stmmac_fs_dir);
+#endif
+}
+
+module_init(stmmac_init)
+module_exit(stmmac_exit)
+
+MODULE_DESCRIPTION("STMMAC 10/100/1000 Ethernet device driver");
+MODULE_AUTHOR("Giuseppe Cavallaro <peppe.cavallaro@st.com>");
+MODULE_LICENSE("GPL");